Research on new accident cluster analyses techniques, was presented at the 2006 ICTCT conference in Minsk. The methodology allows computing spatial concentrations of road accidents based on distances along the road network, including distances between accidents on different, but intersecting roads. A dangerousness index is computed for road segments indicating a weighted number of traffic accidents in the neighbourhood. In the following stage of the research, the statistical significance of the clusters was tested using a Monte Carlo simulation. This paper focuses on the bottlenecks encountered when trying to implement this approach in different conditions: for the research of the accident concentrations in Qing Dao (China), and for the development of tools for the Flemish Road Administration (Belgium). First, the state of progress of the research is briefly presented. The analyses were performed on the accidents in Brussels. Then, problems are presented, encountered when trying to implement the measure in a very different environment (Qing Dao). Second, the progress is reviewed, made since the start of the development of a GIS for traffic accidents in the Flemish Road Administration in 1994. The efforts shifted from the development of a route network database, towards tool development for accident location, and finally towards tools for accident clustering and prioritisation procedures for black zones. The tool developed for accident location on secondary roads in the Flemish Road Administration, used an address matching technique developed in a previous research project on accidents in Brussels. The methods was improved with a quality control system developed in collaboration between the regional (Flemish) and the provincial administration. The further research in turn benefited insights gained during the tools development, such as for snapping of accidents to the network. The clustering technique currently used by the Flemish Road Administration, is based on Euclidian distances along the route, and identifies black zones based on the number of accidents occurring within a distance of 1 km from each given accident. Accidents on intersections are counted for each route passing through the intersection. A number of attributes are then calculated per black zone, in order to prioritise in which zones intervention in the infrastructure is most needed. When trying to implement the network based clustering technique, a number of practical problems are encountered. Most of them are related to the detail of the network data, and their influence on the connectivity between points on the network, such as the modelling of separated lanes, of specific types of intersections,... There are some similarities between problems encountered for the implementation of the method in Qing Dao, and in the Flemish regional administration. The paper illustrates the trade-off between the accuracy of the accident clustering –and the associated network data quality requirements– and the environment where it is to be applied. (Author/publisher) This publication may be accessed by Internet users at: http://www.ictct.org/workshop.php?workshop_nr=25
Samenvatting