Glossary
October 16, 2026

Europe's Weather Agency Tests an AI Program That Forecasts Smoke, Dust and Air Pollution Ten Days Ahead

The European Centre for Medium-Range Weather Forecasts, which makes some of the world's most trusted forecasts, introduced an experimental AI program that predicts what's in the air, from wildfire smoke and desert dust to the pollution that makes breathing hard. It looks twice as far ahead as today's system and runs on a fraction of the computing power. It's still a test, and the agency says its current system remains the standard.

The European Centre for Medium-Range Weather Forecasts, announced a new AI program today that predicts air quality around the world up to ten days ahead.

The program is called AIFS-COMPO. It forecasts what's floating in the air: smoke from fires, dust blown off deserts, salt from the sea, and pollutants such as the tiny particles and gases that can make the air unhealthy to breathe. It also tracks ozone, including the thinning of the ozone layer over Antarctica each year, which its ten-day forecasts still catch.

The agency's air-quality service already makes forecasts like these with a traditional program that calculates the physics and chemistry of the air step by step. That program looks five days ahead and needs thousands of computer processors working together. The new AI program learned instead from about two decades of the agency's own past air-quality records, and it looks ten days ahead. In early tests, the agency says, it was about as accurate or better for several important kinds of pollution, at a small fraction of the computing cost. For the haze of fine particles in the air, checked against ground instruments over a summer of tests, its errors were lower than the current system's.

The agency is careful about what it is and isn't. It's experimental, not the official forecast. It can't start on its own: each forecast begins from a snapshot of today's air made by the traditional system, which combines satellite readings. It covers fewer pollutants, shows less local detail, and doesn't yet account for fires or factory emissions changing day to day, which the team is working to add. "It is a very exciting moment for us," said Laurence Rouil, who directs the agency's air-quality service, while adding that "the operational CAMS forecasting system remains the benchmark for accuracy and reliability."

For now, the first maps are posted on the agency's website, and the underlying data will be shared once the program has been tested further. An outside review of the program by a tech writer noted the same thing the agency does: the AI is fast, but it still leans on the old system to get started.

Sources

  1. ECMWF — ECMWF introduces AIFS-COMPO, a new AI system for atmospheric composition forecastingecmwf.int · primary
  2. arXiv — AIFS-COMPO: A Global Data-Driven Atmospheric Composition Forecasting System (the research paper)arxiv.org
  3. remio — ECMWF's AIFS-COMPO extends pollution forecasts but still depends on physicsremio.ai
Follow the timeline
Get an email when new entries are added.
© 2026 Sugarpine