1
Eddy covariance measurement of ammonia fluxes 1
comparison of high frequency correction 2
methodologies 3
R M FERRARAa1
B LOUBETb P DI TOMMASI
c T BERTOLINI
c V MAGLIULO
c P CELLIER
b 4
W EUGSTERe G RANA
d 5
a Agricultural Research Council ndash Centre for Research for agrobiology and pedology (CRA ndash 6
ABP) Piazza Massimo DrsquoAzeglio30 ndash 50121 Firenze Italy phone +39 055 2491245 fax 7
+39 055 241485 rossanaferraraentecrait 8
b INRA UMR 1091 Environnement et Grandes Cultures F- 78850 Thiverval-Grignon 9
France 10
c CNR ndash ISAFoM Via Patacca 85 Ercolano (Na) Italy 11
d Agricultural Research Council - Research Unit for cropping systems in dry environments 12
via C Ulpiani 5 70125 Bari Italy 13
e ETH Zurich Institute of Agricultural Sciences Universitaumltsstr 2 8092 Zurich Switzerland
14
15
Key words transfer function closed-path analyser ogive inductance QC-TILDAS 16
17
Summary 18
Several methods are available for evaluating and correcting the underestimation of trace gas 19
fluxes measured by eddy covariance (EC) method and due to the design and the setup of the 20
employed equipment Different frequency correction factors (CF) to apply to the measured 21
EC ammonia (NH3) fluxes have been applied using the following approaches (i) the spectral 22
theoretical transfer function (CFTheor) with and without phase shift (ii) the in-situ ogive 23
method (CFO) and (iii) the inductance correction method (CFL) The NH3 fluxes were 24
measured in an experimental field located in south Italy above a sorghum crop submitted to 25
Mediterranean semi-arid climate and fertilized with urea A fast analyser based on Tunable 26
1 Corresponding author present address Agricultural Research Council - Research Unit for cropping systems in
dry environments via C Ulpiani 5 70125 Bari Italy Phone +390805475007 Fax +390805475023
2
Infrared Laser Differential Absorption Spectroscopy (TILDAS) coupled with a quantum 27
cascade (QC) laser was used to measure NH3 concentration at high frequency The results 28
showed that correction of the values of NH3 EC fluxes over natural surfaces are necessary for 29
taking into account the dumping of high frequency contribution due to the coupling of 30
ammonia fast analyser by QC-TILDAS and sonic anemometer In particular the calculated 31
flux losses ranged between -23 (inductance method with the CF threshold selected to 25) 32
and -43 (experimental-ogive method) while the two theoretical transfer function 33
approaches gave comparable loss estimates ie -30 and -31 for the methods with time lag 34
and phase shift respectively 35
36
3
INTRODUCTION 37
Ammonia (NH3) is naturally present in the troposphere with ambient concentrations varying 38
from 5 ppt in remote regions (Sutton et al 2001) to 500 ppb near sources (Krupa 2003) 39
Actually NH3 plays a key role in the environment as the main base neutralizing atmospheric 40
acids and as a precursor of particulate matter which causes human health hazards visibility 41
problems and modifies the radiative forcing of the climate system (Asman et al 1998 IPCC 42
2007) Due to its negative environmental impacts (Galloway et al 2003 Sutton et al 2009) 43
NH3 has become the object of many investigations by the scientific community in 1999 the 44
UN Convention on Long-range Transboundary Air Pollution to Abate Acidification 45
Eutrophication and Ground-level Ozone was extended by the Gothenburg Protocol to include 46
NH3 (httpwwwuneceorgenvlrtap) and its emissions started to be more regulated and 47
monitored 48
Ammonia is mainly emitted by anthropogenic sources It is estimated that the majority of NH3 49
emissions arises from agricultural activities in central Europe agriculture is responsible for 50
about 90 of NH3 release (Erisman et al 2008) However large uncertainty exists on NH3 51
dynamics in the soil-plant-atmosphere continuum due to the scarcity of reliable direct flux 52
measurements This can mainly be attributed to difficulties in measuring fluxes of such a 53
reactive compound (Harper 2005) In particular the first issue is the presence of NH3 in 54
atmosphere in gaseous particulate (eg (NH4)2SO4 and NH4NO3) and liquid (ie NH4OH in 55
cloud and fog droplets) forms where the partition of these three phases is highly variable 56
(Fehsenfeld 1995) Another difficulty is the tendency of NH3 to form strong hydrogen bonds 57
with water and thus to be readily adsorbed on any material exposed to ambient air producing 58
memory effects in the measurements These troubles ndash in addition to the contamination due to 59
NH3 produced by people (Sutton et al 2000) ndash have delayed progress in the development of 60
NH3 measurement techniques with respect to other trace gases 61
Considering its reactive nature and the dependence of NH3 exchanges by environmental 62
factors the micrometeorological methods are the preferred ones due to their non intrusive 63
character (Denmead 1983 Kaimal and Finnigan 1994) The aerodynamic gradient method 64
coupled to wet chemical analysis is the most widely used (Wyers et al 1993 Neftel et al 65
1998 Erisman et al 2005 Kruit et al 2007 Milford et al 2009 Spirig et al 2010) but it is 66
very time consuming Moreover the lack of fast sensors for NH3 has delayed the application 67
of the eddy covariance (EC) method which is considered the most direct and least error-prone 68
approach (Denmead 2008) The evolution of techniques for NH3 monitoring starts with the 69
4
passive filters and denuders developed by Ferm (1979) nowadays the most widely used 70
method for its simplicity and inexpensiveness However the drawbacks of this approach are 71
the relatively low time resolution (minute to hours) and the laboratory intensive labour 72
required for preparing them and for manual off-line sample analysis Coupled to the chemical 73
analytical methods progress in optical methods have been achieved with the development of 74
spectroscopic devices characterized by high time resolution and sensitivity to NH3 but which 75
are still expensive Von Bobrutzki et al (2010) recently provided inter-comparisons for NH3 76
measurement techniques in field highlighting that many challenges still have to be faced for 77
measuring NH3 accurately Nevertheless Tunable Infrared Laser Differential Absorption 78
Spectroscopy (TILDAS) using a quantum cascade (QC) laser are now available for 79
monitoring NH3 concentration with a sufficient sensitivity and robustness to be employed in 80
the field However the application of the EC method for monitoring NH3 is considered a 81
challenge by the scientific community as demonstrated by the scarce literature available on 82
this topic (Shaw et al 1998 Famulari et al 2005 Whitehead et al 2008 Brouder et al 83
2009 Sintermann et al 2011) in which are highlighted the difficulties to its routine 84
applicability In particular the employment of a setup involving a fast closed-path analyser 85
could introduce considerable damping of high frequency fluctuations and consequent 86
underestimation of the fluxes that have to be corrected for (eg Massman 2000) 87
Several methods are available for evaluating and correcting these underestimations when EC 88
equipment is used for trace gas fluxes (Massman and Clement 2004 for CO2 for example) In 89
particular Whitehead et al (2008) evidenced that the correction for EC fluxes of NH3 using 90
TILDAS is fundamental to accurately estimate the real fluxes at different time scales 91
However up to now no attempt has been made to evaluate using different methodologies 92
the correction of the underestimation of EC ammonia fluxes when a QC-TILDAS is used for 93
measuring NH3 fluctuations in the cropped field Furthermore no EC data of NH3 turbulent 94
fluxes are available for crop grown in arid environments 95
In this study we report EC measurements of NH3 fluxes performed employing a fast closed-96
path analyser for NH3 concentration the QC-TILDAS-76 SN002-U developed by Aerodyne 97
Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) under custom specifications 98
The trial was carried out over a field in an intensive agricultural area of south Italy in a semi-99
arid to arid Mediterranean climate The plot was fertilized with urea pellets to stimulate NH3 100
emission and to test the device over a wide range of concentrations The aim was to assess the 101
performance of the QC-TILDAS used for the first time in an EC system for measuring NH3 102
fluxes over an irrigated cropland in Mediterranean environment The challenge was to verify 103
5
the adequacy of the system in terms of setup and quality of the data In order to reach a robust 104
assessment of the needed correction three possible methods for correcting the flux losses due 105
to the design and the setup of the employed equipment were considered in this analysis In 106
particular we compare different frequency correction factors (CF) to apply to the measured 107
EC NH3 fluxes obtained by the following approaches (i) the spectral theoretical transfer 108
function (CFTheor) (ii) the in-situ ogive method (CFO) (Massman and Clement 2004) and 109
(iii) the inductance correction model (CFL) developed by Eugster and Senn (1995) 110
1 MATERIAL AND METHOD 111
11 Field site 112
The experiment was carried out from 17 to 30 July 2008 (DOYs 199 to 211) in Rutigliano 113
(41degN 17deg54 E 122 m asl) near Bari in southern Italy All experimental activities were 114
performed in a 2 hectares field dimensions were 200 m x 100 m The soil type is classified as 115
loamy clay (Clay 41 Silt 44 Sand 15 ) and the terrain is flat The climate is semi-arid 116
Mediterranean with hot and dry summer and short and temperate winter The mean annual 117
temperature is 157 degC with a mean annual precipitation of 600 mm the main wind direction 118
is north during the summer time In this study most of the analysis will focus on the period 119
22-29 July 2008 (DOYs 204-211) where the NH3 fluxes were the largest 120
12 Crop and urea spreading 121
During the experimental campaign the field was cultivated with Sorghum vulgare (cv Hay 122
Day) sowed on 10th
June 2008 and irrigated by sprinkler irrigation A commercial urea 123
product (46 N content) in granular form was applied The total N supply was around 240 124
kg N ha-1
divided into three applications 30 90 and 120 kg N ha-1
on 1st 16
th and 22
nd July 125
2008 respectively The first application was carried out at crop emergence Before the second 126
application the field was irrigated with 79 mm water while further irrigation (93 mm) was 127
applied before and after the third application The amount of urea was slightly higher than 128
current agronomic practices in the area since the dry conditions (high temperature no rain 129
and no irrigation applied) between the second and the last urea spreading inhibited NH3 130
volatilization so that no fluxes were detected The last urea application combined with 131
irrigation was necessary in order to increase the probability to detect NH3 fluxes 132
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
2
Infrared Laser Differential Absorption Spectroscopy (TILDAS) coupled with a quantum 27
cascade (QC) laser was used to measure NH3 concentration at high frequency The results 28
showed that correction of the values of NH3 EC fluxes over natural surfaces are necessary for 29
taking into account the dumping of high frequency contribution due to the coupling of 30
ammonia fast analyser by QC-TILDAS and sonic anemometer In particular the calculated 31
flux losses ranged between -23 (inductance method with the CF threshold selected to 25) 32
and -43 (experimental-ogive method) while the two theoretical transfer function 33
approaches gave comparable loss estimates ie -30 and -31 for the methods with time lag 34
and phase shift respectively 35
36
3
INTRODUCTION 37
Ammonia (NH3) is naturally present in the troposphere with ambient concentrations varying 38
from 5 ppt in remote regions (Sutton et al 2001) to 500 ppb near sources (Krupa 2003) 39
Actually NH3 plays a key role in the environment as the main base neutralizing atmospheric 40
acids and as a precursor of particulate matter which causes human health hazards visibility 41
problems and modifies the radiative forcing of the climate system (Asman et al 1998 IPCC 42
2007) Due to its negative environmental impacts (Galloway et al 2003 Sutton et al 2009) 43
NH3 has become the object of many investigations by the scientific community in 1999 the 44
UN Convention on Long-range Transboundary Air Pollution to Abate Acidification 45
Eutrophication and Ground-level Ozone was extended by the Gothenburg Protocol to include 46
NH3 (httpwwwuneceorgenvlrtap) and its emissions started to be more regulated and 47
monitored 48
