The influence of surface roughness and shape on the submerged floating tunnel is investigated in wave and current conditions. The hydraulic forces caused by varying tunnel shapes and roughness designs are measured and analyzed.

The ramifications of plastic pollution on the environment are becoming increasingly serious in various forms throughout the world. In this context, rivers are the most important suppliers of plastics entering the marine environment. However, rivers that contain high loads of plastic waste also directly harm the livelihoods of people living near these rivers. An example is the flooding of urban areas in Indonesia due to clogging of the hydraulic and drainage systems, which is caused by the blockage of local hydraulic structures by plastic debris.
At this moment, there is a lack of knowledge on this accumulation process and its underlying dynamics, since observational and experimental research is lacking. Numerical modeling has proven to be a great tool for expanding experimental research. However, no suitable numerical method has been identified yet to model the plastic accumulation process, since traditional mesh-based CFD numerical methods are expected to be not a viable option, due to their inability to model the individual interaction between plastic particles, critical during this process. A possible solution could be the SPH-DEM method, which is a two-way coupled numerical approach that simulates fluid and debris as discrete particles and elements.
The objective of this report was to find out if SPH-DEM could be a suitable numerical method to model the dynamic processes of the plastic debris accumulation against hydraulic structures. To accomplish this, the first goal was to realistically model a turbulent open-channel flow and the buoyancy of individual plastic debris, which would be validated by experimental research. The second goal was to investigate which are the most important (numerical) parameters affecting the mentioned plastic debris accumulation.
In this report, experimental research was carried out in the form of buoyancy tests and flume tests, and numerical research was carried out in the form of the design of numerical simulations. In the buoyancy tests, the rising velocities of four plastic fragments that differed in size and density were measured, which were released multiple times in a graduated cylinder filled with water. In the flume tests, first the water elevation was measured along the the flume, after which the passing ratio’s and carpet lengths were measured for the four different released fragments for three different gate configurations. Two types of numerical models were designed that represented both types of experimental tests, for which several design choices had to be made to compensate for several physical phenomena, which can’t be directly represented in the model design.
The numerical buoyancy test was validated with the rising velocities obtained from the experimental equivalent. It was discovered, that for relatively low resolution modeled fragments, the rising velocity is heavily influenced by numerical diffusion. The smoothing length was identified as an important numerical parameter, which can compensate this effect. Furthermore, it was discovered that the degree of numerical diffusion is dependent on the depth of the fragment in the water. The numerical flume test was validated with the water elevation obtained from the experimental equivalent. For uniform flows, by adjusting the boundary viscosity coefficient, smooth turbulent velocity profiles could be simulated throughout the flume corresponding to theoretical values. However, no single value of was found in which the velocity profiles of the uniform flow and the validated water elevation of the gradually varied flows were both in agreement with their theoretical values. After validation, fragments were added to numerical flume model. Per fragment type and gate configuration,
four different numerical scenarios were executed, where each scenario was defined by a combination of a certain density ½s and restitution coefficient e. Finally, the best corresponding scenarios were used to simulate mixed fragments released in the flow. It was found that the gate opening height , density and restitution of the plastic fragments have the largest influence on the passing ratio’s , carpet length , carpet shape and carpet stability. Furthermore, it was confirmed that individual fragment interactions play a crucial role in the accumulation process. However, the model is mainly limited by its low resolution and the absence of suitable turbulence models. This means that many forms of fragment behavior seen in the experimental research such of buoyancy, trajectory and individual interactions, which are heavily influenced by turbulence, cannot be sufficiently represented in the numerical model. However, it is shown that by adjusting the density the buoyancy behavior can be partly replicated and by adjusting the restitution coefficient the turbulent individual interactions can be partly replicated. In conclusion it can be stated that SPH-DEM is an interesting option to model the dynamic processes of the accumulation of plastic debris against a sluice gate; however, further improvements in computational power and turbulence models are needed to be more widely applied.

Currently multiple projects investigate the effect of lowering the crest of groynes in the Dutch rivers with the aim to lower the backwater effect of groynes under high water levels. There is a major uncertainty however in the modelling of groynes, and therefore in the prediction of the effect of groyne lowering. Currently submerged groynes are modelled subgrid as weirs. The flow processes differ however between groynes and weirs as discharge over a groyne is not purely defined by the contraction and expansion over the groyne, but also by lateral shear and interaction with the surrounding main channel and floodplain. The result is a lower resistance of the cross section than expected from weir modelling. For this reason a numerical investigation was started, comparing the modelling of groynes as different subgrid weirs in a 2D model and the modelling of groynes included in the bed topography in a 3D (non-)hydrostatic model. There were large differences between all different methods of groyne modelling. Furthermore the available experimental data did not suffice to explain the differences.
For that reason a new physical model has been set up in the Delft University of Technology, at the hydraulic laboratory of the Faculty of Civil Engineering and Geoscience. The experiment includes a 1:30 scale model of a representative transect of the Waal river. The model is 5 m wide and 30 m long and includes 6 groynes and 5 groyne fields. It further includes a part of the main channel at one side and a part of the floodplain at the other side of the groyne fields. Gravel is fixed to the bed to ensure hydraulically rough flow conditions. In the experiment many measurements were done. The first aim of the measurements was to obtain a good spread of points to validate numerical models on. The second aim was to gain insight in the different flow processes in order to describe the differences between groyne and weir flow and quantify the resistance groynes have on a flow.
An important observation was the existence of a region of low flow, at the tip of the groyne. The observation indicates a complex three-dimensional flow which effectively redistributes discharge over the transverse. The contribution of this flow to the two-dimensional momentum balance is then directly as a secondary circulation and indirectly by alternating the distribution of discharge. Complex flow furthermore invalidates the hydrostatic pressure assumption. The observed flow needs further investigation with either more accurate measurement devices or in a 3D non-hydrostatic model. Another observation includes the contraction and expansion of flow over the groyne crest. The expansion of flow over the main body of the groyne seems comparable to the expansion of flow over a weir, which can be predicted as Carnot head losses. The observation of expansion losses is however clouded by other flow processes, such as bed shear stress, the lateral flow around the groyne and possible three-dimensional processes induced by the groyne tip, so that the observed head loss over the groyne is only partially explained by weir-like expansion losses.
The comparability between groyne flow and weir flow does not seem to hold at the groyne tip, where no separation of flow is observed. Measurements were performed 10 cm away from the transition between the groyne tip and main body of the groyne. This means that at full scale there is a region of 3 m on either side of this geometrical transition where the transition lies between weir-like overflow and non-weir-like overflow.
With the here performed simulations it is possible to adapt weir head loss formulas on the observed head losses over groyne. It should be possible to validate a three-dimensional non-hydrostatic model, which can further quantify the effect of three-dimensional flow and differentiate the different contributions to groyne head loss. Based on these models proper tuning parameters can be chosen to represent the complex flow in a two-dimensional model and separately model the weir-like expansion.

