Evidence of changing storm intensity, duration and frequency derived from a novel CONUS-wide database of severe storms defined by their discrete space time footprints

Accurate identification of flood-relevant storm trends is vital for catastrophe modeling and disaster resilience yet remains challenging due to spatial and temporal inconsistencies in historical weather data. To address this gap, we develop a severe convective storm database catalogue at 4km hourly resolution across the Contiguous United States (CONUS) from 2002 to 2025. Using NOAA’s Analysis of Record for Calibration (AORC), we accumulated precipitation across 1-, 3, 6-, and 12-hour windows and identified extreme events exceeding half the 1-year return period based on NOAA Atlas 14. A spatiotemporal extension of the DBSCAN clustering algorithm was used to group exceedances into discrete storm events. Overlapping events across accumulation windows were deduplicated to retain the most continuous footprints. Storms were classified using a Random Forest model trained on labeled examples of thunderstorm complexes, Mesoscale Convective Systems (MCSs), frontal systems, and tropical cyclone remnants. Intensity-Duration-Frequency (IDF) curves were developed by storm type and NOAA Climate Division to evaluate trends in storm behavior and frequency. This open-access storm event database directly supports interdisciplinary catastrophe modeling efforts by linking high-resolution atmospheric processes to regional exposure and vulnerability contexts.