Rural Speed Safety Pilot Project
Current crash prediction methods, such as those in the first edition of the Highway Safety Manual (HSM) use traffic volume data along with geometric characteristics to predict the annual average crash frequency of roadway segments and intersections. One limitation of the HSM is the omission of speed-related factors from all aspects of safety predictive methods. Recent research has made little substantive progress in incorporating speed-related factors into crash predictive models. To advance the state of practice, this study begins the work of investigating the association between crash risk and traffic speeds using traffic speed information from rural roadways of two states: Washington (WA) and Ohio (OH).
Certain speed measures were found to be beneficial in quantifying safety risk. Annual-level crash prediction models show that increased variability in hourly operating speed within a day and an increase in monthly operating speeds within a year are both associated with a higher number of crashes. This model also shows that when operating speed difference between weekends and weekdays is greater, all rural facility roadways (rural two-lane, rural multilane, and rural interstate) experience a higher number of crashes. Daily- level crash prediction models show that a segment with high variation in daily average speeds is expected to experience a higher number of crashes than a segment with a lower variation in daily speeds. The strength of this finding is one of the biggest insights gained from this study. An exploratory examination of time before and after crashes shows that speed variation increases significantly before a crash event compared to a non-crash traffic flow condition. Overall, this study shows benefit in incorporating speed data in safety modeling to more effectively identify locations that would benefit from additional safety treatments and countermeasures.
This pilot project establishes a framework of data integration and analytical procedure that will help to address the effect of operating speed measures on safety. The framework developed in this study can be applied to other states. The major outcome of this project is the safety prediction models (annual and daily) that incorporate speed and weather data. Another outcome of this project is a beta version of an interactive decision support tool that shows segment-level annual crash estimates using Washington and Ohio data. Subsequent study may examine some limitations found in this study, particularly some missing data in the current version of NPMRDS and zero inflation in short-term crash prediction, to see if those limitations can be overcome by revised versions of the data, additional data sources, and refinement in modeling.