This paper presents a new model that predicts the number of freeway in
cidents and associated delays based on general freeway segment charact
eristics, traffic volumes, and incident management procedures. The mod
el is intended to be used in planning capacity-enhancing freeway impro
vements and incident management programs. Estimates of incident freque
ncies, severity, durations, and delays are provided for seven standard
incident types, each of which represents a significant fraction of to
tal unplanned incidents and has severity and/or duration characteristi
cs substantially different from the others. In addition to describing
the incident prediction model, the paper addresses the need for a coor
dinated national strategy for collecting incident data, with particula
r attention to urban freeways. It concludes that the incident data sys
tems that have evolved in several urban areas, often in connection wit
h freeway service patrols and incident response team activities, alrea
dy provide a valuable nationwide data resource for understanding incid
ent patterns and their variations. However, better national coordinati
on of locally collected incident data would be helpful for addressing
issues beyond the scope of the local concerns for which virtually all
current systems were originally designed. Specific areas for improveme
nt include the definitions of incident types, descriptions of incident
locations (relative to both the length and breadth of the highway), a
nd data recording the critical times during incidents such as when det
ection, response, and clearing occur.