Comparison of threshold determination methods for the deceleration rate to avoid a crash (DRAC)-based crash...

Area of Science:

• Traffic safety research
• Transportation engineering
• Statistical modeling

Background:

• Traffic conflicts are increasingly used as surrogate safety measures.
• The deceleration rate to avoid a crash (DRAC) is a key indicator, with higher values indicating greater crash risk.
• Maximum available deceleration rate (MADR) varies significantly and impacts crash risk assessment.

Purpose of the Study:

• To compare different threshold determination approaches for DRAC-based crash estimation.
• To evaluate the effectiveness of Bayesian hierarchical extreme value modeling in this context.
• To identify the optimal method for MADR threshold determination.

Main Methods:

• Applied Bayesian hierarchical extreme value modeling.
• Utilized traffic conflict and crash data from four signalized intersections.
• Developed models using specific MADR values and truncated normal distributions for MADR.

Main Results:

• The truncated normal distribution N(8.45, 1.4^2)I(4.23, 12.68) for MADR demonstrated superior accuracy in DRAC-based crash estimation.
• Multisite Bayesian hierarchical models outperformed at-site models in both accuracy and precision.
• Crash frequencies estimated by the models were compared against observed crashes.

Conclusions:

• The truncated normal distribution N(8.45, 1.4^2)I(4.23, 12.68) is recommended for MADR threshold determination in DRAC-based crash estimation.
• Multisite Bayesian hierarchical modeling provides more accurate and precise crash estimations.
• This research offers improved methods for traffic safety analysis using surrogate measures.