On the comparison of turbulence intensities from large-eddy simulation with those from experiment or direct numerical simulation

The relation between the Reynolds stresses from experiment or direct numerical simulation (DNS) and large-eddy simulation (LES) is reviewed. As is well known, the Reynolds stresses can only be reconstructed from a LES when the average contribution from the subgrid-scale model is taken into account. However, in the case of LES using traceless models (e.g., effective viscosity models: Smagorinsky model, dynamic Smagorinsky model, etc.), or even partially traceless models (e.g., mixed models), only the deviatoric Reynolds stresses can be reconstructed. This obvious point is often overlooked in the literature. It has important consequences in all flows with at least one inhomogeneous direction (channels, boundary layers, wakes, shear layers, jets, etc.): as far as the rms turbulence intensities are concerned, one can only properly reconstruct, and thus directly compare with experimental or DNS data, their deviation from isotropy.