基于理论干湿边与改进TVDI的麦田土壤水分估算研究

针对旱情监测及农田灌溉中传统的基于地表温度-植被指数特征空间的温度植被干旱指数(Temperaturevegetation drought index,TVDI)构建方法无法准确反映真实地表的水热能量交换,给土壤含水率估算带来极大不确定性的问题,根据地表能量平衡方程,并引入改进植被覆盖度参数,构建一种理论干湿边端点选取方法及基于地表温度-改进植被覆盖度特征空间的TVDI模型,结合两期MODIS遥感影像数据(DOY088和DOY112)及地面观测数据,对陕西杨凌区的麦田土壤含水率进行估算。结果表明,由理论干湿边计算得到的TVDI与实测土壤含水率相关系数在-0.700左右,均方根误差不大于0.060 cm3/cm3。DOY088和DOY112的土壤含水率估算结果均与土壤含水率实测值有较好的拟合关系,尤其是DOY088的反演结果更接近于实际地表干湿状况,相关系数为-0.715,均方根误差为0.029 cm3/cm3,DOY112的散点分布比DOY088分散。该方法可以避免传统特征空间在干湿边估算中必须包含裸土、部分植被覆盖以及全植被覆盖地表覆盖类型的限制,从而实现真实土壤水分的遥感反演和实际地表干湿状况的监测。

Soil Moisture Estimation of Wheat Field Based on Theoretical Dry-Wet Edge and Improved TVDI

The temperature-vegetation drought index (TVDI) based on the surface temperature-vegetation index feature space has very important scientific and practical significance in drought monitoring and farmland irrigation. But traditional methods cannot accurately reflect the real surface water-heat energy exchange and make the soil moisture estimation have great uncertainty. Based on the equation of surface energy balance and the introduction of improved vegetation coverage parameter, a theoretical dry-wet edge endpoint selection method and a TVDI model based on the surface temperature improved vegetation coverage feature space were constructed, which broadened the application range of TVDI in drought monitoring and soil moisture estimation by improving vegetation coverage parameters to a certain extent to avoid restrictions on the types of surface coverage. MODIS remote sensing image data and ground observations were used to estimate the soil moisture of wheat field in Yangling District, Shaanxi Province. The results showed that the correlation coefficient between the TVDI calculated from the theoretical wet edge and the measured soil moisture was about -0.700, and the root mean square error was not more than 0.060cm3/cm3. In addition, the estimated soil moisture values of DOY088 and DOY112 both had good fitting relationship with the measured soil moisture values, especially the inversion results of DOY088 were closer to the actual surface conditions with correlation coefficient of -0.715 and root mean square error of 0.029cm3/cm3. Meanwhile, the scatter distribution of DOY112 was much more dispersed than that of DOY088. Therefore, this method can avoid the limitation of the surface vegetation coverage that must include bare soil, partial vegetation and full vegetation coverage in the estimation of traditional feature space dry and wet edges, realize remote sensing inversion of real soil moisture and monitor the actual wet and dry conditions of the ground.