Desarrollo de un modelo predictivo basado en machine learning para la biomasa aérea del bosque seco tropical en el sector suroccidental del área de compensación ambiental del proyecto hidroeléctrico el Quimbo

Abstract

The tropical dry forest (bs-T) is among the most vulnerable ecosystems globally due to extensive deforestation and fragmentation. This underscores its critical importance for biodiversity conservation and its essential role in carbon storage. In this context, a predictive model was developed using machine learning algorithms to accurately estimate aboveground biomass in the tropical dry forest near the El Quimbo hydroelectric project. The objective of this study was to identify the model that most accurately predicted the aboveground biomass stock in this region. To achieve this, Sentinel-2 satellite images were utilized alongside various topographic variables, such as the Digital Elevation Model (DEM), aspect, and slope, in addition to field data for aboveground biomass prediction. Three predictive models were implemented: XGBoost, Random Forest (RF), and Artificial Neural Networks (ANN). Among these, the XGBoost model demonstrated the best performance, with a coefficient of determination (R²) of 0.72, a mean absolute error (MAE) of 17.38 t ha⁻¹, a root mean square error (RMSE) of 27.61 t ha⁻¹, and a Huber Loss of 22.62 t ha⁻¹.