Abstract
Direct Numerical Simulation (DNS) has been employed with success in a variety of oceanographic applications, particularly for investigating the internal dynamics of Kelvin Helmholtz (KH) billows. Most previous studies have focused on the evolution of local mixing efficiencies, the quantification of local stresses, and the stability of turbulence, among other topics. However, it is difficult to relate these results directly with observations of ocean turbulence due to the significant scale differences involved (ocean shear layers are typically on the order of 10s to 100s of meters in thickness, compared to DNS simulations, with layers on the order of cms to 10s of cms). An alternative focus of DNS models is to calculate mean values of turbulent quantities (e.g. production, buoyancy flux, and dissipation) throughout the simulation, and compare those results with measurements of naturally observed similarly parameterized flows. This can broaden the parameter space of available measurements to enable the development of enhanced closure techniques and can provide more context to the studies of internal KH dynamics, Previous DNS studies often utilize a fixed domain length and initial velocity perturbations or noise in their initial conditions. Noise is usually used to expedite turbulence and suppress secondary instability. However, a systematic investigation of the effects of varying domain length and the use of initial randomized velocity noise has not been investigated. In this thesis, simulation results show that variation of the domain length can have profound impact on the number of KH billows generated, their temporal evolution, and wavelength. The results validate an expression of KH wavelength derived from the dynamics of internal waves. Results of simulations with varying noise fields indicate that the amplitude of the noise is secondary in importance to the structure of the random field (uniformly distributed) in affecting the resulting turbulence.The use of any random noise in the initial conditions is shown to delay the onset of turbulence, with the structure of the random filed having a more significant impact on the magnitude of turbulent quantities and the duration of active turbulence. Together, these analyses will provide valuable guidance for the initialization of future DNS simulations.