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

Incorporating scalable physics and uncertainty into AI weather prediction

  • Register

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

    • Date
    21 January 2026
    Timeframe
    17:00 - 18:00 CET
    Duration
    60 minutes
    • Days
      Hours
      Min
      Sec

    AI Weather Prediction models have sparked a step change in prediction accuracy for global medium range weather forecasts and have been deployed operationally in a very short period of time. However, deeper analyses of AI NWP predictions have identified systematic artifacts and instabilities in the predictions, including violations of physics conservation and underdispersive ensembles. The MILES group at the US NSF National Center for Atmospheric Research has developed CREDIT, an open foundational research platform for AI Earth system prediction, with the goal of finding ways to modify the full AI weather prediction pipeline to adjust these discrepancies. This talk will cover our recent advances in physics constraints and scale-dependent internal perturbations to produce both improved predictions and understanding.

    Session Objectives:

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

    • Understand the fundamentals of AI weather prediction and the underlying challenges.
    • See how physics constraints can be applied to AI weather prediction models to produce improved accuracy and better representations of the underlying atmospheric state.
    • Learn how calibrated ensembles can be generated at low cost from a pre-trained AI weather model.

    Recommended Mastery Level/Prerequisites:

    Participants should have a basic understanding of meteorology and some familiarity with AI/ML. The session is intended for attendees who want to understand the fundamentals of AI-based weather prediction, the associated challenges, and how physics constraints and calibrated ensembles can be incorporated into modern AI weather models.

    Share this session with your network
    • Session starts in
      Days
      Hours
      Min
      Sec
    Add to Calendar
    Discover more from programme stream
    Discovery - AI for Earth and Sustainability Science
    Incorporating scalable physics and uncertainty into AI weather prediction

    Incorporating scalable physics and uncertainty into AI weather prediction

    21 January 2026 - 17:00 to 18:00

    Related stories

    Are you sure you want to remove this speaker?