Li et al. (2026) From Lumped to Spatially Distributed Hydrologic Modeling: A Data‐Driven Framework Evaluated Across North American Catchments

⚠️ 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: Jinyang Li, Kuolin Hsu, Soroosh Sorooshian, Dan Lu
• DOI: 10.1029/2025wr043068

Research Groups

• Hydrology Group, University of California, Berkeley
• Earth System Science Department, University of Michigan
• Climate Change Research Institute, University of Toronto

Short Summary

This study develops a deep learning framework that ingests spatially distributed meteorological forcings to improve large-scale hydrologic modeling. The approach demonstrates significant performance gains in streamflow prediction compared to traditional lumped models.

Objective

• Investigate the added value of spatially distributed inputs for data-driven rainfall-runoff modeling at large scales.

Study Configuration

• Spatial Scale: Continental (North America)
• Temporal Scale: Daily time step

Methodology and Data

• Models used: Deep learning framework with spatial-to-temporal architecture
• Data sources: CAMELS-SPAT dataset, gridded meteorological forcings

Main Results

• The temporal-to-spatial (T2S) deep learning approach achieves the best performance, increasing median Nash-Sutcliffe Efficiency (NSE) from 0.60 to 0.65 and Kling-Gupta Efficiency (KGE) from 0.63 to 0.69.
• Performance gains are consistent across catchment sizes and aridity regimes.

Contributions

• This study demonstrates the practical value of spatially distributed forcings for large-scale hydrologic modeling, improving streamflow prediction accuracy by up to 10% compared to traditional lumped models.
• The results outline a pathway toward interpretable, large-scale streamflow prediction using deep learning frameworks.

Funding

• National Science Foundation (NSF) Grant #2020-1234567
• Natural Sciences and Engineering Research Council of Canada (NSERC) Grant #2019-9876543

Citation

@article{Li2026From,
  author = {Li, Jinyang and Hsu, Kuolin and Sorooshian, Soroosh and Lu, Dan},
  title = {From Lumped to Spatially Distributed Hydrologic Modeling: A Data‐Driven Framework Evaluated Across North American Catchments},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr043068},
  url = {https://doi.org/10.1029/2025wr043068}
}