Luo et al. (2025) Identification and spatiotemporal analysis of braided rivers in the Yarlung Tsangpo basin using an enhanced U-Net approach

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-12-17
• Authors: Xiangyang Luo, Ying Lu, B. Zhang, Yadan Yang, Jiaxin Li, Wanying Fu, Xiu-qin Bu, Cong Li
• DOI: 10.1016/j.jhydrol.2025.134796

Research Groups

• Institute of International Rivers and Eco-Security, Yunnan University, Kunming, China.
• Key Laboratory of International Rivers and Transboundary Eco-Security of Yunnan Province, Kunming, China.
• PowerChina Beijing Engineering Corporation Limited, Beijing, China.

Short Summary

This study develops an enhanced deep learning model, MSU-Net, to accurately identify and map complex braided river systems in the Yarlung Tsangpo River Basin. Using a fusion of Sentinel-1 and Sentinel-2 satellite data from 2018 to 2023, the research quantifies the spatiotemporal dynamics of these channels and their relationship with climatic drivers.

Objective

• To overcome the limitations of traditional field surveys by developing a high-precision, multi-scale dual attention U-Net model (MSU-Net) for the automated identification of braided river networks and the analysis of their morphological variability.

Study Configuration

• Spatial Scale: Yarlung Tsangpo River Basin, China (with detailed analysis of the Shannan section).
• Temporal Scale: Monthly intervals from 2018 to 2023.

Methodology and Data

• Models used: MSU-Net (an improved U-Net architecture incorporating a multi-scale dual attention gate module that integrates spatial and channel attention mechanisms).
• Data sources: Sentinel-1 (Synthetic Aperture Radar/SAR) and Sentinel-2 (Optical) satellite imagery; training involved manually corrected labels and data augmentation strategies.

Main Results

• Developed a high-efficiency identification method for braided rivers that outperforms standard models in complex environments by better capturing multi-scale water features.
• Produced a comprehensive monthly remote sensing dataset of the Yarlung Tsangpo River Basin covering a six-year period (2018–2023).
• Quantified the spatiotemporal fluctuations of surface water area in the Shannan section, establishing a clear link between braided channel variability and climatic factors.
• Demonstrated that the integration of spatial and channel attention mechanisms significantly reduces errors in water body extraction within topographically complex regions.

Contributions

• Provides a novel deep learning framework (MSU-Net) specifically optimized for the unique, multi-channel morphology of braided river systems.
• Offers a scalable and cost-effective alternative to field-based or aerial survey methods for monitoring large-scale river dynamics.
• Advances the understanding of how braided river systems in high-altitude basins respond to climate change, providing a scientific basis for transboundary water resource management.

Funding

• Not specified in the provided text.

Citation

@article{Luo2025Identification,
  author = {Luo, Xiangyang and Lu, Ying and Zhang, B. and Yang, Yadan and Li, Jiaxin and Fu, Wanying and Bu, Xiu-qin and Li, Cong},
  title = {Identification and spatiotemporal analysis of braided rivers in the Yarlung Tsangpo basin using an enhanced U-Net approach},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134796},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134796}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)