Sajni Malde (University of Sheffield and HR Wallingford) Gaussian process emulators applied to coastal flood risk modelling Hicks J11, 3.30pm
Abstract: HR Wallingford and the School of Mathematics and Statistics are collaborating on an ongoing Knowledge Transfer Project (KTP). This presentation will provide an overview of the project, and in particular, on the use of Gaussian process emulators (GPEs) for approximating computationally expensive numerical models. It will provide a brief description of the theory of GPEs It will then go on to describe a specific example of the application of a GPE to the SWAN wave transformation model. Comparisons between HR Wallingford’s traditional SWANMET look up table approach and the GPE will demonstrate the significant improvements in both accuracy and computational efficiency that can be made using a GPE. The presentation will also describe design selection methods that have been used in order to solve common problems faced when applying these to consultancy projects.