2.2 Solar signals in surface climate

Many different approaches have been adopted in the attempt to identify solar signals in climate records. Probably the simplest has been the type of spectral analysis mentioned above, in which cycles of 11 (or 22 or 90 etc.) years are assumed to be associated with the Sun. In another approach time series of observational data are correlated with time series of solar activity. An extension of the latter method uses simple linear regression to extract the response in the measured parameter to a chosen solar activity forcing. A further development allows a multiple regression, in which the responses to other factors are simultaneously extracted along with the solar influence. More sophisticated statistical techniques, involving e.g. pattern-matching, have also been employed. Each of these approaches gives more faith than the previous that the signal extracted is actually due to the Sun, and not to either some other factor or to random fluctuations in the climate system, and many interesting results emerge. It should be remembered, however, that such detection, while potentially robust in statistical terms, is not based on any understanding of how the presumed solar influence takes place. In the remainder of this section some results are presented from studies involving correlations and regressions of meteorological data with solar activity indices. Some of the mechanisms which have been proposed to explain how these changes take place are the subjects of Sections 3 to 7.

Section 2.1 mentioned how temperature records may be extracted from ice cores and ocean sediments. These media may also preserve information on cosmic ray flux, and thus solar activity, in isotopes such as 10Be and 14C. Thus simultaneous records of climate and solar activity may be retrieved. An example is given in Figure 4View Image which shows fluctuations on the 1,000 year timescale well correlated between the two records, suggesting a long-term solar influence on climate.

View Image

Figure 4: Records of 10Be and ice-rafted minerals extracted from ocean sediments in the North Atlantic. From Bond et al. (2001).

On somewhat shorter timescales it has frequently been remarked that the Maunder Minimum in sunspot numbers in the second half of the seventeenth century coincided with what has become known as the “Little Ice Age” during which western Europe experienced significantly cooler temperatures. Figure 5View Image shows this in terms of winter temperatures measured in London and Paris compared with the 14C ratio found in tree rings for the same dates. Similar cooling has not, however, been found in temperature records for the same period across the globe which (if this signal is reliable) suggests that the Sun’s influence may be geographically non-uniform.

View Image

Figure 5: From a paper by Eddy (1976) suggesting that winter temperatures in NW Europe are correlated with solar activity. Note the coincidence of the “Little Ice Age” with the Maunder Minimum in sunspots.

A paper published by Friis-Christensen and Lassen (1991Jump To The Next Citation Point) caused considerable interest when it appeared to show that temperature variations over the observational period could be ascribed entirely to solar variability. The measure of solar activity used was the length of the solar cycle (SCL) and, as can be seen in Figure 6View Image, this value coincides almost exactly with the Northern Hemisphere land surface temperature record. The SCL values were, however, smoothed with a (1, 2, 2, 2, 1) running filter so that at each point the value given has contributions from four solar cycles both forwards and back in time, i.e. a total of ∼ 50 years. This means that the values given for dates more recent than about 1958 required some extrapolation, and the more recent the more the result depended on assumptions about future behaviour. Thus the upturn in SCL values in the latter part of the twentieth century was essentially construed. This work has been repeated by Laut and Gundermann (2000Jump To The Next Citation Point), using longer records, as shown in Figure 7View Image. There is still a correspondence between the two records in the first half of the twentieth century but the potential for the Sun to be the cause of more recent warming looks far less convincing. The detection and attribution of the causes of twentieth century climate change is discussed further in Section 6.1.

View Image

Figure 6: Northern Hemisphere land temperature, (stars) and solar cycle length (inverted, pluses). Both time series are smoothed using (1 2 2 2 1) filter weightings. From Friis-Christensen and Lassen (1991).
View Image

Figure 7: Northern Hemisphere temperatures (Mann et al., 1999) (thin line) and solar cycle length (inverted, dots and thick line), smoothed as in Figure 6View Image. From Laut and Gundermann (2000).

Many studies have purported to show variations in meteorological parameters in phase with the “11-year” solar activity cycle. Some of these are statistically questionable and some show signals that appear over a certain period only to disappear, or even reverse, over another period. The evidence of a solar influence on climate on solar cycle timescales is, however, becoming increasingly robust.

Solar signals have been detected in sea surface temperatures (SSTs). Figure 8View Image presents some results from an EOF analysis of upper ocean temperatures, from bathythermographs 1955 – 1994, with the upper panel showing the time varying amplitude of the pattern of response shown in the lower panel. Two interesting features emerge from this work, one is that SSTs do not warm uniformly in response to enhanced solar activity: indeed, the pattern shows latitudinal bands of warming and cooling, and secondly that the amplitude of the change is larger than would be predicted by radiative considerations alone, given the known variations in TSI over the same period (see Section 3.2).

