Abstract: This study takes Xinjiang Yanqi basin as the research area,using field data and Landsat 8 OLI data to obtain soil vegetation index MSAVI and land surface temperature (Ts) to construct Ts-MSAVI feature space combined with dry and wet space equations. The temperature vegetation drought index (TVDIm) is calculated using this feature space, retrieve soil moisture in September, October, and November, analyzes the temporal and spatial distribution characteristics of soil moisture. The result shows that:① The TVDI retrieved from remote sensing images is significantly correlated with soil moisture (a=0.05). TVDIm has the highest correlation with 10~20 cm soil moisture (R=0.588) in different soil layer. ② Humidity level in Yanqi basin is featured with semi-arid(0.60.8). Spatial distribution of soil moisture denotes that the arid area is mainly distributed the southern side of the Yanqi basin, the western and northern part of the Yanqi basin is arid, and the humid area is the middle of the Yanqi basin. For this area, the lake side wetland and the small lake of Bosten Lake have the highest soil moisture, and the sandy area of the southern Bosten Lake has the lowest soil moisture.③ The wetland showes the lowest TVDIm value in October. The TVDIm value of the sandy land is the highest in September. The TVDI model has been applied in Yanqi basin to obtain good results, which can be used to accurately estimate soil moisture. The research results can provide important parameters for the ecological environment and water resources of Yanqi basin.

Key words: Yanqi basin, high-resolution remote sensing image, soil moisture, inversion model