Solar Variations and Climate Change
As the Sun provides essentially all the energy that drives the Earth's climate system, it is obvious that solar variations have the potential to directly alter climate. Changes in insolation on a variety of time scales have been suggested as causes of known climate change, from the (Milankovitch) orbital cycles of thousands of years (Hays et al., 1976), to the decadal-to-century scale fluctuations typified by the Little Ice Age (Eddy, 1976). During the next 50 to 100 years, the Earth's climate is expected to warm by anywhere from 1.5° to 4.5°C in response to increasing concentrations of greenhouse gases: carbon dioxide (CO2); methane (CH4); nitrous oxide (N2O); chlorofluorocarbons (CFCs) (Intergovernmental Panel on Climate Change, 1992). If solar irradiance were to vary over the next century, natural climate change might also result. Nevertheless, until recently there has been no proof that variations in the Sun's output do in fact occur (as evidenced by the term solar "constant", which is still in widespread use).
Observations of total solar irradiance by spacecraft radiometers (Willson and Hudson, 1991; Hoyt et al., 1992) have now detected decadal variations on the order of 0.1 percent in apparent association with the Sun's 11-year activity cycle (Figure 2.1); larger variations, of the order of a few tenths percent, occur on shorter time scales and are associated with the Sun's 27-day rotation. The magnitude of the 11-year cycle effect
is compared in Figure 2.2 with anthropogenic radiative forcing of climate by increased greenhouse gases and aerosols and by ozone decreases. During the first half of the 1980s, forcing of the climate system by declining solar radiative output was more than sufficient to offset the estimated net anthropogenic forcing.
Despite the similarity of the climate forcings over the decadal time scales shown in Figure 2.2, the magnitude of the climate system's response to solar forcing could be greater or less than its response to anthropogenic forcing. This is because the translation of radiative forcing to surface
temperature response is highly specific to the altitude, latitude, and history of the forcing (Hansen and Lacis, 1990, Hansen et al., 1993), contrary to the conclusions derived from earlier general circulation model (GCM) comparisons that doubled CO2 and a 2 percent increase in solar irradiance have an equivalent effect (Hansen et al., 1984). The observed irradiance changes do imply the potential for additional solar forcing in the future, making it incumbent on global change research to monitor, understand, and ultimately predict solar effects on climate. It also makes more compelling the search for a solar signature in the historical climate record.
Understanding solar influences on climate requires the interaction of two primary research areas that are currently quite distinct: the monitoring and assessment of solar irradiance variations, which is reviewed first in this chapter, and the perspective of solar variability and climate from both the paleoclimate record and for future global change, which is discussed subsequently. Origins of the solar radiative output variations are addressed in the broader context of the variable Sun in Chapter 6.
Total Solar Irradiance Variability
Knowledge of the Sun's radiative energy output at all wavelengths is ultimately required for global change research. However, current capabilities for precise determination of this variation exist only at wavelengths shorter than about 250 nm, since at longer wavelengths the measurement uncertainties significantly exceed the amplitudes of the solar variations, which are thought to be less than 1 percent. Measurements of total (spectrally integrated) solar irradiance can be made with two orders of magnitude greater precision and currently provide the primary record of solar radiative output variations.
During the first three-quarters of the twentieth century, ground based observations were unable to detect total irradiance variations that were unambiguously solar in origin (Frohlich, 1977; Hoyt, 1979; Newkirk, 1983). The two principal limitations were uncertainties due to instrument calibration and to atmospheric interference and attenuation. However, in
the recent decade, long term solar monitoring by calibrated experiments flown on spacecraft (to overcome the atmospheric effects) has succeeded in measuring solar irradiance variability on the time scales of the Sun's 11-year activity cycle.
