Long-term predictability of mean daily temperature data

Abstract. We quantify the long-term predictability of global mean daily temperature data by means of the Rényi entropy of second order K₂. We are interested in the yearly amplitude fluctuations of the temperature. Hence, the data are low-pass filtered. The obtained oscillatory signal has a more or less constant frequency, depending on the geographical coordinates, but its amplitude fluctuates irregularly. Our estimate of K₂ quantifies the complexity of these amplitude fluctuations. We compare the results obtained for the CRU data set (interpolated measured temperature in the years 1901-2003 with 0.5° resolution, Mitchell et al., 2005)with the ones obtained for the temperature data from a coupled ocean-atmosphere global circulation model (AOGCM, calculated at DKRZ). Furthermore, we compare the results obtained by means of K₂ with the linear variance of the temperature data.