Grass-covered dikes are a widely used measure for flood prevention. Overtopping waves can cause erosion that poses significant risks for flood safety, especially with rising sea levels and increasing storm intensities. During storm events, which involves frequent wave overtopping, erosion occurs progressively as multiple individual waves flow over the dike. This causes the dike profile to change over time as erosion holes begin to appear, which can influence the erosion potential of subsequent overtopping waves.

Current methods to estimate forces and erosion from overtopping waves rely on numerical models or analytical and empirical formulas. Numerical models are generally applicable and provide detailed spatial and temporal results, but are computationally too expensive to simulate an entire storm whilst updating the bed profile. Formulas offer quick results but lack general applicability and fail to account for the effects of erosion holes. No current method efficiently combines computational speed, detailed results and broad applicability, making it challenging to study the effects of gradually increasing erosion on overtopping waves.

This research focuses on the development of a new method that provides rapid and reliable wave overtopping simulations that can integrate erosion. This is done through surrogate modelling, which aims to create a computationally cheap model that emulates a detailed model. The foundation of this surrogate model is a Transformer-based deep learning architecture, which has proven superior in handling spatiotemporal processes.

The surrogate model is created by adapting the Vision Transformer model into a new model that can perform next-frame prediction, which involves the prediction of a next frame based on a sequence of input frames. The developed model can take an input sequence, recognize spatial and temporal patterns, and project them into a predicted future timestep. The model is trained on a dataset of overtopping wave simulations produced by the CFD software OpenFOAM. A masked loss function is applied to enhance the training process by forcing the model to focus on improving the relevant errors.

Using the trained model to generate wave overtopping simulations showed that it can accurately replicate the original CFD simulations. The surrogate model is validated against the original CFD simulations by comparing maximum values and time series for flow velocity and water depth at four different locations along the dike. The results generally show good agreement in capturing the maximum values, as well as the time of arrival and the overtopping duration. Simulating wave overtopping over an eroded dike profile showed promising results, though performance could improve with a larger and more diverse training dataset.

This research demonstrates that a Transformer-based surrogate model can effectively emulate wave overtopping simulations produced by CFD software. The surrogate model's speed and simplicity enables the simulation of a storm and updating of the bed profile. The model has not reached its full potential due to limitations in the training dataset and nuances in the simulation technique that require further refinement. However, it serves as a proof of concept that this surrogate model can provide a new tool in wave overtopping modelling, creating possibilities for new studies such as the influence of erosion on overtopping waves during storm conditions, or for probabilistic calculations that require a large number of simulations.

This report details the development of a Flood Early Warning System (FEWS) for the Tana Basin in Kenya, executed by a multidisciplinary team from the Delft University of Technology. Recognizing the Tana Basin’s vulnerability to flood risks, exacerbated by climatic variability, limited funds, and limited available data, the project proposes a model that combines computationally efficient hydrological and hydrodynamic modelling with robust stakeholder collaboration. The study area comprises the entire Tana Basin, with a specific focus on the flood-prone area near Garissa used for validation. The FEWS developed incorporates local and scientifi-cally derived knowledge to forecast floods, aiming to aid the transition from a technologically intermediate to a technologically advanced FEWS. Through an iterative process of model selection, validation, and stakeholder feedback, the system attempts to integrate the GR4J hydrological model in SuperflexPy and combines this with the Super Fast INundation of CoastS (SFINCS) model. Data sources include global remote sensing datasets like FABDEM & CHIRPS. Furthermore, it uses the water level gauge data provided by the Water Resource Authority of Kenya, as well as TAHMO weather station data.

The report concludes by reflecting on the modelling techniques for both the hydrological and hydrodynamic models and provides recommendations for the further development of a FEWS in the Tana Basin in Kenya. The implementation of the hydrological model was not able to propagate external flows through the network, making it poorly suited for use in the Tana Basin. The hydrodynamic model works decently well in flood conditions but overpredicts flooding during regular flow conditions. Recommendations on stakeholder engagements and data-sharing practices to foster a resilient flood management system in the Tana Basin include more comprehensive Memoranda of Understanding (MoU) and stricter adherence to the Disaster Risk Management Framework of the United Nations.

