Foundation Models Proposed to Bridge Climate Science and Society

Context Connecting physical climate risks with societal actions and their feedbacks has long been a challenge for researchers. Most models treat the Earth system and human responses separately, which limits the ability to anticipate real‑world outcomes. The paper proposes that large‑scale AI systems, known as foundation models, could serve as a common framework for both domains.

Key Facts The authors state that integrating climate risk knowledge with societal responses is challenging but could be advanced by using AI foundation models to unify the fields and aid climate decision‑making. The article appears in Nature Climate Change with DOI 10.1038/s41558-026-02624-x and was published in 2026. Readers can access the full text via Nature+ for 27.99 euros for a 30‑day subscription.

What It Means By feeding climate observations, impact data, and policy information into a single foundation model, researchers could generate forecasts that reflect both physical changes and likely societal reactions. This integrated output may help planners design adaptation strategies that are more aligned with real‑world behavior. The authors caution that the approach remains conceptual and requires validation with observed datasets.

What to Watch Next Future studies will need to test foundation models on historical climate‑impact pairs and policy outcomes to assess whether they reduce uncertainty in regional risk projections. Demonstrating practical skill gains will be key to moving the idea from theory to operational use.