Data-driven AI models are currently reshaping the field of numerical weather prediction (NWP). Instead of relying solely on physics-based NWP models, neural networks can be directly trained on historical weather analyses to generate forecasts with accuracy comparable to the state-of-the-art NWP systems. This development has been enabled by the availability of long, high-quality global reanalyses, such as ERA5 from the EU Copernicus programme, together with advances in machine learning algorithms and high-performance GPU computing.
Professor Marko Laine’s group at the Finnish Meteorological Institute (FMI), as part of FMI’s AI weather forecasting team, was among the recipients of the European Meteorological Society (EMS) Technology Award in 2025. The award was granted to the Anemoi community in recognition of the Anemoi framework, developed through collaboration across the European meteorological community.
Anemoi represents a pioneering effort in integrating modern machine learning methods into weather forecasting and in establishing an open, shared foundation for data-driven meteorology. Alongside its own global AI forecasting system (AIFS), the European Centre for Medium-Range Weather Forecasts (ECMWF) has initiated and coordinated the development of the Anemoi framework as an open, modular, and extensible platform for AI-based weather prediction.
Designed to support both research and operational use, Anemoi fosters collaboration among national meteorological services, research institutes, and academia. The EMS Technology Award in 2025 highlighted the scientific, technological, and community building impact of this effort.
Laine’s team is involved in both the application and development of Anemoi. An important collaborator has been Met Norway, who’s AI model Bris has been the model for FMI’s own AI-based weather model, Aila, designed to complement FMI’s existing forecasting systems and explore the added value of machine learning in operational meteorology, particularly for high-resolution regional forecasting in the Nordic and Scandinavian regions. Aila is currently in semi-operational use and under continuous development.
FMI’s Aila model utilises the Anemoi framework and incorporates a modern graph neural network architecture with an encoder processor-decoder structure. This solution supports the use of flexible model grids, including stretched-grid approaches that concentrate higher spatial resolution over regions of interest while retaining global context. Aila’s training has relied on large-scale, high-quality datasets and advanced computing resources. The model has been trained using approximately 40 years of global ERA5 reanalysis data, complemented by several years of high-resolution regional analysis data over Scandinavia.
FMI’s contribution to the Anemoi community and the development of the Aila AI weather model embody the goals of the FAME Flagship: combining cutting-edge AI research, large-scale computing infrastructures, and strong international collaboration to deliver transformative advances in environmental forecasting. The EMS Technology Award represents an external recognition of this collective achievement and underscores the importance of FMI’s work within the European and global meteorological landscape.
Read more: https://www.emetsoc.org/ems-technology-achievement-award-2025-for-anemoi/
