Climate modeling refers to the physics-based tools scientists use to understand how the Earth system responds to short- and long-term changes in the atmosphere, oceans, land and ice. By simulating the climate system over time, climate modeling is used to project long-term trends in the climate, the most prominent being climate change.
AI is rapidly expanding what is possible in climate modeling. AI systems can now reproduce the results of complex climate models much more quickly and at a lower cost. They can help scientists test and improve models, generate more climate scenarios, and create detailed local forecasts from broader global data. This can make climate projections faster, more accurate, and easier for researchers, policymakers, and communities to access and use.
This brief highlights the challenge of predicting climate trajectories that have never existed before. Current AI models perform best when future conditions resemble the data about past events on which they were trained. As climate change pushes the Earth system into conditions with no historical precedent, this may strain the ability of data-driven models to produce accurate results. These constraints may be addressed as AI systems incorporate more physical laws and become more capable of predicting rare events.
Despite this, the data used by both physical and AI climate models has geographic and historical gaps that may undercut their accuracy. AI systems inherit and can amplify these data gaps. Sustained investment in global observing systems is necessary for AI-driven systems to produce credible results.
This Brief stresses that AI climate modeling poses challenges for the 蘑菇视频. These include how to promote equitable access to AI-enabled climate information, particularly for developing countries; how to support transparency, quality assurance, and trust in AI-generated climate products; how to navigate the growing role of private-sector actors in climate modeling; and how to strengthen the observing systems and physical-based models foundational to AI based climate models and climate science
AI can now simulate centuries of climate in hours, yet its greatest challenge remains predicting a future unlike anything the Earth has experienced before.