Launched in late 1978, and operational until 1993, the Earth Radiation Budget (ERB) experiment on the Nimbus 7 spacecraft has provided the longest solar irradiance data base (Hickey et al., 1988; Hoyt et al., 1992), although its record (Figure 2.1) is limited by the constraint that the top priority for the Nimbus 7/ERB platform was nadir-looking Earth observations, with only a few minutes per orbit of solar observational opportunity. Coincident with the ERB measurements over most of its lifetime are measurements made by the Active Cavity Radiometer Irradiance Monitor (ACRIM I), launched on the Solar Maximum Mission (SMM) in early 1980 (Willson et al., 1981; Willson, 1984; Willson and Hudson, 1991). This experiment was specifically designed for, and dedicated to, long term, high precision solar total irradiance monitoring; it ceased operation in October 1989 when the SMM spacecraft reentered the Earth's atmosphere.
The Nimbus 7/ERB and ACRIM I results provided the first unequivocal proof of intrinsic total solar irradiance variability, and variations have since been detected on every observable time scale (Figure 2.1). ACRIM's high precision is attributable to its active, electrically self-calibrating cavity (ESCC) solar pyrheliometers and its full-time solar pointing, which provided large numbers of observations. Solar variations measured by ACRIM have been corroborated by the ERB data, with the agreement between the two independent data sets improved by accounting for temperature dependent calibration errors and solar pointing limitations in ERB (Hoyt et al., 1992).
The successor experiments to the Nimbus 7/ERB were the Earth Radiation Budget Satellite (ERBS) and the Earth Radiation Budget Experiment (ERBE) on the National Oceanic and Atmospheric Administration (NOAA)-9 satellite (Lee III, 1990; Lee III et al., 1994). The use of an active cavity ESCC mode for solar observations has improved the quality of the data, but there are operational constraints on solar viewing similar to those with Nimbus 7/ERB, with even less frequent data acquisition opportunities. These latter instruments operate only about every second week and therefore have limited ability to characterize solar rotational modulations, which occur over 27-day time scales. ERBE and
ERBS data (Figure 2.1) both show a decline through the solar minimum period and an increase with the increasing solar activity of solar cycle 22. Differences do exist among the different irradiance data bases in the rate of decrease in cycle 21, the actual occurrence of minimum activity, and the rate of increase in cycle 22. These differences are possibly the result of uncorrected sensor degradation.
In September 1991, ACRIM II was launched on the Upper Atmosphere Research Satellite (UARS). Preliminary data (Figure 2.1) indicate that the UARS/ACRIM II irradiance measurements are systematically lower by about 2 W/m 2 than those of SMM/ACRIM I, whereas Nimbus 7/ERB data indicate solar variations of only a few hundredths of a percent. It would have been preferable to overlap the SMM/ACRIM I and UARS/ACRIM II experiments to provide direct cross-calibration, but the UARS launch delay made this impossible. Thus to preserve the continuity of the ACRIM solar irradiance data base, a third party comparison between ACRIM I and ACRIM II is needed, using the Nimbus 7/ERB or ERBE/ERBS experiments. In the latter case, given the infrequent ERBE/ERBS solar observations, the standard error is estimated to be some 30 times larger than a direct ACRIM I/II comparison.
The data shown in Figure 2.1 indicate that the average solar irradiance declined systematically from 1980 until mid-1986 at a mean rate of 0.015 percent per year. The irradiance minimum in 1986 occurs near the activity cycle minimum of September 1986 (as indicated by the sunspot number data in Figure 2.1). The subsequent rapid increase, corresponding to the buildup of solar activity in solar cycle 22, becomes clearly visible in 1988, continuing to the cycle 22 maximum. Declining values in the latter half of 1992 herald the approach of the next solar activity minimum, expected in 1995–1996. Taken together, the solar radiometer data indicate that the amplitude of the recent 11-year irradiance cycle is about 0.1 percent, disregarding the high Nimbus 7/ERB values in the early years of that record, where the uncertainties are large because of the need to remove significant instrumental effects from the measurements (see Hoyt et al., 1992).
While the ACRIM I, Nimbus 7/ERB, ERBS, and ERBE sensors indeed show similar solar cycle variations of about 0.1 percent (aside from the high Nimbus 7/ERB data in 1978–1979), their absolute solar irradiance values range over some 6 W/m2, due to absolute calibration uncertainties.
