Bayesian hierarchical analysis on crash prediction models

Traditional crash prediction models, such as generalized linear regression models, are incapable of taking into account the multilevel data structure, which extensively exists in crash data. Disregarding the possible within-group correlations can lead to the production of models giving unreliable and biased estimates of unknowns. This study innovatively proposes a -level hierarchy, viz. (Geographic region level – Traffic site level – Traffic crash level – Driver-vehicle unit level – Vehicle-occupant level) Time level, to establish a general form of multilevel data structure in traffic safety analysis. To properly model the potential cross-group heterogeneity due to the multilevel data structure, a framework of Bayesian hierarchical models that explicitly specify multilevel structure and correctly yield parameter estimates is introduced and recommended. The proposed method is illustrated in an individual-severity analysis of intersection crashes using the Singapore crash records. This study proved the importance of accounting for the within-group correlations and demonstrated the flexibilities and effectiveness of the Bayesian hierarchical method in modeling multilevel structure of traffic crash data.

The influence of motorcyclist behavior in crash involvement

Introduction: In Singapore, motorcycle crashes account for 50% of traffic fatalities and 53% of injuries. While extensive research efforts have been devoted to improve the motorcycle safety, the relationship between the rider behavior and the crash risk is still not well understood. The objective of this study is to evaluate how behavioral factors influence crash risk and to identify the most vulnerable group of motorcyclists. Methods: To explore the rider behavior, a 61-item questionnaire examining sensation seeking (Zuckerman et al., 1978), impulsiveness (Eysenck et al., 1985), aggressiveness (Buss & Perry, 1992), and risk-taking behavior (Weber et al., 2002) was developed. A total of 240 respondents with at least one year riding experience form the sample that relate behavior to their crash history, traffic penalty awareness, and demographic characteristics. By clustering the crash risk using the medoid portioning algorithm, the log-linear model relating the rider behavior to crash risk was developed. Results and Discussions: Crash-involved motorcyclists scored higher in impulsive sensation seeking, aggression and risk-taking behavior. Aggressive and high risk-taking motorcyclists were respectively 1.30 and 2.21 times more likely to fall under the high crash involvement group while impulsive sensation seeking was not found to be significant. Based on the scores on risk-taking and aggression, the motorcyclists were clustered into four distinct personality combinations namely, extrovert (aggressive, impulsive risk-takers), leader (cautious, aggressive risk-takers), follower (agreeable, ignorant risk-takers), and introvert (self-consciousness, fainthearted risk-takers). “Extrovert” motorcyclists were most prone to crashes, being 3.34 times more likely to involve in crash and 8.29 times more vulnerable than the “introvert”. Mediating factors like demographic characteristics, riding experience, and traffic penalty awareness were found not to be significant in reducing crash risk. Conclusion: The findings of this study will be useful for road safety campaign planners to be more focused in the target group as well as those who employ motorcyclists for their delivery business.

Misericordiae Antemortem - The 1955 Mount Superbus Crash

On the 9th April 1955, RAAF Lincoln Bomber A73-64, on a mercy flight to transfer a critically ill infant from Townsville to Brisbane, crashed at Mount Superbus killing the four crew and two civilians on board. The immediate search and rescue was organised by a group of Brisbane bushwalkers who were camping in the area. Police and RAAF personnel subsequently joined the civilians at the crash site to recover the victims. During their initial search of the crash they located what were believed to be the remains of five adults. The arrival of the RAAF Senior Medical Officer (SMO) the following day revealed that only four adult bodies had been found and the bodies of both civilians, an adult and infant, were missing. Later that day the remains of six victims were recovered from the crash site and conveyed to the Warwick Police Station for identification. The RAAF SMO was responsible for the identifications of the aircrew while the Government Medical Officer, police and coroner were responsible for the identifications of the civilians. Eight days later, further remains of the infant were found by a civilian looking through the wreckage. This paper uses archival records not previously researched from a Disaster Victim Identification (DVI) perspective to stimulate interest among forensic practitioners, criminologists and other interested parties in the history of DVI and how practices in Australia have evolved.

