Experiments and numerical simulation of three-dimensional turbulence in gravity currents

Gravity currents are a ubiquitous and crucial class of geophysical flow, being a key driver of sediment transport in rivers and oceans. The body typically forms the largest part of such flows, yet body structure remains poorly understood. Research into gravity current structure has primarily focused on the head of the flow in unsteady lock-exchange type currents (due to the highly turbulent nature of the head, and the simplicity of the lock-exchange setup). The work presented consists of experimental and numerical investigations into the structure of constant-influx solute-based gravity currents. Particle image velocimetry, particle tracking velocimetry (Shake-the-Box), and direct numerical simulation are used to generate instantaneous whole-field two- and three-dimensional velocity measurements. These are used to discuss large-scale structures within the flow. Results question several common assumptions regarding gravity current dynamics. Through application of Fourier transforms, wavelet transforms, and dynamic mode decomposition, empirical data (from both particle image velocimetry and Shake-the-Box) reveals internal waves, sometimes associated with three-dimensional motions, within the current body. These waves are shown to form a critical layer near the height of the velocity maximum. Wave breaking at this critical layer has the potential to limit dilution of the lower part of the flow, and accelerate the flow downstream at the height of the critical layer. The presence of these waves therefore questions the accuracy of extant models assuming a statistically steady body. Existing numerical research concerning gravity currents has almost always assumed a Schmidt number of approximately unity. Using direct numerical simulation, it is shown that some flow features (such as the presence of structures in the upper part of the body) are highly Schmidt number dependent. Further, it is demonstrated that the difference in Schmidt number may explain the structural differences between the experimental and numerical components of this work.