Large-eddy simulation of the convective atmospheric boundary layer

Large-eddy simulations of a free convective atmospheric boundary layer with an overlying capping inversion are considered. Attention is given to the dependence of the results upon the various factors influencing the simulation: the subgrid model, the domain size, and the mesh resolution. By providing artificial constraints upon the convection the results also provide extra insight into the underlying dynamics. The gross features of the boundary layer, such as the overall energy budget, are not sensitive to the details of the simulations but a number of important factors are revealed. It has been found that near the surface the subgrid diffusivity must be larger than is usually supposed, in order for the vertical velocity skewness to have the correct sign. This region of the flow has a significant subgrid-scale heat flux and it seems that the subgrid model requires improvement in such cases. A revised model which under statically unstable conditions allows the mixing length of the subgrid-scale turbulence to depend on the flow stability is found to give improved results. The domain size and mesh spacings have a significant influence upon the results and need a setting which allows resolution of the main, freely occurring scales of motion. The entrainment at the capping inversion is remarkable in its insensitivity to all factors. Finally, the higher resolution simulations provide a detailed view of the flow structure of the convective boundary layer. Downdrafts cover a large fraction of the surface area, and near the surface the flow converges into smaller areas comprising long narrow regions of updrafts. The plumes which penetrate through the depth of the boundary layer to the inversion mainly occur over the intersections of these long narrow regions of updrafts.