Estimating the Probability of Freeway Congestion Recurrence

Abstract

The 1991 Intermodal Surface Transportation Efficiency Act, along with the National Highway System Designation Act of 1995, mandated that congestion management systems be established in urbanized areas with population of 200,000 or more. The mandate created a need for congestion monitoring. A 1995 informal national survey of transportation agencies by FHWA indicated a need for tools to evaluate the impacts of remedial strategies. Further, some agencies have sponsored research on topics such as congestion thresholds and measurement methods. This paper proposes a model that estimates the probability of congestion recurrence at existing and potential freeway bottlenecks. The objectives were that the model accept data that are commonly collected, evaluate existing conditions, predict future conditions, and assess the impacts of mitigating actions. A logistic regression equation is developed from 163 observations. The equation is bivariate, with the annual average daily traffic volume to hourly capacity ratio ( AADT/C) and the K-factor as explanatory variables, and RECON, a binary congestion indicator, as the response variable. The model correctly classifies 83 percent of the data sites used in its development, but it needs to be tested on other data. An AADT/C value of 8.5 may be a critical threshold separating the noncongested and congested regimes. Future research might experiment with other explanatory variables. A larger database might yield an improved classification rate. The extra data might also enable the development of a nested logit model that estimates the probabilities of no congestion, congestion once a day, and congestion twice a day at a bottleneck.