The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
132 
Then, these obtained maps together with DEM, the land use 
map and soil type map participate in the model to get a rougher 
soil moisture distribution map. This map is resampled into the 
same size as the coherence Thus, a preliminary comparison and 
analysis between coherence image and soil moisture map can be 
done. In this section, the detail methods of acquiring soil 
moisture and InSAR coherence are introduced. 
3.1 Soil moisture 
The distributed hydrologic model which is based on raster data 
model in GIS and popular at presmt is used to build the 
distributed model of soil moisture. There are four models: soil 
évapotranspiration model, raster water flow model, rainfall 
runoff model and soil moisture model. In the model, three main 
factors that affect soil moisture are considered. That is rainfall, 
runoff and soil évapotranspiration. The point data on the rainfall 
provided is interpolated to get rainfall distribution maps of the 
whole area on which the physical processes of rainfall runoff is 
simulated. Other meteorological data are processed in the same 
way of interpolation to calculate daily soil évapotranspiration of 
the area by the function of raster calculator in ArcGIS 
according to the évapotranspiration model. 
Based on the soil water balance principle, the soil moisture 
model uses the other three models to get the final soil moisture. 
When no rainfall happens, soil évapotranspiration model is used. 
In the area where there is rainfall, raster water flow model and 
rainfall runoff model are needed. 
Actual soil évapotranspiration is calculated according to 
Penman equation. 
Raster water flow model is based on DEM. Compare slope 
drops between the processing cell with the nearest eight cells. 
The direction of the line that join the processing cell center with 
the one which has the largest gradient among the eight is 
defined as the water flow direction of the processing cell, and 
the water flow direction of a cell is represented by a number. 1, 
2, 4, 8, 16, 32, 64 and 128 correspond to east, southeast, south, 
southwest, west, northwest, north and northeast respectively. 
The gradient equation is as follows 
Z -Z 
MD = —— - x 100 (i) 
J{X k - Xi ) 2 HY k -Y t ) 2 
0 (P-0.2S) 2 
^ (P + O.&S) 
Q = o 
P>0.2S 
(2) 
P < 0.2S 
Where, Q is daily runoff (mm); P is daily rainfall (mm); S is 
retention parameter which is different with area and time and is 
related with the soil, slope, land use, management and prior soil 
moisture. 
The following equation links S with curve number CN: 
5' = 254(—-1) 
CN 
(3) 
The range of CN is 0~100(does not equal 0). When CN is 100, 
S is 0 and Q equals P. 
3.2 Coherence 
The coherence is the key product of SAR interferometric 
process, which is the magnitude of the complex correlation of 
both amplitude and phase information from two interferometric 
signals. It is an important parameter to measure InSAR data 
quality because the area with high coherence could usually 
obtain qualified interferometric measurements. The coherence 
of two co-registered complex SAR images Ij and I 2 can be 
defined as: 
Where, y stands for the coherence. The brackets () is the 
estimated ensemble average and * denotes the complex 
conjugate. The value of y is in the range [0.0, 1.0]. 
The degree of coherence that is estimated from a complex SAR 
image pair can be considered as the product of different de- 
correlation factors as long as the sources of decorrelation are 
statistically independent. 
Where, MD is the gradient between two cells; (X k , Y k ) and (X i; 
Y ; ) represent the coordinates of the calculating cell and the 
nearest ones respectively; Z k and Z\ represent their elevation 
values. 
Y ~ y System SNR N Processor y Baseline v R egistration ZTemporal 
Curve number method created by Soil Conservation Board in 
United States Department of Agriculture is widely used by all 
countries so far and is accepted to calculate rainfall runoff on 
our study.. In this method, daily rainfall data easily acquired is 
used and runoff of land surface is connected with soil type, land 
use, management, etc. This method links runoff with daily 
rainfall directly and the relationship between runoff and rainfall 
is: 
Where, SNR is the radar system signal-to-noise-ratio. yp r0 cessor is 
de-correlation due to the SAR processor, which is the 
processing stages from the SAR raw data to the single-look- 
complex (SLC) SAR image product. Baseline stands for the 
across-track distance between the two satellite passes, and 
Registration is referring to the accuracy a SLC SAR image pair 
can achieve in the coregistration process. 
The four first terms on the right-hand side of the equation are 
factors that one will desire to minimize (i.e., obtain values close 
to 1.0) so that the estimated coherence in an area is 
corresponding more or less to the amount of temporal