2023 Volume 79 Issue 17 Article ID: 23-17023
Highly accurate weather and wave predictions are necessary for marine activities. Although a dynamical model with many ensemble members can consider long-term forecast uncertainty and provide probabilistic information, it is computationally expensive and cannot cancel out the model biases. In this study, we propose a wind and wave forecasting system combining dynamical ensemble models and observations by long short-term memory (LSTM). The proposed system provides probabilistic forecasts that are more accurate than the base model, which has biases caused by the coarse numerical grid. The Brier score for the short-term forecast up to 10 hours ahead was lower than that of the base model, indicating the effectiveness of the proposed system in planning and deciding on marine construction.