Ammonia is mainly emitted by anthropogenic sources It is estimated that the majority of NH3 49
emissions arises from agricultural activities in central Europe agriculture is responsible for 50
about 90 of NH3 release (Erisman et al 2008) However large uncertainty exists on NH3 51
dynamics in the soil-plant-atmosphere continuum due to the scarcity of reliable direct flux 52
measurements This can mainly be attributed to difficulties in measuring fluxes of such a 53
reactive compound (Harper 2005) In particular the first issue is the presence of NH3 in 54
atmosphere in gaseous particulate (eg (NH4)2SO4 and NH4NO3) and liquid (ie NH4OH in 55
cloud and fog droplets) forms where the partition of these three phases is highly variable 56
(Fehsenfeld 1995) Another difficulty is the tendency of NH3 to form strong hydrogen bonds 57
with water and thus to be readily adsorbed on any material exposed to ambient air producing 58
memory effects in the measurements These troubles ndash in addition to the contamination due to 59
NH3 produced by people (Sutton et al 2000) ndash have delayed progress in the development of 60
NH3 measurement techniques with respect to other trace gases 61
Considering its reactive nature and the dependence of NH3 exchanges by environmental 62
factors the micrometeorological methods are the preferred ones due to their non intrusive 63
character (Denmead 1983 Kaimal and Finnigan 1994) The aerodynamic gradient method 64
coupled to wet chemical analysis is the most widely used (Wyers et al 1993 Neftel et al 65
1998 Erisman et al 2005 Kruit et al 2007 Milford et al 2009 Spirig et al 2010) but it is 66
very time consuming Moreover the lack of fast sensors for NH3 has delayed the application 67
of the eddy covariance (EC) method which is considered the most direct and least error-prone 68
approach (Denmead 2008) The evolution of techniques for NH3 monitoring starts with the 69
4
passive filters and denuders developed by Ferm (1979) nowadays the most widely used 70
method for its simplicity and inexpensiveness However the drawbacks of this approach are 71
the relatively low time resolution (minute to hours) and the laboratory intensive labour 72
required for preparing them and for manual off-line sample analysis Coupled to the chemical 73
analytical methods progress in optical methods have been achieved with the development of 74
spectroscopic devices characterized by high time resolution and sensitivity to NH3 but which 75
are still expensive Von Bobrutzki et al (2010) recently provided inter-comparisons for NH3 76
measurement techniques in field highlighting that many challenges still have to be faced for 77
measuring NH3 accurately Nevertheless Tunable Infrared Laser Differential Absorption 78
Spectroscopy (TILDAS) using a quantum cascade (QC) laser are now available for 79
monitoring NH3 concentration with a sufficient sensitivity and robustness to be employed in 80
the field However the application of the EC method for monitoring NH3 is considered a 81
challenge by the scientific community as demonstrated by the scarce literature available on 82
this topic (Shaw et al 1998 Famulari et al 2005 Whitehead et al 2008 Brouder et al 83
2009 Sintermann et al 2011) in which are highlighted the difficulties to its routine 84
applicability In particular the employment of a setup involving a fast closed-path analyser 85
could introduce considerable damping of high frequency fluctuations and consequent 86
underestimation of the fluxes that have to be corrected for (eg Massman 2000) 87
Several methods are available for evaluating and correcting these underestimations when EC 88
equipment is used for trace gas fluxes (Massman and Clement 2004 for CO2 for example) In 89
particular Whitehead et al (2008) evidenced that the correction for EC fluxes of NH3 using 90
TILDAS is fundamental to accurately estimate the real fluxes at different time scales 91
However up to now no attempt has been made to evaluate using different methodologies 92
the correction of the underestimation of EC ammonia fluxes when a QC-TILDAS is used for 93
measuring NH3 fluctuations in the cropped field Furthermore no EC data of NH3 turbulent 94
fluxes are available for crop grown in arid environments 95
In this study we report EC measurements of NH3 fluxes performed employing a fast closed-96
path analyser for NH3 concentration the QC-TILDAS-76 SN002-U developed by Aerodyne 97
Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) under custom specifications 98
The trial was carried out over a field in an intensive agricultural area of south Italy in a semi-99
arid to arid Mediterranean climate The plot was fertilized with urea pellets to stimulate NH3 100
emission and to test the device over a wide range of concentrations The aim was to assess the 101
performance of the QC-TILDAS used for the first time in an EC system for measuring NH3 102
fluxes over an irrigated cropland in Mediterranean environment The challenge was to verify 103
5
the adequacy of the system in terms of setup and quality of the data In order to reach a robust 104
assessment of the needed correction three possible methods for correcting the flux losses due 105
to the design and the setup of the employed equipment were considered in this analysis In 106
particular we compare different frequency correction factors (CF) to apply to the measured 107
EC NH3 fluxes obtained by the following approaches (i) the spectral theoretical transfer 108
function (CFTheor) (ii) the in-situ ogive method (CFO) (Massman and Clement 2004) and 109
(iii) the inductance correction model (CFL) developed by Eugster and Senn (1995) 110
1 MATERIAL AND METHOD 111
11 Field site 112
The experiment was carried out from 17 to 30 July 2008 (DOYs 199 to 211) in Rutigliano 113
(41degN 17deg54 E 122 m asl) near Bari in southern Italy All experimental activities were 114
performed in a 2 hectares field dimensions were 200 m x 100 m The soil type is classified as 115
loamy clay (Clay 41 Silt 44 Sand 15 ) and the terrain is flat The climate is semi-arid 116
Mediterranean with hot and dry summer and short and temperate winter The mean annual 117
temperature is 157 degC with a mean annual precipitation of 600 mm the main wind direction 118
is north during the summer time In this study most of the analysis will focus on the period 119
22-29 July 2008 (DOYs 204-211) where the NH3 fluxes were the largest 120
12 Crop and urea spreading 121
During the experimental campaign the field was cultivated with Sorghum vulgare (cv Hay 122
Day) sowed on 10th
June 2008 and irrigated by sprinkler irrigation A commercial urea 123
product (46 N content) in granular form was applied The total N supply was around 240 124
kg N ha-1
divided into three applications 30 90 and 120 kg N ha-1
on 1st 16
th and 22
nd July 125
2008 respectively The first application was carried out at crop emergence Before the second 126
application the field was irrigated with 79 mm water while further irrigation (93 mm) was 127
applied before and after the third application The amount of urea was slightly higher than 128
current agronomic practices in the area since the dry conditions (high temperature no rain 129
and no irrigation applied) between the second and the last urea spreading inhibited NH3 130
volatilization so that no fluxes were detected The last urea application combined with 131
irrigation was necessary in order to increase the probability to detect NH3 fluxes 132
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
3
INTRODUCTION 37
Ammonia (NH3) is naturally present in the troposphere with ambient concentrations varying 38
from 5 ppt in remote regions (Sutton et al 2001) to 500 ppb near sources (Krupa 2003) 39
Actually NH3 plays a key role in the environment as the main base neutralizing atmospheric 40
acids and as a precursor of particulate matter which causes human health hazards visibility 41
problems and modifies the radiative forcing of the climate system (Asman et al 1998 IPCC 42
2007) Due to its negative environmental impacts (Galloway et al 2003 Sutton et al 2009) 43
NH3 has become the object of many investigations by the scientific community in 1999 the 44
UN Convention on Long-range Transboundary Air Pollution to Abate Acidification 45
Eutrophication and Ground-level Ozone was extended by the Gothenburg Protocol to include 46
NH3 (httpwwwuneceorgenvlrtap) and its emissions started to be more regulated and 47
monitored 48
Ammonia is mainly emitted by anthropogenic sources It is estimated that the majority of NH3 49
emissions arises from agricultural activities in central Europe agriculture is responsible for 50
about 90 of NH3 release (Erisman et al 2008) However large uncertainty exists on NH3 51
dynamics in the soil-plant-atmosphere continuum due to the scarcity of reliable direct flux 52
measurements This can mainly be attributed to difficulties in measuring fluxes of such a 53
reactive compound (Harper 2005) In particular the first issue is the presence of NH3 in 54
atmosphere in gaseous particulate (eg (NH4)2SO4 and NH4NO3) and liquid (ie NH4OH in 55
cloud and fog droplets) forms where the partition of these three phases is highly variable 56
(Fehsenfeld 1995) Another difficulty is the tendency of NH3 to form strong hydrogen bonds 57
with water and thus to be readily adsorbed on any material exposed to ambient air producing 58
memory effects in the measurements These troubles ndash in addition to the contamination due to 59
NH3 produced by people (Sutton et al 2000) ndash have delayed progress in the development of 60
NH3 measurement techniques with respect to other trace gases 61
Considering its reactive nature and the dependence of NH3 exchanges by environmental 62
factors the micrometeorological methods are the preferred ones due to their non intrusive 63
character (Denmead 1983 Kaimal and Finnigan 1994) The aerodynamic gradient method 64
coupled to wet chemical analysis is the most widely used (Wyers et al 1993 Neftel et al 65
1998 Erisman et al 2005 Kruit et al 2007 Milford et al 2009 Spirig et al 2010) but it is 66
very time consuming Moreover the lack of fast sensors for NH3 has delayed the application 67
of the eddy covariance (EC) method which is considered the most direct and least error-prone 68
approach (Denmead 2008) The evolution of techniques for NH3 monitoring starts with the 69
4
passive filters and denuders developed by Ferm (1979) nowadays the most widely used 70
method for its simplicity and inexpensiveness However the drawbacks of this approach are 71
the relatively low time resolution (minute to hours) and the laboratory intensive labour 72
required for preparing them and for manual off-line sample analysis Coupled to the chemical 73
analytical methods progress in optical methods have been achieved with the development of 74
spectroscopic devices characterized by high time resolution and sensitivity to NH3 but which 75
are still expensive Von Bobrutzki et al (2010) recently provided inter-comparisons for NH3 76
measurement techniques in field highlighting that many challenges still have to be faced for 77
measuring NH3 accurately Nevertheless Tunable Infrared Laser Differential Absorption 78
Spectroscopy (TILDAS) using a quantum cascade (QC) laser are now available for 79
monitoring NH3 concentration with a sufficient sensitivity and robustness to be employed in 80
the field However the application of the EC method for monitoring NH3 is considered a 81
challenge by the scientific community as demonstrated by the scarce literature available on 82
this topic (Shaw et al 1998 Famulari et al 2005 Whitehead et al 2008 Brouder et al 83
2009 Sintermann et al 2011) in which are highlighted the difficulties to its routine 84
applicability In particular the employment of a setup involving a fast closed-path analyser 85
could introduce considerable damping of high frequency fluctuations and consequent 86
underestimation of the fluxes that have to be corrected for (eg Massman 2000) 87
Several methods are available for evaluating and correcting these underestimations when EC 88
equipment is used for trace gas fluxes (Massman and Clement 2004 for CO2 for example) In 89
particular Whitehead et al (2008) evidenced that the correction for EC fluxes of NH3 using 90
TILDAS is fundamental to accurately estimate the real fluxes at different time scales 91
However up to now no attempt has been made to evaluate using different methodologies 92
the correction of the underestimation of EC ammonia fluxes when a QC-TILDAS is used for 93
measuring NH3 fluctuations in the cropped field Furthermore no EC data of NH3 turbulent 94
fluxes are available for crop grown in arid environments 95
In this study we report EC measurements of NH3 fluxes performed employing a fast closed-96
