Objectives Analyzing urban form patterns and their differentiation from a multi-dimensional and large-scale perspective is the basis for studies of urban patterns, processes and effects, but the subject remains a great challenge.

Methods We construct a multi⁃dimensional characterization system of urban form by integrating urban physical space and social space perspectives, which include indices on land use, building distribution, transport network, population distribution and urban function distribution. We further propose a pattern identification method based on principal component analysis and clustering algorithm.

Results Taking 2 475 natural cities in China, the United States, and Europe as examples, we identify six major urban morphological patterns which include high-density (HD), high-aggregation with low-density (HALD), low-aggregation with low-density (LALD), multi-center with high-concentration (MCHC), multi-center with high-mixed (MCHM) and multi-center with low-mixed (MCLM). Chinese cities are mostly HD cities and MCHC cities, and the distribution of patterns has a significant difference between northwest and southeast areas. In the United States, there are mostly HALD cities, LALD cities and MCLM cities, while showing difference between east and west areas. The cities in Europe are mostly MCHM cities with a few other types of cities along the western and southern coastlines.

Conclusions Multi⁃dimensional characterization can classify cities into categories in a comprehensive, refined and reasonable way, and urban form patterns can provide support for cross-regional urban comparison studies, urban pattern and process effects, and urban simulation.