Lighthouses have guided seafarers for centuries and still play an essential role in navigating marine vessels and bringing them to safe harbours. With climate change, the already harsh ocean environment may become more severe. It is expected that future wave conditions can compromise the structural integrity of the lighthouses with their violent impact. The hydrodynamic processes of wave transition before, at, and after the point of impact are complex, highly non-linear, and turbulent. There is no exact description of these processes. Therefore, there is a need to further study them in order to better understand the broken wave load and its effect on the structural integrity of lighthouses. From the very beginning, wave breaking processes are complex. For instance, prior to impact with shoal-mounted cylindrical structures, waves undergo a rapid transition from deep to shallow water. During this transition the waves shoal, start to break, and once broken they hit the cylindrical structure of the lighthouse. It is therefore necessary to approximate the hydrodynamics. Since, it is difficult to measure the processes using only physical models. Consequently, this research focuses on broken wave load analysis using a process-based numerical model. An approach widely used for analysing wave forces on slender cylindrical structures is calibrated for breaking wave loads. Critical factors for determining these loads are the curling factor and slamming coefficient. However, there is still no scientific consensus regarding the exact values for these parameters. Further research, like this study, can help in determining these values. The main goal of this thesis is to assess the pressure distribution on the structure and quantify the values for the curling factor and slamming coefficient for a shoal-mounted slender cylindrical structure. The STORMLAMP (STructural behaviour Of Rock Mounted Lighthouses At the Mercy of imPulsive waves) project's experimental data of Dassanayake et al. (2019) is used as a benchmark case to validate the proposed computational framework. In addition, the performance of two different Volume of Fluid methods, isoAdvector and MULES, are assessed. This study demonstrates that the curling factor and slamming coefficient are significantly different for the broken wave load compared to the breaking wave load. The curling factor is found to be twice as large in the case of a broken wave load. On the other hand, the peak slamming coefficient is five times smaller. Moreover, the isoAdvector solver in comparison to the MULES solver produces results that better resemble the benchmark case results. Although the two solvers do not produce the same peak impact forces, they do both produce similar values for the aforementioned critical parameters. Also, it was found that the isoAdvector solver is faster and more efficient than the MULES solver in the case of a 3D simulation. From the results it can be concluded that the chosen numerical framework represents the benchmark case well, and that the methodology used for breaking wave loads can be applied for broken wave loads with adjustment of the curling factor and slamming coefficient.

Many coral reef islands are low-lying, which in combination with population growth, sea level rise and possibly more frequent extreme weather events is likely to result in increased coastal risk (e.g. Storlazzi et al., 2015). On smaller scales of O(10 km) wave-driven coastal inundation can be accurately predicted with advanced models such as XBeach (Roelvink et al., 2009), at already high computational costs. For larger scales, larger number of islands, for scenario modelling, and for implementation in early warning systems, computationally faster methods are needed. Reduced physics models, which neglect some of the processes (e.g. non-hydrostatic pressure gradient term and viscosity), are a potential solution. However, their accuracy and the best method to force them has not been established. In this research we propose a new methodology to model wave-driven flooding on coral reef-lined coasts. A look-up-table (LUT), composed of XBeach model runs, is combined with a reduced-physics model, SFINCS (Leijnse et al., 2021), to achieve high accuracy predictions at limited computational expense. The LUT consists of pre-run 1D XBeach simulations for several reef profiles from Scott et al. (2020), forced with different offshore wave and water level conditions. Wave conditions close to the shore as predicted by the LUT are used to force SFINCS which then simulates the wave runup, overtopping and flooding. These are forced in SFINCS using random wave timeseries from an interpolated parameterized wave spectrum following Athif (2020). The accuracy of the method is investigated for 6 distinctive cross-shore profiles from Scott et al. (2020), for two wave scenarios (gentle swell and stormy conditions). Results of complete XBeach simulations are compared to LUT-SFINCS simulations with different boundary forcing locations. The sensitivity analysis shows that the most optimal boundary location to initialize the SFINCS model is at a water depth of 0.5 m, shore-ward of the reef edge. Interpolation of the forcing conditions at the boundary is investigated with 9 different interpolation methods. Results reveal that the most accurate method to interpolate spectral parameters (the amount of high frequency wave energy, the amount of low frequency wave energy and the frequency peak of high frequency part of the spectrum) and wave setup at the boundary is the Inverse Distance Weighting method with a power of -2. Errors introduced by the interpolated parameterized boundary conditions lead to runup estimation errors of 10-20% on average with the maximal errors up to 60%. Simulating wave runup with XBeach LUT – SFINCS couple leads to about 50-times higher computational speed compared to the XBeach model simulating hydrodynamic processes along the entire reef profile.

