New Strategy is “ambitious but not unrealistic”

ECMWF’s Director of Research Erland KällénOn 30 June, ECMWF’s Council approved the Centre’s new 10-year Strategy, which covers the period up to 2025. In an interview with the Newsletter, ECMWF’s Director of Research Erland Källén explains how the Strategy was drawn up, why collaboration lies at its heart, and what needs to happen to turn the “ambitious” goals it sets into reality.

What is the purpose of ECMWF’s Strategy?

The purpose is to set our direction of travel for the next ten years. I think it is extremely important for an organisation like ECMWF to have a clear goal in mind that everybody understands. An important reason for our success in the past has been that all members of staff as well as our key stakeholders, the national meteorological services, are clear about this goal. Having such a focus is extremely motivating.

This is a ten-year Strategy, yet a new one is drawn up every five years. Why?

This is because we want to take into account things that are happening around us. In ten years a lot of things can happen. There may be significant scientific or technological developments, such as new supercomputer architectures which require highly scalable computer code. Therefore we have to be able to adjust the Strategy in shorter intervals than ten years. On the other hand, you need the ten years to set a broad direction of travel.

How was the Strategy drawn up?

It started with an internal consultation process within ECMWF. We have a lot of in-house expertise to tap into, and we have a scientific vision of where we want to go. In order to remain in a leading position, I think it is important for the impetus for our goals to come from inside.

But of course it is equally important for our Member and Co-operating States to be involved in the whole process to make sure that what we do meets their requirements. So the second step was to have intense consultations with all our Member and Co-operating States. The Directors of ECMWF travelled to most of our Member States and presented the results of the internal consultation to their national meteorological services to get their opinion.

After that we drew up a first draft version of the Strategy, which we then also submitted to our Committees. The Committees represent both our Member States’ meteorological services and scientific specialists in relevant areas of science. So this consultation mechanism was also very important.

Finally, the input of space agencies was very important, especially of EUMETSAT, our main strategic partner in the area of space observations, and they were also involved in the consultation process.

“A very ambitious goal is to say that we will be able to predict high-impact weather events up to two weeks ahead.”

Can you give an example of how the consultations influenced the outcome?

The emphasis on ensemble prediction is a good example. It was our intention to focus our attention on the ensemble method, on probabilistic forecasting, to produce and present forecasts. When we travelled to our Member States, we found that this approach resonated very well with them: everybody was on board.

This was by no means clear before the consultation process because many users still regard the high-resolution or deterministic forecast as ECMWF’s prime product. We also show the Council our results in terms of deterministic forecast scores. But during the consultation process it became very clear to us that we should focus on the ensemble and that our scores and other follow-up mechanisms should reflect that better than today.

What are the main goals set by the Strategy?

A very ambitious goal is to say that we will be able to predict high-impact weather events up to two weeks ahead. I don’t think anyone else has dared to state this as clearly as we have done in our Strategy. Of course, going out into the future as far as possible has always been the objective, but I think what is new is to say that we will be able to do this in the probabilistic sense up to two weeks ahead.

Another goal is to predict large-scale patterns and regime transitions up to four weeks ahead, and global-scale anomalies up to a year ahead. The meteorological community is putting more and more emphasis on the monthly timescale, and we believe we can meet that challenge and make much better probabilistic predictions on this timescale.

In terms of resolution upgrades, the Strategy specifies the target of a 5 km ensemble by 2025. That is ambitious, too. It is at the limit both in terms of what we can do scientifically and in terms of the computing capacity we envisage having in 2025.

These goals are ambitious but not unrealistic.

"We cannot do all the necessary research ourselves, it has to be supported and embraced by the scientific community in our Member States as well as worldwide.”

How can they be achieved?

An important element is collaboration with our Member and Co-operating States. In the Strategy we have very much stressed the collaborative effort that is needed to reach these goals. We cannot do all the necessary research ourselves, it has to be supported and embraced by the scientific community in our Member States as well as worldwide.

The other collaborative aspect concerns weather observations, which we rely on but don’t do ourselves. The Strategy says that we will continue to rely on the space programmes being planned with EUMETSAT, which is a very important partner organisation for us.

