Time series modelling of temperatures: An example from Kefalonia

Abstract

This paper analyses monthly maximum and minimum temperatures from the Greek island of Kefalonia. The data are from the first decade of the twenty-first century and are analysed using a multiplicative seasonal ARIMA model which allows a simple interpretation in terms of adjusting exponentially moving averages of past values. A component decomposition of the data using this model reveals that there has been essentially no trend movement but that the seasonal patterns slowly evolve over time, with the amplitude of the seasonal cycle of maximum temperatures being over 3 {ring operator}C greater than that for minimum temperatures. There appears to have been a break in the mechanism generating the two series in 2011, probably a consequence of shifting the location of the measuring equipment. © 2013 Royal Meteorological Society.