Compared to motorways, rural roads in Germany have a large total length and account for a high proportion of the network. In addition, there is a great variety of roadside planting, ranging from rows of trees on one side or both sides, groups of trees, avenues, wooded areas, individual trees, shrubs, field hedges to fields and meadows. Rural road safety is closely related to planting in terms of accident consequences. Collisions with a tree next to the roadway are characterised by a particularly high accident severity. The aim of this work is to determine the characteristics influencing accidents on the basis of empirical models using a Bayesian inference approach, with special consideration of roadside planting, and to develop and validate the methodological procedure for this empirical Bayesian model in the process. The procedure developed for the creation of the accident models contributes to extending the procedures and methods of safety investigations in the road sector. Many of the safety investigations to date are based on monocausal analysis methods. It has been shown that scientific proof based on monocausal accident analyses is only insufficiently possible. The gaps in our knowledge of the correlations between the different types and other characteristics of roadside planting and accident occurrence formed the focus of the study. The aim was to derive further qualitative statements and road safety estimates as well as quantitative estimates of accident occurrence depending on possible influencing characteristics. With the developed accident prediction models, more accurate risk assessments of rural road sections are to be carried out. The aim is to improve road safety by implementing targeted, effective measures while protecting both road users and roadside planting. The analysis of the individual data of the official road traffic accident statistics showed that the accident severity is 5 times higher in collisions with a fixed obstacle, 8 times higher in collisions with a tree and 3 times higher in collisions with a crash barrier than without such impacts. The evaluation of detailed data from road traffic accidents (In-Depth Study) provided the latest findings on accident causation and its boundary conditions. This evaluation revealed numerous significant correlations, such as between the respective accident sequences and the depth of vehicle deformation as well as between the diameter of the tree and the highest injury severity of the vehicle occupants. Based on an extensive and detailed data set with over 2,600 lane-related sections and almost 3,000 km total length, the focus of the investigation was on single carriageway with two-lanes outside of network junctions. The differentiated accident models showed that the avenue and the forest area are particularly accident-prone plantation types. Tree rows, forest mantles and individual trees are also conspicuous. It was shown that there is a clear correlation between the mileage-related frequency of officially recorded run-off-road accidents and roadside planting. Decisive and significant design elements influencing accidents caused by driving off the carriageway are, in particular, singularities (curve radius less than 200 m including minimum length of 50 m), but also curve radii of 200 to 400 m as well as relative alignments of neighbouring elements of the site plan in the „area to be avoided“ or „usable area“ according to the guidelines for the Design of Rural Roads 2012. With regard to collisions with a tree, it has been shown that the accident frequency increases significantly as the proportion of roadside planting of avenues and wooded areas increases. Row of trees as well as the combination of row of trees and avenue lead to medium to very high increases in accident frequency. These results form the basis for the revision of technical regulations, such as the manual for the assessment of road safety, the recommendations „Roadside Trees“ and for the recommendations of special solutions for tree and object protection on rural roads. (Author/publisher)
Samenvatting