基于自适应极化分解技术的灌区麦田土壤墒情反演方法

自微波遥感进入土壤墒情研究领域以来,植被覆盖地表的校正一直是微波遥感反演土壤墒情的热点。针对植被层对雷达微波信号的散射容易造成墒情计算误差等问题,以河南焦作广利灌区为研究区,以灌区内的主要作物冬小麦为研究对象,在离散植被的一阶物理散射模型基础上,使用Sentinel-1A SAR雷达提取后向散射系数,通过自适应极化分解技术即在极小的临近空间内利用最小二乘法求解土壤散射和植被多重散射的最优解,再利用Landsat8数据作为辅助数据提取改进的归一化植被水分指数,采用支持向量机的方法反演土壤墒情。结果表明:基于自适应的后向散射系数与麦田土壤墒情的关系在VV极化模式20 cm深的土壤条件下,相关系数为0.827 8;20 cm深度的土壤墒情与冬小麦的相关性较高,相关系数达到0.819 0;自适应极化分解技术反演土壤墒情均方根误差、平均相对误差分别为1.490 7、0.002 2,水云模型反演土壤墒情均方根误差、平均相对误差分别为1.958 5、-0.242 2,自适应极化分解技术反演效果较为理想。该研究可为小麦覆盖下的灌区土壤墒情的评估提供参考。

Adaptive polarized decomposition technique-based method for inversion of soil moisture of wheat field in irrigation district

Since the technique of microwave remote sensing entering the field of the research on soil moisture, the correction of the vegetation covered surface is a hotspot in the aspect of inverting the soil moisture with microwave remote sensing all along. Aiming at the problems of the calculation error of soil moisture caused by the scattering from of vegetation layer on radar microwave signal, etc., the backscattering coefficient is extracted by Sentinel-1 A SAR radar on the basis of the first-order physical scattering model for discrete vegetation by taking Guangli Irrigation District in Jiaozuo of Henan Province as the study area and the main crop——winter wheat therein as the study target, and then the optimal solutions of both the soil scattering and the multiple-scattering of vegetation are derived through the adaptive polarized decomposition technique, i. e. the derivation is made with the least square method in the minimum near-space, moreover, the improved normalized difference water index of vegetation is extracted by taking the data from the Landsat-8 as the supplementary data and then the soil moisture is inverted with the method of support vector machine. The result shows that based on the relationship between the adaptive backscattering coefficient and the soil moisture of wheat field under the soil depth of 20 cm of the VV polarization mode, the correlation coefficient is 0. 827 8 and the correlation between the moisture of the soil with the depth of 20 cm is higher, while the root-mean-square error and the mean relative error inverted with the adaptive polarized decomposition technique are 1. 490 7 and 0. 002 2 respectively, furthermore, the rootmean-square error and the mean relative error inverted with water-cloud model are 1. 958 5 and-0. 242 2 respectively, thus the inversion effect of the adaptive polarized decomposition technique is more ideal. The study can provide references for the assessment on the moisture of the wheat covered soil in irrigation district.