Unraveling the determinants of intra-city commuting flows with a spatially weighted interaction model: Nanjing, China as a case study

Abstract

Modeling origin-destination (OD) commuting flows is crucial for effective land use and transportation planning. While spatial interaction models grounded in the gravity law have been extensively utilized to analyze OD flows, the issue of spatial non-stationarity has often been overlooked. Taking Nanjing, China, as a case study, we generate commuting flows between 1 × 1 km grids based on mobile phone signaling data and investigate the spatial non-stationarity in the effects of land use, transportation accessibility, house prices, and travel distance on commuting flows using a Spatially Weighted Interaction Model (SWIM). Our findings indicate that SWIM significantly enhances fitting effectiveness compared to the global spatial interaction model, thus providing a valuable tool for understanding human mobility flows. A comparison of local and global coefficients reveals that the impacts of residential and commercial land use on commuting flows tend to be concentrated at smaller spatial scales, with notable spatial variations in the effects of housing prices. The spatial distribution of local coefficients implies that commuting flows between subcenters and the main center are primarily influenced by housing prices, school locations, and transit accessibility. Additionally, the influence of shopping services and land use diversity on commuting flows is considerably weaker in subcenters compared to the main center. Further, the polycentric structure limits the increase in commuting distances. These findings provide important insights into the spatial non-stationarity in commuting network and underscore the need for tailored urban planning strategies accounting for such heterogeneity.