Ten-day Weather Forecasts using a 4-45 km Variable-resolution Global Model: fvGFS

A variable-resolution Global Forecast System (GFS) using the  Finite-Volume Cubed-Sphere Dynamical Core (FV3), named fvGFS, was built at Geophysical Fluid Dynamics Laboratory, aims to replace the current operational GFS for future use. Based on 73 cases of 2015,  fvGFS with uniform grid (FV3u) shows a very close forecast skill to the current operational GFS model on the anomaly correlation coefficients (ACCs) of 500 mb height and sea level pressure when it was only adopting some minor physical parameterization adjustments. Beyond the uniform grid, fvGFS with stretched grid (FV3s) with locally refinement exhibits higher ACC than that of FV3u over the Contiguous United States (CONUS). FV3u shows a large improvement in precipitation forecasts over CONUS compared to the GFS, while FV3s can be even better in predicting severe events. For a smaller domain over southern great plains, FV3s shows significantly better performance than FV3u and GFS in simulating mean and variation of precipitation. Finally, based on a case study of Hurricane Joaquin, FV3s shows superior capability depicting the hurricane structure and intensity by simply increasing the resolution.