Forest Aboveground Biomass Estimation from Airborne L-Band SAR Data Using Machine Learning

The aim of this study is to evaluate the performance of machine learning to estimate above-ground biomass (AGB) over dense tropical forests using an L-band SAR dataset. Here, we train and validate three machine learning algorithms, namely Random Forest, Artificial Neural Network, and eXtreme Gradient Boosting (XGBoost) using the airborne polarimetric SAR data acquired during the AfriSAR UAVSAR campaign. From the evaluation of model performance, it is observed that these machine learning models were to retrieve AGB values with reasonable accuracies. The RF performed better estimation with $R^{2}=0.97$ and RMSE=28.52 Mg/ha for training sites, however its computational cost was high compared to other models.