Soil moisture retrievals under crops using compact polarimetric surface scattering and vegetation radiative transfer models

Abstract

Accurate soil moisture estimation from Synthetic Aperture Radar (SAR) observations under crop growth con­ditions is challenging because of the complexity of extracting the soil contributions from the total backscatter. Compact Polarimetric (CP) SAR with reduced design requirements represents a viable alternative to fully polarimetric SAR but remains underutilized. The study introduces two soil moisture retrieval algorithms using the canopy Water Cloud Model (WCM) and vegetation Single Scattering Radiative Transfer (SSRT) applied to Cband CP data. Both models are calibrated with the CP-Advanced Integral Equation Model (CP-AIEM). The proposed algorithms, i.e., CP-AIEM-WCM and CP-AIEM-SSRT, are applied to RADARSAT Constellation Mission (RCM) CP data acquired over crop fields of AAFC-Lennoxville, Estrie (2022 and 2023) and in the Monteregie (2024), in Quebec (Canada). Soil moisture retrievals from soybean, corn and wheat fields perform better for lower (≤30◦) compared to higher incidence angles (>30◦) in terms of root-mean-square error (RMSE) (0.06–0.08 vs 0.09–0.10 m3/m3) and inversion rate (IR) (70–89 vs 30–60%). Among study crops, corn fields yield estimates with higher accuracy compared to soybean and wheat, with the lowest RMSE (0.06 m3/m3) under low incidence angle observations. Site-specific cross-validation was performed by splitting training and testing datasets to evaluate transferability of the algorithms. We observed that both algorithms perform well with similar RMSE at all incidence angles and crop type cases, with slight reduction (5–20%) in IR under cross-validation, thereby supporting the hypothesis of spatial transferability.