基于MTVDI与DDI二元回归模型对毛乌素沙地腹部土壤表层水分的研究

采用温度植被干旱指数法(MTVDI)与荒漠化指数法(DDI)，利用2016年4月、9月的Landsat数据对毛乌素沙地腹部的土壤水分进行反演，并与实测的土壤水分进行对比检验，将所反演的土壤含水量图划分为4个等级，基于此分析了2个时期毛乌素沙地腹部的旱情土壤水分分布变化。结果显示：(1)4月份MTVDI指数与0~10 cm、10~20 cm、20~30 cm土层土壤含水量的R²值分别为0.656、0.646、0.637，整体高于9月份R²值0.457、0.436、0.431，MTVDI能够较好地反映毛乌素沙地腹部土壤表层水分，且精度较高；(2)荒漠化指数DDI与MTVDI结合建立二元线性回归模型监测区域土层0~10 cm深度含水量，平均相对误差为10.95 %；(3)4月份，研究区0~10 cm表层土壤含水量5%~10%区域占总面积的53.72%以上，达到了6 256 km²，含水量偏低，需要加强当地水资源管理。

Study on the surface soil water in Mu Us Desert using binary regression models of MTVDI and DDI

By using the Temperature Vegetation Drought Index (MTVDI) and Desertification Index (DDI) and two sets of Landsat data of Mu Us Desert collected from April and September of 2016, we modeled surface soil water status and compared with the measured soil moisture values. The simulated soil moisture content diagram was divided into four categories based on the analysis of surface soils water conditions in the two periods. The results showed that: (1) In April, the R² values of the MTVDI index values and the measured soil moisture contents from 0 to 10 cm, 10 to 20 cm and 20 to 30 cm were 0.656, 0.646 and 0.637, respectively, which were higher than the value of R² of 0.457, 0.436 and 0.431 respectively in September. (2) The combined modeling of DDI and MTVDI was a better binary linear regression model to monitor the water content of 0~10 cm in the soil, with an average relative error of 10.95%. (3) In April, low water content of 5%~10% of the soil moisture in topsoil was observed in about 6 256 km² of land, which accounted for over 53.72% of the total area. Therefore, local water resource management should be enhanced.