An analysis of the fluctuations in the long-term annual mean air temperature data of Turkey

This paper presents a time series analysis of fluctuations in long-term annual mean air temperature series of Turkey, which were tested for non-randomness by the autocorrelation function and subjected to power spectrum methods. Power spectrum estimates are computed for the purpose of finding evidence of a cyclic component within the series. The analysis covered the period of 1930--1995 and included 71 stationsin seven geographical regions. For most of the series, the correlograms revealed no apparent patterns of cyclicity. The arithmetic average of the positive correlation values is about 11 years, and that corresponds well with the sun-spot cycle effect on the temperature values. A further analysis of the series by power spectrum indicated littleevidence for statistically significant differences from the white-noise spectrum. In most cases the spectral densities above the white-noise level tended to peak at frequencies corresponding to 2.4--3.1 and3.3--4.3 years, being particularly pronounced in the east, Mediterranean, central, and south eastern Anatolia regions. The 3.3--4.3-year peaks in the central and south-eastern Anatolia regions were significantly different from white noise at the 95 per cent level. Roughly two groups of regions with different spectral shapes are identified. The coastal regions almost exhibited a white-noise spectrum whereas theinland regions tended to approximate a grey-noise spectrum. [ABSTRACT FROM AUTHOR]