An investigation of incident frequency, duration and lanes blockage for determining traffic delay

Journal of Advanced Transportation, Fall, 2009 by Yi "Grace" Qi, Hualiang "Harry" Teng, David R. Martinelli

As far as incident duration is concerned, the related issues dealt with in the existing studies vary broadly. The study by Golob et al. (1987) was concerned with the probability distribution of duration. It was validated that the duration of truck related accidents follows the log-normal distribution. Skabardonis et al. (1998) studied the relationship between incident duration and several influencing factors by employing linear regression models. Realizing the limitation of linear regression models, Jones et al. (1999) developed advanced hazard-based duration models to identify the influencing factors of accident duration. Along the same lines, Nam and Mannering (2000) considered more than one type of hazard-based duration model when they modeled incident detection/reporting time, response time, and clearance time, separately. Khattak et al. (1994) focused their efforts on on-line incident duration prediction.

In general, there are not many studies in which the number of lanes blocked by an incident has been discussed. Furthermore, few of them have been devoted to identifying the influencing factors for the number of lanes blocked by an incident. As for the ordered probit model that was employed in this study, it can identify the interdependency between the explanatory variables and the dependent variables. Because this model can recognize the ordinal nature of a dependent variable, it is especially appropriate for modeling the number of lanes blocked by an incident, which are ordered in nature. In point of fact, this model has been applied in several studies in the transportation field. One of them is Khattak (2000) where the effects of driver, vehicle and roadway factors on the propensity for driver injury in rear-end crashes were examined. Kockelman (2001) is another recent study where ordered probit models were employed to analyze the effects of factors such as gender, age and speed on driver injury levels.

Incident Data Description

The investigation presented in this study was based on incident data

collected in a survey conducted by NYSDOT Region 11 from December 6 1995 to February 29, 1996. The survey was designed by NYSDOT Region 11 and was distributed to NYCDOT, NYPD, and other related agencies who were responsible for incident response. The objective of the survey was to collect data for as many incidents as possible in order to compare with those collected by TRANSCOM, an agency that was responsible for disseminating incident information around the Tri-State area. In the survey, incident data was collected for eleven roadways that were owned by the State.

After receiving the survey, the incident data was keyed into a computer. To ensure the quality of the input data, the data was double-checked. Since incident data was collected from several different agencies, it is possible that some agencies may have filled out several survey forms for one incident. Thus, it is necessary to search for and exclude these duplicates from the database. To do that, all the incidents were geocoded using a GIS tool based on the incident location marked on the network map displayed on each survey form. A program was written to search for duplicates based on criteria such as whether they occurred in close approximation in time and space and whether they have demonstrated similar incident patterns. As a result, a very small number of incidents were identified as duplicates and then excluded from the database. It was perceived by the administrator of the survey that the Shore Parkway (SHP), Brooklyn-Queens Expressway (BQE), and Gowanus Expressway (GOW) were the three roadways for which the collected incidents can be viewed as the most complete. Thus, it is for these three roadways that incidents were analyzed in this study. Figure 1 shows the location of these three roadways in NYC.