In the eastern United States, the empirical probability distribution of extreme daily precipitation comprises heavy rainfall events stemming from North Atlantic tropical cyclones. At many locations, these events influence estimates of extreme value statistics (e.g., 100-year event), thus have important bearing on the sizing of flood protection infrastructure and, in general, flood risk management and preparedness. Despite the recognized importance of properly identifying the physical processes that generate extreme precipitation among mixed populations to obtain more accurate precipitation frequency estimates, the topic has received little attention. We address this issue by i) systematically characterizing the heterogeneous sub-population or mixtures in the daily precipitation extremes by separating the tropical cyclone precipitation and non- tropical cyclone precipitation, and ii) modelling tropical cyclone precipitation and non- tropical cyclone precipitation as mixed population that weight the individual extreme event-generating phenomenon appropriately. Long-term precipitation records over the period of 1951–2020 for 667 USHCN stations across the eastern United States were used for our analysis. Results from our study are important for proper characterization of extreme value distribution for prediction and attribution of extreme events in a changing climate, as well as implications for improved design and reassessment of flood-controlling infrastructure.
