Accurate prediction of typhoon-induced surge is an important issue for coastal disaster prevention. The purpose of this study is to develop a long-lead-time prediction model for storm surge by using adaptive neuro-fuzzy inference system approach (ANFIS). We carefully select effective controlling parameters, including the maximum wind speed, central pressure deficit, radius of maximum wind, location of typhoon center to the tidal station (i.e., distance and angle). In addition, the forward speed and direction of typhoon are considered to improve the long-lead-time prediction. The prediction performances are examined and discussed. Overall, predictions up to t + 9 hour ahead of lead time are satisfactory (with correlation coefficient CC > 0.7).
Coastal areas with natural resources and economic potentials are highly developed regions in a country. Over the world, more than one billion people are working and living in these areas. The continuously increasing population might approach 2 to 5 billion by 2080 (IPCC, 2007). However, coastal environment system would face destructive hazards induced by the combined influences of land/river, ocean and atmospheric forcing. One of such threats is typhoon-induced surges.
Storm surge has always been an important topic in both science and engineering due to its severe social impacts. In 2018, for example, Hurricane Michael made landfall near Mexico Beach, Florida, United States. It brought heavy rain, high winds, and an extreme storm surge up to 14 feet, resulting in at least 60 deaths and 15 billion economic losses. The catastrophic tragedies might repeat and even get worse, as typhoons and storm surges are expected to surpass historical records under a changing climate, i.e., global warming (Webster et al., 2005; Emanuel, 2005).
Accurate and efficient surge prediction with better understanding of its variation (e.g., maximum or possible range) plays a critical role in disaster mitigation. The basic idea for storm surge prediction is to capture its essential components (e.g., see Flather, 2001). Two main mechanisms are (i) pressure setup (i.e., a pressure drop of 1 mb leads to a 1-cm rise in sea level) and (ii) wind setup (i.e., strong onshore wind leads to significant sea level rise). To date, the prediction tools can be generally divided into three classes:
empirical formulas,
hydrodynamic models, and
artificial intelligence approaches.
