Araujo, Luciana Spinelli 
  
RELATIONSHIP BETWEEN SAVI AND BIOMASS DATA OF FOREST AND SAVANNA 
CONTACT ZONE IN THE BRAZILIAN AMAZONIA 
Luciana Spinelli ARAUJO, Joäo Roberto dos SANTOS, Yosio Edemir SHIMABUKURO 
National Institute for Space Research — INPE, Brazil 
Remote Sensing Division 
flucian: jroberto; yosio@ltid.inpe.br } 
Working Group VII/3 
KEY WORDS: Biomass, Tropical Rain Forest, Savanna, Vegetation Index, Amazonia. 
ABSTRACT 
The Brazilian Amazonia presented the important dynamic changes in land use/land cover in the last two decades, 
mainly at the forest and savanna transition zones, due to human activities related to selective logging, cattle raising 
farms, and agricultural settlements. The general objective of this study is to verify the potential application of Soil 
Adjusted Vegetation Index (SAVI) derived from Thematic Mapper (TM)/Landsat data to relate with biomass values of 
forest and savanna formations. The thematic classification of synthetic SAVI image by the Iterated Conditional Mode 
(ICM) algorithm was used to show the spatial distribution of vegetation cover and its biomass. The results of this study 
are an important task to support a program for monitoring Amazonia region using remote sensing data. 
1 INTRODUCTION 
The burning procedure has been used in tropical areas as an easy and economical procedure for cleaning the terrain 
surfaces for agricultural or cattle raising activities. This extensive practice, when performed in forested areas, 
contributes for the climatic unbalance in a regional and global level, loss of soil potentiality in a medium term, and also 
loss of local biodiversity. Forest fires, starting in the savanna areas (as occurred in the state of Roraima, northern Brazil) 
and advancing in the direction of open and dense tropical forest, have evidenced the need for monitoring the existing 
biomass in this transition vegetation zones. Satellite data, provided by optical and microwave sensors (Santos et al., 
1999: Araujo et al, 1999) with the optimization of the information extraction procedures, has been utilized as a tool for 
forest inventory. ! 
Inside this context, the present work has the objective to evaluate the synthetic image, generated by the SAVI (Soil 
Adjusted Vegetation Index) model using TM Landsat data, for estimating biomass of forest and savanna formations. 
Linear regression analysis was utilized to verify the relationship between SAVI and biomass variables. The spatial 
distribution of vegetation types associated with biomass values was generated. 
  
2 STUDY AREA 
The study site is located at central-north of Roraima 
State (Brazilian Amazonia), with geographical 
coordinates 2° 30" to 2° 50" N and 61° 00" to 61° 30" 
W, representing a zone of abrupt contact zone between 
forest and savanna physiognomic formations (Figure 1). 
The climate of this region is Awi type, according to 
Kóppen, presenting a tropical rain dominance and a 
well-defined dry season, with an average yearly rainfall 
of 1,800 mm. 
    
  
Figure 1. The study area showed in 
the synthetic SAVI image. 
  
  
  
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 77 
A 
3 
D OQ E VE A C UU uda 
as