« Previous | Contents | Next »
Listen
MOTORCYCLE ACCIDENTS AND CASUALTIES IN SCOTLAND 1992 - 2002
CHAPTER FIVE: CLUSTERS OF FATAL AND SERIOUS CASUALTIES FROM 2000-2002
MAAP
5.1 MAAP is TRL's accident analysis package and database software. The package has an analysis function for identifying hazardous locations where the density of accidents is higher than in other areas based on the coordinates provided for accident locations in the STATS19 information.
5.2 The cluster function operates by taking each individual accident location and measuring the distances to its nearest neighbours. The "clusters" identified can be quite large if an insensitive search distance is used. The program then lists and displays the 'worst' sites where numbers of accidents occur together at a distance less than the sensitivity distance set in the program. The clusters are ranked on the basis of their scores which are counts of the numbers of accidents within the clusters. The searches can be carried out with the values of the various accident severities given different weights or scores. Thus fatal accidents may count ten, serious five, slight one and damage only accidents a half for example. This feature enables the operator to distinguish between clusters where there are a lot of minor accidents from sites where there are a few but more severe accidents which from a cost benefit point of view may be a higher priority for investigation and safety treatment.
5.3 The sensitivity of the search can be changed such that a sensitivity of 50 metres for example, will identify a cluster of accidents where the distance between all the accidents in a cluster is less than 50 metres. This means that the utility will find extended areas in which all the accidents marked in a cluster, are located at least 50 metres apart, and not simply a circular area of radius 50metres with most accidents in it. A cluster located at a search resolution of 50 metres may extend for several hundreds of metres. The ability to change the sensitivity is useful since the background density of accidents in urban areas will be higher than in rural areas since more travel occurs on a denser road network in the built-up situation. Thus the sensitivity in rural areas can be reduced to identify hazardous stretches of road rather than discrete locations. Using an insensitive search in urban areas will tend to identify very large areas, which is not particularly useful from a road safety point of view.
5.4 The most recent three years of more serious accidents were chosen for further analysis because more detailed Police records were likely to be available for these accidents. Thus, a relatively small subset of all accidents was being searched and therefore relatively insensitive searches were used. The clustering module in MAAP is being used in this instance to find larger or extended areas of high accident occurrence rather than discrete clusters.
5.5 Initially a variety of search sensitivities was tested to identify the size and numbers of clusters obtained. From the initial runs it was realised that searching all KSI TWMV accidents would give biased results with almost all clusters being identified in urban areas. Thus it was decided that built-up and non built-up accidents were distinguished on the basis of the speed limit of the road on which the accident occurred. The definition used for built-up/non built-up was that listed in Road Accidents Scotland 2002 page 7, where accidents occurring on a road with a speed of 40mph or less are designated as "built-up". Early results from the in-depth analysis also indicated that the patterns in motorcycle accidents differed markedly between the built-up and non built-up situations, again indicating that two separate searches were necessary to obtain a representative sample of clusters from more non built-up areas.
5.6 The counts of the weights of serious accidents were given a score of 5 and fatal accidents a score of 10. This was done to obtain sites with a slight bias towards those containing motorcycle rider fatalities.
Clusters in built-up areas
5.7 The searches in built-up areas were performed with a sensitivity of 75 metres. This search parameter gave a good range of clusters which varied in count and location throughout Scotland. These sites are shown in Annex 3.
Clusters in non built-up areas
5.8 The searches in non built-up areas were performed with a sensitivity of 2000 metres. This setting was higher than that used in the built-up cluster search and this was necessary since the pattern of crash locations is much less dense in non built-up locations. Again this sensitivity gave a good range of clusters which varied in count and location throughout Scotland. These sites are also shown in Annex 3. The use of a relatively high cluster sensitivity means that those clusters identified are not so likely to be associated with specific localised features, e.g. junctions, but are more likely to be associated with routes.
« Previous | Contents | Next »