Abstract: Land surface temperature （LST） is a very important environmental factor to impact the energy and water exchange between the atmosphere and the ecosystem. The spatial distribution of LST in small and medium-scale region has a great significance on the stability of ecosystems and agricultural production and construction of the arid basins. In this study，the Shiyang River Basin in Gansu Province is chosen as the study area. It is a typically arid inland basin and its ecological environment is very fragile. The LST is retrieved from MODIS data by split-window algorithm，and the normalized difference vegetation index （NDVI） is calculated by its first band and second band. At the same time，the land use/cover is acquired from TM data by human-computer interactive interpretation. Combined with land use/cover and DEM data，the spatial distribution of LST is identified and the relation between NDVI and LST is analyzed in different land use types. The result shows as follows：（1）The LST in the north and the east is higher than that in   the south and the west of the study area，respectively. It performed the trend that the LST reduced gradually from the northeast to the southwest. （2）There is a negative correlation between LST and altitude，and the correlation coefficient is -0.928. Due to the differences in heat capacities，different land use/cover types，LST also has differences. The quantitative difference is 16.42 K between the highest LST of unused land and the lowest LST of forest land. （3）For single type of land use/cover，LST and [NDVI] have a significant negative correlation between cropland，forest land，grassland，and construction land. For the whole configuration of land use/cover in the study area，the relation between LST and NDVI displays ‘trapeziform’. The feature space of NDVI-LST contains a lot of geo-information，which reflects the relationship between the basin surface temperature and vegetation cover. And from the feature space of NDVI-LST，we can see that the green vegetation has a great importance on regulating the temperature of regional climate.

Key words: MODIS, split-window algorithm, LST, LUCC