The current inaccuracies of the total solar irradiance measurements, which are typically ± 0.2 percent or larger (Willson, 1984; Luther et al., 1986), are more than twice the downward trend seen from 1980 to 1985. That the uncertainties in measurements made by state-of-the-art solar radiometers are significantly larger than their long term precision, and than the changes caused by solar variability, has important consequences for the continuation of the irradiance data base. In the absence of a third party comparison between ACRIM I and ACRIM II, a decade of solar monitoring would have been terminated, since the solar radiometers lack the accuracies to measure real solar changes smaller than a few tenths percent, twice the 11-year irradiance cycle.
Instruments such as ACRIM and ERB record the variation in the total electromagnetic energy from the Sun without identifying the wavelengths of the radiation at which the variations are occurring. About 99 percent of the total solar irradiance signal is from radiation at wavelengths longer than 300 nm, radiation that penetrates to the troposphere and the Earth's surface. However, shorter wavelength, more variable solar UV radiation (Figure 1.1), which is absorbed primarily above the trophosphere (Figure 1.2), contributed approximately 20 percent of the decline in the total solar irradiance from mid-1981 to 1985 (Lean, 1989). It is not known whether the entire solar spectrum varies in phase with solar activity, or how energy might be redistributed within the spectrum. Percentage variations at longer wavelengths are expected to be much smaller than those at UV wavelengths, on the order of a few tenths percent and not necessarily in phase with the activity cycle (Figure 1.1). However, these longer wavelength spectral irradiance variations have yet to be observationally defined.
Implications from Observations of Solar Surrogates
The direct correlation of solar radiative output with solar activity over the 11-year solar cycle is a major discovery from the ACRIM and ERB long term solar monitoring programs. Variations in total solar irradiance occur continuously, on time scales of days to months, in response to episodes of activity throughout the 11-year solar cycle and the modulation of active region emission by the Sun's 27-day rotation. These variations reflect the inhomogeneous emission of radiation on the solar disk. Solar radiation is depleted in active region sunspots and enhanced in active
region faculae (Wilson et al., 1981; Sofia et al., 1982; Foukal and Lean, 1986; Chapman et al., 1986). From the minimum to the maximum of the 11-year activity cycle there is an increase in active regions, both sunspots and faculae, on the solar disk. Total solar irradiance is thought to be positively correlated with the 11-year solar activity cycle because excess facular brightness, especially from the background active network of bright emission outside of the largest active regions, more than compensates for the sunspot deficit (Foukal and Lean, 1988; Willson and Hudson, 1991). Global perturbations in temperature and/ or diameter may also be occuring (Kuhn et al., 1989; Ribes et al., 1989; Kuhn and Libbrecht, 1991; Sofia and Fox, 1994).
To understand the forcing of the climate system by solar irradiance changes, it is necessary to have empirical models capable of extrapolating the radiative output variations to epochs beyond present solar cycles. Knowing that total solar irradiance is enhanced at times of maximum activity, and that these variations appear to arise from the competing effects of two different types of active regions (dark sunspots and bright faculae), suggests that past variations may be reconstructed from historical indicators of solar activity. Empirical parameterizations have been developed to investigate this possibility. The most successful models (Chapter 6) are based on regressions between the ACRIM I or ERB results (corrected for sunspot effects) with specific solar activity indices (derived from the solar He I 1083 nm, Ca II 393.4 nm, and H I 121.6 nm lines) that are considered better surrogates for the total irradiance brightness source than are the classical solar activity indicators, the Zü rich sunspot number and the 10.7 cm radio flux (Foukal and Lean, 1988; Livingston et al., 1988).
