| dc.description.abstract | The main objective of this dissertation is to advance the models and methods utilized to model a community as a system of systems (SoS) accounting for their interdependencies. In line with this objective, this dissertation contributes to the disaster resilience literature by, first, developing a set of probabilistic models for the business recovery and residential building stock restoration, and second, proposing a modeling approach based on agent-based modeling (ABM) to develop a SoS model of a community. The model developed using the proposed approach can ultimately be used to assess the resilience of a community and make decisions to enhance its resilience. The next few paragraphs summarize the main steps in this dissertation to achieve the proposed objective.
First, a modeling approach based on Bayesian linear regression is proposed to develop predictive models for different attributes of business recovery, including cease operation days, revenue recovery, customer retention, and employee retention. This stepwise modeling approach includes three main steps namely data collection, development of model forms, and model selection. This modeling approach is applied on the data collected from Lumberton, NC, after the 2016 Hurricane Matthew and predictive models were developed for the business recovery in this community. The developed models can be further used in risk analysis studies on businesses in communities with similar characteristics as the Lumberton community. One of the notable findings of this study was the significance of housing recovery on customer retention of the businesses.
Second, an existing analytical framework to generate distributions for the recovery time of different building archetypes subject to tornado hazard was validated and calibrated in this dissertation. Because of the lack of empirical restoration time data, researchers developed an analytical framework based on the performance-based engineering (PBE) approach to develop time distributions for the restoration of different building archetypes damaged in tornado events. In this dissertation, an empirical dataset was developed using the observations from a longitudinal field study in the city of Joplin, MO, after the 2011 Joplin tornado. Time distributions developed using this dataset were compared to the outcomes from the analytical framework, and a number of modifications were proposed to calibrate the analytical framework to better represent the real-world conditions in the aftermath of tornado events.
Third, a modeling approach based on ABM is presented to develop a quantitative model of a community accounting for its interdependent systems. Agents in this context are discretized entities making their decisions based on a set of micro-behaviors, while their internal interactions form different systems in the community model and external interactions between different systems shape the complex behavior of the community. The application of this study is presented in the virtual community of Centerville by defining different agent types for its various entities, including the Electric Power Network (EPN), Water Supply Network (WSN), education system, businesses, healthcare system, households, and people. A broad review of the literature was conducted to define agents and their interactions, while verification and validation were performed to assure the credibility of the outcomes. The developed agent-based model was first utilized to assess the resilience of the education system, which was one of the least-studied components in the quantitative disaster literature. Ultimately, resilience measures were proposed for the community as well as its systems. The proposed community resilience measure, which needs input from decision makers of a community in order to be calculated, is employed to assess the resilience of Centerville in its current condition, while the effect of different mitigation strategies on the resilience of the community was evaluated using the calculated resilience measures after implementing such decisions into the quantitative community model. | |
