The use of artificial neural networks to retrieve sea-level information from remote data sources

Abstract

The knowledge of near-shore sea-level variations is of great importance in applications such as ocean engineering and safe navigation. It also plays an essential role in the practical realisation of the height reference surface in geodesy. In the cases of gaps in tide-gauge records, estimates can be obtained by various methods of interpolation and/or extrapolation, which generally assume linearity of the data. Although plausible in many cases, this assumption does not provide accurate results because shallow-water oceanic processes, such as tides, are mostly of a non-linear nature. This paper employs artificial neural networks to supplement hourly tide-gauge records using observations from other distant tide gauges. A case study is presented using data from the SEAFRAME tide-gauge sta-tions at Hillarys Boat Harbour, Indian Ocean, and Esperance, Southern Ocean, for the period 1992 to 2002. The neural network methodology of sea-level supplementation demonstrates reliable results, with a fairly good overall agreement between the retrieved information and actual measurements.