EGU 2026
Published:
I attended my second EGU last week and thoroughly enjoyed it. I co-convened exciting session along with Jack Atkinson, Milan Klower and Alex Connolley on Advancing Earth System Models using Machine Learning, which featured talks on learning data-driven parameterisations, online training, and the efficiency of AI models. It was busy, so feels like a success!
As well as that, I presented a poster on epistemic and aleatoric uncertainties in weather and climate models which summarised my recent paper. I had a lot of interesting discussions about these two types of uncertainties, and how and why we should consider them separately.
I noticed this year there were a LOT of AI talks, and each AI-related session was super busy. One of my highlights was also the uncertainty quantification and calibration session, where I noticed less focus on AI (although of course, Gaussian processes are still popular!). My hope is that in the future, we can bring the ML and UQ communities closer together and learn from each other, to really improve climate modelling and reduce uncertainties. One of my lowlights was scoring only 2/10 a quiz identifying whether gravity waves are real satellite observations or AI-generated (poster by Adam Hayes)!
Overall, I greatly enjoyed discussing and learning from others. I had a great week and hope to attend again in the future!