The combination of continuous sea level rise, and more frequent extreme river discharges will become a rapidly developing problem for the Rhine-Meuse delta over the course of the next century. The current Dutch delta management policy mostly revolves around the consistent improvement of the existing dikes to account for the increasing water levels in the region. An alternative solution to this problem is the Delta21 project. The Delta21 project proposes the construction of a large scale storage lake, and a high capacity pumping station at the Dutch coast. The project is meant to allow for the simultaneous closure of the storm-surge barriers on the Dutch coast, while actively discharging river water that would normally be pilling up behind the closed barriers. The Delta21 project foresees it will be able to save up to 5billion euros by the end of the century in dike reinforcement costs. In order to enable the flood safety operation of the Delta21 project, the project also requires the construction of a new storm-surge barrier in the Haringvliet ebb-tidal delta, roughly 10km away from the current Haringvlietsluices. The objective of this thesis is the conceptual design of the new Delta21 storm surge barrier, the New Haringvliet barrier. The aim of this design process is to identify critical design components, and requirements that could potentially be hurdles in the development of the New Haringvliet barrier, as well as the overall Delta21 project. For the hydrodynamic design of the barrier, a schematized 1D hydrodynamic model was used to get a first insight into the impact the new barrier will have on the on the tidal range in the tidal delta. As well as how much it could potentially impact the reintroduction of the tide to the Haringvliet. The reopening of the Haringvliet would be a part of Delta21's effort to restore some of the local ecological value to the region as a compensation measure for the project. Additionally, the model was also used to determine the effects of the construction of the barrier on the capability for the system to handle high river discharge rates. The findings of the assessment show that the selection of a small tidal opening in the New Haringvliet barrier will significantly negatively impact the capability for the system to discharge high river water, as well as limit the amount of tidal range that will be available in the system. It was determined that constructing the barrier with a tidal opening with an effective cross-section of 10000m2 will provide the best balance between hydrodynamic performance of the system, and barrier construction costs. An analysis of the possible Delta21 construction sequences was used to determine the preferred construction method for the barrier. In this analysis, an economic construction sequence alternative, and an environmental construction sequence alternative were developed, and consecutively evaluated using a multi-criteria analysis. This MCA revealed that the construction of the New Haringvliet barrier using prefab construction elements and techniques was the desired construction method for the project. The results of hydrodynamic and construction sequence assessment, were used as the basis for the barrier's spatial design. The structure was designed with a total of 20 vertical lift gates, each spanning 80m. This would bring the total tidal opening to 10400m2. Vertical lift gates were chosen for the design due to their great reliability and overall cost effectiveness, which are both important considerations for a barrier of this magnitude. On the southern side of the barrier a shipping lock will be the barrier's primary navigational opening. The final phase of the conceptual design focuses on the design of the gated barrier section, and includes the validation of the transport stability of the caisson, the design of the vertical lift gates, and the design of the scour protection and it's filter layers.

Because of the fortunate absence of great flood events in The Netherlands during the second half of the twentieth century, no physical data was added to the structural understanding of flood fragility on walls. Considering that during the same period the cavity wall became standard in the construction of houses, a certain inconsistency can be recognized between the expected and observed (structural) behaviour of a modern wall during flood events. Nevertheless, failure, or collapse, of residential buildings is included in the Dutch flood protection standards as an important cause of damage and one of the main reasons for loss of life. It is generally believed that the developments in terms of building materials translate to a safer residence i.e. less prone to flood actions. This consensus is also found in the trade-off for certain mitigation measures, especially in areas where horizontal evacuation is difficult. An improved understanding of the flood fragility of a modern cavity wall is thus needed to prove the general consensus or to adjust the current mitigation measures to comply with the allowed fatality risks.

A window-featured cavity wall section was constructed at the Flood Proof Holland facility, using calcium silicate bricks, fired clay bricks, and a weaker mortar to partly account for the virgin effect of any newly-built wall. Both inner and outer walls were connected with adequate wall ties. The wall section was subjected to several hydrostatic pressures at both sides. These experiments were performed to physically grasp the deformations corresponding to certain flood scenarios affecting cavity wall sections from ordinary terraced houses. Additionally, the effect of the window was investigated; both on the stability and its contribution to the water height inside a residence.

Computations showed that the cavity wall in a one-way bending configuration starts to show significant cracks between 1.3 and 1.6 meters of outside water level. Because of the brittleness of the masonry, this would imply failure. It was further found that the non-linear deformations would reach 4 millimeters. Considering that for these water levels the internal moments were still far from their maximum capacity, the results suggest that failure occurred due to cracks that were forced to form because of the deformations. This indicates that modern cavity walls are still quite vulnerable for floods and their flood actions. The influence of the hydrostatic pressure, however, can be decreased considerably by a water level inside the residence that acts as a counter force. Contrary to what was expected, the window does not contribute to this inside water level, since its leakages turn out to be negligible. To keep the damages to a minimum and preserve the overall stability, it is advised to seal the residence to a height of 1.0 meter; floods that exceed this sealing should not be countered anymore and rather be allowed to enter the residence.

One of the preferred features of plastic is its high durability. This merit is its main disadvantage in the natural environment. When poorly disposed of, plastic waste can enter rivers from surface waters and subsequently end up in the ocean. Here, plastic spreads out over the entire ocean and ultimately ends in ocean gyres. Plastic material in the ocean causes environmental damage. Marine animals can get entangled in plastic debris, plastic accumulation occurs when consuming seafood and the material spreads toxins in the seawater. It has also caused economic damage: marine plastic accumulation led to €5 to 18 billion to key economic sectors for 87 coastal countries in 2020 (Deloitte, 2021). Attempts have been made to capture buoyant plastics from the ocean with interception devices, with mixed success.

 

The placement of an interception device in a river is thought to be a better approach to the problem. A standard interception design is that of a floating boom. Plastic debris is transported by the river flow and these objects guide buoyant plastics into a collection tray. Currently, these interception designs are still in their infancy. A difficulty in the optimisation of these interception devices lies in the limited knowledge of the trajectories of plastic waste. Computational fluid dynamics (CFD) modelling can help to create a better understanding of these aspects. This tool can provide a rapid assessment of different flow conditions with different types of plastic debris modelled. Further knowledge into this field can ultimately help into better retention of plastics in rivers.

 

Plastic modelling for buoyant particles has been performed in the marine environment (Van Utenhove, 2019) and in the river (Van Welsenes, 2019) with a Lagrangian study. However, both works did not perform a validation study. The combination of a CFD model and the results of physical model tests can be valuable in removing uncertainties related to plastic trajectories and the modelling of these.  One physical model test performed on plastics is the release of plastic particles in a flume (Zaat, 2020) Small films were released in a 2DV physical model, representing plastic bags. It was found that these particles followed an inverted Rouse profile for spherical particles times a shape factor.

