Abstract

Severe damages and large insured losses over Europe related to natural phenomena are mostly caused by extra-tropical cyclones and their related windstorm fields. Thus, an adequate representation of these events in seasonal prediction systems and reliable forecasts up to a season in advance would be of high value for society and economy. In this study, state-of-the-art seasonal forecast prediction systems are analysed (ECMWF, Met Office) regarding the general climatological representation and the seasonal prediction of extra-tropical cyclones and windstorms during the core winter season (DJF) with a lead time of up to four months.

Cyclones are identified using six hourly mean sea level pressure data, whereas wind storms are tracked based on 6 hourly 10m wind speed (ECMWF) and 12 hourly wind speeds in 925hPa (ECMWF & Met Office). Thus, these analyses gain a deeper insight into the representation of extra-tropical cyclones and their related wind storm fields in state-of-the-art seasonal forecasting suites. Spatial distributions of both wind storm and cyclone events show the well-known centres of activity observed in ERA40 & ERA -Interim reanalyses data. However, large biases are found for some areas.

The skill of the different forecast systems is assessed using deterministic as well as probabilistic measures (e.g., Ranked Probability Skill Score (RPSS), Brier Skill Score (BSS), and Anomalous Correlation Coefficient (ACC)). Statistics are carried out for a hemispheric perspective as focussed on several sub-regions (e.g. Northern Europe).

Results reveal small, but positive skill for the North-Atlantic / European region as well as for the North-Pacific region depending on the model suite and variable investigated. Principle reasons for positive skill (predictability) for extreme wind storm frequency are discussed.

Acknowledgments: with special thanks to: Daniel J. Befort, S.B. Wild (Uni Birmingham), A. Weisheimer (ECMWF), J. Roberts, H. Thornton, J. Knight, L. Hermanson, Philip Bett (UK MetOffice)