Using simulated annealing to improve the information dissemination network structure of a foreign animal disease outbreak response

Abstract

Communication is an integral part of emergency response, and improving the information dissemination network for crisis communication can save time, resources, and lives. This thesis focuses specifically on emergency response to a foreign animal disease (FAD) outbreak, an incident in which an animal disease that is not active domestically is introduced and being spreading in the U.S. In a FAD outbreak, timeliness of detection and response are critical. An outbreak of foot-and-mouth disease, a particularly significant FAD, could cripple the agriculture economy and every hour of poor communication could result in the loss of thousands of animals. Improving this and other such crisis communication networks is of high importance. There is a comparatively large amount of prior research that critiques past catastrophic events but very little that aims to quantitatively improve such networks. This research uses communication data from a FAD response exercise in Kansas to develop a reliable network model, contributing a general method for creating an information dissemination network from empirical communication data. The thesis then introduces a simulated annealing heuristic to alter the network structure, reducing the overall information transmission time by almost 90%. Both the application of simulated annealing in network design and the use of discrete event simulation to calculate the heuristic objective function are new contributions to the field of crisis communication and emergency response. This work begins by extracting data from communication logs, grouping the large numbers of stakeholders into more manageable clusters, and developing a simulation model framework that accurately depicts the flow of information in the actual network. Then a simulated annealing heuristic is used to alter the network structure. The goal is to identify an alternative network structure in which the time for information to reach all response participants is minimized. The resultant network structures are analyzed to reveal observations and recommendations for FAD response communication. This research finds that not only can such a network be improved significantly, but the quantitative results support the qualitative observations from early in the data extraction process. This paper adds original methods to the literature and opens the door for future quantitative work in the area of crisis communication and emergency response.