Accident prediction models for urban roads

This paper describes some of the main findings from two separate studies on accident prediction models for urban junctions and urban road links described in [Uheldsmodel for bygader-Del1: Modeller for 3-og 4-benede kryds. Notat 22, The Danish Road Directorate, 1995; Uheldsmodel for bygader- Del2: Modeller for straekninger. Notat 59, The Danish Road Directorate, 1998] (Greibe and Hemdorff, 1995, 1988).The main objective for the studies was to establish simple, practicable accident models that can predict the expected number of accidents at urban junctions and road links as accurately as possible. The models can be used to identify factors affecting road safety and in relation to 'black spot' identification and network safety analysis undertaken by local road authorities.The accident prediction models are based on data from 1036 junctions and 142 km road links in urban areas. Generalised linear modelling techniques were used to relate accident frequencies to explanatory variables.The estimated accident prediction models for road links were capable of describing more than 60% of the systematic variation ('percentage-explained' value) while the models for junctions had lower values. This indicates that modelling accidents for road links is less complicated than for junctions, probably due to a more uniform accident pattern and a simpler traffic flow exposure or due to lack of adequate explanatory variables for junctions.Explanatory variables describing road design and road geometry proved to be significant for road link models but less important in junction models. The most powerful variable for all models was motor vehicle traffic flow.     

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.     

Effects on road safety of new urban arterial roads

This paper presents an evaluation of the effects on road safety of new urban arterial roads in Oslo, Norway, and a synthesis of evidence from similar studies that have evaluated the safety effects of new urban arterial roads in other cities. A before-and-after study was made of four urban arterial road projects in Oslo. The study controlled for general accident trends in Oslo and for regression-to-the-mean. A statistically non-significant reduction of 9% in the number of injury accidents was found for all four projects combined. The effects on safety of new urban arterial roads were found to vary, depending on whether a new arterial road was built, or an existing arterial road upgraded by means of lane additions and reconstruction of junctions to interchanges. New arterial roads tend to induce more traffic, which tends to offset the benefits of a lower accident rate on the new roads. The results for other cities are very consistent with those for Oslo. For a total of seven cases in which new arterial roads were built, a statistically non-significant reduction of 1% in the number of injury accidents was found. Two cases that involved lane additions and converting at-grade junctions to interchanges resulted in a mean accident reduction of 51%, which was highly significant. On the average, the nine arterial road projects from which evidence was summarised resulted in a net induced traffic of 16%, and a net reduction in accident rate (accidents per million vehicle kilometres) of 18%. These effects almost cancel each other, leading to a very small net change in the expected number of accidents.     

Measuring accident risk exposure for pedestrians in different micro-environments

Pedestrians are mainly exposed to the risk of road accident when crossing a road in urban areas. Traditionally in the road safety field, the risk of accident for pedestrian is estimated as a rate of accident involvement per unit of time spent on the road network. The objective of this research is to develop an approach of accident risk based on the concept of risk exposure used in environmental epidemiology, such as in the case of exposure to pollutants. This type of indicator would be useful for comparing the effects of urban transportation policy scenarios on pedestrian safety. The first step is to create an indicator of pedestrians’ exposure, which is based on motorised vehicles’ “concentration” by lane and also takes account of traffic speed and time spent to cross. This is applied to two specific micro-environments: junctions and mid-block locations. A model of pedestrians’ crossing behaviour along a trip is then developed, based on a hierarchical choice between junctions and mid-block locations and taking account of origin and destination, traffic characteristics and pedestrian facilities. Finally, a complete framework is produced for modelling pedestrians’ exposure in the light of their crossing behaviour. The feasibility of this approach is demonstrated on an artificial network and a first set of results is obtained from the validation of the models in observational studies.     

Estimating the number of accidents at intersections from a knowledge of the traffic flows on the approaches

The increase in the number of accidents at intersections, relative to non-junction accidents, may be explained by the basic fact that an increase in the number of vehicles on the road is generally accompanied by an increase in the number of collisions, which rise at a faster rate than single-vehicle accidents. Generally, more than 50% of the collisions occur at intersections. Data from a number of countries studied support the above statement and show that over the years, the number of intersection accidents has increased at a faster rate than other accidents. The study includes a statistical analysis of the general trends in the number of intersection accidents, their severity, and an analysis of types of accidents at intersections. A model was developed which enables the estimation of the expected number of accidents at individual intersections. It was found that vehicle exposure can be used as the basis for estimation. The exposure, denned as the number of occasions for accidents, was calculated through the sum of the products of flow at the 24 points where vehicle paths cross or merge. This measure of exposure, expressed as a traffic flow index, showed good correlation with the number of accidents. On the assumption that the number of accidents at an intersection, in a given time interval, is Poisson distributed, significance tests were made comparing the actual number of accidents with the expected according to the vehicle exposure index. Such comparisons are useful in the determination of accident “black spots”.     

