The improper management and the low availability of nutrients of soils in "Zona da Mata" in Minas Gerais State, Brazil, have led to low productivity of natural pasture, emergence of bare soils and soil losses by erosion. The objective of this work was to evaluate the capacity of ASTER sensor images to identify different levels of degradation in pasture lands. The studied area includes part of Viçosa, Teixeiras, and São Miguel do Anta municipalities, forming a total area around 3,314 ha. Due to natural characteristics of the pasture in this region, four levels of degradation were used: light, moderate, strong, and very strong. The class that showed the highest error in the classification was the very strong, degraded with 53.91% of the classified pixels, not distinguishable from the other classes. The moderate degradation class showed the best classification. From the total evaluated area, approximately 70% corresponded to pasture, 56.46% of which was classified in the strong degradation level, 28.73% to 'Mata/Capoeira' and only 1.54% to coffee plantations. The results permit to conclude that the use of sensor ASTER images was satisfactory to separate degradation levels of pasture lands in the studied area.
degraded pastures; remote sensing; classification
