NvidiaIsha Salian4 min readintermediate
University of Manchester Uses NVIDIA Earth-2 to Forecast Air Pollution Across the UK
Summary
University of Manchester used NVIDIA Earth‑2 CorrDiff and StormCast generative models to downscale UK‑wide air‑pollution simulations. Training on the Isambard‑AI supercomputer (5,448 GH200 chips, 21 EFLOPS) took two days on an eight‑GPU node, producing a 2‑3 km resolution model. Inference runs on a desktop‑class DGX Spark, enabling rapid scenario forecasting and potential real‑time health alerts.…
- Generative downscaling (Earth‑2 CorrDiff) can replace expensive chemistry‑heavy weather models for air‑quality forecasting.
- Training on a national AI supercomputer (Isambard‑AI) took ~2 days for a year‑long, hourly UK dataset at 2‑3 km resolution.
- Inference can be performed on a single DGX Spark workstation, democratizing access to high‑resolution pollution forecasts.
- The team plans to release training data and workflows, aiming for global reproducibility and fine‑grained (street‑level) forecasts.
High‑resolution, fast‑turnaround air‑quality forecasts can inform public‑health interventions and policy decisions, especially as traditional chemistry‑based models are computationally prohibitive for frequent, fine‑grained runs.
5/10
