Empirical Bayes application on low-volume roads: Oregon case study

Area of Science:

• Transportation Engineering
• Road Safety Analysis
• Statistical Modeling

Background:

• Low-volume roads present unique challenges for traffic safety analysis.
• Traditional network screening methods may not adequately capture the characteristics of these roadways.
• Accurate crash prediction is crucial for effective road safety management.

Purpose of the Study:

• To evaluate the Empirical Bayes (EB) method for network screening on low-volume roads in Oregon.
• To compare the contributions of observed and Highway Safety Manual (HSM) predicted crashes in EB estimations.
• To assess the applicability of HSM predictive methodology for rural, low-volume road networks.

Main Methods:

• Analysis of approximately 870 miles of rural two-lane roadways in Oregon.
• Investigation of crash, traffic, and roadway data over a 10-year period.
• Examination of the influence of low traffic exposure on EB expected crash estimations.

Main Results:

• On low-volume roads, predicted crashes significantly contribute to EB expected crash estimations.
• A notable discrepancy exists between observed and HSM-predicted crashes, potentially due to unique road attributes.
• The HSM EB method yielded expected crash numbers reasonably close to observed data.

Conclusions:

• Network screening methods prioritizing risk factors are effective for low-volume roads, particularly with sparse crash data.
• The EB method, integrating HSM predictions, offers a viable approach for safety analysis on these roads.
• Findings support the use of predictive modeling for road safety management where empirical data is scarce.