Estimating above-ground biomass of trees outside forests using multi-frequency SAR data in the semi-arid regional landscape of southern India

Trees outside forests are vital for sustainable resource management and play a crucial role in the sequestration of carbon. This study attempted to estimate the above ground biomass (AGB) of trees outside forests utilizing the datasets of ALOS PALSAR-2 (L-band) and Sentinel-1 (C-band), with a focus on a semi-arid region in Sri Sathya Sai district of Andhra Pradesh, India. Here, we employed random forest (RF) algorithm integrating AGB observed over a large-scale ecological plot and remote sensing technology for generating 3 models (Model-1 (M1), Model-2 (M2), and Model-3 (M3)). Backscattering coefficients (VV and VH) and H-α dual pol decomposition parameter, anisotropy (A) from Sentinel-1 were applied for M1, and the backscattering coefficients (HV and HH) and the band ratio (HV/HH) from ALOS PALSAR-2 data were utilized in M2. M3 is the ensemble of parameters from both sensors. Validating the three models found that the R2 values fall between 0.44 and 0.64, the RMSE between 1.89 t/ha and 2.49 t/ha, and the MAE between 1.56 t/ha and 1.99 t/ha. The results of the study suggest that both Sentinel-1 and ALOS PALSAR-2 data can be employed for AGB estimation in semi-arid regions incorporating machine learning algorithms like RF. The results of the study are crucial for sustainable land management and reducing uncertainty using data from large-area ecological plot and multi-frequency synthetic aperture radar (SAR).