Lin et al. (2026) Bridging global offshore waves and nearshore coastal conditions with deep learning

Identification

• Journal: Communications Earth & Environment
• Year: 2026
• Date: 2026-08-29
• Authors: Xi Lin, Tianping Xu, Wenfang Lu, Guangyu Yang, Zhan Hu
• DOI: 10.1038/s43247-026-04004-x

Research Groups

• School of Marine Sciences, Sun Yat-sen University and Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), Zhuhai, China.
• State Key Laboratory of Environmental Adaptability for Industrial Products, Sun Yat-sen University, Zhuhai, China.
• Guangdong Provincial Key Laboratory of Marine Resources and Coastal Engineering, Guangzhou, China.
• Guangdong Pearl River Estuary National Integrated Monitoring Station for Ecological Quality of Marine Ecosystem, Zhuhai Guangdong, 519070, China.

Short Summary

This study presents a hybrid framework (SWAN-TC) that integrates numerical emulation with observation-driven refinement to enable rapid, high-resolution nearshore wave simulation. The framework combines a deep learning surrogate model (SWAN-T) with an observation-driven residual correction sub-model (SWAN-C), providing a composite solution with both accuracy and efficiency.

Objective

• To develop a computationally efficient model that can transfer the dominant cross-shore transformation between compatible coastal settings while accounting for systematic local discrepancies relative to observations.
• To bridge global offshore waves and nearshore coastal conditions using deep learning.

Study Configuration

• Spatial Scale: The study focused on the 1–10 km coastal band, with a spatial resolution of approximately 0.25°.
• Temporal Scale: The study used hourly forcing records from ERA5 reanalysis data, with a temporal resolution of 3 hours.

Methodology and Data

• Models used: SWAN (version 41.45), a third-generation phase-averaged spectral wave model; Transformer-based surrogate model (SWAN-T); CatBoost-based observation-driven residual correction sub-model (SWAN-C).
• Data sources: ERA5 reanalysis data, buoy observations, GEBCO bathymetry, and coastal geometry.

Main Results

• The SWAN-TC framework demonstrated high accuracy in predicting nearshore wave conditions, with a coefficient of determination above 0.90 for significant wave height (SWH) and mean wave period (Tm02).
• The framework showed robust capability in correcting systematic residuals between model predictions and buoy observations.
• The hybrid framework improved nearshore wave predictions compared to both the baseline surrogate and the WW3 reanalysis product.

Contributions

• This study presents a novel approach for bridging global offshore waves and nearshore coastal conditions using deep learning.
• The SWAN-TC framework provides an efficient and scalable solution for simulating high-resolution nearshore wave conditions, which can support improved coastal hazard management.

Funding

• This research was supported by the Guangdong Provincial Key Laboratory of Marine Resources and Coastal Engineering (Grant No. 2020B1212030001).
• The study also received funding from the National Natural Science Foundation of China (Grant Nos. 41876005 and 42076201).

Citation

@article{Lin2026Bridging,
  author = {Lin, Xi and Xu, Tianping and Lu, Wenfang and Yang, Guangyu and Hu, Zhan},
  title = {Bridging global offshore waves and nearshore coastal conditions with deep learning},
  journal = {Communications Earth & Environment},
  year = {2026},
  doi = {10.1038/s43247-026-04004-x},
  url = {https://doi.org/10.1038/s43247-026-04004-x}
}