The European Centre for Medium-Range Weather Forecasts (ECMWF) has launched the AI Weather Quest, an international machine learning competition endorsed by the World Meteorological Organization (WMO). The competition invites participants to develop AI-based models to improve forecasts in the sub-seasonal range, specifically targeting days 19–25 and 26–31 ahead.
As a close partner of ECMWF, CESOC encourages its research community – spanning Earth system observation, modelling, and computational analysis – to actively engage in this initiative. The AI Weather Quest aligns closely with CESOC’s vision of integrating observational insights and data-driven methods to better understand and predict the Earth system in a changing climate.
Forecasting challenge
Sub-seasonal forecasting represents a significant scientific challenge. This time window lies between medium-range weather forecasting and long-term climate prediction and is often referred to as the “predictability desert.” Improving forecast skill at this scale is essential to support decision-making in sectors such as agriculture, water management, disaster preparedness, and energy.
Participants in the ECMWF AI Weather Quest are invited to submit probabilistic forecasts for any or all of the following variables:
- Near-surface air temperature
- Mean sea level pressure
- Precipitation
Forecasts may be submitted for one or both of the following lead times:
- Days 19–25
- Days 26–31
Two-phase competition structure
The competition consists of two main phases:
- Initial Training Phase (March–August 2025)
This phase provides teams with access to training data, submission tools, and evaluation frameworks. A key component of this stage is the Testing JJA Period (June–August), during which teams can submit real-time forecasts and self-evaluate their performance using ECMWF’s tools. This phase is non-competitive and designed to help participants refine their methods ahead of the official competition. - Competitive Phase (starting September 2025)
In this phase, teams will continue to submit forecasts under competitive conditions. Scores will be published on the AI Weather Quest website’s public results board, and teams will be required to provide documentation outlining their forecasting methodology.
How to participate
The competition is open to AI researchers, Earth system scientists, and interdisciplinary teams from academia, research institutions, and industry. CESOC encourages its affiliated researchers to take part and contribute their expertise to this forecasting challenge.
To support participants, ECMWF will host two live webinars on 7 May 2025 to guide teams through the process of submitting and self-evaluating their forecasts during the Testing JJA Period:
🔗 For more information and to join the Quest: https://aiweatherquest.ecmwf.int
The AI Weather Quest offers a valuable opportunity for the CESOC community to apply AI methods to one of the most complex and relevant forecasting problems – helping to advance the understanding and prediction of the Earth system at sub-seasonal timescales.
—
If you would like to register for membership and for more information, please visit the page: https://cesoc.net/become-member/ or contact us info@cesoc.net.