The modelling of ocean waves is now carried out routinely at meteorological centres around the world. However, little is know about the source of the uncertainty in the predictions of waves produced, and sources can be numerous depending on the specific application. Historically it was felt that the dominant source of uncertainty originated from incomplete knowledge and expression of forcing winds. However more recent studies have focused on the underlying physical processes and their representations, with some authors questioning whether the limitation of the current modelling approach has been reached. Recently, methods for the statistical analysis of complex computer models, including models such as those used for wave prediction, have been developed. In this thesis these methods are applied to perform the first ever uncertainty analysis of a wave model. These new methods are applied to the state of the art wave model Wavewatch IIIr. This thesis principally explores the effect of tuning parameter uncertainty relating to the “Tolman and Chalikov” input and dissipation parameterisation, the discrete interaction approximation scheme for nonlinear wave-wave interactions and uncertainty about wind forcing, on wave simulation output, in a range of idealised cases, and realistically on Lake Michigan. The effectiveness of the statistical methods is first demonstrated in simple cases, before analysis is performed for progressively more complex simulations. In each case, uncertainty measures are computed with respect to simulation output in terms of summary wave statistics, typically including significant wave height and peak period. The analysis reveals nonlinear response and the relative importance of the various input, which in turn shows the active physical processes, and where the greatest sources of uncertainty lie. Both uncertainty about wind forcing and the process of nonlinear wave-wave interactions are found to be dominant in all cases, although energy dissipation is important in growing sea states. Finally, observational wave height data is used to perform a parameter calibration for simulations of stormy conditions on Lake Michigan, leading to improved performance.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:658809 |
| Date | January 2015 |
| Creators | Timmermans, Ben |
| Contributors | Challenor, Peter |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/378996/ |
Page generated in 0.0021 seconds