Commuting flow patterns across a triad of Chinese megacities: Evidence from Poisson gravity modeling using mobile device data

Understanding commuting flow patterns is crucial for effective urban planning in rapidly growing Chinese megacities. Utilizing mobile device data, point-of-interest (POI) data, road network data, house price data, and 1 × 1 km grids, this study systematically investigates the relationship between commuting flows and land use, transportation accessibility, and house price differences. A Poisson pseudo-maximum likelihood (PPML) gravity model is used to fix zero-value and overdispersion issues, and spatial heterogeneity is analyzed by origin- and destination-specific models. Three typical megacities in China, Beijing, Shanghai, and Shenzhen, are selected as case studies. The city-level regression results indicate that network distances, number of commercial POIs, house price differences, and distance to the city center are the main determinants of commuting flow distribution. The effects of industrial and public service land uses show significant differences across cities. Compared to Beijing and Shanghai, Shenzhen's industrial land is more mixed with residential and commercial land. Distance to the city center shows significant positive and negative effects at the origin and destination, respectively, revealing these megacities' morphological polycentricity and functional monocentricity. Specifically, Shenzhen's city center has a weaker influence, reflecting a more balanced functional structure. The results of local models highlight the spatial heterogeneity of the attractiveness of the city center and the distance decay effect. Residents between 10 and 30 km from the city center are most attracted by the center, and the distance decay effect is weaker in high-density areas. These findings emphasize the impact of industrial structure on land use patterns, the undeveloped functional polycentricity in Chinese megacities, and the limitations of high-density development in reducing commuting distances.