UC San Diego

This work utilizes various computational techniques to study the turbulent mechanisms found in stratified shear flows. Three-dimensional DNS was used to investigate the influence of stratification on turbulence and mixing within a shear layer between two currents. Similarities in the development of secondary instabilities during transition to turbulence and discrepancies in flow evolution are seen between the case of uniform stratification considered here and the two-layer density profile of prior works. Vertical contraction of the shear layer is identified in cases with low Richardson number and determined to be the result of the flattening of Kelvin-Helmholtz billows before the flow becomes fully turbulent. Transition layers with enhanced shear and stratification form at the periphery of the shear layer and are found to support turbulent mixing. In an effort to find a less computationally costly tool than DNS, the Dynamic Smagorinsky, Ducros, and WALE subgrid-scale models were chosen for an LES study of the stratified shear layer. This investigation revealed the Ducros model to the least computationally costly LES option and the most reliable with coarsening grid resolution. A subgrid analysis revealed the LES models to be largely unsuccessful in capturing convective turbulence though the mean flow and turbulent kinetic energy were well-captured.

To address the limitations of DNS and LES, a hybrid spatially-evolving DNS model was developed. The wake of a sphere towed in a stratified background was selected for validation. The hybrid model involves extracting planes from a spatially-evolving, body-inclusive simulation and feeding the planes as inflow into a body-exclusive simulation thereby eliminating the need for a highly resolved grid to capture flow near the body. This study revealed that particular attention should be paid to the extraction location, grid resolution, and time step between extractions. Planes must be extracted downstream of the recirculation region behind the body and sufficient grid resolution is required in the body-exclusive simulation to capture small-scale turbulence. Results show the hybrid DNS model to be an effective tool in the study of the stratified turbulent wake. The combination of results presented herein offer computational techniques and cost-saving options for future studies of shear flows.