Soil moisture retrieval from SMAPVEX12 and SMAPVEX16-MB data
Ramata Magagi, Kalifa Goïta
Honqguan Wang (PDF)
Vincent Beauregard (M. Sc. student)
With the collaboration of Andreas Colliander, Alexandre Roy and Tom Jackson
Outline
• Context
• Results from recent investigations with SMAPVEX12 data
- Empirical soil moisture retrievals over wheat fields;
- Polarimetric soil moisture retrievals;
• Current works on SMAPVEX12 and SMAPVEX16-MB data
2
ContextProject designed in the context of the Canadian science plan for SMAP mission
Financial partners:
Canadian Space Agency (CSA)
Natural Sciences and Engineering Research Council of Canada (NSERC)
Collaborators:
Environment Canada (EC)
Agriculture and Agri-Food Canada (AAFC)
University (Sherbrooke, Manitoba, Guelph)
National Aeronautics and Space Administration (NASA), USDA, JPL
CESBIO, France
DLR Microwaves and Radar Institute, Germany3
Study area and data
4
PALS Low altitude Flight lines
X
X
X Ground radiometers
SMAPVEX12 and SMAPVEX16-MB fields
Collection of ground (soil and vegetation) data
network data (SM and temp.)
L-band radiometers
airborne PALS data
airborne UAVSAR data (2012 only)
satellite data
Empirical soil moisture retrievals over wheat fields
from RADARSAT-2 data
5
Empirical Soil moisture estimation over wheat fields
from RADARSAT-2
• Multiple linear regressions
𝒎𝒗 = 𝜷𝟏𝑿𝟏 +⋯+ 𝜷𝒊𝑿𝒊 +⋯+ 𝜷𝒑𝑿𝒑 + 𝜷𝟎 + 𝝐
1) Consider all the variables 𝑿𝒊 (incoherent + coherent)
2) Reduce the dimension by using only selected non-correlated variables
3) Use stepwise regression
6
Empirical Soil moisture estimation over wheat fields (cont.)
• Results from SMAPVEX12
Model Nb of fields Nb of
Observations R2
RMSE
(m³/m³) Variables β p-value
4 8 60 0.598 0.078 𝝈𝑯𝑯𝟎 0.107 0.000
𝝈𝑯𝑽𝟎 -0.117 0.000
𝝈𝑽𝑽𝟎 0.057 0.005
𝝓𝑯𝑯−𝑽𝑽 -0.003 0.000
Hs 2.072 0.000
A -1.637 0.008
5 8 60 0.510 0.084 (Intercepte) 0.662 0.000
𝝈𝑯𝑯𝟎 0.036 0.000
𝝓𝑯𝑯−𝑽𝑽 -0.004 0.000
7
Polarimetric soil moisture retrievals
8
Polarimetric soil moisture retrievals
• Compare 3 model-based polarimetric decompositions
(Freeman-Durden, Hajnsek and An) for soil moisture
retrievals from L-band UAVSAR data;
• Apply the best approach to Multi-angular UAVSAR
9
Freeman-Durden Hajnsek An Classified Image 2012-06-17
0
10
20
30
40
50
60
Surface Dihedral Volume
Perc
ent (%
)
Scattering mechanisms
Freeman
Hajnsek
An
Comparison of 3 polarimetric methods : RGB composite of scattering powers
Dihedral
SurfaceVolume Wang et al. 2017 RSE (under review)
More surface scattering power in An decomposition10
Surface
Dihedral
Volume
Wang et al. 2016 Remote sensing
Wang et al. 2016 IEEE GRSL
Comparison of 3 polarimetric methods (cont.)
Results of soil moisture retrieval from the dominant surface or dihedral scattering component
Best retrieval with Hajnsek method 11
Wang et al. 2017 RSE (under review)
Wang et al. 2016 Remote sensing
Wang et al. 2016 IEEE GRSL
Polarimetric decomposition of Multi-angular UAVSAR data
Surface scattering component
#3160312
#31604 #31605 #31606
Flight lines
(2012-06-17)
14 acquisition days
Wang et al. 2017 IEEE TGRS (under review)Wang et al. 2016 IEEE GRSL
Polarimetric decomposition of Multi-angular UAVSAR data (cont.)
