Land surface temperature (LST) is a crucial parameter for global change investigation, which is significant to ecosystem and biogeochemical researches. However,studies on LST of complex terrain areas are limited. Guizhou Province is a typical karst landform with complex terrain and fragile ecosystems. Based on multi-year MODIS remote sensing data combined with meteorological data in Guizhou Province,the temporal and spatial changes of LST were studied by using correlation analysis,trend analysis,R/S analysis and geographic detector. The results showed that: (1) The MODIS LST displayed a linear correlation with near surface temperature; (2) LST showed a spatial distribution of the south-high and north-low; (3) The interannual fluctuation of LST was large with the maximum difference of 2.06 ℃·a~(-1),while the annual variation was mainly in unimodal distribution; (4) LST decreased with increasing elevation in the regains with >800 m above sea level,and exhibited a nonlinear relationship with elevation in the regions with < 800 m above sea level; (5) There were significant differences in LST among different land-use types,with a maximum difference of 1.3 ℃. In summary,the spatial distributions of LST in the study area were relatively consistent with that of latitude,which was impacted by elevation and latitude and was related to land-use types.