Announce: Atmosphere Working Group

Atmosphere Working Group
, , ,
50

Atmosphere WG meeting 23.09.2026

For a calendar invite, simply click on the relevant calendar link below

Agenda:

  1. Admin: Volunteer for taking today’s meeting minutes to post here on the Forum.

  2. ACCESS Community updates: (5mins)
    Placeholder for any updates from the atmosphere community.

  3. Hybrid atmospheric model development - can ACCESS-NRI play a role? Steve Sherwood, UNSW (10+5mins)
    A few hybrid modelling approaches are being pursued for improving weather and climate forecasting, where “hybrid” in general just means combining machine-learning and traditional modelling approaches somehow. A particular instance is hybrid climate models that replace sub-grid physics with ML surrogates, thereby attempting to keep known physical constraints while exploiting ML to fill gaps. We will review this strategy and how (or why) ACCESS-NRI might contribute to it by generating training data for surrogates and/or developing a sovereign hybrid modelling capability.

  4. ML-Based Weather modelling at the Bureau of Meteorology Catherine de Burgh-Day, BoM (20+5mins)
    Machine learning has emerged as a powerful tool for weather prediction, with major meteorological centres and research organisations around the world investing in ML-based forecasting systems - both pure ML models, and hybrid physics-ML approaches. This presentation provides an overview of the evolving international landscape, highlighting scientific potential, and key developments including hybrid spectral nudging, foundation models, Direct from Obs Prediction (DOP), ensemble prediction, regional and multi-domain models, and progress in spatio-temporal downscaling.
    The presentation will also outline the Bureau’s current activities and future plans in ML-based global and regional weather modelling, including evaluation of AIFS variants, operational prototyping of AIFS, and progress in developing a combined global-regional ML weather models for Australia.

  5. Discussion Andrew Brown (10mins)

    • How can we develop our community to better share ML techniques and expertise?
      e.g., invited talks at Atmosphere and ML WG meeting

  6. Ask the community?? :thinking: Sonya Fiddes (5mins)
    Do you have a research question that you’ve exhausted ideas to find an answer to?
    Here’s your chance to ask our Atmosphere community for their ideas!

  7. Compute/storage resources: Heidi Nettelbeck (2mins)
    Any experiment proposals for Q3/2026 (2 MSU allocation)?
    → See here for how to apply for Atmosphere WG Resources.

  8. Next Atmosphere WG meeting: October 2026