Road accidents and rainfall in a large Australian city
Introduction
There is a perception in the community that precipitation is a road traffic hazard. A number of studies show that precipitation in the form of rain and snow generally results in more accidents compared with dry conditions (Codling, 1974, Satterthwaite, 1976, Sherretz and Farhar, 1978, Brodsky and Hakkert, 1988, Fridstrøm et al., 1995, Levine et al., 1995, Changnon, 1996, Andreescu and Frost, 1998, Edwards, 1999, Eisenberg, 2004). An early study for Melbourne, Australia by Foldvary and Ashton (1962) found that 20% of accidents were rain-related. Palutikof (1991) observed that 16% of road accidents in built-up areas of Great Britain occurred when it was raining. Hit-object collisions and those involving multiple vehicles that are associated with lane-changing actions are more likely on wet roads (Golob and Recker, 2003). A summary of many of these studies is given in Eisenberg (2004). The findings in them are often difficult to compare since they are based on different accident or injury types, different time periods within a day and may or may not involve an adjustment for traffic volume. Some are based on daily fatal or serious accidents, while others include damage-only accidents. For instance, Keay and Simmonds (2005) found in Melbourne, Australia, that rainfall increased the volume-normalised count for injury accidents by 2.4, 1.9 and 5.2% relative to the daily, daytime and nighttime dry mean, respectively. Furthermore, there is some evidence that wet or snowy weather, particularly if coupled with severe storms, can deter motorists from venturing onto the road, i.e. there is a reduction in traffic volume (Knapp and Smithson, 2000). This reduction is generally of the order of a few percent (Codling, 1974, Hassan and Barker, 1999, Keay and Simmonds, 2005). The avoidance of potentially hazardous conditions may be due to a self-assessment by the road user or road weather alerts broadcast via the media (Khattak and DePalma, 1997, Hansen et al., 2001). There is also some evidence that the sudden onset of rainfall or snowfall may increase the accident count to a greater extent than other wet or snowy days. For instance, the first snowfall of winter (Fridstrøm et al., 1995) or the first wet day of a storm (Levine et al., 1995). Other studies (Brodsky and Hakkert, 1988, Eisenberg, 2004) have found that rainfall has an enhanced effect on the accident count after a dry spell, due to an accumulation of road grime creating a slippery road surface. Ivey et al. (1981) observed that the urban wet accident rate increased with the proportion of the time that the road was wet. Estimates of the relative risk of an accident in wet weather (Haghighi-Talab, 1973, Brodsky and Hakkert, 1988, Andrey et al., 2003) have emphasised the additional hazards associated with rainfall. A revealing aspect of rain-related risk was presented by Andrey and Yagar (1993) who found that the accident risk returned to normal as soon as the rainfall had ended, despite the lingering effects of wet road conditions.
This study examines the accidents in Melbourne, the capital city of the state of Victoria, Australia (37.8°S, 145.0°E) and its metropolitan area (MMA), with a population of 3.5 million (June 2002) and covering an area of 8800 km2. We undertake a comprehensive investigation of several aspects of the influence of rainfall on road accidents in the MMA and will look at the differences between dry and wet days (and daytime and nighttime periods) as well as the effect of the amount of rainfall, i.e. rain classes. A measure of the relative risk of an accident in wet conditions will be examined and we will consider the effect of extended periods without rainfall, i.e. dry spells.
The data and methods are described in Sections 2 Data, 3 Methods. We will present the results with a discussion for the response of wet versus dry days (rain effect) and rainfall amount (rain class effect) in Section 4.1, the relative risk of an accident in Section 4.2 and the impact of dry spells in Section 4.3. Finally, our conclusions are given in Section 5.
