Above-ground biomass estimation of Indian tropical forests using X band Pol-InSAR and Random Forest

Abstract

Accurately assessing the biomass of tropical forests is a challenge for the scientific community. This study has determined the canopy height and above-ground biomass of Indian tropical forest plantation by the space-borne TanDEM-X Polarimetric Interferometric Synthetic Aperture Radar data using the physical scattering model and machine learning. The complex coherence estimated using TanDEM-X datasets of December month was found most appropriate for inversion of canopy height. The best canopy height accuracy (R2 = 0.79, RMSE = 1.66 m) was achieved by the TanDEM-X pair with vertical wavenumber (Kz) 0.18 at a 27.31° incidence angle in Random Volume over Ground model. The Random Forest model selects 12 optimal parameters from 42 variables (e.g. Canopy height, alpha angle, and coherence, etc.) for spatial prediction of AGB. Observed and predicted AGB showed R2 = 0.83 and RMSE = 27 Mg ha−1. The canopy height explained the maximum variation in AGB than other TanDEM-X variables in the ML algorithm.