Assessing RRFS vs. HRRR in Predicting Widespread Convective Systems over Eastern CONUS
Joe Grim - RAL/NCAR/UCAR
This study focuses on the skill of the operational High Resolution Rapid Refresh (HRRR) version 4 and its eventual successor, the experimental Rapid Refresh Forecast System (RRFS) model (summer 2022 version), at predicting the number and coverage of widespread convective events over the eastern U.S. during summer 2022. Thirty-two widespread convective events were selected using observations from the Multi-Radar/Multi-Sensor (MRMS) composite reflectivity: eight events from each of four different mature convective modes: MCSs, QLCSs, clusters, and cellular. These cases had similar size distributions to the entire summer, albeit skewed toward more storms more than 400 km2, due to the focus on widespread convective events. For each event, an evaluation area was manually selected to include all the observed and model storm objects throughout the observed lifecycle of the system. The events were assessed on four primary statistics: total storm area, total storm count, storm area ratio (an indicator of mean storm size), and storm area PDF, each normalized to weigh all storm events equally.
Both HRRR and RRFS overpredicted total storm area coverage, regardless of the composite reflectivity threshold used to identify convective areas, with RRFS overpredicting more than HRRR, although storm convective coverage evolution was better represented by RRFS. Both models underpredicted storm counts using a 35 dBZ threshold, while at 40 dBZ RRFS storm counts were very nearly identical to observations and HRRR underpredicted. Again, RRFS had better timing of the evolution of storm counts. Both models (especially RRFS) overpredicted storm area ratio, especially for the lesser organized systems.
1 окт 2024
