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

Rain was a factor that influenced all three variables, while snow was not. AM and PM peaks had different influences on incident duration and lanes blockage. AM peaks did not influence either incident duration or lanes blocked, while PM peaks adversely impacted the number of lanes blocked an incident. During the period of night, more lanes were blocked in an incident and it took more time to clear an incident. Weekday was not an influencing factor for any of these three variables.

Whether there was property damage, injuries and/or fatalities, disabled vehicles, debris, or the incident happened at a site of unscheduled road repair influenced incident duration in the same order of extent as the number of lanes blocked in an incident.

Auto, bus, truck trailer, van, and standard truck influenced incident duration roughly in the same order of extent as the number of lanes blocked in an incident. Also, the number of vehicles involved in an incident influenced incident duration and the number of lanes blocked by an incident in a similar way.

Incident duration did not significantly vary over these three roadways while incidents which happened on the SHP had significantly fewer lanes blocked than the other two roadways.

Tow trucks influenced incident duration to a greater extent than the involvement of NYPD or NYCDOT.

In addition to the observations derived above, the following needs have been identified for the future. First, the influencing factors for these three variables can be better identified based on more than one class of incident, such as property damage, injury and/or fatality, disabled vehicle, and unscheduled road repair incidents. As mentioned in the analysis above, more incidents occurred on rainy days because rain reduces visibility and friction and old vehicles are hard to maintain at a low speed on rainy days. Note that more accidents may have occurred due to the reduction of visibility and friction and more disabled vehicle incidents may have happened due to rain. This perception needs to be validated based on the data, which can only be done by calibrating models for these types of incidents to see the exact impact of rain on incident frequency. Second, the influencing factors for incident duration can also be better identified based on separate models for each of the three roadways. Note that these three roadways have their own unique characteristics which may have determined that one type of incident has been more likely to happen on a given roadway than the others. For example, the SHP is a non-truck route, while the GOW and BQE carry the truck flow passing through NYC from NJ to Long Island. Thus more severe accidents may have happened on the GOW or BQE. By developing models for different incident types and roadways where incidents happened, the identification of influencing factors can be made more convincing. Third, incident data should be collected for a longer period of time. The data used in this study was collected for about three months covering primarily the winter season. Traffic patterns in other seasons may be different from winter and may cause incidents with different patterns. By the time this study was conducted, there was no official procedure existing for collecting incidents from multiple agencies in NYC. It is recommended that such efforts for collecting incident data be enhanced. Fourth, the impact of secondary accidents can be investigated with more incident data to be collected in future. In the data used this study, there were accidents that occurred in a very close proximity in space and time. However, it was very difficult to confirm from the survey whether there were correlated each other. With information on the secondary accidents to be collected in future, their occurrence can be investigated, which would be very useful for traffic management at incident sites. Fifth, the models developed in this study can be incorporated into Equation (1) by which the factors that significantly influence traffic delay can be identified. Elasticity of these factors versus traffic delay can be derived if these three variables (frequency, duration and severity) are integrated in one equation.