path analyser for NH3 concentration the QC-TILDAS-76 SN002-U developed by Aerodyne 97
Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) under custom specifications 98
The trial was carried out over a field in an intensive agricultural area of south Italy in a semi-99
arid to arid Mediterranean climate The plot was fertilized with urea pellets to stimulate NH3 100
emission and to test the device over a wide range of concentrations The aim was to assess the 101
performance of the QC-TILDAS used for the first time in an EC system for measuring NH3 102
fluxes over an irrigated cropland in Mediterranean environment The challenge was to verify 103
5
the adequacy of the system in terms of setup and quality of the data In order to reach a robust 104
assessment of the needed correction three possible methods for correcting the flux losses due 105
to the design and the setup of the employed equipment were considered in this analysis In 106
particular we compare different frequency correction factors (CF) to apply to the measured 107
EC NH3 fluxes obtained by the following approaches (i) the spectral theoretical transfer 108
function (CFTheor) (ii) the in-situ ogive method (CFO) (Massman and Clement 2004) and 109
(iii) the inductance correction model (CFL) developed by Eugster and Senn (1995) 110
1 MATERIAL AND METHOD 111
11 Field site 112
The experiment was carried out from 17 to 30 July 2008 (DOYs 199 to 211) in Rutigliano 113
(41degN 17deg54 E 122 m asl) near Bari in southern Italy All experimental activities were 114
performed in a 2 hectares field dimensions were 200 m x 100 m The soil type is classified as 115
loamy clay (Clay 41 Silt 44 Sand 15 ) and the terrain is flat The climate is semi-arid 116
Mediterranean with hot and dry summer and short and temperate winter The mean annual 117
temperature is 157 degC with a mean annual precipitation of 600 mm the main wind direction 118
is north during the summer time In this study most of the analysis will focus on the period 119
22-29 July 2008 (DOYs 204-211) where the NH3 fluxes were the largest 120
12 Crop and urea spreading 121
During the experimental campaign the field was cultivated with Sorghum vulgare (cv Hay 122
Day) sowed on 10th
June 2008 and irrigated by sprinkler irrigation A commercial urea 123
product (46 N content) in granular form was applied The total N supply was around 240 124
kg N ha-1
divided into three applications 30 90 and 120 kg N ha-1
on 1st 16
th and 22
nd July 125
2008 respectively The first application was carried out at crop emergence Before the second 126
application the field was irrigated with 79 mm water while further irrigation (93 mm) was 127
applied before and after the third application The amount of urea was slightly higher than 128
current agronomic practices in the area since the dry conditions (high temperature no rain 129
and no irrigation applied) between the second and the last urea spreading inhibited NH3 130
volatilization so that no fluxes were detected The last urea application combined with 131
irrigation was necessary in order to increase the probability to detect NH3 fluxes 132
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
4
passive filters and denuders developed by Ferm (1979) nowadays the most widely used 70
method for its simplicity and inexpensiveness However the drawbacks of this approach are 71
the relatively low time resolution (minute to hours) and the laboratory intensive labour 72
required for preparing them and for manual off-line sample analysis Coupled to the chemical 73
analytical methods progress in optical methods have been achieved with the development of 74
spectroscopic devices characterized by high time resolution and sensitivity to NH3 but which 75
are still expensive Von Bobrutzki et al (2010) recently provided inter-comparisons for NH3 76
measurement techniques in field highlighting that many challenges still have to be faced for 77
measuring NH3 accurately Nevertheless Tunable Infrared Laser Differential Absorption 78
Spectroscopy (TILDAS) using a quantum cascade (QC) laser are now available for 79
monitoring NH3 concentration with a sufficient sensitivity and robustness to be employed in 80
the field However the application of the EC method for monitoring NH3 is considered a 81
challenge by the scientific community as demonstrated by the scarce literature available on 82
this topic (Shaw et al 1998 Famulari et al 2005 Whitehead et al 2008 Brouder et al 83
2009 Sintermann et al 2011) in which are highlighted the difficulties to its routine 84
applicability In particular the employment of a setup involving a fast closed-path analyser 85
could introduce considerable damping of high frequency fluctuations and consequent 86
underestimation of the fluxes that have to be corrected for (eg Massman 2000) 87
Several methods are available for evaluating and correcting these underestimations when EC 88
equipment is used for trace gas fluxes (Massman and Clement 2004 for CO2 for example) In 89
particular Whitehead et al (2008) evidenced that the correction for EC fluxes of NH3 using 90
TILDAS is fundamental to accurately estimate the real fluxes at different time scales 91
However up to now no attempt has been made to evaluate using different methodologies 92
the correction of the underestimation of EC ammonia fluxes when a QC-TILDAS is used for 93
measuring NH3 fluctuations in the cropped field Furthermore no EC data of NH3 turbulent 94
fluxes are available for crop grown in arid environments 95
In this study we report EC measurements of NH3 fluxes performed employing a fast closed-96
path analyser for NH3 concentration the QC-TILDAS-76 SN002-U developed by Aerodyne 97
Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) under custom specifications 98
The trial was carried out over a field in an intensive agricultural area of south Italy in a semi-99
arid to arid Mediterranean climate The plot was fertilized with urea pellets to stimulate NH3 100
emission and to test the device over a wide range of concentrations The aim was to assess the 101
performance of the QC-TILDAS used for the first time in an EC system for measuring NH3 102
fluxes over an irrigated cropland in Mediterranean environment The challenge was to verify 103
5
the adequacy of the system in terms of setup and quality of the data In order to reach a robust 104
assessment of the needed correction three possible methods for correcting the flux losses due 105
to the design and the setup of the employed equipment were considered in this analysis In 106
particular we compare different frequency correction factors (CF) to apply to the measured 107
EC NH3 fluxes obtained by the following approaches (i) the spectral theoretical transfer 108
function (CFTheor) (ii) the in-situ ogive method (CFO) (Massman and Clement 2004) and 109
(iii) the inductance correction model (CFL) developed by Eugster and Senn (1995) 110
1 MATERIAL AND METHOD 111
11 Field site 112
The experiment was carried out from 17 to 30 July 2008 (DOYs 199 to 211) in Rutigliano 113
(41degN 17deg54 E 122 m asl) near Bari in southern Italy All experimental activities were 114
performed in a 2 hectares field dimensions were 200 m x 100 m The soil type is classified as 115
loamy clay (Clay 41 Silt 44 Sand 15 ) and the terrain is flat The climate is semi-arid 116
Mediterranean with hot and dry summer and short and temperate winter The mean annual 117
temperature is 157 degC with a mean annual precipitation of 600 mm the main wind direction 118
is north during the summer time In this study most of the analysis will focus on the period 119
22-29 July 2008 (DOYs 204-211) where the NH3 fluxes were the largest 120
12 Crop and urea spreading 121
During the experimental campaign the field was cultivated with Sorghum vulgare (cv Hay 122
Day) sowed on 10th
June 2008 and irrigated by sprinkler irrigation A commercial urea 123
product (46 N content) in granular form was applied The total N supply was around 240 124
kg N ha-1
divided into three applications 30 90 and 120 kg N ha-1
on 1st 16
th and 22
nd July 125
2008 respectively The first application was carried out at crop emergence Before the second 126
application the field was irrigated with 79 mm water while further irrigation (93 mm) was 127
applied before and after the third application The amount of urea was slightly higher than 128
current agronomic practices in the area since the dry conditions (high temperature no rain 129
and no irrigation applied) between the second and the last urea spreading inhibited NH3 130
volatilization so that no fluxes were detected The last urea application combined with 131
irrigation was necessary in order to increase the probability to detect NH3 fluxes 132
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
5
the adequacy of the system in terms of setup and quality of the data In order to reach a robust 104
assessment of the needed correction three possible methods for correcting the flux losses due 105
to the design and the setup of the employed equipment were considered in this analysis In 106
particular we compare different frequency correction factors (CF) to apply to the measured 107
EC NH3 fluxes obtained by the following approaches (i) the spectral theoretical transfer 108
function (CFTheor) (ii) the in-situ ogive method (CFO) (Massman and Clement 2004) and 109
(iii) the inductance correction model (CFL) developed by Eugster and Senn (1995) 110
1 MATERIAL AND METHOD 111
11 Field site 112
The experiment was carried out from 17 to 30 July 2008 (DOYs 199 to 211) in Rutigliano 113
(41degN 17deg54 E 122 m asl) near Bari in southern Italy All experimental activities were 114
performed in a 2 hectares field dimensions were 200 m x 100 m The soil type is classified as 115
loamy clay (Clay 41 Silt 44 Sand 15 ) and the terrain is flat The climate is semi-arid 116
Mediterranean with hot and dry summer and short and temperate winter The mean annual 117
temperature is 157 degC with a mean annual precipitation of 600 mm the main wind direction 118
is north during the summer time In this study most of the analysis will focus on the period 119
22-29 July 2008 (DOYs 204-211) where the NH3 fluxes were the largest 120
12 Crop and urea spreading 121
During the experimental campaign the field was cultivated with Sorghum vulgare (cv Hay 122
Day) sowed on 10th
June 2008 and irrigated by sprinkler irrigation A commercial urea 123
product (46 N content) in granular form was applied The total N supply was around 240 124
kg N ha-1
divided into three applications 30 90 and 120 kg N ha-1
on 1st 16
th and 22
nd July 125
2008 respectively The first application was carried out at crop emergence Before the second 126
application the field was irrigated with 79 mm water while further irrigation (93 mm) was 127
applied before and after the third application The amount of urea was slightly higher than 128
current agronomic practices in the area since the dry conditions (high temperature no rain 129
and no irrigation applied) between the second and the last urea spreading inhibited NH3 130
volatilization so that no fluxes were detected The last urea application combined with 131
irrigation was necessary in order to increase the probability to detect NH3 fluxes 132
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
6
13 NH3 fluxes by the eddy covariance method 133
The fluxes of NH3 were measured with the eddy covariance method (eg Kaimal amp Finnigan 134
1994 Lee et al 2004 Mahrt 2010) The EC method is based on estimating the transport of a 135
scalar by turbulent motions of upward and downward moving air contained in the atmosphere 136
The vertical flux (F) of scalar at the measurement height is given by 137
wwF
(1) 138
where w is the component of the wind speed normal to the surface and β is the scalar 139
concentration and the usual Reynolds decomposition is used where prime denotes fluctuating 140
quantities w is the covariance between and w With the usual assumption of steady state 141
horizontally homogeneous flow there is no vertical flux of dry air ie 0wcd where cd is 142
the dry air concentration Since the dry air concentration fluctuates due to heat and water 143
vapour fluxes w is not necessarily equal to zero The Webb Pearmann Leuning (WPL) 144
correction (Webb et al 1980 Leuning 2007) gives the expression of w The impact of this 145
term depends on the concentration of the gas and the typical magnitude of the flux (Denmead 146
1983 Liebethal and Foken 2004) for trace gas with small concentration and large fluxes 147
such as NH3 over a source these corrections are negligible 148
The turbulent flux w as a time-averaged quantity can also be viewed as a frequency 149
averaged quantity ie the integral of the co-spectral density (Cowβ) (Kaimal and Finnigan 150
