In the temperaturevegetationdrought index (TVDI)， the impacts of the land surface temperature and the vegetation index on the remote drought monitoring are simultaneously taken into account, and also both merits are combined. Therefore, the effect of vegetation coverage fraction is effectively reduced and the accuracy in remote sensing of drought is obviously improved. Land surface temperature is retrieved based on the Planck radiation function，in which the surface emissivity is used to be a gray body emission. Highest and lowest land temperatures corresponding to each vegetation index are extracted to make up NDVI-Ts space. In this paper, the remote sensing data of the Earth Observing Satellite (EOS) MODIS are used to analyze and reveal the particular morphological characteristics of NDVI-Ts space over Guizhou complex mountains. Soil surface dry conditions on July 25, 2006 and August 19, 2007 are inversed and simultaneously verified with the soil moisture information from local weather stations. The results show that TVDI is significantly related to soil moisture. Because of high temporal and spectral resolution as well as moderate spatial resolution of EOS/MODIS remote sensing data, the method is suitable for the detection and early warning of soil drought in largescale and complex terrains.