Road alignment is assumed to influence traffic flow performance (level-of-service) as well as accident outcome. In the cause and effect relationships used by the Swedish Road Administration (SRA), sight distance is used as a measure of alignment standard and predicted accident rate in different road environments is adjusted with regard to sight distance. Since the sight distance is hard to measure, it is desirable to find an alternative to this variable. By compiling an analysis register, it has been possible to investigate the relationship between alignment and accidents on links. The alignment of each road link is described horizontally as the sum of successive absolute angle changes per kilometer and vertically as the sum of absolute changes in height. The alignment measures have been used for classification: three levels for horizontal curvature and three for vertical. Three different measures regarding accidents on links have been studied. Horizontal alignment turned out to have minor influence on accident outcome, at least for two-lane roads with speed limit 90. In most road environments the accident measures had higher values for links in the group ">/= 30 m/km" compared to the other two levels. A close co-variation was found between total casualty rate and accident rate for head-on and overtaking accidents, for both horizontal and vertical alignment. For some road environments it was possible to use 10 levels for the alignment instead of 3. Again it was found that it is mainly the vertical alignment that influences accident outcome and especially at high values for this alignment measure.
Abstract