Predicting Excessive Rainfall with Large and Small Ensembles: Does Adding Members Add Value?

Operational convection allowing ensemble systems are typically run with 10 or fewer forecast members. This configuration decision results in a high chance of observed events occurring outside the entire envelope of solutions, particularly in cases of intense, short-duration rainfall. A 31-member ensemble of Model for Prediction Across Scales (MPAS) forecasts is run for several mesoscale flash flooding events in the eastern United States during 2022 to 2024. The complete set of members and subsets thereof are examined to assess 1) how the forecast PDF varies with respect to ensemble size, 2) attributes of spread (e.g., location, timing, intensity) in large and small ensembles, 3) whether any value added by a larger ensemble in these events is sufficient to justify the added computational cost. Results will be presented in the forms of both quantitative statistical analyses and qualitative assessment of how forecasts from larger ensembles might improve decision support services in high-impact events.