Date of Award
Doctor of Philosophy (PhD)
Atmospheric & Oceanic Sciences
Turbulence is inherently chaotic and unsteady, so observing it and modeling it are no easy tasks. The ocean's sheer size makes it even more difficult to observe, and its unpredictable and ever-changing forcings introduce additional complexities. Turbulence in the oceans ranges from basin scale to the scale of the molecular viscosity. The method of energy transfer between scales is, however, an area of active research, so observations of the ocean at all scales are crucial to understanding the basic dynamics of its motions. In this collection of work, I use a variety of datasets to characterize a wide range of scales of turbulence, including observations from multiple instruments and from models with different governing equations.
I analyzed the largest scales of the turbulent range using the global salinity data of the Argo profiling float network. Taking advantage of the scattered and discontinuous nature of this dataset, the second-order structure function was calculated down to 2000m depth, and shown to be useful for predicting spectral slopes. Results showed structure function slopes of 2/3 at small scales, and 0 at large scales, which corresponds with spectral slopes of -5/3 at small scales, and -1 at large scales. Using acoustic Doppler velocity measurements, I characterized the meter- to kilometer-scale turbulence at a potential tidal energy site in the Puget Sound, WA. Acoustic Doppler current profiler (ADCP) and acoustic Doppler velocimeter (ADV) observations provided the data for an analysis that includes coherence, anisotropy, and intermittency. In order to more simply describe these features, a parameterization was done with four turbulence metrics, and the anisotropy magnitude, introduced here, was shown to most closely capture the coherent events. Then, using both the NREL TurbSim stochastic turbulence generator and the NCAR large-eddy simulation (LES) model, I calculated turbulence statistics to validate the accuracy of these methods in reproducing the tidal channel. TurbSim models statistics at the height of a turbine hub (5m) well, but do not model coherent events, while the LES does create these events, but not realistically in this configuration, based on comparisons with observations.
Each of the datasets have disadvantages when it comes to observing turbulence. The Argo network is sparse in space, and few measurements are taken simultaneously in time. Therefore spatial and temporal averaging is needed, which requires the turbulence to be homogeneous and stationary if it is to be generalized. Though the acoustic Doppler current profiler provides a vertical profile of velocities, the fluctuations are dominated by instrument noise and beam spread, preventing it from being used for most turbulence metrics. ADV measurements have much less noise, and no beam spread, but the observations are made at one point in space, limiting us to temporal statistics or an assumption of "frozen turbulence" to infer spatial scales. As for the models, TurbSim does not have any real-world forcing, and uses parameterized spectra, and coherence functions and randomizes phase information, while LES models must make assumptions about sub-grid scales, which may be inaccurate. Additionally, all models are set up with idealizations of the forcing and domain, which may make the results unlike observations in a particular location and time. Despite these difficulties in observing and characterizing turbulence, I present several quantities that use the imperfect, yet still valuable observations, to attain a better description of the turbulence in the oceans.
McCaffrey, Katherine, "Characterizing Ocean Turbulence from Argo, Acoustic Doppler, and Simulation Data" (2014). Atmospheric & Oceanic Sciences Graduate Theses & Dissertations. 2.