Files
Abstract
This paper evaluates the Highway Safety Manual (HSM) crash prediction model using data on twolane
rural horizontal curves in North Carolina. 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 results showed that annual average daily
traffic, curve radius, and curve length were the most important factors in determining crash
prediction accuracy, but that average or default values may be used for other parameters with less
risk to accuracy.