Riparian forests in front of levees can dampen incoming waves and reduce erosion, thereby decreasing the risk of flooding. This makes them a sustainable addition to stand-alone dikes, potentially decreasing the overall costs for future dike reinforcement measures. However, there are currently no guidelines for the design, monitoring, and maintenance of dike-forest combinations as significant uncertainties remain. For instance, most studies on the effectiveness of forests in attenuating waves focus on mild wave climates or rely on scaled tests using simplified tree mimics. As a result, prototype-scale studies on wave damping by forests under extreme conditions—essential for design and assessment purposes—are lacking.

This thesis investigated wave damping by riparian forests, with a specific focus on pollard willow trees, which are commonly found along the riverbanks in the Netherlands and other parts of Europe. The primary aim was to reduce the uncertainties associated with the application of forest-dike combinations. The foundation and source of innovation for this thesis is the data from real-scale flume tests, conducted on a 40-m-long live pollard willow forest, subjected to significant wave heights up to 1.5 metres. The analysis of these tests revealed several important areas for further investigation. First, the vertical frontal-surface area (Av) distribution of leafless trees should be measured in detail, as leaves were found to minimally affect wave damping. Second, flume studies at various scaled often form the basis of calibration and validation of analytical, numerical and empirical wave-vegetation models, however, the extent to which small-scale tests accurately represent wave-vegetation interaction at real scale remains unknown. The data from the real-scale tests made it possible to design scaled tests (with 3D-printed tree mimics) and compare the results between both scales. Lastly, during real-scale tests, the live tree branches were observed to sway by nearly 180 degrees under the highest water levels and wave conditions, highlighting the importance of and need for further research into branch flexibility.

Numerical models of vegetation largely underestimate the vegetation surface by assuming that vegetation consists of only stems and a single branch order and by neglecting tapering of branches. In Chapter 2, we investigate methods to obtain accurate Av distributions over the height of live willow trees. One method used a combination of manual measurements and tree allometry relations to create tree models (resulting in a detailed representation of Av). This method was compared to the results of a relatively more practical method: Terrestrial Laser Scanning. The findings showed a large variation of (calibrated) bulk drag coefficients between measuring methods and highlight the importance of reliable frontal-surface area estimations and consequently for reliable wave attenuation predictions.

Until now, no prior studies have compared real-scale and scaled tests with woody vegetation. We therefore conducted scaled tests with complex 3D-printed willow tree mimics to explore scale effects in scaled tests with vegetation (Chapter 3). The maximum measured wave damping (30%) was shown to be roughly 1.5 times higher than the real-scale tests (20%) for water levels just above the knot of the trees. The amount of wave height damping decreased for larger water levels, following the same trend as that of the real-scale tests. The largest effects were attributed to increased viscous damping (due to smaller branch Reynolds numbers), and non-exact flexibility scaling. These notable deviations illustrate that real-scale tests, though expensive, may still be needed to validate the results of scaled tests for woody vegetation. Alternatively, accounting for these discrepancies can increase the reliability of scaled tests for wave damping studies on woody vegetation and reduce the need for more expensive real-scale tests.

Additionally, scaled tests with flexible conical shapes were conducted to study the effects of flexibility on wave damping in greater detail (Chapter 4). The first-mode cone deflection was determined at ~0.7 times the length of the cone to avoid higher-order modes in the measurements. The findings showed that cone deflections greater than 5 degrees had a large spread in force reduction and resulted in a significant decrease in measured forces of up to 50% compared to their rigid counterparts. This work demonstrated that the effective length principle, which has already been successfully applied to grassy vegetation such as salt marshes and seagrass, is a promising dimensionless parameter for predicting force reduction in conical shapes--and could potentially be extended to tree canopies.

Lastly, the experimental data was used as input for analytical wave damping models, which allowed us to discuss the opportunities for riparian forest-dike solutions in the Netherlands (Chapter 5). The outcome of our probabilistic study suggested that pollard forests in front of existing dikes offered the greatest benefit in mitigating failure caused by the erosion of grass on the outer-slope of the dike due to wave impact. We also discussed that the height of the trunk, which determines the location of the knot—where the frontal surface area, and consequently wave damping, are greatest—can serve as a key design parameter for forest-dike systems.

