Convection-Permitting Model-Based Storm Catalogs and Their Utility in Estimating Extreme Precipitation Statistics

Storm catalogs containing precipitation observations from a number of historical extreme storms provide the basis for estimating extreme precipitation statistics, such as the intensity-duration-frequency (IDF) curves and probable maximum precipitation (PMP) estimates. Recent advances in convection-permitting regional climate models have demonstrated great potential for resolving extreme precipitation events and their changes at kilometer scales under global warming. These models can augment observation-based reconstructions of historical storms, especially in complex terrain areas. They also offer great promise for tracking, classifying, and analyzing storms by explicitly accounting for the physical and dynamical processes associated with these storms.

Based on the new continental-scale convection-permitting reanalysis product CONUS404 (4-km, hourly), this study presents an extreme precipitation frequency analysis that combines stochastic storm transposition (SST) with storm catalogs derived from the 42-year (1980-2021) CONUS404 data. This study focuses on exploring the utility of model-based storm catalogs in estimating IDF statistics for the city of Charlotte, North Carolina. To enhance these storm catalogs, we developed a convolutional neural network (CNN) framework to correct biases in CONUS404 precipitation, using the radar-gauge merged Stage IV precipitation data as ground truth labels for training the CNN model. The bias-corrected storms are then supplemented with other CONUS404 atmospheric fields to determine the weather systems and storm types associated with these storms. With these enhancements, the final storm catalogs are coupled with the SST technique to derive IDF statistics for different precipitation durations at various return periods. The results highlight changes in extreme rainfall statistics over the study area in the past decades, especially for 50-, 100-, and 200-year storms. By specifying storm types, we quantify the relative roles of different storm types in shaping the regional climatology of extreme precipitation. Compared with NOAA’s Atlas 14, we further discuss the potential and limitations of state-of-the-art convection-resolving climate models for understanding extreme precipitation frequency statistics and their potential changes.