Hybrid models generalize better to warmer climate conditions than process-based and purely data-driven models
4 affiliations
- 1Climate Change, Extremes and Natural Hazards in Alpine Regions Research Center CERC, Davos Dorf, Switzerland
- 2Department of Mathematics, ETH Zurich, Zurich, Switzerland
- 3WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, Switzerland
- 4Institute for Atmospheric and Climate Science, ETH Zurich, Zurich, Switzerland
Abstract
Deep-learning based rainfall-runoff models, in particular long short-term memory networks (LSTM), have been shown to outperform traditional hydrological models at various tasks, both when used as purely data-driven models and when combined with process-based models in a hybrid setting. These tasks include predictions in ungauged basins (PUB) and regions (PUR), tasks which have traditionally been challenging for conceptual hydrological models. While the spatial generalizability of deep-learning based models has received a lot of attention, it is less clear how they generalize to unseen and warmer climate conditions, i.e. how suitable these models are for hydrological climate impact studies. To address this research gap, we assess the ability of three types of models including (1) fully data-driven (LSTMs), (2) conceptual (Hydrologiska Byråns Vattenbalansavdelning (HBV)), and (3) hybrid (LSTM-HBV) models to simulate streamflow under conditions warmer than those used to train the models by running a differential split sample test. That is, we trained the models using data from the historical period 1960–1990 and evaluated them on both data of this period as well as of the warmer period 2000–2023. We find that LSTMs, while being the most accurate during the 1960–1990 period, have inferior generalizability to the warm period compared to the hybrid and conceptual models. In addition, we show that when generalizing to the warm period, hybrid models have similar accuracy as LSTMs, independently of whether the entire streamflow distribution or extreme events such as floods and droughts are considered. However, for snow-dominated catchments, all models suffer from similar reductions in accuracy when simulating streamflow under unseen climate conditions and the LSTM is the most accurate model for all periods. A detailed look at the snowmelt simulations of the hybrid and conceptual model suggests that better process-representation might be needed to accurately capture the dynamics of snow-melt and -accumulation processes, which are highly sensitive to changes in temperature. We conclude that the hybrid models effectively combine the high accuracy of LSTMs when predicting in ungauged basins with the good generalizability under changes in climate of conceptual hydrological models. This makes them a suitable choice for hydrological climate change impact assessments, particularly in ungauged basins.
Related
- Probabilistic multi-fidelity climate model parameterization for better generalization
- Improved human development index may mitigate drought risk under the changing climate and socio-economic conditions
- Passive retrofit strategies for climate-resilient high-rise social housing: evidence from sydney under current and future extreme climate conditions
- Conditional Climate–Cost Trade-Offs and Transport Break-Even Conditions for Waste Flexible-Film Recycling Under Feedstock Quality Constraints
- Climate health action in Latin America and the Caribbean: Priorities identified through a multi-stakeholder engagement process
BibTeX
@article{bohl2026hybrid,
title = {{Hybrid models generalize better to warmer climate conditions than process-based and purely data-driven models}},
author = {Jan Philipp Bohl and Raul Roger Wood and Corinna Frank and Paul C. Astagneau and Jonas Peters and Manuela I. Brunner},
journal = {Hydrology and earth system sciences},
year = {2026},
month = jul,
doi = {10.5194/hess-30-4667-2026},
url = {https://doi.org/10.5194/hess-30-4667-2026},
note = {CC BY},
}
RIS
TY - JOUR
TI - Hybrid models generalize better to warmer climate conditions than process-based and purely data-driven models
AU - Jan Philipp Bohl
AU - Raul Roger Wood
AU - Corinna Frank
AU - Paul C. Astagneau
AU - Jonas Peters
AU - Manuela I. Brunner
PY - 2026
DA - 2026/07/27
JO - Hydrology and earth system sciences
DO - 10.5194/hess-30-4667-2026
UR - https://doi.org/10.5194/hess-30-4667-2026
AB - Deep-learning based rainfall-runoff models, in particular long short-term memory networks (LSTM), have been shown to outperform traditional hydrological models at various tasks, both when used as purely data-driven models and when combined with process-based models in a hybrid setting. These tasks include predictions in ungauged basins (PUB) and regions (PUR), tasks which have traditionally been challenging for conceptual hydrological models. While the spatial generalizability of deep-learning based models has received a lot of attention, it is less clear how they generalize to unseen and warmer climate conditions, i.e. how suitable these models are for hydrological climate impact studies. To address this research gap, we assess the ability of three types of models including (1) fully data-driven (LSTMs), (2) conceptual (Hydrologiska Byråns Vattenbalansavdelning (HBV)), and (3) hybrid (LSTM-HBV) models to simulate streamflow under conditions warmer than those used to train the models by running a differential split sample test. That is, we trained the models using data from the historical period 1960–1990 and evaluated them on both data of this period as well as of the warmer period 2000–2023. We find that LSTMs, while being the most accurate during the 1960–1990 period, have inferior generalizability to the warm period compared to the hybrid and conceptual models. In addition, we show that when generalizing to the warm period, hybrid models have similar accuracy as LSTMs, independently of whether the entire streamflow distribution or extreme events such as floods and droughts are considered. However, for snow-dominated catchments, all models suffer from similar reductions in accuracy when simulating streamflow under unseen climate conditions and the LSTM is the most accurate model for all periods. A detailed look at the snowmelt simulations of the hybrid and conceptual model suggests that better process-representation might be needed to accurately capture the dynamics of snow-melt and -accumulation processes, which are highly sensitive to changes in temperature. We conclude that the hybrid models effectively combine the high accuracy of LSTMs when predicting in ungauged basins with the good generalizability under changes in climate of conceptual hydrological models. This makes them a suitable choice for hydrological climate change impact assessments, particularly in ungauged basins.
SN - 1027-5606
SN - 1607-7938
ER -