2.4 Prediction of Skew Surge by a Fuzzy Decision Tree

Storm surge resulting from mid-latitude weather systems have the potential to cause considerable damage and fatalities as a result of coastal flooding (Lavery and Donovan, 2005; Lescrauwaet et al., 2006; Jonkman and Vrijling, 2008). The real-time, accurate prediction of storm surge is of primary importance in flood forecasting and warning systems. Whilst existing deterministic, hydrodynamic forecast models are highly successful at predicting storm surge (for example, Horsburgh et al. (2008)) the development of new models and further improvement to existing models is expensive. Therefore, opportunities exist for alternative, complementary approaches to improve upon predictive accuracy and our understanding of the system dynamics. Artificial intelligence (AI) and probabilistic techniques have been applied to the real-time or hindcast prediction of water level including meteorological effects with some success. For example, artificial neural network (ANNs) models have been applied to this problem in the Gulf of Mexico (Tissot et al., 2001; Cox et al., 2002; Tissot and Zimmer, 2007), Western Australia (Makarynskyy et al., 2004), Taiwan (Lee, 2008b,a), the Bohai Sea (Liang et al., 2008) and the North Sea (Ultsch and Röske, 2002; Prouty et al., 2005). A real-time forecast system now successfully operates in the Gulf of Mexico (Tissot et al., 2008). Alternative approaches have included Support Vector Machines (SVMs) applied to the Adriatic Sea (Canestrelli et al., 2007) and Taiwan (Rajasekaran et al., 2008); Genetic Programming (GP) applied to the Gulf of Mexico (Charhate and Deo, 2008); and, chaos and fuzzy Naive-Bayes models applied to the North Sea (Siek et al. (2008) and Randon et al. (2008) respectively). For mid-latitude, coastally-trapped storm surges in shelf seas with moderate tidal ranges, such as those that occur in the North Sea, Adriatic Sea and Sea of Azov, it has been shown that surges tend to cluster on the rising limb of the tide due to a small phase shift in turn driven by the change in water depth due to the meterologically-driven surge (Horsburgh and Wilson, 2007). Storm surge predictions by data-driven techniques, such as those referred to, tend to use the residual water level record, which displays this clustering and may include a leaked tidal component in the signal resulting from non-linear interactions. It is therefore proposed that instead an alternative metric, the skew surge record, be used to train and test a storm surge predictor. Skew surge, is defined as the difference in elevation between the maximum observed water level in a tidal cycle and that predicted by tide tables and is considered to be the most appropriate measure of storm surge for flood warning purposes (since the peak total water level is simply reconstructed from the skew surge plus the high water level for that tidal cycle). The components of total water level are schematised in Figure 1.