Many of the major features of the irradiance data have been reproduced by a regression model using the equivalent width (EW) of the solar He I line; these models do not reproduce the high levels of irradiance measured by the radiometers in 1979– 1980 near the maximum of solar cycle 21. Also , there are inconsistencies between the Nimbus 7/ERB measurements and model around the time of the cycle 22 activity maximum. Either the empirical relationships differ between solar minimum and solar maximum, and perhaps from one solar cycle to the next, or the irradiance observations are too high because of instrumental artifacts.
When empirical models of total solar irradiance variability developed from the extant spacecraft data are extrapolated over the past century, the long term variations arising from magnetic sunspot and faculae features alone have been no greater than 0.1 percent (Foukal and Lean, 1990). However, as discussed below, limits of solar variability, such as inferred from observations of Sun-like stars, provide circumstantial evidence for a brightness component that has been slowly increasing the total solar irradiance since the Maunder Minimum, a time of reduced solar activity from about 1645 to 1715. With changes in this additional brightness component superimposed on the 11-year cycle variations, a reduction of 0.24 percent is estimated for the Maunder Minimum, relative to the mean of the contemporary 11-year irradiance cycle (Lean et al., 1992a).
Solar observations made by telescopes in the seventeenth century also suggest increased solar diameter and equatorial surface rotation during the Maunder Minimum, compared with the modern Sun (Eddy et al., 1976; Nesme-Ribes et al., 1993). Using apparent solar radius as a surrogate for solar irradiance leads to speculation of a reduction as large as 1 percent during the late seventeenth century (Nesme-Ribes et al., 1993). In addition to uncertainties about the amplitude of solar irradiance values in the Maunder Minimum, there are also differences in reconstructions of the relative temporal variations in the irradiance since then -- over the past 300 years. While derivations based on different solar surrogates -- such as the apparent solar radius record, the length of the sunspot cycle, the sunspot decay rate, or the mean activity level of the 11-year cycle -- do agree about the overall increasing levels of solar activity during the past 300 years, phase differences in specific episodic increases and decreases of activity may be as large as 20 years (Hoyt and Schatten, 1993).
Relatively continuous, direct records of solar activity exist only since the telescopic discovery of sunspots in the early 1600s. For estimating changes in solar activity over the past several thousand years, other indicators have been proposed, such as variations in cosmogenic 14 C in tree rings and 10 Be in ice cores (Beer et al., 1988; Suess and Linick, 1990; Beer et al., 1991; Stuiver and Reimer, 1993; Stuiver and Braziunas, 1993). Historical solar activity variations inferred from these cosmogenic
isotopes prior to the industrial era are similar (McHargue and Damon, 1991), even though the physical connections between the proxies and solar activity the indirect. For example the 14 C record is connected to solar activity as follows. Changes in the solar wind in response to solar activity variations modulate the heliospheric magnetic topology. During times of minimum solar activity, cosmic rays are swept out of the heliosphere less effectively by the solar wind than during maximum solar activity. Thus at solar minima an increased flux of galactic cosmic rays reaches the Earth's atmosphere. This leads to increased production of 14 C, which accumulates in the biosphere where it is available for uptake by trees. The similarity between the recent 14 C record and the envelope of the sunspot record of solar activity is evident in Figure 1.3. Although many uncertainties exist in interpreting such phenomena, these records offer the potential for gaining improved understanding of solar behavior in the extended past, relevant to global change issues (Wigley and Kelly, 1990; Damon and Sonett, 1991).
Evidence from Observations of Sun-Like Stars
The sun is a rather common star, and its behavior is thought to be typified by that of stars of similar age, mass, radius, and composition. Routine monitoring of the activity of a selection of Sun-like stars during the past decade has indeed revealed rotational and activity cycles on time scales similar to those seen in the Sun (Radick et al., 1990). Also, observations of Ca II emission in Sun-like stars indicate that 4 out of 13 stars monitored monthly since 1966 exhibited no activity cycle, implying that extended periods of inactivity, as exemplified in the modern solar record by the Maunder Minimum, may be common (Baliunas and Jastrow, 1990). This conjecture is roughly supported by the occurrence of minima that punctuate the 14 C geophysical record of solar activity.
