GSDNet: A Deep Learning Model for Downscaling the Significant Wave Height Based on NAFNet

Accurate wave height information is essential for marine engineering and safety. This study proposes GSDNet, a deep learning model based on the Nonlinear Activation Free Network (NAFNet), for downscaling significant wave height (SWH). The model is designed to generate high-resolution wave fields from low-resolution inputs. Experiments demonstrate that GSDNet outperforms traditional interpolation methods and other deep learning architectures in terms of accuracy and computational efficiency. The model successfully reconstructs fine-scale wave features and shows robust performance across different sea states, providing a powerful tool for obtaining high-resolution wave data.

Journal of Sea Research