The Netherlands is a country that is being threatened by water, both from the rivers and from the sea. The Dutch have built dikes to keep their lands from inundation. To ensure the strength and stability of these dikes, they are being assessed on the basis of several failure mechanisms. One of these failure mechanisms is Backward Erosion Piping, or piping for short. In piping, the current underneath a dike is strong enough to take soil particles with it. Tests on piping in tidal subsoil were conducted in the summer of 2021, where a pipe was found to have grown at greater depth than expected The occurrence of this deeper piping has rarely been seen before, let alone described. This lack of knowledge poses a potential safety risk, as it may underestimate the vulnerability of certain subsoil configurations. Therefore, the objective of this thesis is to develop a comprehensive understanding of deeper piping and identify the key parameters influencing its formation. To achieve this objective, a definition of deeper piping and its differentiation from conventional piping is established. Sub-mechanisms governing deeper piping are examined by analysing the forces responsible for grain movement and the forces that maintain grain stability. A Finite Element Model of the subsoil is constructed to quantify the driving forces within the subsoil, which, when combined with resisting forces, enables the determination of whether deeper piping can occur in a given subsoil configuration. To investigate the factors contributing to deeper piping, a series of simulations are conducted using this Finite Element Model. By varying the parameter values while keeping other factors constant, the influence of each parameter on the occurrence of deeper piping was examined. The analysis revealed that several key parameters significantly affect deeper piping formation, including cohesion force (P), cohesion anisotropy (?P ), permeability and thickness of the top layer (X0 and 70, respectively), permeability of underlying layer (X1), permeability anisotropy (?X ) and representative grain diameter Q_R]. Also, it was found that the entrance configuration plays a large role in deeper pipe formation. These findings provide valuable insights into the mechanisms underlying deeper piping and enhance our ability to identify subsoil configurations that are prone to this phenomenon. These findings enhance the identification of subsoil configurations prone to deeper piping, thereby improving risk assessment and mitigation strategies associated with this failure mechanism.

This study focuses on optimizing the use of high-performance computing on public cloud infrastructure, along with information theory, for assessing water systems. These assessments are computationally intensive and can benefit from parallel computing and the evaluation of the collected data with information theory. A case study of a water system analysis for the Vlietpolder was conducted to test various cloud configuration settings using an embarrassingly parallel batch computation. The hydrodynamic simulations involved D-HYDRO 1D2D models with different precipitation events and model resolutions. The modelling results were quantified using normalized Shannon’s Entropy to facilitate the comparison of system configurations, evaluating the batch computation process to determine whether enough simulations have been performed and comparing individual simulations.

The study showed that public cloud infrastructure provides comparable computational performance to local computers and servers, and offers opportunities for vertical and horizontal scaling for parallelization. The study also provides insight into the impact of allocated resources, node size, and node type on cloud infrastructure performance. Furthermore, the quantified information derived from the simulations can be utilized to evaluate the batch
computation output and support cost-benefit analyses for selecting configuration settings and model decisions given modeling scenarios.

The study concludes that combining cloud infrastructure and information theory can enhance hydrodynamic modelling for batch computations in water system analysis. The findings provide insights into the potential benefits of utilizing public cloud infrastructure for large-scale computations of hydrodynamic simulations.

The applicability of the Sellmeijer design rule to Limburg dike sections near Well, Hout-Blerick, Buggenum, and Thorn was investigated in this thesis. This was accomplished by building a finite element numerical model, in COMSOL Multiphysics, to assess the piping in the hydrogeological systems of the research locations. The proposed model is composed by combining several groundwater and piping model principles, schematizing dike cross-sections, and calibrating the model. The schematization choices that define the model geometry of the average pipe cross-section are critical. It was decided to apply a fracture flow pipe cross-section and the piping assessment was performed iteratively for multiple pipe height values, because the true pipe height is unknown. The pipe height is expressed as a function of the number of grains and the representative grain size.

According to a deterministic assessment, which was performed analytically using multiple versions of the Sellmeijer design rule and numerically using the proposed FEM model, piping does not occur at the research locations Well, Hout-Blerick and Thorn. Piping is theoretically possible at Buggenum, but only with a very small pipe height, making it appear very unlikely. The assessment also revealed that the critical head determined by using revised Sellmeijer design rule (2021) is conservative for the examined dike sections in Limburg. Furthermore, it is discovered that the original Sellmeijer design rule with the new geometry factor (1988/2011) produces critical heads that are similar to the results of the revised Sellmeijer design rule (2021) multiplied by 1.8, which is the current design rule used to assess piping in Limburg.

The Sellmeijer design rule (1988/2011) was stochastically evaluated to determine how the design rule changed when calibrated to Limburg parameter values. 1000 unique randomly generated dike cross-sections were evaluated on piping using the proposed FEM numerical models. The unique dike cross-sections are made up of a random combination of model parameters that were sampled from a uniform distribution using Latin Hypercube Sampling (Olsson et al., 2003). The randomly generated dike cross-sections that demonstrate piping for a realistic pipe height, were used to re-calibrate the Sellmeijer scale factor. Using a linear regression, it was demonstrated that the scale factor corresponding to the broadened application range (including Limburg parameter values), of the Sellmeijer design rule, is 1.56 times greater than the original scale factor. This means that the original Sellmeijer design rule with the new geometry factor (1988/2011) and the additional factor of 1.56, can be applied to (Limburg) dike sections, for which the model parameters fall within the newly set parameter application ranges (e.g. 100 µm ≤ d70 ≤ 900 µm).

Indonesia faces the challenge of attaining renewable energy goals and reducing carbon emissions by 29% by 2030. Despite a renewable energy goal of 23% of the national energy mix by 2025, only 14% of electricity production is projected to be generated from renewable sources by 2023. Accelerated deployment of renewable energy solutions is required to achieve these goals.

Indonesia has abundant renewable energy sources, such as hydropower, solar, and wind. Yet, a challenge arises from the difference between the electricity demand and supply patterns of these sources, which do not match throughout the day. Fluctuating energy supply patterns and variable energy demand necessitate using efficient Energy Storage Systems (ESS) to bridge the gap.

With its extensive coastline, Indonesia can potentially explore single reservoir Seawater Pumped Hydro Storage (SPHS), a variant of Pumped Hydro Energy Storage (PHES), as an alternative to solve these challenges. Similar to an enormous rechargeable battery, the reservoir in an SPHS system functions as an energy storage system. The system works by pumping up the seawater to the reservoir to store surplus energy during periods of ample supply and discharging it to generate electricity through a hydropower plant during periods of high electricity demand. This research aims to identify the ideal locations for SPHS systems in coastal areas of Indonesia. A Python GIS algorithm was developed to automate the selection process. The identified SPHS sites are then evaluated economically to ascertain their viability. The study concludes by comparing the carbon reduction potential of these systems to Indonesia's carbon emission reduction goals.

The research reveals 609 potential SPHS sites across Indonesia, with a total peak power that can be regenerated of technically potential sites of 29 Gigawatt-peak (GWp). Among these, 297 locations are deemed economically feasible, contributing a potential peak power that can be regenerated of 15 GWp. The peak electricity demand in Indonesia is approximately 44 GW, typically occurring at 8 p.m. The technical potential of SPHS promises an 11% reduction in carbon emissions from the energy sector, while the economically feasible sites could achieve a 6% reduction in carbon emissions projected in 2030.

