Evaluation of a Cumulus Scheme in the FV3-LAM and of Experimental MPAS Forecasts for the Summer 2025 Flash Flood and Intense Rainfall Experiment

Abstract

A convection-allowing model (CAM) ensemble was run during the NOAA Hydrometeorological Testbed’s 2025 Flash Flood and Intense Rainfall Experiment. One of the goals was to evaluate the role and effects of a cumulus scheme on precipitation forecasting in CAMs, where four members of the ensemble have a scale-aware simplified Arakawa–Schubert (saSAS) cumulus scheme activated, while the other four have it turned off. The ensemble was run once daily for the 24 days of the experiment. The ensemble consisted of eight FV3-LAM members and three experimental MPAS members run on 3-km grids and employed a variety of physics combinations. Contingency table metrics and the fractions skill score indicate that activating the saSAS scheme in the FV3-LAM does not provide benefit towards summer precipitation forecasting at varying thresholds, and, in some cases, actually degrades forecast skill. The RUC land surface model interacts more strongly with the cumulus scheme than other physics schemes. Additionally, we find that MPAS members perform comparably to their FV3 counterparts on day 1, but degrade thereafter, suggesting that more evaluation of the MPAS core and optimization of the accompanying physics with MPAS is needed.