Pedestrian risk decrease with pedestrian flow. A case study based on data from signalized intersections in Hamilton,Ontario

A unique database provided information on pedestrian accidents, intersection geometry and estimates of pedestrian and vehicle flows for the years 1983-1986 for approximately 300 signalized intersections in Hamilton, Ont., Canada. Pedestrian safety at semi-protected schemes, where left-turning vehicles face no opposing traffic but have potential conflicts with pedestrians, were compared with pedestrian safety at normal non-channelized signalized approaches, where right-turning vehicles have potential conflicts with pedestrians. Four different ways of estimating hourly flows for left- and right-turning vehicles were explored. Hourly flows were estimated for periods of 15 min, hours, two periods a day (a.m. and p.m.) and the 'daily' period (7 h). Parameter estimates were somewhat affected by the time period used for flow estimation. However, parameter estimates seem to be affected far more by the traffic pattern (left- or right-turning traffic), even though approaches were selected such that the situation for left- and right-turning turning traffic was similar (no opposing traffic, no advanced green or other separate phases and no channelization). Left-turning vehicles caused higher risks for pedestrians than right-turning vehicles. At low vehicular flows right turns and semi-protected left turns seemed to be equally safe for pedestrians. When risks for pedestrians were calculated as the expected number of reported pedestrian accidents per pedestrian, risk decreased with increasing pedestrian flows and increased with increasing vehicle flow. As risk decreases with increasing pedestrian flows, promoting walking will have a positive effect on pedestrian risk at signalized intersections.     

Driving performance assessment: effects of traffic accident location and alarm content

According to accident statistics for Taiwan, the two most common traffic accident locations in urban areas are roadway segments and intersections. On roadway segments, most collisions are due to drivers not noticing the status of leading vehicle. At intersections, most collisions are due to the other driver failing to obey traffic signs. Using a driving simulator equipped with a collision warning system, this study investigated driving performance at different accident locations and between different alarm contents, and identified the relationship between crash occurrences and driving performance. Thirty participants, aged 20-29 years, were recruited in this study. Driving performance measures were perception-reaction time, movement-reaction time, speed and a crash. Experimental results indicated that due to different demands for processing information under different traffic conditions, driving performance differed at the two traffic accident locations. On a roadway segment, perception-reaction time for a beep was shorter than the time for a speech message. Nevertheless, at an intersection, a speech message was a great help to drivers and, thus, perception-reaction time was effectively reduced. In addition, logistic regression analysis indicates that perception-movement time had the greatest influence on crash occurrence.     

Quantifying safety benefit of winter road maintenance: Accident frequency modeling

This research presents a modeling approach to investigate the association of the accident frequency during a snow storm event with road surface conditions, visibility and other influencing factors controlling for traffic exposure. The results have the premise to be applied for evaluating different maintenance strategies using safety as a performance measure. As part of this approach, this research introduces a road surface condition index as a surrogate measure of the commonly used friction measure to capture different road surface conditions. Data from various data sources, such as weather, road condition observations, traffic counts and accidents, are integrated and used to test three event-based models including the Negative Binomial model, the generalized NB model and the zero inflated NB model. These models are compared for their capability to explain differences in accident frequencies between individual snow storms. It was found that the generalized NB model best fits the data, and is most capable of capturing heterogeneity other than excess zeros. Among the main results, it was found that the road surface condition index was statistically significant influencing the accident occurrence. This research is the first showing the empirical relationship between safety and road surface conditions at a disaggregate level (event-based), making it feasible to quantify the safety benefits of alternative maintenance goals and methods.     

Accidents, mileage, and the exaggeration of risk

The usual interpretation of accidents per mile as a measure of risk exaggerates the apparent risk of low-mileage groups--for example, teenagers and the elderly. The assumption of a linear proportional relationship between mileage and accidents is shown not to fit obtained data. Neither would it be expected to fit hypothetical data derived from a "standard driver" or a group of equally competent drivers driving different numbers of miles. People driving low mileages tend to accumulate much of their mileage on congested city streets with two-way traffic and no restriction of access, while high-mileage drivers typically accumulate most of those miles on freeways or other divided multilane highways with limited access. Because the driving task is simpler, the accident rate per mile is much lower on freeways, and beyond a certain point, a person driving half as many miles as another would be expected to have considerably more than half as many accidents. This and other considerations lead to the suggestion that an induced exposure approach would be a more valid method of correcting accident rates for mileage.     

