SEOUL, May 13 (AJP) - Researchers at the Korea Advanced Institute of Science and Technology (KAIST) developed an artificial intelligence framework designed to analyze the integrated impacts of climate change on global economies and energy systems, KAIST said Wednesday.
The international research team, led by Professor Jeon Hae-won and Professor Oh Hye-yeon, introduced a foundation model that processes earth observation data, economic scenarios, and policy indicators within a shared virtual space. This system allows for the simultaneous analysis of physical climate phenomena and their resulting socio-economic effects.
Existing climate research often separates physical weather predictions from economic impact assessments, which leads to delays in policy decision-making due to fragmented data systems. The new AI framework utilizes a mixture of experts structure, where specialized AI modules collaborate to improve the accuracy and reliability of long-term forecasts.
The team also released a prototype tool called the Machine Learning-Integrated Assessment Model (ML-IAM) v1.0. This high-speed emulator can process thousands of different policy scenarios within minutes, whereas traditional integrated assessment models often require several hours to analyze a single scenario.
Testing showed that the AI emulator achieved 97 percent accuracy when compared to 15 different international integrated assessment models. The researchers stated that the tool can simulate the immediate effects of policy changes, such as increasing carbon taxes or expanding renewable energy infrastructure.
"The climate-AI model is expected to bridge the gap between climate scientists and policymakers," Professor Jeon Hae-won said Wednesday. "The high-speed AI emulator will become a core technology for providing practical climate solutions by enabling near real-time policy analysis."
The research was conducted in collaboration with institutions including Peking University, Imperial College London, and the International Institute for Applied Systems Analysis. The findings were published in the journal Nature Climate Change on April 28, 2026, while the technical details of the emulator were presented as a preprint in Geoscientific Model Development on January 9, 2026.
"AI technology must contribute to solving the climate crisis that threatens human survival beyond being a mere commercial tool," Professor Oh Hye-yeon said Wednesday. "This international joint research demonstrates that AI can serve as a global public good to address social challenges."
(Reference Information)
Journal/Source: Nature Climate Change
Title: Artificial Intelligence to Support Cross-Disciplinary Climate Change Research
Link/DOI: https://bit.ly/4fi4MpR
Journal/Source: Geoscientific Model Development
Title: ML-IAM v1.0: Emulating Integrated Assessment Models With Machine Learning
Link/DOI: https://bit.ly/4u8Yc9W
Copyright ⓒ Aju Press All rights reserved.