Analyzing Clearance Time of Urban Traffic Accidents in Abu Dhabi, United Arab Emirates, with Hazard-Based Duration Modeling Method

Abstract

Traffic incidents generate adverse impacts in many areas, including traffic flow, air pollution, fuel consumption, and secondary crashes; therefore, it is vital for traffic incident responders and operators to know how they can improve the efficiency of traffic incident management. This paper presents the results of an investigation of the effects of traffic accident characteristics on accident clearance time using fully parametric hazard-based duration models with emphasis on the accelerated failure time metric. Accident characteristics and clearance times were obtained from Abu Dhabi, the capital of the United Arab Emirates. The data were obtained from the Federal Traffic Statistics System and the records of the Abu Dhabi Collision Investigation Branch for the period from May 2009 to April 2010. For the purpose of this study, clearance time was defined as the length of time from the arrival of the collision investigator on the accident scene to the departure of the collision investigator from the scene. According to the goodness-of-fit test conducted in the study, the Weibull model without gamma heterogeneity was used. The estimation results show that various accident characteristics significantly affect clearance time; these characteristics include month, location, weather condition, accident type, number of causalities, and number of vehicles involved. These results highlight some weak points in the current practices of clearing accidents in Abu Dhabi. Accordingly, this paper suggests some mitigation measures.

Analysis of the Different Duration Stages of Accidents with Hazard-Based Model

“…Some studies showed that the distributions are different [4,6,7,[19][20][21][23][24][25]. The differences in the findings that were described may have resulted from several factors, including small sample sizes (several hundred accident records), poor quality of accident data, differences in the factors that affected accident duration, and differences among the countries from which the data was obtained.…”

“…An interesting result was observed with regard to the effect that occurred when a taxi was involved in the accident: the duration lengthened when a taxi was involved. Using accident clearance time data from the city of Abu Dhabi, Alkaabi et al [21] developed the Weibull AFT model (without gamma heterogeneity), which showed that accident clearance times were longer and involved more casualties and vehicles when accidents occurred during out-of-peak hours, during January and March, and under severe weather conditions.…”

Simultaneous equation modeling of freeway accident duration and lanes blocked

“…Typically, an incident is expected to be more likely to clear if it has continued for longer. This distinct feature of hazard-based duration models has inspired many studies to apply same method, but use different or mixture probability distributions that specify hazard functions or unobserved heterogeneity (Qi and Teng, 2008;Qi et al, 2009;Chung, 2010;Chung et al, 2010;Al Kaabi et al, 2011;Al Kaabi et al, 2012;Li et al, 2015). Hazard-based duration models have also the capability to account for spatial dependence of incident duration (Xie et al, 2015) and include random effects of explanatory variables (Tavassoli Hojati et al, 2014).…”

“…For example, Nam and Mannering (2000) adopted a two-year dataset from Washington State's IRT program, and developed models for the duration of each phase. Al Kaabi et al (2011) used a one-year dataset from Abu Dhabi's federal traffic statistics and collision reports to develop models specifically for the period of accident clearance; in a follow-up study, Al Kaabi et al (2012) studied the response time on Abu Dhabi's highways. Chung (2010) and Chung et al (2010) defined accident duration as the time between notification about an accident to its clearance, and developed models of the factors that significantly affect this duration.…”

“…Three distributions-Weibull, Log-logistic, and Log-normal-are frequently used to examine accident duration data (Chung, 2010;Chung et al, 2010;Al Kaabi et al, 2011). The AFT model is defined as a Weibull, Log-normal, and Log-logistic model if the ε in Eq.…”