04 
and 
sus 
lata 
> of 
tric 
ote 
  
und 
and 
ated 
ig a 
os et 
"on, 
that 
).77) 
, the 
and 
was 
and 
| HH 
fitted 
io R? 
and 
erent 
p the 
t al, 
| this 
elow 
. 350 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXX V, Part B7. Istanbul 2004 
  
The model in Equation 1 is linear and thus not affected by 
saturation of backscatter response with increasing biomass 
levels that has been reported as a general problem of forest 
monitoring by radar (LeToan et al. 1992; Imhoff, 1995; Santos 
et al. 2002). 
Despite some difference in the response curves of the 
logarithmic and exponential functions when relating biomass 
and radar backscatter values, the P-band signal alone would 
saturate around 100 ton/ha, as reported by Santos et al. (2003); 
since here we also count with interferometric height (Equation 
1), saturation of the radar response signal as a problem for 
tropical forest biomass estimation does not occur any more. 
In Figure 2 the variance of the forest biomass estimate is 
graphically displayed in function of the observed data range of 
interferometric height (h ;,,) and backscatter radar (6° yy). The 
standard errors range between ca. 20-40 ton/ha. The biomass 
estimates obtained by applying equation 1 are superior to 
common approaches, because they are not only based on 
backscatter signals, and because the variance of the estimate 
does not increase exponentially with high quantities of 
biomass. 
  
  
  
  
  
  
  
  
Figure 2. Variance of biomass estimation as fuction of height 
interferometry and basckscatter. 
The segmentation of the DHM yields a division of the study area 
in landscape patches by polygons, that would reflect individual 
forest stands or management units of farms. As described above, 
the mean interferometric height and mean P-band radar 
backscatter are calculated for all polygons. A simple binary 
classification is performed to distinguish forest cover and 
agricultural/pasture areas. All polygons with mean interferometric 
height > 2.6 m are classified as secondary succession or primary 
forest; all polygons with hj, < 2.6 m are taken to correspond to 
nonforest landuse, i.e. agriculture, pastures, etc. 
Nevertheless, the biomass map was realized based on equation 1, 
which describes only SAR data signals from the primary and 
secondary vegetation. Polygons with nonforest classes are 
masked out. The procedure yields a map that display the spatial 
disctribution of standibng alive above ground biomass in forest 
cover at the test-site in the Tapajós region (Figure 3). 
  
  
  
  
  
  
  
  
  
5 ton/ha 350 ton/ha 
Figure 3. Section of SAR interferometric height image and 
biomass distribution map of land cover classes in the Tapajós 
region. In the above figure, the dark areas correspond to 
deforested areas (mainly pastures), the brightest grey levels are 
primary forest, intermediate variations in tone reflect forest 
successional stages. 
  
4. CONCLUSIONS 
This work shows a new approach to mapping and inventory 
tropical forest biomass, which has become possible by 
improved data quality related to the SAR data (considering both 
backscatter signals and interferometric data). It largely enhances 
biomass model precision due to the integration of 
interferometric height measurements. It was also observed that 
an adequate calibration of DSM and DEM is important for 
height inferences, this is particularly true for initial and 
intermediate successional stages, because the heights to be 
measured are lower than for advanced stages. The segmentation 
approach based on hierarchical region growth (representation by 
pyramidal levels) applied on the DHM image demonstrated 
potential for analysis and improvement of thematical 
stratification by regions. The general problem of excessive 
image segmentation, frequently observed in radar imagery, does 
not occur when using the segmentation approach adopted here. 
This study is a contribution to the Governamental Program 
" . f - 4 QU 
Science and Technology for the Management of Ecosystems 
from the Ministry for Science and Technology (MCT), which is 
looking for alternatives (remote sensing data) and for the 
improvement of technological knowledge which might help in 
the inventory and monitoring of forest resources in Brazil. 
REFERENCES 
Balzter, H. 2001. Forest mapping and monitoring with 
interferometric synthetic aperture radar (INSAR). Progess in 
Physical Geography, 25(2):159-177. 
Chambers, J.Q.; Santos, J.; Ribeiro, R.J.: Higuchi, N., 2001, 
Tree damage, allometric relationships, and above-ground net 
primary production in central Amazon forest. Forest Ecology 
and Management, 152: 73-84. 
1135