Investigating the association between land use and urban vibrancy is crucial for sustainable urban development. To that end, according to the four conditions proposed by Jacobs (1962), suitable units should be identified before investigating this association. In this study, we consider spatial-temporal human activity patterns and propose a novel community-detection approach integrating graph neural network (GNN) and gated recurrent unit (GRU) to identify suitable geographic units. Then, the impact of land-use-relevance variables on community vibrancy is measured by the geographically and temporally weighted regression (GTWR) model. Using the cell phone signaling data from Beijing, we initially identify 66 communities, primarily comprising intra-travel (the origin and destination are in the same community) and a small proportion of inter-community travel, showing distinct vibrancy patterns and different land use characteristics. Then, we develop a GTWR model to reveal the associations between land use and community vibrancy, revealing that land-use-based variables, such as mixed land use and densities of points of interest (POIs), have spatial-temporal-heterogeneous impacts on community vibrancy, implying the heterogeneity among the development focus of communities, such as population density and functional zone structure. Consequently, we propose heterogeneous development strategies for communities, which could offer valuable suggestions for enhancing urban vibrancy.
Read full abstract