IDENTIFYING HAZARDOUS ACCIDENT SITES (CETERIS PARIBUS

Locations with excessive accident histories are intended to be treated by authorities to improve safety. Therefore the initial correct identification of such 'hazardous' sites is critical. Treatments may be applied to individual locations, road features and longer complete sections of road. All locations are essentially analysed the same way but the basis of alternative philosophies which are used in this process has been confused. This paper discusses several reasonable definitions of a hazardous site which are in use. These include sites which have high accident frequencies, high accident rates or sites which have an uncharacteristic accident pattern. Simple examples are used to illustrate mathematical weaknesses of alternative definitions. It is shown that the current practice of analysing accident histories does not routinely identify all sites which may be comparatively hazardous. Sites are not always compared uniformly because there are important differences between sites which are not considered. Several suggestions are made to overcome the selection weakness. These include segregating accident types and approaches, and mathematical modelling of the accident exposure relationship with minimum constraints. The use of accident rates should only be employed as a last resort because it results in a badly biased selection. The use of accidents modelled against exposure provides the most accurate method of identifying hazardous sites. (A) For the covering entry of this conference, see IRRD abstract no. 843369.