Artificial intelligence and satellite data advancing climate modeling

A new approach, developed by a team led by Prof. Veronika Eyring of the German Aerospace Center (DLR) and the University of Bremen, integrates artificial intelligence (AI) with these models, aiming to improve both their accuracy and efficiency. The research has been published in two "Perspectives" articles in 'Nature', discussing future research directions for AI in climate science.

AI has the potential to address current limitations in climate models by improving the simulation of complex processes that existing models struggle to resolve. This approach could significantly reduce the computational load while enhancing precision, a vital step toward projecting climate changes more effectively. As noted in the 'Nature Geoscience' article, "AI-empowered next-generation multiscale climate modelling for mitigation and adaptation," machine learning is central to refining how these systems represent atmospheric processes and their interactions with oceans and land.

A new method for improved accuracy
Earth system models synthesize large amounts of data to simulate future climate scenarios. However, limitations in spatial resolution often result in inaccuracies. High-resolution models, which could address this, are currently too computationally expensive to run over long time scales. The novel AI approach allows for the integration of high-resolution climate processes into lower-resolution models, reducing systematic errors and improving predictions. The research team combines AI, satellite data, and climate modeling across different scales, which provides a more accurate representation of the Earth's climate.

"Satellite-based Earth observation data are invaluable for climate and environmental research," said Eyring. "We can and should use this resource much more intensively to calibrate, evaluate and improve global climate models. By combining AI with Earth system models and Earth observations, we will be able to project the complexity of Earth's future climate and extreme events with unprecedented accuracy."

Collaborative advancements in climate modeling
The AI-enhanced models developed through this collaboration between Germany, Spain, and the US, could also lead to more realistic and adaptable digital twins of the Earth system. These models allow users to interact with simulations and are scalable for various applications.

Prof. Gustau Camps-Valls from the University of Valencia highlighted the significant strides that AI allows in modeling, explaining, "Integrating machine learning techniques with traditional climate modelling allows us to make substantial strides in understanding complex climate interactions and in improving the models. AI is not just assisting us. It is an essential part of redefining what our models can achieve."

The development marks a major step forward for climate science, as AI models could not only project climate change impacts with increased precision but also assist in creating sector-specific strategies to address these changes. Accurate projections are essential for developing strategies to reduce greenhouse gas emissions and prepare society for future climate challenges. Dr. David Lawrence from the NSF National Center for Atmospheric Research, co-author of the study, noted that the approach will be a "crucial tool for planners and decision-makers worldwide."

The expanding role of machine learning
The use of machine learning in climate science is still developing. An international research team, including Eyring, examined the potential of AI to broaden the boundaries of climate research. A second article, published in 'Nature Climate Change' on August 23, 2024, explores challenges such as uncertainty quantification, causality, and the generalization of methods for a changing climate. The research calls for collaboration between machine learning experts, Earth observation specialists, and other industries to accelerate actionable climate science.

Research Report:AI-empowered next-generation multiscale climate modelling for mitigation and adaptation