Introduction

For the design, safety assessment and rehabilitation of coastal structures reliable predictions of wave overtopping are required. Several design formulae exist for simplified types of dikes, rubble-mound breakwaters and vertical breakwaters. Nevertheless, often no suitable prediction methods are available for structures with non-standard shapes.

The XGB-Overtopping model is a conceptual design tool to estimate wave overtopping discharges for a wide range of coastal structures. The XGB-Overtopping model outperforms existing Artificial Neural Networks. Resampling techniques are applied for the assessment of the uncertainties of the predictions. Only one schematisation is used for all types of coastal structures, where not only dikes, rubble-mound breakwaters or vertical breakwaters are defined, but also other non-standard structures are included. Besides the effect of the most common parameters (i.e. wave height, wave period and crest freeboard) also the effects of many other wave and structural characteristics are considered. XGB-Overtopping is the successor to the Overtopping Neural Network (Van Gent et al., 2007). The model is based on about 10,000 physical model tests (Steendam et al., 2004) at several institutes (Deltares | Delft Hydraulics, The Netherlands; Aalborg University, Denmark; Danish Hydraulic Institute, Denmark; Hydraulic Research Wallingford, UK; Leichtweiss Institute für Wasserbau, Germany; Modimar, Italy; University of Edinburgh, United Kingdom; Universidad Politécnica de Valencia, Spain; Ghent University, Belgium; and other sources from countries including Iceland, Japan, Norway and U.S.A).

The employed prediction method is based on Gradient Boosting Decision Trees (GBDT). Because of the ensemble of Decision Trees in GBDT, it is a type of machine learning method that allows for a flexible resolution depending on the local density of training data. Hence, it is suitable to handle the large density differences in overtopping data. Details of the XGB-Overtopping model and the methodology followed for the development of the prediction tool are described in Den Bieman et al. (2021).

The predictions based on the XGB-Overtopping model can be used for the conceptual design of coastal structures; they may not be used in the final design stage, since the results should be verified based on dedicated physical model tests (Deltares facilities) for the particular wave conditions and structure geometry of the structure to be built. Nevertheless, the XGB-Overtopping model can be seen as the best generic model to predict wave overtopping discharges for the conceptual design of coastal and harbour structure.

References

Den Bieman, J.P., M.R.A. van Gent and H.F.P. van den Boogaard (2021), Wave overtopping predictions using an advanced machine learning technique, Elsevier, Coastal Engineering, 103830, https://doi.org/10.1016/j.coastaleng.2020.103830

Van Gent, M.R.A., H.F.P. van den Boogaard, B. Pozueta and J.R. Medina (2007), Neural network modelling of wave overtopping at coastal structures, Elsevier, Coastal Engineering, 54, 586-593, https://doi.org/10.1016/j.coastaleng.2006.12.001

Steendam, G.J., J.W. van der Meer, H. Verhaeghe, P. Besley, L. Franco and M.R.A. van Gent (2004), The international database on wave overtopping, World Scientific, Proc. 29th ICCE, Vol.4, 4301-4313, https://doi.org/10.1142/9789812701916_0347