Climatic conditions in uence peak discharges in rivers and change sea levels; therefore, attention to the safety of dikes is of ever growing importance. Macro instability is one of the dike failure mechanisms that can inundate the hinterland. Soil heterogeneity plays an important role in assessing dike safety, especially for slope stability, because it is a major source of uncertainty. To assess a dike network for safety, numerical simulations for a full probabilistic analysis can be computationally expensive. Therefore, this study investigates how to build a state-of-the-art data-driven framework from a numerical model to predict the safety margins from the macro stability of dikes. Inputs and outputs of tens of thousands D-Stability simulations were used to create a training dataset. The most relevant features were selected based on global sensitivity analysis and the representation of soil heterogeneity in the framework. The maximisation of Shannon's information entropy and the generation of the training dataset was achieved by employing a smart sampling strategy for the input parameters. The sampling strategy consists of a Latin hypercube optimised uncorrelated uniform distributed dataset combined with a correlated dataset for optimal training eciency. The uncertainty due to soil heterogeneity is represented by a Gaussian random eld with a trend. This trend is commonly determined from a geotechnical cone penetration test. With a CPT, it also is possible to nd the vertical scale of uctuation, which is parametrised by the correlation length of the uctuations in soil strength. The second-order Markov correlation function is used to represent the correlation of the random elds. The Gaussian random eld is later mapped onto 16 stacked horizontal layers to model the heterogeneous soil properties. The surrogate model consists of an ensemble of thirteen machine learning models. The most important model is a multi-layer perceptron feed forward articial neural network. The other models are histogram based gradient boosting regression trees. Random search and Bayesian optimisation are used as hyperparametrisation techniques to optimise the prediction capability is of the individual ML algorithms. Weights for each model are determined based on optimisation for error reduction for maximum performance. The surrogate predicts the factor of safety (FOS) as well as the coordinates of the slop failure circles and line of depth from the Uplift-Van method. The surrogate model ensemble that predicted FOS is quite accurate with respect to the numerical FOS of D-Stability, and yet the prediction of the failure plane is still slightly worse. A case study was used to demonstrate the performance of the framework. Despite the uncertainty of the subsoil, due to the soil heterogeneity, the surrogate was able to accurately predict the failure probability. However, the prediction of the far end circle coordinates showed lower performance due to propagating errors. Concluding, application of the framework is possible for dike reinforcement optimisation, risk-based dike safety assessment, length effect, and effcient Monte Carlo simulations.

Second-order Gaussian kernels have been utilized to develop three algorithms that could automatically extract ridge lines for hydrodynamic modelling. Isotropic second-order Gaussian kernels produce inaccurate lines at crossings and junctions. To avoid the malfunctioning of Second-order Gaussian kernels, one default and two alternative algorithms were developed. The first, default algorithm is based on isotropic kernels and non-maximum suppression. For the first alternative algorithm, isotropic and anisotropic kernels have been applied for the filter process. The third algorithm uses skeletonization instead of non-maximum suppression. A verification was applied to analyzed the performance of the algorithms. The Matthews correlation coefficient (MCC) of the default algorithm and the alternative algorithm that included anisotropic kernels was found to be 0.17. For the algorithm based on skeletonization a value of 0.08 was obtained. Hence it has been concluded that the algorithms that utilized non maximum suppression instead could more accurately detect ridge lines than the model based on skeletonization. However, the latter generated lines that contained less discontinuities. Furthermore this algorithm turned out to be computationally less demanding in comparison to the other two algorithms.

The production and consumption of electricity need to be balanced at all times. Due to the ever-growing shift towards renewable energy generation, this poses an increasingly difficult challenge. Currently, supply is regulated to maintain balance. However, there is potential to improve reliability and save costs by shifting the balancing to the demand side, known as demand response. The flexibility of water systems can play a role in this, thereby benefiting from cheaper price fluctuations and reducing operating costs.

This research investigates the IJmuiden pumping station, which drains water from the Noordzeekanaal-Amsterdam-Rijnkanaal system into The North Sea. The primary focus of the control of this system is ensuring safe water levels as it runs through areas of high economic value. The flexibility of the range of safe water levels allows costs to be minimized by selecting favourable moments to consume electricity. This simultaneously contributes to the stability of the electrical grid. This research explores the potential for a Reinforcement Learning controller for such an optimization problem, as there are some drawbacks to the Model Predictive Control methods that are currently widely used. The research objective is formulated as follows:

To optimize the control of the IJmuiden pumping station using Reinforcement Learning while complying with local water level restrictions and compare it to the state-of-the-art Model Predictive Control methods in terms of constraint violation, energy costs, and computational speed.

The Reinforcement Learning controller will use a deep Q-learning algorithm that chooses the most cost efficient control in IJmuiden while respecting the water level restrictions. To do so, the model makes decisions based on electricity prices and details about the state of the water system for the current time step as well as a forecast of 48 hours ahead. This data is provided as an input to the model.

The inputs of the model consist of historical data, meaning that the associated uncertainties are not included. The water system that the model can interact with is represented by a linear reservoir model. Therefore, the water system is influenced dynamically by the actions taken by the model. The possible actions are determined by the state of the water system.

