Measurements of a network of mobile radars during the ... · PDF fileMeasurements of a network...
Transcript of Measurements of a network of mobile radars during the ... · PDF fileMeasurements of a network...
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
Measurements of a network of mobile radars
during the experimental campaign of
the HydroRad project
Kalogiros J. 1, Anagnostou M.
1,2, Anagnostou E.
3, Marzano F.S.
2,4, Picciotti E.
5, Cinque G.
5,
Montopoli M. 4, Bernardini L.
5, Volpi A.
6, Telleschi A.
6
1NOA National Observatory of Athens, Lofos Nymfon, Athens, Greece, e-mail: jkalog@ noa.gr
2Department of Information Engineering, Sapienza University of Rome, Rome, Italy
3University of Connecticut, Storrs, Connecticut, USA, e-mail: [email protected]
4CETEMPS, University of L’Aquila, L’Aquila, Italy, e-mail: [email protected] 5HIMET, Strada Statale 17 Ovest 36, L'Aquila, Italy, e-mail: [email protected]
6ELDES, Via di Porto 2/b, Firenze, Italy, e-mail: [email protected]
John Kalogiros
1. Introduction
Small size X-band polarimetric radars (hereafter called mini-radars) constitute a low-cost solution to the problem of
hydrologic and flood forecast in urban, small-scale basins and coastal areas where operational weather radars cannot provide
adequate coverage. Short-wavelength radar systems are more attractive also for research purposes because of their small size.
Their limitations are the smaller range due to low power and the significant signal attenuation at X-band in heavy rain. In the
framework of HydroRad project (Picciotti et al. 2012) an innovative dual-polarization X-band mini-radar system was
developed and facilitated by a suite of software tools including rain estimation, nowcasting, precipitation classification and
integration with hydrological and meteorological models for applications in weather and flood forecasting. The integrated
system is described in detail by Picciotti et al. (2012), while this work presents validation of the mini-radar rainfall estimates
against in situ observations and evaluates differences against a benchmark high-resolution dual-polarization X-band radar
(XPol). These mini-radars are easy to deploy and, thus, ideal for the setup of radar networks to cover areas with complex
terrain. The overall system was tested in an experimental campaign in Moldova, where mini-radar data were collected
synergistically with XPol and in-situ rain measurements from raingauges and a disdrometer (Fig. 1).
Fig. 1 One of the mini-radars (left) and the XPol mobile radar with the nearby disdrometer and raingauges (right) in the
Moldova field campaign.
Fig. 2 The basin of the Bic river in central Moldova. The locations of the radars are shown.
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
The Moldova Operational Field campaign (MOF) took place in the time period September to October 2011. The
experimental area was the basin of the river Bic with its tributaries around the Moldovan capital Chisinau (Fig. 2). The
terrain is characterized by low hills with elevation up to 300 m. The three mini-radars were installed in locations around
Chisinau in order to cover the basin, while XPol and the 2D-video disdrometer were installed at Chisianu suburbs. Six pairs
of tipping raingauges were installed at different positions along the river. The range of the radars was 60 km with a resolution
of 120 m. Radar observations included the horizontal reflectivity Zh, the differential reflectivity Zdr, co-polar correlation
coefficient ρhv and the differential phase shift Φdp. The specific differential phase shift Kdp is estimated as half the gradient of
Φdp along the radar ray. XPol was scanning in the azimuth sector from 319o to 125
o due to blockage by nearby buildings. The
mini-radars had also small azimuth sectors of partial beam blockage due to terrain. PPI scans at low elevation angles (up to
3.5o) of the radar antenna were performed as well as RHI scans in selected azimuth angles in order to estimate the vertical
structure of the rain field. The time period for a full volume scan was about 3 minutes. The disdrometer data was used for the
analysis of droplet size distribution, shape (axis ratio) and orientation of rain droplets, and the theoretical estimation
(scattering calculations) of polarimetric radar products. In the following sections, radar data processing and rainfall
estimation validation results for the mini-radar system are presented. Comparisons against the XPol rainfall estimates provide
a basis to evaluate performance differences due to differences in the radar characteristics.
