DEVELOPMENT OF A LARGE-SCALE AGENT-BASED ECONOMIC SIMULATOR FOR HIGH-RESOLUTION SIMULATION OF POST-DISASTER ECONOMIES

抄録

Although most natural disasters such as major earthquakes are spatially localized, the strong interdependencies among economic entities can make their impacts ripple through the nation for several years, producing substantial economic losses. Enhancing disaster resilience necessitates a comprehensive assessment of the long-term economic impacts of candidate recovery plans considering the complex interrelationships of economic entities, and other real-world constraints such as lifeline access, transportation, and government policies. This requires fine-grained modeling of the economy and infrastructure as an integrated system. To attain this objective, we developed a high-performance computing (HPC) extension for an agent-based economic model, making it possible to efficiently simulate large-scale economies consisting of hundreds of millions of agents at a 1:1 scale. This paper presents the validation of the developed system for the Japanese economy by reproducing the past observations at the national and sectoral levels. As a demonstrative application, we present a simulation of a post-disaster economy considering hypothetical disaster scenarios.