View Image

Figure 8: Below: pattern of response in sea surface temperatures derived for solar variability. Above: time series of the magnitude of this pattern. From White et al. (1997Jump To The Next Citation Point).

An analysis of a different SST dataset, dating back to the mid-19th century, is shown for the Pacific Ocean in Figure 9View Image. A composite of data from peak years of solar activity cycles has been constructed and the composite of the remaining years subtracted. The results also suggest considerable geographical variation with a mid-latitude pattern resembling the negative phase of the Pacific Decadal Oscillation (PDO) and a tropical signal similar to the negative phase of the El Niño-Southern Oscillation (ENSO). However, the signals derived in Figure 8View Image and Figure 9View Image appear to be in the opposite sense and it remains to be seen which, if either, is a true response to the Sun.

View Image

Figure 9: Anomalies of sea surface temperature in peak years of the solar cycle. From van Loon et al. (2007).

These results are suggestive that the Sun may exert some influence on the phases of various climate “modes of variability”. Of these ENSO dominates the tropics and the PDO the North Pacific. The North Atlantic exhibits an oscillation in surface pressure referred to as the North Atlantic Oscillation (NAO) which has a strong influence on the climate of Western and Northern Europe. Correlations have been demonstrated between the NAO and the electric field strength of the solar wind (Boberg and Lundstedt, 2002) and between the NAO and solar 10.7 cm radio flux (Haigh and Roscoe, 2006Jump To The Next Citation Point). These studies imply that Western Europe experiences warmer and wetter winter weather when the Sun is more active, with more and stronger storms crossing the Atlantic on a more northerly track. For the Northern and Southern Annular Modes (NAM and SAM, representing the main modes of variability in high latitudes) Haigh and Roscoe (2006Jump To The Next Citation Point) found that the combined influence of the Sun and the Quasi-Biennial Oscillation (QBO) of tropical stratospheric winds provided a better indicator, than solar influence alone, of the strength of the circumpolar circulation in both hemispheres (see Table 2). Further discussion of links with the QBO and possible reasons for these are presented in Sections 2.3 and  6.2 respectively.



Table 2: Multiple regression analysis of the surface Northern Annular Mode (NAM) (in winter) and Southern Annular Mode (SAM). The annular modes are the first empirical orthogonal functions of 90-day low-pass filtered anomalies, poleward of 20° in each hemisphere, of 1000 hPa geopotential height in normalised units. Positive values indicate a stronger equator-to-pole temperature gradient and more zonal flow. Columns show regression coefficients for linear trend (N.H. only), stratospheric chlorine (S.H. only), El Niño-Southern oscillation (ENSO), volcanic (stratospheric) aerosol loading, solar variability (10.7 cm index) and the Quasi-Biennial Oscillation (QBO). Sol*QBO indicates that an index composed of a product of the solar and QBO indices was used in place of those two factors individually. The data cover the period 1958 – 2001. Colours indicate the statistical significance levels of the values, derived using a Student’s t-test: 99%, 95%, 90%, 80%, < 80%. From Haigh and Roscoe (2006).








Trend Cl ENSO Vol Sol QBO Sol*QBO








NAM (DJFM) 0.23 –0.82 1.08 0.81 0.64
0.24 –0.85 0.96 –1.16
SAM 1.01 –0.82 –0.65 –0.04 0.13
1.01 –0.71 –0.75 –0.89









Another study of correlations between solar activity and climate was carried out by Marsh and Svensmark (2000Jump To The Next Citation Point) who showed that between 1983 and 1994 low latitude low cloud cover varied in phase with galactic cosmic ray (GCR) intensity. An update to this (see Figure 10View Image) found the correlation continued but only if assumptions where made with regard to temporal variations in the calibration of the satellite data used in constructing the cloud dataset. These have still to be validated. Physical mechanisms which might induce a causal relationship between cloud cover and cosmic rays are discussed in Section 7 but it is worth noting here that, if there is a link between cloud and solar activity, then this result alone does not show that it is due to GCRs any more than any other solar- modulated input to the Earth. Indeed, work by Kristjansson and Kristiansen (2000) suggests that there is a stronger correlation between cloud cover and solar UV radiation.

View Image

Figure 10: Variation of low cloud cover (ISCCP-D2 data) and cosmic rays between 1984 and 2002. The green curve shows data obtained by applying assumed satellite recalibrations. Redrawn from Marsh and Svensmark (2003) by Gray et al. (2005).
View Image

Figure 11: Time series of annual mean 30 hPa geopotential height (km) at 30° N, 150° W (thin line with circles), its three-year running mean (thick line with circles) and the solar 10.7 cm flux (dashed line with squares). From Labitzke and van Loon (1995).

  Go to previous page Go up Go to next page