WeatherGenerator

The WeatherGenerator Horizon Europe project aims to develop a foundation model of the Earth System. This advanced AI system is meant to improve weather and climate forecasts and related downstream applications in Renewable Energy, Health and Disaster Resilience to better prepare society before extreme climate shocks.

The WeatherGenerator project, coordinated by the European Centre for Medium-Range Weather Forecasts (ECMWF), is a collaborative effort involving 16 European organizations, specializing in high-performance computing, machine learning, and Earth system modeling. By integrating machine learning techniques, WeatherGenerator aims to improve the accuracy and efficiency of weather and impact predictions, thereby contributing to better preparedness and response to weather-related events.

The Weather Generator project develops a foundation model of the Earth system using AI to analyze vast atmospheric and land surface data. Here, the EarthNet Initiative builds the land surface foundation model component by exploiting terabytes of satellite data. This model enables detailed forecasting of environmental changes, such as vegetation responses to droughts and high-resolution heat wave predictions. Additionally, it helps understand carbon exchange in ecosystems, examining how plants absorb CO₂ through photosynthesis and how soil releases it via respiration. These insights contribute to climate research and inform climate agreements by assessing the land’s role as a carbon sink.

Visit WeatherGenerator.eu
For further information

EarthNet Team Contributions


WeatherGenerator-Land
We explore self-supervised learning of land surface dynamics with our land surface foundation model WeatherGenerator-Land.
Vegetation modeling for Food Security
Leveraging WeatherGenerator-Land for drought impact forecasting in a food security context on sub-seasonal to seasonal time scales.
Temperature Forecasting for Health
Downscaling forecasts of heat waves to very high resolution land surface temperature with WeatherGenerator-Land.
Biosphere Fluxes
Diagnosing the exchange of energy, water and carbon between the land surface and the atmosphere using WeatherGenerator-Land embeddings.

Involved team members

Markus Reichstein

Markus Reichstein
Director


    Vitus Benson

    Vitus Benson
    PhD Candidate


    • Interests: Beers&Coding&Sunsets
    Sebastian Hoffmann

    Sebastian Hoffmann
    PhD Candidate


      Marieke Wesselkamp

      Marieke Wesselkamp
      PostDoc


      • Affiliation: Max-Planck-Institute for Biogeochemistry
      • Interests: Earth System Sciences, Forecasting, Machine Learning.
      Markus Zehner

      Markus Zehner
      Data Scientist


        Mélanie Weynants

        Mélanie Weynants
        PostDoc


          Christian Reimers

          Christian Reimers
          Project Group Leader


            Nuno Carvalhais

            Nuno Carvalhais
            Research Group Leader


              Partners

              Overview of all the WeatherGenerator Projects: Co-ordinated by ECMWF, partners involving: Forschungszentrum Jülich, MetNorway, Max Planck, KNMI, Meteo France, Met Office, CMCC, eScience Center, buluttan, KAJO, Latest Thinking, Statkraft, ETH Zürich, SMHI, MeteoSwiss. Funded by European Union.