Karlsruhe Institute of Technology researchers are developing an artificial intelligence world model that unifies global climate, weather, and environmental simulations. The WOW project aims to advance climate modeling, from the study of shifting global patterns to more accurate assessments of local ecosystem impacts. KIT's Institute of Theoretical Informatics, coordinated by Professor Peer Nowack, leads the five-year initiative with six million euros in funding provided by the Carl Zeiss Foundation.
Current numerical models facilitate estimates of interrelated changes across broad geographic and temporal scales. However, modeling the Earth system with the necessary complexity has proven challenging. According to Nowack, artificial intelligence can accurately replicate comprehensive physics-based models and can be refined using direct observational data. In recent years, AI models have yielded superior results in weather forecasting. WOW will apply this transformative methodology to broader environmental modeling.
The project investigates how distinct AI models representing Earth system processes can be linked via shared latent spaces – internal data representations within AI frameworks. This coupling strategy provides efficiency and global consistency and is expected to enhance connections among climate simulators, weather predictors, and local impact models for phenomena such as wildfires and floods. The research team, comprised of computer scientists and environmental scientists, will develop modular approaches to connect initially autonomous AI sub-models, establishing an integrated chain from global changes to local effects.
WOW will deepen understanding of nonlinear atmospheric, hydrological, and land interactions. Professor Almut Arneth from KIT's Institute of Meteorology and Climate Research notes that the project will explore how variability in one part of the Earth system, such as droughts or altered cloud formation, affects climate dynamics. This can reveal previously unknown connections across climate processes.
The AI world model may lead to improved assessments of risks and underpin decisions for climate adaptation and mitigation. Dr. Markus Gotz from KIT's Scientific Computing Center highlights that efficiently coupled AI models will support faster, more accurate analyses, broadening applications in other complex scientific domains. Funding for WOW spans five years, with six million euros provided by the Carl Zeiss Foundation.
