Assessing the potential of polarimetric decomposition of Sentinel-1 SAR for the estimation of mangrove forest biomass

Abstract

Mangrove forests provide ecosystem services that can support the welfare of local communities. Therefore, promoting their conservation not only protects their environmental functions but also optimizes their economic potential. Remote sensing approaches, particularly active systems such as synthetic aperture radar (SAR), have emerged as a valuable tool for monitoring forest ecosystems. These systems can capture Earth's surface features regardless of atmospheric conditions. However, the backscatter approach becomes unreliable when the AGBD reaches a certain point. This leads to incomplete information being obtained as the reflected signal becomes too strong and overloads the receiver. Therefore, this study explores the potential of polarimetric decomposition as an option to traditional backscatter approaches. Decomposed polarimetric parameters from Sentinel-1 were used for biomass estimation in mangrove forests in West Kalimantan Province. Specifically, in Kubu Raya and North Kaong districts (Indonesia) for the year 2020. The decomposed polarimetric parameters of Entropy, Anisotropy, and Alpha Angle obtained from the H/A/α decomposition integrated with backscattering parameters were used as dependent variables, which were varied following the parameter usage scenarios (both individual and grouped). Meanwhile, GEDI data were used to train the prediction model instead of observational data. The predictive ability of the model using only SAR-derived explanatory variables resulted in an RMSE of about 45 Mg/ha and an R-squared of about 0.2.