What the study found
River stage dynamics, meaning the period and amplitude of water-level fluctuations, were identified as the primary drivers of dissolved oxygen supply to the aquifer system. The study also found that some factors had important but more localized effects, while others had negligible influence.
Why the authors say this matters
The authors conclude that combining deep learning with global sensitivity analysis can help efficiently evaluate complex environmental systems and support model simplification and diagnosis. The study suggests this approach is useful for assessing the controls on dissolved oxygen in riparian aquifers, where dissolved oxygen helps regulate redox conditions and related subsurface processes.
What the researchers tested
The researchers used a global sensitivity analysis framework that combined Bayesian network-based and variance-based methods. To reduce the computational cost of analyzing a complex numerical model, they built surrogate models with deep learning, specifically multi-layer perceptrons and convolutional neural networks, and applied the framework to a high-resolution model of riparian dissolved oxygen transport.
What worked and what didn't
The analysis showed that river stage dynamics were the dominant controls on dissolved oxygen supply. Hydraulic conductivity, riverine dissolved oxygen concentration, and the maximum dissolved oxygen reaction rate also mattered, but their effects were localized and depended on transport pathways such as river water infiltration, entrapped air dissolution, and diffusion through the unsaturated zone. In contrast, porosity, longitudinal dispersion, and van Genuchten soil parameters had negligible influence.
What to keep in mind
The abstract describes a model-based study of riparian aquifers, so the findings are tied to the tested framework and the high-resolution model used. The abstract does not describe additional limitations beyond the challenge of uncertainty and computational demand in complex coupled flow and transport models.
Key points
- River stage dynamics were identified as the main drivers of dissolved oxygen supply in the aquifer system.
- Hydraulic conductivity, riverine dissolved oxygen concentration, and maximum dissolved oxygen reaction rate had localized effects.
- Porosity, longitudinal dispersion, and van Genuchten soil parameters showed negligible influence.
- The study combined Bayesian network-based and variance-based global sensitivity analysis methods.
- Deep learning surrogate models were used to make the sensitivity analysis computationally feasible.
Disclosure
- Research title:
- River stage dynamics drive dissolved oxygen in riparian aquifers
- Authors:
- Heng Dai, Yijie Yang, Fangqiang Zhang, Alberto Guadagnini, Jing Yang, Xiaochuang Bu, Lunche Wang, Songhu Yuan, Ming Ye
- Institutions:
- China University of Geosciences, Politecnico di Milano, Chang'an University, Hubei University, Florida State University
- Publication date:
- 2026-01-30
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.