Comparing Storm Environments and Forecast Performance in the HRRR, RRFS, and the NSSL MPAS Models

NSSL is exploring the suitability of the NCAR Model Prediction Across Scales (MPAS) as the next-generation dynamical core in the NSSL Warn-on-Forecast System (WoFS). The WoFS is a 3-km, rapidly-updating ensemble targeted for transition to the Unified Forecast System by 2030. The MPAS is a community model that prioritizes realistic simulation of convection. The MPAS uses a Voronoi mesh that confers advantages over traditional latitude-longitude grids in both global settings (e.g., uniform resolution over globe; no polar filtering required) and regional configurations (e.g., flexible local refinement avoids abrupt transitions between nests). To assess strengths and weaknesses of the MPAS versus the Advanced Research version of the Weather Research and Forecasting model (ARW) and the Finite-Volume Cubed-Sphere model (FV3), we will compare forecast output among five deterministic models: the ARW-based High-Resolution Rapid Refresh (HRRR); EMC’s deterministic, CONUS-domain prototype of the FV3-based Rapid Refresh Forecast System (RRFS) that is tentatively scheduled to replace the HRRR and several other mesoscale model systems in 2024; and three regional MPAS models run by NSSL. The three MPAS models differ only in their initializations and microphysics schemes: the MPAS-HT-NSSL is initialized from the HRRR and uses the Thompson scheme; the MPAS-RT-NSSL is initialized from the RRFS and uses the Thompson scheme; and the MPAS-HN-NSSL is initialized from the HRRR and uses the NSSL two-moment scheme. Comparing the MPAS-HN-NSSL and MPAS-HT-NSSL will allow us to test our hypothesis that the NSSL two-moment microphysics produces better thunderstorm forecasts in the MPAS, as it does in the ARW-based WoFS. Comparing the MPAS-RT-NSSL and RRFS, and the MPAS-HT-NSSL and HRRR, will illuminate the systematic impacts of the different dynamical cores given the similarity of the physics packages used in the models.

Storm environments and forecast performance will be compared among the five models using a variety of analysis and verification methods. Effective resolution will be measured using vertical velocity and kinetic energy spectra. Probability-matched composite means will be used to visualize systematic inter-model differences in storms and near-storm environments. Where possible, the NSSL Multi-Radar / Multi-Sensor (MRMS) composite reflectivity will be used for verification. Model soundings and surface fields will be examined for systemic intermodel differences and verified using the rawinsonde and ASOS networks. This is likely to be a multi-year effort; early results will be presented at the conference.