Kim, T. and Lee, W.-D., 2023. Prediction of wave conditions using a machine learning framework on the east coast of Korea. Journal of Coastal Research, 39(1), 143–153. Charlotte (North Carolina), ISSN 0749-0208.
Currently, technology for rapid and accurate wave information prediction is required because of the increase in human and property damage caused by high waves as a result of climate change on the east coast of Korea. Over the past few years, the volume of data produced has become large as the world enters the era of the fourth industrial revolution, and research on machine learning models that can use these data is actively being conducted. In this study, three machine learning models (XGBoost, support vector regression, and linear regression) were used to make accurate estimates of waves on the east coast of Korea, and the model suitable for the data was selected. The input data that were used to train the models included air barometric pressure, wind, and wave data collected using a deep-sea buoy (Donghae), and the data were used to construct models for predicting offshore wave heights and wave periods. Of the three models, XGBoost exhibited a Nash-Sutcliffe efficiency (NSE) of 0.89 and a root-mean-square error (RMSE) of 19.7 cm for wave height and an NSE of 0.81 and an RMSE of 0.66 s for wave period. XGBoost was selected as suitable for predicting offshore wave heights and wave periods at Maengbang on the east coast of Korea, and excellent and reliable consistency was observed between the observed and the predicted values. Because machine learning models have short computation times and high accuracy, they can become an alternative to conventional physics-based wave models.