| Summary: | The knowledge of sea level variations is of great importance in geoenvironmental and ocean-engineering applications. Estimations of sea level change with different warning times are of vital importance for the population of low-lying regions and islands. This contribution describes some recent advances in the application of a meshless artificial intelligence technique (neural networks) to the tasks of sea level retrieval and forecast. This technique was employed because it has been proven to approximate the non-linear behaviour in a geophysical system. The data used were taken from several SEAFRAME stations, which provide records for the Australian Baseline Sea Level Monitoring Project. A feed-forward, three-layered, artificial neural network was implemented to retrieve and predict sea level variations with different lead times. This methodology demonstrated reliable results in terms of the correlation coefficient (0.82-0.96), root mean square error (about 10% of tidal range) and scatter index (0.1-0.2), when compared with actual observations.
|
|---|