基于双指数联合模型的土壤含水量反演——以河北省为例

Soil water content inversion based on double-index combined model:Taking Hebei Province as an example

针对温度植被干旱指数特征空间的非线性现象,研究以Modis数据为数据源,以河北省为研究区,通过引入表观热惯量,提出双指数联合(DICIM)的土壤含水量反演模型用以改进TVDI指数特征空间的非线性问题。研究分别采用TVDI和DICIM模型对6月上中旬的土壤含水量进行反演,对比低植被区土壤含水量反演的空间差异性,并通过误差统计验证模型的反演能力。结果表明:在低植被区,DICIM模型反演的土壤含水量比TVDI反演的土壤含水量特征明显,反演值更加接近于实测值。经误差统计分析发现基于DICIM模型反演的10cm深度的土壤含水量值相比于TVDI指数反演的土壤含水量值平均绝对误差低0.26%~0.50%,均方根误差低0.28%~0.73%,相对均方根误差低0.73%~5.54%,平均相对误差低1.31%~3.27%,且基于DICIM模型的反演值与10cm深度土壤含水量实测值的相关系数R值都在0.65左右。可见,提出的DICIM模型综合了ATI和TVDI模型的优势,提高了TVDI指数的反演能力。

In view of the nonlinear problem of temperature vegetation drought index,the MODIS data in the Hebei Province are studied to improve the accuracy of nonlinear fitting of dry and wet edges in traditional TVDI feature space by introducing the apparent thermal inertia model and a double-index combined soil water content inversion model.The TVDI and DICIM indices are used to invert the soil water content in early and mid-June,respectively.The spatial differences of soil water content inversion in low vegetation areas are compared,while the model′s inversion ability is verified by error statistics.The result shows that in the low vegetation area,the feature of DICIM inversion is more obvious than TVDI,and the inversion value is closer to the measured value.Besides,according to error statistics,the mean absolute error of DICIM is 0.26%to0.50%lower than that of TVDI for the inversion value of soil water content at a depth of 10 cm.The root mean square error is 0.28%to 0.73%lower than that of TVDI,the relative root mean square error is 0.73%to 5.54%lower than that of TVDI,and the average relative error is 1.31%to 3.27%lower than that of TVDI,respectively.The correlation coefficient Rvalues based on the DICIM inversion value and the measured soil moisture content at a depth of 10 cm is approximately 0.65.It can be seen that the proposed DICIM model combines the advantages of the ATI and TVDI models and improves the inversion ability of the traditional TVDI model.