Researchers from the Institute of Atmospheric Physics (IAP) at the Chinese Academy of Sciences, in collaboration with researchers from Seoul National University and Tongji University, advocated for a collaborative approach between artificial intelligence (AI) and physics in climate modeling, moving beyond the notion of an "either/or" scenario.

Their perspectives paper, published in Advances in Atmospheric Sciences, illustrates methods for imposing both soft and hard physical constraints on AI models to ensure consistency with known atmospheric dynamics.

AI is bringing remarkable changes to atmospheric science, particularly with the introduction of large-scale AI weather models such as Pangu Weather and GraphCast. However, along with these advances, questions have arisen about the consistency of these models with fundamental physics principles.

Previous studies have shown that Pangu-Weather can accurately reproduce certain climate patterns, such as tropical Gill responses and extra-tropical teleconnections, through qualitative analysis. However, quantitative investigations have revealed significant differences in wind components, such as divergent winds and ageostrophic winds, within current AI weather models. Despite these findings, there is still concern that the importance of physics in climate science is sometimes overlooked.

"The qualitative assessment finds that AI models can understand and learn spatial patterns in weather and climate data. On the other hand, the quantitative approach highlights a limitation: current AI models struggle to learn certain wind patterns and instead rely solely on total wind speed," said Prof. HUANG Gang of IAP, corresponding author of the study. "This underscores the need for a comprehensive dynamic diagnosis of AI models. Only through a holistic analysis can we improve our understanding and impose necessary physical constraints."

"While AI excels at capturing spatial relationships within weather and climate data, it struggles with nuanced physical components such as divergent winds and ageostrophic winds. This underscores the need for rigorous dynamic diagnostics to enforce physical constraints," said Prof. HUANG.