The thesis offers an overview of key parameters and their associated uncertainties, contributing to the ongoing integration of (riparian) forests into dike design and assessment methodologies.@en

Known for its low-lying and flood-prone geography, the Netherlands has a long history of flooding. Flood events in the 1990s, which led to large-scale evacuations, shifted the country's focus from simply strengthening dikes to creating more space for rivers to flow and storing water to prevent similar disasters. With changing climate conditions, rising water levels, land subsidence and sediment supply issues, the safety of primary flood defences needs to be regularly reassessed. The latest safety assessment, using a new methodology that assesses dikes based on the probability of flooding, found that many dikes do not meet the required safety standards. These dikes will need to be redesigned and reinforced, and there are opportunities for innovative and environmentally friendly solutions.
Previous research has extensively investigated the potential of different ecosystems, such as salt marshes, mangroves and willow forests, to reduce the impact of waves on coastal and riverine structures. Studies have shown that vegetation can significantly dissipate wave energy, thereby reducing the load on flood defences. However, the dynamic and unpredictable nature of vegetation has made it difficult to incorporate it reliably into defences. While there is general agreement on the benefits of vegetation for wave attenuation, its reliability in reducing failure probabilities requires further investigation.
This thesis tests the potential of riparian forests for wave attenuation and dike design optimization. The main objective is to investigate how wave attenuation by riparian forests can improve dike safety and to identify the critical variables that most significantly affect the probability of flooding. Key research questions include how wave attenuation can contribute to dike design and safety assessment, and which variables most influence the probability of flooding.
Vegetated foreshores can reduce the height of incoming waves, with positive implications for dike design and safety assessment. A model was developed to test this by combining current wave impact and run-up models with a wave energy balance that includes energy dissipation by vegetation. The model was subjected to a Monte Carlo simulation to determine the probability of dike failure due to erosion of the outer slope by wave impact and run-up. The model analyzed a dike section near Kampen and showed that riparian forests can significantly reduce the impact of waves on dikes. The results showed that the forest length, frontal-surface area and stiffness of the forest must be large enough to be effective. Strategically placing trees with the largest frontal-surface area at critical points can significantly reduce the probability of dike failure.
In conclusion, this thesis highlights the potential of riparian forests to improve dike safety and contribute to a more resilient and environmentally friendly flood protection strategy in the Netherlands.

Mangrove ecosystems are dynamic environments that provide numerous ecological, economic, and societal benefits, such as shoreline protection, carbon sequestration, and flood protection. These ecosystems are exclusively found in tropical and subtropical regions and thrive at the interface of land and sea. Despite their many benefits, mangroves face significant threats, leading to substantial losses over the past century. The degradation of mangroves compromises flood protection and shoreline stability, increasing the risk of severe coastline erosion and flooding. Current research highlights the role of mangroves in flood protection but presents conflicting theorieson their impact on coastline stability. Two competing theories exist regarding the effect of mangroves on coastline stability, namely the ”land-builders” and the ”land-consolidators” theory. According to the ”land-builders” theory, mangroves actively accumulate sediment and contribute to the expansion of coastlines. In contrast, the ”land-consolidators” theory suggests that mangroves play a more passive role in stabilizing existing landforms and serve as buffers against erosion. The debate on the primary function of mangroves remains unresolved. Furthermore, existing studies often date back over 50 years and are limited by specific hydrodynamic and geomorphic settings, leaving a gap in understanding broad trends. This research aims to bridge this gap through a global assessment of the influence of mangroves on coastline development, complemented by a local assessment to explore detailed interactions in specific case study areas. The study’s objective is to explore the dynamic impact of mangroves on coastline stability through an extensive analysis of historical datasets to identify trends and patterns
in coastline changes associated with mangrove ecosystems. The focus lies on observing trends between coastline stability and the presence of mangroves, the relationship between coastline stability and mangrove forest width, and the influence of the state of mangrove forests (contracting, expanding, stable) on coastline stability.

