A Synthetic Information Approach to Urban-Scale Disaster Modeling

We describe a large-scale simulation of a hypothetical nuclear detonation in an urban region. Simulating such a complex scenario requires modeling the population and its interactions with interdependent infrastructures such as transportation, communications, and healthcare. Our work represents the first model of a behaving human population, resolved to the individual level, where agents make decisions based on their health state, environmental conditions, and (informational) awareness. This "big simulation" approach requires a large amount of carefully curated data at the input, which is combined into a "synthetic information" model. The simulation is database-driven in a novel architecture that enables scaling, and it produces large amounts of data that in turn require advanced analytics in order to extract policy-relevant conclusions. We present results from a spatio-temporal analysis that draw out the connections between spatial variations in population behaviors and health outcomes.