On the commonly accepted assumptions regarding observed motor vehicle crash counts at transport system locations

Readily accepted knowledge regarding crash causation is consistently omitted from efforts to model and subsequently understand motor vehicle crash occurrence and their contributing factors. For instance, distracted and impaired driving accounts for a significant proportion of crash occurrence, yet is rarely modeled explicitly. In addition, spatially allocated influences such as local law enforcement efforts, proximity to bars and schools, and roadside chronic distractions (advertising, pedestrians, etc.) play a role in contributing to crash occurrence and yet are routinely absent from crash models. By and large, these well-established omitted effects are simply assumed to contribute to model error, with predominant focus on modeling the engineering and operational effects of transportation facilities (e.g. AADT, number of lanes, speed limits, width of lanes, etc.) The typical analytical approach—with a variety of statistical enhancements—has been to model crashes that occur at system locations as negative binomial (NB) distributed events that arise from a singular, underlying crash generating process. These models and their statistical kin dominate the literature; however, it is argued in this paper that these models fail to capture the underlying complexity of motor vehicle crash causes, and thus thwart deeper insights regarding crash causation and prevention. This paper first describes hypothetical scenarios that collectively illustrate why current models mislead highway safety researchers and engineers. It is argued that current model shortcomings are significant, and will lead to poor decision-making. Exploiting our current state of knowledge of crash causation, crash counts are postulated to arise from three processes: observed network features, unobserved spatial effects, and ‘apparent’ random influences that reflect largely behavioral influences of drivers. It is argued; furthermore, that these three processes in theory can be modeled separately to gain deeper insight into crash causes, and that the model represents a more realistic depiction of reality than the state of practice NB regression. An admittedly imperfect empirical model that mixes three independent crash occurrence processes is shown to outperform the classical NB model. The questioning of current modeling assumptions and implications of the latent mixture model to current practice are the most important contributions of this paper, with an initial but rather vulnerable attempt to model the latent mixtures as a secondary contribution.

A comparison of perceived crash risk among cyclists and drivers in Brisbane

The Queensland Government has implemented strategies promoting a shift from individual car use to active transport, a transition which requires drivers to adapt to sharing the road with increased numbers of people cycling through transport network. For this to occur safely, changes in both road infrastructure and road user expectations and behaviors will be needed. Creating separate cycle infrastructure does not remove the need for cyclists to commence, cross or finish travel on shared roads. Currently intersections are one of the predominant shared road spaces where crashes result in cyclists being injured or killed. This research investigates how Brisbane cyclists and drivers perceive risk when interacting with other road users at intersections. The current study replicates a French study conducted by co-authors Chaurand and Delhomme in 2011 and extends it to assess gender effects which have been reported in other Australian cycling research. An online survey was administered to experienced cyclists and drivers. Participants rated the level of risk they felt when imagining a number of different road situations. Based on the earlier French study it is expected that perceived crash risk will be influenced both by the participant’s mode of travel and the type of interacting vehicle and perceived risk will be greater when the interaction is with a car than a bicycle. It is predicted that risk perception will decrease as the level of experience increases and that male participants will have a higher perception of skill and lower perception of risk than females. The findings of this Queensland study will provide a valuable insight into perceived risk and the traffic behaviours of drivers and cyclists when interacting with other road users and results will be available for presentation at the Congress.

Comparing cooperative and non-cooperative crash risk assessment

Cooperative Systems provide, through the multiplication of information sources over the road, a lot of potential to improve the assessment of the road risk describing a particular driving situation. In this paper, we compare the performance of a cooperative risk assessment approach against a non-cooperative approach; we used an advanced simulation framework, allowing for accurate and detailed, close-to-reality simulations. Risk is estimated, in both cases, with combinations of indicators based on the TTC. For the non-cooperative approach, vehicles are equipped only with an AAC-like forward-facing ranging sensor. On the other hand, for the cooperative approach, vehicles share information through 802.11p IVC and create an augmented map representing their environment; risk indicators are then extracted from this map. Our system shows that the cooperative risk assessment provides a systematic increase of forward warning to most of the vehicles involved in a freeway emergency braking scenario, compared to a non-cooperative system.

