The extreme value theory approach to safety estimation.

Area of Science:

• Traffic Safety Engineering
• Statistical Modeling

Background:

• Traditional crash-based safety analysis faces challenges due to data limitations like rarity, randomness, and inconsistency.
• Analyzing observable traffic characteristics offers a more frequent and potentially reliable alternative to crash data.

Purpose of the Study:

• To propose and evaluate a novel proactive safety analysis method using extreme value theory.
• To assess the method's applicability to right-angle collisions at signalized intersections without relying on historical crash data for calibration.

Main Methods:

• Application of extreme value theory to estimate road safety based on traffic characteristics.
• Evaluation of the method using right-angle collision data at signalized intersections.
• Comparison of estimated safety with historical crash data and Poisson-based confidence intervals.

Main Results:

• The proposed method demonstrated a promising relationship between safety estimates and historical crash data.
• Crash estimates at 7 out of 12 sites fell within established Poisson-based confidence intervals.
• A simulation indicated that 3-6 weeks of observation are required for reliable safety estimates.

Conclusions:

• The extreme value theory-based method offers a proactive approach to safety analysis, reducing reliance on historical crash data.
• The method shows potential for application to various collision types and locations.
• Sufficient observation periods are crucial for obtaining robust safety estimates with narrow confidence intervals.