One year into the implementation of our 2025 Strategy, a triple anniversary highlights some of the science that plays a key role at ECMWF: 25 years of ensemble prediction, whose future was discussed at our Annual Seminar in September; 20 years of 4D-Var data assimilation, to be celebrated with a symposium on 26 January 2018; and 20 years of seasonal forecasts, which we are marking in style by releasing the new seasonal forecasting system SEAS5 on 5 November 2017.

Advances in these three areas as well as others have enabled substantial improvements in forecasts of severe weather, such as the heat wave that affected southern Europe in August or the tropical cyclones that have hit parts of the Caribbean and the US over the last few weeks. Better modelling of coupled processes, such as ocean–atmosphere interactions, is part of our strategic move towards an Earth system approach. Progress in this area has had a significant impact on the skill of ECMWF’s extended-range forecast. The improvement has been particularly big in the tropics, especially in predicting the Madden–Julian Oscillation (MJO). As a result, the sub-seasonal forecast skill of a wide range of high-impact weather events, including tropical cyclones, has increased. Extended-range forecasting over Europe has also benefited through the impact of the MJO on the North Atlantic Oscillation. Further improvements in Europe are the result of including a sea-ice model.

Ensemble prediction, 4D-Var data assimilation and modelling coupled processes all carry high computational costs. These must be judged against the benefits they bring. As the science progresses, so does our need for more computing power. Our challenge today is to find ways to enable continued advances by our scientists by providing the required computing power whilst respecting legitimate financial and environmental constraints. Our ambitious Scalability Programme helps us in this process by optimising our use of computing resources.

We are only one piece of the puzzle, and it is not our sole responsibility to push the limits of science and technology to improve numerical weather prediction. Many others are involved, and we are working closely with them on the challenges that lie ahead. But a lot is at stake, and experience shows that success comes through each of us bringing our best to the table. The European heat wave of August 2017 was well predicted, and so were the recent tropical cyclones, but there is still room for much improvement, for example in predicting the intensification of hurricanes and severe convection events and in quantifying the uncertainty of our predictions.

Former NOAA Administrator Dr Kathryn Sullivan has described numerical weather prediction as the ability to give humankind foresight. This is an elegant and inspirational way to describe our job. Let’s prove her right and get as close as we can to 20/20 foresight.

Florence Rabier