Mei et al. (2026) Diagnosing Cryospheric Runoff Dynamics: A Distributed Differentiable Hydrological Model With Global Transfer Learning

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

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-08-01
• Authors: Jun Mei, Bin Yong, Gerald Corzo, Yang Hong, Ruochen Sun, Xin Tao, Jiahu Wang, Qingyun Duan
• DOI: 10.1029/2026wr044148

Research Groups

Not specified in the provided text.

Short Summary

The study introduces dCREST, a differentiable physics-informed hydrological model, to improve runoff predictions in alpine regions by optimizing soil structure and utilizing global transfer learning to overcome local data scarcity.

Objective

• To investigate the hydrological role of frozen-ground processes and evaluate whether global pre-training can enhance runoff simulation in data-scarce alpine basins.

Study Configuration

• Spatial Scale: Headwater Area of the Yellow River (HAYR) and globally distributed catchments.
• Temporal Scale: Seasonal runoff partitioning and long-term hydroclimatic responses.

Methodology and Data

• Models used: dCREST (differentiable Coupled Routing and Excess STorage model).
• Data sources: Local observations from the HAYR and global catchment datasets for pre-training.

Main Results

• Conventional single-layer soil structures are insufficient for capturing permafrost-controlled hydrological dynamics.
• An explicit dual-layer soil structure combined with dynamic cryospheric parameterizations significantly improves the simulation of seasonal runoff and long-term responses.
• Global pre-training (transfer learning) increased the average Nash–Sutcliffe efficiency (NSE) from 0.35 (operational system) to 0.63.

Contributions

• Developed a distributed differentiable framework that allows for the simultaneous diagnosis of model structure, parameter dynamics, and spatial transferability.
• Demonstrated that physically constrained transfer learning can effectively mitigate the impact of limited local observations in alpine hydrological modeling.

Funding

Not specified in the provided text.

Citation

@article{Mei2026Diagnosing,
  author = {Mei, Jun and Yong, Bin and Corzo, Gerald and Hong, Yang and Sun, Ruochen and Tao, Xin and Wang, Jiahu and Duan, Qingyun},
  title = {Diagnosing Cryospheric Runoff Dynamics: A Distributed Differentiable Hydrological Model With Global Transfer Learning},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2026wr044148},
  url = {https://doi.org/10.1029/2026wr044148}
}