Travelling Speed and the Risk of Crash relationship in Adelaide South Australia

The main aim of this project was to establish a mathematical curve that defines the relationship between free travelling speed and the risk of involvement in a casualty crash, for sober drivers in an urban setting. Data collected in a case control study (Kloeden, McLean, Moore and Ponte, 1997) were reanalysed using logistic regression modelling. The speeds of passenger vehicles involved in casualty crashes were compared with the speeds of passenger vehicles not involved in crashes but travelling in the same direction, at the same location, time of day, day of week, and time of year.

Both absolute travelling speeds and speed differences were used in the modelling process and allowance was made for uncertainties in the reconstructed case speeds.

An absolute speed curve was found to provide a good fit for speeds between 60 and 80 km/h whereby the risk of casualty crash involvement approximately doubled for each 5 km/h increase in travelling speed. Although the data were relatively sparse outside this speed range, we assumed that the curve could be used for speeds down to 26 km/h and for speeds above 80 km/h in our hypothetical analysis, since the curve modelled the available data and its general shape (exponentiated second order polynomial) is not unexpected given the physics of road crashes and injury biomechanics.

Such considerations also indicate that speed is a risk factor in and of itself. That is, the observed differences in crash risk between vehicles travelling at different speeds is primarily due to the actual travelling speeds and not other factors such as the type of drivers who choose to travel at different speeds or with the variance in travelling speeds.

A speed difference risk curve was also fitted to the data and found to produce comparable results to the absolute speed risk curve.

The secondary aim of the project was to examine the effect of hypothetical speed reductions on this set of crashes and urban crashes in general, using the derived mathematical risk curves, to allow some insight to be gained into the possible effects of changing the speed behaviour of urban drivers (although a number of unproved assumptions, as stated, had to be made to do this).

It was estimated that illegal speeding in Adelaide 60 km/h zones accounts for around 25 per cent of all casualty crashes in those zones. That is, if we could reduce the maximum speed of all vehicles in Adelaide 60 km/h speed zones to 60 km/h, we would expect casualty crashes in those zones to fall by around 25 per cent.

Moreover, nearly 60 per cent of the benefit of eliminating speeding would be achieved by eliminating speeding among those travelling between 61 and 75 km/h. This is because there are many more drivers who travel in this speed range than at faster speeds. Their relative risk of casualty crash involvement is lower than those travelling above 75 km/h, but their contribution to the total number of casualty crashes is the product of the number of these drivers and their relative risk of involvement in a casualty crash.

Examination of the estimated hypothetical effects of slowing all vehicles down by the same amount indicate that very small reductions in travelling speed (even 1 km/h or less) can be expected to have a meaningful impact on casualty crash numbers.

Estimates were also made for a hypothetical reduction in the general urban area speed limit from 60 km/h down to 50 km/h using two sets of assumptions. Casualty crashes in these speed zones would be expected to drop by around 21 per cent using a speed fine avoidance method and by 28 per cent using a speed distribution movement method. While similar reductions on local streets would be expected from a reduction in the speed limit to 50 km/h, if the speed limit reduction was limited to local streets, the relatively small proportion of casualty crashes on local streets means that the effect on all casualty crashes in the metropolitan area would be much smaller than a change in the general urban area speed limit.

Crash reduction studies of cars

Early research has provided a mix of results. Comprehensive reviews incorporating meta-analysis by Elvik (1996) and Koornstra et al. (1979) have systematised these results. Elvik showed that studies of DRL in vehicle fleets provided larger and more consistent effects than studies of traffic systems, and that outcomes are sensitive to the type of measures used. Koornstra et al. carried this analysis a step further by statistically excluding factors such as number of vehicles already using DRL at the start of a trial to estimate the intrinsic effects of DRL, ie the effects which would occur when use rate is increased from zero to 100% of vehicles. The effect of this re-analysis was to make many studies which were previously non-significant significant, and show that by far the majority of studies found a reduction in multiple-party daytime crashes associated with DRL. Koornstra et al. proposed a set of relationships relating the extent of crash reductions with DRL to latitude, based on the longer hours of twilight in the higher latitudes. It should be noted that many later studies do not fit this model well.

Studies that have been conducted since the reports of Elvik (1996) and Koorsntra et al. (1997) have generally confirmed the reduction in crash rates associated with DRL, although the size of the reduction varies. Tofflemire and Whitehead (1997) compared crash rates in the same year for Canadian cars with and without DRL, eliminating factors such as weather, economic climate, and enforcement as a possible source of differences. They found an overall reduction of 5.3%, principally due to a reduction in crashes involving vehicles travelling in opposite directions.

A North America study by the National Highway Traffic Safety Administration (NHTSA, 2000) reinforced the sensitivity of outcomes to analysis methods, but showed an estimated effect of DRL was a reduction of 7% in multiple-party daytime crashes, which was statistically significant. Bergkvist (2001) reports upon a study conducted for General Motors Corporation by an independent consulting firm. The study involved a comparison of the crash rates of specific GM, Volvo, Saab and Volkswagen vehicles before and immediately after DRL became standard equipment on these models. The results suggest a reduction in the incidence of target vehicle-vehicle crashes in excess of 5% and a reduction in vehicle-pedestrian collisions of approximately 9%. Farmer and Williams (2002) compared makes and models which included DRL as a standard feature with the same makes and models purchased before this was the case. The reduction in daytime crashes was 3.2%, which was highly significant. When considered separately, all states but one showed a reduction in crashes for the DRL vehicles. However, only in Texas was the reduction large enough to be statistically significant. Lassarre (2002) reports a trial involving a campaign to encourage voluntary use of headlights during the day. Crash rates in the area where the campaign was run were compared with crash rates in neighbouring areas over the same period, and showed a reduction of 58.7% for fatal crashes and for serious injury crashes on major roads, but no reductions in less serious crashes and no effects on crashes on minor roads.

