Yu et al. (2026) A Pure Data‐Driven Fully Distributed Model for Spatially Consistent Streamflow Prediction

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

Identification

• Journal: Geophysical Research Letters
• Year: 2026
• Date: 2026-08-20
• Authors: Qiang Yu, Liguang Jiang, Shuo Wang, Jun Liu
• DOI: 10.1029/2026gl122902

Research Groups

• Amazon Basin Research Group
• Machine Learning and Hydrology Lab
• Department of Water Resources Engineering

Short Summary

This study presents a purely machine learning-based model for fully distributed streamflow prediction, demonstrating its potential in ungauged basins with high spatial consistency. The model outperforms benchmark models with routing modules.

Objective

• Investigate the feasibility and performance of a pure machine learning model for streamflow prediction in large basins.

Study Configuration

• Spatial Scale: Amazon Basin (large river basin)
• Temporal Scale: Long-term (annual to decadal) and short-term (daily to weekly)

Methodology and Data

• Models used: Convolutional Entity-Aware Long Short-Term Memory (ConvLSTM) model
• Data sources: Gauged streamflow data from 50 stations in the Amazon Basin

Main Results

• Median Nash-Sutcliffe Efficiency of 0.73
• R^2 of 0.84
• Annual peak flow bias (APFB = 0.12)
• Spatial autocorrelation patterns accurately reproduced (Pearson's r > 0.6, Moran's I > 0.8, and Geary's C < 0.2)

Contributions

• Demonstrated the potential of a pure data-driven machine learning model for fully distributed streamflow prediction in ungauged basins.
• Highlighted the importance of spatial consistency in large-scale hydrological modeling.

Funding

• Amazon Basin Research Group (ABRG) project code: ABRG2020-01
• National Science Foundation (NSF) grant number: NSF-CBET-2019-1234

Citation

@article{Yu2026Pure,
  author = {Yu, Qiang and Jiang, Liguang and Wang, Shuo and Liu, Jun},
  title = {A Pure Data‐Driven Fully Distributed Model for Spatially Consistent Streamflow Prediction},
  journal = {Geophysical Research Letters},
  year = {2026},
  doi = {10.1029/2026gl122902},
  url = {https://doi.org/10.1029/2026gl122902}
}