Pinning down the predictability of the atmosphere

6 April 2017

ECMWF training course

Weather forecasts invariably come with a degree of uncertainty. The problem for forecasters is to know how big that uncertainty is in a given situation.

Just how predictable are the atmosphere and related parts of the Earth system in a particular place at a particular time? How far into the future can we predict the weather or at least the statistics of weather? How can forecasts reflect this state of affairs? And where does predictability or the lack of it come from?

These questions will be addressed in the last of this year’s season of ECMWF training courses on numerical weather prediction (NWP).

The course, which takes place from 8 to 12 May, unpacks the issues involved in a series of lectures, hands-on exercises and group discussions.

Entitled Predictability and ocean–atmosphere ensemble forecasting, it comes after four other courses on various aspects of NWP aimed at scientists in ECMWF’s Member States and beyond.

These courses have covered advanced numerical methods for Earth system modelling; the parametrization of subgrid physical processes; data assimilation; and satellite data assimilation.

They have been attended by over a hundred participants from 23 countries.

About half of the Research Department along with members of the Forecast Department contribute to the NWP training by giving lectures and practical sessions.

“One of the most enjoyable and rewarding aspects of the task is the interaction with participants,” says course organiser Sarah Keeley.

“Every year we receive feedback that topics that were once a mystery have been taught so well that they are now well understood. It reflects the time, energy and enthusiasm that ECMWF staff put into preparing the training material.”

From the ocean to the stratosphere

The predictability course covers topics ranging from coupled ocean–atmosphere–land processes to probabilistic forecasts and their evaluation, including monthly and seasonal forecasting.

View of the upper atmosphere with ocean, land and clouds

The course emphasises the role of interactions between different parts of the Earth system in our forecasts, from the ocean to the stratosphere. (Photo: MarcelC/iStock/Thinkstock)

“This year we have had such a high demand for the Predictability course, we have moved the training to the Lecture Theatre so that we can accommodate up to 60 participants rather than 32,” Dr Keeley notes.

“We have guest lectures by Professors Bart van den Hurk from the Dutch national meteorological service, KNMI, and Andrew Charlton-Perez from the University of Reading on the importance of the land surface and the stratosphere, respectively.”

The course also looks into some of the factors that limit predictability in weather forecasting.

One of these is that the atmosphere is a chaotic system.

As a result, small errors in the estimate of the initial state on which forecasts are based can quickly lead to large errors in the forecast.

 “We discuss the underlying theory and current research being done to develop forecast systems used to provide forecasts from the medium range right out to the seasonal time scale,” Dr Keeley says.

“The aim of the course is to understand why we need to consider all forecasts in a probabilistic sense, based on the fact that the atmosphere is chaotic.”