Andrew P. Tarko

Tobit analysis of vehicle accident rates on interstate highways

There has been an abundance of research that has used Poisson models and its variants (negative binomial and zero-inflated models) to improve our understanding of the factors that affect accident frequencies on roadway segments. This study explores the application of an alternate method, tobit regression, by viewing vehicle accident rates directly (instead of frequencies) as a continuous variable that is left-censored at zero. Using data from vehicle accidents on Indiana interstates, the estimation results show that many factors relating to pavement condition, roadway geometrics and traffic characteristics significantly affect vehicle accident rates.

Markov switching negative binomial models: an application to vehicle accident frequencies

In this paper, two-state Markov switching models are proposed to study accident frequencies. These models assume that there are two unobserved states of roadway safety, and that roadway entities (roadway segments) can switch between these states over time. The states are distinct, in the sense that in the different states accident frequencies are generated by separate counting processes (by separate Poisson or negative binomial processes). To demonstrate the applicability of the approach presented herein, two-state Markov switching negative binomial models are estimated using five-year accident frequencies on Indiana interstate highway segments. Bayesian inference methods and Markov Chain Monte Carlo (MCMC) simulations are used for model estimation. The estimated Markov switching models result in a superior statistical fit relative to the standard (single-state) negative binomial model. It is found that the more frequent state is safer and it is correlated with better weather conditions. The less frequent state is found to be less safe and to be correlated with adverse weather conditions.

Predicting crash likelihood and severity on freeways with real-time loop detector data

Real-time crash risk prediction using traffic data collected from loop detector stations is useful in dynamic safety management systems aimed at improving traffic safety through application of proactive safety countermeasures. The major drawback of most of the existing studies is that they focus on the crash risk without consideration of crash severity. This paper presents an effort to develop a model that predicts the crash likelihood at different levels of severity with a particular focus on severe crashes. The crash data and traffic data used in this study were collected on the I-880 freeway in California, United States. This study considers three levels of crash severity: fatal/incapacitating injury crashes (KA), non-incapacitating/possible injury crashes (BC), and property-damage-only crashes (PDO). The sequential logit model was used to link the likelihood of crash occurrences at different severity levels to various traffic flow characteristics derived from detector data. The elasticity analysis was conducted to evaluate the effect of the traffic flow variables on the likelihood of crash and its severity.The results show that the traffic flow characteristics contributing to crash likelihood were quite different at different levels of severity. The PDO crashes were more likely to occur under congested traffic flow conditions with highly variable speed and frequent lane changes, while the KA and BC crashes were more likely to occur under less congested traffic flow conditions. High speed, coupled with a large speed difference between adjacent lanes under uncongested traffic conditions, was found to increase the likelihood of severe crashes (KA). This study applied the 20-fold cross-validation method to estimate the prediction performance of the developed models. The validation results show that the models crash prediction performance at each severity level was satisfactory. The findings of this study can be used to predict the probabilities of crash at different severity levels, which is valuable knowledge in the pursuit of reducing the risk of severe crashes through the use of dynamic safety management systems on freeways.

Use of crash surrogates and exceedance statistics to estimate road safety

The limited ability of existing safety models to properly reflect crash causality has its source in cross-sectional analysis applied to the estimation of the intrinsically complex safety factors with highly aggregated and frequently poor quality of data. The adequacy of the data may be improved thanks to the unprecedented progress in sensing technologies and the invention of the naturalistic driving method of data collection. Proposed in this paper is a new modeling paradigm that integrates several types of safety models. The primary improvement results from a more adequate representation of the crash occurrence process by incorporating crash precursor events into the modeling framework. A Pareto-based estimating method for the likelihood of a collision occurrence, given a precursor event, is explained and illustrated with the simple example of road departures.

Speed Factors on Two-Lane Rural Highways in Free-Flow Conditions

The mean free-flow speed and its variability across drivers are considered important safety factors. Despite a large body of research on operating speeds, there is still much to learn about the factors of free-flow speeds, especially on tangent segments of two-lane rural highways. The roadway factors of speed dispersion across drivers are largely unknown. Also, the use of the entire free-flow speed distribution suggested by other authors has not yet been addressed. Consequently, the existing models are not aimed to evaluate the speed variability at a site. This paper presents free-flow speed models that identify factors of mean speed and speed dispersion on tangent segments and horizontal curves of two-lane rural highways. Ten highway variables, six of them functioning as both mean speed and speed dispersion factors, were identified as speed factors on tangent segments. Four highway and curve variables, two of them functioning as both mean speed and speed dispersion factors, were identified as speed facto...

