Abstract
The accuracy of storm surge predictions based on numerical simulation models depends heavily on the precision of typhoon forecasts, e.g. the course, the central atmospheric pressure, etc. In order to avoid the uncertainty of typhoon forecasts, a simplified prediction model of storm surges based on the observed meteorological data was developed by applying an Artificial Neural Network(ANN). The model predicts the storm surges occurring in a few hours on the basis of the present condition of atmospheric pressure, i.e. the distribution, its change for several hours, etc. The model was applied to some points along the Ariake Sea coast and its accuracy was examined through comparisons with the observed data. Then, it was verified that the proposed model can predict small storm surges as well as large ones sufficiently.
