This paper categorizes the entire country into five zones based on Chinese vegetation types: tropical broad-leaved rainforest, temperate coniferous and broad-leaved mixed forest, temperate desert, temperate grassland, and plateau alpine vegetation. In each zone,utilizing MODIS data with a 1km spatial resolution, the relationship between environmental indicators and climate was established by using the Genetic Algorithm (GA) and optimized Back Propagation Neural Network (BPNN) algorithms, which enables the modeling of the national near-surface temperature and its inversion. Results reveal that:(1) compared to the Multiple Linear Regression (MLR) model, the GA-BPNN model exhibits superior temperature estimation, with a coefficient of determination (R²) ranging from 0.90 to 0.94 and a Root Mean Square Error (RMSE) between 0.44-0.71 ℃.(2) In various spatial and temporal distribution patterns, the fitting accuracy of the GA-BPNN model significantly surpasses that of the MLR model in regions characterized by abundant vegetation (e.g., tropical broadleaf rainforest area, temperate coniferous broadleaf mixed forest area) and intricate topography (e.g., plateau alpine vegetation area).(3) Regarding the spatial distribution of air temperatures, the maps derived from the GA-BPNN model provide clear spatiotemporal details. (4) In terms of the spatial distribution of temperature, the temperature distribution map obtained by the GA-BPNN model presents obvious information on spatial and temporal details, and the temperature as a whole shows a pattern of high in the Southeast and Northwest arid zones, low in the Northeast and Tibetan Plateau regions, high in the plains, and low in the mountainous regions, which is in line with the actual situation.