Resumen
Understanding urban form is beneficial for planners and designers to improve the built environment. The street network, as an essential element of urban form, has received much attention from existing studies. Recently, an open dataset containing 8910 global urban street networks and 25 different form indicators has been produced, but the urban forms of cities across the globe have rarely been recognized based on analyzing such a large dataset, which was the main purpose of our study. We employed correlation analysis, principal component analysis and hierarchical clustering methods for analyzing this dataset. We also compared the spatial pattern of clustering results with those using terrain and land-cover data. Results show that: (1) Most of these indicators are highly correlated with at least another indicator, and six principal components (i.e., size, terrain-variation, regularity, long-street, circuity and altitude) were found. (2) Seven clusters (i.e., regular, long-street, large size, irregular, varied-terrain, high-circuity and high-altitude) of cities were identified; cities of the same cluster can be spatially aggregated and also distributed across different regions. (3) Most of these clusters can be interpreted using terrain and land-cover data, which indicates that the urban forms of most cities across the globe are related to geographical factors. The clustering results may be used not only to compare street networks and their urban forms at a global scale but also to understand the formation and development of an urban street network.