Impact of Ensemble Design in the Rapid Refresh Forecast System Using Initial Condition and Stochastic Perturbations

As NOAA transitions to the Unified Forecast System (UFS), development continues on a future convective-allowing Rapid Refresh Forecast System (RRFS) ensemble that uses a single dynamic core (dycore) and will eventually use a single physics suite. The intent is to replace the High Resolution Ensemble Forecast (HREFv3), which is NOAA’s current operational convective-allowing, ensemble forecast system and is a multi-dycore and multi-physics ensemble. To account for model, physics, and initial condition uncertainties, it is important that an ensemble system sample the full envelope of possible outcomes and thus exhibit adequate forecast spread. Although the HREFv3 generally produces sufficient spread, due to its multi-dycore and multi-physics nature its members tend to cluster by dycore/physics pairing, resulting in multi-modal statistics. An ensemble based on a single dycore and single physics suite can remedy this problem, but such ensembles tend to lack sufficient spread. In this work, we present results from an RRFS-like forecast system based on the NOAA Unified Forecast System (UFS) Short-Range Weather (SRW) App that uses the Global Ensemble Forecast System (GEFS) to generate initial condition (IC) perturbations for its members and optionally uses two types of stochastic physics perturbations – Stochastically Perturbed Parameterizations (SPP) and Stochastically Perturbed Parameterization Tendencies (SPPT). In order to assess the ability of this ensemble design to provide sufficient spread to account for all forecast error, we run a set of 36-hour forecasts initialized at 00Z on 13 days from late April to mid-May 2022, and we evaluate the impact of IC perturbations with and without SPP+SPPT on several deterministic- and ensemble-based verification metrics (e.g., bias, frequency bias, rank histogram, Brier score, reliability, AUC, spread/skill). These metrics are generated automatically by the SRW App’s workflow using the Developmental Testbed Center’s (DTC) enhanced Model Evaluation Tools (METplus) verification software. This work is part of the DTC’s efforts to inform the development of the RRFS (managers and modelers) as NOAA moves toward implementation of RRFSv1 in the near future.