Post on 27-Nov-2021
Retrieval of vegetation biophysical Retrieval of vegetation biophysical parameters by inverting parameters by inverting
hyperspectralhyperspectral, , multiangularmultiangularCHRIS/PROBA Data from SPARC CHRIS/PROBA Data from SPARC
20032003
D'UrsoD'Urso G., Dini L., Vuolo F., Alonso L.G., Dini L., Vuolo F., Alonso L.
ITAP, Albacete
Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas, Madrid
Universitá degli Studi di Napoli �Federico II�, Italy
University of Thessaly, Greece
INRA-CSE, Avignon
Laboratoire du Télédétection et SIRS, Tunisia
Meteo-France
European Spatial Agency (ESA)
University of Valencia - Remote Sensing Unit
University of Valencia - Global Change Unit
University of Valencia - Solar Radiation Unit
University of Valencia � GPDS
University of Castilla-La Mancha
Institute of Regional Development (IDR), Albacete
SPARC SPARC campaigncampaign((BarraxBarrax, 12, 12--14 14 JulyJuly 2003)2003)
AimAim ::
Assessment of retrieval accuracy by using :- RT models vs. empirical approaches (i.e. veget. Indexes)- multi-angular and/or super spectral info
Retrieval of canopy parameters (in particular LAI) from E.O. data for :- calculation of crop transpiration and soil evaporation (P-M approach)- soil water balance simulations (input forcing)
FIELD DATA
LAI measurements
113 Elementary Sampling Units
(24 data samples each ESU)
b
bb
b bb
bb b
b bbb
bbb b
bbbbbbbb
bb bbb
b
bb
bb bb
bbb
bb
bb
bbbbbb
bbbbb
b
bbbbbbbbbbb
bbbbbb
bb
b
bbb
bbb
bbbbbb
bbbbb
bbbbbb
b
bb b bb
b
bbb
bbb
LAI measurements
7 types of crop:
alfalfacornsugarbeetonionsgarlicpotatopapaver
0%
2%
4%
6%
8%
10%
12%
0.6
0.9
1.2
1.5
1.8
2.1
2.4
2.7 3
3.3
3.6
3.9
4.2
4.5
4.8
5.1
5.4
5.7 6
6.3
LAI
avg = 3.07; std = 1.45
LAI = 1.32
LAI = 2.49
LAI = 3.72
Alfalfa, LAI = 3.72
Sugarbeet, LAI = 3.78
Corn, LAI = 3.84
Chlorophyll MeasurementsGood correlation between laboratory and field measurements for different crops
i.e. 4000+ valid chlorophyll measurements
0.010
0.020
0.030
0.040
0.050
0.060
0.070
1 10 100 1000
Chlorophyll Units
2 )
Clor.A1 (mg/cm2)Clor.C1 (mg/cm2)Clor.B1 (mg/cm2)Clor.W1 (mg/cm2)Clor.G1 (mg/cm2)Clor.ON1 (mg/cm2)Clor.P1 (mg/cm2)
Chl
orop
hyll
(mg/
cm2 )
S6S6
S5S5
S3S3
S0S0
WW
AA
Sampling Points for Radiometric Calibration
Sampling Points for Radiometric Calibration
targets: soil, vegetation
radiometers inter-comparison at Las Tiesas14 july
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
300 400 500 600 700 800 900 1000 1100
wavelength (nm)
refle
ctan
ce
LAI,LIDF,HOT,Esky
ModelsModels parametersparametersPROSPECT requires:PROSPECT requires:!! Leaf Leaf mesophyllmesophyll structure, Nstructure, N!! Chlorophyll Chlorophyll a+ba+b content, Cab content, Cab (mg cm(mg cm--22))!! Equivalent water thickness, Equivalent water thickness, CwCw (gcm(gcm--2)2)!! Dry matter content, Cm Dry matter content, Cm (gcm(gcm--2).2).