The trained model was tested on 2 years of unseen data (data that was not used during training). Using the same test data, control plans were generated using Model Predictive Control. The Reinforcement Learning model was very successful in ensuring safe water levels. However, this did result in approximately 50\% higher energy costs. The use of the gate was close to optimal but the pumping was not clearly correlated with favourable prices and power consumption. The trained model was robust, with consistently accurate results with regards to respecting the water level constraints.

The most significant difference with the Model Predictive Control was the computation time. The Reinforcement Learning model was able to create a control plan approximately 300 times faster. This opens doors for further development of the model and increased complexity. A more accurate model of the water system can be used to take into account temporal and spatial effects and individually representing the six pumps in IJmuiden.

There are still many steps before such a model can be used for operational control, but the method has potential for such an application. Many aspects of the model can be improved as well as making adjustments to increase the usability for control operators.

One-third of the Dutch dike system consists of peat dikes. Drought causes these dikes to crack and fail more easily. Visual dike inspections are therefore inefficient, especially considering the increasing climate changes of the future. Research has shown that a fiber optic sensor cable (FOS), used in distributed temperature sensing (DTS), can help measure soil thermal responses, but the question remains whether it is also suitable as a replacement for visual inspections of dikes. This study helps to answer this question by coupling a finite element method (FEM) model with measured DTS information collected at Flood Proof Holland (Delft). In addition, (thermal) images are used to calibrate the FEM model. The measurement period was 20 summer days. The measured data consisted of webcam images, thermal images, and temperature time series of a FOS cable. The meteorological data was obtained via a weather station located in Rotterdam. The FEM model, calibrated and validated with the measured data, helped to find the thermal response of the system in situations of which no data was available, for example having various crack dimensions, cable positions or climatic conditions. The more distinguished the material properties of air and soil are (days with high water content and/or high radiation), the better the crack detection via DTS. After correction for the overestimation influence of radiation on the black FOS cable on sunny afternoons, the thermal response of the crack is corresponding to the air temperature more compared to the thermal response of grass. Crack detection via DTS turned out to be possible by x,T- and t,T-plots (diurnal temperature variation) and a regression plot with the daily peak-to-peak amplitude of the air temperature on the one hand and the daily peak-to-peak amplitude of the cable on the other. The advantage of the regression plot is that only one DTS thermal time series is needed to determine if the segment is most likely cracked or not. Furthermore, the peak-to-peak axes allow for diurnal climatic condition indications, and with it the heat storage and release of the system: rainy, cloudy days are in lower axis regions whereas sunny, clear days are in higher axis regions. However, due to a small measurement period, this method is only fully proven for sunny afternoons. Future studies have to map the thermal processes for other situations too.

Flood simulations can give insight into the consequences of flood scenario's and can help to create hazard- and risk maps to support decision-making in flood risk management and in crisis management. 2D hydrodynamic simulations give accurate descriptions of the propagation of a flood and rely on advanced numerical methods to solve a set of physics-based mathematical equations. A drawback of these models is that they can be computationally expensive with run times in the order of hours or days depending on the time and spatial resolutions. In this study we explore the use of deep learning techniques in a surrogate model for 2D flood simulation. We propose and test a deep learning-based surrogate modelling framework that can be used to train a deep learning-based surrogate model. Once trained, the surrogate model can be used as a substitute for the hydrodynamic model with the advantage of being much more efficient in terms of run time and can be of great value in for example crisis situations. For training, a data set of expensive 2D hydrodynamic simulations was created using the SOBEK software program. Such simulations require a lot of input data, such as input parameter maps specifying the terrain over the computational grid and boundary conditions. To make training data-efficient, a sampling strategy was used for the input of the flood simulations. Three deep learning architectures were trained and tested. The first two architectures are feed-forward networks and the third architecture is of recurrent network type. These networks contain convolutional neural network (CNN) architectures with an encoder-decoder structure to make patch-level predictions of the flood characteristics in time. These patches contain a small section of the flood prone area and an encoder network is used to extracts coarse feature maps from this data that is then refined by a decoder network to create a prediction of the flood propagation. Using patches has the advantage of making a surrogate model able to create flood simulations over prone areas without restrictions on size or shape by tiling the output patches with flow predictions. Also it allows the surrogate to focus only on regions where the flood has reached and not on the regions where no water has arrived. It was found that with the recurrent architecture, the surrogate model was most capable of emulating the ground truth flood simulations in the test simulation. This trained network architecture was used in a case study where the surrogate was applied to create flood simulations in a small dike ring in the Netherlands. This shows that the surrogate modelling framework can be used to train a deep learning-based surrogate and, once trained, can be used to create flood simulations similar to hydrodynamic simulations. However, two main challenges were identified in using such data-driven deep learning-based surrogates. Firstly, keeping the predictions of the flood characteristics accurate enough to avoid large error propagation. Secondly, accurately generating large amounts of data from relatively little information present in the boundary condition and terrain.

