A Deep Learning-Based Bias Correction Method for Predicting Ocean Surface Waves in the Northwest Pacific Ocean

Abstract

<p>Ocean waves, especially extreme waves, are vital for air-sea interaction and shipping. However, current wave models still have significant biases, especially under extreme wind conditions. Based on a numerical wave model and a deep learning model, we accurately predict the significant wave height (SWH) of the Northwest Pacific Ocean. For each day in 2017-2021, we conducted a 3-day hindcast experiment using WAVEWATCH3 (WW3) to obtain the SWH forecasts at lead times of 24, 48, and 72hr, forced by GFS real-time forecast surface winds. The deep learning-based bias correction method is BU-Net by adding batch normalization layers to a U-Net, which could improve the accuracy. Due to the use of BU-Net, the mean Root Mean Squared Errors (RMSEs) of the SWH forecast from WW3 at lead times of 24, 48, and 72hr are reduced from 0.35m to 0.21m, 0.39m to 0.24m, and 0.43m to 0.30m, corresponding to drop percentages of 40%, 38%, and 30%, respectively. During typhoon passages, the drop percentages of RMSEs reach 45%, 42%, and 35% for three lead times. Therefore, combining numerical models and deep learning algorithms is very promising in ocean wave forecasting.</p>