The Effect of Geometric Elements on Interurban Accident Rates

This paper presents the results of two separate studies into the relationship between accident rates and road design elements. Both studies used the Irish National Road Database which contains detailed information on each uniform section of the entire National Road Network including traffic volumes and accidents. These studies were the first research uses of this database. They enabled the calculation of reliable interurban accident rates for different types of road in Ireland which were not previously available. Other objectives included the identification of the road design factors which best contributed to safety and an examination of the reliability of the accident reporting system in Ireland. The accident reporting process in Ireland is initially examined. Since the National Road Database had not previously been subjected to any detailed scientific analysis, it was found that a considerable data cleaning effort was required. However, the most substantial errors arise due to the underreporting of accidents, a problem which is encountered internationally. Comparisons of the unadjusted accident rates for each road type with the equivalent adjusted rates are used to emphasize the large differences which can occur due to the underreporting of injury accidents. Differences between the derived accident rates for the two studies (which covered different time periods) highlight the necessity of including vehicle kilometers, accident totals and road lengths in any such comparisons. The effects of these are considered and also the influence of traffic volume on accident rates. The principal conclusion is the high accident rates for undivided two-lane roads. However, three-lane roads (roads with a climbing lane) had the highest fatal accident rate. Divided roads had substantially lower accident rates than undivided roads while motorways were at least three times safer than dual carriageways. The National Road Database is also used to determine the influence of a number of road environment variables on road safety, paying particular attention to the road cross sections recommended in road design standards. A primary aim was to rank the road variables in terms of 'importance' in accident occurrence, and therefore to identify priority areas for achieving accident reductions. As this study did not set out to create a predictive model, multiple linear regression was used to model the data. The analysis was limited to two lane undivided interurban road sections. Comparisons are made with a similar study carried out in 1976 in order to investigate whether the relative effects of geometric factors have varied over time. A detailed examination was also carried out into the impacts of volume and lane and shoulder width on accidents on undivided two lane roads. Apart from vehicle kilometers of travel, lane and shoulder width and the number of roadside developments were shown to have the greatest influence on the number of accidents on undivided two-lane roads. A lane width from 3.50 to 3.75 m was optimal while the safest shoulder width was in the range 2.5 m to 3.0 m.