Search
Buy your pass
OnlineDiscovery - AI for Earth and Sustainability Science Badge available

When rivers meet neural nets: The promise and limits of Machine Learning in hydrology

  • Register

    Watch

    Watch and get your discovery badge

    * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

    • Date
    20 October 2025
    Timeframe
    16:00 - 17:00 CEST Geneva
    Duration
    60 minutes

    Machine learning has rapidly transformed hydrology, offering new ways to simulate, predict, and understand the movement of water through landscapes. From data-driven rainfall–runoff models to hybrid architectures that couple physical and neural representations, these approaches promise to overcome long-standing limitations of traditional process-based models. Yet, while neural networks can rival or even surpass established hydrological models in predictive performance and it is often unclear how well they generalize beyond the conditions on which they were trained. This talk presents concrete use cases where machine learning provides genuine value for classical hydrological problems, and highlights where it currently reaches its limits. Examples include national-scale flood and drought forecasting systems that explicitly account for predictive uncertainty. I will show how deep learning can capture hydrologically meaningful patterns from large-sample datasets, yet struggles with extrapolating to extreme events and unseen regimes. By reflecting on both the strengths and the limitations of current approaches, the talk aims to open a dialogue on how physics-based understanding, uncertainty quantification, and transparent evaluation can help make AI a more reliable and insightful tool for hydrological research and practice. 

     

    Learning Objectives:

    By the end of the session, participants will be able to:

    • Understand the principles and motivations behind data-driven and hybrid (physics–ML) approaches in hydrology, and how they compare to traditional process-based models.
    • Analyze the strengths and limitations of machine learning in hydrological prediction, including its performance in ungauged basins, extrapolation to extremes, and integration into national flood and drought forecasting systems that account for uncertainty.
    • Evaluate how physical constraints, uncertainty quantification, and transparent model interpretation can enhance the reliability and sustainability of AI applications in Earth system science. 

    Recommended Mastery Level / Prerequisites: 
    Intermediate to Advanced – suitable for graduate students, early career researchers, and professionals in climate science or environmental data science. 

    Discovery badge available
    Showcase it on your LinkedIn profile or your professional portfolio
    Verified badge is available for CHF 10
    Learn more
    Share this session with your network
    Discover more from programme stream
    Discovery - AI for Earth and Sustainability Science
    Risks and solutions when using non-stationary training data in earth...

    Risks and solutions when using non-stationary training data in earth...

    8 December 2025 - 17:00 to 18:00

    Related stories

    AI and the future of inclusive mental health care

    During the 2025 AI for Good Global Summit, the International Health Charity Association (IHCA) organized a workshop on empowering equal...
    Global Summit | Health

    Launching the ITU AI and Space Computing Challenge: Innovating for...

    The International Telecommunication Union (ITU), in partnership with Zhejiang Lab and leading global collaborators, has officially launched the 2026 AI...
    Global Summit

    Building trust through responsible AI governance

    At the AI for Good Global Summit, the World Internet Conference (WIC) convened a workshop on developing responsible AI, moderated...
    Global Summit

    Are you sure you want to remove this speaker?