The study by Zaat (2020) was replicated in the CFD modelling environment of OpenFOAM in this thesis. Subsequently, the retention of plastic particles was investigated. Three hydrodynamic conditions were investigated, with uniform inlet flow velocities of 0.10, 0.55 and 0.90 m/s. The $k-\epsilon$ turbulence model was used. An Euler-Euler approach was applied with a discrete and continuous phase. This modelling approach is computationally less expensive compared to a Euler-Lagrangian framework and can be easily established for different cases. The drag models Gibilaro, Wen-Yu and Syamlal-O'Brien were tested to investigate which concentration profile best fits an inverted Rouse profile for rising particles. The interception device was modelled as a square obstruction of 10\% of the water depth and the model was assumed as a rigid-lid. The retention of particles by the system was explored for neutrally buoyant particles, light material and high-density polyethylene (HDPE) and polypropylene (PP) films.

 

This study found relative differences in near-surface flow and the analytical approach of 3 to 4%. The Gibilaro model was best applicable and model coefficients for the virtual mass, lift and turbulent dispersion of 0.5, 1.6 and 1.0 were found. The concentration profiles followed the inverted Rouse profile and that for films closely for the medium and high flow cases, but not for the low flow case. A lower retention of plastics was observed for increasing flow velocities. In the extreme case of low flow and light particles a build-up of particles near-surface can be observed and runs against the streamwise direction.

Several limitations were present in this study. Only two parameters of the plastic particles were adjustable, which were the particle diameter and density. In reality, plastic debris has different shapes will experience drag differently, especially the friction drag. The rigid-lid model is valid only if the velocity head is smaller than that of 10% of the water depth, which is not the case for the high flow velocity case.

 

The results show that on the basis of a rise velocity and a density a representative diameter can be determined from Stokes’ law. The Euler-Euler modelling approach provides an accurate assessment and gives a proper representation of the particles for the Rouse distributions. This is done without handling the particles individually and thus results can be found more quickly. Both for HDPE and PP inverted Rouse profiles were found that followed the theoretical profile with an underestimation of 10\%-20\% near-surface. The retention was influenced mainly by the size of the wake. It is evident that the interception system can work for buoyant particles. The object can be placed during low flow conditions for the highest efficiency. For high discharge in rivers, more particles will be in suspension and thus further away from the water surface, which makes this design less preferred during high flow conditions.

 

Recommendations for further research are to perform more physical model tests, implement a non-spherical drag model applicable for plastic material and to apply a three-phase model. This removes uncertainties related to the model and to the current knowledge in particle dynamics. The addition of physical model tests would benefit the current knowledge the greatest. It is only possible to expand the complexity of the CFD model with a larger availability of these physical model tests.

The Meuse river has been made navigable for ships by the construction of a cascade of seven weirs. At only two weirs hydropower facilities exist. This implies the full hydropower potential of the Meuse cascade is not utilized. Rijkswaterstaat only provides additional permits when fish friendly turbines are used. By using pump-turbines the river sections upstream the weirs might be usable as energy storage reservoirs.
The objective of this thesis is to study if the full hydropower- and energy storage potential of the Meuse cascade can be utilized by the use of pump-turbines within Dutch environmental regulations. In order to meet the objective a conceptual design of a pumped-storage hydropower plant has been compiled. By using this design, an assessment of utilizing the full hydropower- and energy storage potential of the cascade has been carried out. For the assessment a numerical model has been constructed and applied.
The cumulative installed turbine capacity for the cascade is 81 MW. The annual energy yield on hydropower is 225 GWh. The annual energy yield on pumped-stored power is 77 GWh. Up to 75.000 households can be powered with hydropower and pumped-stored power. When the Juliana canal is included in the assessment up to 131.000 households can be powered using an installed capacity of 134 MW.

The aim of this research is to find the influence of rapid drawdown on dike stability with the Delta21 project as case study. The Delta21 project is a proposed plan for a waterdefence system in front of the Haringvliet, The Netherlands. Its main element is a water retaining lake with an extensive pumping and turbine system. Whenever other storm surge barriers close, water from inland rivers can be pumped to the sea. The water retaining lake can act as an energy storage system as well by using the turbines to generate electricity at moments of high energy prices. The pumping system has the capacity to reduce the water level with 17.5m in 12hours. The dike surrounding the lake has a gradual slope as the height/width ratio below the water line is 1/20. It consists fully of a homogeneous and permeable fine sand. The estimation of material properties were based on limited CPT data, samples from the adjacent Maasvlakte II and empirical relationships. The rapid decrease of the water level in the lake causes rapid draw down conditions in the dike. Hereby the external hydrostatic pressure is reduced due to unloading effect of removing water, while a delay in the dissipation of pore pressures inside the dike is produced. This causes stress relaxation and slope instability. The research was performed using a numerical model verified with physical modeling in the form of centrifuge tests. For the numerical model the finite element model PLAXIS2D was used. A fully coupled flow deformation analysis was chosen, which can solve the full interaction between deformation, consolidation and groundwater flow simultanously in the same phase. It considers a reduced permeability and degree of saturation in the unsaturated zone. The Hardening Soil small strainmodel was selected as constitutive model due to the potential of capturing hysteresis in the stress-strain regime. Hysteresis could occur as a result of multiple cycles of water level fluctuations. To gain better understanding of the capabilities and weaknesses of the numerical model, it was verified with several physical tests from literature. The first test is a centrifuge test with failure due to increasing hydrostatic pressures within the dike. The moment of occurrence and shape of the failure contour were compared with the numerical model. The second test was a centrifuge test where a dike experiences rapid draw down conditions with the purpose to validate the pore pressure and saturation level. Thirdly, a full scale test of a dike with a changing water level was added, because full scale tests offer the closest resemblance to situations in the field. The field test was modeled to double check pore pressures and saturation. The research aimed to further investigate the development of a failure during rapid draw down, a centrifuge test was performed at 100g. Rapid draw down conditions were simulated on a sand dike with slope 1:1.6 to generate failure. By tracking grains with Particle Image Velocimetry (PIV), correctness of estimating displacement by the numerical model was verified. The addition of pore pressure sensors could give further information on pore pressure development. The numerical model proved to accurately predict positive pore pressures and the moment at which macroscopic failures occur due to drawdown. Negative pore pressures, saturation and development of the failure, in terms of size and deformation could not be accurately approximated. With understanding of the capabilities and weaknesses of the numerical model, the dike of the Delta21 project was modeled. The original design proved to be stable with a FoS of 6.20. Increasing slope angles were modeled resulting in a decreasing FoS. The influence of hysteresis was found to be very small after modelling multiple cycles of changing water level. The positive effect on stability the addition of a high permeable filter layer was modeled. Finally the risk of erosion due to outflow was analytically calculated and proved to be closely related to the numerical results.

