The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
For the ASTER scene used in our study (collected on 
11/24/2001) the sun information checked from ASTER HDF 
head file was as follows: 
Sun elevation:40.902706( complementary angle of zenith) 
Sun azimuth: 161.336042 
The images before and after C-correction were showed in Fig.2. 
furthermore, the spectral profiles along a random line across 
mountain area in the before and after C-correction ASTER 
image were also given in Fig.3.(two random lines shared the 
same location and distribution). 
With the sun information listed above, combining the aspect and 
slope calculated from DEM (section 3.1),the C-correction was 
applied to ASTER data with 9 bands processed in section 3.1. 
Figure 2 ASTER images of pre-C corrected (a) and post-C corrected (b) (bands 132 for RGB) 
100 200 _ 300 
Line 
400 
band I bic. i band2 
I ’ 
I 
100 200 . 300 400 
Line 
3000 
9) 
I 2000 
> 
1000 
bandl — band: band2 
c>) 
Figure 3 Spectral profiles along a random line across mountain area in the pre-C (a) and post-C (b)corrected ASTER image (two 
random lines shared the same location and distribution). 
3.3 Linear spectral mixture model 
When using LSMM, the spectra signals of a pixel are expressed 
linear combination of finite number of endmembers weighted 
by their abundances. According to the restriction on abundances, 
a number of approaches have been developed to analyze 
LSMM(Ichoku & Kamieli, 1996), such as unconstrained 
method, augmented matrix method, sum-to-one constrained 
method and full constrained method(Xin Miao et al,2006), 
where the augmented matrix method and the sum-to one 
constrained method confine the sums of endmember 
abundances to be one or close to one (Smith et al., 1990), and 
he fully constrained method further requires the endmember 
abundances to be positive (Brown et al., 1999, 2000; 
GarciaHaro et al., 1996; Settle&Drake, 1993;Shimabukuro & 
Smith, 1991), here we chose the fully constrained LSMM, and 
the basic algorithms for a pixel were as follows: 
n 
Li A. ^ ' fkiRkX + £ iA subject to 
k = 1 
n 
X f ki = * and f ki ~ 0 (4) 
k = 1