EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Differentiable Hydrologic Modelling Improves Streamflow Prediction with Spatially Varying Parameters

Penghao Shao, Kairong Zhang, Yiping Miao, Deping Jian, Zhaohui He, Wenhao Jia, Jianyu Fu, Qing Jia, Wenzhuo Wang, Jitao Zhang, Gang Chen, Ningpeng Dong

Water · 07.11.2026 · 18 (21) · 2773

DOI: 10.3390/w18212773 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Water, online erschienen am 07.11.2026. Verfasst von Penghao Shao und 11 weiteren Autor:innen. Die Arbeit nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, CSHE, State Key Laboratory of Water Cycle and Water Security.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Distributed hydrological models represent spatially varying forcing and flow paths but often use spatially uniform process parameters. We developed a differentiable distributed model that couples Xin’anjiang runoff generation on a 3 km grid with Muskingum routing on a 1 km river network. Spatially uniform and spatially varying parameters were trained and compared within the differentiable framework to assess the benefit of spatial parameterization. Six leave-one-year-out experiments (2017–2022) in the Luomie Basin, China, showed better performance with spatially varying parameters in every held-out year. For the pooled evaluation record, Nash–Sutcliffe efficiency increased from 0.757 to 0.884, Kling–Gupta efficiency from 0.811 to 0.868, and root-mean-square error decreased by 31%, from 86.70 to 59.86 m3 s−1. Additional component comparisons showed stronger dry-season benefits from spatial runoff parameters and stronger high-flow benefits from spatial routing parameters. Permutation analysis highlighted terrain slope for several runoff parameters and upstream area for the routing weighting factor. These results show improved streamflow prediction for withheld years in this basin and associations between the learned parameter fields and terrain, climate, vegetation, and network structure.

Förderung

  • National Natural Science Foundation of China
  • CSHE
  • State Key Laboratory of Water Cycle and Water Security
  • Science and Technology Project of China Huaneng Group
  • Yunnan Key Laboratory of Water Security

Zitieren

APA

Shao, P., Zhang, K., Miao, Y., Jian, D., He, Z., Jia, W., Fu, J., Jia, Q., Wang, W., Zhang, J., Chen, G., & Dong, N. (2026). Differentiable Hydrologic Modelling Improves Streamflow Prediction with Spatially Varying Parameters. Water, 18(21), 2773. https://doi.org/10.3390/w18212773

Vancouver

Shao P, Zhang K, Miao Y, Jian D, He Z, Jia W, et al. Differentiable Hydrologic Modelling Improves Streamflow Prediction with Spatially Varying Parameters. Water. 2026;18(21):2773. doi:10.3390/w18212773

BibTeX
@article{Shao2026,
  author = {Shao, Penghao and Zhang, Kairong and Miao, Yiping and Jian, Deping and He, Zhaohui and Jia, Wenhao and Fu, Jianyu and Jia, Qing and Wang, Wenzhuo and Zhang, Jitao and Chen, Gang and Dong, Ningpeng},
  title = {Differentiable Hydrologic Modelling Improves Streamflow Prediction with Spatially Varying Parameters},
  journal = {Water},
  year = {2026},
  volume = {18},
  number = {21},
  pages = {2773},
  doi = {10.3390/w18212773}
}
Kostenlos starten →