Transfer station choice in a multimodal transit system: An empirical study

Connectivity between modes and lines constitutes a core aspect for multimodal transport system user experiences. Using a large-scale transit smart card dataset, we reveal transfer station choices in bus–rail intermodal trips frequently deviate from the “nearest-station” heuristics, motivating research to better explain real transfer behaviors and define more realistic station catchment areas. The investigation is centered on modeling transfers from bus to rail using multinomial logistic regression discrete choice models applied to 200k observations from four weeks of transit smart card observations. The results show that: (1) for trips requiring an intermodal transfer, riders select the nearest rail station in only 40% of cases; (2) commuters’ transfer station selections are jointly influenced by the trip attributes of all involved modes, with the bus ride distance contributing the most, followed by the walking distance; and (3) commuters’ sensitivity to travel distances exhibits nonlinear trends: sensitivity to bus ride distance dampens gradually as a function of trip length, while the sensitivity to walking distance exhibits a concave shape peaking at around 0.6 km. Catchment area modeling reveals that (4) location attributes (e.g., distance from the city center) play a greater role in defining the station catchment area than other attributes (e.g., land use and built environment). Specifically, the catchment area radius is larger for stations located further away from the city center or from the nearest interchange station. Moreover, (5) stations with more amenities (e.g., higher density of shops and greater accessibility) are likely to attract commuters from a larger area. This study strengthens the connections between station selection patterns and station catchment area analysis, to better guide intermodal transit design and operations, such as improving predictions of passenger flows at transfer stations.