Tom Narock, J Michael Johnson, Arash Modaresi Rad

Urban flooding causes billions of dollars in damages annually, with severe flood events becoming more frequent and destructive as our climate changes. While extreme weather is a primary driver of flooding, its consequences depend on the interconnectedness of urban systems - referred to as the Urban Multiplex, which includes the power grid, transportation network, natural surface water and groundwater systems, sewerage and drinking water systems, intertwined with the socioeconomic and public health sectors. One component of this multiplex - a reliable building
inventory - is critical for assessing the number of people affected by flooding, the propagation of shocks throughout the economy, and for forecasting detailed socioeconomic risk from flooding. Yet, a major discrepancy exists in the way data about buildings are collected across the U.S. There is no harmonization in what data are recorded by city, county, or state governments, let alone at the national scale. We demonstrate how existing open source datasets
can be spatially integrated and subsequently used as training for machine learning models. These machine learning models can then predict building occupancy type, a currently lacking component of flood risk assessment. Multiple
machine learning algorithms are compared and an application to the 100-year flood in North Carolina is provided. Results indicate that a 100-year flood will disproportionately impact Mecklenburg, Wake, Dare, and Brunswick counties.