Abstract:Land surface temperature (LST) is an important indicator of surface energy exchange and ground characteristics, the remote sensing technology is becoming a convenient tool for LST retrieval on regional or global scale,and the research of LST retrieval based on thermal infrared remote sensing images with different sampling algorithms and different data is endless, while the results of Landsat data retrieval are particularly prominent. In this study, the land surface temperature be retrieval from Landsat 8 image with the split window algorithm. We discussed the different LST retrieval results which based on empirical values and MODIS thermal infrared data, and the thermal infrared radiance temperature of Beijing is compared, and the surface temperature classification is calculated, the trend of surface temperature distribution in Beijing was analyzed. The results show that the retrieval temperature of Landsat 8 data based on the split window algorithm is more close to the actual temperature, moreover the precision is superior to MODIS products. The spatial distribution pattern of surface temperature in Beijing is restricted by the structure and reflectivity of ground objects, the distribution of high-temperature area mainly distributed in the middle - east, and the distribution of middle-or-low temperature was basically consistent with that of woodland and water distribution.