Our mission is to improve weather forecasting into the medium range, and to do this by tackling those scientific issues that are challenging for individual countries to address on their own. In other words, we were created to embody the strength of a common goal shared by a community.
Since the birth of ECMWF, collaboration with partners has been key, primarily with our Member States and partners in Europe, including satellite agencies EUMETSAT and ESA. A key partnership which has grown over the years is also the one that connects us to the European Union, with whom we operate Copernicus services and projects in the field of climate change, air quality, but also groundbreaking work on the concept of Digital Twins of the Earth in the context of the Destination Earth initiative.
But collaboration extends to the rest of the world. In particular, cooperation with US scientists has largely influenced the development of our activities, a few examples being the development of our numerical model in the 70s-80s, the direct assimilation of satellite radiances, or the development of a global reanalysis of the atmosphere in the 90s.
Although the evolution of the forecasting system has been progressive in the last five decades, there have been some notable breakthroughs, sometimes disruptive, which enhanced the breadth and reach of our activities. A few are introduced below:
- Ensembles: realising that the value of forecasts is much larger if there is a proper estimate of their uncertainties, ensembles have been the method of choice to deliver forecasts, starting in 1992, and culminating in 2023, when the deterministic forecast became just one control forecast member of a global ensemble, at the same resolution.
- Use of satellite observations: building on the theory of optimal control, 4D-Var blends models and time-distributed observations. Combined with the use of satellite radiances rather than retrievals (in collaboration with NCEP), and the advent of advanced satellite data at the end of the 90s, it brought a large improvement in the forecasts, especially in the Southern Hemisphere. Pioneering work continued in the following decades, notably with the increasing use of satellite data in cloudy/rainy conditions, then over all types of surfaces (land, sea-ice...).
- Reanalysis: developed in collaboration with prominent US scientists, it is essential for documenting climate change and also of prime importance to produce re-forecasts and advanced forecast products. Today, with the advent of machine learning in weather prediction, reanalysis plays a critical role as learning basis of such models.
- Earth System: Weather prediction is not any longer based solely on an understanding of the atmosphere. Today, our model runs a full Earth system for weather, which not only includes the atmosphere, but also the land, ocean, sea ice, atmospheric composition, and their interaction with each other.
- Longer time ranges: as the quality of the models improved and they became coupled with other components of the Earth System such as the ocean, ECMWF has gradually tackled longer time-ranges, going from medium to seasonal and then sub-seasonal timescales.
- Hybrid computing architectures: starting with our first supercomputer, the well-known Cray 1, installed in our headquarters in Reading In 1978 (the first Cray in Europe!), supercomputers have steadily increased in power and are now adopting hybrid architectures, with a mix of CPUs and GPUs. This requires us to completely revisit how we write our numerical codes, making them more portable and flexible.
In 2025, it is clear that the needs of society for accurate weather and climate monitoring information have never been greater. Reliable advance warning of weather events is essential for the protection of lives and properties, and environmental monitoring is a crucial input for finding appropriate adaptation and mitigation strategies. It is also evident that Machine Learning (ML) is a revolutionary technology that can be adopted by both big tech companies and organisations like ECMWF and academia.
Fully embraced by the organisation, ML is proving to bring invaluable benefits to our area of work. ML models, once trained, can do a lot of our work faster and cheaper than we normally can with solely physics-based models. We are also very well placed to train such models as we produce the data which is commonly used for such purposes.
Building on our past 50 years, and looking to the future, this is how we can imagine ECMWF in 10 years from now:
- Innovating at the cutting edge of physical, computational and data science for environmental monitoring and prediction.
- Delivering forecast tools and products of the highest possible quality, exploiting data-driven methods anchored on physics-based modelling.
None of this will be possible without scientists and organisations working together, sharing their expertise and data and going beyond egos. Weather and climate do not abide by any man-made boundaries, and science must follow the same rule.
As we are after all the European weather centre, the focus is clearly put on Europe and on our strong partnership with our Member States. However, the collaboration with the rest of the world is certainly thriving, with cooperation agreements with the US, China, Brazil, Australia and Canada. Today I want to specifically mark the strong partnership with the American scientists and institutions which have been part of our fifty years’ history. They have helped both sides advance science for the benefit of society.
It is a fascinating time to be doing the work we do, where nations, continents, public and private sectors work together for the common good.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner