Observation-based probabilistic reanalysis of storm surge and sea level extremes for the United States

Reliable probabilistic estimates of storm surge and sea level extremes at gauged and ungauged locations with robust uncertainty quantification are key for successful risk assessment and cost-effective adaptation, now and in the future. However, existing observational estimates are often unavailable or uncertain along most coastal regions because of data scarcity. Here, we provide a fully observational-driven probabilistic dataset of storm surge extremes for the U.S. coastline covering 1950–2020. Non-stationary extreme storm surge distributions at gauged and ungauged sites are estimated by analyzing the U.S. hourly tide gauge data network using a spatiotemporal Bayesian Hierarchical Model. In addition, an extended version of such dataset is also produced where many additional storm events are incorporated to account for key extremes not contained in standard tide gauge records. The resulting distributions are combined with deterministic tidal data to estimate return periods of extreme sea levels and their uncertainty. Comparisons against published estimates are provided indicating that storm surge and sea level extremes along most U.S. coast have been underpredicted. The data provided can support coastal managers to make more confident decisions, particularly along many U.S. coastal regions that are vulnerable and where long-term in-situ water level monitoring is limited or non-existent.