Quality assessment between calibrated highway safety manual safety performance functions and calibration functions for predicting crashes on freeway facilities

Crash prediction models are commonly used for network screening in highway safety management process, where potential impacts of highway safety treatments are quantified. The Highway Safety Manual (HSM) provides crash prediction models for various types of highway facilities that are often referred to as safety performance functions (SPFs). Freeway facility SPFs in the HSM were developed using data gathered from the states of California, Maine, and Washington. When applying these HSM-default SPFs to a local jurisdiction, the HSM recommends calibration of HSM-default SPFs or development of jurisdiction-specific SPFs to improve the accuracy of crash predictions. This study first calibrated the HSM-default freeway SPFs and for further accuracy and comparison purposes calibration functions were developed using Kansas freeway data. The performance of calibrated HSM-default SPFs was then compared with developed calibration functions concerning the accuracy in crash prediction. Freeway facility calibration dataset included 521 freeway segments, 351 entrance-related speed-change lanes, and 366 exit-related speed-change lanes. Cumulative residual plots and several other goodness-of-fit measures were used to assess the quality of calibrated HSM-default SPFs and calibration functions. Calibration functions fitted better compared to calibrated HSM-default SPFs for Kansas freeway data. The methodology used in this study could be beneficial and practiced to any jurisdiction. Calibration functions could be used as an alternative to jurisdiction-specific SPFs or a replacement for HSM-default SPFs, which are frequently used in comparing alternatives, in calculating economic benefits of project improvements, and in estimating economic effectiveness of crash reduction in highway safety-related decision making. (C) 2019 Periodical Offices of Changan University. Publishing services by Elsevier B.V. on behalf of Owner.