Limits of predictive skill in global, medium-range weather forecasting

By Prof. Erland Källén, Director of Research, ECMWF

Medium-range weather prediction skill has steadily increased over the past few decades. There are two main factors contributing to the improved skill level: A better accuracy of the initial state and an improved accuracy of the prediction model. The relative importance of these has been assessed and it is concluded that they contribute about equally to the skill improvement.

The initial state accuracy is both determined by the available observations as well as the assimilation method used to calculate initial states. For future improvements it will be necessary to continue improving both assimilation methods and model accuracy. In some areas also better observations are required, this is particularly true over tropical regions where better wind observations are needed.

The Aeolus satellite, planned to be launched by ESA in 2017, has the potential to provide these observations. Impact studies have shown that wind observations from Aeolus can give a substantial improvement in medium-range forecast quality, in particular for flow situations where initial state errors in the tropics develop into large forecast errors in mid-latitudes.