Evaluation of real-time, storm-scale forecasts over the conterminous Unites States produced with global MPAS configurations
Dr. Craig Schwartz
NCAR
17 December, 2pm
Abstract:
In this presentation, I will discuss real-time forecasts produced at NSF NCAR between 2023 and 2025 that used global configurations of the Model for Prediction Across Scales (MPAS).
I will primarily focus on verification statistics of real-time, 5.5-day MPAS ensemble forecasts that had 3-km horizontal cell spacing over the conterminous United States and 15-km horizontal cell spacing over the rest of the globe. Objective verification indicated that probabilistic 3–5-day MPAS forecasts of severe weather (based on updraft helicity) were skillful relative to climatology and better than severe weather forecasts issued by the Storm Prediction Center. Further objective verification revealed that for 3–5-day forecasts, the 3-km MPAS ensemble produced better depictions of precipitation placement than coarser ensembles (like the GEFS) only during weak synoptic-scale forcing, with lesser benefits of 3-km ensemble forecasts during strong synoptic-scale forcing situations.
Additionally, I will briefly overview real-time, 60-h, deterministic MPAS forecasts that NSF NCAR produced on a globally uniform 3-km mesh in spring 2025. This demonstration was unprecedented, marking the first ever real-time, storm-scale (3-km), global numerical weather prediction model forecasts. Plans for future real-time global storm-scale numerical forecasts at NSF NCAR will also be discussed.