协同多传感器SAR数据的农田土壤水分反演研究

Synthetic Aperture Radar (SAR) has been extensively studied for soil moisture retrieval owing to its high sensitivity to soil dielectric properties. The research focused on the soil moisture retrieval over agricultural area using the synergistic multi-sensor SAR data, including retrieval model construction, model solving and comprehensive evaluation system. Against the ill-posed problem for soil moisture retrieval using SAR data, the multi-sensor SAR observations were developed and employed for soil moisture retrieval over agriculture areas. Firstly, against the influence of surface roughness for soil moisture retrieval, the synergistic soil moisture retrieval model was developed and structured on the basis of the radar backscatter theory and multi-sensor SAR observations. Therefore, the underdetermined solution for soil moisture retrieval was transformed into the well-posed solution, and the influence of surface roughness was eliminated. Then, against the shortcomings of look-up table method, the iterative method based on the least square theory was introduced to solve the equations. The soil moisture and surface roughness can be derived with effective SAR observations as inputs. In addition, the field measurements, observation station data, model retrieval rate, soil texture information and precipitation information are applied to evaluate and validate the retrieval model qualitatively and quantitatively, and thus to construct the comprehensive evaluation system for soil moisture retrieval model.

合成孔径雷达凭借其对土壤水分响应敏感的特性，在土壤水分反演中具有重要的研究意义。项目围绕多传感器SAR数据协同反演农田土壤水分展开模型构建、模型求解及综合评价体系研究。面向SAR反演土壤水分存在的“病态”求解问题，从多传感器SAR联合观测的角度，对农田区域开展土壤水分定量反演研究。首先，针对表面粗糙度对土壤水分反演的影响，基于雷达散射理论构建协同多传感器SAR观测的联合约束模型，将土壤水分反演面临的欠定求解问题转化为适定方程组求解。其次，针对查表法存在的计算复杂度高以及可能的多解性和收敛性问题，提出基于最小二乘原理的迭代法进行模型求解，在满足有效观测的前提下实现土壤水分和表面粗糙度的联合解算，从而有效克服表面粗糙度的影响。此外，通过实测土壤水分、站点观测数据、模型反演率、土壤质地信息及降水信息从定量和定性两个角度开展模型的可靠性和实用性评定，构建土壤水分反演模型综合评价体系。

项目面向农田土壤水分动态监测开展了多维度SAR协同反演模型构建与模型求解关键技术研究，突破了SAR反演土壤水分存在的病态求解问题，实现了不同物候条件下农田土壤水分的可靠监测。影响雷达后向散射系数的土壤表面参数主要包括粗糙度和土壤含水量，其中表面粗糙度对雷达观测信息具有直接的贡献，会显著降低雷达回波信号对土壤水分的响应敏感度，因此表面粗糙度是影响SAR反演土壤水分的重要误差因子。针对该问题，项目不局限于从剔除表面粗糙度影响的角度展开研究，而是通过协同多维度SAR 观测信息联合构建土壤水分反演模型，从土壤水分和表面粗糙度联合求解的角度实现了土壤水分的可靠反演，求解过程有效克服了表面粗糙度的影响。在研究过程中引入了测量学理论，通过增加有效观测量的角度对未知参数进行求解，将土壤水分反演中面临的欠定方程求解问题转化为适定方程组求解，从而实现了对土壤水分和表面粗糙度的协同反演。即通过多角度、多极化、多频率SAR观测联合构建观测方程组，采用代价函数最小化准则和最小二乘迭代法实现了观测方程组求解，综合了多维SAR观测信息对土壤表面特性的描述，通过联合约束求解获取了全局土壤表面参数信息。项目协同利用多传感器、多极化、多时相SAR数据，从数学角度解决了土壤水分反演存在的“病态”求解问题，充分顾及了不同物候阶段农田作物覆盖的差异性，突破了多维度SAR联合反演土壤水分模型构建与模型求解技术，在河北农田试验区开展了长时序应用分析与测试，获取了试验区可靠的土壤水分信息，整体土壤水分反演精度优于0.05cm3/cm3，实现了不同物候条件下农田土壤墒情的精确动态监测，为我国粮食安全和农业强国建设等国家重大战略实施提供了基础支撑。