Lagrangian observations and numerical modelling of hydrodynamics, turbulence, and sediment transport in a tidal river
Permanent link to Research Commons versionhttps://hdl.handle.net/10289/16120
Fluvial-to-marine transition zones in rivers play a critical role in transporting materials, including nutrients and sediments, from land to ocean. These riverine environments located close to estuaries constitute complex dynamical systems, which are affected by the influences of both tidal variations from downstream and freshwater inputs from upstream, with the interaction of multiple physical processes controlling velocities, salinities, mixing, and sediment transport. As such, accurate numerical modelling of the hydrodynamics and sediment movement in these regions can prove challenging. However, it is essential to address this challenge as models are now regularly utilised as a tool to underpin coastal management projects, such as predicting pollutant dispersal, assessing navigability, or undertaking hazard management. Turbulence has been shown to influence many processes involved in the movement of particles and, particularly, sediment transport. Turbulence is notably a key control of flocculation processes (aggregation and disaggregation) of fine cohesive particles in suspension, in particular in marine settings. The variability of turbulence along a stretch of river affects the size of flocs, and thus also the settling velocity and movement of sediment within the water column. In this thesis, I use numerical modelling to explore hydrodynamics, turbulence, and sediment transport processes within the tidally driven, heavily sediment-laden and meandering Kaipara River, New Zealand. In particular, I investigate the performance of Delft3D-FLOW in modelling flow speeds, turbulence, and sediment transport along a ~15-km stretch of the tidally influenced river in a Lagrangian (flow-following) frame of reference by comparing predictions to a unique Lagrangian dataset which offers exceptional spatial resolution of quantities along the river. Specifically, I examine (1) the ability of the model to reproduce the Lagrangian observations with the commonly used k-epsilon turbulence closure scheme, (2) a comparison of model predictions of hydrodynamics when two turbulence closure schemes are implemented, namely k-epsilon and k-L and (3) the Lagrangian dynamics of sediment transport along the river. Lagrangian observations and modelling of turbulence along a tidally influenced river Lagrangian datasets of hydrodynamic and sediment variables were collected along a 15 km-stretch of the river during the ebb tide using `FlocDrifter' platforms, deployed from different locations. Two platforms were also fixed in an Eulerian frame-of-reference, collecting time series upstream of the river and towards the middle of the studied domain. Delft3D was used, with the two-equation k-$\epsilon$ turbulence closure scheme to simulate flows along the river. The model calibration was classified as excellent when predictions were compared with Eulerian measurements; however, on comparison of model predictions with the Lagrangian observations, discrepancies were revealed. Overall, the model could predict flow speeds and the general patterns and the right order of magnitude of dissipation rates of turbulent kinetic energy epsilon along the river. Nevertheless, the model did not always correctly reproduce the observed epsilon, particularly around abrupt meander bends. While such errors in epsilon could be partially connected to errors in predictions of speeds, the omission or the lack of accuracy of other processes (e.g., wind-driven mixing, secondary flow) were likely to explain some of the errors in model predictions. Overlaps in the drifter tracks indicated that the bathymetry and the geometry of the river were the primary controls on the along-river structure of flow and turbulence. The vertical and cross-sectional distributions of dissipation rates of turbulent kinetic energy, turbulent kinetic energy and vertical eddy viscosity were examined and showed that, generally, turbulence varied more in the vertical direction rather than the longitudinal direction, in agreement with previous studies. Overall, results demonstrated that the flow-following Lagrangian observations allowed identification of variability across multiple length and time scales, which would not necessarily be captured using an Eulerian frame of reference. Moreover, the comparison with Lagrangian observations offers a more stringent validation of model results over large (riverine) spatial scales than a traditional Eulerian approach. Evaluation of the performance of two turbulence closure schemes in a Lagrangian frame of reference along a fluvial-to-marine transition zone The well-known `closure problem' of turbulence means that numerical computation of turbulent quantities requires assumptions and simplifications of the flow dynamics. Statistical approaches, which have been found to be relatively computationally efficient, are commonly used. These approaches are based on the Reynolds-Averaged Navier-Stokes (RANS) equations, which are usually first simplified, and then solved through use of a turbulent closure scheme. This work compares numerical model predictions of flow speeds and dissipation rates of turbulent kinetic energy to a unique set of Lagrangian observations, collected along a meandering tidal river (the Kaipara River in the North Island of New Zealand). Two turbulence closures schemes (the k-Land the k-epsilon schemes) are implemented in the model, with both producing very similar predictions. Indeed, both the one-equation k-L and the two-equation k-epsilon turbulence closure schemes could reproduce the overall along-river structure and the right order of magnitude of flow speeds and epsilon. Nevertheless, the smaller-scale patterns were not accurately captured by either closure scheme, with larger differences to observations occurring at bends. Moreover, the k-L turbulence closure scheme performed slightly better than the k-epsilon turbulence closure scheme in predicting the along-river changes in dissipation rates of turbulent kinetic energy. This result suggests that, while most modelling studies use the `default' more sophisticated formulation for turbulence closure, simpler closure schemes can perform similarly and sometimes better, hence saving computational resources. Controls on the flow-following distribution of suspended sediment concentrations in a tidally driven river Sediment movement, deposition, and erosion within rivers and coastal environments are complex physical processes which vary widely depending on forcing regimes; however, accurate representation of these processes in numerical models is crucial to predict the geomorphological evolution of these regions. We use observations and numerical modelling to explore the controls on the distribution of suspended sediment concentration along a tidal river. Model runs with variable implementations of advection, erosion, and deposition processes revealed that the main source of suspended sediments was resuspension from the river bed, rather than advection from upstream which formed a smaller secondary contribution. The comparison with the high-resolution Lagrangian dataset provided a particularly strenuous test of model performance, and the model was found to do an excellent job of reproducing the rate and magnitude of the downstream increase in SSCs, especially when considering a number of simplifications made in model setup and boundary conditions. While patterns were very similar, small (5%) differences in the predictions of SSC were found between the k-epsilon and k-L closure schemes. Abrupt local increases in SSC were observed in transitions from bends to straight sections of the river, and conversely, sudden local decreases in SSC were observed on the entrance to bends. These flow-following changes in SSC over medium spatial scales (O(km)) were found to be strongly correlated with shear velocities within the model. However, some of the km-scale variability seen in the observations was not reproduced by the model, with differences in these cases attributed to other physical processes which were not incorporated (e.g., additional freshwater inputs and wind- or wave-driven resuspension). Model predictions remained remarkably reliable even for locations where flocs were present in the observations, despite not including a parameterisation for the flocculation process. Results indicated that conditions were sufficiently energetic (shear velocities -> floc settling velocities) to prevent settling of even the larger flocs.
The University of Waikato
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.
- Higher Degree Theses