Spatial variations in urban public ridership derived from GPS trajectories and smart card data

Understanding urban public ridership is essential for promoting public transportation. However, limited efforts have been made to reveal the spatial variations of multi-modal public ridership (such as buses, metro systems, and taxis) and the underlying controlling factors. This study explores multi-modal public ridership and compares the similarities and differences of the associated factors. Daily bus, metro, and taxi ridership patterns are first extracted from multiple sources of big transportation data, including vehicle (bus and taxi) GPS trajectories and smart card data. Multivariate regression analysis and geographically weighted regression analysis are used to reveal the associations between these data and demographic, land use, and transportation factors. An empirical study in Shenzhen, China, suggests that employment, mixed land use, and road density have significant effects on the ridership of each mode; however, some effects vary from negative to positive across the city. The results also indicate that road density, income, and metro accessibility do not have significant effects on metro, transit or bus ridership. These findings suggest that the effects of the associated factors vary depending on the mode of travel being considered and that the city should carefully consider which factors to emphasize in formulating future transport policy. •Three mode public ridership (bus, metro, and taxi) from multi-sourced big data are explored.•Ordinary least square regression model (global) and GWR model (local) are compared.•Similarity and difference of associated factors are distinguished.•Five or six variables have different effects on different ridership.•The influences of associated factors vary from negative to positive across urban space.