Araujo, Luciana Spinelli 
  
3 DATA ACQUISITION AND METHODS 
The TM Landsat, path / row 232/058, acquired on January 17, 1996, was used in this study. The digital processing of 
this TM image (bands 3, 4, and 5) was performed using SPRING software (INPE, 1999). Initially, the TM image was 
geometrically corrected using the topographic chart ‘Maloca do Sucuba', MI-53, scale 1:100,000 of DSG (Diretoria de 
Serviço Geográfico). 
The TM digital numbers were converted to reflectance values according to Markham and Barker (1986) before 
generating the synthetic SAVI image (Huete, 1988). This synthetic image has the purpose to minimize the influence of 
soil in the spectral characterization of the canopy cover, and is expressed by the following equation: 
SAVI = [(TM4 - TM3) / (TM4 + TM3 + IJ'd»-*L) 
where, 
TM4 = Reflectance value of band 4 (near infrared) of TM sensor; 
TM3 = Reflectance value of band 3 (red) of TM sensor; and 
L = constant. 
. The L value is a function of the level of surface coverage in the different facies, varying from 0 (close canopy cover) to 
1 (open canopy cover). To characterize the different facies, it was generated SAVI images using L values of 0.25, 0.5, 
and 0.75. Following this procedure, the integration of spectrum textural of synthetic image (vegetation index values) 
with the biomass values was performed. The vegetation index values extracted from the synthetic images have the same 
geographic location of the stands where the biomass values were measured on the field. Since the analysis of the 
relationship between satellite data and field measurement was done, it was performed a supervised classification 
utilizing ICM (Iterated Conditional Modes) algorithm of ENVI software, including a series of sample tests to evaluate 
the resulting map including vegetation cover types associated with class intervals of biomass values. The results found 
in the classification process were evaluated using Kappa statistics analysis (Landis and Kock, 1977). 
In a summary description of field work, besides an analysis of landscape, it was performed a forest inventory, acquiring 
DBH and height (H) measurements, in a 250 m x 10 m transects for primary forest (9 transects) and 100 m x 10 m 
transects in areas of secondary succession (10 transects), as detailed by Araujo (1999). From these DBH and height 
values, utilizing the specific allometric equations (Brown et al., 1989; Uhl et al., 1988), the biomass values were 
calculated. In the savanna areas (32 transects), a clear cut and weighting all individuals of arboreal and/or shrub for 
biomass estimation were done in a 50 m x 5 m sample units. The contribution of herbaceous stratum for the biomass 
estimation was also obtained by clear cut in 5 plots of 1 m? each, for each one of the sample units. This herbaceous 
materials was weighted and dried in a estufa, in order to determine the dry weight. 
4 RESULTS 
Visually, the synthetic images generated by SAVI model, distinguish the forested and non-forested areas. It was 
observed no significant difference among the SAVI images generated using three different L values, which showed a 
certain similarity in the correlation of this vegetation index with biomass, either for forest and savanna. From this 
analysis, the value 0.5 for the constant L was adopted to generate the synthetic SAVI image, comprising the different 
densities of vegetation cover in this transition zone. In this image, it is noted low SAVI values (0.13) for park savanna 
and grass/shrub savanna areas, due to: low foliage cover (with higher exposition of soil), the condition of leaf material, 
with higher percent of non-photosynthetic capacity, and consequently, low biomass (mean values varying from 7.4 and 
4.5 ton/ha, respectively). The areas in forest regeneration process, which canopy cover is more homogeneous, with 
pioneer species in a plenty development stage, i.e., photosynthetically more active, present value of 0.47 of SAVI (mean 
values of biomass — 43.8 ton/ha). From a determined growth stage, the SAVI values tend to decrease again. This can be 
observed when the values for primary forest are analyzed, with SAVI = 0.36, a little lower than of the secondary 
succession, in spite of having higher biomass (130.6 ton/ha) and have a structural characteristic more complex, 
composed by several stratum and a higher diversity of species. As observed by Bernardes (1998) in forested areas, after 
certain level, there is a decreasing of SAVI value with increasing biomass. 
  
78 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.