Figure 1 - Area of study (TM Band 5 for 07/15/94 is shown).
Brighter areas correspond to deforested areas. Dark grey areas are forest.
bone" pattern, with small properties (around 100 ha; 1 ha = 104
m2) scattered along a network of roads; some areas of bigger
cattle farms can also be found in the state, some of them with
more than 10,000 ha. Part of the deforested area is abandoned in
many regions of the State; Alves & Skole (1996) estimated that
abandonment within an area of important recent occupation in
Rondônia varied between 22% and 48% of the total deforested
land between 1986 and 1992, while approximately 42% of
abandoned areas existing in 1986 still remained in place after 6
years, to form older stages of secondary vegetation.
This paper presents a technique for characterizing changes in
land cover and land use based on the use of multitemporal
Landsat Thematic Mapper (TM) imagery and a segmentation
algorithm for image classification. Preliminary results of the
14
processing of two TM scenes covering an area of important
deforestation in Rondônia are discussed. The aims of this work
are twofold: firstly, to develop a technique appropriate for
detecting changes in areas of the typical Rondônia "fish-bone"
occupation, with a majority of small properties and some larger
cattle-raising developments; secondly, to perform an analysis of
land use dynamics in the region, with special emphasis on the
characterization of areas of secondary vegetation.
AREA OF STUDY
The area of study corresponds to TM scene WRS-TM 231/067
(figure 1). It has an area of approximately 34,000 km? and is
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
delimi
W; 10
federa
State
the S
locate
highw
Follov
corres
Submc
covers
4 to 6-
The m
some
cacao
the 19
Satell
This s
dates
consid
and 1
acquir
The in
used i:
works:
the fol
Image
The t
descri!
segme
Six La
were |
referer
Positic
region
scene
scene
The «
al. (19
same
algorit
segmei
that is
unsupe
