An agent-based model for large scale parallel simulation of natural disasters in towns and cities : civilians and power

Abstract

Natural disasters are a major cause of damage to the populace and infrastructure of cities and towns, affecting nearly 250 million people per year. Most frameworks for modeling the effects of disasters focus only on the areas a disaster will hit or on a wide-scale view of an area. The downside of these approaches is that they fail to account accurately for the human factor. Our approach to modeling these disasters uses an agent-based model to simulate events down to the level of the individual person, to allow for a more accurate simulation of the exact effects of a disaster. Additionally, the agent-based model allows us to see what effects changing evacuation plans and populace preparedness has on the overall loss of life in a disaster. This thesis is meant to provide a useful and fast parallel framework to aid in simulating disasters and optimizing contingency plans.