e Ie 
el. 
iC 
1€ 
Batistella, Mateus 
LULC classifications were derived from manipulation of the geometrically corrected LANDSAT TM image. Visual 
analysis was conducted to achieve a general concept of the study area, including vegetation, land cover patterns, and 
possible data limitations, such as cloud cover and noise. Image enhancement techniques were applied to improve the 
visual interpretability by increasing the varied apparent distinction between features. A portion of the scene including 
both settlements was selected for LULC classifications. The image was clustered in 50 classes using ISODATA 
algorithms. An edited signature file derived from statistical separability analysis between these classes was used to 
classify the entire area. 
To carry out the comparative analysis between fishbone and topographically constrained designs of colonization, two 
subsets of equal form and dimension were selected, each one representing a portion of the studied municipalities with 
similar age of occupation. As the exact cartographic limits for the settlements were not available we focused on two 
rectangular areas including the majority of Machadinho d'Oeste and Vale do Anar (Figure 2). 
  
25 0 25 50 km 
Figure 2. Landsat TM scene showing the orthogonal road network (fishbone”), including Vale do Anari, and the 
different settlement design of Machadinho d'Oeste in Rondónia, Brazilian Amazon. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 151 
a ee i d 
A 
[i 
5 
4