Correlation between model uncertainties wave height and wave period

Kennis voor Keringen, een onderzoeksproject van Rijkswaterstaat onderzoekt de onzekerheden in de sterkte en belastingen.
Dit onderzoek heeft zich gericht op de modelonzekerheden voor golven. Het advies is om de correlatie in modelonzekerheden golven tussen golfhoogte en golfperiode in het vervolg mee te nemen.
Per gebied van Nederland worden default waarden gegeven. Beheerders en andere gebruikers kunnen de informatie uit dit rapport in de beoordeling (in een toets op maat) of in het ontwerp gebruiken samen met de huidige software van het Beoordelings- en Ontwerpinstrumentarium. Sinds mei 2021 is deze functionaliteit beschikbaar in de software.

As part of the new safety assessment approach for the Dutch flood defences, the uncertainties in the hydraulic loads are taken into account. Amongst these uncertainties are the uncertainties with respect to the translation of stochastic variables (e.g. wind speed and water level) to wave conditions at the toe of the flood defences: so-called model uncertainties. The implementation of the model-uncertainties for wave conditions assumed (for various reasons amongst which restrictions in time and planning for implementation) no correlation between the model-uncertainty of the wave height and wave period. The data used to derive the model-uncertainties for the coastal water systems (Deltares, 2013a) showed however that this assumption is not valid. This study aims at providing the correlation between the already derived model uncertainties for wave height and wave period.
The derivation of the correlation between the model uncertainties for wave height and wave period is (in its basis) quite simple. We reproduce the found model-uncertainties to ensure we are using the same dataset. Finally, we use standard techniques to derive the Pearson correlation coefficient. This is done per application area, similar to the derivation of the model-uncertainties themselves. Using the Crude Monte Carlo tool as described in Deltares (2017), we subsequently apply the correlation model and compare the results with the present probabilistic model. This analysis addresses the required crest height.
The computed correlations between éHmo and éTp were 0.34 and 0.57 [-] except for Lake Marken (remainder), which was negligible (0.01 [-]). The found correlations seem reasonable, as wave physics would prevent either fully or not correlated model-uncertainties. The impact of adding correlation on the HBN was relatively limited (especially compared to the impact of the bias correction): in the order of 0.1 to 0.2 metres for coast/lakes and even less for narrow river stretches. As expected, the inclusion of correlation leads to less high (more realistic) wave steepness in the design points of the probabilistic calculation. It is advised to the thoroughly evaluate impact of this correlation together with other potential improvements of the existing load model (several of which are discussed herein) before making any changes.