Andrade, Nilo Sergio de Olive 
  
STUDY OF THE VEGETAL COVERING AND LAND USE DYNAMICS IN THE REGION OF JI- 
PARANÄ/RO USING CLASSIFICATION TECHNICS 
Nilo Sergio de Oliveira Andrade *, Luciana Spinelli Araujo ** Izaya Numata ** Mario Valerio Filho ”” 
* Instituto de Estudos Avancgados (IEAv) — Centro Técnico Aeroespacial (CTA) 
Praga Marechal Eduardo Gomes, 50 — 12228-9000 — Sao José dos Campos, SP 
andrade@ieav.cta.br 
** INPE — Instituto Nacional de Pesquisas Espaciais 
Caixa Postal 515, 12201-097 Sao José dos Campos, SP, Brasil 
{lucian.izaya} @ltid.inpe.br 
** IP&D - Instituto de Pesquisa e Desenvolvimento - UNIVAP 
Caixa Postal 8088, 12244-000 Sáo José dos Campos, SP, Brasil 
mvalerio@univap.br 
  
KEY WORDS: Remote Sensing, land use/cover, visual analysis, image classification. 
ABSTRACT 
The Brazilian Amazon has been object of anthropic activities (agronomic and cattle raising) that have been changing 
areas of forests into areas destined to the agriculture and pasture, contributing to the alteration of the environment. The 
evaluation of the vegetal covering and the land use is therefore indispensable for a rational planning that will overcome 
problems of uncontrolled development and deterioration of the environmental quality. However, high costs and 
difficulties of obtaining data in a short period characterize the conventional techniques used for that end. In this context, 
the analysis process and interpretation of images is an effective technique for the evaluation of that dynamics. The aim 
of this paper is to present the results obtained by the visual analysis and interpretation of Landsat Thematic Mapper 
(TM) imagery (Path/Row = 231/067) for years 1995 and 1997 using remote sensing techniques for mapping the classes 
of vegetal covering and land use in the study area, as well as the quantification of the area deforested in that period. It is 
also showed the results obtained with supervised and unsupervised classifiers. The major change occurred in the area 
was the mature (primary) and secondary forest that became bare soil and pasture, representing 69.79 Km” (6.35%) of 
the study area. The second more representative alteration was the bare soil that became pasture, representing 33.47 Km? 
or 3.05% of the total area. 
1. INTRODUCTION 
Since the 70’s, optical remotely sensed data has been used for monitoring of natural areas. These data have been used 
for land cover/land change mapping, allowing the identification of several cover classes like crop fields, bare soil, 
pasture, secondary forest and mature forest. 
This work presents a simplified analysis of landscape changes in an area of high deforestation rates in the state of 
Rondónia. The aim of these analyses is to contribute to a better understanding of the results that can be obtained using 
image classification techniques - visual, supervised and unsupervised. 
The study was split into two parts. In the first part, the images were visually analyzed by using interpretation keys, and 
allowed the determination of six classes: mature forest, secondary forest, regrowth, pasture, bare soil e burned areas, as 
well as the quantification of the area and percentage of class modification in the 1995-1997 period. 
During the second part, a region of the images were analyzed using the supervised and non-supervised classification 
techniques and the results were then compared with the results obtained during the visual analysis. The values obtained 
with this comparison will be presented in this work. 
2. STUDY SITE 
The study area is located at Rondónia State, western Brazilian Amazon, with the boundary coordinates: from 09? 41' to 
09° 54' of latitude South and from 62° 16' to 62? 41' of longitude west, representing a total area of 1.099,30 Km”. 
The region presents a slightly undulated terrain, with an average annual rainfall of about 2,200mm, mean annual 
temperature of 23,6? C (H. Schimitz, unpublished data), and a dry season from late April to late August. After the 70's 
decade the settlement of small farmers began along the BR-364 highway. As a consequence forest has been clear cut to 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 
63