Recent Improvements in Rapid Refresh and High-Resolution Rapid Refresh Forecasts of Severe Convective Storm Initiation

The 13-km Rapid Refresh (RAP) and 3-km convection-allowing High-Resolution Rapid Refresh (HRRR) are hourly-updating forecast models that support severe convective storm forecasting and other short-range forecasting interests. This study examines the performance of the RAP and HRRR in forecasting storm initiation preceding several recent severe thunderstorm events, and finds that instances of excessively widespread storm initiation in previous RAP and HRRR versions may be partly attributed to the overremoval of convective inhibition within these models.

To improve RAP and HRRR forecasts of daytime convective initiation, experimental modifications to the Mellor–Yamada–Nakanishi–Niino (MYNN) planetary boundary layer (PBL) and the Grell–Freitas–Olson (GFO) shallow-convection parameterization schemes are tested. These modifications are found to better simulate the effects of nonlocal mixing, thereby facilitating the retention of capping inversions and convective inhibition atop deepening boundary layers. Furthermore, the improved coupling of subgrid-scale clouds to the RRTMG radiation scheme is found to produce additional forecast skill in the RAP and HRRR.

Results from retrospective simulations of selected severe thunderstorm events will be shown to illustrate these performance improvements in the RAP and HRRR. Finally, ongoing physical parameterization work at ESRL will be summarized.