Large parts of the Netherlands are vulnerable to flooding. Due to the great impact that a flood would have for the Dutch society, The flood safety standards are very strict, which means that the required failure probability of flood defences is very low. To keep the flood risk within acceptable limits of the hinterland, flood embankments are assessed on whether the probability of failure matches the stringent norms using the methods and rules which are stated in the legal assessment instrument (WBI). Within the WBI the probability of failure of the flood embankment equals the sum of the probability of occurrence of individual failure mechanism. This thesis specifically focuses on the failure mechanism: sliding of the landside slope cover due to wave overtopping of clay flood embankments. Due to wave overtopping the relative permeable top layer of the clay embankment becomes saturated. This leads to a shallow subsurface flow parallel to the landside slope cover which is unfavourable for the geotechnical stability of the landside slope cover. The stability assessment of this mechanism is currently assessed with the Edelman & Joustra formula. This formula method is not based on a probabilistic or semi-probabilistic analysis and therefore does not give an failure probability. Instead it uses safety factors. Levee managers therefore believe that the outcome is overly conservative (Waterschappen, et al., 2021). This thesis addresses the question: How can the stability assessment of the Edelman & Joustra formula can be optimized for clay flood embankments in the Netherlands within the legal assessment instrument? To answer this question, new semi-probabilistic safety factors were derived based on a full probabilistic assessment of the Edelman and Joustra formula. The outcomes show that the current approach is not always as conservative as is often assumed. Application of the newly derived safety factor allows levee managers to be more flexible in how to account for the contribution of this failure mechanism to the total levee failure probability. The newly developed method therefore prevents unnecessary rejection and over-dimensioning of flood embankments, so the available resources for flood embankment improvements can be used better.

Coastal areas around the world have always been densely populated areas. However, sea-level-rise and an increase in single extreme events due to climate change, threaten the coastal areas and their inhabitants. Governmental organizations, coastal managers and various private parties thrive for better insights into long-term shoreline behavior for sustainable decision-making. Currently, these insights are gathered by process-driven models that are often time expensive, require local input and calibration, and are often limited by equations. However, satellite imagery proved to be a promising data source to derive historic shoreline behavior on a global scale. Machine Learning (ML) algorithms are suggested to create an extra in-depth understanding of these satellite images. The increase in the availability of both satellite data and ML algorithms opens possibilities for shoreline forecasting on a global scale. This research aims to improve forecasting of shorelines by creating a global model in which cross-time series information can be used. To encourage a meaningful exchange of time series information, a clustering forecasting approach is proposed.This research builds upon the yearly SDS data behind the Shoreline Monitor of Deltares. The SDS dataset captures shoreline evolution and behavior on a global scale by quantifying it in time series. The SDS dataset consists of transects every 500 m along the global coast and contains 33 years (1984-2016) of shoreline evolution. First, in order to cluster transects, the features for clustering were defined. Here, clustering features were divided into time series features and hydraulic and geomorphic features. Whether transect clustering could improve shoreline forecasting, based on these time series and hydraulic and geomorphic features, was explored. For this thesis, it was decided to focus on time series features for further clustering usage. Eventually, a shape-based transect representation was chosen as feature for input of the clustering algorithm. Around 350,000 globally distributed sandy transects were assigned to nine different main clusters, with a semi-unsupervised clustering approach. These nine clusters captured global sandy shoreline behavior ranging from extreme erosion to extreme accretion and proved to be a practical tool to quantify the (distribution of) shoreline behavior on different spatial scales. Subsequently, a second supervised step was developed to create sub-clusters to gain more insights into the trends of the nine clusters. With this supervised sub-clustering step, three sub-clusters were generated for each of the five largest clusters, allowing for the distinction between accelerating and decelerating behavior in the last decade of time series per cluster. Global, country-level and local case studies showed the potential of these sub-cluster refinements. Hereafter, the nine main clusters were separately used as input for four different forecasting algorithms. This resulted in predictions per cluster with a more reliable and more accurate seven-year forecast for >95 % of the transects, sometimes improving accuracy up to factor 15, compared to recently published work. Furthermore, a multi-method approach was created to determine the most reliable forecast, on a local scale by combining overall accuracy and information gained during clustering. Time series clustering enhances the forecasting of transects based on overall accuracy and local reliability. The multi-method approach for selecting the appropriate prediction on a local level, could be a practical tool for governments to apply. Besides, a vulnerability assessment explores the possible application of the predictions and underlines the need for global databases regarding the coastal zone. However, further research should consider incorporating hydraulic and geomorphic features to strengthen the clustering of transects. Hence, shoreline forecasting based on historical data should be considered as a valuable tool for sustainable decision-making regarding coastal zones.

