Svensmark and Friis-Christensen (1997) demonstrated a high degree of correlation between total cloud cover, from the ISCCP C2 data set, and cosmic ray flux between 1984 and 1991. Changes in the heliosphere arising from fluctuations in the Sun's magnetic field mean that galactic cosmic rays (GCRs) are less able to reach the Earth when the Sun is more active so the cosmic ray flux is inversely related to solar activity. Svensmark and Friis-Christensen analysed monthly mean data of total cloud using only data over the oceans between 60°S and 60°N from geostationary satellites. They found an increase in cloudiness of 3 to 4% from solar maximum to minimum and speculated that (a) increased GCR flux causes an increase in total cloud and that (b) the increase in total cloud causes a cooling of climate. Svensmark and Friis-Christensen (1997) also extended this analysis to cover the years 1980 to 1996 using cloud data from the DMSP and Nimbus-7 satellites and showed that the high correlation with GCR flux is maintained. However, it was not possible to intercalibrate the different data sets so the validity of the extended data set as a measure of variations in absolute total cloudiness is open to question.
Svensmark (1998) showed that, at least for the limited period of this study, total cloud varies more closely with GCRs than with the 10.7 cm solar activity index over the past solar cycle. On longer time-scales he also demonstrated that Northern Hemisphere surface temperatures between 1937 and 1994 follow variations in cosmic ray flux and solar cycle length more closely than total irradiance or sunspot number. There has been a long-term decrease in cosmic ray flux since the late 17th century, as evidenced by the 10Be and 14C cosmogenic isotope records (Stuiver and Reimer, 1993; Beer et al., 1994), and this mirrors the long-term increase in TSI. However, the TSI reconstruction of Hoyt and Schatten (1993), which is based on solar cycle lengths, does not appear to track the cosmogenic isotope records any more closely than that of Lean et al. (1995), which is based on sunspot cycle amplitude (Lean and Rind, 1998). Such use of different solar indices may help to identify which physical mechanisms, if any, are responsible for the apparent meteorological responses to solar activity.
Kuang et al. (1998) have repeated Svensmark and Friis-Christensen's analysis of ISCCP data and showed high correlations with an El Niño-Southern Osciallation (ENSO) index difficult to distinguish from the GCR flux. Farrar (2000) showed that the pattern of change in cloudiness over that period, particularly in the Pacific Ocean, corresponds to what would be expected for the atmospheric circulation changes characteristic of El Niño. Kernthaler et al. (1999) have also studied the ISCCP dataset, using both geostationary and polar orbiter data and suggested that the correlation with cosmic ray flux is reduced if high latitude data are included. This would not be expected if cosmic rays were directly inducing increases in cloudiness, as cosmic ray flux is greatest at high latitudes. Kernthaler et al. (1999), Jørgensen and Hansen (2000), and Gierens and Ponater (1999), also noted that a mechanism whereby cosmic rays resulted in greater cloud cover would be most likely to affect high cloud as ionisation is greatest at these altitudes. Even if high cloud did respond to cosmic rays, it is not clear that this would cause global cooling as for thin high cloud the long-wave warming effects dominate the short-wave cooling effect. Kristjánsson and Kristiansen (2000) have additionally analysed the ISCCP D2 dataset, 1989 to 1993, and found little statistical evidence of a relationship between GCRs and cloud cover with the possible exception of low marine clouds in mid-latitudes. They also noted that there was no correlation between outgoing long-wave radiation, as represented in ERBE data, and GCRs. Thus the evidence for a cosmic ray impact on cloudiness remains unproven.
A further consideration must be potential physical mechanisms whereby cosmic rays might enhance cloudiness. Cosmic rays are the principal source of ionisation in the free troposphere. Furthermore, ionisation rates and atmospheric conductivity are observed to vary with solar activity. Svensmark and Friis-Christensen (1997) propose that the correlation between cosmic rays and cloud cover that they observed is due to an increase in efficiency of charged particles, over uncharged ones, in acting as cloud condensation nuclei. There is evidence for this occurring in thunderstorms (Pruppacher and Klett, 1997) but it is not clear to what extent this affects cloud development. There is also evidence that ions are sometimes critical in gas-to-particle conversion but again there is no evidence that this has any impact on cloud formation.
In a series of papers, Brian Tinsley has developed a more detailed mechanism for a link between cosmic rays and cloudiness (e.g., Tinsley, 1996). This is based on the premise that aerosols ionised by cosmic rays are more effective as ice nuclei and cause freezing of supercooled water in clouds. In clouds that are likely to cause precipitation the latent heat thus released then causes enhanced convection which promotes cyclonic development and hence increased storminess. There is some laboratory evidence to suggest that charging increases ice nucleation efficiency (Pruppacher, 1973) although there is no observational evidence of this process taking place in the atmosphere. Furthermore, only a small proportion of aerosol particles are capable of acting as ice nuclei, depending on chemical composition or shape. There are also laboratory studies (Abbas and Latham, 1969) which indicate the existence of "electrofreezing", but again no evidence in the real atmosphere. Thus Tinsley's mechanism is plausible but requires further observational and modelling studies to establish whether or not it could be of sufficient magnitude to result in the claimed effects (Harrison and Shine, 1999).
We conclude that mechanisms for the amplification of solar forcing are not well established. Variations in ultraviolet and solar-induced changes in O3 may have a small effect on radiative forcing but additionally may affect climate through changing the distribution of solar heating and thus indirectly through a dynamical response. At present there is insufficient evidence to confirm that cloud cover responds to solar variability.
Other reports in this collection