It has long been known that substantial variations occur over the 11-year sunspot cycle in the short wavelength radiation that is absorbed in the upper atmosphere. The solid line in Figure 1, adapted from data in a recent review by Lean [1991], shows current estimates of the fractional variation between solar maximum and minimum inferred from direct measurements. The dash-dot line shows an estimate of the variation across the visible spectrum, which contains the vast bulk of the sun's radiant energy, and which is the principal driving force for global climate. The horizontal dashed line shows the approximate variation that was directly measured in the sun's total irradiance (long known as the solar ``constant'') over the 11-year period encompassing the maxima of solar cycles 21 and 22, using the accepted numerical reference system. The measurements made by the Active Cavity Radiometer Irradiance Monitor (ACRIM I) on board the Solar Maximum Mission satellite have been summarized by Willson and Hudson [1991], while the corresponding measurements made by the Earth Radiation Budget (ERB) experiment on Nimbus-7 have been discussed by Hoyt et al. [1992]. Although these two independent data sets differ in the absolute magnitude of the irradiance, the time-varying components are in relatively good agreement. It is both frustrating and comforting that the fractional spectral irradiance variations in the visible shown in Figure 1 are so small---frustrating because they are near the limit of our ability to measure them, and comforting because it is doubtful if life could have survived the much larger variations that occur at shorter wavelengths.
The total irradiance variations seen to date have been tentatively explained in terms of a combination of decreases due to sunspot blocking and increases due to bright faculae and plages [Foukal and Lean, 1990], both of which are positively correlated with solar magnetic activity. The two effects tend to cancel each other, and the net positive correlation of the total irradiance is explained as a slight excess of facular brightening over sunspot darkening. An important question that remains unanswered is the extent to which this model can be applied to infer irradiance variations over longer time periods by using such proxies for magnetic activity as the sunspot number or sunspot area, as has been done by Foukal and Lean [1990]. Other sources of luminosity variation that are not related in this simple way to solar activity cannot be ruled out, and some indication that other sources exist is suggested by the fact that the irradiance variations seen by the early spacecraft measurements in 1980 showed substantially higher irradiances than would have been predicted from those measured at later times [Willson and Hudson, 1991]. Unfortunately, the only continuous solar observations that extend over the important climatic time scale of decades to centuries are those of sunspots, yielding a measure of magnetic activity. If irradiance variations have occurred that were not connected to magnetic activity, there is very little hope that we would ever be able to prove their existence. A long-term program of monitoring the solar irradiance from space will provide the answers as far as the sun itself is concerned, but the climatic effects of solar variations are likely to be increasingly difficult to isolate in the future, as anthropogenic forcing becomes stronger.
Discussion of the causes of the irradiance variations is
beyond the scope of this article, but their magnitude is of
critical importance to the question of climatic effects. As shown
in Figure 1, the total variation between the peaks of solar cycles
21 and 22 was about 0.1%, which is expected to produce a
corresponding variation of about 0.2
C in globally
averaged equilibrium surface temperature [Hansen and Lacis, 1990].
Because of the thermal inertia of the oceans, however, the time
needed to approach equilibrium is much longer than 11 years [e.g.,
Reid, 1991], so that the actual temperature response to the
observed variation is likely to be considerably smaller, and
probably insignificant from a climatic point of view.
For this reason, attention has focussed on longer-term variations in solar irradiance, for which there is as yet no direct evidence. Eddy [1977] first pointed out that the Maunder Minimum of solar activity in the late 17th century, when sunspots were apparently absent for several decades, coincided roughly with the coldest phase of the Little Ice Age in Europe and North America, and that this might be explained by a dependence of solar irradiance on the overall level of solar activity. The proxy relationships observed during solar cycle 21 and the behavior of other sun-like stars [Baliunas and Jastrow, 1990] have been used by Lean et al. [1992] to estimate the solar irradiance during the Maunder Minimum as somewhere between 0.15 and 0.35% lower than the present solar-cycle mean value. An independent estimate by Baliunas and Jastrow [1993] gave a range of 0.1 to 0.7% based purely on observations of solar-like stars, discussed by Lockwood et al. [1992]. The use of other stars to infer solar variability has been questioned by Schatten [1993], however, who has pointed out that the observed irradiance is likely to be a function of the heliographic latitude of the observer, being a minimum near the solar equatorial plane, where the Earth is located. Since other stars are observed at random latitudes relative to their spin axes, the variations observed might not be directly relevant to the local situation.