1994 Aubinet et al 2000 Massman 2000) 151
0
)( dffCow w
(2)
152
To catch all relevant turbulent fluctuations β and w should be measured with a sampling 153
frequency much larger than the frequency of the eddies carrying the energy of the signal (fx) 154
(Horst 1997 Massman 2000) Usually a sampling frequency of 10 Hz is considered 155
sufficient above a vegetated surface However flux loss is inevitable with any EC system 156
especially employing closed-path analysers to which the air sample is transported into the 157
measuring cell by means of a tube The mixing effects in the tube and additional possible 158
chemical or micro-physical processes occurring in the system produce low-pass filtering 159
effects on the measured covariances (Ibrom et al 2007) Nowadays a variety of methods are 160
available to correct these underestimated measured fluxes by means of frequency response 161
correction factors (eg Ammann et al 2006 Shimizu 2007 Aubinet et al 2000 Massman 162
2000 Eugster and Senn 1995) However in this work since the adopted closed-path analyser 163
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
7
for NH3 concentration measurements is relatively new and it is the first time it is being used 164
under semi-arid climates we compare a variety of methods to correct the NH3 fluxes 165
measured over an agricultural plot 166
14 NH3 concentration measurements 167
141 The QC-TILDAS analyser 168
1411 The instrument 169
The QC-TILDAS employed during this research is the compact QC-TILDAS-76 SN002-U 170
developed by Aerodyne Research Inc (ARI Billerica MA USA) (Zahniser et al 2005) 171
This spectrometer combines QC laser designed for pulsed operation at near room temperature 172
(Alpes Lasers Neuchacirctel Switzerland) an optical system together with a computer-controlled 173
system that incorporates the electronics for driving the QC laser along with signal generation 174
and a data logger Molecular transition selection for given laser is achieved by temperature 175
tuning typically within a small range around 40 ordmC using a two-stage Peltier element The 176
laser frequency is swept across the full molecular transition at 96734 cm-1
which means that 177
the scan is done over a narrow range of wave numbers chosen to contain an absorption feature 178
of the trace gas of interest The optical system collects light from the QC laser and directs it 179
through an astigmatic Herriott type multi-pass absorption cell (76 m effective path length) 180
onto a TE-cooled photovoltaic detector (Vigo Systems) A detailed description of the device 181
can be found in Zahniser et al (2005) 182
1412 The inlet design and the sampling 183
Ellis et al (2010) report the characterization of the QC-TILDAS by laboratory tests In 184
particular they reported a detailed analysis on the performance of the QC-TILDAS in 185
function of the inlet design To achieve accurate atmospheric measurements of NH3 the 186
sampling setup has to reduce gas interactions with surfaces then it is crucial the configuration 187
of the devicersquos inlet through which the air is drawn into the optical cell The inlet used is a 188
short glass tube containing a critical orifice needed to keep the air flow rate constant and to 189
reduce the pressure in the optical cell to approximately 540 kPa a compromise between 190
minimising line broadening and adsorption effects maximising time response and achieving 191
good sensitivity (Warland et al 2001) The setup of the inlet is done to remove particles in 192
the air sample relying on inertia to remove more than 50 of particles larger than 300 nm in 193
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
8
particular the flow is split into two branches 90 of the flow makes a sharp turn and is 194
pulled into the optical cell by a pump (VARIAN TriScroll 600 Series) the other 10 is 195
pulled by a second air pump (VARIAN mod SH110) This inlet design is optimized to 196
eliminate the need of filters and any interference they may cause The operation of the 197
spectrometer is fully automated and computer-controlled through a Windows-based 198
proprietary software (TDL Wintel) developed at ARI Concentrations are real-time 199
determined from the light absorption spectra through a non-linear least-squares fitting 200
algorithm (Levenberg-Marquardt) that uses spectral parameters from the HITRAN (high-201
resolution transmission) molecular absorption database (Rothman et al 2003) The pressure 202
and temperature in the multi-pass cell are continuously monitored to supply information 203
needed for spectral fitting The software calculates absolute concentrations so no calibration 204
was performed in the field However laboratory tests demonstrate the need to carry out a 205
post-proceeding calibration in order to overcome the failure of the analyzer to reproduce the 206
values of the ammonia standard Moreover we would justify the calibration factor found for 207
the QC-TILDAS using ethilene standard which has absorbing frequency close to ammonia 208
one but it is not affected by stickiness Part of the problem with absolute spectroscopic values 209
for NH3 in our instrument was due to using pulsed-QCLs rather than CW-QCLs with the 210
laser linewidth considerably greater than the molecular linewidth the spectral analysis 211
required much effort to get absolute values Thus a post-proceeding calibration was done and 212
zeroing was routinely carried out in order to optimize the calculated spectra following 213
Whitehead et al (2008) 214
1413 The response time 215
The decreasing of the mixing ratio of the ammonia was well represented by two exponential 216
decay functions which in Ellis et al (2010) is written as 217
2
02
1
010 expexp
ttA
ttAyy (3) 218
where A1 and A2 are constants and their sum is equal to the stable mixing ratio of NH3 prior to 219
the calibration flow being switched off y0 is equal to the NH3 mixing ratio reached at the end 220
of the decay t is time and t0 is the time at the start of the fit τ1 and τ2 are time decay 221
constants the first of which is fast and corresponds to the gas exchange time of the system 222
(04 s in our case) and the second being much slower and corresponding to interactions of 223
NH3 with surfaces on the interior surfaces of the inlet sampling lines and absorption cell (15 224
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
9
s) In this case the response time was checked in laboratory and it is controlled by pumping 225
speed and by adsorptiondesorption on the walls In particular it is affected by the flow rate 226
the length and the heating of the sampling tubes Therefore acting on these parameters would 227
allow optimizing the instrument for measuring the fluxes by EC technique 228
Moreover for a mixing ratio of 350 ppb the amount of noise (1) in a 10 Hz measurement 229
obtained by Ellis et al (2010) by an Allan variance analysis was 315 ppb (22 microg NH3 m-3
) 230
Based on statistics obtained for sensible heat CO2 and temperature w w was estimated 231
to be around 023 plusmn 013 (mean plusmn SE) and taking ~ 2 microg NH3 m-3
and w ~ 125 u would 232
mean that the flux detection limit of the setup would be around 25 u (in microg NH3 m-2
s-1
) 233
hence typically around 750 ng NH3 m-2
s-1
However most of the turbulent signal is 234
transported by eddies with a frequency around 02 Hz (evaluated as the co-spectral peak 235
frequency for air temperature (Ta) CO2 and water vapour (H2O)) At this frequency the Allan 236
variance of the QC-TILDAS found by Ellis et al (2010) is around 02 microg NH3 m-3
and hence 237
the detection limit should be on the order of 025 u (typically 75 ng NH3 m-2
s-1
) As reported 238
by Ellis et al (2010) there is an increase in the relative importance of NH3 surface 239
interactions at mixing ratios lower than 100 ppb which further increased when sampling 240
humid air as opposed to dry nitrogen Then the heating in the sampling line to prevent 241
condensation is crucial to perform accurate NH3 measurements There was no line heating 242
used in our study since we assumed that this was not needed since the air temperature 243
encountered in the field was very high (up to 33degC) On the other hand under such 244
conditions it was required the use of an air conditioner for the box containing the QC-245
TILDAS to limit instrumental drift induced by changes in the optical alignment as well as in 246
the laser power beam width and temperature 247
142 The ALPHA badges and ROSAA system 248
The NH3 concentration measured with the QC-TILDAS were compared against passive 249
diffusion samples (ALPHA badges Tang et al 2001) These ALPHA badges were 250
positioned at the same height as the QC-TILDAS inlet and changed twice a day to get daily 251
and nightly concentrations Three replicates were used to evaluate the uncertainty in the 252
measurements The background concentration bgd was estimated with three sets of badges put 253
at three locations at more than 100 100 and 300 m away from the field and changed weekly 254
The badges were prepared with filters coated with citric acid The filters were extracted in 3 255
ml double deionised water and analysed with a FLORRIA analyser (Mechatronics NL) In 256
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
10
addition during the trial NH3 concentration was monitored by means of the new device 257
ROSAA developed at INRA (Loubet et al 2010 2011) It relies on the coupling of NH3 258
capturing by means of an acid stripping solution (wet-effluent denuders) to convert NH3 in 259
NH4+ and the analysis of the NH4
+ concentration by means of a detector based on a semi-260
permeable membrane and a unit to perform temperature controlled conductivity 261
measurements (ECN Netherlands) In the ROSAA system the three denuders were sampling 262
at three heights (at z = 05 07 and 14 m) above ground and the averaged concentration was 263
measure every 30 minutes 264
15 The eddy covariance system description of the equipment 265
The EC system was equipped with the QC-TILDAS and a three axis ultrasonic anemometer 266
(Gill R2 Gill Instruments Ltd UK) The analogue signals from the QC-TILDAS were passed 267
to the sonic anemometer which uses an analogue-to-digital converter to digitise the signal at 268
10 Hz The sonic anemometer acquired data at 168 Hz internally and reported average values 269
at 208 Hz The digitised signals from the QC-TILDAS were then combined with the wind 270
velocity information and sent to the serial port of a personal computer NH3 fluxes were 271
computed offline with a time step of half an hour The software package EddySoft (Kolle and 272
Rebmann 2007) was used for the data acquisition The same software package was used for 273
post processing EC data and to obtain the cospectra useful for applying the TF method while 274
custom Matlabreg and R scripts were used to process and analyse data for the ogive and the 275
inductance methods respectively Trace gas and heat fluxes were calculated as well as other 276
micrometeorological parameters The components of the wind velocity were rotated every 30 277
minutes to set w and v to zero No detrending was applied and block averaging was instead 278
used (Leuning 2007) 279
The sonic anemometer was mounted on a 15 m tall mast so that the height of the EC sensors 280
was about 07 m above the top of the crop and adjusted to follow crop development Ambient 281
air close to the sonic anemometer head was continuously sampled through a 25 m long inlet 282
line (PFA 38rdquo outer diameter) leading to the QC-TILDAS The flow of air through the tube 283
was measured before and after the T connection used for splitting the flow the flow rate 284
towards the sampling cell was about 9 litresminute resulting in a laminar flow regime with a 285
Reynolds number (Re) of about 2000 below the turbulence limit (for pipe flow Re=2300 286
Shimizu 2007) which was the primary cause of the high frequency loss in the measuring 287
system (Eugster and Senn 1995) The limits in the power supply lines in our experimental 288
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
11
site have prevented the use of pumps able to provide a turbulent regime into the sampling 289
tube 290
The eddy covariance system together with ROSAA and ALPHA samples equipments were 291
located close to the centre of the field which provided an acceptable upwind fetch (around 292
120 m) in the prevalent wind direction during the trial anyway the footprint of EC 293
measurements was found to be always inside the experimental field 294
16 Spectral correction methods for low-pass filtering in a closed-path EC system 295
As mentioned before all eddy covariance systems tend to underestimate the true atmospheric 296