2. Radar data processing
At X-band frequencies attenuation of radar signal by rain can be quite significant with values greater than 10 dB in heavy
rain. There are many dual-polarization rain attenuation correction algorithms like the ZPHI algorithm (Testud et al. 2000),
which is based on Φdp measurements, and its extension with the additions of a Φdp - Zdr constraint (Bringi et al. 2001). In this
study a new attenuation correction algorithm (Kalogiros et al. 2012b) was used which is based on the new rain microphysics
parameterizations described in Kalogiros et al. (2012a) with minimum parameterization error (up to 5%). These
parameterizations use the theoretical Rayleigh scattering limits corrected by a multiplicative rational polynomial function of
reflectivity-weighted raindrop diameter to approximate the Mie character of scattering. The rain attenuation algorithm was
applied to the three mini-radars and the XPol radar data from MOF. Calibration of the reflectivity measurements from the
mini-radars was carried out through comparison with the radar observables estimated from the disdrometer data (not shown
here). XPol was already calibrated from a large dataset collected during the past five years in Athens, Greece. XPol
reflectivity was also used to further check the calibration of the mini-radars.
Due to the use of high elevation angles of the radar antennas to avoid beam blockage by terrain features melting layer
effects (bright band) were observed at these elevation angles during widespread (stratiform) rain. For MOF data a low
elevation angle (1.5o) was used for rainfall estimation in order to avoid melting layer effects and to minimize effects due to
ground clutter. Figure 3 presents a high elevation angle (3o) PPI from XPol. The bright band is seen as a circular zone of high
reflectivities values in the range from 40 to 50 km corresponding to altitudes of about 2000 and 2500 m, respectively. The
white line shown is the base of the melting layer which was detected using ρhv as the rain-to-mixed phase classification
criterion. After detection a correction algorithm for the average apparent vertical profile of reflectivity in the PPI scan was
applied (Kalogiros et al. 2012c). The reflectivity corrected for the vertical profile of reflectivity (VPR) is indicated with Zhc.
It can be seen that the bright band has been removed in the corrected reflectivity. The corrected reflectivity at 3o elevation
angle is quite similar to the reflectivity field at 1.5o elevation angle (not shown here) with differences at some azimuth angles
due to partial blockage of XPol beam at the lower elevation angle by nearby terrain.
Fig. 3 Example of VPR correction of XPol PPI scan at an antenna elevation of 3.0o on 9/10/2011 showing the melting layer
(bright band).
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
Fig. 4 PPI scans from R3 mini-radar at the same time period as in Fig.3 at different elevation angles of the antenna.
Fig. 5 Statistical difference between R3 mini-radar and XPol measurements versus range of the mini-radar during the 8-
9/10/2011 rain event for two different elevation angles of the antennas.
Even though the highest elevation of the mini-radars scanning was 3.5o a melting layer signature was not detected in these
data. Figure 4 presents PPI scans at 0.5o and 3.5
o from R3 mini-radar at about the same time as XPol data in Fig. 3. Both
PPIs show the same horizontal structure of reflectivities but with lower average value at the higher elevation. The reason for
this difference from XPol is the wide beam of the mini-radars (half power width 3o compared to 1
o of XPol). When the mini-
radar beam crosses the melting layer at a range of about 35 km it is 1800 m wide. In the vertical direction the beam width
corresponds to an integration (filtering) of the vertical profile of reflectivity within this width according to the beam pattern.
The horizontal beam width is the same but it simply results in reduced horizontal resolution, which can be improved with
angular oversampling in PPIs and inverse filtering. In the vertical direction the beam is essentially covering most of the rain
layer and the melting layer above it. This filter in the vertical direction results in hiding the bright band at this elevation
angle. The bright band could probably be observed in mini-radar data if its altitude was lower and the elevation angle was
higher. Thus, in MOF data no VPR correction could be applied to the mini-radar data.
