Inter-model Storm-scale Comparisons from the 2017 Spring Forecasting Experiment

The 2016-2018 NOAA Hazardous Weather Testbed (HWT) Spring Forecasting Experiments (SFE) have featured the Community Leveraged Unified Ensemble (CLUE), a coordinated convection-allowing model (CAM) ensemble framework designed to provide empirical guidance for the development of operational CAM systems. The 2017 and 2018 CLUEs included 81 members that all used 3-km horizontal grid spacing, allowing direct comparison of forecasts generated using different dynamical cores, physics schemes, and initialization procedures. This study leverages the 2017 CLUE output to evaluate and compare the ability of various experimental and operational CAMs (initialized at 00 UTC) to realistically represent and predict thunderstorms at forecast hours 18-26. A major focus is identifying relative strengths and weaknesses of the ARW, NMM-B, and FV3 dynamical cores. The NSSL Multi-Radar/Multi-Sensor (MRMS) product suite is used to verify model forecasts and climatologies of observed variables. In the case of unobserved variables, the intermodel comparisons still prove useful for illuminating impacts of model design choices on storm prediction.

Results from a wide range of grid- and object-based metrics reveal operationally important model differences that correlate with dynamical core or can otherwise be attributed to differences in model design. For example, models initialized with direct radar data assimilation more accurately depict the diurnal cycle of storm frequency; rotating storms in the ARW models exhibit the largest northeastward motion biases; and the FV3 models have the least surrogate severe skill. The outcomes of this and similar comparisons will be crucial for improving existing CAMs and designing next-generation CAM systems.