Danish road authorities have since the 1970s been handling hazardous road locations - i.e. black spots – with great success. However, recent developments have undermined this approach as present-day accidents are far more widely scattered over the road network than in the 1970s, 1980s and 1990s. This situation is due to generally significant improvements in traffic safety, which have reduced the number of accidents since the 1970s, and due to the fact that the reporting rate of accidents has fallen. 10 years ago almost 20% of the persons injured in road accidents were recorded by the police and included in the official accident recording system in Denmark, whereas today only about 10 % of the injuries are recorded by the police. Consequently, many municipalities are faced by the paradox that they are unable to identify any black spots on the basis of accident recordings in rural areas even though more than 2/3 of the persons killed in road accidents are killed on rural roads. In order for the road authorities to continue to spend traffic safety budget money most effectively there is a requirement for workable non-accident based approaches. A PhD project at Aalborg University in Denmark is addressing this problem in seeking to develop a workable method based on road characteristics to identify hazardous road locations in the secondary rural road network in Denmark. This approach will leave road administrations independent of accident recordings in their site specific road safety work. As part of this PhD project the influence of road characteristics on accident density on secondary rural roads in Denmark is analyzed. The analysis isperformed on 180 km rural two-lane roads in the municipality of Aalborg. The relationship between road characteristics and accident density is modeled using Poisson regression. As road characteristics AADT, carriageway width, shoulder width, presence of hard shoulder, intersection density, road access density, presence of bicycle path and the use of centerlines and lines along the shoulders are used. As accident data records from the official accident database in the years 2003-2011 are used. Poisson regression is used to analyze data rather than the negative binomial regression, since the Poisson model fits data better and data is not significantly over dispersed. The results show that AADT, intersection density and road access density have significant influence on accident density on secondary rural two-way roads. Furthermore, the results indicate that carriageway width and the extent of road markings – no road markings, only centerline, or both centerline and lines along the shoulders – also affect the accident density. (Author/publisher)
Samenvatting