Predicting sea level variations with artificial neural networks at Hillarys Boat Harbour, Western Australia

Abstract

In the present study, the artificial intelligence meshless methodology of neural networks was used to predict hourly sea level variations for the following 24 hours, as well as for half-daily, daily, 5-daily and 10-daily mean sea levels. The methodology is site specific; therefore, as an example, the measurements from a single tide gauge at Hillarys Boat Harbour, Western Australia, for the period December 1991-December 2002 were used to train and to validate the employed neural networks. The results obtained show the feasibility of the neural sea level forecasts in terms of the correlation coefficient (0.7-0.9), root mean square error (about 10% of tidal range) and scatter index (0.1-0.2).