Exploring the heterogeneous nonlinear effects of the built environment on metro ridership considering bike-sharing catchment areas

Existing research on the relationship between the built environment and metro ridership mostly uses pedestrian catchment areas (PCA) such as the 800 m buffer, ignoring bike-sharing catchment areas (BSCA). The heterogeneous nonlinear effects of the built environment in different types of stations also tend to be overlooked. Here, we used bike-sharing trip data to identify the BSCA of each metro station in Beijing. A Fuzzy C-means (FCM) clustering-based explainable machine learning framework was proposed to explore the nonlinear effects of the built environment, including land use, transport, accessibility and socio-economic factors on four kinds of metro ridership under PCA and BSCA, respectively. The heterogeneity of the nonlinear relationship among different types of stations was also investigated. We found that BSCA-based models outperform PCA-based models. The nonlinear effects of the built environment on metro ridership change with various catchment areas. These effects tend to have various trends and thresholds in different types of stations. Furthermore, residential and employment-oriented stations are more sensitive to catchment areas. This paper emphasizes the role of BSCA and heterogeneous planning targets in different types of stations in TOD (Transit-oriented development) planning.