摘要：Abstract Given the rapidly changing climate, accurate spatiotemporal information on the evolution of extreme rainfall events is required for flood risk assessment and the design of resilient infrastructure. Consequently, various research efforts have focused on investigating the appropriateness of various parametric and non‐parametric approaches in modeling the observed changes in the frequency of extreme rainfall over time. Yet, the assumption of stationarity, or the change of model parameters when accounting for nonstationary rainfall, may magnify estimation uncertainty of rain rates associated with low exceedance probabilities. Moreover, the use of climate model results may yield inconclusive outcomes, given the existence of epistemic uncertainties in the frequency of extreme events developing on smaller spatial scales or over complex terrain. Herein, we employ a parametric approach based on multifractal scaling arguments, along with high‐resolution (4‐km) hourly precipitation estimates covering a 40‐year period over CONUS, to derive Intensity‐Duration‐Frequency curves and investigate the spatiotemporal evolution of extreme rainfall over a wide range of characteristic temporal scales and exceedance probability levels. Considering the robustness of the multifractal models even when fitted to short rainfall records, we uniquely apply the framework to sequential 10‐year segments of data, where the rainfall process can be reasonably assumed stationary. The obtained results reveal that existing infrastructure may be severely impacted by the intensification of precipitation extremes due to climate change, with the observed trends being significantly influenced by the topography and rainfall climatology of each region, while depending on the averaging durations and return periods of interest.
