Turbulence in open channels : an experimental study of turbulence structure over boundaries of differing hydrodynamic roughness

River channels modify their boundaries by entrainment and transfer of sediment. The flow represents a turbulent boundary layer which must be examined to understand the relation between channel morphology and its modifying flow. Discrete roughness elements with variable spacings were used, in flume experiments to simplify the study of the relation between flow resistance and boundary morphology. The patterns were intended to model gravel bed channels. The results are compared with wind tunnel and atmospheric boundary layer studies In order to examine the influence of the free surface and the applicability of the extensive results in air to open channel flow. Streamwise and vertical velocities were measured throughout the depth of flow at constant Reynolds number over a rough boundary. The roughness density was modelled by plastic blocks and the hot film measurements were subsequently digitally analysed. Density is defined as the ratio of the surface area of blocks, in plan view, to the total area of the bad. Results on six densities between 1/8 and 1/80, based on plan areas, are presented in detail. Three layers were identified in the flow: an outer shear layer, a wake layer extending to two roughness heights above the bed, and a wall layer below the top of the roughness elements. No extensive range of linear correlation of shear velocity with the logarithm of a roughness density function was found, possibly due to the surface between the roughness blocks being hydrodynamically rough. The greatest resistance to flow was presented by density 1/12: at higher densities 'skimming' flow occurred indicated by a shift in the 'bed' to the top of the roughness elements. The shape of the energy spectrum varied across the flow depth, the free surface acting as a boundary constraining the range of inertial transfer. Most of the turbulent energy was contained below 7.25 Hz while the cospectrum of u and v indicated that most of the stress lay below that frequency. No significant turbulent transfers occurred beyond 5 Hz while the coherence was below. 0.2 beyond 10 Hz. No significant inertial subrange was detected in the spectra or the cospectra. The moments of the velocity derivative indicated a strongly intermittent process and exhibited a marked change at the free surface, the kurtosis there being nearly twice that found elsewhere in the flow. The probabilistic structure of the velocity fluctuations showed that the series were non-Gaussian in form and not strictly self-similar. All the series yielded Hurst coefficients of approximately 0.8. An explanation of this result is postulated in terms of the superimposition of several intermittent processes. Measurements in the flow approaching a single block indicated that vorticity amplification occurred. When obstacle width was increased to one third of channel width the isolated block behaved as a two dimensional obstacle (i.e. like a bar). Results from measurements in four natural gravel channels indicated that the flume reproduced successfully the spectral and probabilistic structure of the flow. The results are applied in a phenomenologica1 explanation of the shape of the Shields entrainment function and in an explanation of the differences in entrainment and transport between sand and gravel. The results on the flow around an isolated block are applied to the construction of a statistical flow stability model for gravel entrainment. The hierarchy of processes {low frequency 'waves', bursts, wake shedding and dissipation), operating intermittently and acting non-linearly may be viewed as instabilities which are responsible for the generation and dissipation of turbulence. Such instabilities are discussed in relation to the hierarchy of morphological features observed in natural channels.Arts, Faculty ofGeography, Department ofGraduat