Abstract
The Socheongcho Ocean Research Station (S-ORS) is a fixed marine observation platform in the Yellow Sea, producing wave parameters (significant wave height, periods, and directions) using the Miros Wave and Current Radar (MWR). However, wave parameters measured with the MWR are prone to overestimation in environments with low winds and waves due to insufficient backscattered radar pulses. In this study, wave data observed using the Waverider Buoy near the S-ORS in 2015 and 2018 were compared with the MWR wave data to verify their quality. A deep neural network (DNN) was used for calibration, with the 2015 MWR data used as training data and the 2018 data as testing data. Different wave and wind parameters were input into the DNN model. Consequently, significant improvements were observed in all wave heights, periods, and directions when the DNN model was applied compared to traditional filters. The root mean squared errors (RMSE) of significant wave height, period, and direction were initially 0.20 m, 3.23 s, and 66.34°, respectively, but decreased by 20%, 80%, and 33% after calibration. Correlation coefficients increased, with wave period and direction improving from 0.09 to 0.84 and from 0.31 to 0.63, respectively. Our study significantly contributes to the literature by improving the quality of remotely sensed time-series wave data using a deep learning approach.
© 2025 American Meteorological Society. This is an Author Accepted Manuscript distributed under the terms of the default AMS reuse license. For information regarding reuse and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
