Spherical Diffusion Model Accelerates Century-Scale Climate Forecasting to 25 Hours Using AI
Researchers at UC San Diego and the Allen Institute for AI developed Spherical DYffusion, a generative AI model that integrates physics-based data to simulate 100 years of climate patterns in just 25 hours. This hybrid approach leverages diffusion probabilistic models tailored to spherical data domains, significantly speeding up complex climate projections while maintaining scientific validity.
This breakthrough illustrates the power of combining domain-specific physics knowledge with generative AI to tackle computationally intensive simulations. For climate scientists and AI practitioners, it suggests a workflow where AI accelerates traditional modeling, enabling faster scenario testing and decision-making.
The UC San Diego computational climate group and the Allen Institute’s AI division jointly achieved this. Their model drastically cut simulation time from weeks or months to less than a day, a game changer for climate research.
Step 1: Obtain the Spherical DYffusion codebase from UCSD’s or Allen Institute’s repositories. Step 2: Prepare your climate data formatted for spherical coordinate input. Step 3: Run the model on a high-performance GPU cluster to generate century-scale climate projections within 25 hours. Reference https://today.ucsd.edu/story/nine-breakthroughs-made-possible-by-ai for details.