fluxes due to physical limitations of the flux measurement instrumentation which causes 297
losses of eddy flux contributions at low and high frequencies These losses can be quantified 298
and corrected for by multiplying the measured covariance (subscript m) with a frequency 299
response correction factor (CF ge1) 300
mwCFw (4) 301
The correction factor can be estimated by means of spectral correction methods that can be 302
divided into two families the spectral theoretical transfer function (TF) methods and the in-303
situ methods (Massman and Clement 2004) The general approach consists of three steps 1) 304
the identification of a spectral transfer function (theoretical or experimental) 2) its use to 305
calculate a correction factor and 3) the parameterisation of the correction factor with 306
meteorological variables mainly the wind speed which is directly linked to the turbulence 307
Here three methods have been employed in order to estimate CFs to apply at our measured 308
EC NH3 fluxes (i) Theoretical TF analysis CFTheor (ii) A variant of In situ experimental TF 309
method the Ogive method CFO (iii) the inductance method CFL These methodologies are 310
the most analysed in the literature for evaluating the most accurate CFs for gas trace fluxes 311
over natural surfaces 312
Common to all approaches is the requirement of a measured or evaluated reference 313
cospectrum for scalar as well as a measured cospectrum and the estimate of the flux loss as a 314
function of frequency In any case an important aspect relative to any EC system is the time 315
delay between the instant open-path w measurement and the closed-path scalar concentration 316
measurement This time lag is due to the transport time in the tubing system and the phase 317
shift due to low-pass filtering (Ibrom et al 2007) It depends on the tube length and the flow 318
rate it may change slightly due to changing flow rates in the tube caused by pumping 319
variation or density changes Usually this delay can be calculated empirically by determining 320
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
12
the maximum of the covariance function between wrsquo and rsquo (Moncrieff et al 1997) which 321
includes the time lag due to the longitudinal (streamwise) separation distance from the air 322
sample inlet to the sonic anemometer depending on wind speed and direction The 323
individually determined time delays for each 30 minutes set of data can be used for 324
calculating the cospectra in this case the phase shift due to low-pass filtering and longitudinal 325
sensor separation has already been corrected for (Ibrom et al 2007) So two subcategories 326
can be identified in the theoretical transfer function analysis with (CFTheor_time_lag) and without 327
the time lag (CFTheor_phase_shift) applied to computation of the covariances and cospectra 328
329
161 Theoretical TF analysis 330
Following Moore (1986) theoretical transfer functions characterizing the measurement 331
process have been developed The correction factor in this case is summarized by Massman 332
and Clement (2004) using the following equation 333
dffCofHfT
fT
dffCo
w
wCF
w
b
b
w
m
TF
)()(sin
1
)(
0
2
2
0
(5)
334
where mw is the measured covariance and w is the true flux H(f) is the product of all 335
the appropriate transfer functions associated with HF losses (
n
i
iTFfH1
)( ) )( fCow is the 336
one-side reference cospectrum Tb is the block averaging period 22sin1 bb fTfT is the 337
TF for the low frequency spectral attenuation The reference cospectrum is either taken from 338
literature (Kaimal and Finnigan 1994) or a site specific reference cospectrum usually derived 339
from sensible heat flux measurements The requirements to apply this method are the TFs 340
and a model of the cospectra which is the weakness of this approach since smooth cospectra 341
are not typical of the atmospheric surface layer (Massman and Clement 2004) In particular 342
in this work we used the reference cospectral models (eq 6-9) given by Moncrieff et al 343
(1997) where udzffk )( is normalized frequency f is the natural frequency u is 344
horizontal wind speed z-d is height above the zero-plane displacement d For stable 345
atmospheric condition the reference cospectrum is 346
12)(
kww
kw
fBA
fffCo
(6)
347
where 348
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
13
750
4612840
L
dzAw (7)
349
and 350
11342 ww AB (8)
351
While for unstable atmospheric conditions the reference cospectrum is 352
540
7261
9212)(
3751
k
k
kw ffor
f
fffCo (9a)
353
540
831
3784)(
42
k
k
kw ffor
f
fffCo (9b)
354
Moore (1986) developed the first theoretical TFs for EC systems taking into account the 355
attenuation due to the dynamic frequency response of the sensors the scalar path averaging 356
the sensor separation the sensor response mismatch while Moncrieff et al (1997) gave a 357
complete description of TFs of EC systems with closed-path analysers Then considering the 358
wide literature on the used theoretical TFs (see Shimizu 2007 for a review) and the weight of 359
each TF on the total one for our equipment set-up here we report only details about TFs 360
related to (i) the lateral separation (ie perpendicular to the mean wind direction) between 361
the sonic anemometer and QC-TILDAS inlet (TFs Moore 1986) (ii) the tube damping effect 362
(TFtube Lenschow and Raupach 1991) (iii) the volume averaging in a measurement cell 363
(TFvol Massman 2004) (iv) the total phase shift (TFphase_shift Massman 2000) In detail 364
(i) Following Moore (1986) the effect of laterallongitudinal separation s between two 365
sensors can be represented by the following cospectral transfer function 366
)99exp(99 51nTFs (10)
367
where n=fsu As suggested by Massman (2000) the horizontal separation between the two 368
sensors has been set to 030 m 369
(ii) The function TFtube describing the attenuation of the fluctuation concentrations down the 370
sampling tube in which there is a laminar flow for covariance calculation (Lenschow and 371
Raupach 1991 Aubinet et al 2000) is 372
ts
t
tubeUD
LfrTF t
12exp
222 (11)
373
Here rt is the tube radius and Ds the molecular diffusivity of the scalar around 02 cm2s
-1 for 374
NH3 (Massman 1998) Moreover the attenuation may vary with time Leuning and Judd 375
(1996) showed that the attenuation to water vapour by aged and dirty tubes is much greater 376
than that by new and clear tube In the case of laminar flow through the tube the theoretical 377
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
14
cut off frequency fco_t at which the TF equals 2-12
is given by 378
tt
sttco
Lr
DUf
2_ 6490 (12)
379
about 14 Hz in our EC system To minimizing HF losses due to nonuniform velocity profiles 380
turbulent flow would be preferred (Lenschow and Raupach 1991) This requires a powerful 381
pump with high power consumption and a tremendous start-up current (13 A at 230 V AC for 382
an Edwards Tri-Scroll pump for example pers comm W Eugster) that was unrealistic at our 383
field site Hence we had to use a less powerful pump and thus need to account for the extra 384
loss in frequency response due to laminar flow in the intake hose (Moncrieff et al 1997) The 385
possible tube damping correction does not correct potential absorptiondesorption of moisture 386
by hygroscopic dirt particles in the tube however a strong pressure drop along the intake 387
hose reduces risk of condensation on the walls of the tube On the other hand HF due to wall 388
interaction in the sample tube could be reduced by minimizing lengthrsquos tube and maintaining 389
high flow rate In general the line loss is a function of tubing type and how dirty the tube is 390
(Sukyer and Verma 1993) 391
(iii) As reported by Massman (2004) in a closed-path detector the flow velocity within the 392
detection chamber can be very different than the wind speed of the ambient atmosphere near 393
the tube inlet so that the transfer functions need to be expressed in terms of the volume 394
flushing time constant of the detection chamber τvol 395
2
2sin 2
vol
volvolTF
(13)
396
where f 2 and τvol is equal to the time needed to fill the chamber and its minimum value 397
is given by the ratio between the volume of the chamber (05 l) and the maximum flow rate 398
(iv) However if covariances and the cospectra are calculated without the time lag that means 399
without the method of maximum covariance function or if there is no clear time lag due to 400
lag depending on the frequency of the eddies (sensor separation in the wind direction) or if 401
there are complex adsorption and desorption processes it is necessary to take into account the 402
phase shift (Massman 2000) The TF associated with the total phase shift (Massman 2000) 403
is 404
t
tlon
t
tlonshiftphase
U
L
u
l
U
L
u
lTF sincos_ (14)
405
where llon is the longitudinal sensor separation (function of wind direction) and β is the 406
intrinsic time constant of the scalar sensor This latter was estimated by using the Boylersquos law 407
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
15
applied to the measurement chamber working at 33 kPa and using the measured flow rate it 408
is about 01 s and moreover it means that the sampling frequency of 10 Hz is adequate for 409
the employed QC-TILDAS in the EC system 410
For the theoretical TF approach cospectra of the vertical wind speed and T and NH3 411
respectively were calculated using the post processing EddySpec package of Eddysoft 412
software using the following setup number of frequencies 4096 for calculating the cospectra 413
for half-hourly runs multiplied by frequency tapering the time series with a Hamming 414
window (Kaimal and Finnigan 1994) and averaging into 20 logarithmically spaced bins 415
Either the time lag or the phase shift was used in the functions of the two approaches 416
CFTheor_time_lag and CFTheor_phase_shift respectively 417
418
162 In situ ogive method 419
The limitation of the theoretical TF approach is that it could miss additional chemical or 420
micro-physical process occurring in the EC system a restriction that is expected to be 421
overcome by the in situ experimental methods Moreover the experimental approach is based 422
on the assumption of cospectral similarity between scalar fluxes and unlike the theoretical TF 423
approach these methods do not require smooth models of atmospheric cospectra (Massman 424
and Clement 2004) The correction factor can be estimated as the ratio between the 425
attenuated flux and a reference flux (usually the sensible heat flux) Aubinet et al (2000) give 426
the procedure to estimate this experimental frequency transfer function but since the 427
cospectra derived from sensible heat fluxes are often not well-defined then direct application 428
of this approach can produce implausible values 429
Starting from the hypothesis of similarity between all scalars cospectra Ammann et al (2006) 430
developed a method for estimating HF losses Applied to EC measurement of NH3 fluxes the 431
method consists in fitting the ogive for NH3 OgwNH3 (the ogive is the integrated cospectra 432
CowNH3 from low to high frequencies) against the ogive of air temperature OgwTa The fitting 433
was made with an iterative linear least square method in which the points being too far away 434
from the regression line were eliminated progressively (robustfit function under Matlabreg) 435
The high-frequency damping correction (CFO) is simply evaluated as the value of the fitted 436
NH3 ogive (OgwNH3) at the highest frequency The frequency band over which the linear 437
regression is performed is determined as the frequency for which OgwT larger than 01 and 438
OgwNH3 is smaller than 08 This choice was arbitrary and may not be appropriate if the 439
damping of the NH3 signal was below the range chosen The ogives were calculated using 440
detrended signals 441
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
16
442
163 The inductance method 443
Eugster and Senn (1995) propose a correction model for EC data affected by systematic error 444
introduced by the combination of all sources of damping which also includes chemical 445
reaction (in the case of NO2) besides the possible damping due to laminar flow in the 446
sampling tube and sensor separation This damping loss is described by an analogy to 447
inductance L in an alternating current circuit which can be used for taking into account the 448
reduction of spectral density in the inertial subrange This model follows the approach of 449
Moore (1986) using electronic components in a current circuit instead of a gain function 450
As reported by Eugster and Senn (1995) the covariance of the measured (ie damped) scalar 451
flux is given by 452
02220
