Enhancing urban bikeability is crucial in building environmentally friendly and sustainable cities. While numerous studies have proposed various metrics, they generally overlooked the potential yet crucial impact of the urban visual environment on cycling behavior. To address this gap, we propose a Visual Bikeability Score (VBS). This score is derived by modeling the semantic elements within street view images against the actual volume of cycling activity. We apply this model to cycling on both routine and non-routine roads during the morning and PM peak. Based on this, we identify mismatches between the cycling environment and cycling behavior. Through Random Forest and SHAP analyses, we then uncover the underlying socio-economic factors that contribute to these mismatches. Our results indicate that VBS effectively evaluates the visual bikeability of urban spaces, particularly highlighting the critical role of beautiful urban landscapes. Based on the VBS analysis, we further categorize roads into “forgotten,” “matchable,” and “over-cycling” types. The analysis of socio-economic characteristics reveals that routine routes are consistently associated with work-related characteristics, whereas non-routine routes show a stronger correlation with socio-demographic and ethnic characteristics. This study provides targeted and broadly applicable strategies to mitigate the mismatch between the cycling environment and behavior, and to enhance urban cycling participation.
Hou et al. (Fri,) studied this question.