To estimate wave attenuation by mangrove forests, trees are often schematized as uniform cylinders (over the height) representing a tree’s stem. However, trees have more complex geometrical features that influence the interaction between waves and tree structures. Recent studies have shown that the frontal surface area distribution over the height Av(z) is a crucial parameter for estimating wave attenuation by vegetation (Kalloe et al., 2022; van Wesenbeeck et al., 2022) because it can account for complex geometrical features. This study aims to model the complex geometry of the Avicennia marina mangrove species. The research question then becomes:

How to construct geometrical tree models of Avicennia marina vegetation and how to obtain the projected frontal surface area distribution over the height, Av(z)?

In this study, manual measurements were performed to obtain tree structure parameters. Both canopy and root measurements were conducted for Avicennia marina vegetation with varying characteristics (age and density). Two saplings (approximately 1.5 years old) with variable canopy densities and two young trees (around five years old) with varying canopy densities were measured. We stored canopy measurements in a so-called tree data structure consisting of nodes (representing the branches) that contain information about the individual branches, such as branch orders, -lengths and -diameters and edges, which are links that establish direct relations from one branch to another. After that, we developed a search tree algorithm to loop through all tree data, compute parameters of interest and store the results in assigned data arrays. Parameters of interest to obtain after the data analysis were branch dimensions and diameter ratios of branches (between various branch classes). This information was necessary to create a blueprint for constructing geometrical tree models.
Tree modelling started with the construction of a root (pneumatophore) model. Additionally, the frontal surface area of these roots was computed and displayed as a function of the height. In addition to the root model, we constructed two canopy models. The first model is based on relations between branches and can be called deterministic or dependent (canopy model 1). The second canopy model (canopy model 2) generates branch dimensions based on the normal distribution of branch diameters and is, therefore, more probabilistic with branches independent of each other. Both canopy models construct branches based on information travelling from preceding branches. The models account for the proper placement of branches within 3D space by rotations and translations. The algorithm computes and outputs the projected frontal surface area over the height of a system of branches for a given direction (XZ-plane projection or YZ-plane projection). We validated the two canopy/tree models against validation measurements. Three trees were divided into seven vegetation layers, and branches were registered that intersected with the layers. As a result, estimation could be made on the frontal surface area per layer.

This study found that the frontal surface areas due to roots and canopies are significant and are well represented by the parameter Av(z). The contribution of the roots and canopy cannot be neglected in modelling wave dissipation by vegetation. The contribution of smaller branches to the total Av is significant, proving that considering only the tree stem is an underestimation. Moreover, the research shows that it is possible to develop a geometrical tree model based on a set of measurements and design rules that follow from observations, at least for the considered tree species. We could extrapolate and interpolate our results to generate tree models for a wide range of tree ages. Additionally, it is possible to create a forest by generating multiple trees.

Both tree models (1 and 2) showed little differences in Av(z) during the model validation process. Moreover, the validation measurements led, in general, to an overestimation of the total frontal surface area. Consequently, model validation was inconclusive regarding both tree models. However, we found that constructing tree models according to tree/canopy model 1 is preferred because it resembles an L-system more and is more practically applicable in computer models. We can improve future tree models by collecting more measurements regarding branch structures and root distribution as a function of tree age. This will result in increased model reliability.

Mangrove is a coastal vegetation type primarily located in the tropical regions between 5 North and 5 South. This coastal vegetation is capable of reducing the force of incoming waves. This is a result of the obstruction created by the roots, stems and canopies against waves to propagate through. Because of this capability, mangrove vegetation offers coastal protection against extreme annual flood events for 15 million people. Recent studies identified that in 2050, the present-day 100-year extreme sea level will occur annually in a large part of the tropical region. The increase of these extreme events is primarily driven by the projected sea level rise. The hypothesis is that before this chronic flooding is observed, storm-induced flooding might already be observed. The expected increase in the probability of these extreme events is defined as non-stationarity of extreme events. The hypothesis therefore is that the non-stationarity of extreme events could already be observed in historical data.

