Identification of Hazardous Road Locations on the basis of Floating Car Data: Method and first results

A central aspect of road safety work relies on identification of hazardous road locations (HRL). However, the said identification is based on police-reported accidents, which show massive underreporting. Thus, in Denmark, only 14% of the serious injury accidents were reported in 2007, and the proportion is decreasing. Hence, HRL identification and enhancement are carried out more or less at random. Also, they are retrospective, i.e. accidents must occur before road safety enhancements can be made. In place of that procedure, a predictive model based on serious jerks (the derivative of deceleration) found in GPS data from driving cars is under development. Other studies have found a connection between serious jerks and conflicts. This paper focuses on a small-scale study based on a distance driven of 38,000 km and 2 million observations. It is found that to be useful for HRL identification observations, should include a clear indication of when deceleration is initiated as well as when deceleration ends. Also, to avoid erroneous results due to speed bumps etc. a measurable reduction of the driving speed has to occur within few seconds prior to the jerk. Furthermore, the speed prior to jerks has to be above a certain level to enable distinction from results involving for example the passing of kerbs and departure from driveways. However, large-scale studies are required to assess if the approach is sufficiently reliable and to establish a threshold for including jerks in the HRL identification. These studies were initiated towards the end of 2012.