The Impacts of the Grell-Freitas Scheme on Short-Range Forecasts of the April 19, 2023 Convective Event

The Unified Forecast System (UFS) is NOAA’s community-based Earth modeling system that includes the Short Range Weather (SRW) Application, which serves as the foundation for NOAA’s next-generation regional Rapid Refresh Forecast System (RRFS). A known issue with the RRFS is its tendency to produce excessively strong convection. To counteract this bias, convective parameterizations like the Grell-Freitas (GF) scheme have recently been implemented within the SRW App. This study compares forecasts with and without GF for the April 19, 2023 convective event with the aim of determining how well the GF counteracts the RRFS's high convective bias. This event was chosen because it posed numerous challenges for weather forecasters due to the presence of marginal synoptic forcing and uncertainties in the strength of a capping inversion. Forecasts were initialized at 12 UTC on April 19 using initial conditions from the High Resolution Rapid Refresh (HRRR) and lateral boundary conditions from the Rapid Refresh (RAP). Qualitative analysis shows that in the No-GF forecast, there is spurious convection in eastern Texas/Oklahoma before the initiation of severe weather in Oklahoma, while in the GF forecast, there is no such spurious convection in this location during this period. In each forecast, supercellular convection initiates at the same time as observations, but is shifted southwest, potentially due to errors in initial conditions. In both forecasts, these storms track across central Oklahoma, but the No-GF forecast predicts weaker storms that dissipate much earlier than what was observed due to cold pools in eastern Oklahoma from the earlier spurious convection. The GF forecast aligns better with observed rainfall than No-GF while also having lower frequency bias for most rainfall thresholds, reflectivities greater than 40 dBZ, and echo tops above 30 kft. The GF forecast also has slightly more skill for these forecast variables at higher thresholds, although results are noisy due to just one simulation being run for each physics suite. These results show that GF counteracts the high convective bias of RRFS for this case, indicating that implementation of GF within NOAA’s CAM-based modeling systems may improve forecasts and help operational forecasters when dealing with marginally convective environments.