Detection of logging in Amazonian transition forests using spectral mixture models

Techniques to detect the area affected by logging in the Amazon basin have yet to be tested on transitional forest environments, which contribute significantly to the total logging in the region. Logging in transitional forests is selective and leaves small clearing in the forests where timber is temporarily stored. These areas, called log landings, can be detected automatically in soil fraction images, generated through linear mixture modelling. Based on a harvesting radius from these log landings, it is possible to estimate the area affected by logging. This method was tested in Amazonian transition forest, using Landsat TM and ETM satellite sensor data from the years 1992, 1996 and 1999. Additionally, a methodology to record areas of old and repeated logging is described.