During intense periods of drought, the development of cracks is observed in peat and clay dikes. Asset managers of the dikes increase the inspection frequency in times of drought to be able to monitor these cracks. Significant development of the cracks contributes to the development of different failure mechanisms. In this study, the occurrence of the cracks is predicted at a large spatial scale. An inspection database in which the observations from the last three years are stored is used as the basis. The database contains hundreds of observed cracks including the location and time in which they were observed. The database was extended with attributes such as the precipitation deficit, the peat width at the surface, the orientation of the dike body, the subsidence of the dike body and the soil stiffness. Decision tree algorithms were then used to classify which circumstances will lead to cracks and which circumstances will not. From the resulting decision trees it was deduced that high precipitation deficits, low soil stiffness and the peat width can be used as the main predictors for the occurrence of cracks. Both subsidence of the foundation and the dike body being orientated to the sunny side are also contributors, although less prominent. Time-independent cracking criteria were then used to classify which regions are prone to cracking. Dikes which are rich in peat with a low stiffness were thus highlighted. The Mathews correlation coefficient was used as performance criteria resulting in a 0.3 value for the obtained tree. Application of a random forest increased the coefficient to 0.8. An important conclusion is that proper monitoring of the peat width, soil stiffness and precipitation deficit may result in better asset management. On behalf of more classification validity, it is advised to register negatives during inspections as well. Besides, making a clear distinction between drought-induced cracks and cracks due to macro-instability will reduce the likeliness of the model predicting cracks not related to drought.

Punchiná reservoir is part of the San Carlos Hydroelectric Power Plant, situated in the Guatapé watershed. The Guatapé river is an affluent of the Samaná Norte river, which in turn is an affluent of the Magdalena river. San Carlos Hydroelectric Project uses the waters from the rivers San Carlos and Guatapé and discharges the turbined flow directly into the Samaná Norte river by a tunnel. Currently, there is flow downstream of the Punchiná dam only on the days where the spillway operates, significantly impacting the riverine ecosystem. Additionally, claims have been made about how hydropeaking causes floods in villages downstream, particularly in La Pesca village. This town is located in the confluence of Samaná Norte and Magdalena river, on the left bank of Samaná Norte river mouth. The present report deals with the multi-objective optimization of the Punchiná reservoir of San Carlos Hydroelectric Project in Colombia by considering the objectives of maximizing hydropower revenues, maximizing the ecological discharge at Guatapé river, downstream of the dam, and reducing the flood risk at La Pesca village. Four numerical methods were coded in Python to solve the reservoir routing. To solve the multi-objective optimization, the non-dominated sorting genetic algorithm II (NSGA-II) using the Pymoo framework in Python was set up, along with the use of an Explicit Euler numerical method for modelling the river routing. The simulation was performed for 3 periods (high, average and low flow conditions) within the years 2010-2017. After multiple optimization scenarios, it can be concluded that the hydropower and environmental flows are competing objectives, i.e., allocating water for environmental flow purposes from the Punchiná reservoir will always result in a reduction of the hydropower revenues. Hence, it is recommended that an incentive system is developed so that the ecosystemic services are compensated to persuade the generating companies into including ecological objectives into their optimal operation curves. In addition, suggestions on considering a bypass tunnel to let the discharge flow into Guatapé river dry trajectory while adding a turbine to take advantage of this flow are also given. The results also show that the flood mitigation objective does not result in a competing objective against the hydropower and environmental flow objectives when there are average flow conditions in the Magdalena river. Floods commonly occur during extreme weather periods whereas the optimization of the Punchiná reservoir is performed for monthly average flow conditions at Magdalena river. Thus, to assess the hydropeaking effect in the water levels at La Pesca site, it is recommended that the reservoir optimization should also include extreme flow conditions at Magdalena river when experienced.

