Evaluation of an Experimental 15-Member FV3-LAM Ensemble for the Summer 2024 Flash Flood and Intense Rainfall Experiment

A 15-member experimental ensemble using the Finite-Volume Cubed-Sphere Limited Area Model (FV3-LAM) dynamic core is evaluated for precipitation forecasting during summer 2024. The ensemble was run once daily out to an 84-h lead, for a total of 30 days, in support of the National Oceanic and Atmospheric Administration's (NOAA) Hydrometeorological Testbed's Flash Flood and Intense Rainfall Experiment. It employs a variety of physics combinations, boundary conditions, radar data assimilation, and cumulus convective schemes. Evaluation via contingency table metrics suggests that activating cumulus convective schemes, in conjunction with Global Ensemble Forecast System (GEFS) boundary conditions, may have a detrimental effect on precipitation forecasts beyond day 1. Additionally, greater biases are noted in members using the Noah-MP land surface scheme than in others. Radar reflectivity data assimilation improves ensemble performance on day 1, while a realignment of features to a common position among members (spatial aligned mean) improves the performance of ensemble consensus products. An experimental Model for Prediction Across Scales (MPAS) member run alongside the FV3 ensemble shows competitive performance out to day 2, with slightly worse performance during days 3 and 4. The results helped inform ongoing Rapid Refresh Forecasting System (RRFS) development during and after the experiment. Recommendations based on these results are subsequently made for future model development and forecast experiments.