基于大涡模拟评估GRAPES模式对对流边界层的模拟性能

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

为检验GRAPES半拉格朗日动力框架在大涡尺度上的模拟性能,为未来发展千米及其以下高分辨尺度的数值模式奠定基础,并构造GRAPES大涡模式以检验和发展边界层湍流参数化提供科学工具。通过在GRAPES模式中加入Smagorinsky-Lilly小尺度湍涡参数化,并将模式分辨率提高至50 m,构建GRAPES大涡模式(GRAPES_LES),以便分析GRAPES模式在大涡尺度上的适用性。同时利用广泛应用的已有大涡模式UCLA_LES作为参考,通过对干对流边界层湍流的模拟分析及与UCLA_LES模拟结果的对比,得出如下主要结论:GRAPES半拉格朗日动力框架能够模拟出与已有的大涡模式相似的边界层湍流特征;同时,通过分析也证明GRAPES存在由于采用半拉格朗日平流计算而带来过度耗散的问题:当使用相同的滤波尺度(Smagorinsky常数)时,GRAPES_LES模拟出的速度场更为平滑,小尺度湍流结构过于光滑,通过对湍流能量的能谱分析更清楚地表明了这一点。进一步,对不同的Smagorinsky常数(对应不同的滤波尺度)进行了敏感性试验,表明可以通过改变滤波尺度,有效地缓解半拉格朗日框架隐含的耗散问题,得到更接近UCLA_LES所模拟的湍流特征。

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.