Abstract: Quantitative analysis of the relationship between urban thermal environment and urban morphological characteristics can provide a basis for urban territorial space planning and improvement of urban ecological environment. Local Climate Zones (LCZ) can well reflect the difference in local climate in cities. Therefore, the quantitative relationship between local climate and urban morphology can be constructed based on LCZ to study the urban thermal environment. In this paper, the Landsat-8 remote sensing data is used to invert the surface temperature, and the deep learning method is used to classify the local climate zone based on the Sentinel-2 data. The results show large differences in the surface temperature between different LCZ categories, and the urban morphology had a greater impact on the surface temperature. Among them: (1) LCZ10 (heavy industry) had the highest temperature, LCZG (water body) had the lowest temperature, and the temperature difference of each category in summer was greater than that in winter; (2) The surface temperature of dense buildings (LCZ1-3) was higher than open buildings (LCZ4-6), and the surface temperature of low-rise buildings was higher than high-rise buildings; (3) In urban morphology, the water proportion (WSF), building height (BH) and building proportion (BSF) had a greater impact on the surface temperature, of which the water proportion (WSF) and building height (BH) were significantly negatively correlated with land surface temperature, and building proportion (BSF) was positively correlated with land surface temperature.