Evaluation of Safety Identification Methods for Low-Volume Roads Using Monte Carlo Simulation

Abstract

Two-lane rural roads characterized by low traffic volume (fewer than 3,000 vehicles per day) represent a specific road typology of interest in safety analysis. Correct identification of black spots (sites in which the expected number of crashes is abnormally high) is crucial for the road administration when the location and ranking of improvements are chosen. Simply observing unusually high crash counts does not necessarily indicate a true safety problem. In black spot identification on two-lane rural roads with low traffic volume, the reduced number of observed crashes is a critical issue that can emphasize weak results carried out with different safety indicators and control checks. Generally it is assumed that the empirical Bayesian estimate represents the best approach for black spot identification, but the precision of the estimation among different procedures is difficult to define quantitatively. The simulation approach is able to identify crash data similar to empirical data with the unique possibility to define a priori hazardous sites and therefore assess whether a method can correctly identify them. The Monte Carlo simulation presented here was able to define the accuracy and efficiency of the procedures on the basis of observed frequency of crashes, crash rate, empirical Bayes estimation, and potential for safety improvement. The results show that an appropriate selection of exposure values can mitigate the lower performance of some indicators, but the best practice is to use the empirical Bayes approach to identify true positives.