The wide beam of mini-radars has also additional effects in their measurements. Figure 5 shows the effect of wide beam on
reflectivity, differential reflectivity, correlation coefficient and specific differential phase shift by comparing the
measurements from R3 mini-radar with XPol measurements versus the range of the mini-radar for all the data of the
widespread rain event on 8-9 October 2011. Only data with correlation coefficient ρhv higher than 0.8 (i.e., good rain signal)
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
were included in the statistical comparison. The wide beam results in partial beam blockage by the terrain at elevation angles
below 3.5o. The average difference of R3 mini-radar reflectivity from XPol data is about 4 dB at the elevation angle of 3.5
o.
The near zero difference at 1.5o is due to the calibration of the mini-radar using the data from this elevation. On average it
was found that due to this partial beam blockage the R3 mini-radar reflectivities at 0.5o, 1.5
o and 2.5
o elevations were lower
from the reflectivity at 3.5o elevation angle by 10.5 dB, 4 dB and 1dB, respectively. These difference values depend also on
the terrain around the mini-radar because different terrain will cause different partial blockage. In addition to reflectivity the
rest of radar measurements were also affected by the wide beam effect. The differential reflectivity was showing a significant
difference by XPol data at the highest elevation and ranges longer than 45 km and it is probably due to the extension of the
beam well above the rain layer at these distances. The correlation coefficient shows a significant reduction with range at low
elevations and less at the highest elevation probably due to signal de-correlation by the reflection from the ground, too. The
specific differential phase shift is about 30% of XPol measurements at the low elevation angle and very small at the higher
elevation. However, as there was no trend in the differences with XPol versus range the low specific differential phase shift
may be also due to noise problems in the mini-radars.
3. Rainfall estimation
Rainfall estimates based on classical weather radar observations have quantitative limitations mainly due to the lack of
uniqueness in the relationship of the single radar measurable (reflectivity) to the associated rainfall intensity. The polarization
diversity capability of modern weather radars is expected to moderate this effect using polarimetric relations for estimation of
the rainfall rate, which combine Zh, Zdr and Kdp (Matrosov et al. 2002, Anagnostou et al. 2004, Park et al. 2005). In this study
three rainfall estimators were evaluated. The first one is a classic Z-R estimator with steady coefficients which were evaluated
from historic XPol data:
R=3.36x10-2
Zh0.58
,
(1)
where R is rainfall rate in mm h-1
units and Zh is in linear units instead of dBZ. The second estimator is a polarimetric one,
which is based the Nw normalization approach with constants estimated from electromagnetic scattering simulations
(Kalogiros et al. 2006). Nw is the intercept parameter (units mm-1
m-3
) of rain droplet size distribution (DSD), which is
approximated with a normalized Gamma distribution (Bringi and Chandrasekar 2001). Nw is obtained also from polarimetric
relations found from simulations. The first polarimetric rainfall estimator is:
Rp1=1.305x10-3
Nw(Zh/Nw)0.58
. (2)
The third rainfall estimator is a new polarimetric estimator which minimizes the approximation error using the theoretical
Rayleigh scattering limit with the addition of a rational polynomial function of reflectivity-weighted droplet diameter to
approximate the Mie character of scattering (Kalogiros et al. 2012a, Anagnostou et al. 2012):
Rp2=0.8106FR(µ)ΝwD04.67
fR(D0) . (3)
D0 and µ are the median volume diameter and the shape parameter of the DSD, respectively, FR is a function of µ, which is
included in the Gamma approximation of the DSD, and fR is a third degree rational polynomial of D0 with constants
evaluated by the simulations. For the estimation of rainfall rate by the radars the 1.5o elevation angle was used. Data from
two rain events were used. Figure 6 shows an example of reflectivity maps from R3 mini-radar and XPol during the first rain
event (8-9 September 2011) which was of convective type with isolated strong rain cells. The positions of the three mini-
radars and the six raingauge sites relative to XPol are also shown in the figure. The mini-radar reflectivity field agrees well
with the near concurrent field from XPol, which shows the capability of the mini-radar to capture well the spatial variability
of rain. The second rain event (8-9 October 2011) was of stratiform rain type with widespread moderate rain.
