Update status of soil moisture algorithm

1
Update status of soil moisture algorithm
H.Fujii (JAXA/EORC) and T.Koike(UT)

• 2008.7.15
• Joint AMSR Science Team Meeting,
• July 14-16, Telluride, CO

2
Defect of Koike's Algorithm
"There are significant weaknesses in the JAXA
product over vegetated sites" (T.J. Jackson, The
GLI / AMSR workshop 2008 in Atami Japan,
Jan.23,2008)
Fig Scatter plots of AMSR-E Monthly Soil
moisture (y-axis) versus in-situ data (x-axis) at
the Little River ,GA,USA
monthly volumetric soil moisture in
60

• Match-up
• Data Period
• January, 2003 -- December 2007 (60 months)
• In-Situ
• data source SCAN 2027 Little River, GA, U.S.A
• sensor depth 5cm
• sampling method observed data which has closest
  timestamp to AMSR-E observation time within 1
  hour
• AMSR-E
• data source JAXA product ver.5 Level2
• sampling method closest FOV to the SCAN point
  within 5km range.

AMSR-E
0
0
60
In-situ (SCAN2027, LR, 5cm depth)
3
Update log of Koikes algorithm

• March, 2004 algorithm ver.3.01 (product
  ver.2-5)
• Radiative transfer model for soil layer
• Fractional vegetation cover , fc TB
  (1-fc)TBbare fcTBvege
• March, 2008 algorithm ver.5.11 (product ver.6)
• Developing of the global data set of fractional
  vegetation cover, fc, (3 years average) from
  Terra/Aqua MODIS(LP DAAC).
• Reconfiguration of vegetation parameter set , b
  and w, based on the result of Ground Based
  Experiment.
• Jun, 2008 algorithm ver.5.33 (prototype
  product)
• Expansion of data period of fc global data set
  ( Jun 2002 - May 2008)
• Reconfiguration of roughness parameter set , Q
  and h

4
Overview of Koike Algorithm
Terra/Aqua MODIS Vegetation Indices (LP
DAAC) MOD13A2/MYD13A2 (1km, 16days, v5)
NDVI
Global Data Set of fractional vegetation cover,fc
Period Jun 2002 May 2008 Data size 8 days,
0.05x0.05 grid Resolution equivalent size to
AMSR-E 10GHz FOV fc model Carlson and replay,
1997
Soil Moisture Algorithm
5
AMSR-E Global Soil Moisture Map
6-yeas average of July (2002-2007) , Descending
6
Comparison of AMSR-E Soil Moisture and
Precipitation map

• AMSR-E Soil Moisture
• L2, Descending
• Volumetric Soil Moisture

• Precipitation
• Total amount of precipitation for 48 hours
  before AMSR-E observation.
• Data source GSMaP MVK hourly
  (JST-CREST/GSMaP )

7
Validation 1
Seasonal Changes of EVI in 2003
1.0
0.8
0.6
EVI
0.4
0.2
0.0
2003 Jan.
Apr.
Jul.
Oct.

1. Gaize, Tibet, China. (CEOP)location 32.30N,
   84.05E
2. Balranald Bolto, NSW, Australira.
   (CEOP)location 34.658S, 143.549E
3. Little River, GA, USA. (SCAN, NRCS)location
   31.50N, 82.45W

8
Validation 1 New product (prototype)
1.0
0.8
0.6
EVI
0.4
0.2
0.0
60
Number of Data Des. 214 Asc. 200
(a) Tibet Gaize sensor depth 3cm
(1.5 - 4.5cm)
30
40
0
60
0
Number of Data Des. 217 Asc. 223
20
(b) Balranald Boltosensor depth 0-7cm
30
40
0
Volumetric Soil Moisture
Daily Precipitation mm/day
60
0
Number of Data Des. 200 Asc. 198
20
(c) Little Riversensor depth 5cm
(3-7cm)
30
40
0
2003 Jan.
Apr.
Jul.
Oct.
9
Validation 1 JAXA product (ver.5)
1.0
0.8
0.6
EVI
0.4
0.2
0.0
60
Number of Data Des. 93 Asc. 167
(a) Tibet Gaize sensor depth 3cm
(1.5 - 4.5cm)
30
40
0
60
0
Number of Data Des. 206 Asc. 220
20
(b) Balranald Boltosensor depth 0-7cm
30
40
0
Volumetric Soil Moisture
Daily Precipitation mm/day
60
0
Number of Data Des. 187 Asc. 184
20
(c) Little Riversensor depth 5cm
(3-7cm)
30
40
0
Apr.
Jul.
Oct.
2003 Jan.
10
Validation 2 Monthly Soil Moisture at LR.
Algorithm Orbit Bias RMSE MAE
(a) JAXA ver.5 Des. 11.5 4.8 11.5
(a) JAXA ver.5 Asc. 9.0 4.2 9.0
(b) new (Prototype) Des. 3.1 2.5 3.5
(b) new (Prototype) Asc. 2.9 2.4 3.5
Scatter plots of AMSR-E monthly averaged soil
moisture(y-axis) versus in-situ data at Little
River , GA in U.S.A.. The data period is from
January 2003 to December 2007 (60 months).
unit volumetric soil moisture in
11
Summary

• The new global data set of fractional vegetation
  cover, fc, is introduced into the algorithm
• The roughness parameter calibrated at Mongolia is
  used in the Ratiative Transfer Model.
• The algorithm validated at Little River and the
  result showed that the accuracy was improved by
  comparing the old algorithms.
• Near future plan.
• Masking method for uniformly and dense vegetated
  area.
• Behavior of the algorithm in Sahara

12
(No Transcript)
13
(No Transcript)
14
Impact evaluation of interannual variation of
vegetation cover on AMSR-E Soil Moisture
retrieval.
July, 2003, Descending, Monthly average.
(a) fc actual time
(b) fc 6 years average
(c) Difference ( b a )
15
Monthly Soil Moisture Map in Africa (July, 2003)