Discovering the nonlinear association between the built environment and metro ridership: Global and local perspectives

Ridership serves as a critical measure of the interaction between urban spaces and metro systems, and a detailed analysis of its influencing factors is pivotal for the advancement of sustainable transportation. While prior studies have examined the nonlinear relationship between built environment features and metro ridership, they have given limited attention to incorporating spatial effects arising from station proximity or accounting for spatial heterogeneity in these associations within their nonlinear models. In addition, the integration of nonlinear relationships at both the citywide and local levels has not been fully explored. To address these gaps, this study develops a GW-XGBoost model, and proposes a comprehensive framework that integrates global and local perspectives. Five categories of built environment factors are examined and interpretation approaches are applied to elucidate how these factors are association with ridership. Employing Shanghai Metro as a case study, our findings reveal that GW-XGBoost outperforms benchmark models, underscoring the critical nature of nonlinear and spatial effects in ridership modeling. Empirical findings indicate that station context and land use explain over half to the ridership predict, with betweenness centrality, commercial land, residential land being the most correlated factors. Additionally, as land use intensity, betweenness centrality, station entrances and exits, and the number of bus stops and lines increase, their associations with ridership follow a rising trend until a saturation point is observed. In contrast, variables including the green view index and elderly ratio show declining or inverse effects. Moreover, the relationship between built environment factors and ridership shows pronounced spatial heterogeneity, varying from the urban core to suburban areas with varying patterns. The proposed framework is transferable to other cities and provides valuable insights for urban planning and transportation management.