Wang et al. (2026) RemoteWaterNet: a lightweight and efficient algorithm for remote river surface velocimetry

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-01-13
• Authors: Xiaochao Wang, Yu Xiao, Chongli Di
• DOI: 10.1016/j.jhydrol.2026.134940

Research Groups

• School of Mathematical Sciences, Tiangong University, Tianjin, PR China
• Institute of Surface-Earth System Science, School of Earth System Science, Tianjin University, Tianjin, PR China

Short Summary

This study introduces RemoteWaterNet, a lightweight deep learning framework for robust and efficient remote river surface velocimetry. It achieves a 26.33% improvement in accuracy and a 92.38% reduction in model parameters compared to existing methods, making it highly suitable for real-time environmental monitoring.

Objective

• To develop a robust and efficient lightweight deep learning framework (RemoteWaterNet) for remote river surface velocity measurement, addressing the limitations of existing non-contact imaging methods, particularly their lack of robustness and stability under diverse environmental conditions and varying flow characteristics.

Study Configuration

• Spatial Scale: River systems, focusing on local river sections for surface velocity measurement.
• Temporal Scale: Real-time and continuous monitoring of river surface velocities.

Methodology and Data

• Models used: RemoteWaterNet, a deep learning framework integrating simplified image preprocessing with a pre-trained optical flow model (SEA-RAFT) for iterative refinement and unit conversion.
• Data sources: Multiple datasets for extensive training and fine-tuning, and eight field datasets for experimental validation. These are image-based datasets.

Main Results

• RemoteWaterNet demonstrates superior generalization capabilities across diverse environmental conditions.
• The proposed method improves accuracy by 26.33% compared to existing surface velocimetry techniques.
• It significantly reduces model parameters by 92.38%, enhancing efficiency.
• The framework is highly suitable for real-time environmental monitoring applications due to its lightweight and efficient nature.

Contributions

• Proposes RemoteWaterNet, a novel lightweight and efficient deep learning framework for remote river surface velocimetry.
• Significantly advances the application of deep learning-based optical flow models in hydrological measurements.
• Offers valuable new insights for the practical monitoring and management of river systems.
• Demonstrates enhanced robustness and generalization under diverse environmental conditions compared to existing methods.
• Achieves high computational efficiency, making it suitable for real-time data acquisition and processing.

Funding

• Not specified in the provided text.

Citation

@article{Wang2026RemoteWaterNet,
  author = {Wang, Xiaochao and Xiao, Yu and Di, Chongli},
  title = {RemoteWaterNet: a lightweight and efficient algorithm for remote river surface velocimetry},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.134940},
  url = {https://doi.org/10.1016/j.jhydrol.2026.134940}
}