Numerical and experimental study on remote sensing of turbulent free-surface flows

In water flows, understanding pollutant mixing helps to predict the spread of diseases and/or harmful chemicals whilst understanding sediment transportation helps to predict where blockages may occur within the water transportation infrastructure. Both are important turbulence-driven properties that are exacerbated during floods. Past studies have shown that the surface of water flow can give information regarding the turbulent water flow underneath (although the exact relationship is still not clear). It has been hypothesised that this could also be used to deduce overall flow properties such as flow rate and turbulence level.

However, in many published open channel computational fluid dynamics (CFD) simulations, the dynamic water-air surface boundary is modelled as a "rigid lid" to reduce computational load. In essence, the small vertical fluctuations at the water surface are assumed to not exist; with only transverse and streamwise velocities permitted (slip condition). This is done even in shallow water flows where a rough bed may influence the water surface. The purpose of this research is to investigate the impact of the rigid lid assumption (if any).

To achieve this, a suite of shallow water experimental tests was designed and conducted to replicate natural flows over rough beds. A range of sensing techniques was used to measure the flow and the water surface. For the flow, stereo Particle Imaging Velocimetry (PIV) and Acoustic Doppler Velocimetry (ADV) were deployed. To mitigate the typical dynamic distortion of a vertical PIV laser plane by the fluctuating water surface, the laser plane here was projected from beneath a bed of hexagonal-packed, translucent spheres. To measure the water surface, novel 3D infrared mapping sensors from the gaming industry (Microsoft Kinect) were tested and used alongside traditional conductance wave probes.

The results obtained from these physical experiments were used to validate various CFD models developed in an open-source code called OpenFOAM. In rigid lid simulations, the flow turbulence was modelled using the Wall Adapting Large Eddy (WALE) method. The normalised velocity profiles and turbulence profiles were in good general agreement with the experiments and published literature. However, the Reynolds' shear stresses from the CFD was forced to zero at the rigid lid. This is despite turbulent structures being identified immediately beneath the free-surface through U-level, Proper Orthogonal Decomposition (POD), Quadrant analysis, Q-criteron and Lambda-2.

This suggests that the sub-surface turbulence has an important impact on the water surface behaviour and that the rigid lid may not be suitable, especially at relative submergence $<$3. The CFD model forms one of the first OpenFOAM publications on Open Channel flows over rough beds and drastically reduces the barriers to using OpenFOAM in this research field.