Friday, June 1, 2012
Numerical simulation of stratified turbulence
Georgios Matheou, JPL
Stably stratified turbulent flows are prominent in many geophysical and engineering applications. Using direct numerical simulation, stationary and homogeneous shear driven turbulence is studied in various stratifications, ranging from neutral to very stable. To attain and maintain a stationary flow, the mean shear is throttled so that the net production stays constant for all times. This results in a flow that is characterized solely by its mean shear and its mean buoyancy gradient, independent of initial conditions. The method of throttling is validated by comparison with experimental spectra in the case of neutral stratification. With increasing stratification comes the emergence of vertically sheared large-scale horizontal motions that preclude a straightforward interpretation of flow statistics. However, once these motions are excluded, simply by subtracting the horizontal average, the underlying flow appears amenable to the standard methods of turbulence analysis. A direct acknowledgement of the confining influence of the periodic simulation domain leads to a meaningful physical interpretation of the large scales. Once an appropriate confinement scale is identified, many features, including horizontal spectra, fluxgradient relationships and length scales, of stratified sheared turbulence can be readily understood, both qualitatively and quantitatively, in terms of MoninObukhov similarity theory. As an application, the similarity-theory framework is used to interpret the scaling of the vertical diapycnal diffusivity in stratified ocean turbulence. Utilizing elements of the similarity theory, a buoyancy-adjusted extension of stretched-vortex subgrid-scale model is proposed. The model remains free of flow-adjustable parameters and is consistent with features of anisotropic mixing frequently observed in stratified flows. The extended stretched-vortex model is applied to the large-eddy simulation of a stably stratified atmospheric boundary layer. The LES predictions exhibit good resolution independence even for grids that can be considered coarse.