Investigating Spatial Autocorrelation and Spillover Effects in Freeway Crash-Frequency Data

Area of Science:

• Transportation Engineering
• Traffic Safety Analysis
• Spatial Statistics

Background:

• Micro traffic safety analysis often overlooks spatial dependencies between freeway segments.
• Understanding spatial autocorrelation and spillover effects is crucial for accurate crash frequency modeling.

Purpose of the Study:

• To investigate spatial autocorrelation and spillover effects in micro traffic safety.
• To propose a hybrid Poisson regression model incorporating these spatial effects for freeway crash frequency analysis.

Main Methods:

• Developed a hybrid Poisson regression model using conditional autoregressive (CAR) prior for spatial autocorrelation and adjacent segment variables for spillover effects.
• Employed Bayesian estimation in WinBUGS for model fitting and comparison.
• Utilized one-year crash data from Kaiyang Freeway, China.

Main Results:

• Significant spatial autocorrelation and spillover effects were found to coexist in freeway crash data.
• The hybrid model demonstrated superior fit (lower Deviance Information Criterion - DIC) compared to models with only one spatial effect.
• Daily vehicle kilometers traveled and segment alignments (horizontal and vertical) were identified as significant factors influencing crash occurrences.

Conclusions:

• Simultaneously accounting for spatial autocorrelation and spillover effects enhances traffic safety model performance.
• The findings underscore the importance of considering spatial dependencies and segment characteristics in freeway safety analysis.
• The proposed hybrid model provides a robust framework for identifying critical crash contributing factors.