Macroscopic spatial analysis of pedestrian and bicycle crashes

Area of Science:

• Transportation Science
• Spatial Statistics
• Public Health

Background:

• Modeling pedestrian and bicycle crashes is crucial for urban planning and road safety.
• Aggregate models often overlook spatial dependencies between geographic areas.
• Understanding factors influencing crash risk requires robust statistical approaches.

Purpose of the Study:

• To investigate the impact of spatial correlation on modeling pedestrian and bicycle crashes.
• To compare Bayesian spatial models with non-spatial models for crash prediction.
• To identify key predictors for pedestrian and bicycle crashes at the Traffic Analysis Zone (TAZ) level.

Main Methods:

• Utilized a Bayesian spatial framework, specifically the Poisson-lognormal model.
• Modeled pedestrian and bicycle crashes within Traffic Analysis Zones (TAZs).
• Accounted for spatial correlation among TAZs in the analysis.

Main Results:

• Bayesian models incorporating spatial correlation outperformed those that did not.
• Identified distinct sets of significant predictors for pedestrian and bicycle crashes.
• Nine variables, including roadway characteristics, income, dwelling units, population density, vehicle ownership, parking costs, and employment, were significant for pedestrian crashes.

Conclusions:

• Spatial correlation is a critical factor in aggregate-level modeling of pedestrian and bicycle crashes.
• The findings underscore the importance of incorporating spatial dependencies for accurate traffic safety analysis.
• Distinct socio-demographic and roadway factors influence pedestrian versus bicycle crash risks.