Geographic Information System (GIS) database is built by ArcGIS software based on observation data in Lanzhou city. GIS data set was analyzed, including testing data distribution, the trend analysis of spatial data, data outlier analysis. According to the analysis results, the appropriate spatial analysis model is chosen to simulate and predict thermal field. Cross validation method is used to verify simulation accuracy and validity of model. Simulation map of urban thermal in is output by optimal model and parameters eventually. Results show that urban thermal field presents regular distribution in summer. The spatial distribution characteristics of thermal field have directly relations with land use type, urban energy consumption, population density, building volume rate, and thermodynamic properties of underlying surface.