Rapid urbanization has altered the natural hydrological cycle of the Tub Ma catchment, inducing major floods after heavy rainfall events. For that reason, structural mitigation measures were built, including floodwalls along the channel and a pumping station at the outlet. However, no operational rules were available for operating the river outlet structures and inundation was still common. To prevent or reduce flooding by improving the operational management of the outlet pumping system, an early warning system was required. The objective of this research was to assess whether informed decision-making resulted in an operational pumping strategy that prevents or reduces fluvial flooding. First, the question was asked whether water level monitoring could result in a timely and reliable warning service for an early warning system. The second question was whether timely operation of the outlet pumping station prevented fluvial flooding. For the fulfillment of the data requirements in this thesis, a fieldwork trip was organized first. During a short period of dedicated data survey, water level time series and river geometries were obtained. Water levels from three gauging stations were statistically tested as precursors for fluvial flooding in Tub Ma. Consequently, a hydraulic model was built to assess the flood mitigation capabilities of the outlet pumping station. Lastly, synthetic flood design hydrographs were run through the model for different pumping station operation strategies, with and without notification from a flood early warning system. The results from this thesis demonstrated that monitoring water levels in the upstream tributaries of Tub Ma was a useful information source for an early warning system that forecasts flood waves. This water level based early warning enabled a maximum potential lead time of 13-15 hours for flood mitigation. Nevertheless, connecting flood early warning to the operation of the pumping station did not reduce or prevent flooding. The Tub Ma river had a flow regime of limited baseflow and rapidly rising discharge, and as a result, the discharge capacity of the pumping station ahead of flood waves was physically limited by water level and inflow. Therefore, improved operation of the pumping station was not the solution for preventing or reducing fluvial flooding. On the basis of the results from this research, structural measures were suggested for flood mitigation, like temporary storage, a bypass channel, and measures that slow down the run-off processes. Also, it was recommended to examine the value of a flood early warning system by using 13-15 hours of lead time within a broader scope of disaster risk reduction in Tub Ma.

The Oscillating Water Column (OWC) device is one of many available technologies for generating electricity from water waves. It is a caisson-like structure, which houses a water column in an enclosed chamber with an inlet at the bottom. Waves enter the device, exciting the water column. The trapped air above is forced out through a turbine, generating electricity. The technology is relatively young compared to other sources of renewable energy, such as solar or wind energy. This makes the produced energy expensive, expressed in the Levelised Cost of Energy (LCOE). Integrating devices into breakwaters already resulted in a large reduction of the costs, as they now can be shared. During the design of such devices, main attention is given to the geometrical design. This should be based on local wave conditions to ensure the largest energy production. Having the eigenperiod match with the incident wave period enhances the resonance effect in the chamber. This research explores the potential to reduce the costs of OWC devices integrated in breakwaters through means of structural optimisation. The structural are built quite robust, leaving the thought there is room for improvement. The device in Civitavecchia, Italy was used as a case study. A new method was adopted to investigate the potential. A numerical model was constructed in STAR-CCM+, including both a fluid domain and a solid domain. The two are one-way coupled, from water to structure. The exerted wave pressures are mapped on the structure directly, resulting in a stress state in the OWC structure. Due to the transient nature of the simulations, all results are available through time. It was found, the main walls shaping the OWC device could be built with 35% less material. Variants of the original geometry were tested as well to see how these influence the results. Furthermore, the structural behaviour was investigated. Clear trends were found in position and timing of governing load combinations in the different walls. One of the most important findings was that the governing load case always was in the direction of the transverse width. The walls were found to behave more like beams in this direction. The bending moments could therefore easily be calculated using standard beam equations and the net horizontal pressure on the wall as load. This leads to the beginning of a new simplified design method where the full numerical structural analysis is not required anymore.

Hydraulic structures can be prone to impulsive wave impact, which is a highly stochastic and uncertain process. This type of impact, defined by extreme pressure peaks and a very short duration, is not only caused by breaking waves, but also by non-breaking standing waves on structures with an overhang, such as culverts and steel gates. In this study, the capabilities of the Lagrangian numerical tool Smoothed Particle Hydrodynamics (SPH) are validated by means of experimental research conducted at the Hydraulics lab of the Delft university of Technology, in which two short overhang configurations were subjected to multiple non-breaking wave conditions. SPH distinguishes itself by discretizing the numerical domain in particles instead of a grid, unlike traditional computational fluid dynamics (CFD). In doing so, it excels in free surface modelling and complex wave-structure interaction. In this thesis, the theory of pressure-impulse is applied, which is defined as the integral of the pressure over the impact duration. This method is more stable than using pressure peaks and can be used to obtain the reaction forces on hydraulic structures. However, the theory is in development and subject of recent literature. This research includes the theoretical pressure-impulse model, which is based on the Laplace equation and solely includes the vertical impact by assuming a circular profile under the overhang with a constant impact velocity. Furthermore, a new conceptual model is introduced, which is based on integration by the particle velocities in both horizontal and vertical direction as described by Linear Wave Theory. The assumptions of the velocity fields of both models are assessed by the SPH method. As a result, modifications are proposed to the conceptual model and validated with the experiment. SPH shows good agreement with the experiment in terms of wave generation, pressure distribution and pressure-impulse profile. However, the lack of air in the numerical model result in overestimations of the pressure peaks. The more air is entrapped in the experimental wave impact, the higher the deviation. That said, the impact duration also becomes longer the more air is entrapped while the SPH model shows a somewhat constant and shorter duration. As a result, the pressure-impulse profiles shows corrective behavior over the vertical, mitigates the absence of air and thus greatly increases the accuracy and stability of the results. Finally, an analytical validation is performed in which the theoretical models for overhang configurations and design formulae for vertical walls are compared to SPH and the experiment.