Comparison of moped, scooter and motorcycle crash risk and crash severity

The increased popularity of mopeds and motor scooters in Australia and elsewhere in the last decade has contributed substantially to the greater use of powered two-wheelers (PTWs) as a whole. As the exposure of mopeds and scooters has increased, so too has the number of reported crashes involving those PTW types, but there is currently little research comparing the safety of mopeds and, particularly, larger scooters with motorcycles. This study compared the crash risk and crash severity of motorcycles, mopeds and larger scooters in Queensland, Australia. Comprehensive data cleansing was undertaken to separate motorcycles, mopeds and larger scooters in police-reported crash data covering the five years to 30 June 2008. The crash rates of motorcycles (including larger scooters) and mopeds in terms of registered vehicles were similar over this period, although the moped crash rate showed a stronger downward trend. However, the crash rates in terms of distance travelled were nearly four times higher for mopeds than for motorcycles (including larger scooters). More comprehensive distance travelled data is needed to confirm these findings. The overall severity of moped and scooter crashes was significantly lower than motorcycle crashes but an ordered probit regression model showed that crash severity outcomes related to differences in crash characteristics and circumstances, rather than differences between PTW types per se. Greater motorcycle crash severity was associated with higher (>80 km/h) speed zones, horizontal curves, weekend, single vehicle and nighttime crashes. Moped crashes were more severe at night and in speed zones of 90 km/h or more. Larger scooter crashes were more severe in 70 km/h zones (than 60 km/h zones) but not in higher speed zones, and less severe on weekends than on weekdays. The findings can be used to inform potential crash and injury countermeasures tailored to users of different PTW types.

A data mining driven crash risk profiling method for road asset management

This thesis takes a new data mining approach for analyzing road/crash data by developing models for the whole road network and generating a crash risk profile. Roads with an elevated crash risk due to road surface friction deficit are identified. The regression tree model, predicting road segment crash rate, is applied in a novel deployment coined regression tree extrapolation that produces a skid resistance/crash rate curve. Using extrapolation allows the method to be applied across the network and cope with the high proportion of missing road surface friction values. This risk profiling method can be applied in other domains.

Pre-crash traffic flow trend analysis on motorways

Crashes on motorway contribute to a significant proportion (40-50%) of non-recurrent motorway congestions. Hence reduce crashes will help address congestion issues (Meyer, 2008). Crash likelihood estimation studies commonly focus on traffic conditions in a Short time window around the time of crash while longer-term pre-crash traffic flow trends are neglected. In this paper we will show, through data mining techniques, that a relationship between pre-crash traffic flow patterns and crash occurrence on motorways exists, and that this knowledge has the potential to improve the accuracy of existing models and opens the path for new development approaches. The data for the analysis was extracted from records collected between 2007 and 2009 on the Shibuya and Shinjuku lines of the Tokyo Metropolitan Expressway in Japan. The dataset includes a total of 824 rear-end and sideswipe crashes that have been matched with traffic flow data of one hour prior to the crash using an incident detection algorithm. Traffic flow trends (traffic speed/occupancy time series) revealed that crashes could be clustered with regards of the dominant traffic flow pattern prior to the crash. Using the k-means clustering method allowed the crashes to be clustered based on their flow trends rather than their distance. Four major trends have been found in the clustering results. Based on these findings, crash likelihood estimation algorithms can be fine-tuned based on the monitored traffic flow conditions with a sliding window of 60 minutes to increase accuracy of the results and minimize false alarms.