In the four Sun-like stars observed to be inactive, Ca II emissions were almost always lower than in the stars that exhibited activity cycles (Baliunas and Jastrow, 1990). While et al. (1992) have shown that the Sun's contemporary Ca II emission corresponds to that of the brighter half of the cycling stars observed by Baliunas and Jastrow (1990) and does not overlap the range of lower Ca II emission typical of noncycling stars. Lean et al. (1992a) investigated the implications of these stellar observations
for the Sun's radiative output by utilizing current understanding of the origin of the variations in total solar irradiance and in the Ca II emission from the Sun and stars. Their results, shown in Figure 2.3, suggest that during the Sun's Maunder Minimum the total solar irradiance might have been about 0.24 percent less than its mean value between 1980 and 1990.
Such a decrease is consistent with inferences about the level of solar radiative output during the Maunder Minimum reported by Wigley and Kelly (1990) from the climate record, and also with stellar observations that provide compelling evidence for variabilities of 0.2 percent to 0.5 percent in the luminosity of Sun-like stars (Lockwood and Skiff, 1990; Lockwood et al., 1992). Foukal (1994) notes that the larger luminosity changes observed in Sun-like stars do not necessarily imply equally larger changes in the Sun, at least in the present epoch, since these changes are actually consistent with current understanding of modulation by photospheric magnetism. Also, the variability amplitudes detected in stars likely depend on the observer's viewing angle relative to the stellar spin axis (Schatten, 1993).
Solar Forcing of Climate Change
Variations in solar irradiance may affect the Earth's climate through a direct influence on the global mean temperature or in more subtle ways. The magnitude of climate change that can be associated directly with the changes in total solar irradiance measured during the recent solar activity cycle (about 0.1 percent, see Figure 2.1) is small compared to past climate excursions. Current GCMs estimate that a 2 percent increase in the solar irradiance would produce about 4°C global warming (Hansen et al., 1984). Assuming this result is the right order of magnitude, and that it scales linearly, the 0.1 percent irradiance variation observed by spaceborne radiometers in solar cycle 21 would produce an equilibrium temperature change of about 0.2°C. However, the change from maximum to minimum activity of the 11-year cycle occurs over about five years, too little time to allow for full equilibrium response of the climate system. Furthermore, where averaged over the solar cycle, the effect is reduced by the periodic nature of the forcing, the radiative change during the second half effectively
balancing that during the first half. Thus, direct solar-cycle induced climate change is expected to be less than the variability of ±0.2°Cthought to be inherent in the climate system in the absence of external forcing (Hansen and Lebedeff, 1987). To the extent that feedbacks of the climate system are not symmetric, it is possible that solar cycling could produce a net climate forcing that would accumulate over longer times.
From the standpoint of future global climate change, if temperature changes of about 0.2°C were the maximum expected from solar forcing, it might be concluded that solar forcing could be safely ignored, relative to a doubling of greenhouse gases, for which the predicted temperature increase is in the range 1.5°–4.5°C. Yet the climate record suggests that larger effects may have resulted from solar forcing in both the distant and recent past, and that even today unexpected sensitivities may exist (e.g., Wigley and Kelly, 1990; Damon and Sonett, 1991; Labitzke and van Loon, 1993; Thompson et al., 1993). The challenge of expeditious detection of global warming, well in advance of the 50 to 100 year time for greenhouse gas doubling, requires that small (< 1°C) temperature increases be adequately separated into natural and anthropogenic components. Furthermore, as a result of the expected cooling by aerosol and ozone changes, the net anthropogenic climate forcing may be only about half that expected for greenhouse gases alone (Hansen et al., 1993), making the direct solar influence a potentially larger component of the net climate change signal.
Solar-induced changes in the stratosphere could have a variety of indirect influences on the troposphere and climate (e.g., NAS, 1982). Investigations with general circulation models (Kodera, 1991; Rind and Balachandran, 1994) suggest that variations in solar ultraviolet energy input modify the ozone and temperature structure of the stratosphere, affecting the latitudinal temperature gradient. This modifies stratospheric wind speeds and the ability of long-wave energy to propagate out of the troposphere. Altered tropospheric stability affects various tropospheric dynamic processes, including the Hadley cell intensity at low and subtropical latitudes, and low pressure systems in the extratropics.