Shanghai is one of the numerous megacities worldwide that experience severe flood events triggered by torrential rainfall. To deal with the undesirable consequences of these events and mitigate the flood hazard, the research of flood reduction measures is necessary. In this effort, the hydrodynamic modelling is a useful tool.
In this master thesis a 2D model was developed for the simulation of urban flood events in Jingan District in the downtown of Shanghai. The main objectives of the thesis were the production of the flood hazard maps for numerous rainfall events and the assessment of the proposed flood mitigation measures. Delft3D Flexible Mesh was used as a tool to produce the inundation maps. Also, several data were considered regarding the grid for the numerical calculation, the surface elevation, the local drainage system and the rainfall events. SOBEK was used for the set up and the preprocessing of the sewer system. For the simulations, data from three historical rainfall events were used: August 2005, August 1997 and September 2013 and five rainfall events with return periods of 1, 3, 5, 10 and 50 years were generated by using the Chicago hydrograph.
For the validation of the model, the rainfall event of August 2005 was used. Although the validation of the model was not proven due to the simplifications that were made in the input data and the lack of data, the model showed that some processes can be simulated, and inundation maps can be produced. By comparing the results that occur with and without the inclusion of the drainage system, it was concluded that the local drainage system should be included in the analysis for the assessment of the flood hazard in an urban area, since its presence plays an important role in the flood reduction. The results showed that the maximum inundation depth can decrease by around 45%. As flood reduction measures, the creation of water storage areas and the increase of the drainage capacity were considered. The water storage areas covered around 10%, or less, of the block areas with available space, leading to a water depth reduction that depends on the location in the map. For the increase of the drainage capacity, the value of 1 m3/s was assigned in a single and in multiple locations in a specific area of investigation. The results showed a percentage of water depth reduction around 15.9% and 45.5%, respectively. For the same location the percentage of water depth reduction due to the water storage areas was 22.5%.
Finally, uncertainties were introduced in the model due to the assumptions and the simplifications that were made in the input data. However, this model can work as a base for future researches to accomplish more realistic results, by improving the current model and adding more updated and precise data.

Extra-Tropical Cyclones (ETCs) are major storm system ruling and influencing the atmospheric structure at mid-latitudes. These events are usually characterized by strong winds and heavy precipitation and cause considerable storm surges with threatening wave systems for coastal regions. The possibility to simulate these storms or to increase the amount of significant data available is crucial to optimize risk assessment and risk management for construction projects and territorial plans which might get damaged by events of this kind. The project addresses the possibility to learn the distribution of cyclones atmospheric fields of pressure, wind and precipitation in the North Atlantic by training a Generative Adversarial Network (GAN). The ETCs tracks are extracted from the ERA5 reanalysis dataset in the domain with boundaries 0°-90°N, 70°W-20°E and period going from 1st January 1979 to 1st January 2020. A GAN tries to learn the distribution of a training set based on a game theoretic scenario where two network competes against each other, the generator and the discriminator. The former is trained to generate new examples which are plausible and similar to the real ones by having as input a vector of random Gaussian values. The random vectors domain is called latent space. The latter learns to distinguish whether an example is coming from the dataset distribution or not. The competition set by the game scenario makes the network improve until the counterfeits are indistinguishable form the original. The generative models trained on the ETCs dataset are validated to understand if they are able to generate new samples of fields presenting similar atmospheric characteristics to those of the original dataset. To train the GAN two different loss function are considered, the Wasserstein distance and the Cramèr distance. The Cramèr Gan (CGAN) shows better performance in representing the distribution of the atmospheric fields, generating images that on average look similar to the original ones. The Wasserstein GAN (WGAN) behaviour shows poor performance in representing the precipitation in general, but it is able to similarly reproduce the values distribution for what concerns pressure and wind. The images generated by the WGAN have many differences compared to the original ones and are very blurry with particular data structures that looks like artefacts built by the network. The atmospheric structure of the images generated by the CGAN is investigated by considering 4 cyclones as case study and comparing the frames of their tracks to those of synthetic tracks generated by linear interpolation. The linear interpolation is performed between the random vectors generating the most similar images to the initial and final snapshot of the original track. The interpolated images show interesting features in the similarity with the original track, which suggest that the network has learned a representation of the ETCs fields that is promising for further investigation.

Flexible river training structures such as tetrahedron frames (`porcupines') can be attractive for control of braided river channel networks in regions where permanent control structures (e.g. groynes or dams) are too expensive or potentially inefficient, such as systems with highly dynamic flow regimes and morphology. Porcupines provide hydraulic resistance, generating energy loss and reducing velocities that may further encourage sediment deposition. Porcupine systems have seen increasing implementation, especially for bed or bank protection, but were also recently implemented in a channel-control pilot project on the Ayeyarwady River in Myanmar. Many porcupine systems to date are designed by trial and error due to lack of quantitative design criteria. The 2019 pilot project used large-scale numerical modelling of resistance areas to evaluate the potential impacts of porcupine structures; however, how to best incorporate the impacts of porcupine structures on flow and sediment transport in numerical models has not been systematically evaluated. Improved models or better estimates of uncertainties could facilitate future porcupine system design. Therefore, this study examines methods for incorporating porcupine fields into numerical models and interpreting those models to make informed design choices. Data and observations from the 2019 pilot project implementation, a 2018 porcupine flume experiment, and extensive literature review have been used to examine the expected hydrodynamic and morphodynamic impacts of porcupine systems that need to be accurately captured in large-scale numerical models. A 2DV model using two resistance formulations representing porcupines was elaborated and tested against experimental porcupine performance. The Uittenbogaard (2003) rigid cylinder model was found to replicate well porcupine behavior for less dense systems under limited flow conditions; however further verification is needed for more dense systems, variable flow regimes, wider boundary conditions and mobile beds. The 2DV model was collapsed to 1DH, using two different resistance formulations, to examine the strengths and weaknesses of large-scale models in predicting porcupine performance. The Baptist (2005) resistance model was found to best represent porcupines; however, parameterization was not straightforward from porcupine geometry, indicating that further studies will be needed to confidently translate porcupines into `rigid cylinders' for input into the model. In addition, neither the 2DV nor 1DH models were able to capture the hydrodynamics at the leading and trailing edges of porcupine fields, which were found to be critical for design considerations from the pilot study data. Therefore, these important limitations need to be taken into account when interpreting model results and examining how to improve them in the future.

