基于改进作物散射模型的陕西杨凌区麦田土壤水分反演研究

本文提出一种改进作物散射模型反演麦田土壤水分,该模型根据冬小麦等低矮植被的散射特性,在原模型的基础上保留植被层直接散射部分以及植被与地表相互耦合作用的信息,同时加入裸土地表的直接散射部分,并根据经验权重将两部分信息分离开,构建出适用于冬小麦等低矮植被的后向散射模型,并结合RADARSAT-2雷达数据以及陕西杨凌农田试验区的地面实测数据,计算得到改进模型的经验参数,进而对模型进行验证分析。研究结果表明：改进作物散射模型的模拟精度相对于未改进的作物散射模型有显著的提高,R_²在HH和VV极化下都达到80%以上。为了验证改进的作物散射模型算法及土壤水分反演的有效性,本研究将改进作物散射模型与TVDI光学指数模型、简化的MIMICS模型的土壤水分反演结果进行对比分析,改进的作物散射模型反演精度比TVDI和简化的MIMICS模型要好,R_²达到84.3%,均方根误差为0.028 cm_³/cm_³,简化的MIMICS模型反演结果比TVDI要好,但是精度不高,R_²为66.9%,均方根误差为0.043 cm_³/cm_³。改进的作物散射模型对地表植被比较敏感,可以有效的将冬小麦对雷达信号散射影响和裸土层散射贡献区分开,为植被覆盖下地表土壤水分的反演创造条件,给大面积大范围的地表土壤水分反演提供强有力的技术支撑。

Soil Moisture Inversion Based on Modified Crop Scattering Model Over Wheat Fields in Yangling District, Shaanxi Province

In this paper, a modified crop scattering model is proposed to retrieve soil moisture in wheat fields. According to the scattering characteristics of low vegetation such as winter wheat, the model retains the direct scattering contribution of the vegetation layer and the coupling component between vegetation and surface ground on the basis of the original model. The direct scattering contribution of the underlying ground is considered in the modified model. The scattering effects between vegetation and surface bare soil are separated by introducing an empirical weight to construct the modified scattering model which is suitable for low vegetation such as winter wheat. Combined with RADARSAT-2 radar data and ground measurements of farmland experimental area over Yangling District, Shanxi province. The structural parameters of the modified model are calculated and the model is verified and analyzed. The results show that the simulation accuracy of the modified crop scattering model is significantly improved compared to the original crop scattering model with R2 of more than 80% in both HH and VV polarizations. In order to verify the effectiveness of the modified crop scattering model and soil moisture retrieval, this study compares the results of soil moisture retrieved by the modified crop scattering model with that of the TVDI optical model and the simplified MIMICS model. The accuracy of the modified model is better than that of TVDI and simplified MIMICS model with R2 and root mean square error (RMSE) of 84.3% and 0.028 cm3/cm3, respectively. The simplified MIMICS model produces a better inversion result than TVDI but the accuracy is not high. The R2 is 66.9% and the RMSE is 0.043 cm3/cm3. The modified crop scattering model is sensitive to surface vegetation, which can effectively distinguish the influences of winter wheat and bare soil on radar signal. It can create conditions for the inversion of surface soil moisture under vegetation-covered area and provide strong technical support for large-area and large-scale surface soil moisture retrieval.