Baliunas and Jastrow [1993] conclude that a reduction in
irradiance of 0.4%, in the middle of their calculated range, would
be enough to explain the cold average temperatures of the Little
Ice Age, as estimated by Wigley and Kelly [1990]. Hoyt and Schatten
[1993] have used a variety of possible proxies for solar irradiance
to estimate a value for the Maunder Minimum period that is about 5
W m-2, or about 0.36% below current values, in general agreement
with other estimates. Rind and Overpeck [1993] used a general
circulation model to estimate the regional temperature changes
caused by a decrease of solar irradiance by 0.25%, in the middle
of the range estimated by Lean et al. [1992]. They found a global
average reduction of 0.45
C with no clear latitudinal
variation, and with the largest effects over the continental land
masses. Changes in advection resulting from the change in
land-ocean temperature contrasts actually led to significant
warming in some regions. Although averages over different time
slices of the model output gave similar temperature patterns, the
authors point out that the magnitude of the temperature variations
is comparable to that of the model standard deviations, so they
cannot be taken as definitive.
There are thus good reasons for believing that changes in the solar ``constant'' have been at least a significant contributor to the rise in global temperature since the late 17th century, and possibly to earlier climate variations on the century time scale [Reid, 1991; Baliunas and Jastrow, 1993]. Their contribution to more recent climate change, however, remains speculative and controversial, and the evidence is largely circumstantial, based on correlations between the global temperature record and proxies for solar irradiance variations. Reid [1991] pointed out the similarity between the overall level of solar activity, as expressed by the envelope of the 11-year sunspot cycle, and globally averaged sea-surface temperatures over the period from 1860 to the present. Using a simple one-dimensional model of the ocean's thermal structure, he calculated that variations in solar irradiance of several tenths of a percent on the time scale of decades to centuries would be needed to explain the effect, assuming that no other causes of temperature change were in operation. Scientists in Denmark [Friis-Christensen and Lassen, 1991] reported an even more striking correlation between northern-hemisphere land temperatures and the length of individual solar cycles, which range roughly between 9 and 14 years. Hoyt and Schatten [1993] have also shown a correlation between the same surface temperatures and an irradiance model based on several proxies for the period since 1750.
A high degree of correlation between surface temperatures and solar irradiance, or its proxies, can be expected only if solar variability has been the dominant source of climate change. This is probably not the case, especially in recent decades, when anthropogenic effects must have been making significant contributions. Furthermore, the irradiance variations that have been estimated from proxy sources in recent decades appear to be too small to explain the observed temperature variations, based on our current understanding of the climate system. Attempts have been made to include other sources of climate variability in model calculations to evaluate the relative importance of solar forcing by comparison with the temperature record. Schlesinger and Ramankutty [1992] in particular concluded that intercyclic variations in solar irradiance have probably been significant contributors, but that the dominant role has been played by increases in greenhouse gas emissions.
Further developments in this area probably depend on whether more decisive evidence can be obtained on the past behavior of solar irradiance or on the sensitivity of the climate system to the rather small variations that are well established. While the state of our understanding is too insecure to allow firm estimates of how the sun's irradiance may behave in the future, it is worth noting that Wilson [1992] and Schatten and Pesnell [1993] have predicted that solar cycle 23, peaking around the turn of the century, will be a large one. If it is one of the greatest on record, as predicted by Wilson [1992], any long-term variation in irradiance that depends on the long-term average level of solar activity would be expected to be considerably larger during the next few years than it has been in recent years, when the peak amplitudes of cycles 21 and 22 were nearly identical [Reid, 1991]. If so, a solar contribution to global warming might occur that would add to the expected contribution from increased greenhouse-gas loading. In general, contemporary solar activity has reached historically high levels as measured by sunspot number and cosmogenic isotope concentrations. The rise since the Maunder Minimum period 300 years ago has roughly paralleled the climb in global temperatures from the depths of the Little Ice Age; whether or not this trend will continue remains to be seen.