SailHSailH requires:requires:!! Leaf Area Index: LAILeaf Area Index: LAI!! Leaf inclination distribution function: Leaf inclination distribution function:
LIDF LIDF !! Leaf Leaf relectancerelectance and and trasmittancetrasmittance
(PROSPECT)(PROSPECT)!! Soil spectral reflectance, which is Soil spectral reflectance, which is
assumed to be Lambertianassumed to be Lambertian!! Solar zenith (Solar zenith (qqss) and azimuth angle () and azimuth angle (YYss) ) !! View zenith (View zenith (qqvv) and azimuth angle () and azimuth angle (YYvv))!! Fraction of incident diffuse skylight Fraction of incident diffuse skylight
expressed in terms of visibility, expressed in terms of visibility, EskyEsky!! KuuskKuusk hot spot size parameters, shot spot size parameters, s
FORWARD SIMULATION
Alfalfa measured groundreflectance (ASD ASD FieldSpecFieldSpec) and PROSPECT/SAILH simulated reflectance byusing different background measured spectra.
Data from SPARC 2003
Soil reflectance: ground measurement
FIELD MEASUREMENTSFIELD MEASUREMENTS
0
0,1
0,2
0,3
0,4
0,5
0,6
300 400 500 600 700 800 900 1000 1100
wavelength (nm)
refle
ctan
ce
soil (5-1.001)soil (5-2.001)soil (5-3.001)soil (5-4.001)soil (5-5.001)
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
300 400 500 600 700 800 900 1000 1100
wavelength (nm)
refle
ctan
cesoil (6.001)
soil (6-2.001)
soil (6-4.001)
soil (6-5.001)
Best fit soil
RRMSE=0.0207
Soil Analsysis for CHRIS/PROBA reflectance simulation on Barrax site.
CHRIS max soilreflectance
CHRIS min soilreflectance
Ground measurement
CHRIS mean soilreflectance
1 Chris/proba bands
Ref
lect
ance
0.7
0.6
0.5
0.4
0.2
0.3
0.1
0101 20 30 5040 60
+55: VZA= 55.99°; VAA= 26.11°+36: VZA= 38.78°; VAA= 37.98º0: VZA= 19.40°; VAA= 102.40º-36: VZA= 39.15°; VAA= 165.44º-55: VZA= 56.24°; VAA= 177.06º
SUN ZENITH ANGLE= 22.4°SUN AZIMUTH ANGLE= 134.7°
[0º]
+55: VZA= 57.29°; +36: VZA= 42.44°; VAA= 339.44º0: VZA= 27.4°; VAA= 285.27º-36: VZA= 42.53°; VAA= 231.22º-55: VZA= 57.4°; VAA= 216.91º
SUN ZENITH ANGLE= = 19.8°SUN AZIMUTH ANGLE= 148.3°
12th of July Acquisition
14th of July Acquisition
+36° 0°
12/07/2003
+36° 0°
14/07/2003
Alfalfa: Forward
+55° 0°
Potatoes: Forward
+55° 0°
12/07/2003
14/07/2003
MODELINVERSION
Inversion algorithm - 1PEST-ASP using Gauss-Marquardt-Levenberg estimation techniquePEST runs the PROSAILH model, compares the model results with the target values (observed reflectance values), adjustsselected parameters using optimisation algorithm and runs the model as many times as is necessary in order to determine the optimal set of adjustable parameters
Inversion algorithm - 2
Parametersestimate
( LAI )
• LUT (look-up table) using RRMSE (relative mean square error)
+55+36 0 -36 -55
+55+36 0 -36 -55
[ ]
[ ]∑∑
∑∑
= =
= =
−= 5
1
62
1
2
5
1
62
1
2
),(
j imeas
j iestmeas
ij
(j,i)ρ (j,i)ρRRMSE
ρ
(Privette, 1994)
PEST-ASP theory
( )( ) ( )( )0000 bbJccQbbJcc t −−−−−−=Φ• Objective function :
Where:
• b0 : parameters vector to be upgraded
• b : parameter vector upgraded
• c0 = PROSAILH ( b0 ) : model calculated observations vector
• c : experimental observation vector
• J : Jacobian matrix of PROSAILH
• Q : Observation weigths matrix
( ) ( )01
0 ccQJIQJJbb tt −+=− −α• Algorithm by which the system parameter vector is estimated :
Marquardt-LevenbergWhere:
I : Identity matrix
α: Marquardt parameter
Parameters rangePROSPECT N=[1 � 3 ]Cab=[10 � 110]Cw=0.022Cm=[0.001 � 0.02]SAILHLAI=[0.3 � 8]HOT=[0.0001 � 1]LIDFS=[1 2 3 4 5 6 7 8 9 10 11 12 13]Esky=0.13
Initial parameters vector estimatePROSPECTN=1Cab=10Cw=0.022Cm=0.001SAILHLAI=0.4HOT=0.05LIDFS=1Esky=0.13
PEST-ASP settings
Contours of equalΦΦΦΦ
Initial parameterestimation
Parameter # 1
Parameter # 2
PEST-ASP results ILAI PEST ASP
0,00
1,00
2,00
3,00
4,00
5,00
6,00
7,00
8,00
9,00
0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00 9,00
Measured LAI
Estim
ated
LA
I
LAITheor.