Section snippets
Data
We make use of the datasets for the 16-year period from 1987 to 2002. These comprise the traffic volume and road accidents datasets, which were obtained from VicRoads, a state government authority responsible for managing Victoria's major road network and the rainfall dataset for the central Melbourne weather station was provided by the Australian Bureau of Meteorology. The traffic volume data was recorded at a site on the Monash Freeway, a major southeastern arterial road and augmented with
Methods
We employ a range of statistical techniques using the Minitab statistical software package (2000). In most analyses we consider the volume-normalised count (VNC) defined as:where C is the accident count and V is the traffic volume normalised to a mean of unity. With this approach the level of V as well as any day of the week (DOW) variation of V with C is incorporated. Our use of the traffic volume allows the adjustment of the accident count for the exposure of a vehicle to others on the
Rain (class) effects
The response of the accident count (volume-normalised) to rainfall for whole days is presented in Fig. 3. The rain class effect is depicted graphically with the rain effect annotated on each panel. The majority of effects are significant at the 5% level (solid circles). For all days in 1987–1991 there are 7.7 more accidents (volume-normalised) on wet days compared with dry days—a number greater by 16.7% of the dry day mean. The rain effect is larger in the second (wettest) epoch (19.0%) and
Conclusions
We have analysed the impact of rainfall on road accidents from several viewpoints and found that the effect of rainfall is multifaceted. A subdivision into three epochs was made due to a large non-linear trend over 1987–1996. In this paper we have revealed a number of associations between rainfall characteristics and road accident incidence.
There is an association of more accidents with rainfall for all epochs and subdivisions of a day. We found 1992–1996 to be the wettest of the epochs and
Acknowledgements
The authors would like to express their appreciation to Julie Maley from VicRoads for her assistance in providing the traffic volume and road accident databases. They are also grateful to Shoni Dawkins of the Australia Bureau of Meteorology for providing rainfall data over 1997–2002.
References (28)
- et al.
A temporal analysis of rain-related crash risk
Accid. Anal. Prev.
(1993) - et al.
Risk of a road accident in rainy weather
Accid. Anal. Prev.
(1988) The mixed effects of precipitation on traffic crashes
Accid. Anal. Prev.
(2004)- et al.
Measuring the contribution of randomness, exposure, weather, and daylight to the variation in road accident counts
Accid. Anal. Prev.
(1995) An investigation into the relationship between rainfall and road accident frequencies in two cities
Accid. Anal. Prev.
(1973)- et al.
Predicting wet weather accidents
Accid. Anal. Prev.
(1981) - et al.
The association of rainfall and other weather variables with road traffic volume in Melbourne, Australia
Accid. Anal. Prev.
(2005) - et al.
The impact of adverse weather conditions on the propensity to change travel decisions: a survey of Brussels commuters
Transport. Res. Part A Policy Pract.
(1997) - et al.
Daily fluctuations in Honolulu motor vehicle accidents
Accid. Anal. Prev.
(1995) An assessment of seasonal and weather effects on the frequency of road accidents in California
Accid. Anal. Prev.
(1976)
Weather and traffic accidents in Montreal, Canada
Clim. Res.
Weather as a chronic hazard for road transportation in Canadian cities
Nat. Hazards
Effects of summer precipitation on urban transportation
Clim. Change
Cited by (79)
A study on road accident prediction and contributing factors using explainable machine learning models: analysis and performance
2023, Transportation Research Interdisciplinary PerspectivesWhen left is ‘right’! The impact of driving-side practice on road fatalities in Africa
2021, Transport PolicyAn exploratory analysis of factors associated with traffic crashes severity in Cartagena, Colombia
2020, Accident Analysis and PreventionHourly associations between weather factors and traffic crashes: Non-linear and lag effects
2019, Analytic Methods in Accident ResearchCitation Excerpt :In particular, the maximum RRs of rainfall (lag 0) are 1.19 (95% CI: 1.04–1.35), 1.36 (95% CI: 1.20–1.55), 1.34 (95% CI:1.16, 1.55) respectively for rainfall of 0.5–2.0 mm/h, 2.0–5.0 mm/h and over 5.0 mm/h. This finding is consistent to that of previous studies (Brodsky and Hakkert, 1988; Keay and Simmonds, 2006). It could be attributed to the reduction in visibility, and generation of water film on the pavement surface.