CAIRNE Community Invited to Join ECMWF’s AI Weather Quest

The European Centre for Medium-Range Weather Forecasts (ECMWF) has launched an international machine learning competition – the AI Weather Quest – aimed at improving sub-seasonal weather forecasts through artificial intelligence. Endorsed by the World Meteorological Organization (WMO), the competition offers a unique opportunity for AI researchers to apply their methods to a complex, real-world challenge with broad societal relevance.

The CAIRNE community, representing some of Europe’s leading artificial intelligence research laboratories, is particularly well-positioned to contribute. The competition is open to all, and no prior background in weather forecasting is required.

Why sub-seasonal forecasting?

The focus of the AI Weather Quest is on days 19–25 and 26–31, a timeframe sometimes referred to as the “predictability desert.” It lies between traditional weather forecasting (up to ~15 days) and seasonal climate prediction (months ahead), where current systems struggle to deliver reliable forecasts. Improving accuracy at this range could bring significant benefits for decision-making in sectors such as energy, agriculture, water management, and disaster preparedness.

Structure of the competition

The AI Weather Quest is structured around two key phases:

Initial Training Phase (March–August 2025)

This phase gives teams the opportunity to:

  • Access the datasets and resources provided by ECMWF
  • Understand the forecasting targets, submission format, and scoring system
  • Participate in community exchanges and technical onboarding

A key part of this phase is the Testing JJA Period (June–August 2025), where teams can submit their forecasts in real time and self-evaluate their performance using transparent metrics. Although results during this period are not scored competitively, the experience is relevant for preparing for the second phase.

Competitive Phase (September 2025 onwards)

In this phase, teams will be challenged to submit real-time forecasts for near-surface air temperature, mean sea level pressure, and precipitation: but results will now be scored and compared on a public results board. Participating teams will be asked to document their methodology to promote transparency and scientific exchange.

How to participate

The competition welcomes participants from AI research labs, companies, startups, and independent researchers working in machine learning or related fields. ECMWF encourages cross-disciplinary collaboration and the formation of diverse teams.

Learn more and register on the official competition platform: https://aiweatherquest.ecmwf.int

Join Pre-Competition Webinar

We invite you to join one of our interactive launch webinars on July 22nd, 2025 to get ready for the competition:

The AI Weather Quest offers the CAIRNE community a concrete opportunity to demonstrate the power of European AI in advancing climate resilience – and to help shape the next generation of weather forecasting tools.