Sunspots and Temperatures

Power spectrum analyses of temperature series at New York City (1822-1956), the Netherlands (1735-1944), and central England (1698-1955) show that the only significant energy in these spectrums appears at the annual period of 12 months. No significant energy was found corresponding to any known sunspot periodicity. Correlation coefficients between January and July temperatures for the years of each series are uniformly low, ranging from 0.09 to 0.17 for the three series. This indicates a lack of correlation between winter and summer temperatures contrary to what might be expected if temperature variations were controlled to some extent by sunspot variations. It is concluded that no demonstrable connection exists between midlatitude temperatures and sunspots. Effects depending on temperature variations, such as sea level variations, are also independent of sunspot variations. Introduction. There have been many suggestions through the years of a possible relationship between sunspots, air temperature, and sea level [e.g. Willerr, 1949; Landsberg et al, 1959; Fairbridge, 1963]. To test one aspect of this hypothesis, long temperature records in central England from 1698 to 1955 [Manley, 1953], the Netherlands from 1735 to 1944 [Labrijn, 1945], and New York City from 1822 to 1956 [U.S. Dept. Commerce, 1958] were subjected to power spectrum analysis using a hamming spectral window by methods described by Blackman and Tukey [1958]. Schove's [1955] sunspot data for 1700 to 1953 were also subjected to power spectrum analysis. Although the relationships have been postulated for the tropics, mid-latitude data were chosen here for two reasons: (1) the only temperature series sufficiently long to resolve an l 1year cycle by spectrum analysis are in mid-latitudes, whereas temperature records for the tropics rarely extend for 50 years; (2) if the x Contribution 846, Lamont Geological Observatory. sunspot-temperature effect is real, it ought to be observable at any latitude but would tend to be masked by the midlatitude cyclone noise. However, power spectrum analysis, on a sufficiently long series of temperatures, would detect the relevant periodicities. Data preparation. The original temperature data are in the form of monthly means. In addition to the raw monthly values of temperature, 12-month running means were calculated and analyzed. This procedure has the effect of completely suppressing the 12-month periodicity and reducing the amplitude of all shorter periods to almost zero while leaving the amplitude of the longer-period constituents virtually unchanged. For exa.mple, at a period of 11 years the amplitude of a spectral component is only reduced by 4%. By using the 12-month running means I avoid contaminating the longer-period part of the spectrums with side lobe energy from the predominant 12-month periodicity. 'Prewhitening' of the data, which in this case was accomplished by the use of these running means, was used to reduce the energy appearing in the side lobes.