Safety at rural three- and four-arm junctions : development and application of accident prediction models. Dissertation Helsinki University of Technology, Espoo.
C 18862 /82 / ITRD E205332
Espoo, Technical Research Centre of Finland VTT, 1995, 104 + 42 p., 65 ref.; VTT Publication ; No. 233 - ISSN 1235-0621 / ISBN 951-38-4771-3
|Samenvatting||The purpose of the work was to study the safety of main road junctions with the help of accident prediction models. The districts of the Finnish Road Administration undertook an inventory of all at level junctions of major roads in the summer of 1988. The inventory included a total of 2,700 junctions. The number of police-reported accidents and of their victims in 1983-1987 were studied with accident prediction models as a commission from the Road Administration. The variation in the number of accidents was explained with traffic volumes and the variables collected in the junction inventory. Only the junctions with no major changes within the study period were included in the models. We made separate models for 915 three-arm and 847 four-arm junctions. The work was based on the theory of generalised linear models. The goal was to estimate in a reliable manner the expected value of the number of accidents. The actual number of accidents was assumed to vary around the expected values according to Poisson or negative binomial distributions. Separate models were made for injury accidents involving motor vehicles only, injury accidents involving unprotected road users, all injury accidents, the number of accident victims, single accidents, crossing accidents, turning accidents, and rear-end accidents. The most important variables were those describing the magnitude and distribution of motor vehicle volumes. The number of accidents and their victims were generally found to be proportional to the total number of vehicles passing through the junction. The risk of accidents seems to increase as the traffic share of the minor road increases. Several variables describing the road environment were also in the models. A new method for determining the soundness of accident prediction models was developed during the work. The method is based on estimating the amount of systematic variation in the accident data, i.e. the variation that can be explained, and separating from it the variation due to pure randomness. Accident models explained most of the systematic variation in the number of accidents. The systematic variation, however, generally accounts for just 30% of the total variation in the number of accidents. A method is described of obtaining a reliable estimate for the expected number of accidents at a single junction based on the accident models and the observed number of accidents in the past.|
|Full-text||Beschikbaar Niet beschikbaar, klik om contact op te nemen voor een digitalisatie verzoek|
|Suggestie?||Neem contact op met de SWOV bibliotheek voor uw opmerkingen|