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00
Measured LAI
Estim
ated
LA
I AlfalfaPotatoesCornSugarbeetOnionTheor.
Potatoes
Alfalfa
RRMSE = 0.40
PEST-ASP results II
Alfalfa1 < N < 2
LIDFS Plagiophile
Meas. Estim.
LAI:3.24 3.30
Cab: 63 27(25-35)
+55 +36 0-36 -55
A priori A priori knowledgeknowledge
PEST-ASP results II
Potatoes1 < N < 1.6
HOT < 0.0027
LIDFS Uniform
Meas. Estim.
LAI: 5.20 5.10
Cab: 35 14(10-14)
+55 +36 0-36 -55
A priori A priori knowledgeknowledge
PEST-ASP results IIA priori A priori knowledgeknowledge
SugarbeetN = 1.5
HOT < 0.6Meas. Estim.
LAI: 4.08 4.05
+55 +36 0-36 -55
PEST-ASP results IIA priori A priori knowledgeknowledge
0
1
2
3
4
5
6
7
8
0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00
Measured LAI
Estim
ated
LA
I
- 15%
+15%
RRMSE = 0.11
Potatoes
Sug. Beet
Alfalfa
585000
Simulatedspectral profiles
Points in the parameter space are uniformely taken:
PROSPECT N=[1.5 1.7 … 2.5]Cab=[10 15 … 70]Cw=0.011Cm=[0.002 0.004 … 0.02]SAILHLAI=[1 1.2 … 6.8]HOT=[0.05 0.15 … 0.5]LIDFS=[Planophile; Plagiophile; Extremophile; Erectophile; Spherical]Esky=0.13
Geometry of illumination and observation was fixedby time of acquisition and Chris/PROBA orbit
CHRIS/PROBA 12/07/2003 Barrax
Image extracted spectral profile
Estimated parameters (LAI)
RRMSE minimum
Look-up table theory and settings
0
1
2
3
4
5
6
7
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00
Measured LAI
Estim
ated
LA
I
LAITheor.
Look-up table results
0
1
2
3
4
5
6
7
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00
Measured LAI
Estim
ated
LA
I AlfalfaPotatoesOnionCornSugar beetTheor
Potatoes
Alfalfa
RRMSE = 0.55
+ 55 + 36 0- 36 - 55
AlfalfaLAI Estim. 1.8
LAI Meas. 1.9
Look-up table Alfalfa
Look-up table Potatoes
+55+36 0 -36 -55
PotatoesLAI Estim. 2.6
LAI Meas. 5.6
Conclusion
• LAI WAS ESTIMATED WITH AN ACCURACY OF AROUND 15% FOR CROPS CLOSE TO THE TURBID MEDIUM HYPOTHESIS
• PEST IS A GOOD TOOL TO ESTIMATE LAI OF SOME CROPS (ALFALFA AND POTATOES) WITH LITTLE A PRIORI KNOWLEDGE. PROBABILY FOR DIFFERENT CROPS (CORN, ONION) IT IS NEEDED TO ADD MORE A PRIORI KNOWLEDGE TO AVOID “ILL-POSED INVERSION PROBLEM”
• LUT PROBABILY NEED FINEST AND BETTER PARAMETERS SPACE SAMPLING TO BETTER ESTIMATE BIOPHYSICAL PARAMETERS
Future steps
• WE HAVE TO BETTER DEFINE A PRIORI KNOWLEDGE FOR CORN, ONION, WHEAT
• WE NEED TO BETTER UNDERSTAND THE INFLUENCE OF THE SOIL IN THE RADIOMETRIC SIGNAL FOR MULTI-ANGULAR AND HYPERSPECTRAL SATELLITE DATA ON VEGETATION WITH DIFFERENT LAI.
• WE HAVE TO TEST DIFFERENT INVERSION ALGORITHMS (NEURAL NETWORKS, GENETHIC ALG.)