J. Ambros, J. Sedoník, Z. Křivánková
In order to conduct network screening (hotspot identification), safety performance functions (also known as accident prediction models) are required. However, their development is demanding, since they require knowledge of traffic volume data for all evaluated segments and intersections. In addition, the whole screening process as well as its output is separated into segments and intersections, which may not be the most practical step from the perspective of a road agency. In this regard, using a number of intersections per segment length is a potential simplification which allows omitting separate modelling for intersections; however, its performance has been rarely tested. The aim of the paper is to follow the original UK application of the concept, using it in different geographical conditions and also adding an assessment of the method consistency, which is important for the quality of network screening. The method was found feasible: predictions from a simplified model were closely correlated with predictions based on a combination of segment and intersection models, and consistency in terms of overlapping between two rankings of the final segment lists was also sufficient. The simplified approach may thus increase the efficiency of network screening and enable wider practical application for Czech regional agencies.
Keywords: road safety; network screening; segments; intersections; regions