A surrogate-aided approach for wave predictions in the Gulf of Mexico

Reliable coastal structure design, safe navigation, sediment transport predictions and future planning all depend on the access to accurate and long-term wave data. Wave measurements are not usually available everywhere and on a long-term basis; and those data which are available usually include periods of data gaps. Hence, numerical simulations or surrogate models are to be adopted to obtain such data in locations of interest. It is while the most important role in obtaining reliable simulated wave data is played by well-established and reliable simulated wind field data sets over the study area. Numerical weather prediction models can be used as a tool to estimate marine surface winds and for providing input parameters to the sea surface wave and ocean circulation numerical models. It is while, the model results which are to be verified with observational data usually show remarkable bias from the real situation in certain circumstances; something which can be attributed to the very dynamic and unstable nature of the atmospheric phenomena affecting simulation results.

The present study aims to develop a surrogate-aided model for wave predictions over the Gulf of Mexico (GoM) as a semi-enclosed body of water. The emphasis on ‘semi-enclosed’ is to avoid swell waves, generated elsewhere and under different meteorological conditions. To this end, a three year set of wind field data over the GoM was adopted from ECMWF Global Model together with real field wave observations from offshore and nearshore areas of the Northern GoM (NGoM). An Artificial Neural Network (ANN) approach was employed to establish a surrogate-aided model to provide reliable wave estimations. The ANN-based model was trained, assuming wind data at certain locations as the input to the network and wave observations at selected station as the target. The model has been trained for two years out of the three and validated for the third year. The results obtained from the surrogate model are in fair agreement with the observations.

Keywords: wave generation, surrogate-aided model, the Gulf of Mexico, Artificial Neural Network