Linear spectral mixture model as a tool for monitoring deforestation and timber exploitation in the Brazilian Amazon

Abstract

The objective of this study is to show the operational capacity of a spectral mixture model using TM/Landsat data for the characterization/monitoring of the annual deforestation and the timber logging exploitation process in the Amazon. In the methodological procedure, the original TM bands were initially converted to vegetation, shade and soil fraction images, derived from the linear spectral mixture model. After the selection of fraction images, the scene segmentation was made using a region growing algorithm, and then an unsupervised classifier (per region) as applied. Afterwards, the thematic polygons were manually edited to generate the final maps. An analysis was made on the proportion of vegetation, shade and soil components, for primary forest, selective logging, regrowth, and deforestation areas, for the timeframe 1997-2001. This analysis demonstrates, through the ternary diagram, that the variations in the spatial attributes of these component fractions were caused by a land cover/land use change process. A set of images and maps, showing the temporal identification of deforested and timber logging exploitation areas is shown, as a result of the operational use of this technique. The spatial distribution of these landscape changes provides subsidies to environmental agencies for the control and enforcement of specific conservation policies referring to the Amazon forest resources.