Mapping flood memory: How risk perception and social vulnerability drive flood insurance patterns in the U.S

Floods cause severe damage to private property in the United States, but the vast majority of flood-exposed residents are not insured. To gain detailed knowledge on how to improve insurance coverage for people in need, we investigate the temporal dynamics of flood insurance adoption patterns across U.S. counties by analyzing National Flood Insurance Program data through the lens of social memory and risk perception. Using change point detection methodologies, we identify significant shifts in insurance purchasing behavior following flood events and quantify both the magnitude ( salience ) and memory ( time-to-forget ) of these post-flood responses. We show that social memory of flood events, measured through insurance participation rates, may be considerably shorter and more heterogeneous than previously suggested in the literature. By including demographic, environmental, and institutional variables, we disentangle the multi-modal distribution of salience and memory. We identify three components driving insurance adoption: social vulnerability, risk perception, and flood damage patterns. Subsequent cluster analysis of these components reveals five distinct county profiles, with notable findings regarding areas where high social vulnerability coincides with low risk perception. The results demonstrate spatiotemporal variations in community responses to flood risks and suggest that only community-specific policy interventions can successfully maintain equitable and sustainable insurance coverage.