Logistic regression analysis of pedestrian casualty risk in passenger vehicle collisions in China

A large number of pedestrian fatalities were reported in China since the 1990s, however the exposure of pedestrians in public traffic has never been measured quantitatively using in-depth accident data. This study aimed to investigate the association between the impact speed and risk of pedestrian casualties in passenger vehicle collisions based on real-world accident cases in China. The cases were selected from a database of in-depth investigation of vehicle accidents in Changsha-IVAC. The sampling criteria were defined as (1) the accident was a frontal impact that occurred between 2003 and 2009; (2) the pedestrian age was above 14; (3) the injury according to the Abbreviated Injury Scale (AIS) was 1+; (4) the accident involved passenger cars, SUVs, or MPVs; and (5) the vehicle impact speed can be determined. The selected IVAC data set, which included 104 pedestrian accident cases, was weighted based on the national traffic accident data. The logistical regression models of the risks for pedestrian fatalities and AIS 3+ injuries were developed in terms of vehicle impact speed using the unweighted and weighted data sets. A multiple logistic regression model on the risk of pedestrian AIS 3+ injury was developed considering the age and impact speed as two variables. It was found that the risk of pedestrian fatality is 26% at 50 km/h, 50% at 58 km/h, and 82% at 70 km/h. At an impact speed of 80 km/h, the pedestrian rarely survives. The weighted risk curves indicated that the risks of pedestrian fatality and injury in China were higher than that in other high-income countries, whereas the risks of pedestrian casualty was lower than in these countries 30 years ago. The findings could have a contribution to better understanding of the exposures of pedestrians in urban traffic in China, and provide background knowledge for the development of strategies for pedestrian protection.     

Modeling secondary accidents identified by traffic shock waves

The high potential for occurrence and the negative consequences of secondary accidents make them an issue of great concern affecting freeway safety. Using accident records from a three-year period together with California interstate freeway loop data, a dynamic method for more accurate classification based on the traffic shock wave detecting method was used to identify secondary accidents. Spatio-temporal gaps between the primary and secondary accident were proven be fit via a mixture of Weibull and normal distribution. A logistic regression model was developed to investigate major factors contributing to secondary accident occurrence. Traffic shock wave speed and volume at the occurrence of a primary accident were explicitly considered in the model, as a secondary accident is defined as an accident that occurs within the spatio-temporal impact scope of the primary accident. Results show that the shock waves originating in the wake of a primary accident have a more significant impact on the likelihood of a secondary accident occurrence than the effects of traffic volume. Primary accidents with long durations can significantly increase the possibility of secondary accidents. Unsafe speed and weather are other factors contributing to secondary crash occurrence. It is strongly suggested that when police or rescue personnel arrive at the scene of an accident, they should not suddenly block, decrease, or unblock the traffic flow, but instead endeavor to control traffic in a smooth and controlled manner. Also it is important to reduce accident processing time to reduce the risk of secondary accident.     

Safety effects of blue cycle crossings: a before-after study

This paper presents a before-after accident study of marking blue cycle crossings in 65 signalised junctions. Corrections factors for changes in traffic volumes and accident/injury trends are included using a general comparison group in this non-experimental observational study. Analysis of long-term accident trends point towards no overall abnormal accident counts in the before period. The safety effect depends on the number of blue cycle crossings at the junction. One blue cycle crossing reduces the number of junction accidents by 10%, whereas marking of two and four blue cycle crossings increases the number of accidents by 23% and 60%, respectively. Larger reduction and increases are found for injuries. Safety gains at junctions with one blue cycle crossing arise because the number of accidents with cyclists and moped riders that may have used the blue cycle crossing in the after period and pedestrians in the pedestrian crossing parallel and just next to the blue marking was statistically significant reduced. Two or four blue cycle crossings especially increase the number of rear-end collisions only with motor vehicles involved and right-angle collisions with passenger cars driving on red traffic lights.     

Models of perceived cycling risk and route acceptability

Perceived cycling risk and route acceptability to potential users are obstacles to policy support for cycling and a better understanding of these issues will assist planners and decision makers. Two models of perceived risk, based on non-linear least squares, and a model of acceptability, based on the logit model, have been estimated for whole journeys based on responses from a sample of 144 commuters to video clips of routes and junctions. The risk models quantify the effect of motor traffic volumes, demonstrate that roundabouts add more to perceived risk than traffic signal controlled junctions and show that right turn manoeuvres increase perceived risk. Facilities for bicycle traffic along motor trafficked routes and at junctions are shown to have little effect on perceived risk and this brings into question the value of such facilities in promoting bicycle use. These models would assist in specifying infrastructure improvements, the recommending of least risk advisory routes and assessing accessibility for bicycle traffic. The acceptability model confirms the effect of reduced perceived risk in traffic free conditions and the effects of signal controlled junctions and right turns. The acceptability models, which may be used at an area wide level, would assist in assessing the potential demand for cycling and in target setting.     

