The invention discloses an urban local climate region classification method based on multi-source data, and the method comprises the following steps: 1, obtaining a Gaofen-1 image and urban 3D building data in a research region range, and carrying out preprocessing; 2, extracting parameters for local climate partitioning, including the building height, building surface score, normalized vegetationindex, vegetation coverage, permeable surface score, water surface score and impermeable surface score, establishing regular grids of different scales, and extracting LCZ partitioning parameters of corresponding spatial scales based on the regular grids of different scales; 3, establishing a random forest classification model; 4, based on the established random forest classification model, carrying out local climate region classification of different scales; and 5, selecting an optimal local climate partition scale through visual interpretation and quantitative precision evaluation. Accordingto the method, the advantages of two local climate partitioning methods are effectively combined, and the precision and high efficiency of local climate partitioning are improved.