Citation

Saleem T, Persaud BN, Shalaby A, Ariza A. Transp. Res. Rec. 2014; 2432: 142-148.

Copyright

(Copyright © 2014, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.3141/2432-17

PMID

unavailable

Abstract

Safety prediction models are designed to estimate the safety of a road entity and, in most cases, they link traffic volumes to crashes. A major problem with such models is that, because crashes are rare events, crash statistics cannot account for many of the possible contributing factors. Using traffic conflicts to measure safety can overcome this problem because conflicts occur more frequently than crashes do and can be either measured in the field or estimated with microsimulation models. This study developed crash prediction models from simulated peak hour conflicts for a group of urban four-legged signalized intersections in Toronto, Ontario, Canada, and evaluated their predictive capabilities. Case studies with two microsimulation packages, VISSIM and Paramics, demonstrated the use of microsimulation for estimating safety performance. For a further demonstration of the approach's versatility, VISSIM was used with precalibrated parameter values, while substantial effort was devoted to calibrating Paramics parameters with the crash data. For the assessment of the predictive capability of the crash-conflict models, specifically the models' ability to capture the safety impacts of geometric and operational variables, the effects of a hypothetical left-turn treatment on crashes and conflicts were explored and compared with results of an empirical Bayes study that evaluated actual treatments in Toronto. For this task, the predictive ability of the models for intersections with various ranges of average annual daily traffic and with various combinations of left- and right-turn lanes was also assessed. The results indicate that use of simulated conflicts is a viable, promising approach for intersection safety performance estimation. Free first page