-B3, 2012 
0 (maximum 
ibility) based on 
VI results were 
oil moisture data 
tween these two 
y Edge 
   
let Edge 
ve 
NDVI 
I adapted from 
ll data obtained 
0 investigate the 
to facilitate this 
g the rain gauge 
results indicated 
11 in most cases 
after rain events, 
ain several days 
For example, at 
17, 73 and 225 
sed again after 
3a) The same 
where the TVDI 
the rain events, 
81, 137 and 241 
  
Bainfali iem) 
  
  
  
: 8 
300 350 
and (b) Suao. 
  
  
32 Spatio-temporal evolution of surface soil moisture 
Figure 4 shows the spatio-temporal evolution of TVDI from 
January to December 2009. In general, the TVDI results (values 
from 0 to 1) show a large degree of variation in TVDI values over 
space and time. Low TVDI values (0 — 0.4) were generally 
observed at water surfaces and high elevation mountain areas. 
Higher TVDI values (0.4 - 0.6) were observed in the plains and 
lower elevation parts of the study area. High TVDI values (0.6 - 
0.8 and 0.8 - 1.0) were observed mainly in coastal and residential 
areas. The regression analysis between TVDI results and 
AMSR-E soil moisture data (Figure 5) was performed to find the 
transformation function between the two datasets. This was done 
to convert the TVDI to the real soil moisture data that had the 
same unit with AMSR-E soil moisture data (i.e. g cm”). The 
results indicated that there was good agreement between the two 
January 11 February 25 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
   
datasets. The mean correlation coefficient was -0.82 and the root 
mean squared error (RMSE) to quantify the difference between 
the two datasets was 39.8. These results confirmed that the 
AMSR-E soil moisture data (25-km resolution) can be used to 
transform TVDI results to surface soil moisture data with a 
higher spatial resolution (1-km resolution). 
Figure 6 showed the spatial distributions of surface soil 
moisture (achieved by regression analysis between TVDI and 
AMSR-E soil moisture). In general, the spatio-temporal 
evolution of TVDI-derived soil moisture was comparable with 
those from AMSR-E data throughout the year. The areas of low 
soil moistures were generally concentrated in the west part of the 
study area, where most residents settled in the cramped plains. 
The areas of low soil moistures were expanded from April, but 
likely returned to wet conditions by the end of September due to 
the monsoonal influence. 
March 16 April 28 
ga = 
     
  
November 04 
Figure 4. Spatiotemporal evolution of monthly TVDI during 2009.