ul 2004 
(4) 
jjomass. 
opolitan 
jS were 
ermined 
Ve used 
| model. 
nce. 
(5) 
(6) 
(7) 
pixel 
ctively. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
Next, the tree height was estimated from Equation (8) and the 
stereoscopic image made from the ASTER image. The ASTER 
has the stereoscopic band (backward look). The elevation values 
made from the ASTER were called by DSM (Digital Surface 
Model). Thus, 
  
  
  
Tree height ASTER-DSM DEM (8) 
2500 
2000 
1500 X ee - m 
  
  
Pixel Value 
S 
gl 
= 
c 
  
500 MASA fl Nn 
Pre 
0 10000 20000 30000 40000 
  
  
  
  
  
  
Distance [meters] 
Figure 4. DSM and DEM 
Figure 4 shows the cross sections of ASTER-DSM and DEM. 
We added pixel values of each band, NDVI, and land cover 
ratio each pixel in a digitized vegetation map. Figure 6 shows 
digitized vegetation map. NDVI was converted to 0 to 255, 
because NDVI include negative numbers. We carried out the 
regression analysis for the digital numbers each band in the 
satellite images vs. the timber volume, the whole volume, and 
the biomass. Figure 5 shows relationship between digital 
.numbers each band and timber volumes. The correlation 
coefficients increased by reducing the effects of haze and shade. 
Table 2 shows determination coefficients between the forest 
volumes and estimated tree heights. Table 3 shows the 
determination coefficients between estimated biomass and each 
band, NDVI, and vegetation area summed up from land cover 
ratio in each pixel. Vegetation Area in Table 3 is the vegetation 
area summed up by the land cover ratio in each pixel. Similarly, 
we calculated the determination coefficients for timber volumes 
and whole volumes. Finally, the biomass showed the highest 
determination coefficient. The difference of the determination 
coefficient each season was recognized. Moreover, vegetation 
area summed up by mixed pixel decomposition was effective 
for estimating forest volume. 
  
Timber volume| Whole volume Biomass 
ASTER 0.695 0.681 0.714 
  
  
  
  
  
  
  
Table 2. Determination coefficients of tree height and volume 
I 
174 77 : 
T7 77 7 
7 .765 7 
40000 4 
0000 4 
209000 4 
Timber Volume 
10000 4 
  
Hy om Bayt 
mA 
m u 
JE 
9 1000 2000 3000 * 
Sum of Landsat/TM3 
  
  
Figure 5. Relationship between digital numbers and timber volumes 
  
Figure 6. Digitized vegetation map 
3.5 Comprehending Geological Formation 
We comprehended geological formation. Slope angles and 
directions was calculated from DEM. 
3.6 Categorizing Vegetation 
Next, we categorized water area, urban area, coniferous trees, 
evergreen broad-leaved trees, and deciduous broad-leaved trees 
from Landsat/TM images. Water area, urban area, and forest 
area were classified with the maximum likelihood method. 
Moreover, in forest area, we categorized coniferous trees and 
broad-leaved trees. We added TM2 and TMS on the winter 
season images. Higher values of this image showed broad- 
leaved trees and lower values of that showed coniferous trees. 
In the broad-leaved area, we also categorized the evergreen 
broad-leaved trees and the deciduous broad-leaved trees. 
Similarly, we added TM2 on the summer season image and 
TMS on the spring season image. Higher values of this image 
showed deciduous broad-leaved trees and lower values showed 
evergreen broad-leaved trees. Figure 7 shows land cover 
classification and vegetation types. 
  
Table 3. Determination coefficients of Biomass and bands 
853