Monitoring Native Forests in Southern Brazil

摘要

The monitoring of tropical rain forest in southeastern Brazil, along the Atlantic coast is made using a set of ten satellite images taken from December 2002 through February 2003 by the HRV sensor aboard SPOT 5 spacecraft. The methodology is based on the analysis of images classified by the Gaussian Maximum Likelihood (GML) classifier. Classification results were validated by ground truth surveyed at field trips. Images were geocoded using control points extracted from topographic maps at scale 1:50,000. Sixteen land-cover classes were implemented into the classification process. The image data was subjected to a low-pass filtering procedure, as a pre-processing step, prior to the GML classification. Final results include a set of 45 maps of the region, and areas for five forest classes. In a longer perspective, the aim is to monitor the conditions across the forested areas concerning additional deforestation and/or re-growth areas, what calls for a new study in the near future, using data collected during the period 2006-2007.