Pre-crash and non-crash traffic flow trends analysis on motorways

Crashes that occur on motorways contribute to a significant proportion (40-50%) of non-recurrent motorway congestions. Hence, reducing the frequency of crashes assists in addressing congestion issues (Meyer, 2008). Crash likelihood estimation studies commonly focus on traffic conditions in a short time window around the time of a crash while longer-term pre-crash traffic flow trends are neglected. In this paper we will show, through data mining techniques that a relationship between pre-crash traffic flow patterns and crash occurrence on motorways exists. We will compare them with normal traffic trends and show this knowledge has the potential to improve the accuracy of existing models and opens the path for new development approaches. The data for the analysis was extracted from records collected between 2007 and 2009 on the Shibuya and Shinjuku lines of the Tokyo Metropolitan Expressway in Japan. The dataset includes a total of 824 rear-end and sideswipe crashes that have been matched with crashes corresponding to traffic flow data using an incident detection algorithm. Traffic trends (traffic speed time series) revealed that crashes can be clustered with regards to the dominant traffic patterns prior to the crash. Using the K-Means clustering method with Euclidean distance function allowed the crashes to be clustered. Then, normal situation data was extracted based on the time distribution of crashes and were clustered to compare with the “high risk” clusters. Five major trends have been found in the clustering results for both high risk and normal conditions. The study discovered traffic regimes had differences in the speed trends. Based on these findings, crash likelihood estimation models can be fine-tuned based on the monitored traffic conditions with a sliding window of 30 minutes to increase accuracy of the results and minimize false alarms.

The consistency of crash involvement recall across time

Introduction Road safety researchers rely heavily on self-report data to explore the aetiology of crash risk. However, researchers consistently acknowledge a range of limitations associated with this methodological approach (e.g., self-report bias), which has been hypothesised to reduce the predictive efficacy of scales. Although well researched in other areas, one important factor often neglected in road safety studies is the fallibility of human memory. Given accurate recall is a key assumption in many studies, the validity and consistency of self-report data warrants investigation. The aim of the current study was to examine the consistency of self-report data of crash history and details of the most recent reported crash on two separate occasions. Materials & Method A repeated measures design was utilised to examine the self-reported crash involvement history of 214 general motorists over a two month period. Results A number of interesting discrepancies were noted in relation to number of lifetime crashes reported by the participants and the descriptions of their most recent crash across the two occasions. Of the 214 participants who reported having been involved in a crash, 35 (22.3%) reported a lower number of lifetime crashes as Time 2, than at Time 1. Of the 88 drivers who reported no change in number of lifetime crashes, 10 (11.4%) described a different most recent crash. Additionally, of the 34 reporting an increase in the number of lifetime crashes, 29 (85.3%) of these described the same crash on both occasions. Assessed as a whole, at least 47.1% of participants made a confirmed mistake at Time 1 or Time 2. Conclusions These results raise some doubt in regard to the accuracy of memory recall across time. Given that self-reported crash involvement is the predominant dependent variable used in the majority of road safety research, this issue warrants further investigation. Replication of the study with a larger sample size that includes multiple recall periods would enhance understanding into the significance of this issue for road safety methodology.

Random breath testing in Queensland and Western Australia : Examination of how the random breath testing rate influences alcohol related traffic crash rates

In this paper we explore the relationship between monthly random breath testing (RBT) rates (per 1000 licensed drivers) and alcohol-related traffic crash (ARTC) rates over time, across two Australian states: Queensland and Western Australia. We analyse the RBT, ARTC and licensed driver rates across 12 years; however, due to administrative restrictions, we model ARTC rates against RBT rates for the period July 2004 to June 2009. The Queensland data reveals that the monthly ARTC rate is almost flat over the five year period. Based on the results of the analysis, an average of 5.5 ARTCs per 100,000 licensed drivers are observed across the study period. For the same period, the monthly rate of RBTs per 1000 licensed drivers is observed to be decreasing across the study with the results of the analysis revealing no significant variations in the data. The comparison between Western Australia and Queensland shows that Queensland's ARTC monthly percent change (MPC) is 0.014 compared to the MPC of 0.47 for Western Australia. While Queensland maintains a relatively flat ARTC rate, the ARTC rate in Western Australia is increasing. Our analysis reveals an inverse relationship between ARTC RBT rates, that for every 10% increase in the percentage of RBTs to licensed driver there is a 0.15 decrease in the rate of ARTCs per 100,000 licenced drivers. Moreover, in Western Australia, if the 2011 ratio of 1:2 (RBTs to annual number of licensed drivers) were to double to a ratio of 1:1, we estimate the number of monthly ARTCs would reduce by approximately 15. Based on these findings we believe that as the number of RBTs conducted increases the number of drivers willing to risk being detected for drinking driving decreases, because the perceived risk of being detected is considered greater. This is turn results in the number of ARTCs diminishing. The results of this study provide an important evidence base for policy decisions for RBT operations.