Overview » Terms and Conditions for the AI Weather Quest

Terms and Conditions for the AI Weather Quest

Updated on 10 August 2026 – Effective from 13 August 2026

ARTICLE 1 – THE AGREEMENT

The ECMWF AI Weather Quest (“AI Weather Quest” or “the Competition”) is organised by the European Centre for Medium-Range Weather Forecasts (ECMWF), an intergovernmental organisation with headquarters at Shinfield Park, Reading, RG2 9AX, United Kingdom, and for the purposes of this Competition, also acting through its duty station at Robert-Schuman-Platz 3, 53175 Bonn, Germany, and regulated by these terms (“the Agreement”).

ARTICLE 2 – OBJECTIVES OF THE COMPETITION

The AI Weather Quest aims to advance sub-seasonal to seasonal weather predictions by using Artificial Intelligence (AI) / Machine Learning (ML) techniques.

ARTICLE 3 – PARTICIPATION

Participation in the Competition is free of charge and can take place at any point during the Competition. The Competition is open only to natural persons (“Participants”), either registering alone or jointly in teams with other Participants, filling in the required information and accepting the Agreement. Each person must (i) at the time of application be minimum eighteen (18) years old or have the age required by applicable law in their country of residence to validly enter this Competition, whichever is higher, (ii) have access to all necessary hardware and software to participate in the Competition and (iii) have sufficient AI/ML experience.

Participants proposing to work jointly with other Participants must designate a representative (“team leader”). On behalf of the Participants, who have designated him/her, the team leader registers the respective team of Participants for participation in the Competition and provides name, surname, email address, country of residence and nationality of each team member. The team leader shall act as a representative of the team, and respectively of all the Participants being the team members in this team, and the primary point of contact with the Competition organisers. To the same effect, the registration to participate in the Competition as a team by the team leader, constitutes an acceptance of this Agreement by each one of the respective team members and the obligations of a Participant as outlined hereto are applicable to each one of the respective team members. Furthermore, it is the team leader’s responsibility to ensure all team members are informed and agree to abide by the Competition’s rules, terms, and policies.

Each team is limited to ten (10) team members/Participants (including the team leader). Each individual may participate in only one team or alone. Teams may from time to time be supported by additional contributors, who are not to be listed as part of the team and shall not be considered Participants. Nevertheless, for the purposes of the Competition the Participants and the team as a whole shall be fully responsible for any such additional contributors.

The team leader can edit the registration details of the team they are representing (notably add or remove team members) or withdraw the team from the Competition directly through their login page on the AI Weather Quest website at any point during the Competition.

ARTICLE 4 – COMPETITION’S TERM AND FRAMEWORK

The AI Weather Quest started in March 2025. It began under an initial framework (“Initial Framework”) and, from 13 August 2026, continues under an extended framework (“Extended Framework”) as a long-term competition, benchmarking and learning initiative until terminated by ECMWF in accordance with Article 7. Information and results relating to the initial framework (from 17^th March 2025 until 13^th August 2026) are available in the Results section of the AI Weather Quest website. The Extended Framework preserves the core weekly, real-time sub-seasonal forecasting challenge established under the Initial Framework, while adding new forecast variables, transitional testing arrangements and associated evaluation and publication arrangements, as outlined below.

1. Initial framework

The Initial Framework applied from 17 March 2025 to 12 August 2026 and consisted of an Initial Training Phase followed by an Initial Competition Phase.

1.1. Initial Training Phase (17^th March 2025 – 13^th August 2025)

The Initial Training Phase took place from 17 March to 13 August 2025.

During this phase, participants could develop and train up to three (3) AI/ML-based models in order to submit weekly probabilistic forecasts for up to three (3) variables and two lead times.

A designated testing period within the Initial Training Phase for June, July, August 2025 (“Testing JJA period”) allowed Participants to trial the process before the Initial Competition Phase.

1.2. Initial Competition Phase (14^th August 2025 – 12^th August 2026)

The Initial Competition Phase took place from 14 August 2025 to 12 August 2026.