Safety impacts at intersections on curved segments

Indiana geometric design policy, consistent with national standards, allows for the design of intersections on superelevated curves if other solutions are prohibitively expensive. Consequently, the Indiana Department of Transportation (DOT) has built a number of such intersections. Following a series of fatal crashes at one of these intersections, Indiana DOT made a decision to avoid designing intersections on segments with steep superelevation. This design restriction calls for expensive alternatives, such as realigning roads or adding grade separations. This research was done to determine whether superelevated intersections were more hazardous than similar intersections located on tangents and, if so, to determine what combination of factors made this true. The research focused on two-way stop-controlled intersections where the mainline was a high-speed four-lane divided highway located on a superelevated curve. An attempt was made to analyze as many factors as possible by using appropriate comparison techniques. Negative binomial models were developed to determine the statistical relationship between crash occurrence and intersection geometric characteristics, including curvature of the main road. Crash severity and the joint impact of curvature with weather and lighting conditions were examined by using binomial comparisons of proportions. Research findings show significant increases in crash frequency and severity at intersections located on superelevated curves.

Reconciling Speed Limits with Design Speeds

In recent years, context-sensitive highway design has been promoted to ensure that designers consider the environmental, scenic, aesthetic, historic, community, and preservation aspects of the road. Context-sensitive design may lead to situations where the current design standards cannot be met because of restricting local conditions. Indiana has road sections designed and built some time ago. In a considerable number of roads with the statutory limit of 55 mph (90 km/h), the road geometry does not meet the current standards. At individual intersections and on curves, advisory speeds are posted together with warning signs. Although this solution increases the safety of road users and allows for traveling at reasonably high speeds outside of these segments, the final solution is to upgrade their geometry to the desirable level. This report presents models that predict user-selected percentile free-flow speeds on two-lane rural and four-lane rural and suburban highways. The percentile speeds are computed as a linear combination of the mean speed and the standard deviation in panel data models with random effects. The developed percentile speed models involve more design variables than typical speed models, and separate the mean speed factors from the speed dispersion factors. These benefits ease the model interpretation and its use in highway design. The study results should help designers bring the predicted speed to the desired speed as close as possible given the budget constraints. Engineering judgment can then be applied to balance safety and construction cost in highway improvement projects.

Pedestrian injury analysis with consideration of the selectivity bias in linked police-hospital data

Evaluation of crash-related injuries by medical specialists in hospitals is believed to be more exact than rather a cursory evaluation made at the crash scene. Safety analysts sometimes reach for hospital data and use them in combination with the police crash data. One issue that needs to be addressed is the, so-called, selectivity (or selection) bias possible when data used in analysis are not coming from random sampling. If not properly addressed, this issue can lead to a considerable bias in both the model coefficient estimates and the model predictions. This paper investigates pedestrian injury severity factors using linked police-hospital data. A bivariate ordered probit model with sample selection is used to check for the presence of the selectivity bias and to account for it in the MAIS estimates on the Maximum Abbreviated Injury Scale (MAIS). The presence of the sample selection issue has been confirmed. The selectivity bias is considerable in predictions of low injury levels. The pedestrian injury analysis identified and estimated several severity factors, including pedestrian, road, and vehicle characteristics. Male and older pedestrians were found to be particularly exposed to severe injuries. Rural roads and high-speed urban roads appear to be more dangerous for pedestrians, particularly when crossing such roads. Crossing a road between intersections was found to be particularly dangerous behavior. The size and weight of the vehicle involved in a pedestrian crash were also found to have an effect on the pedestrian injury level. The relevant safety countermeasures that may improve pedestrian safety have been proposed.

Tool with Road-Level Crash Prediction for Transportation Safety Planning

The growing use of packages based on geographic information systems (GISs) in transportation planning is assisting in the development and implementation of methods that facilitate consideration of safety. This paper presents a method of predicting safety for planning alternatives. Although applicable to large road networks, the method predicts crashes at the road facility level (intersections and segments). Thus it is suitable for joint evaluation of modifications of the road network, changes in network traffic flows, and improvements in road geometry considered by planners. A complete set of crash prediction models was developed by the authors for seven types of road segments and four types of road nodes on the basis of crashes reported on Indiana highways from 2003 to 2005. A mainstream research method has been used: negative binomial regression with a stepwise method for variable selection and the Akaike information criterion. The obtained equations are transparent to transportation planners and allow efficient computations for large road networks. The crash prediction equations have been implemented in the GIS-based planning package TransCAD as an add-on tool. The calibration procedure included in the tool allows a planner to search for optimal values of calibration factors if calibration of crash prediction models is needed.

Effective and Fair Identification of Hazardous Locations

The two fundamental objectives of a safety management system are to prevent as many crashes as possible and to reduce the difference in risk faced by individual road users. An index of crash frequency and an index of crash cost are proposed to address these two criteria in the hazard identification phase of safety management. An index of crash cost is used to incorporate severity in identification criteria. Safety performance functions based on negative binomial distribution are used to predict the typical crash frequency at the location. The proposed methods can be used to rank locations and to evaluate the degree of hazard at an individual location without referring to other locations. These methods determine the evidence of hazard at all types of locations (intersections and segments), and they can use crash data from periods shorter than 1 year. Indices of crash frequency and cost were evaluated and found helpful for safety management that reduces the number of crashes and risk variability across a road network.