Developing Analytical Procedures for Calibrating the Highway Safety Manual Predictive Methods

Performance-based safety goals and objectives are more attainable with the use of the Highway Safety Manual (HSM). However, the safety performance functions (SPFs) in the HSM may not be accurate when used with local jurisdictions. Each SPF and crash modification factor (CMF) assume a set of base site conditions that might not be realistic or representative of local highways. The calibration procedures provided in Part C, Appendix A, of the HSM should therefore be modified to accommodate local data availability as well as roadway, traffic, and crash characteristics. Furthermore, a set of base conditions applicable to local highways should be determined. Results show that the HSM models underestimate the total average number of crashes on highway segments for all rural highway facility types in South Dakota. To quantify highway safety performance better, this study analyzed the underlying factors contributing to the underestimation and proposed procedures to improve crash prediction accuracy. The HSM calibration was performed with crash data from rural two-lane, two-way highways and rural multilane divided and undivided highways during a 5-year period (2008–2012). The procedures included establishing new base conditions, developing jurisdiction-specific SPFs, and converting CMFs to new base conditions. The comparison results show that the customized models outperform the HSM models in predicting sites with crashes.