Spatial and Temporal Heterogeneity of Multi-Scale Built Environment Influences on Cycling Behavior

Effective bike-sharing planning requires understanding how built-environment associations with usage vary across locations and day types. Existing research remains limited by comparing the relative contributions and spatial variation in macro-scale and street-view attributes within one framework, distinguishing an attribute’s general contribution from geographic context and its interaction with location, or directly assessing weekday–weekend differences. Using Manhattan Citi Bike data and 24 environmental indicators, this research applies an explainable machine learning approach incorporating GeoShapley to examine nonlinear, spatially varying associations and distinguish environmental variable contributions, geographic context, and their interactions. Geographic context ranked first, and station supply had the largest environmental variable contribution in both periods. Metro accessibility, building density, and commercial land use were more important on weekdays, particularly in central and southern Manhattan; residential land use, cycling route density, and station supply were more important on weekends. Among street-view features, visible buildings were more important on weekdays, while sidewalk visibility had a broader positive contribution on weekends; greater sky visibility was generally associated with lower cycling volume. These findings show how built environment contributions vary across locations and day types and inform context-sensitive bike-sharing and cycling infrastructure planning.