UC Irvine

For a diverse area of research, spanning from financial markets to weather and climate systems to experiments conducted in turbulent flows, the most common form of data belongs to the category of time series. In the context of turbulence research, the time series analysis techniques have mostly focused on spectral approach, where Fourier wavenumbers or frequencies are associated with eddy length or time scales. In this thesis, I propose an event-based approach as an alternative to the spectral one. This approach is used to address three important but fundamental problems in wall turbulence. First, identification of coherent structures in the flow from single-point time series measurements. Second, unravelling the scales associated with intermittency in wall turbulence. Third, the quantification of small-scale anisotropy in wall-bounded turbulent flows. To address these three objectives, different novel time-series analysis techniques have been introduced. Regarding the first objective, I introduce a level-crossing method and show how specific features of a turbulent time series associated with coherent structures can be objectively identified, without assigning a priori any arbitrary threshold. For the second and third objectives, I introduce a scale-dependent event framework, where the turbulent time series is considered to consist of a chronicle of events with finite size and duration across multiple scales of the flow. Overall, this dissertation provides a novel contribution towards data-driven modelling of turbulent flows with widespread applications, especially for atmospheric systems exhibiting complex temporal interactions at scales from seconds to decades.