This study examined non-stationary extreme events between 1987 and 2018 for a total of 5809 hydrodynamic environments distributed around the global mangrove coastline. These mangrove environments defined the hydrodynamic wave characteristics and water levels. The results showed that 87.5% of the mangrove environments have a positive trend in the location parameter of the Generalized Pareto Distribution for the extreme events. The shift of the location parameter indicates an increase of the impact of the extreme events. These extreme events were defined as multivariate extreme events consisting of the significant wave height, the mean wave period and the skew-storm surge, respectively the parameters H_s, T_m and S. This combination is based on the fact that the coastal protection offered by mangroves is depending on the water level and the energy within the long-period waves.

The methodology proposed is capable to perform a non-stationary multivariate extreme value analysis and observe the evolution of the extreme events in these three dimensions between 1987 and 2018. The study showed that the average of the extreme events has been increasing at 70.6%, 72.6% and 64.6% of the mangrove environments in the discussed three dimensions. The average multivariate extreme event in mangrove environments increased by 0.06 m, 0.16 s, and 0.7 cm in three dimensions respectively H_s, T_m, and S. Furthermore, the number of extreme events increased, on average, between the first and second half of the time period, from 4.16 to 4.32 extreme events per year.

The global impact assessment of the non-stationarity of these multivariate extreme events was translated to the 27440 mangrove locations along the global coastlines. Statistical upscaling methods allowed the application of a numerical expensive hydrodynamic model to propagate the offshore conditions to onshore. The impact of the non-stationarity of extreme events was applied to a theoretical framework, introducing the possibility of defining the vegetation width as an optimization parameter. To meet the same safety standard for the 1/40 year design condition in 2018 concerning 1987, the vegetation width of the theoretical framework had to increase on average by 16.8 m. After post-processing the hydrodynamic runs, the results showed that the coastal safety offered by mangrove vegetation is depending on the combined impact of these three parameters, emphasizing the importance of expanding the mangrove vegetation to withstand the non-stationary multivariate extreme event.

The main recommendations from this study are based on the results. First, the study showed that the required vegetation width to ensure coastal safety has increased. It is therefore highly questionable if a stationary extreme value analysis is still valid in these environments, especially when the currently neglecting sea level rise is taken into account. Second, the observed increase in multivariate extreme events and the number of extreme events may reduce the persistence of mangroves and therewith their ability as coastal protection. It is therefore recommended to monitor the condition of the global mangrove forests more closely and to take action when mangrove vegetation is decreasing.

The wave attenuation of vegetation plays an increasingly important role in flood control in coastal areas. Past studies have found that the interaction of waves with vegetation mainly depends on hydraulic conditions and vegetation characteristics. Therefore, it is necessary to build vegetation models and quantify vegetation characteristics related to wave attenuation. Mangroves are one of the typical tropical intertidal vegetation. This study aimed to test the potential of using a low-cost, convenient smartphone-based structure-from-motion with multi-view stereo-photogrammetry (SfM-MVS) to accurately measure mangrove parameters related to wave attenuation.

SfM-MVS is a computer vision technique in which the point cloud coordinates of an object can be calculated from a series of 2D photos to generate a 3D point cloud model. This study first tested the optimal photography distance （about 25cm） and optimal weather conditions （cloudy and no sunlight） using the smartphone-based SfM-MVS method. Then, based on the optiaml usage conditions, 3Dpoint cloud models of 10 individual mangrove samples were reconstructed using this method, then the mangrove parameters related to wave attenuation were estimated according to the model: linear parameters (tree height, crown diameter, stem diameter, branches diameter) and the frontal area at a certain height. The true values of these parameters were measured using traditional hand measurements to evaluate smartphone-based SfM-MVS-derived parameter estimates.

