Suggestions on How Agencies Should Apply the Highway Safety Manual to Two-Lane Road Curves

Horizontal curves are relatively risky portions of the highway system in the US and elsewhere. Fortunately, curves are also places where highway agencies have many options and opportunities for making safety improvements. The publication of the Highway Safety Manual (HSM) offers agencies a tool to evaluate the safety of a horizontal curve for safety with fewer field visits. This paper evaluates the HSM crash prediction model using data on two-lane rural horizontal curves in North Carolina. The objectives of this paper include providing highway agencies with practical advice on how they can validate and apply the HSM to the analysis of horizontal curves. An analysis of the local conditions calibration factor for the HSM predictive model in North Carolina found that a large number of sites (approximately 300) are required to meet HSM recommendations. The large number of sites is partly due to the finding that random segments provided a more accurate outcome than a subset of high crash locations. The primary purpose of this study was to analyze the differences in predicted collisions based on field data collection versus using average or default values. Furthermore, these study results can provide highway agencies with practical guidance on how they can apply the new HSM methods to the analysis of horizontal curves for their own local or state conditions. The results showed that annual average daily traffic, curve radius, and curve length were the most important factors in determining prediction accuracy but that average or default values may be used for other parameters with less risk to accuracy.