Bayes classifiers for imbalanced traffic accidents datasets

Traffic accidents data sets are usually imbalanced, where the number of instances classified under the killed or severe injuries class (minority) is much lower than those classified under the slight injuries class (majority). This, however, supposes a challenging problem for classification algorithms and may cause obtaining a model that well cover the slight injuries instances whereas the killed or severe injuries instances are misclassified frequently. Based on traffic accidents data collected on urban and suburban roads in Jordan for three years (2009–2011); three different data balancing techniques were used: under-sampling which removes some instances of the majority class, oversampling which creates new instances of the minority class and a mix technique that combines both. In addition, different Bayes classifiers were compared for the different imbalanced and balanced data sets: Averaged One-Dependence Estimators, Weightily Average One-Dependence Estimators, and Bayesian networks in order to identify factors that affect the severity of an accident. The results indicated that using the balanced data sets, especially those created using oversampling techniques, with Bayesian networks improved classifying a traffic accident according to its severity and reduced the misclassification of killed and severe injuries instances. On the other hand, the following variables were found to contribute to the occurrence of a killed causality or a severe injury in a traffic accident: number of vehicles involved, accident pattern, number of directions, accident type, lighting, surface condition, and speed limit. This work, to the knowledge of the authors, is the first that aims at analyzing historical data records for traffic accidents occurring in Jordan and the first to apply balancing techniques to analyze injury severity of traffic accidents.     

Predicting wet weather accidents

The development of a wet weather safety index, WWSI_e, is presented. WWSI_e is an empirical formulation based on prediction equations for wet accident rates. These equations were derived by multiple regression techniques from a survey of 68 highway segments in Texas. These segments covered a range of wet accident rates from zero to 40 accidents per year per mile. In general, much higher values of wet accident rates are observed in urban areas than in rural areas. Also, the sensitivity to pavement skid resistance is much higher in urban than in rural areas. The findings and developments reported warrant a restructuring of many state programs to reduce wet weather accidents. The Wet Weather Safety Index and associated equations represent a practical method of predicting wet accident rates as a function of traffic, road geometric, and pavements surface characteristics. These predictive equations may be cautiously used to determine accident reduction due to specific remedial measures. These new developments can be integrated into a comprehensive plan to reduce wet weather accidents, a plan which should greatly increase the effectiveness of those resources devoted to this objective.     

Predicting cycling accident risk in Brussels: A spatial case–control approach

This paper aims at predicting cycling accident risk for an entire network and identifying how road infrastructure influences cycling safety in the Brussels-Capital Region (Belgium). A spatial Bayesian modelling approach is proposed using a binary dependent variable (accident, no accident at location i) constructed from a case–control strategy. Control sites are sampled along the ‘bikeable’ road network in function of the potential bicycle traffic transiting in each ward. Risk factors are limited to infrastructure, traffic and environmental characteristics.     

Characteristics of rear-end accidents at signalized intersections using multiple logistic regression model

Multi-vehicle rear-end accidents constitute a substantial portion of the accidents occurring at signalized intersections. To examine the accident characteristics, this study utilized the 2001 Florida traffic accident data to investigate the accident propensity for different vehicle roles (striking or struck) that are involved in the accidents and identify the significant risk factors related to the traffic environment, the driver characteristics, and the vehicle types. The Quasi-induced exposure concept and the multiple logistic regression technique are used to perform this analysis. The results showed that seven road environment factors (number of lanes, divided/undivided highway, accident time, road surface condition, highway character, urban/rural, and speed limit), five factors related to striking role (vehicle type, driver age, alcohol/drug use, driver residence, and gender), and four factors related to struck role (vehicle type, driver age, driver residence, and gender) are significantly associated with the risk of rear-end accidents. Furthermore, the logistic regression technique confirmed several significant interaction effects between those risk factors.     

Some implications of an event-based definition of exposure to the risk of road accident

This paper proposes a new definition of exposure to the risk of road accident as any event, limited in space and time, representing a potential for an accident to occur by bringing road users close to each other in time or space of by requiring a road user to take action to avoid leaving the roadway. A typology of events representing a potential for an accident is proposed. Each event can be interpreted as a trial as defined in probability theory. Risk is the proportion of events that result in an accident. Defining exposure as events demanding the attention of road users implies that road users will learn from repeated exposure to these events, which in turn implies that there will normally be a negative relationship between exposure and risk. Four hypotheses regarding the relationship between exposure and risk are proposed. Preliminary tests support these hypotheses. Advantages and disadvantages of defining exposure as specific events are discussed. It is argued that developments in vehicle technology are likely to make events both observable and countable, thus ensuring that exposure is an operational concept.