Existing methodologies for urban rail demand estimation, such as the traditional four-step model, were primarily developed to calculate route-based mode transfer volumes for alleviating road congestion and assessing benefits. However, these approaches are complicated, time-consuming, and limited in their ability to predict boarding and alighting volumes at individual stations, as well as to validate and utilize the results. This study addresses the need for a new methodology that leverages transportation big data to provide more precise and faster station-level demand estimation.
Using trip chain analysis that combines ‘Alttul transportation card(now K-Pass)’ and regular transit card data, the actual travel routes of urban rail users in the Seoul metropolitan area were reconstructed. Based on this, station influence zones were defined and classified into six types according to factors such as population density, office space, and proximity to neighboring stations. A predictive model incorporating these spatial and transportation characteristics was then developed to estimate time-dependent boarding and alighting volumes. The model was applied to the ‘Ui-Sinseol Line’ (operational) and the ‘Myeonmok Line’ (planned) for validation. Results showed high accuracy for trips originating from residential areas and demonstrated the model’s potential as a complementary tool to existing demand estimation methods.
The proposed model offers practical utility by allowing cross-verification of traditional four-step model results and providing rapid station-level demand estimates. It can serve as a foundational dataset for policy applications such as new line planning, station area development, bus route coordination, and land use planning. While prediction accuracy improves when development plans are more defined, the model is especially effective for feasibility assessments of projects with established planning frameworks. Further research, including handling transfer stations and incorporating future development plans, will enhance the model’s applicability and effectiveness.