Vertical diffusion and cloud scheme coupling to the Charney–Phillips vertical grid in GRAPES global forecast system

Abstract

AbstractSpatial aspects of the physics–dynamics coupling in the Global and Regional Assimilation and Prediction System (GRAPES) for global medium‐range numerical weather prediction (GRAPES_GFS) are studied. As a Charney–Philips (CP) grid is used for the dynamics but all physical processes are calculated on a Lorenz grid in GRAPES_GFS V2.2 and its previous versions, interpolation has to be used for potential temperature and moisture between full and half levels in the physics–dynamics coupling. Besides interpolation error, a computational mode appears in the solutions of the vertical heat diffusion equations. An idealised test using a simple one‐dimensional heat conduction equation shows definitively that inconsistency of vertical grid in the dynamics–physics coupling and an improper boundary condition can together produce the computational mode. Based on 1st‐order K‐closure, a modified parametrization scheme for implementation of the planetary boundary layer (PBL) scheme on the CP grid (PBL_CP) has been developed, together with a cloud scheme implementation on the CP grid (CLOUD_CP), such that there is no need for interpolation for both the PBL and cloud scheme coupling to the dynamics. By using the PBL_CP scheme in a case‐study, the computational mode in potential temperature and moisture predictions is shown to have been successfully eliminated and the corresponding vertical profiles appear to be reasonably smooth. Because of the improvements in potential and absolute temperature and moisture predictions, the prediction of low‐level cloud water has also been improved greatly. Meanwhile, the prediction of water vapour at high levels is more reasonable with CLOUD_CP. Consistent with these improvements in a case‐study, an overall and significant enhancement is found in 8‐day forecasts of absolute temperature, moisture, vector wind and stratocumulus with the revised model in a 4D‐Var (four‐dimensional variational) cycle experiment over a period of 3 months.Key points A revised PBL scheme called PBL_CP was developed on a Charney–Phillips grid in the GRAPES_GFS model to avoid interpolation in the coupling of vertical heat diffusion to the dynamical equations. Implementation of the PBL_CP scheme in GRAPES_GFS eliminates the computational mode in potential temperature and moisture prediction in the PBL. The revised model containing PBL_CP and cloud implementation on the CP grid has improved the GRAPES_GFS forecasts of absolute temperature, moisture, vector winds and stratocumuli significantly.