Professor of Atmospheric Dynamics Heini Wernli from the ETH Zurich says studying the detailed physical processes involved in severe weather events can help us understand why some forecasts go wrong.
A key moment in his scientific career was the windstorm Lothar in December 1999, which caused huge damage in Switzerland as well as other parts of Europe.
Heini Wernli, who was appointed to a three-year ECMWF Fellowship earlier this year, found the storm to defy textbook explanations. Intrigued, his team went on to develop a framework which can help forecasters make sense of this kind of event.
More recently, he and his research group have focussed on the organised movement of warm moist air in extratropical cyclones called a ‘warm conveyor belt’.
Why is it that forecasts involving warm conveyor belts sometimes go wrong? How can we improve our understanding of the microphysical cloud processes in warm conveyor belts and of their interaction with the larger-scale flow?
Heini Wernli says exploring these questions with ECMWF scientists should benefit not just the Centre but also his research group, who will be able to work with “the best model in the world” and with colleagues who understand that model better than anybody else.
How did you come to be a meteorologist specialising in atmospheric dynamics?
I studied physics at the ETH Zurich and chose atmospheric physics as an option. I immediately felt this was a great subject and went on to do my Masters and my PhD in this area. During the first week of my PhD I attended a training course at ECMWF, which was the starting point for developing strong bonds with the Centre.
During my PhD and throughout my career, working with ECMWF data has been absolutely essential. I am interested in how weather works, why a particular storm gets dangerous while other extratropical cyclones just bring some rain and bad weather.
To figure this out, it is useful to work with real data. ECMWF’s climate reanalysis has proved a treasure chest in that regard – first ERA-40 and then ERA-Interim. We can do so many things with it. Today, whenever somebody in my research group has an idea, they say: ‘It would be good to test that with ERA-Interim.’ That just shows you how much we work with the Centre’s data.
What are your expectations of the Fellowship programme?
I was extremely honoured to be nominated as a Fellow. I see it both as an honour for what our research group has been doing over the last few years, but also as a great opportunity for our group to strengthen our cooperation with ECMWF experts. I think that is really important.
Over the last two or three years, several of my postdocs have also come here for workshops and other events, so we have a number of linkages. Being able to work with ECMWF is also extremely valuable to younger colleagues in my group. Conversely, some people at ECMWF are interested in our diagnostics.
Thanks to the Fellowship, my group will be much closer to more people here at ECMWF. We will benefit from a better understanding of what the driving questions are and how our way of looking at weather and model error can help.
Your research group, which comprises nearly 20 scientists, aims to improve our understanding of high-impact weather systems. Can you give an example?
A storm that enormously influenced my scientific career was the Lothar storm in December 1999. To many people in Switzerland it felt like a once-in-a-lifetime event. There was huge destruction but people felt unprepared and asked: what can we do about it?
It was fascinating that most of what I had learnt from lectures and textbooks did not work for that storm. It did not have an upper-level trough and there was no disturbance at the tropopause. The cyclone remained small as it moved extremely rapidly from Newfoundland towards central Europe. The sea level pressure drop was not dramatic. And still the damage was enormous.
So this was something new. We looked at some theoretical studies and further developed the concept of diabatic Rossby waves, which helped us to understand that particular storm as well as similar ones.
Has it been possible to translate this improved understanding into forecast model development?
Doing that is not straightforward. In this particular case the problem did not so much lie with the models. For example, 20% of ensemble members in ECMWF forecasts predicted a storm that was even worse than what happened.
Instead, there was a problem with forecasters interpreting the model output: if you see something in a model that is nothing like what you have learnt about the behaviour of cyclones, then maybe you do not believe it.
But if you know that this kind of event has occurred before and that some people have a theoretical understanding of it, then you are more likely to take it seriously as a forecaster.
When it comes to improving forecasting models, one of the main goals of my research group at the moment is to try to understand how the latent heating in clouds due to condensation and freezing influences the dynamics.
Hanna Joos from my group is collaborating with Richard Forbes from ECMWF on analysing the microphysics in so-called warm conveyor belts, air streams in extratropical cyclones in which a lot of cloud processes occur. Using the concept of potential vorticity, they are trying to understand the impact on forecasts when different versions of ECMWF’s cloud microphysics scheme are applied to this kind of situation.
This is interesting for us as an application of the more theoretical concept of potential vorticity, and it is interesting for Richard to get a new perspective on his improvements in the microphysics.
What is a warm conveyor belt and why is it important?
Extratropical cyclones are usually associated with cloud bands produced by warm moist air moving poleward. As the air is pushed poleward, the temperature gradient forces it to ascend, giving rise to a lot of condensation, cloud formation and rainfall. This coherent movement of rising warm moist air is called a warm conveyor belt.
What we have found out over the last few years is that when this conveyer belt arrives at 10 km altitude, near 60° north, it kind of hits the jet stream. Cloud systems such as the conveyor belt can thus have a direct impact on the dynamics at the tropopause, which is directly relevant for medium-range weather prediction.
In the conveyor belt, we have a very good understanding of the phase transition from water vapour to liquid, but there are a lot of unknowns related to, for instance, the temperature at which the liquid droplets start to freeze, and whether snow is produced or tiny ice particles that are lifted higher. For example, in the presence of one kind of aerosol the cloud droplets may freeze earlier than in the presence of another kind of aerosol, and so on.
This is tricky, and models represent these processes in a simplified and approximate way. We are now interested in how these uncertainties influence what happens when the warm conveyor belt arrives at upper levels, which in turn has an influence on the weather a few days down the line.
Can this kind of work drive further changes in the schemes that represent such processes in forecasting models, the so-called parametrization schemes?
I think this kind of work can be used to test different options in parametrization schemes. We know that in several cases of poor forecasts warm conveyor belts have been involved.
Colleagues at ECMWF have ideas on rerunning these cases with modified parameters in the physics scheme. And then we can use our tools, our trajectory packages and so on, to compare and analyse the different simulations.
What are the limitations of this approach in terms of model improvement?
I worked on warm conveyor belts for my PhD, but at the time I did not realise that they were important for forecast quality. About ten years ago, I was looking at the ten worst forecasts that I could find, and I could see that warm conveyor belts were involved. That brought me back to this research topic.
We then tried to develop an objective measure of how well a conveyor belt is represented in a forecast. This is a bit complicated because a conveyor belt is not just a two-dimensional field. Our approach looks at air parcel trajectories.
Using this method, we could show that the quality of warm conveyor belts in ECMWF forecasts has improved over the last ten years, and of course we can also see that in some forecasts the errors are large. And when the errors in the conveyor belt structure are large, then everything else in the forecast is not that good, either.
The next question is: what should we change in the model to improve that? And here we come up against a limitation, or a challenge. By just doing verification and diagnostics, you can tell whether one model or model version is better than another. But it is much easier to say whether something is good or bad than to also give an indication of why it is good or bad.
What benefits do you expect the ECMWF Fellowship to bring you and your group?
Working with ECMWF allows us to analyse numerical forecasts together with scientists at the Centre. Their expertise allows them to pick the right experiments, and they know where it makes sense to change something.
So it is really a win-win situation, for ECMWF because we can – hopefully – offer an additional layer of understanding, and for us because we can work with the best model in the world and with people who understand that model better than anybody else.