CESOC/ELLIS Workshop on “How Resolution Matters: Shaping the Future of AI Climate Modelling”

Artificial intelligence is rapidly transforming Earth system science, yet a fundamental question remains at the core of AI-based climate modelling: how much resolution do training models actually require? To address this challenge, the upcoming CESOC/ELLIS workshop “How Resolution Matters: The Future of AI Climate Modelling” will bring together researchers and experts in atmospheric physics, numerical modelling, and machine learning on November 6, 2026, at the InnoDom in Cologne. Supported by the HClimRep project, the CESOC ML Working Group, and the ELLIS Unit NRW, this event offers an ideal platform to explore how next-generation AI models can achieve both high accuracy and computational sustainability.

Efficiency and Physical Precision in Harmony

In traditional numerical climate modelling, higher resolution typically ensures a better representation of fine-scale atmospheric dynamics and weather events, but it incurs significant energy and computational costs. Emerging AI approaches show that training models at coarser resolutions and applying downscaling techniques can capture essential large-scale climate patterns while significantly reducing computational requirements. However, critical scientific questions remain regarding whether these coarse models maintain physical accuracy under a changing climate, particularly given the complex interactions between large-scale dynamics and small-scale processes.

Interdisciplinary Exchange for Sustainable AI Strategies

To address these uncertainties, the workshop promotes interdisciplinary collaboration between atmospheric science and artificial intelligence. The program includes perspective talks, lightning presentations, interactive breakout sessions, and plenary discussions aimed at challenging existing assumptions, bridging domain knowledge, and defining new research directions. Whether attending in person in Cologne for networking or joining online, participants are encouraged to actively shape the future of AI-driven climate research. Registration is open, with deadlines set for October 20 for in-person attendance and November 3 for virtual participation.