Data-driven modelling continues to gain momentum within the groundwater research community, reflecting strong interest in machine learning, AI and innovative approaches integrating data-driven techniques with physical process understanding. Data-driven approaches, including time-series models, statistical methods, machine and deep learning techniques and emulators, are transforming how we study, manage and forecast groundwater systems. By learning from observations and other sources, these methods can complement or even substitute detailed process-based models. Advancements in physics-informed methods, spatiotemporal deep learning, probabilistic machine learning, foundation-model approaches and generative AI are rapidly expanding possibilities for groundwater science.
We welcome novel methodological developments and practical applications addressing real-world groundwater management problems along the following topics:
Methodological Advances
-Modelling techniques and data-driven methods for predicting groundwater quantity/quality in space or time, such as machine learning, deep learning, statistical methods and time series models
-Hybrid approaches combining machine learning with conventional physical based models, including physics-informed neural networks
-Machine learning-based emulation of physical based models for enhanced modelling and data assimilation
-Applications of generative AI and large language models for synthetic data generation, knowledge extraction and model development for groundwater research and management workflows
Uncertainty and Interpretability
-Uncertainty quantification and sensitivity analysis using machine learning techniques, including probabilistic forecasting, quantile regression and deep ensembles
-Explainable and interpretable machine learning approaches to improve hydrogeological system understanding
Data Handling and Transferability
-Approaches for analysing big data and complex datasets from groundwater systems
- Solutions for data scarcity, multi-source/scale/fidelity data integration, non-stationary time series and irregular time steps
-Transferability and regionalization methods for spatial transfer to ungauged sites and temporal extrapolation to unseen conditions
In session you will engage with researchers working at the intersection of data science, process-based modelling and groundwater quantity/quality. Join us to share your research and advance this dynamic field through knowledge exchange and discussion.