Numerical climate models generally become more realistic as their spatial resolution increases, but it remains unclear whether this improvement primarily results from the models themselves or from the richer variability available in high-resolution data. AI-based climate models, in contrast, often perform better and remain more stable at coarser resolutions, while recent work suggests that downscaling can successfully recover realistic small-scale features. This raises an exciting question: can AI models learn the influence of fine-scale processes without requiring computationally expensive high-resolution training? At this workshop, we will bring together experts in atmospheric physics, numerical modelling, and AI for climate science to discuss these challenges and explore strategies for the next generation of AI climate models.
Location: InnoDom Cologne, Weyertal 109, 50931 Cologne-Lindenthal
Contacts: KM Florentine Weber, Martin Schultz