Influence of Walking Accessibility for Metro System on Pedestrian Safety: A Multiple Membership Multilevel Model
Manman Zhu,N. N. Sze,Haojie Li
DOI: https://doi.org/10.1016/j.amar.2024.100337
IF: 14.556
2024-01-01
Analytic Methods in Accident Research
Abstract:In the past decades, many cities have adopted transit-oriented development approach for urban planning. Studies have explored the effects of built environment, street network and accessibility on the perception and behaviour of pedestrians. However, the relationship between pedestrian safety and walking accessibility is less studied. In this study, influences of land use, socio-demographics, pedestrian network, and transport facilities on pedestrian crash frequencies in the areas around metro stations would be evaluated. Additionally, walking accessibility for individuals with and without physical disabilities would be accounted for. Since data at different spatial scales, i.e., zone level (individual) versus catchment area level (group), are used, the hierarchical approach is adopted for the crash frequency model. Furthermore, some zones are nested within the catchment areas of more than one metro station, the multiple membership approach should be adopted, accounting for the possible correlation. Different from the conventional multiple membership multilevel model, multiple membership weights would be assigned in accordance with the walking distances between zones and stations. Last but not least, temporal instability in the parameter estimation is also explored. Results indicate that pedestrian crash frequencies increase with population density, working population, traffic volume, walking trip, footpath density, node density, barrier-free facilities, bus stop, residential area, commercial area, and government and utility area. In contrast, pedestrian crash frequencies decrease with average gradient and walking accessibility. Findings should shed light on the street design that can enhance walking accessibility and public transport use, without compromising pedestrian safety. Moreover, issues of spatial crash analysis, including hierarchical data structure, and between- and within-group variances, are addressed.