Evaluation of FV3-SAR Initialized by Multiscale Hybrid EnVar Analyses for Convection-Allowing Hazardous Weather Forecasting

The Finite Volume Cubed Sphere Model (FV3) has recently replaced the GFS as the new dynamical core within NOAA’s Next Generation Global Prediction System (NGGPS). Additionally, the FV3 has become the foundation for a unified modeling framework at NOAA that will be designed for applications across all time and spatial scales. As part of the unification and consolidation effort, the FV3 stand-alone regional (FV3-SAR) model that runs the FV3 dynamical core on a regional grid with external lateral boundary conditions, for applications to convection-allowing modeling (CAM), has been developed. The FV3 will eventually replace the dynamical cores for regional model and analysis systems such as the North American Mesoscale Model (NAM) and the High-Resolution Rapid Refresh (HRRR) model.

Given these recent developments, the FV3-SAR is now being integrated with the operational convective-scale Gridpoint Statistical Interpolation (GSI)-based hybrid ensemble-variational (EnVar) data assimilation (DA) system through collaboration between NOAA EMC, GSD, and the University of Oklahoma’s MAP laboratory. In this study, we will present results on the efforts of this integration into our ensemble DA system. There are two main components to the testing of the FV3-SAR into the hybrid EnVar system. First, extensive tests are conducted comparing 10-member free forecasts of the FV3-SAR initialized by interpolated final analyses produced by the current optimal configuration of the convective-scale EnVar system (using the WRF-ARW dynamical core). These forecasts will be compared with forecasts produced by the WRF-ARW core over a set of ten retrospective cases. The goal is to evaluate the FV3 for skillfulness of precipitation forecasts, identify areas of improvement, and understand its behavior for ensemble-based CAM initialized by a full convective-scale ensemble analysis. Several experiments testing different configurations of the FV3-SAR model are performed to find an optimal configuration; for example, testing different physics suites, stochastic physics, and model dynamics options. The second component of this work is the implementation of the FV3-SAR core during the DA cycling period for our convective-scale EnVar system. This integration includes further tuning of DA parameters including localization and inflation, as the change to FV3-SAR introduces different model error characteristics that will need to be reflected in the DA parameters. These experiments will be conducted over a set of ten retrospective cases and compared to the WRF-ARW configuration. Additional efforts and results of advancing the hybrid EnVar in FV3-SAR context are planned to be presented in the conference.