Fig. 6 PPI scans of horizontal reflectivity Zh measured from the XPol and the R3 mini-radar at an elevation angle of 1.5
o.
The locations of the three mini-radars and the six pairs of raingauges relative to XPol are also shown.
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
Fig. 7 Time series of mini-radars rainfall estimates (accumulation in 30 minutes) from the three algorithms of Eqs. (1)-(3).
Fig. 8 Histograms of difference between mini-radars and XPol estimates of total rainfall accumulated in the rain event using
the three algorithms of Eqs. (1)-(3).
Fig. 9 Statistical difference between mini-radars and XPol rain estimates (accumulation in 30 minutes time periods) versus
range of the mini-radars during the 8-9/10/2011 rain event for the 1.5o elevation angle of the antennas.
ERAD 2012 - THE SEVENTH EUROPEAN CONFERENCE ON RADAR IN METEOROLOGY AND HYDROLOGY
Figure 7 presents time-series comparison between the mini-radar and XPol rainfall estimates (accumulation in 30 minutes)
from the three rainfall algorithms Z-R, Rp1 and Rp2 and the raingauge measurements. It was not possible to compare the mini-
radars to all the raingauges because of partial beam blockage by the terrain in the direction of some raingauges. The time
series comparison shows that the rainfall estimates by the mini-radars follow well the temporal evolution of rain and that the
classic estimator Z-R generally underestimates the rainfall rate. It should be noted that the constants in Z-R relation were
estimated from historic XPol data in Athens, Greece. Histograms of the difference between mini-radar and XPol rainfall
estimates are shown in Fig. 8. This comparison was made after interpolation of the rainfall maps for each mini-radar and
XPol on a common grid of 90 km radius with 1 km spatial resolution and XPol position at the center of the grid. It can be
seen that the bias of the mini-radar rainfall estimates is small. Tails in the histograms are probably due to altitude differences
of the measurement volume of the mini-radars and XPol and beam blockage effects on the radar data by the terrain in some
azimuth sectors. The differences of the rainfall estimates from the polarimetric algorithm show more spread compared to the
Z-R algorithm, which is mainly due to noise in the Kdp observations by the mini-radars.
Figure 9 shows the statistics of differences in rainfall estimates (accumulation in 30 minutes time periods) against range for
the three algorithms between the mini-radars and XPol for an elevation angle of 1.5o during the stratiform rain event of 8-9
October 2011. In general, the mini-radars underestimate rainfall rate at ranges longer than 40 km, which probably has to do
with the wide beam effect of mini-radars described in section 2. The underestimation by R3 at short ranges is probably due to
the underestimation of Kdp by the mini-radar (Fig. 5). The Rp2 polarimetric algorithm performs better than the other two
algorithms, but it is probably more affected by the noise in mini-radar data as it was seen in the histogram comparison in Fig.
8. It should be noted that even though the estimates Z-R algorithm between mini-radars and XPol are similar this algorithm
underestimates rain as concluded by the comparison with raingauge data in Fig. 7.
4. Conclusions A preliminary comparison of polarimetric mini-radars, which were developed in the framework of HydroRad project,
against a benchmark polarimetric X-band radar showed promising results. The mini-radars measure accurately the spatial
variability of rain, but there are problems with partial beam blockage and significant vertical averaging especially at long
ranges due to their wide beam. New polarimetric algorithms for attenuation correction, rainfall and rain microphysics
estimation were applied satisfactorily to the mini-radar data, but further improvement in noise level which affected the
differential phase measurements is needed. Despite these limitations at longer ranges these X-band systems provide a reliable
low cost solution for weather and flood monitoring in small scales. Networks of mini-radars can cover broader areas in
complex terrain where high-power operational systems cannot achieve this.