The results show that the estimation of the linear parameters of the mangroves generally achieves high accuracy (RMSE of tree height, canopy diameter, stem diameter and thick branch diameter are all lower than 15%). There were significant negative biases in the estimates of tree height and crown diameter, and no significant biases in the estimates of stems and branches. The results of linear regression show that there is a strong positive correlation between the estimated values of all parameters and the true values. There are large errors and negative biases in the estimates of the frontal area at different heights. Overall, RMSE of frontal area at a certain height is 45.58% with a bias of -38.83%. This result showed that the SfM-MVS model resulted in a large underestimation of the frontal area at each height. A series of analyses showed that the main reason for the large error in the frontal area at a certain height was that the terminal branches with the lowest branch order (smallest diameter) were difficult to visualize in the SfM-MVS model. And this part of the branches actually occupies a large proportion of the frontal area.

This study demonstrates that smartphone-based SfM-MVS is capable of estimating mangrove tree height, canopy diameter, stem diameter, and thick branch diameter. Its accuracy is comparable to other existing methods such as TLS, ALS, camera-based SfM-MVS. However, the resolution is insufficient for very thin branches (<=5mm), which makes the method inaccurate in estimating the frontal area at a certain height. Factors such as photography distance and ambient lighting can affect the accuracy of the model. Compared with the currently used remote sensing technologies such as TLS, the smartphone-based SfM-MVS has the advantages of low cost, high flexibility, and low difficulty for those who lack professional knowledge or training, which makes it an alternative with great potential.

The risk of flood risk is increasing with global trends such as rising sea levels, land subsidence, increasing populations in coastal areas and economic growth. Traditional solutions to reduce flood risk are a levee or a dam; however, the research on hydraulic engineering is increasingly promoting nature-based solutions for flood protection. Salt marshes and mangroves can attenuate waves and could thus help to reduce failure probability. This research focuses on the attractiveness of mangroves in flood risk reduction. The essence of the reasoning followed in this research is that mangroves in front of a levee enable a lower levee and therefore reduce costs while maintaining the same failure probability. There are three aspects related to this assessment: the biological behavior of mangroves; the physical behavior of waves and storm surge in a mangrove forest; and the economics of flood risk, levees, and mangroves. Mangroves grow under specific conditions, such as high temperatures, saline water with fresh water input, between mean sea level and high tide, and with low wave impact. Mangroves attenuate waves, with the attenuation rates depending on mangrove height, stem diameter and density, in addition to the water depth, wave height and wave period. Lastly, the costs of restoring mangroves and building levees have been considered. In this thesis a method has been developed to define the optimal configuration of a levee-mangrove system. The goal is to reach a desired safety level while keeping costs as low as possible. Costs consist of building a levee, restoring mangroves, and the expected annual damage and maintenance. This depends on the ratio of mangrove and levee costs, the ability of mangroves to attenuate waves, and levee characteristics. Although all variables have equal significance in defining the mangrove forest width, the variety in levee and mangrove costs and significant wave height are the variables most likely to affect the optimal mangrove forest width. Mangroves are only useful to decrease the height of waves, thereby reducing wave run-up. This wave-attenuating effect decreases exponentially with the mangrove forest width. all Based on the literature, when considering common values for levee costs and mangrove restoration costs, levee characteristics and wave attenuation, wave attenuation is strong enough such that the optimal mangrove forest width is larger than zero. The costs of restoring mangroves depend on the required measures to enable mangroves to grow back. Hydrological restoration involves returning the mangrove area to the natural condition where mangroves are able to grow. This could, for example, be restoring a fresh water source that has been blocked by human intervention. If hydrological restoration is required and can be done cheaply, then this is the most effective option. Planting mangroves can also be a cost-effective option. Sheltering the mangrove area by using a permeable structure requires more financial resources. If large-scale filling or excavation is required, costs may increase significantly, and it is likely to be too expensive to use mangroves in flood-risk reduction. However, this cost assessment may change if the ecosystem services of mangroves are taken into account. As a rule of thumb, for mangroves to be economically effective, restoration costs [USD/m/m] in case of a 1 m wave height should remain below 0.003 times the variable levee costs [USD/m/m]. This value increases linearly with the wave height. For levee costs of 1 million USD/km/m, of which 0.6 million USD/km/m variable costs, the mangroves restoration costs should not exceed 0.18 USD/m/m, or 18,000 USD/ha. For Kaback, Guinea, the optimal mangrove forest width and height of the levee are assessed. These are 900 meter and 1.1 meter respectively, with the levee located just behind the mangroves. 900 meter is the maximum width possible, considering the physical conditions where mangroves grow. Since mangroves are growing back naturally already, costs for mangroves restoration is limited to fixed costs. The model developed in this study can be used to create a ``mangrove opportunity map”. This map can indicate—based on physical attributes—the costs, hydraulic conditions such as storm surge and wave height, and whether mangrove restoration for flood risk reduction is effective. Governments or organizations such as the World Bank can use this method to explore whether mangrove-based strategies are an option. Further research can focus on improving the cost estimates on levees and mangroves. In this study, costs are assumed to be linear with the levee height or mangrove area. In practice, levee costs might especially increase exponentially with the levee height, which would make mangroves more effective.

