Numerical weather prediction is changing fast, and this Newsletter bears witness to that. Consider, first of all, the introduction of ECMWF operational ensemble forecasts driven by artificial intelligence and machine learning at the beginning of this month. The ensemble edition of the Artificial Intelligence Forecasting System (AIFS ENS) had already been available as an experimental version for some time, but we have now made it operational. Ensembles are used to provide a range of likely scenarios in our forecasts, so this is an essential step. It completes the operationalisation of the AIFS, which started earlier this year with the AIFS Single, but it is of course not the end of the story. Just as our traditional Integrated Forecasting System (IFS) is continually upgraded, so will the AIFS evolve, notably by increasing its resolution. While the AIFS still uses initial conditions derived in the conventional way, via data assimilation, a more radical approach is to use artificial intelligence to derive forecasts from observations directly. This method, which we reported on in the winter 2024/25 Newsletter, is still being worked on. Some of the news articles in this Newsletter also bear witness to the growing role of machine learning in hybrid numerical weather prediction, with articles detailing methods to combine the best of the physics-based and data-driven methods in a hybrid system, for both forecasting and data assimilation.
Change is also coming to us as an organisation, with the announcement that Florian Pappenberger will take over the role of ECMWF's Director-General from January 2026. As Director of Forecasts and Services and Deputy Director-General, Florian has played a pivotal role in the introduction of machine learning into our forecasts. He has in particular guided the operationalisation of the AIFS, while also continuing to oversee the implementation of new versions of the traditional IFS. The future of ECMWF will be in safe hands when I retire at the end of this year.
More novelties and changes are set out in other articles of this Newsletter. One reports on progress in the use of satellite observations of visible and near-infrared radiances to help establish the initial conditions of forecasts, while another describes the impact of Mode-S aircraft data. These are also examples of the benefits that come from close collaboration with Member and Co-operating States, as some of this work builds on progress made at the German National Meteorological Service (DWD), and Mode-S data benefit from the data processing and distribution performed by the Royal Netherlands Meteorological Institute (KNMI) and the UK Met Office. Our computing services must also move with the times. As the last feature article explains, the continuing increase in data volumes and the potential of cloud computing and machine learning have created new challenges, to which we are responding.
Finally, as he is about to retire, I would like to acknowledge the fantastic work performed by Georg Lentze, the editor of these Newsletters, whose science communication skills have been greatly appreciated at ECMWF.
Florence Rabier
Director-General