Beyond station: Rethinking built environment effects on metro-integrated electric bike-sharing from an origin–destination perspective

Emerging electric bike-sharing (EBS) systems, by enabling higher speeds and longer travel distances, offer an effective solution to the first- and last-mile challenges of metro systems. However, existing studies have predominantly focused on the integration of conventional bike-sharing with metro systems, leaving the role of EBS in enhancing metro connectivity insufficiently understood. Moreover, most existing studies adopt a station-centric perspective, implicitly assuming that first- and last-mile travel is shaped only by the built environment within station catchment areas. This perspective may be less applicable to EBS, whose longer travel range often extends beyond such catchment areas. To address these limitations, this study adopts an origin–destination (OD) perspective to examine how built environment characteristics at station areas and non-station trip ends, measured respectively at the station and grid levels, together with travel distance, jointly shape metro-integrated EBS usage. Using EBS trip data from Hefei, China, this study employs eXtreme Gradient Boosting (XGBoost) combined with SHapley Additive exPlanations (SHAP) to examine the nonlinear and interaction effects of these factors. The results show that metro-integrated EBS usage exhibits a clear commute-oriented pattern, with pronounced peaks during the morning and evening rush hours. Spatially, residential-related feeder trips, including morning access and evening egress, are concentrated in the northeastern part of the city, whereas workplace-related feeder trips, including morning egress and evening access, are more concentrated in the western and southern areas. Travel distance emerges as the most influential determinant across all feeder contexts. Station-level attributes are more important in residential-related contexts, whereas grid-level attributes are more important in workplace-related contexts. With respect to nonlinear effects, station-level built environment attributes exhibit broadly similar patterns across feeder contexts, whereas grid-level built environment effects show pronounced context-dependent variation between residential-related and workplace-related contexts. In addition, interaction analysis reveals a clear distance-attenuation pattern, with station-level built environment effects weakening as travel distance increases. These findings provide valuable insights for optimizing EBS operations and designing context-sensitive built environment interventions to strengthen EBS–metro integration.