Effects of catchment measurement on the associations between determinants and metro-ridesourcing integration

Constructing metro-integrated ridesourcing catchment and understanding its determinants are essential for advancing multimodal urban mobility. However, existing studies rarely utilize text-inclusive ridesourcing trip data to identify metro-ridesourcing integration, and most extract explanatory variables based on a fixed-radius catchment of metro stations. This study leverages origin-destination address textual information from ridesourcing trip data in Tianjin, China, to identify metro-integrated ridesourcing trips and applies a hierarchical clustering method to generate station-specific catchments for access to and egress from metro stations during morning and evening peak periods. Machine learning models are employed to examine the relationship between integration demand and various attributes, with model performance comparison between station-specific and fixed-radius catchments. Results show that models based on station-specific catchments outperform those using fixed-radius catchments. Key findings reveal that road network density is significantly associated with metro-ridesourcing integration, exhibiting distinct threshold effects. GDP displays a nonlinear positive relationship with integration demand. Land-use mix shows a positive correlation with integration demand, particularly during the evening peak. Ridesourcing trip distance exhibits the strongest positive association within the first 5 km. Metro station ridership is positively related to ridesourcing demand, with a higher saturation threshold for inbound compared to outbound flows. This finding offers policymakers new insights into metro-ridesourcing integration, supporting efforts to improve connection efficiency and promote multimodal transport planning.