This paper describes relationships between roundabout geometry and accident rates for one hundred roundabouts from throughout Queensland, Australia. Regression analyses are undertaken on "single vehicle," "approaching rear end," and "entering/circulating" accident categories. After various trials, it was found to be most important that accident models incorporate appropriate explanatory variables. The models described here were based on the concept of exposure and propensity, and on observed driver behavior. The driver behavior model developed in this study uses the 85th percentile speeds on each geometric element and the location of vehicle paths through the roundabout. The single-vehicle accident model includes traffic flow, length of driver path on the geometric element, 85th percentile speed on the previous geometric element, and radius of the geometric element. This model demonstrates the importance of limiting the difference between the expected drivers' speeds through successive geometric elements. The approaching rear-end vehicle accident model includes approaching and circulating traffic flow, and approach speed. This model demonstrates the importance of limiting the approach speed. Similarly, the entering/circulating vehicle accident model includes approaching and circulating traffic flow, and relative speed between entering and circulating vehicles. This model demonstrates the need to minimize the relative speed between entering and circulating vehicles. A brief discussion on the application of this research to other intersection types and roadways is given.
Abstract