The return period of wind storms over Europe

• Published in: International Journal of Climatology
• Main Authors: Della‐Marta, Paul M., Mathis, Hubert, Frei, Christoph, Liniger, Mark A., Kleinn, Jan, Appenzeller, Christof
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2008
• Subjects: North Atlantic
• Online Access: http://dx.doi.org/10.1002/joc.1794; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002%2Fjoc.1794; https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/joc.1794

Abstract Accurate assessment of the magnitude and frequency of extreme wind speed is of fundamental importance for many safety, engineering and reinsurance applications. We utilize the spatial and temporal consistency of the European Centre for Medium Range Forecasts ERA‐40 reanalysis data to determine the frequency of extreme winds associated with wind storms over the eastern North Atlantic and Europe. Two parameters are investigated: 10‐m wind gust and 10‐m wind speed. The analysis follows two different view‐points: In a spatially distributed view, wind‐storm statistics are determined individually at each grid‐point. In an integral, more generalized view, the wind‐storm statistics are determined from extreme wind indices (EWI) that summarize storm magnitude and spatial extent. We apply classical peak over threshold (POT) extreme value analysis techniques (EVA) to the EWI and grid‐point wind data. As a reference, a catalogue of the 200 most prominent European storms has been compiled based on available literature. The EWI‐based return periods (RP) estimates of catalogue wind storms range from approximately 0.1 to 300 years, whereas grid‐point‐based RP estimates range from 0.1 to 500+ years. EWIs sensitive to the absolute magnitude of wind speed rank the RP of wind storms in the 1989/1990 and 1999/2000 extended winter season similarly to the RP derived from the distributed approach. The RP estimates derived from EWIs are generally higher when calculated using only land grid‐points compared to RPs derived using whole domain. Both the uncertainties in EWIs and grid‐point‐based RPs show a greater dependence on the wind parameter used than on the uncertainty associated with the EVA for RPs less than 10 years, whereas for RPs greater than 10 years the effect of the different datasets is lower. The EWIs share up to approximately 50% of the variability of the local grid‐point RPs. Copyright © 2008 Royal Meteorological Society