Automated Landsat Classification of Tropical Forest Disturbances for Large Scale Identification in the Amazon

Abstract

Under-reporting of selective logging and forest disturbance has posed an issue for forest health and deforestation estimations in the tropics due to difficulty of detection by satellite data. Several studies have proposed methods for the delineation of these areas using a variety of data and models. In the interest of supporting the study of large-scale ecosystem and climate dynamics in this region for the purposes of prioritization of critical focal points in climate mitigation policy and incentive programs, I hope to test one such methodology over different regions across Amazonia. Landsat-derived Multiple Soil Adjusted Vegetation Index (MSAVI) and alternately Multiple Soil Adjusted Vegetation Index with Aerosol Resistance (MSAVIaf) have been shown to be relatively successful in obtaining green fractional percentage (fc) when incorporated into a linear mixture model. I plan to replicate this technique, followed by a multi-temporal comparison of fc to locate disturbed and degraded areas, and then expand the analysis over an extended study area to review the potential for automation to a state-wide or national scale.