Calibration of Highway Safety Manual prediction models for freeway segments, speed-change lanes, ramp segments, and crossroad ramp terminals in Kansas

Abstract

Crash prediction models in the Highway Safety Manual (HSM) are used to quantify the safety experience of existing and new roadways. Safety performance functions (SPFs) or crash prediction models are statistical formulas developed on limited data from a few selected states, Kansas not being one of those states. Therefore, the HSM recommends calibration of HSM-default SPFs, or development of local SPFs, to enhance accuracy of predicted crash frequency. This dissertation demonstrates the HSM calibration procedure and its’ quality assessment for freeway segments, speed-change lanes, ramp segments, and crossroad ramp terminals in Kansas. The study used three years of recent crash data, the most recent geometric data, and HSM-recommended sample sizes for all facilities considered for the calibration. The HSM methodology overpredicted all fatal and injury (FI) crashes and underpredicted all property damage only (PDO) crashes for freeway segments. The HSM methodology consistently underpredicted both FI and PDO crashes for both entrance- and exit-related speed-change lanes. The HSM methodology overpredicted all FI crashes, underpredicted multiple vehicle PDO crashes, and overpredicted single vehicle PDO crashes for entrance ramp segments. In the case of exit ramp segments, the HSM methodology underpredicted all multiple vehicle crashes and overpredicted all single vehicle crashes. The HSM methodology overpredicted all FI crashes and underpredicted all PDO crashes for both signal- and stop-controlled crossroad ramp terminals. Cumulative residual plots and coefficient of variation were used to evaluate the quality of calibrated HSM-default SPFs. Results of calibration quality assessment indicated that estimated calibration factors were satisfactory for all freeway and ramp facilities considered in this study. However, for further accuracy and comparison purposes, calibration functions were developed to improve the fit to local data. Calibration functions were better fitted compared to calibrated HSM-default SPFs for freeway and ramp facilities in Kansas. Challenges faced, how those challenges were addressed, and data collection techniques used in this study are discussed. In summary, estimated calibration factors and developed calibration functions of this study would greatly improve making accurate decisions related to freeway and ramp safety in Kansas.