Where do built environment attributes most effectively influence bike sharing usage?

Abstract

Built environment policies have proven to be an impactful strategy to promote bike sharing usage. However, most studies on spatial impact of built environment attributes failed to consider nonlinear relationships in their local models. Furthermore, most studies ignored holidays during weekdays in their models. With the data from New York, we explore the spatially nonstationary built environment effects on bike sharing usage on workdays and non-workdays considering nonlinear relationships with the geographically weighted gradient boosting decision tree approach. We found that the contributions of built environment attributes are spatially heterogeneous across the study area and differ between workdays and non-workdays. For example, proportion of park land use has a pronounced influence on central Manhattan and southern Brooklyn on non-workdays but exerts no notable impact on workdays. These findings provide targeted implications for designing location-specific built environment and transportation policies to encourage bike sharing usage.