Wu et al. (2026) Comparative analysis of SAR-based soil moisture inversion methods for crop-covered under cloudy, rainy, and irrigation conditions

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-03-14
• Authors: Shangrong Wu, Hanxiao Meng, Yiqing Zhu, Hu Zhong, Hong Cao, Han Gao, Yingbin Deng, Guipeng Chen, Qian Song
• DOI: 10.1016/j.ejrh.2026.103337

Research Groups

• State Key Laboratory of Earth Surface Processes and Disaster Risk Reduction, Innovation Research Center of Satellite Application (IRCSA), Faculty of Geographical Science, Beijing Normal University, Beijing, China
• State Key Laboratory of Efficient Utilization of Arable Land in China, The Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing, China
• College of Oceanography and Space Informatics, China University of Petroleum (East China), Qingdao, China
• Guangzhou Institute of Geography, Guangdong Academy of Sciences, Guangzhou, China
• Institute of Agricultural Economics and Information, Jiangxi Academy of Agricultural Sciences, Nanchang, China

Short Summary

This study developed a scenario-adaptive framework for soil moisture inversion in crop-covered areas, integrating Radarsat-2 SAR and HJ-2A/B optical data with Random Forest (RF) and the Water-Cloud Model (WCM). It found that direct optical-SAR fusion via RF achieved the highest accuracy (R² = 0.90) under clear conditions, while the VWC-coupled WCM was optimal (R² = 0.61) for cloudy, rainy, or irrigation scenarios.

Objective

• To develop a robust SAR-based soil moisture inversion model adaptable to complex meteorological (cloudy, rainy) and irrigation scenarios in rapeseed-covered areas of Jiangxi Province, China.
• To construct a scenario-adaptive soil moisture inversion strategy by systematically evaluating direct and indirect methods, integrating multi-source remote sensing data (SAR and optical) with machine learning (Random Forest) and semi-empirical models (Water-Cloud Model).

Study Configuration

• Spatial Scale: Regional scale, focusing on Gao'an City, Jiangxi Province, China, a major rapeseed-producing area.
• Temporal Scale: Rapeseed growth period from mid-October to early May of the following year, with four-phase remote sensing acquisitions and ground observation campaigns conducted between November 2023 and March 2024.

Methodology and Data

• Models used: Random Forest (RF) algorithm for direct inversion and canopy parameter retrieval, Water-Cloud Model (WCM) for indirect soil moisture estimation, Particle Swarm Optimization (PSO) and Trust Region Reflective (TRF) algorithm for WCM parameter calibration.
• Data sources:
  • Satellite: Radarsat-2 (fully polarimetric C-band SAR, HH, HV, VH, VV polarization, 8 m spatial resolution), HJ-2A/B (multispectral optical, 16 m spatial resolution).
  • Observation: In-situ measurements from 32 ground sampling points across 4 campaigns, including soil moisture content (at 5 cm depth using TDR 350 sensor), Leaf Area Index (LAI using LAI-2200C), and Vegetation Water Content (VWC calculated from fresh/dry weight).
  • Reanalysis/Other: Daily precipitation data from a meteorological station in Gao'an City.

Main Results

• Under clear, non-irrigation conditions, the direct soil moisture inversion using optical-SAR data fusion with the RF algorithm achieved the highest accuracy (training R² = 0.90, nRMSE = 0.11; validation R² = 0.90, nRMSE = 0.10).
• Under cloudy, rainy, or irrigation conditions, the indirect soil moisture inversion method based on SAR-derived Vegetation Water Content (VWC) coupled with the WCM demonstrated the best performance among indirect methods, with a validation R² of 0.61 and nRMSE of 0.21.
• VH cross-polarization SAR features and optical features like Normalized Pigment Chlorophyll Index (NPCI) and Plant Pigment Ratio (PPR) showed high importance in soil moisture and VWC inversion, attributed to their sensitivity to vegetation-soil interaction and water stress.
• Direct inversion methods generally yielded higher accuracy but showed limitations in capturing extreme soil moisture values and had lower scenario applicability (CI Score).
• Indirect inversion methods, particularly SAR-based VWC coupling, demonstrated better ability to capture extreme values and exhibited stronger applicability across various complex meteorological and irrigation scenarios (highest CI Score).
• Spatial analysis showed that optical and optical-SAR synergistic methods exhibited lower spatial variability (smoother distributions), while SAR-based indirect inversion had higher spatial variability, reflecting its sensitivity to subtle moisture gradients and localized waterlogging.

Contributions

• Establishes a novel scenario-adaptive soil moisture inversion framework for crop-covered areas, specifically rapeseed, addressing the challenges of complex meteorological and agricultural management conditions.
• Fills a technical gap in precise soil moisture monitoring for southern rapeseed regions by integrating multi-source remote sensing data (SAR and optical) with machine learning (Random Forest) and semi-empirical models (Water-Cloud Model) in a "multi-source, multi-model, multi-scenario" collaborative approach.
• Provides practical references and technical support for agricultural management, such as guiding precision irrigation and providing waterlogging warnings, in similar agricultural zones.
• Systematically compares the performance, spatial characteristics, extreme value capture capability, and scene applicability of various direct and indirect soil moisture inversion methods.

Funding

• National Key Research and Development Program of China (2021YFD1600503)
• National Natural Science Foundation of China (42271374)
• Youth Innovation Program of the Chinese Academy of Agricultural Sciences (Y2023QC18)

Citation

@article{Wu2026Comparative,
  author = {Wu, Shangrong and Meng, Hanxiao and Zhu, Yiqing and Zhong, Hu and Cao, Hong and Gao, Han and Deng, Yingbin and Chen, Guipeng and Song, Qian},
  title = {Comparative analysis of SAR-based soil moisture inversion methods for crop-covered under cloudy, rainy, and irrigation conditions},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103337},
  url = {https://doi.org/10.1016/j.ejrh.2026.103337}
}