Deep Learning and Machine Learning Based Prediction of Significant Wave Height along the Grand Coast of Dakar, Senegal

Recently, deep learning (DL) has become an essential tool for processing large datasets and is playing a crucial role in scientific research. It significantly contributes to protecting the marine environment and forecasting oceanic phenomena. This study applies an autoregressive integrated moving average model (ARIMA) time series model to forecast significant wave heights ( SWHs ) at two beaches in Dakar, Senegal: Malika and Yoff. Additionally, the long short-term memory network (LSTM) is used for comparative analysis. Both models estimate SWH for future predictions ranging from 12 hours to 60 days. The study utilized ERA5 reanalysis data, comprising 52584 elements of SWH and 368088 elements across seven other physical parameters. This hourly data, spanning from January 1, 2018, to December 31, 2023, was used to train and evaluate the models. For a 30-day forecast at Yoff, the LSTM model achieved highly accurate results, with a root mean square error (RMSE) of 0.0403 m and a correlation coefficient (CC) of 0.9750. In comparison, the ARIMA model yielded an RMSE of 0.0407 m and a CC of 0.9745. The results demonstrate that the wave energy flux at Dakar is significant. This work enhances ocean safety in West Africa, by utilizing advanced DL techniques.