In 2003, the Washington State Department of Transportation (WSDOT) initiated a research effort to develop and test a methodology for estimating congestion delay on Seattle area freeways. The initial phase of the research developed a preliminary methodology for both measuring congestion and assigning that congestion to specific lane blocking incidents. In Phase 2 of this study, that methodology was automated and extended to include an examination of the effects of incidents occurring on the shoulders of the freeway, spillback from incident-caused congestion on one freeway that affects a second freeway that feeds traffic onto the initially congested facility, special events, and weather.
Because funding for this effort was extremely limited, only three corridors (with both directions analyzed independently) and two months of data were analyzed. The project found that the Phase 1 methodology works well in assigning traffic congestion that occurs in the immediate time frame and geographic location of incidents to those incidents. It also showed that this process can be automated, making the data preparation effort now the primary determinant of the cost of additional work of this kind.
Unfortunately, even with the addition of shoulder incidents, special events, and weather to lane blocking events, the automated process is unable to assign a significant proportion of the non-recurring delay occurring on Seattle freeways. The detailed analysis of freeway performance showed that much of the delay caused by specific incidents occurs in places or at times removed from the incident itself. The result is that much of the observed non-recurring delay can not be assigned to specific causes by the current analytical process. New analytical procedures will need to be developed in Phase 3 of this work.
The Phase 2 analysis also confirmed that the congestion effects of specific incidents and other traffic disruptions are highly dependent on the background traffic conditions at the time of the disruption. While the Phase 2 analysis did not produce all of the results desired, the additional analyses showed that the summary statistics output from the Phase 2 software are useful as input to other analytical procedures.