An international team of scientists, led by professor Veronika Eyring from the German Aerospace Center (DLR) and the University of Bremen, and with participation from the University of Valencia, has published an innovative approach that integrates artificial intelligence (AI) with Earth system models to improve the accuracy and speed of climate simulations. The proposal, published in Nature Geoscience, is crucial for informed decision-making regarding climate change mitigation and adaptation.
The proposed approach integrates hybrid models that combine Earth system models with machine learning techniques, enabling faster and more accurate simulations, essential for the design of global and regional climate policies. By combining physics with AI, these models improve the generalisation capabilities in climate projections and provide a more accurate representation of key Earth system processes.
This work is the result of collaboration between the principal investigators of the European Research Council (ERC) Synergy Grant project Understanding and Modelling the Earth System with Machine Learning (USMILE). They are professor Veronika Eyring along with professors Pierre Gentine (Columbia University, USA), Gustau Camps-Valls (University of Valencia, Spain), Markus Reichstein (Max Planck Institute for Biogeochemistry, Germany) and David M. Lawrence (National Center for Atmospheric Research, USA).
"Integrating machine learning techniques with traditional climate modelling allows us to significantly advance our understanding of complex climate interactions and to improve our models. AI doesn’t just assist us-it is essential in redefining what our models can achieve", explains Gustau Camps-Valls, professor of Electronic Engineering and coordinator of the Image and Signal Processing research group at the University of Valencia.
"The integration of AI into climate models is a transformative step towards more accurate and useful projections", notes Veronika Eyring. This advancement addresses historical challenges in climate projection and enhances the representation of small-scale processes and feedback mechanisms that are crucial for understanding climate dynamics. Furthermore, the publication marks a milestone in climate forecasting, with significant implications for climate policy and strategies aimed at reducing greenhouse gas emissions.
Nature Geoscience is a leading journal that publishes high-quality research across all’areas of Earth sciences, including climate science, geology and oceanography. It aims to advance knowledge and promote interdisciplinary research to tackle the most pressing environmental challenges.
: Eyring, V., P. Gentine, G. Camps-Valls, D. M. Lawrence and M. Reichstein. ’AI-empowered Next-generation Multiscale Climate Modelling for Mitigation and Adaptation’, Nat. Geosci., https://doi.org/10.1038/s41561’024 -01527-w , 2024
: ’024 -01527-w
Figure (in the annex) : Schematic of the proposed AI-powered multi-scale climate modelling approach to advance mitigation and adaptation. This approach connects different scales and complexity of processes, providing faster and more accurate climate information at both regional and local levels. Based on Figure 2 of Eyring et al. (2024).
New AI models improve the accuracy and speed of climate simulations
Advert
