Evaluating MPAS Configurations for Convective-Storm Forecasting

The National Severe Storms Laboratory (NSSL) and the Global Systems Laboratory (GSL) are developing regional configurations of the Model for Prediction Across Scales (MPAS) for convective-storm forecasting in the Warn-on-Forecast System (WoFS) and Rapid Refresh Forecast System version 2 (RRFSv2). The NSSL has been running forecasts daily over a CONUS domain with 3-km cell spacing since early 2023. Subjective evaluations during the 2023 and 2024 Hazardous Weather Testbed - Spring Forecasting Experiment (HWT-SFE) periods and objective metrics computed by the GSL and the Storm Prediction Center (SPC) have shown comparable skill to operational convection-allowing models (CAMs).

Spurred by these positive results, the NSSL and GSL have explored ways to further improve MPAS convective-storm forecasts. Deterministic forecasts with 6 different configurations, currently with varied initial conditions and varied suites of physical parameterizations, are run daily in real time, complemented with objective and subjective evaluations. The initial conditions for these forecasts come from the operational HRRRv4, the operational RAPv5, and the experimental RRFSv1. An experimental MPAS ensemble was also run in spring 2024.

During the past year, we have been increasingly focused on retrospective experiments for developing MPAS for convective-storm forecasting. Proposed updates to physical parameterizations are being tested for a May 2024 retrospective period with numerous severe weather events. We are also evaluating forecast sensitivity to namelist parameters, including microphysics parameter settings and numerical filtering. At the conference, we’ll report on the major forecast sensitivities that have been identified and summarize forecast performance for convective storms and their environments for our current model configurations.