HR Wallingford has a long history of developing and applying wave overtopping methods. Most recently we have led the development of the well-known EUROtop manual. Wave overtopping is a complex phenomenon subject to significant uncertainty. To assist in the analysis of uncertainty, a Bayesian statistical method was applied in conjunction with an Artificial Neural Network (ANN) approach for wave overtopping. Evolutionary optimisation techniques were applied within the ANN fitting process. The outcome from this component of the project was a new overtopping model, BAYONET, Kingston et al (2008), that captures the primary uncertainties of the overtopping process providing greater confidence in the output results from the flood risk analysis. This model has already been applied on projects for the Environment Agency where it has been shown to significantly improve estimates of coastal flood risk.
Flood risk management often involves the consideration of many different structural and non-structural mitigation options (e.g. defence improvements, offline storage, flood warning, property flood proofing). Moreover, there are often a range of measures against which the mitigation measures can be assessed, cost, benefits, in terms of economic damage and fatality reduction and environmental impacts (both positive and negative). It is often a requirement to assess these mitigation measures over long time scales (>50 years) to account for future variability, climate change and increased urbanisation, for example. Multi-objective optimisation techniques were applied, together with a comprehensive risk analysis model, to explore the performance of different flood risk mitigation strategies. Real Option analysis techniques were used to account for uncertainties relating to climate change, for example, Woodward et al. These methods will enable more robust climate change adaptation strategies to be developed.
Manual calibration of numerical models can be time-consuming and subjective. In this component of the study, automated calibration techniques were explored using evolutionary optimisation algorithms. It was evident that these techniques can be used to improve the robustness of the calibration process as well as increase efficiency in terms of the amount of time taken for calibration.
|Authors||Ben Gouldby, Michelle Woodward, Greer Kingston, Tim Pullen, Nigel Tozer, Peter Hawkes.|
|Keywords||Optimisation, wave overtopping, flood risk management, calibration, real options.|
HR Wallingford (2007) Machine learning techniques: Review and potential applications at HR Wallingford, IT 542
HR Wallingford (2009) BAYONET: Bayesian wave overtopping neural network, IT 587
HR Wallingford (2009) Automated calibration of wave models, IT 603
Kingston G, Robinson D and Gouldby B (2008) “Reliable prediction of wave overtopping volumes using Bayesian neural networks”, Proc. of FLOODrisk 2008, 30 Sep - 2 Oct, Oxford, UK In: Samuels et al. (eds). Flood Risk Management: Research and Practice. Taylor & Francis Group, London.
Woodward, M, Kapelan Z, Gouldby B (2013) “Adaptive flood risk management under climate change uncertainty using real options and optimisation, Risk Analysis, accepted
Woodward, M, Gouldby B, Kapelan Z, Hames D (2012) “Multi-objective optimisation for improved management of flood risk”, ASCE J. Water Resour. Plann. Man., Accepted
Woodward M, Gouldby B, Kapelan, Z, Khu, S-T. and Townend I. (2012) “Real options in flood risk management decision making”, J. Flood Risk Man. In 4 (4), pg 339-349.