During this phase, Participants submitted weekly global probabilistic forecasts for at least one of the following variables:

Near-surface (2m) temperature (tas)
Mean sea level pressure (mslp)
Precipitation (pr)

Forecasts provided quintile probabilities at a 1.5-degree latitude/longitude resolution for one or both of the following lead times:

Days 19 to 25
Days 26 to 32

Forecasts were evaluated against climatology and displayed through weekly-updated public leaderboards. The Initial Competition Phase was organised into the four 13-week SON 2025, DJF 2025, MAM 2026 and JJA 2026 periods, with model methodology summaries and period results shared publicly.

2. Extended Framework (from 13 August 2026 onwards)

The Extended Framework applies from 13 August 2026 onwards. It continues the weekly submission and evaluation of forecasts for near-surface temperature, mean sea level pressure and precipitation and introduces forecasts for the Madden–Julian Oscillation and tropical storm days.

The Extended Framework consists of a Bridge Period followed by the Extended Competition Phase.

2.1 Bridge Period (13 August–11 November 2026)

The Bridge Period takes place from 13 August to 11 November 2026 (SON 2026) and constitutes a transitional period between the Initial Framework and the full implementation of the Extended Framework.

During this Bridge Period, Participants may test the submission process for the additional forecast variables introduced under the Extended Framework:

Madden–Julian Oscillation phase probabilities
Tropical storm days

These trial submissions are intended to allow Participants to familiarise themselves with the applicable forecast formats, technical requirements and submission processes. They will not be included in the official public evaluation, leaderboards or Competition results.

Forecasts for near-surface temperature, mean sea level pressure and precipitation will continue to be submitted, evaluated and published during the Bridge Period.

2.2 Extended Competition Phase (from 12 November 2026 onwards)

The Extended Competition Phase begins on 12 November 2026 and continues until the Competition is terminated by ECMWF in accordance with Article 7.

Under the Extended Framework, participants create, and submit under their own team’s name and up to three (3) model(s) name, forecasts for at least one of the following variables:

Near-surface (2m) temperature (tas)
Mean sea level pressure (mslp)
Precipitation (pr)
Madden-Julian Oscillation phase probabilities (MJO)
Tropical storm days (TS)

Forecasts of near-surface temperature, mean sea level pressure and precipitation must provide global, quintile probabilities at a 1.5-degree latitude/longitude resolution, whilst tropical storm day forecasts are basin-based tercile probabilities of tropical storm activity. Forecasts of these variables will be assessed at the following lead times (inclusive):

Days 19 to 25
Days 26 to 32

For MJO predictions, probabilistic forecasts of each MJO phase, as well as an inactive MJO state, at a 22 and 29-day lead time are required.

Day 1 marks the official forecast start date of each forecast workflow. Participants commit to using only data available up to this point in the forecasts. However, Participants have four (4) days to upload their forecasts to a platform specified by ECMWF to this effect, providing sufficient time to accommodate varying levels of resource capability.

Forecasts will be publicly accessible, including but not limited via an ECMWF-hosted sub seasonal AI forecasting portal. For every week, evaluation results will be computed by ECMWF and featured alongside both teams and model’s names on the Competition’s website.

The Competition is subdivided into 13-week periods:

September, October, November (SON)
December, January, February (DJF)
March, April, May (MAM)
June, July, August (JJA)

Participants may compete in any or all these periods.

Participants may leverage as inputs or reference points in their AI/ML model development processes (e.g., for re-forecasting techniques) open-source ECMWF forecast datasets or other forecast datasets, available to the Participants. For avoidance of doubt, ECMWF will not provide or arrange for any ECMWF or other forecast data to be made available to the Participants as part of the Competition.

The forecasts submitted in the Competition must be generated by Participants’ own AI/ML models, demonstrating original modelling efforts. Direct replication or reuse of ECMWF or third-party forecasts without significant transformation is considered not in line with the Agreement, and any Participant that fails to meet this criterion may be disqualified and excluded from further participation.