Estuaries are dynamic partially enclosed water bodies that are constantly or periodically influenced by the ocean and at least occasionally impacted by river discharge. This creates unique but fragile ecosystems that have to be managed with care in order to be recreationally, economically and ecologically valuable. One of the management issues is water quality, which is mainly influenced by hydrodynamic processes and other processes affecting the transport of dissolved or suspended materials. The understanding of the hydrodynamic processes of an estuary and its physical drivers is crucial for management. One of these fragile ecosystems that have been prone to many human interventions in the past, is the Leschenault Estuary. In this thesis, a 3D-numerical model is developed in D-Flow FM to unravel the governing hydrodynamics of the Leschenault Estuary. Besides, field measurements provided data used as input for the numerical model and more information about the dynamics of the Leschenault Estuary. Most importantly, a methodology is proposed to improve the efficiency in drawing relevant conclusions from observed and modelled data. Efficient and easy-to-apply classification methods are therefore considered that could be powerful management tools. To validate this methodology, a scenario has been considered where the Preston River is aligned to the Bunbury Port. The governing hydrodynamic processes in the Leschenault Estuary are internal circulation, stratification and turbulence, which are predominantly driven by the freshwater discharge and the tides. However, the dominant physical processes are highly dependent on the seasonal conditions and the specific locations. In general, three different seasonal conditions were distinguished: normal summer, normal winter and peak river discharge conditions. In summer, freshwater discharge is reduced, which increased the impact of tidal stirring and vertical mixing. In winter, the high river flow generated more stratified flows and under peak discharge conditions some areas of the estuary presented salt-wedge regimes. The Leschenault Estuary can be spatially subdivided in four distinct regions (southern, central, northern and riverine basin), based on the governing hydrodynamic processes. The southern basin is the most dynamic and can not be specified by a single regime due to the influence of the ocean, the Preston River, the Collie River and the northern regions. The central and northern basins were classified as partially-mixed throughout the year and showed weakly to strongly stratified water bodies, depending on the seasonal conditions. Furthermore, a classical estuarine circulation develops under normal winter conditions. In summer, the northern regions become hypersaline, generating an inverse circulation in the central and northern basins. The Collie River is characterized by a partially-mixed water body, with high stratification. Under peak discharge conditions, the salt wedge can be temporally forced out of the river and partially out of the estuary. Winds, waves and Coriolis have a significant influence on the hydrodynamics of the central and northern basin, due to their shallow and stagnant waters. The driving forces of the salt transport were obtained by the decomposition of the salt flux. The main physical processes affecting the salt transport between the estuary and the ocean along the transect of the 'Cut' over the whole year and in winter were freshwater discharge, topographic trapping and Stokes drift. Stokes drift was dominant in summer, followed by freshwater discharge and topographic trapping. This indicates that regardless the season the advective terms were dominant drivers of the salt transport. It also indicates the seasonality of the salt flux at the 'Cut'. The dominant salt flux components at the central basin was the Stokes drift, followed by freshwater discharge and topographic trapping. The Preston River alignment was compared with its current location to provide useful information for management and to evaluate the efficacy of the adopted classification methods. The scenario results were significantly different than the initial model results and the Preston River alignment had a substantial impact on the hydrodynamics of the Bunbury Port and the southern basin. Bunbury Port circulation became more stratified and its seasonal variation was increased, while the southern basin became less dynamic and partially-mixed, with low to high stratification. The Collie River and central and northern basins however remained almost unaffected. Further, the physical drivers of the salt transport did not vary much but the role of the Stokes drift became relatively more pronounced at the `Cut', due to the decreased influence of freshwater discharge and topographic trapping. At the central basin, the Preston River alignment had a negligible effect on the physical drivers of the salt transport. However, larger spatial and temporal variability of salinity and temperature distributions were observed. It is therefore recommended to conduct an additional ecological valuation of this intervention. Valuable insights were presented in this thesis that have been critically validated. The used methods have proven to be efficient and valuable tools for management.

In recent years, there has been an increasing need for solutions against the threats that SLR and climate change represent for coastal systems, especially in low lying areas like the SW Dutch coast. Simultaneously, awareness of the impacts that human activities have on natural environments has surged. Integrated and sustainable solutions are necessary. The Delta21 plan proposes an integrated plan for flood-protection, energy storage, and natural ecosystem restoration for The Netherlands, specifically for the Haringvliet estuary. This large-scale intervention will produce large disturbances in the system. Despite the resilient character of tidal basin systems \parencite{Wang2009}, the evaluation of the intervention's effects, in terms of the initial response and the variations in the mechanisms driving morphological changes is compulsory. Assessing the bed level changes is especially important, considering the valuable intertidal and subtidal ecosystems present in the Haringvliet mouth. It is essential to guarantee their preservation despite the large disturbances. The evaluation of the Haringvliet mouth response is performed employing a long-term morphodynamic 2DH computational simulation to forecast the response of the system. The models implemented for the simulation are the Delft3D-FLOW and Delft3D-WAVE. To reduce computational times, the modeling approach applies input reduction and acceleration techniques. The hydraulic forcing is schematized (morphological tide, waves, and discharges) while maintaining the seasonal character of the input. A variable morphological acceleration factor (MF), depending on the wave condition is applied. The variable MF allows us to employ low values during strong-episodic conditions and larger during more regular conditions. The approach is especially beneficial in the presence of oscillating forcings that are also enhanced by the MF. Finally, the impact of the D21 plan on the Haringvliet mouth is assessed by studying the alterations to the hydrodynamic regime, the related net sediment transport pathways, and the observed bed level changes. The analysis of the resulting morphological evolution focuses on the variations of the intertidal and subtidal zones and the mechanisms behind them. We showed that the tidal regime shifts again towards a long-basin regime once the D21 is implemented. The intervention also causes the emergence of new subtidal and intertidal areas, mostly within the Tidal Lake, resulting from the redistribution of the material from the main channel. The original intertidal and subtidal areas, formed by the Hinderplaat and Slikken van Voorne, are almost undisturbed. This behavior guarantees the preservation of the intertidal and subtidal ecosystems. The results also show that despite the initial dredging activity necessary for the D21 plan implementation, which decreases the extent of shallow areas, the emergence of intertidal and subtidal features balances the negative effect as long as no further dredging is performed. The TL shows a net sediment export of material at both control cross-sections (the Haringvliet Dam and the new TL inlet). The findings of the morphological development agree with the known morphology and processes of mixed-energy tidal inlet systems, despite the large-scale intervention. This study provided the opportunity to evaluate the emergence of typical tidal features in a highly disturbed system, where the tidal inlet is shifted further offshore. To achieve an accurate forecast, anthropogenic forcing signals also had to be considered and extrapolated to long-term morphodynamic simulations. The forecasting of large-scale anthropogenic interventions in sensible systems, such as tidal estuaries, is performed successfully by adapting an existing model. The development of this type of study will be increasingly relevant in the future for the evaluation of measures against SLR during conceptual project stages.

