Pukyong National University

Pukyong National University Research Team Improves Long-Term Ocean Wave Height Prediction Accuracy with an AI Model Integrating Wind Speed Data

― Enhanced long-term wave forecasting expected to support marine safety and smart ocean technologies

A research team at Pukyong National University has successfully improved the accuracy of Significant Wave Height (SWH) estimation under long-term data conditions by developing an artificial intelligence (AI)-based model that integrates X-band marine radar imagery with wind speed data.

The research team, led by Na-Yoon Kang, a doctoral student in the Department of Computer and Artificial Intelligence Engineering, and Professor Won-Doo Jang of Pukyong National University, in collaboration with Professor Young-Jun Yang of the Department of Naval Architecture and Ocean Engineering at Tongmyong University, developed a deep learning model named SWH-WindNet. The team evaluated the model’s performance in estimating significant wave height using long-term observational datasets that combine X-band marine radar imagery with wind speed information.

The proposed model demonstrated superior predictive performance compared with existing models that use radar imagery alone, recording a higher correlation coefficient and a lower root mean square error (RMSE). By incorporating wind speed as an external environmental variable, the model achieved an average correlation coefficient of 0.9461 and an RMSE of 0.3887 meters, outperforming a range of state-of-the-art 2D and 3D convolutional neural network (CNN) and transformer-based models.

The research team also evaluated the model’s seasonal performance using a full year of observational data collected at Sokcho Beach. While the performance of existing models varied significantly across seasons, the proposed model maintained stable prediction accuracy throughout the year. In winter, in particular, the model reduced RMSE by approximately 11.8 percent, demonstrating that the integration of wind speed data effectively mitigates performance degradation caused by seasonal environmental changes.

This study verified the generalizability of a radar-based significant wave height estimation model using long-term marine observation data and introduced an AI model that effectively integrates wind speed information. The technology is expected to serve as a key foundation for marine safety and smart ocean applications, including route planning, vessel traffic service (VTS) systems, autonomous ships, and long-term ocean monitoring systems.

The paper presenting the findings, titled “Significant Wave Height Estimation Using X-band Radar Imagery with Wind Fusion: Performance Enhancement in Long-Term Data Environments,” was published online in June in Ocean Engineering (IF 6.3, JCR Q1), an SCIE-indexed international journal in the field of ocean engineering published by Elsevier. <Pukyong Today>