Paleoclimate, like the recent climate, displays numerous examples of potential interactions between solar radiative forcing and climate. On the longer time scales associated with variations in the Earth's orbital parameters, the insolation gradient between low and high latitudes may be modified, or its seasonal variation changed, which could lead to the growth of ice sheets and altered ocean circulation patterns. Indeed, ubiquitous in climate records are cycles with periods of about 23,000, 41,000, and 100,000 years that are widely attributed to variations in the distribution of insolation with orbital parameters -- the so-called Milankovitch forcing.
For the purposes of the U.S. Global Change Research Program, discussed below are three times scales associated with different mechanisms for solar variability: the relatively recent climate (a hundred to a few thousand years), the weather (tens of years), and orbital variations (many thousands of years).
Solar Irradiance Changes and the Relatively Recent Climate
There has been much speculation that climate changes over the past few thousand years have been the product of variations in the Sun's radiative output. Eddy (1976) pointed out the coincidence in time between the Maunder Minimum of solar activity and the lowest temperatures of the Little Ice Age in Europe and North America (see Figure 1.3). He also presented qualitative evidence that other century-scale variations in climate over the past millennium coincided roughly with variations in solar activity deduced from anomalies in the 14C cosmogenic isotope record. Whereas the long term trend in records of cosmogenic isotopes such as 14C and 10Be reflects, primarily, changes in the Earth's magnetic field that affect the interaction of cosmic rays with the Earth's atmosphere, the wiggles superimposed on the smooth long term trend are believed to occur because of the modulation of the local cosmic ray intensity by magnetic fields embedded in the solar wind, which varies in response to solar activity (Damon and Sonett, 1991; Beer et al., 1991; Stuiver and Braziunas, 1993). Thus, enhanced solar activity corresponds to 14C minima, and the mechanism proposed by Eddy for the apparent relationship between climate and the 14C wiggles involved changes in the total solar irradiance linked to the long term envelope of the 11-year sunspot cycle and reflected in the 14C record.
The extent to which cosmogenic isotope variations really indicate terrestrially relevant variations in solar energy outputs, either radiative or particle, and the scaling of the relationship over long times, is poorly known; the paleoclimate record is similarly somewhat uncertain (Bradley and Jones, 1993). Although results are mixed (Wigley and Kelly, 1990; Crowley and Howard, 1990; Damon and Jirikowic, 1994), there is some suggestion of a relationship; during the past 10,000 years, six of the seven strongest maxima in the 14C wiggles correspond closely to climate minima,
reflecting an apparent 200 year cycle in the 14C record that has been recently identified also in the 10Be ice core record.
Reid (1991) has extended the possible relationship between solar variability and climate parameters down to the decadal time scale by pointing out the similarity between the record of globally averaged sea-surface temperature anomalies and the solar cycle envelope over the past 130 years (Figure 2.4). Accounting for the observed sea-surface temperature changes in terms of solar forcing requires a change in solar irradiance of about 0.6 percent. An even more striking relationship between northern-hemisphere surface temperatures and the variation of solar cycle length over the past 130 years (Figure 2.4) has been pointed out by Friis-Christensen and Lassen (1991). The solar cycle envelope and the solar cycle length are related to each other, and both are indicators of the long term variability of solar activity (with longer cycles having smaller amplitudes). The striking correspondence of solar variability indicators with climate parameters provides strong circumstantial evidence for a physical connection, most probably through a variation in solar irradiance, but possibly through an indirect mechanism.
