A Data-Driven Foresighted Agent-Based Model for Municipal Level Projection of Urban and Rural Migration Response to Recurrent Flooding in Virginia

Worsening changes in climatic stressors such as sea-level rise and flooding, together with socio-economic trends such as changing population, urbanization, and resource extraction present great challenges for coastal communities. Despite the attempts to adopt different adaptation approaches, recent research suggests the potential for major climate change impacts, including human displacement in coastal areas, by the end of the century. Coastal municipalities can receive more support by forward-looking modeling analyses to develop risk-informed planning. Most data-driven approaches assume that future changes will follow historical trends, which is not necessarily true, particularly in the context of climate change. In addition, they often neglect the differences among people's behavior and decision making in the face of different types of climate impacts. This transdisciplinary study presents a novel approach for future flood-induced migration that considers household-level heterogeneous decision-making details while being parametrically based on long-term population projections. To do so, we consider urban and rural coastal areas in Virginia and Maryland as one of the most affected regions by sea-level rise and flooding, and propose a framework for stochastic agent decision rules that can replicate georeferenced U.S. county-level population projections by Socioeconomic Data and Application Center (SEDAC). This framework is used to investigate flooding impacts on migration decisions using physical flood-exposure data for extreme and nuisance flooding from the North Atlantic Coast Comprehensive Study (NACCS). The results demonstrate that the proposed approach not only incorporates the heterogeneity in human behavior into flood-induced migration assessment, but also considers a forward-looking source of behavior in tuning its parameters.