Intersection accident estimation: the role of intersection location and non-collision flows

Area of Science:

• Traffic Engineering
• Transportation Safety
• Statistical Modeling

Background:

• Existing models for relating traffic flow to accidents often lack specificity in model form and statistical analysis.
• Accurate prediction of intersection accidents is crucial for effective traffic safety management.

Purpose of the Study:

• To review and evaluate practical models for accident prediction based on traffic flows.
• To develop and apply generalised linear models (GLMs) for predicting individual accident types at intersections.
• To investigate the influence of intersection location and non-collision flows on accident occurrence.

Main Methods:

• Review of existing accident prediction models, focusing on model form and parameter estimation techniques.
• Development of generalised linear models (GLMs) for predicting specific accident types at New Zealand intersections.
• Application of covariate analysis to assess the impact of intersection location and non-collision flows.

Main Results:

• GLMs predicting individual accident types based on conflicting flows are superior to models predicting total accidents based on approach flows.
• Intersection location significantly influences the occurrence of different accident types.
• Interactions between turning flows and non-collision flows are important factors in accident mechanisms.

Conclusions:

• Generalised linear models provide a more accurate and nuanced approach to predicting intersection accidents, particularly when differentiating by accident type and considering flow interactions.
• Network-level accident predictions using these models show fair agreement with observed data, despite poorer agreement at individual intersections.
• The findings support the use of GLMs for targeted safety interventions and network-wide traffic management strategies.