Tech

Spherical DYffusion Model: A Century of Climate Forecasts in 25 Hours

April 9, 2026 · From the archive, written in the Professor's voice before our October 2026 redesign from today.ucsd.edu

What happened

Scientists at UC San Diego and the Allen Institute for AI have developed 'Spherical DYffusion,' a hybrid generative AI and physics-based model that simulates 100 years of global climate patterns in just 25 hours. This represents a substantial acceleration over traditional climate modeling methods, combining deep learning diffusion techniques with spherical data projections.

Why it matters

This breakthrough illustrates how integrating generative AI with domain-specific physics models can exponentially speed up complex simulations. For practitioners, it encourages adopting hybrid models that leverage AI's pattern recognition alongside established scientific principles to enhance both speed and interpretability in forecasts.

Who's doing it

The collaborative team from UC San Diego and the Allen Institute for AI has demonstrated this model's capabilities, offering a new tool for climate scientists to conduct rapid, long-term environmental assessments with higher temporal resolution.

Try it

1. Review the Spherical DYffusion model details at UC San Diego's news portal (https://today.ucsd.edu/story/nine-breakthroughs-made-possible-by-ai).
2. Acquire the model codebase (check linked repositories or contact authors for access).
3. Run the model on suitable high-performance computing resources, inputting relevant climate data to generate accelerated long-term climate projections, verifying outputs against known historical patterns.

Read the original at today.ucsd.edu