Date of Award

Spring 2012

Document Type


Degree Name

Doctor of Philosophy (PhD)


Civil & Environmental Engineering

Committee Director

Asad Khattak

Committee Member

Mecit Cetin

Committee Member

R. Michael Robinson


The objective of this study is to understand the nature of primary and secondary traffic incidents, assess their impacts and explore the implications in traffic operations, safety, and planning. To achieve the objective, a queue-based secondary incident identification method was developed and applied based on detailed incident, traffic and geometric data sets from Hampton Roads, Virginia. This identification method can overcome the limitations in earlier studies and identify secondary incidents in both road directions. An innovative event categorization defines the term "primary-secondary incident event", as one characterized by a primary incident and one or more associated secondary incidents in both directions to capture traffic impact and incident adversity.

To observe distributing pattern differences of primary-secondary incident events, two major interests: event frequencies in different categories and durations of primary incidents have been analyzed spatially and temporally. Frequencies of primary-secondary incident events and duration distributions of primary incidents both show considerable spatial and temporal differences across different event categories. The hotspots (i.e. locations that have higher frequency of primary-secondary incident events) were identified.

To understand the occurrence of primary-secondary incident events, two proportional odds models were estimated to explore associations with various factors. In particular, the partial proportional odds model can relax parallel lines assumption and capture unequal contributions of explanatory variables across the event categories. The model suggests that with multiple-vehicle involvement, lane-blockage in a primary incident makes unequal contributions to the occurrence of different primary-secondary incident events, and they are particularly prone to multiple secondary incidents.

This study sought to answer how soon does a secondary incident happen after a primary incident; how far is the secondary from the primary incident; and what factors are associated with near versus far secondary incidents. The appropriate methods and models have been developed to examine the spatio-temporal patterns of cascading incident events and identify associated factors. Time gaps were found to be positively associated with crashes, longer duration of primary incidents, and heavier traffic. In terms of distance, primary crashes, fires, lane-blockage and longer duration are associated with secondary incidents that occur at longer distances after its primary incident. The study found that distance and time vary systematically with characteristics of primary incidents.

Regarding the clearance time of primary-secondary incident events, the event duration is defined and such events were further categorized as either contained events (i.e. clearance time of the secondary is earlier than that of primary incident) or extend events (i.e. clearance time of the secondary extends that of primary incident). The associated major factors were estimated and identified through rigorous statistical models. These two types of events show substantially different incident characteristics and operational response patterns. Primary incident characteristics are dominant in contained events while secondary incident characteristics play a substantial role in extended events, requiring substantial resources from response agencies.

To quantify the total delay associated with primary-secondary incident events, the joint impacts of primary and secondary incidents have been taken into account. Shock wave analysis and microscopic simulations were used to understand and evaluate the associated critical parameters. Three critical contributing factors were evaluated: time gap, physical distance and traffic demand level. The analysis shows the traditional method which treats each incident independently will over- or under- estimate the actual delay of primary-secondary incident events. For those secondary incidents that end after their associated primary incidents, total delays increase as time gap increases and distance decreases.