• Results
Retrieval rate RMSE
Multi-incidence angle increases the retrieval rate and decreases the RMSE
165 170 175 180 185 190 195 200 2050
10
20
30
40
50
60
70
80
90
100
DOY 2012
Retr
ieval ra
te (
%)
-1
-2
-3
-4
Multi-
13
Wang et al. 2017 IEEE TGRS (under review)Wang et al. 2016 IEEE GRSL
Current works with SMAPVEX12 and SMAPVEX16-MB
data
14
Preliminary analyses
Field campaigns Data analyses
SMAPVEX16-MB L-band Radiometers :
Ground vs soil moisture
Ground vs low altitude PALS
Low altitude PALS vs soil moisture
SMAPVEX12 and SMAPVEX16-MB Comparison of low altitude PALS
Modelling results
15
SMAPVEX 12 & 16-MB PALS TB (Low altitude) to soil moisture
Objective : Understand the effects of soil and vegetation conditions
on TB for soil moisture retrieval
Ground radiometer TB versus soil moisture• Field #105: Wheat Field #202: Canola (Low vegetation)
16
TBH
TBV
TBH
TBV
Less vegetation over Field #202 Higher sensitivity to soil moisture
Ground radiometer versus PALS
• Field #105: Wheat
17
TBH
TBV
TB of radiometer over wheat field Higher than PALS
Ground radiometer versus PALS (cont.)
• Field #202: Canola
18
TBH
TBV
TB of radiometer over canola field Lower than PALS
Preliminary analysesSMAPVEX 12 & 16-MB PALS TB (Low altitude) to soil moisture
0 0.1 0.2 0.3 0.4 0.5 0.6100
150
200
250
300
June:R=-0.67
July: R=-0.24
Ground measured soil moisture (m3/m3)
PA
LS
passiv
e T
BV (
K)
June
July
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
June:R=-0.62
July: R=-0.44
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BH (
K)
June
July
24 days observation gap from June to July(very different vegetation status )
24 days observation gap from June to July(very different vegetation status )
Low vegetation
High vegetation (24 days later on)
Low vegetation
High vegetation (24 days later on)
More vegetation effect in July 14-22than in June 08-20
TBH TBV
Sensitivity of SMAPVEX16 PALS passive signal to soil moistureCanola
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m
3)
PA
LS
passiv
e T
B H (
K)
43
202
203
204
224
234
235
237
June
8 FieldsDifferent color & shape represent different fields
Field #
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
passiv
e T
BH (
K)
43
202
203
204
224
234
235
237
July
8 Fields
Field #
20
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
June:R=-0.063
July: R=-0.46
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BH (
K)
June
July
0 0.1 0.2 0.3 0.4 0.5 0.6100
150
200
250
300
June:R=-0.16
July: R=-0.43
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BV (
K)
June
July
Sensitivity of SMAPVEX16 PALS passive signal to soil moisture
Diversity in soil conditions particularly for soybean;
For June and July, clusters of points.
Soybean
TBH TBV
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m
3)
PA
LS
passiv
e T
B H (
K)
41
71
72
101
113
210
223
231
232
236
June
10 FieldsDifferent color & shape represent different fields
Field #
0 0.1 0.2 0.3 0.4 0.5 0.6100
150
200
250
300
Ground measured soil moisture (m3/m3)
PA
LS
passiv
e T
BH (
K)
41
71
72
101
113
210
223
231
232
236
July
10 Fields
Field #
21
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
June:R=-0.36
July: R=-0.36
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BV (
K)
June
July
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
June:R=-0.47
July: R=-0.36
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BH (
K)
June
July
Less sensitivity to soil conditions
TBH TBV
Sensitivity of SMAPVEX16 PALS passive signal to soil moistureWheat
DTB over wheat fields not large as for other crops
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m
3)
PA
LS
passiv
e T
B H (
K)
31
32
62
91
104
105
233
June
7 Fields
Different color & shape represent different fields
Field #
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
passiv
e T
BH (
K)
31
32
62
91
104
105
233
JulyField #
7 Fields
Preliminary analyses
Field campaigns Data analyses
SMAPVEX16-MB L-band Radiometers : Ground vs soil moistureGround vs low altitude PALSLow altitude PALS vs soil moisture
SMAPVEX12 and SMAPVEX16-MB Comparison of low altitude PALS
Modelling results
23
SMAPVEX 12 & 16-MB PALS TB (Low altitude) to soil moisture
Objective : Understand the effects of soil and vegetation conditions on TB for soil moisture retrieval
Soil moisture conditions in 2012 and 2016