To attribute changes in climate to solar variability, it is first necessary to estimate the magnitudes of changes in both climate and solar radiation and to show that they are consistent. Yet the global temperature changes that have accompanied climatic events during the past millennium are not known in detail. Various figures have been given for the surface temperature excursions of this period, ranging from 0.4–0.6°C (Wigley and Kelly, 1990) to 1°–1.5°C (Crowley and North, 1991). To translate this temperature change into an equivalent solar radiative forcing effect requires that the climate sensitivity be known. Using the value associated with current GCMs of 1°C/(W/m2) and assuming that the duration of the events was sufficiently long (about 100 years) to allow for a full climatic response, accounting for the entire range of historical climate variability requires changes in total solar irradiance ranging from 0.4 to 1.5 percent. These variations are larger than those observed by the ACRIM I, Nimbus 7/ERB, and ERBS solar radiometers over the past decade (Figure 2.1).
It is not known whether changes in Solar irradiance larger than have been observed in the contemporary Sun are plausible on longer time scales although estimates of the likely variability of 0.24 percent due to change in sunspots, faculae, and bright magnetic network radiation on century
time scales are somewhat smaller (Lean et al., 1992a and Figure 2.3) than the 0.4 to 1.5 percent needed to explain the paleoclimate record. If the climate sensitivity is greater (one inference from Milankovitch GCM studies; Rind et al., 1989; Phillipps and Held, 1994; discussed below) or the global temperature change smaller than indicated, the required solar variability would be reduced. Furthermore, although GCM climate simulations estimate a mean global temperature reduction of 0.46° C for a solar irradiance reduction of 0.25 percent (Rind and Overpeck, 1993), some regions of the Earth's surface may cool and others warm by as much as 1° C as a result of advective changes caused by differential heating of the land and oceans.
The problem of assessing direct solar radiative forcing of climate change is additionally complicated because the extent to which total solar irradiance variability arises from radiative changes at ultraviolet rather than at visible wavelengths (Lean, 1989) determines the altitude of its direct impact on the global system. If this impact shifts to altitudes mostly above the troposphere, total solar irradiance forcing of surface temperature would be reduced. On the other hand, the amplitude of irradiance variations in the visible and infrared portions of the solar spectrum that directly heat the surface, though thought to be small (e.g., Figure 1.1), is not currently known.
While solar radiative changes are probably not the sole driving force of the historical climate record, they nevertheless will need to be understood and quantified in order to unravel the contribution of solar forcing. Indeed, circumstantial evidence points to a solar forcing contribution to the temperature changes observed over the past century (Kelly and Wigley, 1992; Schlesinger and Ramankutty, 1992) that decreases the predicted temperature change associated with a doubling of atmospheric CO2 by nearly half (Lacis and Carlson, 1992).
From the perspective of the U.S. Global Change Research Program, it is important to know how solar irradiance variations can be expected to vary in the future and, in particular, the likelihood that events such as another Little Ice Age, will occur in the coming century. Were the only variations in solar radiative output an 11-year cycle with peak-to-peak amplitude of about 0.1 percent, solar forcing could be expected to modulate the net anthropogenic climate forcing as shown in Figure 2.2. But another scenario is that additional solar forcing might arise from
longer-term irradiance variations superimposed on the 11-year activity cycle, such as the speculated long term increase in irradiance from the Maunder Minimum to the present Modern Maximum.
Lacking a detailed modeling capability for, and adequate knowledge of, solar processes on which to base predictions, researchers have utilized spectral analysis to develop predictive tools. Phenomena such as sunspot numbers have periodicities on the order of 100, 55, and 11 years, along with the solar magnetic cycle of 22 years (e.g., Berger et al., 1990). Ice core records as well as other climatic data suggest periods of about 80 and 180 years (Johnsen et al., 1970), possibly related to solar activity (Otaola and Zenteno, 1983). Extrapolation into the future of two cycles evident in the 14 C record, at 208 years (the Suess cycle) and 88 years (the Gleissberg cycle), suggests that the increasing solar activity that has followed the Maunder Minimum may continue into the early twenty-first century (Damon and Sonett, 1991), with a decline commencing around 2040. But extrapolation of these cycles into the future and prediction of solar effects is a highly questionable procedure, given our lack of knowledge of the fundamental processes involved (see Chapter 6).
