Speaker
Description
With the promotion of sustainable transportation, cycling has gained significant attention as an eco-friendly and healthy mode of travel. Cyclists, unlike motor vehicle users, are directly exposed to street environments without physical protection, making the quality of street environment crucial to their riding experience. Previous research has shown that streetscape features, including greenery visibility and sky openness, significantly influence cycling behavior. However, there are two main research gaps: firstly, previous studies have predominantly focused on the impact of street environments on aggregate-level cycling volume, with insufficient attention to individual-level route choice behavior; secondly, existing literature has explored dynamics of cycling volume at a relatively crude temporal resolution, such as weekday/weekend or peak/off-peak distinctions, while the dynamics of cycling route preferences at different periods of the day remain far from understood.
This study endeavours to unveil cyclists' authentic environmental preferences by analyzing their route choice behaviors: 1) unraveling the nonlinear relationship between street environment factors and cycling route choices and exploring thresholds of key factors to provide actionable design guidelines for creating bicycle-friendly street environments; 2) investigating the diurnal dynamics of cyclists' environmental preferences to offer precise route recommendations for Mobility-as-a-Service (MaaS) systems.
Leveraging a large-scale dataset of 4,739,419 bike-sharing trajectories from the central city of Shanghai, we analyzed route preference patterns within each origin-destination (OD) pair. We investigated comprehensive street environment factors, including nighttime illumination intensity obtained from the Earth Observation Group (EOG) at the Colorado School of Mines, and twelve categories of streetscape element proportions (e.g., pedestrian paths, buildings, and sky) calculated from Baidu Map high-resolution streetscape images. Additionally, road network features (e.g., real-time traffic congestion, route length) and surrounding environmental variables (e.g., land use diversity) were considered as control variables. A Path Size Logit model, combined with XGBoost was employed to capture non-linear relationships between street environment factors and route preferences, and SHAP (SHapley Additive exPlanations) values were used to interpret the model results and identify critical thresholds of streetscape factors. Moreover, in order to explore temporal dynamics, the sample was divided based on departure time into four categories: morning peak (7:00-9:00), daytime (9:00-17:00), evening peak (17:00-19:00), and night (19:00-24:00), conducting subgroup analyses for each.
Key findings include: (1) Significant threshold effects of critical street environment factors were identified. For instance, greenery visibility had negligible impact in the low range (10%-20%) but showed exponential positive influence once it surpasses 30%. (2) Cyclists' environmental preferences varied significantly by period of day: during the morning peak, route choices were primarily influenced by route length and traffic congestion, highlighting an efficiency-driven preference. In contrast, during the evening peak, despite the commuting focus, the importance of greenery visibility increased, reflecting a heightened concern for environmental comfort in a fatigued state after a day's work. At night, illumination intensity became the predominant factor affecting route choice, possibly related to perceived safety.
These findings provide a quantifiable design index for creating bicycle-friendly street environments, such as ensuring greenery visibility above 30% for main cycling corridors. Furthermore, the revealed dynamic environmental preference patterns provide a theoretical foundation for MaaS platforms to enhance intelligent services. It is advisable to adjust the weights of environmental factors according to different periods of the day, like prioritizing route efficiency during morning peaks, emphasizing the greenery and lighting conditions in the evening and night, respectively. These empirically-based insights not only offer precise quantitative standards for urban street design but also lay the groundwork for MaaS platforms to develop dynamic, personalized route recommendations, advancing urban transportation systems towards greater intelligence and human-centricity.
| Keywords | Streetscape; Bike-sharing; Route preferences; Temporal dynamics; Trajectory data |
|---|---|
| Best Congress Paper Award | Yes |