Including vegetation in flood defense systems has the potential to be a more cost-effective solution than conventional dike reinforcement measures, as vegetation contains wave damping properties. However, more insight is required on the complex physical processes due to wave-vegetation interaction in order to predict the amount of wave damping. Past research found that these processes are dependent on both flow conditions and vegetation characteristics. Therefore, reliable quantification of these vegetation characteristics is
of importance to understand these processes and thus to improve predictions for wave dissipation due to vegetation. This study aims to gain insight on practical methods for quantifying relevant vegetation characteristics
for wave damping. Data is used from full scale physical experiments conducted in the Delta Flume at Deltares, with 40 meters of willow forest. The vegetation characteristics are quantified in this study both by manual measurements on the willow trees and by executing Terrestrial 3D laser scans (TLS) of the whole forest. The frontal area of vegetation is found in literature to be a relevant parameter for determining the wave attenuation, therefore the focus in this study lies on schematizing this parameter over the vertical, which is seen as representations of the average willow tree. This schematization is referred to in this report as “tree model”. In this study, four tree models are obtained. The four tree models include: tree model 1a (manual measurements on the primary branches, excluding side branches), tree model 1b (branching method based on Strahlers ordering scheme, including side branches), tree model 1c (adjusted branching method) and tree model 2 (from terrestrial laser scanner). The reconstructed area from the TLS point cloud is determined by using Matlab built-in alpha shape function. The tree models are compared in terms of frontal area distribution over the vertical and of their effect on the corresponding wave attenuation. The latter comparison is achieved by using the numerical wave model SWAN and confronting the results with the measured wave attenuation from the physical experiments. With regards to the manual measuring methods, the frontal area of the average tree from tree model 1a serves as a lower limit, while tree model 1b gives the upper limit. The TLS outcome (tree model 2) underestimates the frontal area as computed by the other tree models. In particular, the underestimation is 70% when compared with tree model 1b, and 30 % when compared to tree model 1a. Adjustments on tree model 1b leads to tree model 1c. This tree model accounts for the tapering form of the branches and results in a total frontal area in between tree model 1a and tree model 1b. In this study, the representative area for the willow trees can be best captured with tree model 1c. This tree model results in a drag coefficient (CD) of 1.15 averaged over the tests with leafless willows and showed a negative correlation with the Keulegan-Carpenter number (KC). However, the capabilities of the TLS should be analyzed further and this study encourages to use a larger data set of trees in order to find a relation between the laser penetration and corresponding mismatch. Gathering of vegetation parameters by hand is in fact not economically attractive forwoody riparian vegetation, as these trees are characterized by complex canopy structures and high elevations. The TLS can serve as a practical tool for obtaining these relevant vegetation parameters for applying willows in hybrid solutions.