The design of offshore and coastal hydraulic structures is very much dependent on the hydraulic boundary conditions, such as significant wave height, mean wave period and wave direction, among other parameters. A proper design value of these parameters is required during the design process, based on the corresponding safety philosophy and the lifetime of the structure. In this context, an extreme event is characterised by a combination of unfavourable parameters. However, the interdependencies between the parameters are not always accounted for in the design process, despite the fact that some parameters are clearly related. This potentially leads to an overly conservative or optimistic design.

To complicate matters further, a given sea state might consist of a combination of wind wave and swell systems, sometimes coming from different directions and with different spectral shapes. Different combinations of crossing wave systems might lead to the same total significant wave height, mean wave period and mean wave direction. Only analysing the total wave parameters might oversimplify the situation in the presence of combined wave systems. In this thesis a methodology has been developed to establish extreme offshore wave conditions given the presence of these combined wave systems.

A time series that partitions the total wave into a wind wave- and swell component is used as input for the analysis. The location of interest being off the coast of southern Brazil, where combined sea states are observed regularly. The main objective is to compute design values for all wave parameters of interest. With these design values a number of extreme offshore sea states are described in terms of a single total wave system and equivalent combinations of two wave systems. The former resulting in a single-peaked wave spectrum and the latter in an equivalent double-peaked wave spectrum. The extreme offshore sea states are transformed to the nearshore and compared. The single-peaked and equivalent double-peaked wave spectra may result in very similar values for the wave energy nearshore, albeit with different spectral shapes and directions. For the investigated directional combination, this means that the more elaborate approach with two wave systems potentially affects the design of coastal infrastructure if it is sensitive to spectral shape and direction, although the uncertainty of the result is not quantified. Equivalent wave systems could be compared for other directional combinations in future research to investigate if the more elaborate approach results in a more cost-effective design of coastal infrastructure.

The quality of the multivariate vine copula model, used to compute the set of design values for the wave parameters of interest, is assessed in multiple ways. It is recommended not to pick a single set of design values at a point of high joint probability density. Instead it is suggested to use conditionalised samples from the vine copula model to determine the most unfavourable combination of load parameters, which has to be evaluated case-by-case.

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.

A low-lying country as The Netherland is prone to coastal flooding, and its risk may be enhanced by global-warming induced climate change. Sea level rise has been historically considered as the key factor in coastal retreat, but waves also play an important erosive role along the coast, which can also be affected by the changing climate. During the last years, important advances have been achieved in climate modeling, with a very detailed characterization of the different components of the climate system, for the present and for different future scenarios. However, characterization of future ocean waves is still a matter of discussion and ongoing research. In this thesis, a statistical downscaling methodology based on weather types has been chosen to model the present wave climate and explore potential changes in future waves. These changes are quantified in terms of the impact of these variations in the longshore sediment transport. The methodology is applied to Noordwijk, selected as a representative location of the central Dutch coast. The statistical downscaling methodology is based on a classification procedure of the predictor into similar atmospheric patterns over the wave generation areas, namely the weather types. Then, the wave data is grouped according to the occurrence of the weather types. The predictor is built from the sea level pressure fields (SLP) and the squared SLP gradients, while the predictant wave climate is characterized by significant wave height, wave peak period and mean wave direction of wind sea and swell components, resulting in unimodal or bimodal sea states. The chronology of the weather types is modeled using an autoregressive logistic model, which incorporates the seasonality, the interannual variability and the persistence observed from the historical data. For each weather type, wave parameters are modeled using the categorical distribution for the sea-state type, non-parametric kernel density functions for the central-mass regime and two generalized Pareto distributions for the lower and upper tails of wave height and wave period data. The statistical dependence between wave parameters for each sea state is included using a vine-copula approach, where the bivariate dependence of H_s and T_p is modeled using the AC skew t copula and the remaining relations are considered to be Gaussian. The effect of climate change is studied using SLP predictors from the global circulation model ACCESS1.0, under the RCP8.5 scenario and period 2070-2099. The statistical model is applied to identify changes in the occurrence probability of the weather types in the future. The importance of these changes are quantified in terms of the wave-induced longshore sediment transport using the process based model Unibest TC. The longshore sediment transport distribution for each weather type is computed and afterwards the changes in gross and net transports are estimated using the present and future probabilities of the weather types. In terms of the weather types, the results of this work suggest changes in the occurrence probability of the weather types in the future, with variations of ±60% with respect to present climate, and no relevant changes in the sequence of weather types, neither in terms of the transition probability matrix nor in terms of the persistence of each weather type. In terms of longshore sediment transport, significant changes are detected in the transport associated with some of the weather types. For the remaining weather types, the changes are in the order of the model uncertainty. Taking into account the contributions from all weather types, a net increase of the southward directed net longshore sediment transport with respect to the historical period is detected. This increase is driven by a decrease in northwards transport and a lower decrease in southwards transport. This result is in line with a poleward shift of the North Atlantic storm track reported in other studies. Most importantly, weather-type based climate classification has in this thesis been succesfully proven to be a reliable tool to analyze the wave climate at the location of interest. Furthermore, the statistical downscaling also provides a climate emulator that captures the climate dynamics at different time scales, which can be used for stochastic simulations of the atmospheric and wave climate, either for the recent past or future projections.