Surface soil moisture (SM) retrieval over agricultural areas from polarimetric synthetic aperture radar (PolSAR) has long been restricted by vegetation attenuation, simplified polarimetric scattering modelling, and limited SAR measurements. This study proposes a modified polarimetric decomposition framework to retrieve SM from multi-incidence and multitemporal PolSAR observations. The framework is constructed by combining the X-Bragg model, the extended double Fresnel scattering model and the generalised volume scattering model (GVSM). Compared with traditional decomposition models, the proposed framework considers the depolarisation of dihedral scattering and the diverse vegetation contribution. Under the assumption that SM is invariant for the PolSAR observations at two different incidence angles and that vegetation scattering does not change between two consecutive measurements, analytical parameter solutions, including the dielectric constant of soil and crop stem, can be obtained by solving multivariable nonlinear equations. The proposed framework is applied to the time series of L-band uninhabited aerial vehicle synthetic aperture radar data acquired during the Soil Moisture Active Passive Validation Experiment in 2012. In this study, we assess retrieval performance by comparing the inversion results with in-situ measurements over bean, canola, corn, soybean, wheat and winter wheat areas and comparing the different performance of SM retrieval between the GVSM and Yamaguchi volume scattering models. Given that SM estimation is inherently influenced by crop phenology and empirical parameters which are introduced in the scattering models, we also investigate the influence of surface depolarisation angle and co-pol phase difference on SM estimation. Results show that the proposed retrieval framework provides an inversion accuracy of RMSE<6.0% and a correlation of R≥0.6 with an inversion rate larger than 90%. Over wheat and winter wheat fields, a correlation of 0.8 between SM estimates and measurements is observed when the surface scattering is dominant. Specifically, stem permittivity, which is retrieved synchronously with SM also shows a linear relationship with crop biomass and plant water content over bean, corn, soybean and wheat fields. We also find that a priori knowledge of surface depolarisation angle, co-pol phase difference and adaptive volume scattering could help to improve the performance of the proposed SM retrieval framework. However, the GVSM model is still not fully adaptive because the co-pol power ratio of volume scattering is potentially influenced by ground scattering.

长期以来，通过植被衰减，简化的极化散射建模和有限的SAR测量，一直限制了通过极化合成孔径雷达（PolSAR）在农业地区获取地表土壤水分（SM）。这项研究提出了一种改进的极化分解框架，可以从多事件和多时间PolSAR观测值中检索SM。该框架是通过将X-Bragg模型，扩展的双菲涅耳散射模型和广义体积散射模型（GVSM）组合而构建的。与传统的分解模型相比，提出的框架考虑了二面散射的去极化和不同的植被贡献。假设在两个不同的入射角下，SM对PolSAR观测是不变的，并且在两次连续测量之间植被散射没有变化，则可以通过求解多元非线性方程来获得包括土壤和作物茎的介电常数在内的分析参数解。 。提出的框架适用于在2012年土壤水分主动被动验证实验中获得的L带无人飞行器合成孔径雷达数据的时间序列。在这项研究中，我们通过将反演结果与原位测量结果进行比较来评估检索性能比较了豆类，油菜籽，玉米，大豆，小麦和冬小麦地区，并比较了GVSM和Yamaguchi体积散射模型之间SM检索的不同性能。鉴于SM估算固有地受作物物候和散射模型中引入的经验参数的影响，我们还研究了表面去极化角和共极化相位差对SM估算的影响。结果表明，所提出的检索框架提供了RMSE <6.0％的反演精度，且相关性为ř ≥0.6与转化率大于90％。在小麦和冬小麦田中，当表面散射占主导地位时，SM估计值与测量值之间的相关性为0.8。特别是，与SM同步获取的茎介电常数也与豆，玉米，大豆和小麦田上的作物生物量和植物水分含量呈线性关系。我们还发现，表面去极化角，co-pol相差和自适应体积散射的先验知识可以帮助改善所提出的SM检索框架的性能。但是，GVSM模型仍然不能完全适应，因为体积散射的同极化功率比可能受到地面散射的影响。