A Diagnostic for Evaluating the Representation of Turbulence in Atmospheric Models at the Kilometric Scale

Abstract

urbulence is well represented by atmospheric models at very fine grid sizes, from 10 to 100 m, for which turbulent movements are mainly resolved, and by atmospheric models with grid sizes greater than 2 km, for which those movements are entirely parameterized. But what happens at intermediate scales, Wyngaard?s so-called terra incognita?Here an original method is presented that provides a new diagnostic by calculating the subgrid and resolved parts of five variables at different scales: turbulent kinetic energy (TKE), heat and moisture fluxes, and potential temperature and mixing ratio variances. They are established at intermediate scales for dry and cumulus-topped convective boundary layers. The similarity theorem allows the determination of the dimensionless variables of the problem. When the subgrid and resolved parts are studied, a new dimensionless variable, the dimensionless mesh size , needs to be added to the Deardorff free convective scaling variables, where h is the boundary layer height and hc is the height of the cloud layer. Similarity functions for the subgrid and resolved parts are assumed to be the product of the similarity function of the total (subgrid plus resolved) variables and a ?partial? similarity function that depends only on . In order to determine the partial similarity function form, large-eddy simulations (LES) of five dry and cloudy convective boundary layers are used. The resolved and subgrid parts of the variables at coarser grid sizes are then deduced from the LES fields.The evolution of the subgrid and resolved parts in the boundary layer with is as follows: fine grids mainly resolve variables. As the mesh becomes coarser, more eddies are subgrid. Finally, for very large meshes, turbulence is entirely subgrid. A scale therefore exists for which the subgrid and resolved parts are equal. This is obtained for in the case of TKE, 0.4 for the potential temperature variance, and 0.8 for the mixing ratio variance, indicating that the velocity structures are smaller than those for the potential temperature, which are smaller than those for the mixing ratio. Furthermore, boundary layers capped by convective clouds have structures larger than dry boundary layer ones as displayed by the scaling in the partial similarity functions.This new diagnostic gives a reference for evaluating current and future parameterizations at kilometric scales. As an illustration, the parameterizations of a mesoscale model are eventually evaluated at intermediate scales. In its standard version, the model produces too many resolved movements, as the turbulence scheme does not sufficiently represent the impact of the subgrid thermal. This is not true when a mass-flux scheme is introduced. However in this case, a completely subgrid thermal is modeled leading to an overestimation of the subgrid part.