Wave overtopping is an important aspect for the design of rubble mound breakwaters. Acceptable overtopping limits are usually based on the mean overtopping discharge or the maximum overtopping volume (EurOtop, 2018). For temporary situations, such as breakwaters under construction, overtopping limits are not clearly defined. This complicates the choice for a working crest level, where the safety of staff and equipment on the partially constructed breakwater is the main concern.
In this study, OpenFOAM in combination with the waves2foam toolbox (Jacobsen et al., 2012) is used to model wave overtopping over a breakwater under construction to identify the flow characteristics (i.e. the flow depth and velocity) caused by overtopping waves. The aim of the study is to provide a method to define a safe working level for rubble mound breakwaters under construction, based on the tolerable flow depth and velocity. The safety of the staff and equipment is expressed in limit functions. The method of Sandoval (2016) is applied to define the limit function for workers safety on the breakwater under construction (based on combinations of flow depth and velocity). The safety of the hydraulic excavators is mainly based on the overtopping flow depth, e.g. the flow depth should not exceed the cabin-level.
2D physical model test results on a partially constructed breakwater are processed in order to obtain benchmark data to validate the numerical model. Once the numerical model was validated, a numerical test program was set up that varied the wave height, wave steepness and the relative freeboard, in order to investigate the dependence of the overtopped flow depth and velocity on the varied conditions. A JONSWAP spectrum was used to simulate 8 different wave fields of 250 waves. On the crest of the numerical breakwater, the flow depth and velocity are measured by means of equally spaced gauges and probes in order to describe the flow behavior.
Overtopping events due to the irregular wave train are statistically analysed in terms of flow depth and velocity. The empirical 2% exceedance values are considered as design parameters as these can be compared to existing literature on impermeable and smooth dikes, to highlight the effects of the partially finished and permeable structure.
Empirical relations between the investigated parameters and the overtopping flow depth and velocity on the core crest edge of the breakwater are presented. It is found that the maximum flow depth reduces exponentially over the crest. The reduction appears less strong when saturation of the crest material starts to play a role. The flow velocity increases over the crest for the cases with larger initial flow depths and/or flow detachment from the breakwater crest edge, in contrast to the formulas for dike cross sections. This latter phenomenon was explained by the acceleration caused by the redirection of the flow on the breakwater crest. From the numerical results, two distinct areas of the unconstructed breakwater crest can be identified: a “transition zone” where the flow is detached from the breakwater crest and a “flow zone” with attached flow where the limit functions are valid.
Based on the numerical findings, empirical guidance is given for the land-based workability of breakwaters under construction. It is shown how the newly developed empirical formulas can be applied in practice to determine a safe working level. It is noted that the maximum overtopping flow depth and flow velocity do not necessarily occur at the same moment. A method is introduced that shows how the safety can be assessed using a combinational number for the flow depth and flow velocity (based on numerical modelling) and the stability limit for people. It is shown how a safe working level can be found based on exceedance values of safety limits.

Concerns regarding plastic pollution arise as large quantities of plastic enter the ocean and affect wildlife and possibly human health. Rivers form a dominant pathway for macro plastic particles into the ocean. Plastic pollution in rivers can result in blockage, where it affects water quality and ecology. This increases risks of flooding and provides health concerns. To prevent further dispersion of plastics into the ocean and reduce health and flood risks, reduction of plastic concentrations in rivers is desired. Buoyant macro plastics are a dominant polluter and understanding their cross-sectional distribution in rivers could promote areas for efficient extraction or monitoring. As field measurements are expensive and time consuming it is desired to accelerate research with modelling techniques. To improve modelling of buoyant macro plastics and translate cross-sectional distributions it is necessary to model relative motion of buoyant macro plastic particles to the flow and one another. Previous research has shown that the understanding of macro plastic forces and transport is limited. The range of macro plastic properties complicates modelling approaches to macro plastic transport. To simplify, a classification of macro plastics is introduced according to cornerstones in behaviour. A basic force model is constructed (derived from riverine wood transport studies) to provide an understanding of the effect of size, density, wind and discharge on transport of different macro plastic types. This new understanding is applied to a variety of flow features induced by a range of river elements (river bend, channel widening, river confluence and groyne) to develop a hypotheses on the cross-sectional distribution of the different macro plastic objects in these environments. These hypotheses are tested in a numerical transport model. A theoretical background study on particle response times and Stokes numbers is performed to identify a relation between the particle response time and relative particle dispersion (compared to the flow). This provides a possibility for translating the transport of the different macro plastic objects into the numerical model. To define numerical modelling input of different buoyant macro plastic objects, particle response times are calculated and translated to horizontal particle dispersion coefficients in D-WAQ PART (software package of choice). Each of the previously formulated river elements is modelled as well in the numerical model. The modelling results show that river bends cause an increase in particle concentrations along the river bend. Channel widening increases the concentration along the boundaries. A river confluence increases the concentration along the tributary side of the channel. A groyne increases the macro plastic concentration along the groyne-free side of the channel. These results show that particle response times and their relation to Stokes number and particle dispersion propose an interesting tool for the modelling of relative particle motion towards the flow and the different macro plastic objects. This study identifies the significance, diversity and complexity of macro plastic particles in riverine environments. It highlights the need for classification and simplification to facilitate modelling and still be able to obtain relevant results. Describing macro plastics according to cornerstones in behaviour provides an efficient tool to simplify modelling. Coupling of different macro plastic classes to a range of riverine environments and understanding the differences in their cross-sectional distribution provides practical use.