Uncertainty and sensitivity analysis of a coastal flood risk modelling chain

Uncertainty and sensitivity analysis of a coastal flood risk modelling chain.
Gouldby, B.P.and Liu, Y. and Forster, A. and Hornsby, J. and Mitchell, C.
In: EVAN 2019, 17-19 September 2019, Paris, France. (2019)

Full text not available from this repository.

Abstract:This paper describes the application of uncertainty and sensitivity analysis techniques to a coastal flood risk modelling chain to a site on the south coast of England. The modelling chain comprises multivariate extreme value modelling of sea conditions. Whilst this technique is now well-established, it is well-known that significant uncertainties arise when extrapolating historical datasets to extremes. Whilst these uncertainties can, to a certain extent, be evaluated through the statistical model fitting process, the resulting confidence limits are rarely utilised in practice. The analysis described here evaluates the uncertainty associated with the statistical extrapolation to extremes. This uncertainty is then propagated through a modelling chain that comprises: Wave transformation; Wave overtopping; Flood inundation; Economic damage. Each model component within a model chain has uncertainty associated with it. This includes uncertainty relating to the input data and the formulation of the model component itself, sometimes referred to as model structural uncertainty. To date, however, the overall uncertainty associated with the output of a chain of coastal flood models is not well understood. In this study the uncertainty associated with the multivariate extreme value model has been combined with uncertainty from these other model components, to provide estimates of uncertainty on flood risk. Sensitivity analysis is related to uncertainty analysis. The objective of the sensitivity analysis undertaken in this context is to gain an insight into which sources of uncertainty (both model components and data) within the modelling chain are most important in terms of contributing to the overall output uncertainty. So, for example, at the site analysed, it is possible to answer questions like “is the uncertainty associated with the multivariate extrapolation to extremes more influential than the uncertainty associated with the wave transformation model?” This information can then be used to support decisions relating to prioritisation of data collection and model component improvement activities. A generic technique, Variance Based Sensitivity Analysis (VBSA), has been applied to the modelling chain. The analysis shows that at this site the uncertainty associated with the wave overtopping model dominates all other sources.
Item Type:Conference or Workshop Item (Paper)
Subjects:Floods > General
Coasts > General
ID Code:1761
Deposited On:17 Sep 2019 11:56
Last Modified:17 Sep 2019 11:56

Repository Staff Only: item control page