)(41
1)(
0dffCo
LfdffCow
LL wwL
(15)
453
where )(0 fCo
Lw is the undamped cospectrum The value of L is modified in order to fit the 454
measured cospectrum The independent variables in this correction model are the measuring 455
height above zero-plane displacement z-d the mean horizontal wind speed u the Monin-456
Obukhov stability parameter ζ the natural frequency f and the inductance L The value for 457
L can be derived from spectral and cospectral analysis and is assumed not vary significantly 458
as long as the measuring system is not altered 459
The damped cospectra can be derived combining equation (14) with the empirical formula of 460
Kaimal et al (1972) for stable unstable and neutral stratification as modified by Eugster and 461
Senn (1995) (equations 29 and 30 in their paper) The approximated solutions of Eugster and 462
Senn (1995) for CF range between 10 (no damping at all) and infin (no signal at all) but they 463
suggest to restrict the fit to the range between CF=10 and CF=25 that means all 464
measurements which represent less than 40 of the real fluxes were discarded 465
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
17
2 RESULTS AND DISCUSSION 466
21 Micrometeorological conditions 467
During the trial the air temperature ranged from 16degC to 335degC and averaged 247degC while 468
the soil surface temperature reached more than 546degC at hourly scale The sky was clear 469
most of the period apart from the 23-24072008 where there was little rain (82 mm) The 470
global solar radiation was following the fair weather diurnal course throughout the period and 471
reached 950 W m-2
at maximum The daily average relative humidity was around 60 with 472
variations ranging between 25 and 93 473
The prevalent wind direction during the experimental period (84 of the runs) was NW the 474
wind speed ranged from 05 to 89 ms-1
and averaged 34 ms-1
while the friction velocity u 475
ranged between 001 and 099 ms-1
with an average of 030 ms-1
(Figure 1a) The energy 476
balance for the period of the trial is reported in Figure 2b at hourly scale In particular it is 477
clear the direct impact of the net radiation (Rn) on the sensible heat flux (H) and on the latent 478
heat flux (LE) The pattern of LE were strongly correlated with the available energy Rn with 479
distinct peaks immediately following irrigation which provided water to evapotranspirate 480
from plant and soil 481
22 Ammonia concentrations 482
The field NH3 concentrations measured by the diffusive samplers (ALPHA badges) over 483
several hours ranged from 3 to 118 microg NH3 m-3
while the 30 minutes ROSAA concentrations 484
varied from around 0 to around 90 microg NH3 m-3
The ROSAA concentrations averaged over 485
the same intervals as the ALPHA badges slightly underestimated the badges concentration by 486
8 on average (ROSAA = 092 ALPHA R2 = 082) The uncalibrated QC-TILDAS 487
concentrations strongly underestimated the ALPHA badges concentrations by 67 but 488
without any offset (QC-TILDAS = 034 ALPHA R2 = 090) This underestimation is larger than 489
what Ellis et al (2010) found with the same instrument in a laboratory calibration Since the 490
correlation was very good the regression against the ALPHA badges was used to calibrate the 491
QC-TILDAS concentration data The correction factor was found to be 286 plusmn 009 The 492
calibrated QC-TILDAS concentration was in quite good agreement with the ROSAA 493
concentrations both in terms of magnitude and dynamics (Figure 2) In Figure 3 a detailed 494
comparison between calibrated values of QC-TILDAS NH3 concentrations and ROSAA 495
values is shown Since we did not find a significant offset in the regression of QC-TILDAS 496
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
18
measurements with respect to the ALPHA badges the same correction factor (286 plusmn 009) 497
could be used for EC fluxes for which only the span calibration is relevant 498
The background concentration measured with ALPHA badges over several days varied from 499
19 plusmn 03 to 7 plusmn 15 microg NH3 m-3
and averaged 27 plusmn 12 microg NH3 m-3
over the whole 500
experimental period The background was therefore very small compared to the NH3 501
concentration in the field during the period 22 to 30 July 2008 (see Figure 2) 502
503
23 Quality of the eddy covariance signal 504
231 Integral Turbulence Test 505
In order to test the validity of the eddy covariance measurements and to test the development 506
of the turbulent conditions the integral turbulence characteristic the so-called flux variance 507
similarity was analyzed as first proposed by Foken and Wichura (1996) The ratio of the 508
standard deviation of vertical wind component and sonic temperature and their respective 509
turbulent scaling variable (u and T respective) are shown in Figure 4 as a function of the 510
Monin-Obukhov stability parameter ζ (defined by zLMO where LMO is the Monin-Obukhov 511
length) The experimental data are compared to the theoretical model under unstable and near 512
neutral conditions (Kaimal and Finnigan 1994) 513
2
1
c
MO
w
L
zc
u
(16)
514
2
1
c
MO
T
L
zc
T
(17)
515
where the constant c1 and c2 are reported by Foken et al (2004) Because of the relative large 516
errors on integral turbulence characteristic in near neutral condition temperature data were 517
discarded when sensible heat flux density was less than 100 W m-2
The experimental data 518
nicely follow the theoretical behaviour for the vertical wind component whereas values 519
slightly higher than the model are observed for the sonic temperature variance This additional 520
turbulence was likely due to the high surface heterogeneity of a not completely closed canopy 521
and to the low measurement height 522
523
232 Cross-correlations wrsquorsquo 524
Before the analysis of the fluxes we looked at both the time and frequency response of the 525
signal An example of the cross-correlation functions between wrsquo rsquo (CO2 and NH3) is given 526
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
19
in Figure 5 The delay time (or lag time) can clearly be identified by the location of the 527
maximum of the cross-correlation function between wrsquo and rsquo It is close to zero for 528
temperature and CO2 (measured by an open path analyzer Li-7500 Licor Biosciences USA) 529
A positive lag is also clearly seen for the QC-TILDAS NH3 signal The difficulty for 530
estimating the time delay from the cross-correlation function can be seen from the long tail 531
observed for NH3 This tail can be interpreted as NH3 adsorbed and subsequently released 532
from the lines Nevertheless the lag times estimated with the cross-correlation method for 533
CO2 H2O and NH3 were 020plusmn004 s 020 plusmn003 s and 690 plusmn 014 s respectively In the case 534
of NH3 however we have to take into account that this delay is basically the sum of two 535
terms the time delay of the DAC conversion (6 s) and the gas residence time in the tube and 536
device 537
24 Cospectral density of NH3 and other scalars 538
An example of averaged and normalized scalar cospectra for one day of good data is given in 539
Figure 6 for the runs from 800 to 1600 in a clear day We did not use the dimensionless 540
frequency to average our cospectra because wind speed variation within the selection of runs 541
was small The cospectrum of sensible heat CowT compares very well to the theoretical one 542
while the cospectrum of NH3 CowNH3 obtained taking in to account the time lag falls off 543
faster than the theoretical one for frequencies higher than the cut-off frequency of fx ~ 03 Hz 544
Moreover for higher frequencies the NH3 cospectrum is also more noisy with values 545
sometimes being negative we hence show the absolute values as suggested by Mammarella 546
et al (2010) This allows inspecting the fall-off in the inertial subrange which is the result of 547
all high frequency losses In the following we analyse the correction of these HF losses by the 548
considered approaches Moreover Figure 7 is a clear example of the data with an 549
underestimation of all the measured scalar cospectra during the trial in particular both the 550
cospectra relative to the CO2 and the sensible heat show a loss at high frequency like the NH3 551
cospectra 552
241 High Frequency loss corrections 553
Figure 8 shows the trends of the theoretical TF due to the dynamic response time reported and 554
all the TF expressed in equations from 9 to13 (TFs TFtube TFvol TFphase_shift) moreover their 555
product to give the total theoretical transfer function to apply at cospectra is reported In 556
particular two total theoretical TFs are plotted the TF_tot=TFdyn_responseTFsTFtubeTFvol to 557
apply at data elaborated taking into account the maximum correlation function method for the 558
computation of the time lag and the TF_tot_phase_shift=TFdyn_responseTFsTFtube 559
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
20
TFvolTFphase_shift where the total phase shift is involved For the latter the TFphase_shift is 560
responsible for negative values of the overall transfer function (Massman 2000) The 561
individual cut-off frequencies can be estimated for the two theoretical approaches 562
considering the frequency at which the total TF equals 2-12
these values are around 06 and 563
02 Hz for TF_tot and TF_tot_phase_shift respectively (see Figure 8) Then the resulting 564
CFTheor_phase_shift are higher than the CFTheor_time_lag ones (around 15) with a wider range 565
between 16 and 45 566
On average the correction factor found by the ogive method was 136 plusmn 037 Figure 9 shows 567
an example of temperature CO2 and NH3 ogive as well as the fitted NH3 ogive and the 568
corresponding estimation of the high-frequency attenuation for NH3 It shows a typical 569
situation where the CO2 and H2O HF loss was around 15 while the NH3 HF loss was around 570
20 The corresponding HF correction factor for NH3 for this particular example would be 571
around 108 572
The damping computed using the inductance method CFL has a wide range starting from 1 573
however following suggestion by Eugster and Senn (1995) a threshold of 25 has been used 574
in order to reject data where we have not measured at least 40 of the expected flux An 575
example of the application of the method is reported in Figure 10 where measured NH3 576
cospectrum is plotted with the idealized empirical one given by Kaimal et al (1972) and the 577
damped cospectrum obtained using in equation (14) the value of L which gives the best fit of 578
the measured cospectrum 579
For invariant set up of instruments the frequency correction factors should depend on (i) the 580
lag time which affects the shape of the measured spectra (ii) the atmospheric stability which 581
affects the shape of the reference cospectra and (iii) the wind speed which affects the 582
distribution of cospectra density across the frequency range In order to compare the above-583
described correction methodologies applied to our set of data relationships between the 584
correction factors and the wind speed have been searched separating stable and unstable 585
conditions (Table 1) The relationships for the theoretical TF have been obtained using all the 586
data while for the inductance method because of the relatively large scatter of the individual 587
data they were grouped in wind speed classes of 1 m s-1
width for which the median were 588
determined and fitted describing the overall dependence of the damping factor on wind speed 589
under stable and unstable condition using only damping factor less than 25 as suggested by 590
Eugster and Senn (1995) For stable conditions the relationships in Table 1 show lower r2 591
than the ones under unstable conditions due to the lower number of data By means of these 592
relationships the EC NH3 fluxes have been corrected and the ones relative to 5 days of the 593
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
21
second week of the trial (24-29 July) are plotted in Figure 11 594
Table 1 Frequency correction factors (CF) in function of wind speed (u) under stable and unstable atmospheric 595 conditions 596
Unstable Stable
Theoretical TF with
time lag
950
211080
2
__
r
uCF lagtimetheor
320
441080
2
__
r
uCF lagtimetheor
Theoretical TF with
phase shift
960
381550
2
__
r
uCF shiftphasetheor
850
300303
2
__
r
uCF shiftphasetheor
Experimental-Ogive
990
501110
2
r
uCFO
470
2310250
2
r
uCFO
Inductance
810
171300
2
r
uCFL
760
810210
2
r
uCFL
597
598
599
In general the correction applied following the different approaches agree quite well among 600
them except where the inductance method gives larger correction (CFL gt 25) with respect to 601
the theoretical ones 602
