Storm-Based Verification and Intercomparison of Forecasts during May and June 2023 from the HRRR, RRFS, NSSL MPAS, and C-SHiELD Models

Most current convection<allowing models (CAMs) in the United States use one of three dynamical cores: the Advanced Research version of the Weather Research and Forecasting (ARW) Model, the Finite<Volume Cubed-Sphere (FV3) model, or the Model for Prediction Across Scales (MPAS). The NSSL Warn<on<Forecast System (WoFS), a rapidly updating storm<scale ensemble targeting 0–6<h lead times, uses the ARW. However, with NSF NCAR ending support of ARW, and NOAA’s transition to the Unified Forecast System (UFS), WoFS will need a new dynamical core. Given recently identified limitations of the FV3 for storm<scale prediction, NSSL began running three daily MPAS CAMs to assess MPAS as a next<generation dynamical core for both WoFS and Environmental Modeling Center (EMC)’s contiguous United States (CONUS)<domain Rapid Refresh Forecast System (RRFS). The three MPAS models differ only in mi-crophysics and driving model}the RRFS or the High<Resolution Rapid Refresh (HRRR)}and all five models share similar physics. These models, along with GFDL’s FV3<based CONUS System for High<Resolution Prediction on Earth-to<Local Domains (C<SHiELD), were evaluated and compared in the 2023 Hazardous Weather Testbed Spring Forecasting Experiment (SFE). We complement the subjective SFE findings using a storm object framework to verify and compare the models’ 0000 UTC forecasts of convective environments and storm morphology. We analyze 1–36<h forecast times over May–June 2023. Our results are consistent with the SFE findings, supporting consideration of MPAS for the WoFS and other next<generation CAMs and illuminate the systematic impacts of dynamical core, driving model, and microphysics scheme in CAMs.