Constructing proper shoulders may improve road safety on two-lane roads. Previous research reported crash reductions following shoulder widening. This study aimed to examine the relationship between shoulder characteristics and crash occurrences on two-lane rural roads in Israel. The study database combined information on crash numbers, traffic volumes and road infrastructure characteristics of 3594 road sections. To examine a relationship between shoulder characteristics and crashes, given other road characteristics, two types of statistical models were developed: case-control and negative-binomial regression models, for several crash types. We found that the impacts of shoulder width and other road characteristics on crashes were generally consistent across various models and crash types, where a non-monotonous link between the shoulder width and crashes was typically observed. For various crash types, the models showed an increase in crash risk with an initial extension of the total shoulder, up to 2.2 m, and a consequent decrease in crashes with a further shoulder widening, over 2.2 m, by 2-6% and 1-4%, respectively, for each 0.1 m of shoulder extension. An increase in the width of unpaved shoulders, over 0.9 m, was associated with increased crash risk, in injury and total crashes, by 5% for each 0.1 m of shoulder extension. Lowest crash risks were found for total shoulder widths of about 3 m or more, but also for narrow total shoulders, below 1 m. Conversely, medium total shoulders, of 1.8-2.4 m in width, and unpaved shoulders of over 1 m, were associated with an increase in crash risk and, hence, are not recommended for use. The tools developed in the study may assist in decision-making during the design stages of a new road or upgrading existing road sections, on two-lane local roads.
Keywords: Case-control; Safety modelling; Shoulders; Two-lane Roads.
Copyright © 2018 Elsevier Ltd. All rights reserved.