“New information from satellite data will be important for achieving our strategic goals.”

The collaboration with ESA is also important, and then of course we also want to influence the American, Japanese and Chinese satellite programmes to make sure that the satellite data we get in the future will be useful to us.

Today more than 95% of all the observations we get come from satellites. New information from satellite data will be important for achieving our strategic goals.

What will have to happen in research and development?

In the area of research, there are three main goals. One is the fact that we want to use and further develop the ensemble method in all aspects of the forecasting system. This is necessary to achieve the goal of predicting severe weather up to two weeks ahead.

A second goal is improved Earth system modelling. We need to consider more parts of the Earth system, and the ones that we have included need to be better described. The description of clouds and radiation will have to be improved, and we believe that a better description of atmospheric composition will also lead to better weather forecasts.

A third main area is scalability, in other words adapting the forecasting system to the changing architectures of high-performance computers. Higher resolution is essential to improve the accuracy of the initial state and of our forecasts, but we need greater computing power to achieve it. That means we also need to develop our software to be able to use that kind of computing power. If we fail to deliver on scalability, we will not be able to increase the resolution at the required pace. So scalability is an absolutely essential part of our research programme. 

What needs to change in the area of Earth system modelling?

If you look at weather prediction in historical perspective, it started out 60 years ago with scientists only looking at the mean flow in the atmosphere at a height of five kilometres, disregarding all the sub-grid scale physics, things such as radiation and clouds. When clouds and radiation began to be taken into account, this was still with a very primitive description of the Earth’s surface.

Today we have a better description of the thermodynamics and dynamics of the atmosphere and we have a better coupling with the land surface. But there are still a lot of things we need to develop in sub-grid scale physics. One example is clouds and radiation, another is the exchange of moisture between the surface and the atmosphere.

Ocean coupling will also have to improve. In the past, we have regarded the oceans on the timescales relevant to medium-range prediction as completely static, which they are not. They are part of the dynamics of the whole system. And we have to get better at including the oceans in the data assimilation and at representing atmosphere–ocean couplings, including sea ice, at all time ranges. We have come to realise that, even just a couple of days ahead, the interaction between the atmosphere and the ocean in the tropics makes a big difference, and we have to get better at including that.

“We need to consider more parts of the Earth system, and the ones that we have included need to be better described.”

In addition to scalability, what are the plans for high-performance computing?

The Strategy says that we want to be a centre for high-performance computing in meteorology. We want to continue to make it possible for Member States to use computing capacity on our machines, including for the optional provision of boundary conditions for limited-area modelling.

For future high-performance computing, the Strategy makes it clear that environmental sustainability is an important goal. Using computers in an environmentally sustainable manner means both making them more energy-efficient and making sure that the energy is generated in an environmentally sustainable way.

What role does environmental prediction play in the Strategy?

The EU-funded Copernicus Earth observation programme is now a large part of our organisation, in the form of the Atmosphere Monitoring Service and the Climate Change Service. We also contribute to the Emergency Management Service through flood computations and there might be scope for further development, taking on third-party activities that complement what we do.

Importantly, these activities should be in line with our core goals, with the Centre’s Convention. I have already mentioned the link between atmospheric composition and weather forecasting. Climate reanalysis is also extremely important to weather prediction, for example as a source of verification, initialisation and calibration.

Outside ECMWF, scientists use reanalysis as a dataset to improve our knowledge of atmospheric processes. We can in turn use the insights gained to improve our forecasting system.

How important will training and education be?

In accordance with our Convention, we should continue to be a centre for advanced training and education in NWP. We want to continue to do that by providing courses and software, such as the OpenIFS. A new element in the Strategy is that we want to deepen training collaboration with our Member States.

This is important because ECMWF has unique collective expertise in NWP. We should take advantage of that by providing appropriate training and education. We also want to be able to recruit people in the future who can continue to develop our models, and scientists and others in the Member States are our recruitment base.

What does the Strategy say about funding?

There will continue to be a diverse set of funding sources to support ECMWF. The Member States and their contributions will continue to provide the core of our funding, but we also have to look for other types of funding, such as research funding from the European Commission, funding by satellite agencies, and other types of third-party funding.