Tech
Manchester Team Brings UK Pollution Forecasting to Desktop AI
University of Manchester researchers used NVIDIA Earth-2 models to create detailed UK-wide air pollution forecasts faster and more affordably.
What happened
Researchers at the University of Manchester retrained NVIDIA’s Earth-2 AI frameworks to model UK air pollution at a 2–3 square kilometre resolution. They generated training data from a year of hourly chemistry-climate simulations and trained the Earth-2 CorrDiff model on Isambard-AI, the UK’s most powerful AI supercomputer. Training took two days on one eight-GPU node. The team also added Earth-2 StormCast for time-dependent forecasts using air-quality observations. The workflow can run on NVIDIA’s DGX Spark desktop system, and the researchers plan to release open-source training data and workflows for other countries and regions.
Why it matters
The model reduces the computing burden of detailed pollution forecasting and makes smaller-scale development possible on desktop AI hardware. It could support policy scenario modelling, healthcare alerts and real-time responses to events such as wildfires.
Source: NVIDIA Blog