This project had its origins in an earlier project which showed that crash rates increased at extremes of horizontal and vertical geometry, and were lowest with narrow sealed shoulders. This paper reports an investigation based on the original database to resolve the issues of the relation between advisory speeds on crashes and the relation of geometry and cross-section to crashes involving motorcycles or trucks. The study linked data relating to road geometry, cross-section, traffic flow and crashes using the ARCINFO Geographic Information System. It was used to generate maps based on GPS data collected by instrumented survey vehicles surveying the routes in question, the other data subsequently being overlaid on these maps. A high correlation between advisory speed and crash rate indicates that crash rate increases as combinations of geometric features become more demanding. Motorcycle and truck crashes showed the same relationship to advisory speeds as did other vehicles, except that motorcycles tended to have high relative crash rates for low advisory speeds and low relative crash rates for high advisory speeds, and trucks tended to have low relative crash rates for low advisory speeds. There were few differences among the relative crash rates for motorcycles, trucks and other vehicles in relation to other geometric variables. These results suggest that advisory speed captures more of the totality of risk to road users than do individual geometric variables, and so reveals relationships which were not evident in the process of examining the relationship between geometry and crashes variable by variable. Analysis by advisory speed should be a key feature of similar studies in future. The results also suggest that there may be different relationships between advisory speed and crashes for different types of vehicle, although these findings may be explicable in terms of exposure in different road environments. Since Gipsi-Trac surveys have been carried out for much of Australia’s main road system, and since some jurisdictions already geocode crashes, it may be possible to conduct a definitive study of the relationship between crashes and infrastructure characteristics using routinely-available outputs when motorcycle traffic data becomes available. (Author/publisher)
Abstract