Desirable bikeshare routes: Nonlinear impacts of micro-level street environments

Facilitating an attractive and safe street environment (SE) for cyclists can yield multifaceted benefits for sustainable, healthy cities. However, due to limited urban-scale trajectory data, the influence of micro-level SE (e.g., trees, shrubs, perceptions, and road cracks) on cyclists’ route preferences remains less understood. Using over 4,000 Lime Dockless Bikeshare trajectories, this study constructed a Desirable Bikeshare Index (DBI) that measures differences between observed trips and the shortest routes suggested by Google Maps at the street-segment level. Using computer vision, road damage was detected in street view images, alongside objective streetscape features and subjective perceptions (e.g., enclosure) to characterize micro-level SEs comprehensively. Nonlinear associations between SE and DBI were identified using explainable machine learning algorithms, and customer satisfaction theory (e.g., basic, excitement, performance factors) was adopted to interpret the threshold effects. Contrary to earlier findings, green spaces do not always attract cyclists – lush trees draw cyclists, while too many shrubs repel them. Moreover, sidewalks are a must-have (basic factor), while road damage is more common where cyclists aggregate, underscoring the need for timely monitoring and maintenance of bicycle infrastructure. Perceptions of accessibility, order, and richness are excitement factors that explain route desirability. Interestingly, perceived ecology and enclosure promote cycling at low values, then quickly play a negative role beyond mid-to-high values. Our findings underscore the need to combine subjective perceptions with objective streetscapes to delineate a comprehensive micro-level SE. Urban designers and transport planners can leverage revealed thresholds to implement cost-effective infrastructure upgrades that facilitate desirable street environments for cyclists.