Implementation of a Sentinel-2 Based Exploratory Workflow for the Estimation of Above Ground Biomass

Abstract

This work presents a Sentinel-2 based exploratory workflow for the estimation of Above Ground Biomass (AGB) in a Mediterranean forest. Up-to-date and reliable mapping of AGB has been increasingly required by international commitments under the climate convention, and in the last decades, remote sensing-based studies on the topic have been widely investigated. After the generation of several vegetation and topographic features, the proposed approach consists of 4 major steps: 1) Feature selection 2) AGB prediction with k-Nearest Neighbour (kNN), Random Forest (RF), Extreme Gradient Boosting (XGB), and Artificial Neural Networks (ANN); 3) hyper-parameters fine-tuning with Bayesian Optimization; and finally, 4) model explanation with the SHapley Additive exPlanations (SHAP) package. The following results were obtained: 1) before hyper-parameters optimization, the Deep Neural Network (DNN) yielded the best performance with a Root Mean Squared Error (RMSE) of 42.30 <tex>$t/ha$</tex>; 2) after hyper-parameters fine-tuning with Bayesian Optimization, the Extreme Gradient Boosting (XGB) model yielded the best performance with a RMSE of 37.79 <tex>$t/ha$</tex>; 3) model explanation with SHAP allowed for a deeper understanding of the features impact on the model predictions. Finally, the predicted AGB throughout the study area showed an average value of 83 <tex>$t/ha$</tex>, ranging from <tex>$0\ t/ha$</tex> to 346.56 <tex>$t/ha$</tex>.