This has been a long time coming. The EarthCARE satellite, a joint venture between the European Space Agency (ESA) and the Japan Aerospace Exploration Agency (JAXA), was launched in May 2024. Its four instruments provide new insights into the interactions between clouds, aerosols, precipitation and radiation, as well as helping to establish the initial conditions of weather forecasts. EarthCARE is part of the Earth Explorer missions, which address key scientific challenges identified by the Earth observation community. For us, there are two areas which are important: the first is new data from which to infer the properties of clouds and how they interact with solar and thermal-infrared radiation. The results can be used to improve how cloud processes are represented in weather models. The second is the assimilation of the satellite’s data into our Integrated Forecasting System (IFS) to improve the initial conditions of weather forecasts. This is exciting as it will be the first time that radar and lidar data are assimilated operationally in a global data assimilation system. An article in this Newsletter provides an overview of our use of EarthCARE data, and it reports on first positive results in data assimilation. Our forecasts are expected to benefit from the data as early as this year.
Another development highlighted in this Newsletter is the operational release of data-driven forecasting. Our Artificial Intelligence Forecasting System (AIFS) became operational on 25 February with the AIFS Single 1.0 version. It produces a single forecast, and it will be complemented later this year with an AIFS ensemble system. The AIFS still depends on traditional weather forecasting because it uses ECMWF’s ERA5 reanalysis dataset as well as the IFS operational analysis for training purposes and for initialisation. However, on that basis, starting from the operational analysis, it produces a forecast solely by means of machine-learning (ML) methods. And it does so very successfully because it achieves better results than traditional forecasts in a number of headline scores. The AIFS Single 1.0 is the first operational ML weather forecasting model. It has a greater number of forecast variables than other data-driven systems.
The growing influence of AI and ML methods has also resulted in ECMWF collaborating with the Swiss National Supercomputing Centre (CSCS) and MeteoSwiss to establish efficient access to our forecast products and archive to train AI and ML models. An article in this Newsletter describes how the project creates new opportunities for research in meteorology and scientific computing. There is also a look at the next upgrade of the IFS to Cycle 50r1 later this year, with an article on reintroducing an analysis of humidity in the stratosphere. The article demonstrates that this step will improve forecasts of humidity and temperature across all lead times, in particular in the Upper Troposphere Lower Stratosphere (UTLS) layer, but improvements can also be seen further down in the troposphere.
Florence Rabier
Director-General