We are adding the link to the preprint which is now accepted in ICML 2024 workshop and will be presented as a talk and a poster this week in Vienna: Link to the paper .

This study, led jointly by Surya Dheeshjith and Adam Subel, showcases Machine Learning Ocean Emulators accurately predicting global ocean surface conditions over 5-8 years under different CO₂ forcing scenarios. While the models struggle to generalize, fine-tuning with small amounts of additional data from warner climates can significantly improve model performance. The study also shows the robustness of the emulators to noise in the atmospheric forcing. Shubham Gupta, Alistair Adcroft, Carlos Fernandez-Granda, Julius Busecke, and Laure Zanna all contributed to the work.