Soil moisture is a key parameter affecting crop growth. Gaofen-3 satellite is the first C-band synthetic aperture radar produced by China, which provides full-polarization data sources for soil moisture estimation. This article evaluated the potential of estimating soil moisture via GF-3 SAR over agricultural area using different polarimetric decomposition models, namely, the modified Freeman-Durden model (MFDM), the An model and the FDM. Among them, the MFDM is the first attempt to be used for soil moisture retrieval. After removing the volume scattering, the surface and dihedral scattering component were used complementarily to estimate soil moisture. The results show the performance of each polarimetric decomposition models for soil moisture estimation depends on the crop type, crop growth stages and soil moisture conditions. Soil moisture retrievals exhibit an overall underestimation with a root mean square error of 8-11vol. %. This is mainly because of the random orientation assumption in the volume scattering model, which cannot accurately describe the variability of the crop structure. Due to the application of de-orientation process and power constraint, the MFDM shows the best performance both for corn and wheat, with inversion rates of 39%-45%.

中文翻译：

---

高分三号C波段SAR使用不同偏振分解模型反演土壤水分的评价


土壤湿度是影响作物生长的关键参数。高分三号卫星是我国首颗C波段合成孔径雷达，为土壤湿度估算提供全偏振数据源。本文使用不同的极化分解模型，即改进的 Freeman-Durden 模型 (MFDM)、An 模型和 FDM，评估了通过 GF-3 SAR 估算农业区土壤湿度的潜力。其中，MFDM是首次尝试用于土壤水分反演。去除体积散射后，补充使用表面和二面散射分量来估计土壤湿度。结果表明，每种土壤水分估计极化分解模型的性能取决于作物类型、作物生长阶段和土壤水分条件。土壤水分反演表现出整体低估，均方根误差为 8-11vol。 %。这主要是因为体积散射模型中的随机方向假设，无法准确描述作物结构的变异性。由于解取向过程和功率约束的应用，MFDM 对玉米和小麦都表现出最佳性能，反转率为 39%-45%。