A study by Poole (1999) is of particular interest as it is the only fleet study carried out in Australia. DRL were fitted to 80 fleet vehicles based in metropolitan Western Australia. The crash records of the DRL equipped vehicles over a ten month period were compared with the crash records of a matched sample of vehicles not fitted with DRL. Results based on a time-to-crash analysis found that DRL fitted vehicles took more than five times longer than non-DRL fitted vehicles to be involved in a vehicle-vehicle daytime crash.

Potential Benefits and Costs of Speed Changes Rural road freeways

An increase in the speed limit to 130 km/h on rural freeways would save each car 8.4 minutes and each truck 13.8 minutes per 100 km, but would increase the number of fatal crashes by 2.8 per year per 100 km of freeway. Casualty crash costs would increase by 89%, vehicle operating costs would increase by 7% and time costs would decrease by 17%. There would be a net cost increase of $2.35 million per year per 100 km of road, provided it is appropriate to value leisure travel time savings and to value the road trauma increases by the ‘human capital’ approach. If the leisure time savings are not valued, then the net impact would be an economic cost of $7.6 million per year per 100 km of freeway. If road trauma is valued by society’s ‘willingness to pay’ to prevent it, the net cost would be $10.5 million per year per 100 km. Since these alternative valuations of leisure travel time and road trauma are central to the estimated economic output of the increased speed limit on rural freeways, the implications of their choice in making policy decisions needs to be considered carefully.

However, the analysis does indicate that the negative economic impacts of the increased speed limit on rural freeways could be overcome, and even made positive, if trucks were limited on such roads to 100 km/h. A further alternative would be a variable speed limit system, whereby the speed limit is reduced to 100 km/h for cars and light commercial vehicles under adverse road conditions (such as at night or other adverse condition approximately doubling the crash risk for about 20% of the traffic), and is fixed at 100 km/h for trucks at all times. If the increased speed limit under good conditions was no more than 120 km/h, the increase in road trauma would be minimal. This variable speed limit system would still result, however, in an increase in fatal crashes of 0.2 per year per 100 km of freeway, due to the increase in speed limit for 80% of the traffic, albeit during safer daytime conditions. This system would increase casualty crash costs by 7%, increase vehicle operating costs by 1% and reduce time costs by 4%.

Nine different models of bicycle helmets testing

Nine different models of bicycle helmets were tested, with 6 helmets of each model submitted to the laboratory for testing to a combination of the AS/NZS2063 and Snell B95 standards tests. All the tests were performed after conditioning in the ambient environment. The following tests carried out and the findings were as follows:

• The test area coverage required by the two standards was similar for all the sizes of helmets, but the area of the head covered by the Snell certified helmets was greater. The more generous coverage of the Snell B95 helmets was also indicated by the slightly higher weights on average for these helmets.
• The order of testing followed AS/NZS 2063-1996, as this was the worst case from the two standards. The AS/NZS 2063-1996 standard requires the impact testing to be carried out before the retention system strength test on the same helmets. This effectively increases the severity of the retention system test as more deformation of the restraint system is likely to occur.
• The Snell B95 dynamic helmet stability test is significantly more demanding of the helmet design than the AS/NZS 2063:1996 static test. The Snell test takes the retention system of the helmet near to its mechanical limits.
• The flat anvil impact energy attenuation tests are included in both the AS/NZS 2036- 1996 and the Snell B95 standards and were made on the front, rear, and top of the helmets. For these tests both group of helmets returned similar results. Further, it is clear that the higher energy of the flat anvil tests to the Snell standard generate a higher acceleration of the headform in this test for both groups of helmets, by about 25%. The safety margin built into the helmets to both standards easily deals with this increase in drop energy. Therefore the helmets to both standards offer the same impact attenuation protection and this is confirmed by the similarity in liner densities found for the range of models tested.
• The kerbstone and hemispherical anvil impact energy attenuation tests were only performed within the Snell B95 Standard. For both these tests the AS/NZS 2036-1996 approved helmets gave noticeably higher headform accelerations, with the average for these helmets exceeding the requirement of 300g. The test variability was also significantly greater for these tests on the AS/NZS 2036-1996 approved helmets. Two of the AS/NZS 2036-1996 helmets failed these tests with the kerb and hemi anvils.
• The sampled helmets certified to both standards gave very similar results in the load distribution test. If one of the AS/NZS 2036-1996 certified helmets, the Rosebank Ms 16, is removed from the test results, then the test variability is significantly reduced. This helmet still met the requirements of AS/NZS 2036-1996.
• The AS/NZS 2036-1996 certified helmets have greater retention system strength in comparison to the Snell B95 helmets. The average dynamic displacement for the Snell B95 certified helmets was close to exceeding the maximum limit of both AS/NZS 2036- 1996 and Snell B95 standards.
• The liner foam density was consistent in all the helmet sizes and both certification standards, with the exception of the Bell Stryker helmet which was higher.
• The AS/NZS 2036-1996 certified helmets are consistently of lower helmet weight across the whole range of sizes than the Snell B95 certified helmets. This is a direct measure of the larger area of coverage of the head given by the Snell standard, as the foam density was similar for all the helmets tested, except for the Bell.