To evaluate the overall effects of spectral losses on eddy covariance measurements the 603
cumulated values of the NH3 fluxes in the experimental period were calculated taking into 604
account the correction factors following the methods studied in this work The results are 605
reported in Table 2 as percentage of the losses estimated by the applied corrections 606
607
Table 2 Flux losses estimated by the four correction methods described in the section 26 608
Methods Losses ()
Theoretical TF with time lag -31
Theoretical TF with phase shift -30
Experimental-Ogive -43
Inductance -23
609
Calculated flux losses ranged between -23 (inductance method with the CFL threshold 610
selected to 25) and -43 (experimental-ogive method) while the two theoretical TF 611
approaches gave comparable loss estimates However it should be underlined that the 612
inductance method only corrects for high-frequency losses whereas the ogive method also 613
corrects low-frequency differences There is no doubt that the experimental approach takes 614
into account damping effects due to absorption and desorption of ammonia which are not 615
considered by the theoretical ones 616
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
22
It is of interest to compare these results with the other few works where the fluxes are 617
measured with equipment based on EC technique with NH3 fast analysers Actually the 618
values of flux underestimation are in agreement with the work of Whitehead et al (2008) 619
who found -41 using a QCLAS and -37 using a TDLAS during a specific experimental 620
period and applying the cospectral similarity between the ammonia and heat fluxes On the 621
other hand Shaw et al (1998) found a greater underestimation of -48 using a tandem mass 622
spectrometer and a TF approach Sintermann et al (2011) found high frequency attenuation 623
losses from 20 to 40 using a chemical ionisation mass spectrometer 624
625
4 Conclusions 626
In this work we evaluated the correction to be applied to the NH3 eddy covariance fluxes from 627
a crop for taking into account the damping of the high frequencies contribution due to the set-628
up of the equipment when a Tunable Infrared Laser Differential Absorption Spectrometer 629
with a quantum cascade laser source is used This instrument has only recently become 630
available for routine measurements of NH3 concentrations at high frequency as is needed for 631
direct eddy covariance flux measurements Here three methodologies were analysed in detail 632
for determining the correction factor for the NH3 fluxes (1a) the spectral theoretical transfer 633
function with time lag (1b) the spectral theoretical transfer function with total phase shift (2) 634
the ogive method which is a variant of the in-situ spectral experimental transfer function 635
technique and (3) the inductance correction method 636
A comparison between passive diffusion samples (ALPHA badges) and QC-TILDAS clearly 637
showed that this last equipment underestimated the NH3 concentration values therefore they 638
should be corrected before the calculation of EC fluxes (the correction factor was found to be 639
around 3) When this correction factor is applied the concentration values of ammonia 640
measured by the QC-TILDAS are quite similar to that measured by an independent system 641
(ROSAA) based on wet-effluent denuders 642
Secondly the correction factors of the EC fluxes all follow linear functions of the wind speed 643
(CF=a u+b u wind speed) for all the methodologies above mentioned However for 644
unstable atmospheric conditions the most common encountered during the experiment the 645
wind speed dependency of CF is very low for the spectral theoretical TF with time lag 646
(a=008 and b=121) and the ogive method (a=011 and b=150) whereas it is similar for the 647
remaining two methodologies (a=055 030 b=138 117 for the TF with phase shift and 648
inductance respectively) Overall when the calculated CFs are applied to EC NH3 fluxes 649
measured in our experimental field the values of the fluxes obtained are comparable and 650
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
23
basically reflected the conceptual differences in assumptions that must be made under absence 651
of an undisputed reference flux measurement 652
In conclusion correction of the values of NH3 concentration and EC fluxes over natural 653
surfaces are necessary for taking into account the dumping of high frequency contribution due 654
to the coupling of ammonia fast analyser by QC-TILDAS and sonic anemometer 655
Considering that we made measurements just above the crop the contribution of the high 656
frequencies to the whole flux of ammonia is comparatively large therefore the 657
underestimation of the EC fluxes has to be corrected for 658
Moreover considering that we lack an undisputed reference NH3 flux against which the 659
methods could be tested and hence the differences are basically the conceptual differences of 660
the approaches it would not be possible to firmly establish that one is better than the other 661
however there are pro and contra for each of them Hence at present either method could 662
provide realistic results and therefore future research should also focus on the question of 663
how realistic it is to use sensible heat flux for similarity with NH3 fluxes namely on fields 664
with finite extent where fertilizer was applied and from which NH3 fluxes can originate 665
whereas sensible heat flux is also determined by surrounding However in case the sensible 666
heat cospectra of the experimental site are not available the theoretical TF approach or the 667
inductance method could be used to determine the percentage of the flux that is lost by the 668
measurement system 669
670
Acknowledgements 671
This study has been supported by Italian Projects FISR ldquoCLIMESCOrdquo (DD 285ric 222006) 672
and AQUATER (DM n 209730305) European Project NitroEurope supported the 673
experimental field work First author has been also funded by NinE - ESF Research 674
Networking Programme ldquoNitrogen in Europerdquo and ACCENT - Biaflux (Network of 675
Excellence funded by EC FP6 PRIORITY 1163 Global Change and Ecosystems 676
subproject Biosphere-Atmosphere Exchange of Pollutants) The authors thank Ceacuteline Decuq 677
Pasquale Introna Nicola Martinelli and Sylvie Masson for them helping in conducting the 678
experimental work in the field and Dr Domenico Vitale for artwork 679
680
References 681
682
Ammann C Brunner A Spirig C and Neftel A 2006 Technical note Water vapour 683
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
24
concentration and flux measurements with PTR-MS Atmos Chem Phys 6 4643-4651 684
Asman W A H Sutton M A and Schjorring J K 1998 Ammonia emission 685
atmospheric transport and deposition New Phytol 139 27ndash48 686
Aubinet M Grelle A Ibrom A Rannik U Moncrieff J Foken T Kowalski AS 687
Martin PH Berbigier P Bernhofer C Clement R Elbers J Granier A Grunwald 688
T Morgenstern K Pilegaard K Rebmann C Snijders W Valentini R Vesala T 689
2000 Estimates of the annual net carbon and water exchange of forests the EUROFLUX 690
methodology Adv Ecol Res 30 113-175 691
Brodeur JJ Warland JS Staebler RM amp Wagner-Riddle C 2009 Technical note 692
Laboratory evaluation of a tunable diode laser system for eddy covariance measurements 693
of ammonia flux Agric For Meteorol 149 385-391 694
Denmead OT 1983 Micrometeorological methods for measuring gaseous losses of 695
nitrogen in the field in Freney JR Simpson JR (Eds) Gaseous loss of nitrogen from 696
plant-soil systems Martinus NjhoffDr W Junk The Hague pp 133-157 697
Denmead OT 2008 Approaches to measuring fluxes of methane and nitrous oxide between 698
landscapes and the atmosphere Plant Soil 309 5ndash24 699
Ellis R A Murphy J G Pattey E van Haarlem R OBrien J M Herndon S C 2010 700
Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption 701
Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia Atmos Meas 702
Tech 3 397-406 703
Erisman J W Bleeker A Hensen A Vermeulen A 2008 Agricultural air quality in 704
Europe and the future perspectives Atmos Environ 42 3209ndash3217 705
Erisman JW Vermeulen A Hensen A Flechard C Dammgen U Fowler D Sutton 706
M Grunhage L Tuovinen JP 2005 Monitoring and modelling of 707
biosphereatmosphere exchange of gases and aerosols in Europe Environ Pollut 133(3) 708
403-413 709
Eugster W Senn W 1995 A cospectral correction model for measurement of turbulent 710
NO2 flux Boundary-Layer Meteor 74(4) 321-340 711
Famulari D Fowler D Hargreaves K Milford C Nemitz E Sutton M Weston K 712
2005 Measuring eddy covariance fluxes of ammonia using tunable diode laser absorption 713
spectroscopy Water Air Soil Pollut Focus 4 (6) 151-158 714
Fehsenfeld F 1995 Measurement of chemically reactive gases at ambient concentrations in 715
Matson PA Harris RC (Eds) Biogenic trace gases measuring emission from soil 716
and water Blackwell Science Cambridge England pp 206-258 717
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
25
Ferm M 1979 Method for determination for atmospheric ammonia Atmos Environ 13 718
1385-1393 719
Foken T Wichura B 1996 Tools for quality assessment of surface-based flux 720
measurements Agric For Meteorol 78 83-105 721
Foken T Gӧ cked M Mauder M Mahrt L Amiro B Munger W 2004 Post-field data 722
quality control in Lee X Massman W J Law B (Eds) Handbook of 723
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 724
Publishers pp 181-208 725
Galloway JN Aber JD Erisman JW Seitzinger SP Howarth RW Cowling EB 726
Cosby BJ 2003 The Nitrogen Cascade BioScience 53 (4) 341-356 727
Harper LA 2005 Ammonia measurement issues in Hatfield JL Baker JM Viney 728
MK (Eds) Micrometeorology in Agricultural systems Agronomy Monograph 47 729
ASA CSSA amp SSSA Madison Wisconsin USA pp 345-379 730
Horst TW 1997 A simple formula for attenuation of eddy fluxes measured with first-order 731
response scalar sensors Boundary-Layer Meteorol 82 219ndash233 732
Ibrom A Dellwik E Flyvbjerg H Jensen N O and Pilegaard K 2007 Strong low-733
pass filtering effects on water vapour flux measurements with closed-path eddy 734
correlation systems Agric For Meteorol 147 140ndash156 735
IPCC 2007 The Physical Science basis of Climate Change A Report of Working Group I of 736
the Intergovernmental Panel on Climate Change Fourth Assessment Report Summary 737
for Policymakers httpipcc-wg1ucareduwg1wg1-reporthtml 738
Kaimal JCWyngaard JC Izumi Y Cote OR 1972 Deriving power spectra from a 739
three-component sonic anemometer J Appl Meteorol 7 827ndash837 740
Kaimal JC amp Finnigan JJ 1994 Atmospheric boundary layer flows ndash their structure and 741
measurements Oxford University Press Oxford 742
Kolle O Rebmann C 2007 EddySoft documentation of a software package to acquire and 743
process eddy covariance data Max Planck Institute for Biochemistry Jena Germany 744
Kruit RJW van Pul WAJ Otjes RP Hofschreuder P Jacobs AFG and Holtslag 745
AAM 2007 Ammonia fluxes and derived canopy compensation points over non-746
fertilized agricultural grassland in The Netherlands using the new gradient ammonia - 747
high accuracy - monitor (GRAHAM) Atmos Environ 41(6) 1275-1287 748
Krupa S V 2003 Effects of atmospheric ammonia (NH3) on terrestrial vegetation a 749
review Environ Pollut 124 179ndash221 750
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
26
Lee X Massman W Law BE 2004 Handbook of micrometeorology a guide for surface 751
flux measurement and analysis Atmospheric and Oceanographic Sciences Library 752
Kluwer Academic Publishers Dordrecht The Netherlands 753
Lenschow DH Raupach MR 1991 The attenuation of fluctuations in scalar concentration 754
through sampling tubes J Geophys Res 96(D8) 15259ndash15268 755
Leuning R 2007 The correct form of the Webb Pearman and Leuning equation for eddy 756
fluxes of trace gases in steady and non-steady state horizontally homogeneous flows 757
Boundary-Layer Meteorol 123(2) 263-267 758
Leuning R Judd MJ 1996 The relative merits of open- and closed-path analysers for 759
measurement of eddy fluxes Global Change Biol 2 241-253 760
Liebethal C and Foken T 2004 On the significance of the Webb correction to fluxes (vol 761
109 pg 99 2004) Boundary-Layer Meteorol 113(2) 301-301 762
Loubet B Decuq C Personne E Massad RS Fleacutechard C Fanucci O Mascher N 763
Gueudet J-C Masson S Durand B Geacutenermont S Fauvel Y Cellier P 2011 764
Investigating the stomatal cuticular and soil ammonia fluxes over a growing tritical crop 765
under high acidic loads Biogeosciences Discussions 8 10317-10350 766
Loubet B Geacutenermont S Ferrara R Bedos C Decuq C Personne E Fanucci O 767
Durand B Rana G Cellier P 2010 An inverse model to estimate ammonia 768
emissions from fields Eur J Soil Sci 61(5) 793-805 769