Acknowledgment
The research leading to these results has received funding from the European Union’s Seventh Framework Programme
(FP7/2007-2013) under grant agreement n 232156 within FP7-SME-2008-1.
References
Anagnostou, E., Anagnostou M., Krajewski W., Kruger A., and Miriovsky B., 2004: High-resolution rainfall estimation from X-Band
polarimetric radar measurements. J. Hydrometeor., 5, 110–128.
Anagnostou, M. N., J. Kalogiros, F. S. Marzano, E. N. Anagnostou, M. Montopoli, and E. Picciotti, 2012: Performance evaluation of a
new dual-polarization microphysical algorithm based on long-term X-band radar and disdrometer observations. Submitted to the J.
Hydrometeor.
Bringi, V., and Chandrasekar V., 2001: Polarimetric Doppler weather radar. Cambridge University Press, Cambridge, UK.
Bringi, V., Keenan T., and Chandrasekar V., 2001: Correcting C-band radar reflectivity and differential reflectivity data for rain
attenuation: a self consistent method with constraints. IEEE Trans. Geosci. Remote Sens., 39, 1906-1915.
Kalogiros, J., M. N. Anagnostou, and E. N. Anagnostou, 2006: Rainfall retrieval from polarimetric X-band radar measurements.
Proceedings of the 4th European Conference on Radar in Meteorology and Hydrology (ERAD), Barcelona, Spain, pp. 145–148.
Kalogiros, J., M. N. Anagnostou, E. N. Anagnostou, M. Montopoli, E. Picciotti, and F. S. Marzano, 2012a: Optimum estimation of rain
microphysical parameters using X-band dual-polarization radar observables. Submitted to the IEEE Trans. Geosci. Remote Sens..
Kalogiros, J., M. N. Anagnostou, E. N. Anagnostou, M. Montopoli, E. Picciotti, and F. S. Marzano, 2012b: Evaluation of a new
polarimetric algorithm for rain path attenuation correction of X-band radar observations against disdrometer data. Submitted to the
IEEE Geosci. Remote Sens. Lett..
Kalogiros, J., M. N. Anagnostou, E. N. Anagnostou, M. Montopoli, E. Picciotti, and F. S. Marzano, 2012c: Correction of polarimetric
radar reflectivity measurements and rainfall estimates for apparent vertical profile in stratiform rain. Submitted to the J. Appl.
Meteorol. Clim..
Matrosov, S., Clark K., Martner B., Tokay A., 2002: X-band polarimetric radar measurements of rainfall. J. Appl. Meteor., 41, 941-952.
Park, S., Maki M., Iwanami K., Bringi V., and Chandrasekar V., 2005: Correction of radar reflectivity and differential reflectivity for rain
attenuation at X-band, Part II: Evaluation and application. J. Atmos. Ocean. Technol., 22, 1633-1655.
Picciotti, E., Marzano F.S., Cinque G., Montopoli M., Bernardini L., De Sanctis K., Anagnostou E., Kalogiros J., Anagnostou M., Fessas
Y., Volpi A., Telleschi A., Cazac V., and Pace R, 2012: Exploiting X-band dual-polarization mini-radar network and hydro-
meteorological forecast models in Moldova territory during the field campaign of HYDRORAD project. Proceeding of the 7th
European Conference on Radar in Meteorology and Hydrology (ERAD), Toulouse, France, 25-29 June 2012.
Testud, J., Le Bouar E., Obligis E., and Ali-Mehenni M., 2000: The rain profiling algorithm applied to polarimetric weather radar. J.
Atmos. Ocean. Technol., 17, 332-356.