Forecasting the wave height of the Gulf of Mexico using wavelet neural network

Predicting factors like height and frequency of the waves in seas is one of the most essential information needed in the marine environment. In the present paper, the height of the waves in the Gulf of Mexico were analyzed by transforming wavelet into different sub-levels of approximation and details; and the wave height of the upcoming 3, 6, 8 and 12 hours lead time was predicted by artificial neural network. By comparing the real data measured by Buoy with the results of neural network prediction, it was found that coif(5) with 5 subsurface for 3 lead times and 6 Dmey wavelets with 6 and 7 subsurface are appropriate for predicting wave heights for 6, 8 and 12 lead times respectively. Then the meteorological parameters of air pressure, air and water temperature variation, wind direction, and wind speed were also added as model inputs. With this method, it was inferred that the parameters of wind direction and the variation in air and water temperature have no effect on the forecast accuracy. And by adding the current recorded wind speed and air pressure, the forecast accuracy increased compared to the case where only the decomposed subsurface was used. Finally, the predicted values of wave height at the time of the occurrence of Hermine, Choline and Matthew hurricanes in 2016 were evaluated and the results showed that at the time of the occurrence of Hermine and Choline hurricanes, the prediction accuracy of the model was significantly reduced.