A cross-comparison of different techniques for modeling macro-level cyclist crashes

Area of Science:

• Transportation Safety
• Urban Planning
• Traffic Engineering

Background:

• Cyclists are vulnerable road users, necessitating research into factors influencing their safety.
• Sustainable transportation modes like cycling face safety challenges that require investigation.

Purpose of the Study:

• To evaluate and compare macro-level cyclist safety models.
• To identify factors contributing to cyclist crashes using a comprehensive covariate list.

Main Methods:

• Developed four macro-level crash models (CMs) within a full Bayesian framework: Poisson lognormal (PLN), random intercepts PLN (RIPLN), random parameters PLN (RPPLN), and spatial PLN (SPLN).
• Utilized data from 134 traffic analysis zones (TAZs) in Vancouver, incorporating variables on traffic exposure, socio-economics, land use, built environment, and bike networks.

Main Results:

• The spatial PLN model demonstrated the best goodness of fit, emphasizing the impact of spatial correlation.
• Cyclist crashes positively correlated with bike/vehicle exposure, household density, commercial area density, and signal density.
• Negative associations with cyclist crashes were observed for bike network indicators like average edge length, average zonal slope, and off-street bike links.

Conclusions:

• Spatial modeling is crucial for accurately assessing macro-level cyclist safety.
• Understanding the interplay of exposure, urban form, and network design is key to enhancing cyclist safety.