Comparison of common methods for determining hazardous locations for improving road safety

Abstract

Identifying hazardous locations is crucial for maximising benefits from road safety investments. Using an appropriate method for identifying hazardous road locations (HRL) is essential due to limited research on existing approaches. This study evaluated the effectiveness of the four most commonly used approaches to prioritise HRLs such as crash frequency (CF), crash rate (CR) Empirical-Bayes (EB) adjustment and potential for safety improvement (PSI). This study used six years (2010–2015) of severe-crash data collected from 80 highway segments in Toowoomba, Australia. Crash prediction models were created to anticipate crash expectations. The negative binomial technique was found to be suitable for developing the models. These HRL identification techniques were assessed using rigorous quantitative criteria, such as the site consistency test, the total-rank differences test, the method consistency test and the total-score test. Our data demonstrate that the EB approach significantly outperformed the other ranking strategies. In contrast, the CR method consistently underperformed because of its inherent bias towards low-traffic sites. Notably, this technique assumes a linear relationship between CRs and traffic volume, despite earlier research proving the normal nonlinearity of this connection. As a result of this study, road engineers can develop models to predict crash trends and use the EB approach to prioritise treatment sites and identify the most hazardous locations for safety improvements. In conclusion, building on our current findings and prior research, we strongly recommend that the EB adjustment approach be adopted as the standard for determining HRLs unless alternative methods emerge to replace it.