AI and Climate Prediction: Opportunity or Danger?

On 3^rd July 2026, the Oxford Climate Research Network (OCRN) welcomed researchers from across academia and operational forecasting for a lively panel discussion exploring a question at the forefront of climate science: Is the success of AI models for weather forecasting an opportunity or danger for climate prediction? The event brought together experts in atmospheric science, climate modelling, and machine learning to examine how rapidly advancing AI methods could shape the future of climate prediction.

Chaired by Philip Stier (Professor of Atmospheric Physics, University of Oxford), the panel featured Tony McNally (Head of Earth System Assimilation, ECMWF), Kirstine Dale (Chief AI Officer, Met Office UK), and Tim Palmer (Royal Society Research Professor in Climate Physics, University of Oxford). Together, they offered complementary perspectives on the opportunities and challenges presented by AI across weather forecasting and climate prediction.

A recurring theme throughout the discussion was that AI is already transforming weather forecasting. Panellists highlighted the rapid progress of AI forecasting systems and their major developments in speed and efficiency. The conversation, however, focused on how AI and physics-based models can work together, rather than an approach where AI models entirely replace the existing methods. This was seen to combine the strengths of data-driven methods with existing scientific expertise.

The discussion also explored apparent limitations of AI in this sphere. While machine learning has demonstrated impressive results in weather forecasting, climate prediction presents additional challenges, including longer timescales, changing climate conditions, model interpretability and the need to understand why predictions are made. Panellists emphasised the continued importance of observations, robust evaluation, and physical reasoning in building confidence in future climate predictions.

Another key message was the importance of collaboration. Speakers stressed that advances in AI have been driven by expertise from multiple disciplines and argued that the climate community should embrace interdisciplinary ways of working rather than viewing AI and traditional modelling as mutually exclusive approaches. It was highlighted that equipping the next generation of climate scientists with AI skills is an important priority for the field – a priority Oxford is addressing through our Intelligent Earth UKRI AI Centre for Doctoral Training in AI for the Environment.

The engaging audience discussion and networking lunch that followed reflected the strong interest in this rapidly evolving area of research. OCRN thanks our panellists, Chair, and everyone who attended and contributed to such a stimulating conversation. As AI capabilities continue to develop, discussions like these will remain essential in shaping how the climate science community utilizes new technologies while maintaining scientific rigour.