Mapping deforestation and secondary succession forest in the Brazilian Amazon using SPOT-4 VEGETATION data
João
Brito
Carreiras, Instituto Superior de Agronomia, Department of Forestry, Lisboa, jmbcarreiras@isa.utl.pt
(Presenting)
José
Cardoso
Pereira, Instituto Superior de Agronomia, Department of Forestry, Lisboa, jmcpereira@isa.utl.pt
Manuel
Lameiras
Campagnolo, Instituto Superior de Agronomia, Department of Mathematics, Lisboa, mlc@math.isa.utl.pt
Yosio
Edemir
Shimabukuro, Instituto Nacional de Pesquisas Espaciais, Department of Remote Sensing, yosio@ltid.inpe.br
The main purpose of this work is to assess the potential of multitemporal images from the SPOT-4 VEGETATION sensor, 1-km spatial resolution, to obtain a land cover map of the Brazilian Amazon, discriminating primary tropical forest, cerrado (savanna), deforestation, natural/artificial waterbodies, and secondary succession forest. For our study, we used a temporal series of 12 monthly composite images of 2000, derived from the SPOT-4 VEGETATION sensor. A set of 19 Landsat TM/ETM+ scenes was used to obtain areas of known land cover class. The 10-fold cross validation procedure indicates the k-nearest neighbors (k-nn), with k=7, as the best classification algorithm to solve this problem, with an overall accuracy of 0.923. Accuracy assessment of some land cover classes was done by comparison with county statistics of deforestation, primary tropical forest and natural/artificial waterbodies, derived from Landsat TM/ETM+ imagery. The correlation coefficient between county-level proportion of deforestation/secondary succession forest, primary tropical forest, and natural/artificial waterbodies, obtained from coarse resolution (this study) and high resolution (Landsat TM/ETM+ imagery) is high (0.88, 0.95, and 0.97, respectively). The estimated total area of deforestation and secondary succession forest was 834,469 km2 and 57,592 km2, respectively. Deforestation occurred primarily in areas of primary tropical forest (45.5%), in regions of cerrado (32.9%), in transition forest (19.1%), and other vegetation types (2.5%). Deforestation in areas of cerrado partially explains the difference between this study and INPE’s estimation (587,727 km2, up to year 2000).
This study concluded that SPOT-4 VEGETATION data are very adequate for estimating deforestation and secondary succession forest in tropical regions. The land cover map produced may be useful in models of regional carbon flux.
Submetido por Joao Manuel de Brito Carreiras em 25-MAR-2004
Tema Científico do LBA: LC (Mudanças dos Usos da Terra e da Vegetação)