HydroML2025: Machine Learning in Water, Earth, and Climate Sciences
From the ancient water clocks of 4000 BC Egypt and Mesopotamia to the rise of personal computers, smartphones, and autonomous vehicles, automation has fundamentally impacted almost all aspects of human life. Today, Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing research and education across many scientific and engineering disciplines, among which are hydrology, hydrometeorology, climatology, and environmental engineering. Innovative methods and technologies juxtaposed with the quantum computing of the near future hold unprecedented potential for substantially advancing our understanding of subsurface, land-surface, boundary-layer, and atmospheric processes. This is key to enhancing predictions of climate change and extreme events—such as floods, droughts, fires, and hurricanes—and their impacts on infrastructure, human life, and ecosystem health, resilience, and sustainability. However, AI and ML also present many new challenges for Water, Earth, and Climate scientists. These include (but are not limited to):
How to best adapt new methods and technologies to efficiently handle and analyze high-dimensional data sets from Earth-orbiting satellites, in-situ monitoring networks, and high-resolution physics-based models.
How to leverage and synthesize the capabilities of data-driven and physics-informed ML, large language models (LLMs), and generative AI to enhance process understanding of water and energy cycles across spatiotemporal scales.
How to train, constrain, test, and evaluate ML and differentiable models to maximize their physical realism and robustness while accurately portraying higher-order data patterns and signatures.
How to balance deep learning model complexity, accuracy, and interpretability.
How to embed AI technology and ML models within a probabilistic framework for real-time forecasting.
How to quantify epistemic and prediction uncertainty of ML models and yield well-calibrated predictive distributions.
The HydroML 2025 symposium explores the use of AI and ML methods in Water, Earth and Climate research. This 3-day meeting aims to bring together Earth, environmental, and climate scientists and engineers with experts in statistics, optimization, and nonlinear control to explore the application and use of AI and ML methods in water and energy cycles. Specific areas of interest include uncertainty quantification, probabilistic (ensemble) forecasting, real-time prediction and decision making, and ML model malfunctioning and misspecification.
The HydroML 2025 symposium will explore how AI/ML concepts can be used to enhance the predictive understanding of complex systems in hydrological and geological sciences. The overarching goal is to discuss process-based scientific principles that can help integrate AI/ML with earth system science. In essence, the symposium seeks to stimulate discussions that will help develop physically guided AI/ML approaches which are explainable, interpretable, and improve the mechanistic understanding of earth system science. It will foster collaborations among researchers who are both new to the field and already involved, thereby strengthening ties within the community of AI/ML researchers.
This symposium will be the fourth in a series, the first being held at Penn State University in May 2022.
Research Presentations
Breakout Discussion Sessions
Tutorials and Hackathons
Community-Building Activities