Sentinel-2-Based Mapping of Forest Aboveground Biomass in Alpine Areas Using GEDI as Reference Data: An Application in Valle d'Aosta, Italy

Accurate estimation of forest aboveground biomass (AGB) is essential for monitoring carbon stocks and supporting climate mitigation strategies, particularly in heterogeneous mountain environments where field-based inventories are costly and logistically challenging. This study presents a cloud-based machine learning workflow implemented in Google Earth Engine (GEE) to estimate and map forest AGB across the alpine region of Valle d’Aosta (northwestern Italy).

Sparse but high-quality observations from the Global Ecosystem Dynamics Investigation (GEDI) L4A product were used exclusively as reference data. A Random Forest regression model (50 trees) was trained to upscale these reference measurements using wall-to-wall environmental predictors. These predictors included a summer–autumn 2024 Sentinel-2 surface reflectance composite (incorporating Blue, Green, Red, NIR, and SWIR bands) and derived spectral indices (NDVI, EVI, MNDWI, NDBI, and BSI), alongside topographic variables (elevation and slope) from the Copernicus GLO-30 DEM. To reduce spatial mismatches and GEDI geolocation uncertainties, all datasets were resampled to a 100 m grid, and GEDI footprints with relative standard errors greater than 50% or located on slopes steeper than 30° were excluded.

The model successfully captured the spatial variability in biomass across the complex alpine landscape, achieving an R² of 0.84 and an RMSE of 36.73 Mg ha⁻¹. AGB estimation was constrained to forested areas through a high-resolution mask derived from 1 m Global Canopy Height Maps and updated with official regional Land Cover datasets. By combining predicted AGB density with forest area estimates, total forest AGB was quantified at approximately 13.37 million Mg, corresponding to an average density of 171.9 Mg ha⁻¹. Using GEDI spaceborne LiDAR as a reference for Sentinel-2 optical imagery within GEE provides a robust, scalable framework for regional biomass assessment, effectively addressing topographic challenges in mountain environments.