The Netherlands is a country prone to flooding. Recent assessments led to the insight that protection levels of many flood defences should be increased. Integrating reinforcement measures is a difficult task as many dikes are situated in densely populated areas. Conventional reinforcement measures include berm construction or the implementation of sheet pile walls. The first can become very expensive in case houses are situated close to a dike, the latter is rather expensive and irreversibly changes dike composition. Geotechnical failure modes piping and slope instability are most important failure modes for Dutch river dikes. In this thesis a case study is carried out on such a dike that is disqualified for those failure modes. The dike is situated in an urban area with limited space available for reinforcement works. It is studied whether pore pressure measurements behind the dike can be used to improve the reliability estimate for piping. Subsequently it is analyzed whether implementing pressure relief wells can be used to increase dike reliability for both considered failure modes. For the case study an advanced modelling framework was used consisting of groundwater modelling software and a random field generator. The case study was divided into two parts. The first part consisted of a dike section of 100 m based on a dike section at Wijk bij Duurstede. The second part consisted of the same dike section, only now extrapolated over a length of 2 km for which variations in soil conditions become more important. First, an analysis was conducted to define the optimal amount of pore pressure sensors behind the dike. It was found that for a dike section of 100 m a total of four sensors could be used to perform reliability updates, for a dike section of 2000 m it was found that a total of six sensors could be used. For the 100 m section piping failure probability improved from 5.21E-3 per year to 1.62E-4 per year. For the 2000 m section piping failure probability improved from 5.21E-3 per year to 1.89E-4 per year. Pressure relief well implementation behind the dike was considered as a measure to increase dike reliability for both failure modes. The system was designed based on a target reliability level for slope instability. The same modelling framework was applied. An analysis was conducted and it was shown that for the 100 m section a well spacing of 50 m would sufficiently increase dike reliability for slope instability. For the 2000 m section a well spacing of 45 m was found. For the 100 m section failure probability for slope instability increased from 8.81E-5 per year to 1.22E-6 per year, failure probability for piping increased from 1.62E-4 per year to 2.03E-7 per year. For the 2000 m section failure probability for slope instability increased from 8.81E-5 per year to 1.30E-6 per year, failure probability for piping increased from 1.89E-4 per year to 1.02E-7 per year. It was shown that for both trajectories target reliability levels for all failure modes were met. For this case study it was shown that pressure relief wells provide a good design alternative for dike reinforcement in urban areas. A life cycle cost analysis was applied and it was shown that implementation of relief wells is economically attractive compared to traditional design alternatives berms and sheet pile walls. For the first relocation of houses forms an important cost driver, for the latter initial construction cost are high. Furthermore it was shown that implementation of pore pressure monitoring prior to the design of a relief well system yields a positive value of information.

River floods are becoming increasingly devastating because of climate change (more frequent and extreme rainfall), population growth and the increasing economic importance of river basins. This situation requires maintenance and strengthening of flood-defence systems.

Adding certain types of vegetation at precise locations for their positive impact may be a cheaper, more flexible, and more environment-friendly way to strengthen dikes than the traditional increase in height. However, this nature-based (NB) option is not yet widely implemented due to the lack of precise knowledge of the potential of vegetation effects and their uncertainty.

This study uses a probabilistic method to better understand the effects of vegetation by including vegetation in the computation of the failure probabilities of Dutch river dikes. A framework was established to combine all these vegetation effects simultaneously in the computation of the total failure probability, considering different magnitudes of each effect. This enables the consideration of a wide range of vegetation scenarios, from which conclusions were drawn.

Overall, this thesis provides a useful and versatile tool for assessing the influence of vegetation on dikes that has a lot of potential and can be easily enhanced in the future.