Quality-Controlled Databases of US Extreme Rainfall Events Based on Gridded Precipitation Estimates and Convection-Permitting Model Output

Quantitative precipitation estimates (QPEs) can vary widely for extreme events, depending on the radar algorithms used and the rain gauge observations incorporated into the estimate. QPEs also have substantial uncertainties in complex terrain owing to sharp spatial gradients in rainfall and radar beam blockage. These limitations hamper efforts to develop consistent databases of extreme rainfall events across the US.

In this work, we present a series of quality-controlled databases of US extreme rain events, based on exceedances of average recurrence interval (ARI) thresholds at durations from 1 to 72 hours. NOAA’s Stage IV analysis from 2002-2023 and Multi-Radar Multi-Sensor (MRMS) analysis from 2018-2023 are consistently processed on a common grid to remove spurious data points. The extreme events identified in these datasets are analyzed to characterize the spatial and temporal distributions of extreme rainfall, building on previous research. At timescales of 24+ hours, these databases are compared to analyses from the PRISM climate group, and from observations in the CoCoRaHS and Cooperative Observer networks. In general, the NOAA datasets that more heavily rely on radar data reveal greater coverage of extreme rainfall than PRISM, which is weighted more heavily toward gauge observations.

Lastly, the 40-year run of the Weather Research and Forecasting model known as CONUS404 is also processed and analyzed in the same manner as the QPEs. This reveals strengths and weaknesses of CONUS404: it compares favorably to observations for strongly forced weather systems such as atmospheric rivers and frontal systems, but sometimes provides a poor representation of mesoscale convective systems and smaller-scale storms. Overall, these databases will enable further research into extreme precipitation, and may serve as a useful resource for improving forecasts of extreme rainfall using post-processing.