And, there's a challenge in how ML models deal with 'missing data', especially when it's land in an ocean model -- these points have impact on the dynamics even though they are not themselves modeled. Maybe new ML techniques built specifically for geosciences are needed...?

This highlights the importance of robust assessment of ML models to catch systematic bias. ML models are optimised to give good average RMSE, but the way we use these models is so much broader, so our assessment needs to be broad too. There's an inherent optimism-use dichotomy in ML models.

Hello, add me please!!

In this last session, we are honored to have Julie Deshayes as solicited speaker.
@rachelfurner.bsky.social @aida-alvera.bsky.social @redouanelg.bsky.social

Please can you add me!

Hi, I've just joined bluesky, and would love to be added to this list. Thanks for curating it!

And I'm now here too!!