The main rapeseed production region of Jiangxi Province, China, was selected for this study. As a typical southern cultivation zone, the area features frequent cloud-rain weather and routine irrigation, making it a representative site for investigating soil moisture inversion under complex meteorological and agricultural management scenarios. To enhance soil moisture inversion applicability in crop-covered areas, this study integrated Radarsat-2 fully polarimetric Synthetic Aperture Radar (SAR) and HJ-2A/B multispectral data by combining Random Forest (RF) and the Water-Cloud Model (WCM). Under clear, non-irrigation conditions, a direct inversion model was constructed using fused optical and SAR features via RF. Under cloudy, rainy, and irrigation conditions, an indirect approach was employed: RF first retrieved canopy parameters, Leaf Area Index (LAI) and Vegetation Water Content (VWC), which were then coupled with the WCM to estimate soil moisture from radar backscattering. The optimal strategy was identified: direct optical-SAR fusion via RF achieved the highest accuracy (R 2 = 0.90). Among indirect methods, the VWC-coupled WCM outperformed the LAI-coupled scheme, reaching a maximum R 2 of 0.61. Validated by in situ observations and meteorological data, the results confirm the model's robustness. This study establishes a tailored soil moisture SAR inversion framework, filling the technical gap in precise soil moisture monitoring for southern rapeseed regions and providing practical references for similar agricultural zones. • Proposed a scene-adaptive soil moisture inversion framework for rapeseed fields. • Integrated optical–SAR features with Random Forest for direct retrieval (R² = 0.90). • Designed indirect inversion via Water-Cloud Model with canopy parameters VWC and LAI. • VWC-coupled SAR inversion showed superior accuracy under cloudy and irrigated conditions. • Supports precision irrigation and crop monitoring in complex agricultural environments.
Wu et al. (Sat,) studied this question.