123 Transportation Research Record: Journal of the Transportation Research Board, No. 2395, Transportation Research Board of the National Academies, Washington, D.C., 2013, pp. 123–131. DOI: 10.3141/2395-14 Institute for Transportation Research and Education, North Carolina State Uni- versity, Centennial Campus, Box 8601, Raleigh, NC 27695-8601. Corresponding author: S. Aghdashi, saghdas@ncsu.edu. capacity or an abnormal increase in demand” (5). This definition classifies special nonrecurring events (such as work zones or ball games) as incidents. The Highway Capacity Manual 2010 defines incidents as “any occurrence on a roadway that impedes normal traffic flow” (6 ). Finally, the Traffic Management Data Dictionary defines an incident as “an unplanned randomly occurring traffic event that adversely affects normal traffic operations” (7 ). Although these three definitions are similar, they do not exactly reflect the inci- dent definition used in this study. In this study, an incident is defined as an unscheduled and random event that impedes normal traffic flow and results in the closure of the shoulder or lanes as a result of vehicle breakdowns; crashes; the presence of disabled, stopped, or pulled- over vehicles on the roadway or the shoulder; or the presence of debris on the roadway or the shoulder. In addition, five levels of inci- dent severity are defined: (a) shoulder closure, (b) one-lane closure, (c) two-lane closure, (d) three-lane closure, and (e) four-lane closure. Appropriate mathematical models are required to account for the stochastic behavior of incidents on freeway facilities. The proposed mathematical model in this paper is a queuing theory approach that models the freeway as a queuing system in which incidents repre- sent the customers who are served by infinite servers (points on the freeway facility). The incident occurrence rate and its duration are fed into the queuing model to fully characterize the problem. The proposed methodology in this paper provides the analyst with multiple paths to estimate the required probabilities of incident occurrence. The methodology includes, on the one hand, the option of directly computing the probabilities and, on the other, the option of using national default values along with mathematical models to estimate the probabilities (8). Either way, the incident probability for a given time period is always defined as the fraction of time that an incident of a specific level of severity occurs on the facility. The default values used in this methodology are based on national data for incidents and crashes from nine freeway facilities (9). These default values are provided for the analyst to use in the absence of local data. In travel time reliability analysis, along with the probability, the impact of incidents should be considered. The incorporation of incident impacts includes one additional step: the modeling of the incident effect during the study periods, for which the Highway Capacity Manual’s freeway facilities methodology is used to esti- mate the resulting travel time. A study period is defined as the time during which the freeway facility reliability is analyzed and usually covers the morning or evening peak. Incident impacts are realized by inserting the appropriate capacity adjustment factors, the free-flow speed adjustment factors, and the number of remaining open lanes associated with the incident. Estimation of Incident Propensity for Reliability Analysis in the Highway Capacity Manual Seyedbehzad Aghdashi, Nagui M. Rouphail, and Ali Hajbabaie This paper presents the method used to generate the incident probabili- ties required by the freeway scenario generator for travel time reliability analysis in the Highway Capacity Manual. The freeway scenario genera- tor requires the estimation of monthly probabilities of different levels of incident severity during specified study periods. Incident probability in this context is the fraction of time that an incident of a specific level of severity is active somewhere on the freeway facility during the study period for the month considered. The proposed method is designed to recognize and deal with the varying levels of incident and facility data availability at the implementing agencies. A queuing model is proposed for the conversion of incident frequencies into incident probabilities when agencies have access only to frequencies instead of probabilities. The estimation of the probability of incidents is a key ingredient in travel time reliability analyses for freeways (1). The estimation feeds into the scenario generation procedure to enumerate and characterize the full variability of the operational status of a freeway facility (2). These operational statuses are analyzed, and the travel time distribu- tion, which constitutes the basis for the reliability performance over an extended time horizon, is formed (3). In travel time reliability analyses, three basic stochastic events may affect travel time: the variability in demand, the weather, and incidents (3). The demand level is not fixed, and changes over time result in variations in travel time. Similarly, inclement weather occurs in probabilistic events that yield changes in the free-flow speed and capacity of a roadway and, consequently, in travel time (3). Weather events that are considered in freeway travel time reli- ability analysis are taken from the Highway Capacity Manual 2010 (4). To successfully capture travel time reliability, the probabilities of these three stochastic events need to be calculated. This paper provides a methodology for the estimation of incident probabili- ties. Incidents are considered to be stochastic in nature, in that their occurrence and duration are probabilistic with certain distributions and parameters. The location and start time are two other dimensions of incident occurrence, both of which are stochastic as well. There is no clear-cut definition of incidents in the transportation literature. The Traffic Incident Management Handbook defines inci- dents as “any non-recurring event that causes a reduction of roadway