“There are no perfect AI models, but there are useful ones – and that’s what matters!” – Sophie Xhonneux
This was the spark for a lively panel discussion on “AI as an Opportunity for Climate Research?”, hosted by the Sparkasse KölnBonn Stiftung Wissen in the inspiring rooms of the Tyssen Foundation.
Our CESOC Directors Susanne Crewell and Christian Sohler (University of Cologne), Philip Stier (University of Oxford, Director of Intelligent Earth), and Sophie Xhonneux (European Centre for Medium-Range Weather Forecasts – ECMWF) explored the current and future role of AI in climate science, chaired by Carolin Riethmüller.
The discussion highlighted a shared view among the researchers: climate models can used as tools for experimentation before implementing changes in the real world, for instance, testing the effects of building additional houses in Cologne’s city center or predicting localised warming. Understanding what works and what does not in simulation technology remains a key part of climate research.
One particularly engaging topic was what AI still needs to fully support researchers. Crewell emphasised observational data, such as near-surface measurements, drone-based methods, because “AI is only as good as its training data.” She, like Stier, is fascinated by clouds and their complexity. Stier added: “We need smart people! AI is embedded in every field of science, but it cannot replace human ingenuity. And we need funding to make this happen.” Xhonneux highlighted the importance of long-term projects to maintain consistent datasets.
Data governance was another recurring theme: controlling data often means controlling AI. With geopolitical challenges in mind, preserving and securing datasets is critical for reliable research.
Sohler shared even other research ambitions: exploring how AI tools can be integrated into scientific workflows to generate new insights collaboratively.
The overarching message was clear: AI opens incredible possibilities, but it still needs the human brain to guide it. Any scientific tool can be applied for benefit or harm – and thoughtful application remains crucial.
The session concluded with a mix of answered questions, questions still waiting for answers, and new questions ready to be asked – a reminder that the dialogue between AI and climate science is just beginning.