Evaluating safety effects of variable speed limit systems via joint modeling.

Area of Science:

• Traffic Engineering
• Transportation Safety
• Statistical Modeling

Background:

• Variable Speed Limits (VSL) are crucial for Active Traffic Management Systems (ATMS) to improve traffic flow, safety, and reliability.
• Previous safety impact studies often modeled crash outcomes separately, overlooking correlations between collision type and severity.
• This oversight can lead to underestimation of the effectiveness of traffic management interventions.

Purpose of the Study:

• To evaluate the safety impacts of Variable Speed Limits (VSL) implemented on the I-24 Smart Corridor in Nashville, Tennessee.
• To jointly model rear-end, injury, and Property Damage Only (PDO) crashes using a copula framework to account for interdependencies.
• To estimate the Crash Modification Factor (CMF) for VSL implementation using the Difference-in-Differences estimator.

Main Methods:

• Utilized a copula framework (Gaussian, Frank, Clayton) to jointly model crash types and severities.
• Employed the Difference-in-Differences estimator within marginal Negative Binomial (NB) models to calculate CMFs.
• Compared the performance of copula models against separate NB models for crash outcome analysis.

Main Results:

• Copula models significantly outperformed separate Negative Binomial models in analyzing crash data.
• Variable Speed Limits (VSL) implementation resulted in a Crash Modification Factor (CMF) of 0.677 for rear-end crashes and 0.686 for injury crashes.
• A high correlation (0.91) between reduced rear-end and injury crashes suggests a direct safety benefit, while PDO crash changes were not statistically significant.

Conclusions:

• The study confirms the positive safety impact of Variable Speed Limits (VSL) systems.
• Jointly modeling crash types and severities provides a more accurate assessment of treatment effects.
• Findings support the justification of future investments in VSL systems for traffic management.