NVIDIA released Earth-2 CorrDiff on September 15, 2026, saying it can forecast air pollution with high resolution using generative frameworks. It is the company's first climate and weather AI update since the launch of the Earth-2 family.
NVIDIA reported the model ran on Isambard-AI, the UK's national AI supercomputer, achieving a resolution of 2-3 square kilometers. That compares with traditional chemistry-based models that are slower and less detailed.
Earth-2 CorrDiff is built on NVIDIA's AI platforms and targets environmental science applications, particularly air quality forecasting. Availability begins with open-source training data and workflows for researchers globally.
"The fact that this model trained in two days on Isambard-AI — and can now run on a DGX Spark sitting on a desk — changes who can do this science and how quickly," said Niall Robinson, developer relations manager for weather and climate at NVIDIA. The model can now be used for real-time decision making and predictive scenarios.
The announcement follows a collaboration between NVIDIA and the University of Manchester. NVIDIA said the model represents a significant step in open science and environmental modeling, enabling global access to pollution data.
NVIDIA did not say how the model will be scaled to other regions, and raised the open question of how to handle real-time data integration with edge AI devices. The team plans to further increase the model's resolution by incorporating additional open data.
Source: nvidia
