Assessment Of The Performance Of GRAPES Model On The Convective Boundary Layer Based On The Large Eddy Simulation

This paper devotes to examining the GRAPES semi-Lagrangian dynamic core at thelarge-eddy scale, which may lay a foundation for the future development of kilometer-scaleand even more high resolution numerical model. Another purpose of this research is todevelop GRAPES large-eddy simulation model, which will provide a scientific tool fortesting and developing the boundary layer turbulence parameterization. GRAPES model hasbeen well applied in the mesoscale numerical forecast and Global medium-range forecast,but never study its performance on large eddy scale. furthermore, it is necessary to firstunderstand the ability of simulating the model on the large eddy scale when developing thevery high resolution numerical model in the future; At the same time, how to think about theturbulence will play a key role in forecasting and developing the convective boundary layeron cloud scale. It has a great significance in numerical prediction by building GRAPES largeeddy simulation model to understand the characteristics of turbulence in boundary layer andstudying the boundary layer turbulence parameterizations.In order to analyze the ability of GRAPES at large eddy scale, the Smagorinsky-Lillysubgrid closure is introduced into GRAPES dynamic core, and model resolution is increasedto be50m. However, there is a diffusion implicit in the interpolation in semi-Lagrangianadvection, they may affect GRAPES dynamic performance when used in large eddysimulation. So A widely-utilized large eddy simulation model (UCLA-LES model) is appliedhere as a reference. With these motivations in mind, a simple case of a dry convectiveboundary layer (CBL) case is adopted. The use of removing moisture in the simulationenables to focused on the dynamical core and the implied diffusion in this semi-Lagrandianmodel. Some main conclusion summarized here: the GRAPES semi-Lagrangian dynamiccore can produce a close match in the convective turbulence simulation between theexistinglarge eddy simulation. Moreover, the test reveals that, GRAPES-LES need toconsider the excessive diffusion in the GRAPES semi-Lagrangian advection scheme. Whenusing he same filter scale (Smagorinsky constant), the velocity field simulated byGRAPES-LES is more smoothly, and the small scale turbulence structures in velocity field iseither not enough obvious, the turbulent energy spectrum show this problem more clearly.Further, through a sensitivity test by using different Smagorinsky constants (corresponding todifferent filter scales) demostrates that different filtering scale can ease the implicit dissipation problem in semi Lagrangian model, and get a closer turbulent field comparedwith the UCLA-LES.