Likelihood estimation of secondary crashes using Bayesian complementary log-log model.

Area of Science:

• Traffic safety research
• Transportation engineering
• Urban planning

Background:

• Secondary crashes (SCs) significantly increase traffic delays and reduce safety, especially in urban environments.
• Limited understanding of SCs hinders effective mitigation strategies.
• Proactive prevention of SCs is crucial for improving road safety.

Purpose of the Study:

• To develop a reliable risk prediction model for secondary crashes (SCs).
• To utilize real-time traffic flow conditions for SC risk assessment.
• To identify key factors influencing SC occurrence for proactive mitigation.

Main Methods:

• Data collected over three years on a 35-mile I-95 freeway section in Jacksonville, Florida.
• Secondary crashes identified using travel speed data from Bluetooth detectors.
• Bayesian random effect complementary log-log model employed to link SC probability with various factors; Random Forests used for variable selection.

Main Results:

• Key predictors of SC likelihood identified: average occupancy, incident severity, percentage of lanes closed, incident type, clearance duration, impact duration, and time of occurrence.
• Model demonstrates significant relationships between traffic flow, incident characteristics, and SC occurrence.
• Average occupancy and incident severity were found to be particularly influential.

Conclusions:

• The developed SC risk prediction model offers potential for proactive SC prevention.
• Understanding the impact of real-time traffic flow and incident characteristics is vital for enhancing road safety.
• Findings can inform traffic management strategies to mitigate secondary crash risks.