Search

Highlights: [•] We model spatial regressions to examine the determinant factors of rail transit ridership. [•] Rail transit service coverage of 500m gives the best fit for estimating rail transit ridership levels. [•] Development density is positively associated with rail transit ridership within 750m service coverage. [•] Land use diversity does not show a significant association with rail transit ridership. [•] Station-level accessibility measures are the most important factors for promoting rail transit ridership. [Copyright &y& Elsevier]

Distinguishing commuters from non-commuters is important for transportation planning and traffic demand management. A framework to identify commuters using public transit is proposed based on a spatial-temporal clustering algorithm. The framework extracts commute trips by mining spatial-temporal travel patterns depending on whether the travelers' trip chains are complete. The commuting features in terms of travel frequency, time regularity, and spatial regularity are utilized to compare the difference in travel behavior between commuters and non-commuters. The framework is applied to trip data collected from an incentive program implemented by the Metropolitan Transport Commission of Korea. The data records spatial and temporal information on travelers' boarding and alighting points, as well as actual origins and destinations. One trip record constitutes a complete trip chain for travelers who joined the incentive program. The results show that commuters exhibit a higher travel frequency and more regular spatial-temporal travel patterns than non-commuters. It is found that not all commuters travel during morning and evening peaks; some commuters leave work after evening peak. The proposed identification framework based on the spatial-temporal clustering approach is capable of exploiting the inherent spatial-temporal travel patterns of travelers to achieve a reliable identification of commuters. [ABSTRACT FROM AUTHOR]