Prediction of AGB Density in Skikda, Algeria: Leveraging Multisensor Data, ML Models

Aboveground Biomass Density (AGBD) estimation is of extreme significance in carbon cycle studies and climate change research. Remote sensing techniques, including Light Detection and Ranging (LiDAR), Synthetic Aperture Radar (SAR), and optical sensors, are an economical way of AGBD estimation on a large scale. This study integrates multi-sensor data, including GEDI LiDAR, Sentinel-1 SAR, and Sentinel-2 optical imagery, with machine learning (ML) regression models—Classification and Regression Trees (CART) and Gradient Tree Boosting (GTB)—to predict AGBD in Skikda Municipality, Algeria. Using Google Earth Engine (GEE), we processed and extracted key geospatial features such as vegetation indices, backscatter coefficients, and topographical attributes. The GEDI Level 4B (L4B) biomass product was used as reference data for model training and validation. Models’ performance was evaluated using Root Mean Square Error (RMSE) and the coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). GEDI L4B product was used as reference data for model validation and training. Root Mean Square Error (RMSE) and coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) were employed to evaluate models' performance. CART is strongly dependent on topographic variables (slope, elevation, aspect, and B3), while GTB balances spectral, topographic, and polarimetric data. GTB performs better than CART during training (RMSE = 9.02 Mg/ha, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.86) and validation (RMSE = 12.47 Mg/ha, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.71) with better generalization. This led to offering a useful methodology for AGBD monitoring at large scales, to aid forest management by exploiting ML algorithm.