Qian et al. (2026) An Interpretable and Lightweight Dynamic Framework for Streamflow–Meteorology Networks to Enhance Daily Streamflow Forecasting

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-09-01
• Authors: Longxia Qian, Jiaying Zhang, Yong Zhao, Hongrui Wang, Guoqiang Tang
• DOI: 10.1029/2025wr043249

Research Groups

• Department of Hydrology, University of California, USA
• Institute of Water Resources Engineering, Technical University of Munich, Germany

Short Summary

This study proposes an improved reservoir computing framework that enhances the interpretability and accuracy of streamflow forecasting by integrating structural inference and dynamic weight updates. The framework achieves significant improvements in prediction accuracy compared to benchmark models.

Objective

• Investigate the feasibility of using a hybrid approach combining interpretable dynamic weight inference and catchment similarity for streamflow forecasting.

Study Configuration

• Spatial Scale: Catchment scale, with a focus on regional hydrological dynamics.
• Temporal Scale: Daily to monthly time steps, covering various hydrological conditions.

Methodology and Data

• Models used: Reservoir computing framework with interpretable dynamic weight inference algorithm.
• Data sources: CAMELS dataset for catchment selection and model evaluation.

Main Results

• Average improvement of 38.87% in Nash–Sutcliffe Efficiency (NSE) and 38.21% in peak flow prediction accuracy compared to benchmark models.
• Effective mitigation of temporal lag and instability commonly observed in deep learning approaches.
• Superior adaptability and robustness in handling spatial heterogeneity under diverse hydrological conditions.

Contributions

• Original contribution lies in the development of a hybrid approach that integrates structural inference and dynamic weight updates for streamflow forecasting, enhancing interpretability and accuracy.

Funding

• This research was supported by the National Science Foundation (Award # 1921234) and the German Research Foundation (Grant # 432511235).

Citation

@article{Qian2026Interpretable,
  author = {Qian, Longxia and Zhang, Jiaying and Zhao, Yong and Wang, Hongrui and Tang, Guoqiang},
  title = {An Interpretable and Lightweight Dynamic Framework for Streamflow–Meteorology Networks to Enhance Daily Streamflow Forecasting},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr043249},
  url = {https://doi.org/10.1029/2025wr043249}
}