Weekly-updated leaderboards display Ranked Probability Scores (RPSSs) for all variables except for the MJO. For the MJO, leaderboards will be populated with Brier Skill Scores (BSSs).

Mean sea level pressure forecasts are assessed across all grid points.
Temperature forecasts are evaluated over land-dominated regions (>=50% land coverage).
Precipitation forecasts are evaluated over land-dominated regions (>=50% land coverage) while excluding persistently arid areas, i.e., regions where even the wettest climatological quintile has 0 mm of precipitation.
MJO forecasts are evaluated only for the Thursday within the forecasted week.
Tropical storm days forecasts are tercile-based and evaluated only across basins which are active during the forecasted 13-week period:
- JJA and SON: Atlantic Ocean and North-West Pacific Ocean.
- DJF: South-West Indian Ocean and South-East Indian Ocean.

All variables are benchmarked against climatology. Additionally, skill scores associated with global variables are area weighted.

To promote transparency, each participating model must have a model methodology summary, which Participants may update and resubmit whenever the model or its methodology changes. This summary is shared publicly.

Submitted forecasts are displayed on an ECMWF-hosted portal, with submitted data made publicly accessible on day 5 of the forecast submission schedule, following the closure of the submission window. Furthermore, ECMWF publicly highlights top performing teams through hosting exchange webinars from time to time, and through communication via the AI Weather Quest website or any other means. Further details on the evaluation methods and criteria, a detailed schedule for the submission of the forecasts and additional useful information, regarding the Competition can be found on the AI Weather Quest website.

ARTICLE 5 – INTELLECTUAL PROPERTY

Each Participant warrants that the forecasts submitted in the Competition have been generated by Participants’ own AI/ML models and are free from rights of third parties. Each Participant shall own any results that have been created by them during the Competition, such as the submitted in the Competition forecasts and associated model documentation, and all intellectual property rights that may be contained in such results. Each Participant grants to ECMWF a free, non-exclusive, irrevocable, worldwide, transferable, sub-licensable and perpetual license for the results, created by them during the Competition, including but not limited to the right to use, reproduce, distribute, modify, publicly display, and disseminate the results, free of charge and without any additional restrictions. The Participant warrants to be entitled to grant such rights.

For the Participant’s pre-existing intellectual property rights or rights that the Participant may have developed or acquired independently of the performance of the obligations and/or activities with respect to this Competition, the Participant grants ECMWF a free, non-exclusive, irrevocable, worldwide, transferable, sub-licensable and perpetual license to use all such intellectual property rights as are necessary for the use of the results. The Participant warrants to be entitled to grant such licenses. For avoidance of doubt, Participants are not required to share, licence or transfer their own AI/ML models in the context of the Competition and retain full ownership of them.

ARTICLE 6 – COMPENSATION

Participation in the Competition, including any achievements or outcomes, will not result in any monetary prizes, financial compensation, or in-kind rewards from ECMWF. Participants are solely responsible for any expenses incurred in connection with their participation in the Competition. Participants will, at no time, be considered staff members or consultants of ECMWF.

ARTICLE 7 – MODIFICATION AND TERMINATION OF THE AI WEATHER QUEST COMPETITION AND THIS AGREEMENT

ECMWF can terminate, shorten, extend, modify or otherwise change the Competition at any time without cause and with immediate effect. Any changes will be communicated to Participants as soon as possible. In case of termination of the Competition, this Agreement is deemed terminated with immediate effect.

