Zhu et al. (2025) Attention enhanced ResNet for ocean surface wind speed retrieval using CYGNSS observables

Identification

• Journal: Advances in Space Research
• Year: 2025
• Date: 2025-11-11
• Authors: Yongchao Zhu, Qianqian Lu, Maorong Ge, Xiaochuan Qu, Tingye Tao, Kegen Yu, Shuiping Li
• DOI: 10.1016/j.asr.2025.11.023

Research Groups

• School of Civil Engineering, Hefei University of Technology, Hefei, China
• Shanghai Astronomical Observatory, Chinese Academy of Sciences, Shanghai, China
• School of Environment Science and Spatial Informatics, China University of Mining and Technology, Xuzhou, China

Short Summary

This study proposes an Attention-enhanced Residual Network (Att-ResNet) to improve ocean surface wind speed retrieval using Cyclone Global Navigation Satellite System (CYGNSS) bistatic radar data. The Att-ResNet model achieved high accuracy, demonstrating root mean square errors of approximately 1.38 m/s when validated against ERA5 and CCMP wind products.

Objective

• To address the challenge of establishing robust multi-parameter retrieval models for ocean surface wind speed using Global Navigation Satellite System Reflectometry (GNSS-R) observables by developing and evaluating an Attention-enhanced Residual Network (Att-ResNet) leveraging CYGNSS data.

Study Configuration

• Spatial Scale: Global ocean surface, with a focus on high-resolution spatiotemporal monitoring.
• Temporal Scale: Continuous monitoring capability for ocean surface wind speed.

Methodology and Data

• Models used: Attention-enhanced Residual Network (Att-ResNet), with ResNet and AlexNet as backbone architectures for comparison.
• Data sources:
  • CYGNSS (Cyclone Global Navigation Satellite System) bistatic radar data (Delay-Doppler Maps (DDMs), normalized bistatic radar cross-section (NBRCS), incidence angle).
  • ERA5 (European Centre for Medium-Range Weather Forecasts Reanalysis 5) wind products (for validation).
  • CCMP (Cross-Calibrated Multi-Platform) wind products (for validation).

Main Results

• Att-ResNet-retrieved wind speeds exhibited strong spatiotemporal consistency with ERA5 and CCMP data.
• Quantitative evaluations against ERA5 showed a root mean square error (RMSE) of 1.379 m/s, a bias of -0.069 m/s, and an unbiased RMSE (ubRMSE) of 1.377 m/s.
• Quantitative evaluations against CCMP showed an RMSE of 1.390 m/s, a bias of -0.014 m/s, and an ubRMSE of 1.390 m/s.
• The Att-ResNet architecture significantly enhances spaceborne GNSS-R wind retrieval accuracy through its attention-driven feature selection and residual learning mechanisms.

Contributions

• Proposes a novel Attention-enhanced Residual Network (Att-ResNet) for ocean surface wind speed retrieval using GNSS-R observables, specifically CYGNSS data.
• Demonstrates significant improvement in spaceborne GNSS-R wind retrieval accuracy, achieving low RMSE and bias compared to benchmark wind products.
• Establishes an artificial intelligence-driven framework for high-resolution spatiotemporal ocean surface wind monitoring, showcasing the transformative potential of deep learning in advancing GNSS-R applications.

Funding

• Not specified in the provided text.

Citation

@article{Zhu2025Attention,
  author = {Zhu, Yongchao and Lu, Qianqian and Ge, Maorong and Qu, Xiaochuan and Tao, Tingye and Yu, Kegen and Li, Shuiping},
  title = {Attention enhanced ResNet for ocean surface wind speed retrieval using CYGNSS observables},
  journal = {Advances in Space Research},
  year = {2025},
  doi = {10.1016/j.asr.2025.11.023},
  url = {https://doi.org/10.1016/j.asr.2025.11.023}
}