Applying emulators for improved flood risk analysis

Applying emulators for improved flood risk analysis.
Malde, S.and Wyncoll, D. and Oakley, J. and Tozer, N.P. and Gouldby, B.P.
In: FLOODrisk 2016, 18-20 October 2016, Lyon, France. (2016)

Full text not available from this repository.

Abstract:Flood risk is generally recognized to be the product of probability and consequence where the probability relates to probabilities of flood hazards e.g. extreme rainfall, river flows, coastal waves and sea levels or multivariate combinations of these variables. The probability can also relate to the performance of flood defense infrastructure and likelihood of failure. By definition, flood risk analysis involves the integration of multivariate probability distributions over a domain defined by a consequence function. Often, solutions of this risk integral involves the use of Monte-Carlo sampling techniques, whereby tens of 1000’s of potential flood events are generated through statistical sampling techniques. It is, in principal, necessary to evaluate the consequence of flooding for each of these sampled events. There is, however, typically a significant computational time involved in running physical process models that are capable of simulating the consequences of flooding. It can therefore become computationally impractical to evaluate flood risk using this approach. To overcome the computational challenges, meta-modeling approaches can be applied. These approaches include: Piecewise Polynomials, Neural Networks, and Gaussian Process Emulators (GPE). This paper presents an analysis of the benefits of using a GPE of the SWAN spectral wave transformation model within the context of a coastal flood risk analysis modelling chain. Within traditional coastal modelling a “look-up table approach” based on limited, but significant, number of conditions is often used to replicate the SWAN model, due to computational constraints. This approach typically involves specifying and simulating conditions defined across a regular matrix, and then linear interpolation is undertaken to obtain values for intermediate conditions. In this paper we compare the look-up table approach to the GPE and analyse the performance of both approaches in approximating the SWAN model within the context of a coastal flood risk modelling system. In both cases, selecting an appropriate training design set is important, hence the efficiency of selecting a design and the span of the design are taken into consideration in the analysis.
Item Type:Conference or Workshop Item (Paper)
Uncontrolled Keywords:Gaussian Process Emulators; Emulators; Computer models; Flood risk
Subjects:Floods > Flood risk assessment and mapping
Floods > General
ID Code:1306
Deposited On:18 May 2016 14:31
Last Modified:18 Jan 2017 13:08

Repository Staff Only: item control page