Improved selective logging detection with Landsat images in tropical regions

Abstract

Logging is a major form of forest degradation in the tropical regions like Brazilian Amazon. It alters the tropical habitat environments and results in release of carbons as well. The traditional way of logging is through forest clearing, which converts forest to other land uses such as agriculture or rangeland. Recently a new form of forest degradation is selective logging, removing only those good quality tree species. This form of deforestation does not result in land use conversion, but degradation. Logging by means of clear-cutting can be easily detected and monitored from satellite images such as those from Landsat sensors. However, detection and monitoring selective logging is difficult with satellite images because the process only removes a small number of trees per area, resulting in subtle disturbances but substantial removal of biomass. Therefore, traditional classification technique is unable to detect and monitor this type of disturbances effectively. In order to detect selective logging, and to better understand carbon sequestrations, a continuous field ought to be used that can quantify the degree of disturbances due to selective logging, instead of using binary classification techniques. In this paper, we used signal-unmixing techniques in a spectral vegetation index domain as a continuous field measure of forest density, with which selective logging is mapped and quantified. The spectral index used is the MSAVI further modified to enhance its sensitivity to subtle forest degradations in the tropical environments in Brazilian Amazon as well as in Southeast Asia.