Abstract: Surface temperature-vegetation index feature space (hereinafter to be referred as feature space method), constructed by using remote sensing technology, combines these 2 components' physiological and ecological functions and is widely used in regional drought monitoring and evapotranspiration estimation. However, influenced by the antecedent precipitation, the premise of the feature space method, i.e., the study area has extreme drought regions, is hard to satisfy. In addition, different spatial resolution remote sensing data have different abilities to identify the extreme moist or drought condition of the soil. All these facts increase the uncertainty of the feature space method. To explore the effects of spatial and temporal scale on the feature space method, this paper analyzed the continuous changes of feature space fitting borders after the rain using MODIS data. The surface temperature and Normalized Difference Vegetation Index (NDVI) retrieved by Landsat 5 TM data were interpolated into different resolution data, feature space parameters and Temperature-Vegetation Drought Index (TVDI) obtained by these different resolution data were studied. The results showed that the fitting dry edges were far from the theoretical ones because the influence of antecedent rainfall. The continuous changes of the fitting dry edges were in accordance with the soil moisture evolution of the study area. To improve the estimate precision of the feature space method, the value of fitting dry edge at the bare soil (where NDVI=0.1) should be assigned correctly and dynamically. The decrease of the spatial resolution of remote sensing data made the fitting dry and wet edges more far from the theoretical ones and the feature space compressed to the center. These led to some places be mistaken for more drought or more moist. By this token, any discrepancy with the premise of the feature space caused error in the drought monitoring and evapotranspiration estimation. The discrepancy should be corrected based on its mechanism and the demands of the feature space method.