Hybrid models generalize better to warmer climate conditions than process-based and purely data-driven models

Bohl, Jan P.; Wood, Raul R.; Frank, Corinna; Astagneau, Paul C.; Peters, Jonas; Brunner, Manuela I.

doi:10.5194/hess-30-4667-2026

Articles | Volume 30, issue 14

https://doi.org/10.5194/hess-30-4667-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-30-4667-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 30, issue 14

Research article

|

27 Jul 2026

Research article |

| 27 Jul 2026

Hybrid models generalize better to warmer climate conditions than process-based and purely data-driven models

Jan P. Bohl, Raul R. Wood, Corinna Frank, Paul C. Astagneau, Jonas Peters, and Manuela I. Brunner

Data sets

Generalizability of data-driven hydrological models in the European Alps J. P. Bohl et al. http://www.hydroshare.org/resource/0e708f4a9f8440c880408221d3fa86b5

E-OBS Daily Gridded Meteorological Data for Europe from 1950 to Present Derived from in-Situ Observations Copernicus Climate Change Service, Climate Data Store https://doi.org/10.24381/CDS.151D3EC6

SPASS - new gridded climatological snow datasets for Switzerland C. Marty et al. https://doi.org/10.16904/envidat.580

SNOWGRID Klima v2.1 GeoSphere Austria https://doi.org/10.60669/fsxx-6977

Model code and software

Caravan - A global community dataset for large-sample hydrology F. Kratzert https://github.com/kratzert/Caravan/

Analyzing the generalization capabilities of hybrid hydrological models for extrapolation to extreme events E. Acuna Espinoza https://doi.org/10.5281/zenodo.14191623

differentiable parameter learning (dPL) + HBV hydrologic model D. Feng et al. https://doi.org/10.5281/zenodo.7943626

Short summary

To assess climate impacts on streamflow, we need models that can predict streamflow under future conditions. This study compares three model types: data-driven (LSTM - long short-term memory), conceptual (HBV - Hydrologiska Byråns Vattenbalansavdelning), and hybrid (LSTM-HBV). LSTMs perform best overall, but HBV and hybrid models generalize better to warmer climates. Hybrid models are a promising tool for climate impact assessments, combining LSTMs accuracy with better generalizability of traditional models. In snowy regions, all models struggle to generalize.