Linda Mountain

ACCIDENT PREDICTION MODELS FOR ROADS WITH MINOR JUNCTIONS

The purpose of this study was to develop and validate a method for predicting expected accidents on main roads with minor junctions where traffic counts on the minor approaches are not available. The study was based on data for some 3800 km of highway in the U.K. including more than 5000 minor junctions. The highways consisted of both single and dual-carriageway roads in urban and rural areas. Generalized linear modelling was used to develop regression estimates of expected accidents for six highway categories and an empirical Bayes procedure was used to improve these estimates by combining them with accident counts. Accidents on highway sections were shown to be a non-linear function of exposure and minor junction frequency. For the purposes of estimating expected accidents, while the regression model estimates were shown to be preferable to accident counts, the best results were obtained using the empirical Bayes method. The latter was the only method that produced unbiased estimates of expected accidents for high-risk sites.

THE INFLUENCE OF TREND ON ESTIMATES OF ACCIDENTS AT JUNCTIONS

While reliable estimates of expected accidents can be achieved by combining observed accidents and accident model predictions using an empirical Bayes approach, there are a number of obstacles to the widespread adoption of the method. This paper concentrates on problems associated with the available predictive models. Of particular concern is the effect on model predictions of accident trends over time resulting from, for instance, traffic growth or national road safety programmes. Since accident models invariably include traffic flow as an explanatory variable, the effects of flow changes can be included provided that account is taken of the nonlinear relationship between accidents and exposure. It is, however, common to assume that accident risk per unit of exposure is constant over time, whereas national data imply that accident risk is declining. In addition, there is a need, in practice, to rank and evaluate remedial sites in terms of the specific accident types or severities which might be targeted by treatment (for example, wet road accidents in the case of surface treatment). This then raises the question of whether the proportions of accidents of various types varies over time or with traffic flow and site characteristics. Generalized linear modelling was used to develop regression estimates of expected junction accidents (both in total and disaggregated by severity, road surface condition and lighting condition) which allow for the possibility of accident risk varying over time. Accident risk at the sample of some 500 junctions was shown to be declining annually by an average of 6%, with no significant difference in the value of trend between accident types. The factors which affected the proportions of accidents of various types included the method of junction control, speed limit and traffic flow.

Methodology for fitting and updating predictive accident models with trend

Reliable predictive accident models (PAMs) (also referred to as Safety Performance Functions (SPFs)) have a variety of important uses in traffic safety research and practice. They are used to help identify sites in need of remedial treatment, in the design of transport schemes to assess safety implications, and to estimate the effectiveness of remedial treatments. The PAMs currently in use in the UK are now quite old; the data used in their development was gathered up to 30 years ago. Many changes have occurred over that period in road and vehicle design, in road safety campaigns and legislation, and the national accident rate has fallen substantially. It seems unlikely that these ageing models can be relied upon to provide accurate and reliable predictions of accident frequencies on the roads today. This paper addresses a number of methodological issues that arise in seeking practical and efficient ways to update PAMs, whether by re-calibration or by re-fitting. Models for accidents on rural single carriageway roads have been chosen to illustrate these issues, including the choice of distributional assumption for overdispersion, the choice of goodness of fit measures, questions of independence between observations in different years, and between links on the same scheme, the estimation of trends in the models, the uncertainty of predictions, as well as considerations about the most efficient and convenient ways to fit the required models.

Updating outdated predictive accident models

Reliable predictive accident models (PAMs) (also referred to as safety performance functions (SPFs)) are essential to design and maintain safe road networks however, ongoing changes in road and vehicle design coupled with road safety initiatives, mean that these models can quickly become dated. Unfortunately, because the fitting of sophisticated PAMs including a wide range of explanatory variables is not a trivial task, available models tend to be based on data collected many years ago and seem unlikely to give reliable estimates of current accidents. Large, expensive studies to produce new models are likely to be, at best, only a temporary solution. This paper thus seeks to develop a practical and efficient methodology to allow currently available PAMs to be updated to give unbiased estimates of accident frequencies at any point in time. Two principal issues are examined: the extent to which the temporal transferability of predictive accident models varies with model complexity; and the practicality and efficiency of two alternative updating strategies. The models used to illustrate these issues are the suites of models developed for rural dual and single carriageway roads in the UK. These are widely used in several software packages in spite of being based on data collected during the 1980s and early 1990s. It was found that increased model complexity by no means ensures better temporal transferability and that calibration of the models using a scale factor can be a practical alternative to fitting new models.

Updating predictive accident models of modern rural single carriageway A-roads

Abstract Reliable predictive accident models (PAMs) are essential to design and maintain safe road networks, and yet the models most commonly used in the UK were derived using data collected 20 to 30 years ago. Given that the national personal injury accident total fell by some 30% in the last 25 years, while road traffic increased by over 60%, significant errors in scheme appraisal and evaluation based on the models currently in use seem inevitable. In this paper, the temporal transferability of PAMs for modern rural single carriageway A-roads is investigated, and their predictive performance is evaluated against a recent data set. Despite the age of these models, the PAMs for predicting the total accidents provide a remarkably good fit to recent data and these are more accurate than models where accidents are disaggregated by type. The performance of the models can be improved by calibrating them against recent data.