At any time with immediate effect ECMWF can terminate this Agreement, and respectively exclude from the Competition any Participant, who violates any of the Agreement’s terms and/or the Competition’s rules, policies or processes. In particular, ECMWF can terminate this Agreement, respectively exclude from the Competition any Participant, if ECMWF has reasonable grounds to believe that the Participant has submitted any other but their own work as forecasts or has, in any way, shape or form, violated applicable provisions governing intellectual property regulations or criminal provisions in their country of origin or where ECMWF is located in conjunction with the submission of the forecast. For avoidance of doubt, in case of team’s participation, a violation of the Agreement’s terms or the Competition’s rules, policies or processes or any applicable provisions governing intellectual property regulations or criminal provisions by either of the team members or any of the supporting additional contributors, may lead to exclusion from the Competition of the entire team and termination of the Agreement with all the Participants of this team, and respectively with the team.

In any case of a termination, ECMWF does not owe fees or compensation to any of the Participants.

ARTICLE 8 – PERSONAL DATA

The AI Weather Quest Data Protection and Privacy Statement applies to all personal data provided by the Participants in order to register and submit their forecast.

ARTICLE 9 – CONFIDENTIALITY

The teams of Participants have the option to anonymise their identities at any moment during the Competition. While ECMWF will still retain knowledge of the Participants’ identities for administrative purposes, when their results are made public such results will not be linked to their individual identity. However, the public leaderboard will display the team’s name and model’s name.

ARTICLE 10 – RESPONSIBILITY

The AI Weather Quest website or any of its content may be unavailable from time to time for scheduled maintenance or due to unforeseen circumstances.

Other than in cases of intentional or grossly negligent breaches of duty by its staff or representatives or claims resting on an injury to life, body, or health, ECMWF shall not be responsible or liable to any Participant for any loss or damage of any sort incurred in connection with the Competition.

ECMWF shall not be responsible or liable for the accuracy, usefulness, or availability of any content of the Competition website. Any content made available for download or use through or within the Competition website is supplied “as is” without further warranties of any kind, either express or implied, including, but not limited to, the quality, performance, merchantability, or fitness for a particular use or purpose. Subject to the aforesaid, ECMWF shall not be liable for any damages, including but not limited to direct, indirect, special, incidental, punitive, exemplary, or consequential damages arising out of the use or the inability to use content.

ECMWF shall not be liable and accept no representation or responsibility for the functionality or contents of external websites, platforms, services, data, or software products hyperlinked from the Competition website and/or accessed and/or used by the Participants for or in relation to the Competition.

ECMWF disclaims all warranties related to the provision of the Competition.

ARTICLE 11 – VISIBILITY

Participants acknowledge and agree that ECMWF may report on the AI Weather Quest to the general public and name and use footage of non-anonymous Participants in any such communication.

The AI Weather Quest shall be recognised in communications by the Participants.

ARTICLE 12 – APPLICATION OF THE TERMS AND CONDITIONS, DISPUTE RESOLUTION

Any terms, conditions and/or licenses associated with the use of the ECMWF and/or any third party forecast datasets and/or ECMWF or third-party platforms and/or websites for or in relation to the Competition shall be directly applicable to the Participants in addition to this Agreement.
Nothing in or relating to this Agreement shall be deemed a waiver of any of the privileges and immunities conferred upon ECMWF by virtue of its Convention and Protocol on the Privileges and Immunities.
Any dispute between ECMWF and interested parties arising out of, or relating to, the Competition shall be resolved amicably by negotiation. If the dispute cannot be so settled, the parties agree to first enter into mediation in good faith to settle such a dispute and will do so in accordance with the Centre for Effective Dispute Resolution (CEDR) Model Mediation Procedure. Unless otherwise agreed between the parties within fourteen (14) days of notice of the dispute, the mediator will be nominated by CEDR. To initiate the mediation a party must give notice in writing to the other party to the dispute, referring the dispute to mediation. A copy of the referral should be sent to CEDR. Unless otherwise agreed, the mediation will start not later than twenty-eight (28) days after the date of the notice referred to in this Article.
If the dispute cannot be settled in accordance with Article 12.3 above, it shall be finally settled under the Rules of Arbitration of the International Chamber of Commerce by three arbitrators appointed in accordance with the said rules; sitting in London, England. The proceedings shall be in the English language. The right of appeal by either party to any national courts is hereby agreed to be excluded.