Extreme value and cluster analysis of European daily temperature series

Abstract

Time series of daily mean temperature obtained from the European Climate Assessment data set is analyzed with respect to their extremal properties. A time-series clustering approach which combines Bayesian methodology, extreme value theory and classification techniques is adopted for the analysis of the regional variability of temperature extremes. The daily mean temperature records are clustered on the basis of their corresponding predictive distributions for 25-, 50- and 100-year return values. The results of the cluster analysis show a clear distinction between the highest altitude stations, for which the return values are lowest, and the remaining stations. Furthermore, a clear distinction is also found between the northernmost stations in Scandinavia and the stations in central and southern Europe. This spatial structure of the return period distributions for 25-, 50- and 100-years seems to be consistent with projected changes in the variability of temperature extremes over Europe pointing to a different behavior in central Europe than in northern Europe and the Mediterranean area, possibly related to the effect of soil moisture and land-atmosphere coupling.

Keywords:

• daily mean temperature series
• cluster analysis
• Bayesian inference
• return values

Acknowledgements

The authors would like to express their gratitude to two anonymous referees for all helpful comments and constructive criticism. The three authors are supported by ‘Acções Integradas Luso-Espanholas’ under the grants E-83/09 and HP2008-008. The third author also acknowledges the support of grant SEJ2007-64500.