Development of L-band fully polarimetric SAR algorithm for forest biomass retrieval using 7SD and random forest regression

Abstract

Forest biomass is essential for carbon budgeting, biodiversity health, and climate change research. This study developed a novel L-band fully polarimetric SAR method for estimating aboveground biomass (AGB) in a forested region in India. Biomass was estimated using fully polarimetric ALOS-2/PALSAR-2 data sets and field AGB measurements. To estimate forest AGB, the algorithm employs a seven-component scattering power decomposition (7SD) and a random forest regression (RFR) machine-learning approach. By implementing the 7SD model, the seven scattering powers were extracted from ALOS-2/PALSAR-2 inverted for aboveground biomass. The field data were used to validate the biomass estimated by the model. The 7SD model estimated AGB in Shivamogga was consistent with field measured biomass. The root mean squared error (RMSE) and relative RMSE with respect to mean AGB were 21.94 Mg/ha and 19.46, respectively, which are within acceptable ranges. The 7SD model was also used for cross-validation at Tundi Forest, where the relative RMSE with respect to the mean AGB was 22.9. The scattering powers generated from L-band fully polarimetric data can be useful in tropical forest AGB estimation.