Monitoring tropical forest degradation using time series analysis of Landsat and Sentinel-2 data

Tropical forest loss is expected to be contribute 5 to 15% of anthropogenic carbon emissions in the coming century. The wide range of expected emissions is indicative of the large uncertainties that exist in the terrestrial carbon cycle. Total carbon loss from forest conversion consists of loss from deforestation plus loss from degradation. There have been significant improvements in the ability to relate plot-level estimates of carbon stocks to remote sensing-derived calculations of deforestation to estimate total carbon emissions from forest loss. These approaches, however, have been limited in their ability to assess the magnitude, extent, and overall impact of forest degradation. The causes of tropical degradation include selective logging, fuel wood collection, fires, and the development of forest plantations. This study demonstrates a newly developed methodology for detecting subtle changes in forest structure and condition using time series analysis of Landsat and Sentinel-2 data. The research shows how the ability to detect small changes in forest biomass, in addition to changes in forest composition, can be improved by incorporating historical context and multi-sensor data fusion. Results are demonstrated from two climatically unique tropical forests in Thailand and Brazil.