Evacuation Behavior during No-Notice Emergency Events

The study presented in this thesis focuses on creating a disaggregate evacuation demand model for analyzing evacuation behavior in the case of no-notice emergency events. The proposed framework is designed to be compatible with the Agent-based Dynamic Activity Planning and Travel Scheduling (ADAPTS) model for the Chicago Metropolitan Area. The study develops series of statistical and machine learning models designed specifically for each part of the evacuation decision-making process. Incorporation into an activity-based model allows for pinpointing persons and resources’ location in the network, which is of most importance in the case of no-notice emergency events due to the dispersity of family members in the transportation network (which may result in additional trips to pick up family members). The models developed in this study are based on a stated preference survey that was conducted in 2012 from residents of Chicago metropolitan area. The framework provides a decision-support platform to help planners and emergency responders to first assess potential hazards, locating affected area and population, and investigate probable operability of transit systems for transit-dependent population. The framework is also suitable to investigate policies and strategies to re-deploying resources and understand evacuees’ behavior at the time of an event to direct individuals’ decision in favor of the most useful decision in order to prevent economic damages and loss of life.