155 160 165 170 175 180 185 190 195 200 2050
0.1
0.2
0.3
0.4
0.5
0.6
DOY 2012
Gro
und m
easure
d m
v (
m3/m
3)
155 160 165 170 175 180 185 190 195 200 2050
0.1
0.2
0.3
0.4
0.5
0.6
DOY 2016
Gro
und m
easure
d m
v (
m3/m
3)
SMAPVEX12
SMAPVEX16-MB
24
• Ground measurements mv
• Ground measurements VWC
160 165 170 175 180 185 190 195 200 2050
1
2
3
4
5
DOY
VW
C (
kg/m
2)
Canola 2012
Corn 2012
Soybean 2012
Wheat 2012
Canola 2016
Corn 2016
Soybean 2016
Wheat 2016
25
Vegetation conditions in 2012 and 2016
Lower VWC in SMAPVEX16-MB than SMAPVEX12
26
Canola 2012
Corn 2012
Soybean 2012
Wheat 2012
Canola 2016
Corn 2016
Soybean 2016
Wheat 2016
Temperature conditions in 2012 and 2016
Canola 2012
Corn 2012
Soybean 2012
Wheat 2012
Canola 2016
Corn 2016
Soybean 2016
Wheat 2016
Vegetation temperature
Soil temperature
Lower in 2016
Lower in 2016
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BH (
K)
June 2012
July 2012
June 2016
July 2016
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
passiv
e T
BV (
K)
June 2012
July 2012
June 2016
July 2016
Canola
Comparison between 2012 and 2016 PALS data
27
TBH TBV
Low soil moisture during SMAPVEX12 Low contrast between June and July 2012
High soil moisture during SMAPVEX16 High contrast between June and July 2016
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BH (
K)
June 2012
July 2012
June 2016
July 2016
0 0.1 0.2 0.3 0.4 0.5 0.6100
140
180
220
260
300
Ground measured soil moisture (m3/m3)
PA
LS
pa
ssiv
e T
BV (
K)
June 2012
July 2012
June 2016
July 2016
Soybean
Lower VWC & Surface and vegetation temperature for all crops during SMAPVEX16-MB
TB in 2016 should be lower than TB 2012: not always observed
28
Comparison between 2012 and 2016 PALS data
TBH TBV
DT in SMAPVEX12 > DT in SMAPVEX16-MB
Wheat
29
Comparison between 2012 and 2016 PALS data
TBH TBV
Lower VWC & Surface and vegetation temperature for all crops during SMAPVEX16-MB
TB in 2016 should be lower than TB 2012: not always observed
Preliminary analyses of modelling results obtained with
SMAPVEX12 and SMAPVEX16-MB data
Objective : Soil moisture mapping
• Use of SMAPVEX12 PALS data into L-MEB model to estimate bp
parameters (p is the polarization);
• Analyze the performance of bp parameters with SMAPVEX16-MB
PALS data;
• Develop the retrieval approaches (not yet done) 30
(1 R ) (1 )(1 )T (1 )(1 )T Rp p p soil p vege p vege p pTB T
/cos( ) * /cos( )p pb VWC
p e e
Preliminary analyses of modelling results with SMAPVEX12 data
(cont.)
• Optimal b parameters obtained from low altitude PALS TB in 2012
31
• Performance of optimization
Preliminary analyses of modelling results with SMAPVEX12 data (cont.)
Soybean (9 fields, 17 days)
Wheat (4 fields , 17 days)
Canola (1 field , 17 days)
32
TBH
TBH
TBH
TBV
TBV
TBV
The b parameter of canola based only one field
Preliminary analyses of modelling results with SMAPVEX16-MB data (cont.)
100 150 200 250 300100
150
200
250
300
Measured TBH (K)
Sim
ula
ted
TB
H (
K)
R=0.58
RMSE=21.96 K
100 150 200 250 300100
150
200
250
300
Measured TBV (K)
Sim
ula
ted
TB
V (
K)
R=0.58
RMSE=23.30 K
Soybean (10 fields , 12 days)
100 150 200 250 300100
150
200
250
300
Measured TBH (K)
Sim
ula
ted
TB
H (
K)
R=0.23
RMSE=20.36 K
100 150 200 250 300100
150
200
250
300
Measured TBV (K)
Sim
ula
ted
TB
V (
K)
R=0.19
RMSE=13.63 K
Wheat (7 fields , 12 days)
100 150 200 250 300100
150
200
250
300
Measured TBH (K)
Sim
ula
ted
TB
H (
K)
R=0.12
RMSE=34.90 K
100 150 200 250 300100
150
200
250
300
Measured TBV (K)
Sim
ula
ted
TB
V (
K)
R=0.044
RMSE=25.99 K
Canola (8 fields , 12 days)
Calibrated L-MEB underestimates PALS TB in 2016
Roughness or vegetation water content or both??33
TBH
TBH
TBH
TBV
TBV
TBV
• Performance of b parameter
Perspectives
• Refine the data analysis and model calibration;
• Develop the retrieval algorithms of soil moisture;
• Soil moisture disaggregation with RADARSAT-2.
34
Acknowledgments
Canadian Space Agency (CSA)
Natural Sciences and Engineering Research Council of Canada (NSERC)
Environment Canada (EC)
Agriculture and Agri-Food Canada (AAFC)
University (Sherbrooke, Manitoba, Guelph)
National Aeronautics and Space Administration (NASA)
United States Department of Agriculture (USDA)
Jet Propulsion Laboratory (JPL)
CSA-SOAR program
CESBIO, France
DLR Microwaves and Radar Institute, Germany35