Mahrt L 2010 Computing turbulent fluxes near the surface Needed improvements Agric 770
For Meteorol 150(4) 501-509 771
Mammarella I Werle P Pihlatie M Eugster W Haapanala S Kiese R Markkanen T 772
Rannik Uuml Vesala T 2010 A case study of eddy covariance flux of N2O measured 773
within forest ecosystems quality control and flux error analysis Biogeosciences 7 427-774
440 775
Massman 1998 A review of the molecular diffusivities of H2O CO2 CH4 CO O3 SO2 776
NH3 N2O NO and NO2 in air O2 and N2 near STP Atmos Environ 777
32(6) 1111-1127 778
Massman W 2004 Concerning the measurement of atmospheric trace gas fluxes with open 779
and closed-path eddy covariance system the WPL terms and spectral attenuation in Lee 780
X Massman W J Law B (Eds) Handbook of micrometeorology Atmospheric and 781
Oceanographic Sciences Library Kluwer Academic Publishers pp 133-160 782
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
27
Massman W Clement R 2004 Uncertainty in eddy covariance flux estimates resulting 783
from spectral attenuation in Lee X Massman W J Law B (Eds) Handbook of 784
micrometeorology Atmospheric and Oceanographic Sciences Library Kluwer Academic 785
Publishers pp 67 -100 786
Massman WJ 2000 A simple method for estimating frequency response corrections for 787
eddy covariance systems Agric For Meteorol 104 (3) 185-198 788
Milford C Theobald MR Nemitz E Hargreaves KJ Horvath L Raso J Dammgen 789
U Neftel A Jones SK Hensen A Loubet B Cellier P Sutton MA 2009 790
Ammonia fluxes in relation to cutting and fertilization of an intensively managed 791
grassland derived from an inter-comparison of gradient measurements Biogeosciences 792
6(5) 819-834 793
Moncrieff JB Massheder JM de Bruin H Elbers J Friborg T Heusinkveld B 794
Kabat P Scott S Soegaard H Verhoef A 1997 A system to measure surface fluxes 795
of momentum sensible heat water vapour and carbon dioxide J Hydrol 188-189 589-796
611 797
Moore CJ 1986 Frequency response corrections for eddy correlation systems Boundary-798
Layer Meteorol 3717ndash35 799
Neftel A Blatter A Gut A Houmlgger D Meixner F Ammann C Nathaus F J 1998 800
NH3 soil and soil surface gas measurements in a Triticale wheat field Atmos Environ 801
32 499ndash505 802
Rothman LS Barbe A Chris Benner D Brown LR Camy-Peyret C Carleer 803
KCMR Clerbaux C Dana V Devi VM Fayt A Flaud RRG J-M Goldman 804
A Jacquemart D Jucks KW Lafferty J-YM WJ Massie ST Nemtchinov V 805
Newnham DA Perrin CPR A Schroeder J Smith KM Smith MAH Tang K 806
Toth JVA RA Varanasi P Yoshino K 2003 The HITRAN molecular 807
spectroscopic database Edition of 2000 including updates through 2001 J Quant 808
Spectrosc Radiat Transfer 82 5-44 809
Shaw WJ Spicer CW amp Kenny DV 1998 Eddy correlation fluxes of trace gases using a 810
tandem mass spectrometer Atmos Environ 32 (17) 2887-2898 811
Shimizu T 2007 Practical applicability of high frequency correction theories to CO2 flux 812
measured by a closed-path system Boundary-Layer Meteorol 122 417ndash438 813
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
28
Sintermann J Spirig C Jordan A Kuhn U Ammann C and Neftel A 2011 Eddy 814
covariance flux measurements of ammonia by high temperature chemical ionisation mass 815
spectrometry Atmos Meas Tech 4(3) 599-616 816
Spirig C Flechard C R Ammann C Neftel A 2010 The annual ammonia budget of 817
fertilised cut grassland ndash Part 1 Micrometeorological flux measurements and emissions 818
after slurry application Biogeosciences 7 521ndash536 819
Sutton M A Dragosits U Tang Y S and Fowler D 2000 Ammonia emissions from non 820
agricultural sources in the UK Atmos Environ 34 855ndash869 821
Sutton M A Reis S and Baker S M H 2009 Atmospheric ammonia detecting emission 822
changes and environmental impacts Springer 823
Sutton M A Tang Y S Dragosits U Fournier N Dore T Smith R I Weston K J 824
and Fowler D 2001 A spatial analysis of atmospheric ammonia and ammonium in the 825
UK The Scientific World 1 (S2) 275ndash286 826
Suyker Ae Verma SB 1993 Eddy correlation measurement of CO2 flux using a closed 827
path sensor theory and field tests against an open path sensor Boundary-Layer Meteorol 828
64 391-407 829
Tang YS Cape JN Sutton MA 2001 Development and types of passive samplers for 830
NH3 and NOx International Symposium on Passive Sampling of Gaseous Pollutants in 831
Ecol Res 1 513-529 832
von Bobrutzki K Braban C F Famulari D Jones S K Blackall T Smith T E L 833
Blom M Coe H Gallagher M Ghalaieny M McGillen M R Percival C J 834
Whitehead J D Ellis R Murphy J Mohacsi A Pogany A Junninen H Rantanen 835
S Sutton M A Nemitz E 2010 Field inter-comparison of eleven atmospheric 836
ammonia measurement techniques Atmos Meas Tech 3 91ndash112 837
Warland JS Dias GM Thurtell GW 2001 A tunable diode laser system for ammonia 838
flux measurements over multiple plot Environ Pollut 114 215-221 839
Webb EK Pearman G Leuning R 1980 Correction of flux measurements for density 840
effects due to heat and water vapour transfer Quart J Roy Meteorol Soc 106 85-100 841
Whitehead J Twigg M Famulari D Nemitz E Sutton MA Gallagher MW Fowler 842
D 2008 Evaluation of Laser Absorption Spectroscopic Techniques for Eddy covariance 843
Flux Measurements of Ammonia Environ Sci Technol 42 2041-2046 844
Wyers GP Otjes RP and Slanina J 1993 A continuous-flow denuder for the 845
measurement of ambient concentrations and surface-exchange fluxes of ammonia 846
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
29
Atmos Environ 27A 2085-2090 847
Zahniser MS Nelson DD McManus JB Shorter JH Herndon S Jimenez R 2005 848
Development of a Quantum Cascade Laser-Based Detector for Ammonia and Nitric 849
Acid Final Report US Department of Energy SBIR Phase II Grant No DE-FG02-850
01ER83139 851
852
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
30
Figure captions 853
854
Figure 1 Trends of (a) friction velocity u and (b) energy balance components during the 855
trial Net radiation (Rn) latent heat flux (LE) sensible heat flux (H) and soil heat flux (G) are 856
shown Moments of irrigation are shown with blue triangles 857
Figure 2 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) the 858
ROSAA analyser (calibrated) and the ALPHA diffusion samplers The QC-TILDAS 859
concentrations were calibrated as explained in the section 32 860
Figure 3 Ammonia (NH3) concentration measured with the QC-TILDAS (calibrated) against 861
the values measured by the ROSAA analyser The linear regression gives 862
ConcentrationQCL = ConcentrationROSAA times 09958 [09356 -10560] where the 95 863
confidence interval is given in brackets 864
Figure 4 Normalized standard deviation for vertical wind speed σwu and temperature σTT 865
in function of the Monin Obukhov stability parameter (|zLMO|) Continuous lines are relative 866
to the model reported by Foken and Wichura (1996) 867
Figure 5 Typical cross-correlation function for (wrsquo Ta
rsquo) (w
rsquo βCO2
rsquo) and (w
rsquo βNH3
rsquo) for the 868
DOY 207 at 11h30 UT and DOY 2009 at 7h30 UT determined over 60 minutes Here w 869
vertical component of the wind speed Ta the sonic temperature βCO2 the CO2 concentration 870
and βNH3 the NH3 concentration while a prime denote a fluctuating quantity determined by 871
linear detrending 872
Figure 6 Normalized averaged cospectra (ratio of cospectra over covariance) of NH3 (CowNH3) 873
and sensible heat fluxes (CowT) over a 8 h period (from 800 am to 4 pm with U = 2 m s-1
874
and zLMO = -002) The negative values of NH3 cospectrum (CowNH3neg) have been inverted 875
to be represented on the logarithmic scale 876
877
Figure 7 Normalized averaged cospectra of NH3(CoSpwnh3) temperature (CoSpwt) water 878
(CoSpwq) and CO2 (CoSpwc) over a 4 h period (DOY 208 from 800 am to 1200 am with 879
U = 21 m s-1
and zLMO = -002 A high pass filter has been applied at f = 000139 Hz (12 880
min) 881
882
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
31
883
Figure 8 Trend of the single theoretical transfer functions (TF) and the total ones TF_total 884
and TF_total_phase shift to apply to cospectra obtained taking into account the maximum 885
cross-correlation function and the total phase shift respectively (see the section 261 for 886
details about each theoretical TF) The data are relative to a half an hour with U=21 ms-1
and 887
zLMO=-004 888
889
Figure 9 Example ogives for heat CO2 H2O and NH3 fluxes and corresponding estimation of 890
the high frequency damping by ogive fitting (DOY 208 from 0800 am to 1200 am) T 891
stands for temperature ogive init is the initial ogive and corr is the corrected ogive (fit to 892
the temperature ogive as explained in section 262) The High Frequency loss damping is 893
simply one minus the value of the corrected ogive at the higher frequency Here the CO2 894
H2O and NH3 high frequency dampings are 017 = 1 ndash 083 018 = 1 ndash 082 and 037 = 1 ndash 895
063 A high pass filter has been applied at f = 000139 Hz (12 min) 896
897
Figure 10 An example of the inductance method applied to half an hour dataset (DOY 207 898
1000 am) The measured (black dots and lines) the Kaimal et al (1972) (blue) and the 899
damped (red) cospectra are plotted The values of inductance L wind speed u flux FNH3 and 900
time lag are reported 901
Figure 11 Ammonia (NH3) fluxes corrected by means of the correction factors (CF) 902
calculated using the approaches explained in the section 26 (i) theoretical transfer function 903
with time lag (CF_theor_time_lag) and total phase shift (CF_theor_phase_shift) (ii) in situ Ogive 904
method (CFO) (iii) inductance method (CFL) 905
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
32
906
FIGURE 1907
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
33
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
908
909
FIGURE 2 910
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
34
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
911
912
FIGURE 3 913
914
915
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
35
916
917 FIGURE 4 918
919
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
36
920 FIGURE 5 921
922
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
37
923
924
FIGURE6 925
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
38
926
FIGURE 7 927
928
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
39
929
930
FIGURE 8 931
932
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
40
933
FIGURE 9 934
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
41
935
FIGURE 10 936
937
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
42
938
939
FIGURE 11 940
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
0
20
40
60
80
100
120
140
170708 190708 210708 230708 250708 270708 290708
Co
nc
(micro
g N
H3 m
-3)
ROSAA (calibrated)
QCL (calibrated)
alpha Badges
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
00
02
04
06
08
10
u(msminus1
)(a
)
Energyfluxes(Wmminus2
)
170
708
190
708
210
708
230
708
250
708
270
708
290
708
minus20
00
200
400
600
800
(b)
Rn
HLE
GIr
rigat
ion
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
0
20
40
60
80
100
0 20 40 60 80 100
ROSAA concentration
(microg NH3 m-3
)
QC
L c
alib
rate
d c
on
ce
ntr
ati
on
(microg
NH
3 m
-3)
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
10minus5
10minus3
10minus1
101
01
1010
|zL
MO|
σwu
10minus3
10minus2
10minus1
100
01
1010
|zL
MO|
σTT
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
FIGURE 5
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
minus4
3
10minus3
10minus2
10minus1
100
101
10minus4
10minus3
10minus2
10minus1
100
Nat
ura
lfre
qu
ency
(Hz)
fsdot(ScalarCospectra)
Co w
TC
o wN
H3
Co w
NH
3neg
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
10minus2
10minus1
100
101
005
01
015
02
025
fC
oSp
co
var)
20082080800
freq (Hz)
CoSpwtCoSpwcCoSpwqCoSpwnh3KaimalminusFinnigan 036 09
3
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
10minus3
10minus2
10minus1
100
101
minus1
0
minus0
5
00
05
10
Nat
ural
freq
uenc
y(H
z)
TF(minus)
TF
dyn
amic
freq
uenc
y re
spon
seT
F la
tera
l sep
arat
ion
TF
tube
lam
inal
flow
TF
vol
ume
aver
agin
gT
F p
hase
shi
ftT
F to
tal
TF
tota
l pha
se s
hift
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
FigureClick here to download high resolution image
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
Nat
ura
lfre
qu
ency
(Hz)
fCowNH3(f)wrsquoNH3rsquo
10minus3
10minus2
10minus1
100
101
minus0
4
minus0
2
00
02
04
06
Lag=
68
6 s
NH
3
Idea
lized
Dam
ped
L=
025
56+
minus0
0917
(p=
000
66)
u=
257
2m
sminus1F
NH
3=
146
26pp
bm
sminus1
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
230
708
240
708
250
708
260
708
270
708
280
708
290
708
0
500
1000
1500
2000
2500
3000
Dat
e(U
T)
NH3flux(ngmminus2
sminus1
)
EC
cal
ibra
ted
(28
6)C
F th
eor
time
lag
CF
theo
r ph
ase
shift
CF
Llt
25
CF
O
Figure
Top Related