Assessment of the simulative performance of the GRAPES model on the convective boundary layer based on the large eddy simulations

This paper devotes to examining the GRAPES semi-Lagrangian dynamic core at the large-eddy scale, which may lay a foundation for the future development of kilometer-scale and even more high resolution numerical models. Another purpose of this research is to develop GRAPES large-eddy simulation model, which will provide a scientific tool for testing and developing the boundary layer turbulence parameterization. In order to analyze the ability of the GRAPES at large eddy scale, the Smagorinsky-Lilly subgrid closure is introduced into the GRAPES dynamic core, and model resolution is increased to be 50 m.A widely applied large eddy simulation model (UCLA_LES model) is applied here as a reference. Through simulating the turbulence in a dry convective boundary layer, the following conclusions are obtained: The GRAPES semi-Lagrangian dynamic core is able to show a reasonable performance at the large-eddy scale by using a smaller filtering scale comparing with the UCLA_LES. Also, it is revealed that there exists excessive diffusion in the GRAPES semi-Lagrangian model due mainly to the interpolation at upstream points of Lagrangian advection. When using the same filtering scale (Smagorinsky constant) as UCLA_LES, the velocity field simulated by the GRAPES_LES exhibits much smoother distribution, and the fine turbulent structures look not so obvious compared with those by the UCLA_LES. This is also confirmed by analyzing the turbulent energy spectrum. Moreover, sensitivity experiments by using the different Smagorinsky constants demonstrate that smaller filtering scale can compensate for the implicit dissipation in the semi-Lagrangian model, and can produce a good convective turbulence simulation.