Changing seasonality in Europe's air temperature

Climate change is expected to involve not only changes in the mean of climate parameters, but also in the characteristics of the corresponding seasonal cycle. However, the discrimination from an observational record of long-term changes in the mean and low-frequency variations in the seasonal pattern is a challenging task, requiring the application of specific statistical methods. In this work, a time series decomposition method based on autoregression is applied in order to obtain a flexible description of seasonal variability from European temperature records. The method is based on the dynamic linear model representation for an autoregressive process and is particularly useful for isolating time-varying cycles in climate time series, allowing to retrieve fluctuations in the amplitude and phase of the periodic components and to assess their statistical significance. This approach is utilised in the analysis of long time series of daily mean temperature from the ECA (European Climate Assessment) project. Seasonality in Europe's air temperature is characterised by an annual cycle with a stable phase but considerable inter-annual and inter-decadal variability. In particular, the annual amplitude was highest in the 1940's and exhibits a distinct minimum around 1975, coincident with the climatic regime shift of the mid-1970's.