Turbulent interactions between stream flow and near-subsurface flow: a laboratory approach using particle image velocimetry and refractive index matching

Turbulent flow over a rough, porous gravel bed (particle diameter, D = 1.27 cm) is investigated in the laboratory using particle image velocimetry (PIV) and refractive index matching (RIM). This combination allows for detailed examination of flow fields both in the near-bed region and within the pore spaces at two flow Reynolds numbers. In additional experiments, a channel-spanning cylinder and a single spherical grain were mounted to the top of the bed to study the turbulent flow around obstacles adjacent to a highly permeable wall. Supplementary experiments involving a mobile bed (D = 1.3 mm) in a very thin flume are also presented. With no obstacle on top of the bed, the mean velocity flow structure resembles that of a classic boundary layer, but with a significant slip velocity at the bed interface. In addition, the permeability of the interface allows for large instantaneous near-bed streamwise momentum due to vertical momentum exchange by turbulence. In the pore spaces, mean velocities are subhorizontal in direction and 5-10% of the maximum free stream velocity. High Reynolds stresses near the bed, particularly around the crests of spherical roughness elements, suggest turbulence is produced by flow separation and the shedding of vortices from streambed grains, rather than via viscous friction as in a classic smooth wall boundary layer. The geometry and dimensions of turbulent flow structures—determined via multi-point correlations of velocity fluctuations and Galilean decompositions—appear similar to those of hairpin vortices, although the resemblance remains unconfirmed without time-resolved data. The structure of the turbulent flow field is strongly affected by the addition of a cylindrical obstacle. In particular, the cylinder produces strong downwelling into the bed upstream of its location and upwelling in discrete jets downstream. The velocity directions in the pore spaces reflect this upwelling and downwelling in the stream above. Mean velocities in the pore spaces beneath the cylinder are accelerated up to 30% of the maximum free stream velocity. Downstream of the cylinder, flow separation produces a shear layer with high magnitude Reynolds stresses. However, rather than reattaching downstream, the shear layer dissipates in the outer region of flow, reaching the bed outside of the field of view, at ~11.5D downstream, with reduced Reynolds stresses. Furthermore, the streamlines do not reattach; rather, they are, like the Reynolds stresses, disturbed by upwelling fluid downstream of the cylinder, and the recirculation zone only extends 0.75D downstream and is truncated by the upwelling. Overall, the location of turbulence production is shifted away from the wall and to the height of the obstacle. Mesoscale turbulent flow structures appear to be produced by the rolling over of separated flow from the cylinder, while the dimensions of macroscale structures from upstream are strongly diminished by the obstacle. The single spherical grain produces a similar flow pattern to that of the cylinder, but the mean flow is more three-dimensional. Overall, the sphere presents less of a blockage to flow than the cylinder, and because flow can accelerate around the sphere in the streamwise-spanwise plane, the magnitude of fluid forced into the bed via downwelling and expelled via upwelling decreases by 50% in comparison to the cylinder. In addition, the magnitude of Reynolds stresses in the separated flow downstream of the single grain is 20% less than those observed for the cylinder. Similar to the cylindrical obstacle, the mean velocity streamlines and shear layer do not reattach downstream and are disturbed by upwelling flow. The recirculation zone is also affected, being rendered almost nonexistent, with a length of less than 0.25D. No standing horseshoe vortex is observed wrapping around the sphere, as often observed around hemispheres and spheres mounted to smooth walls, but rather turbulent flow structures are limited to hairpin-like vortices produced by the shear layer. The results of these experiments have important implications for hyporheic exchange and sediment transport. In particular, the transfer of momentum across the streambed interface by turbulence not only increases net hyporheic exchange in highly permeable sediments but also provides an explanation for the bedform morphology of gravel bed rivers. Overall, the experiments described in this thesis make it clear that the permeability of gravel systems must be addressed for accurate descriptions of stream flow