Thai Dry-Evergreen Forest’s Biomass Estimation using Machine Learning Models

Abstract

Above ground biomass (AGB) is the key measurement for carbon credit. To quantify AGB over large forests, it is essential to develop a method that can handle the growing forest area. In this study, we investigate the possibility of estimating AGB using satellite data and machine-learning models. Several machine learning models, linear, non-linear, and ensemble methods, are evaluated. Random forest algorithm achieved the best model performance. On validation data, the random forest model can predict AGB with 24.5 Mg per area in terms of RMSE. The results demonstrate that satellite data from Sentinel-1, Sentinel-2, and MODIS have the potential to predict AGB in Thai dry-evergreen forests.