One of the first steps in effectively managing a road network is to identify sites that require safety investigations. It is important that this process of identifying the 'sites with promise' be efficient because scarce resources may be wasted on sites that are incorrectly identified as unsafe and sites that are truly unsafe may not be flagged in this process. In North Carolina, safety warrants have been developed to identify intersections and sections that are possibly deficient. Sites are ranked based on their frequency of crashes by type and severity, and the percentage distribution of relevant crash types.
For the past 8 years or so, as part of the Highway Safety Manual (HSM), methods based on the empirical Bayes (EB) approach have been proposed for identifying sites with promise. In theory, the methods based on the EB approach are supposed to be better than traditional methods (that make use of crash frequency or crash rates) primarily because the EB method has been shown to be effective in controlling for the possible bias due to regression to the mean (RTM), and account for the non-linear relationship between crash frequency and traffic volume. However, very little work has been to compare the different methods in a realistic setting. The intent of the proposed project is to compare the performance of different methods for identifying sites for safety investigations and determine if a particular method is more optimal for identifying locations susceptible to cost-effective safety improvement. Following is a preliminary list of methods that could be evaluated (the final list will be decided following discussion with NCDOT):
- Methods currently used by NCDOT as part of the Highway Safety Improvement Program (HSIP)
- Empirical Bayes method that uses Expected Crashes as the measure of interest
- Empirical Bayes method that uses Expected Excess Crashes as the measure of interest
- LOSS method, where the measure of interest is the difference between observed crashes and predicted crashes from a safety performance function (SPF)
- Excess proportion of specific crash types
Combinations of methods will also be considered in addition to the considering these methods individually. The overall approach will involve the following steps:
- Apply each method (or combination of methods) using NCDOT data and identify sites of interest. It is possible that some sites may get included in one list, while others may be included in multiple lists.
- Compare the sites identified in Step 1 with sites already investigated by NCDOT. Identify sites that may need further investigation. This is expected to be only a handful of sites. Conduct a field investigation and identify countermeasures (if any) for each site identified by the different methods. The field investigations will have to be done by qualified NCDOT staff.
- Based on the cost of these countermeasures and the expected benefits, determine if a particular method (or combination) is more likely to identify locations susceptible to cost-effective safety improvement.