Analytical Optimization of Polarimetric Change Detectors for Mapping Open and Grassland Flooding Using Sentinel-1. Case of Rupununi, Guyana

Abstract

Quantifying flood inundation beneath vegetation is critical for understanding flood dynamics in complex wetland environments, where conventional remote sensing approaches often struggle to distinguish between open-water flooding and vegetation-covered inundation. In this study, we propose four novel detectors for the characterization of both open flooding and flooding beneath vegetation using radar-based signal analysis. The methods are developed from two fundamental signal models: the additive model of optimization of power difference and the multiplicative model of optimization of power ratio. Our objective is to investigate additive signal changes can inform or be combined with multiplicative ones. Based on these models, we developed four detectors through different combinations: First, monomial unitary combination, which represents the product of additive and multiplicative models; Second, monomial combination, a parameterized product-based model; Third, linear combination, based on additive integration; Fourth, mixed combination, a data-driven approach. The proposed detectors demonstrated strong potential for quantifying two types of flooding open-water flooding and grassland flooding beneath vegetation with the data-driven detector achieving the best overall performance.