The Estimation of Forest Aboveground Biomass Using Multiple Satellite Data and Different Machine Learning Models

Forests are critical in reducing carbon emissions and increasing carbon sink areas. Understanding, estimating, and mapping forest biomass is vital for the role of forests in the global carbon budget and climate. The progress in satellite technologies and remote sensing systems has emerged as an alternative method for conducting studies on the estimation of above-ground biomass (AGB), diverging from conventional terrestrial approaches. In this study, the effects of band and vegetation index values on Above Ground Biomass (AGB) estimation were investigated in Oriental beech (Fagus orientalis Lipsky.) and Uludağ fir (Abies nordmanniana subsp. Bornmülleriana mattf.) species within the buffer borders of Küre Mountains National Park in the Western Black Sea region using data from the Sentinel-1 radar and Sentinel-2 optical satellite along with Gradient Boosting and Random Forest machine learning methods. The relationships between AGB values obtained from ground sample plot data and the satellite data were examined. The best results for AGB estimation were achieved using the model that incorporated the Sentinel-1 VH backscatter value, Sentinel-2 Band 4 and 11, Enhanced Vegetation Index (EVI), Difference Vegetation Index (DVI), texture measures derived from Sentinel-2 and the Gradient Boosting method (R2=0.71, RMSE= 20.98 t/ha, MAE=17.47 t/ha).