The changing risk of extreme precipitation is difficult to project. Events are rare by definition, and return periods of heavy precipitation events are often calculated assuming a stationary climate. Furthermore, ensembles of climate model projections are not large enough to fully categorize the tails of the distribution. To address this, we cluster the contiguous United States into self-similar hydroclimates to estimate changes in the expected frequency of extremely rare events under scenarios of global mean temperature change. We find that, although there is some regional variation, record events are projected in general to become more intense, with 500-year events intensifying by 10-50% under 2 °C of warming and by 40-100% under 4 °C of warming. This analysis could provide information to inform regional prioritization of resources to improve the resilience of U.S. infrastructure.
Keywords: climate change; clustering; extreme rainfall; extreme value analysis; record precipitation.