Wigley and Kelly (1990) have attempted to assess limits on the role that solar forcing of climate change may play, relative to that of greenhouse gases, during the next 200 years. Analogous to their approach, and consistent with their results, the predictions shown in Figure 2.5 indicate that were the Sun to experience a period of inactivity such as the Maunder Minimum, commencing in the year 2000, and accompanied by reduction in its radiative output of 0.25 percent, the resultant climate forcing would indeed modulate, but not counter, the predicted anthropogenic climate forcing. As noted previously, determining the actual climate impact of the forcings shown in Figure 2.5 (and Figure 2.2) is difficult because of the specific nature expected for the climate system's response to each of the individual forcings.
Solar Activity Cycles and the Weather
There have been many studies of the possible relationships between weather phenomena and the 11-year solar sunspot cycle or the 22-year solar magnetic cycle. Summaries of the results of these studies prior to the early 1980s have been published by Herman and Goldberg (1978) and NAS
(1982). While statistical relationships have in some cases been significant, thescientific community as a whole has strongly resisted accepting the findingsas proof of a causal relationship, primarily because the mechanisms providingthe linkage have not been apparent.
The subject has received new impetus in the past decade, due both to the observation of total and ultraviolet irradiance variations associated with the 11-year solar activity cycle and to observations of a distinct 10-to-12-year oscillation (TTO) in various atmospheric parameters that appear to be in phase with the solar cycle (e.g., Labitzke and van Loon, 1990;
1993) and related to the quasibiennial oscillation (QBO) in tropical stratospheric winds (Figure 2.6). The connection between the TTO and the 11-year solar cycle remains unproven and the statistical validity of the relationship has been debated (Salby and Shea, 1991). Nevertheless, the relationship is considered sufficiently useful to be incorporated in techniques for seasonal forecasting of U.S. weather (Barnston and Livezey, 1989).
Other studies have indicated correlations between solar activity and weather phenomena even when no stratification by QBO phase is made. For example, the mean latitude of winter storm tracks in the north Atlantic appears to shift equatorward at times of maximum solar activity relative to times of activity minima (Tinsley and Deen, 1991). Periods at 11 and/or 22 years have appeared as prominent peaks in spectral analyses of the Earth's surface temperature (Allen and Smith, 1994), sea level pressure (Kelly, 1977), the length of the Atlantic tropical cyclone season (Cohen and Sweester, 1975), ice accumulation data (Holdsworth et al., 1989), drought incidence in the western U.S. (Mitchell et al., 1979), the areal extent of North American forest wildfires (Auclair, 1992), global northern hemisphere marine temperatures (Newell et al., 1989), and the separation between annual dust layers in an ice core from the Guliya Ice Cap (Thompson et al., 1993).
However, the basic problem remains: without an understanding of the physical causal connection, the suspicion will persist that the results are the product of a posteriori choices (e.g., Baldwin and Dunkerton, 1989, Salby and Shea, 1991) or are simply the product of natural internal variability (James and James, 1989). Understanding the implied relationships between the Sun and the weather, and the role played by the QBO, would be of enormous benefit, both from the practical standpoint of seasonal forecasting and by enhancing the ability to model and deduce the sensitivity of the climate system to a small external perturbation.
Results of a recent set of GCM studies (Rind and Balachandran, 1994; Balachandran and Rind, 1994) indicate that variations in the middle atmosphere temperature and wind structure associated with the QBO and solar UV irradiance variations did impact the troposphere, primarily through alterations in the generation and propagation of the longest tropospheric planetary waves. The resulting longitudinal variations in tropospheric temperature, wind, and geopotential height were similar in
nature and magnitude to those reported by van Loon and Labitzke (1988). The major caveat is the exaggerated UV irradiance variations utilized in the study; nevertheless, knowledge of the response of the troposphere to solar cycle activities that directly affect the middle atmosphere is growing.
In addition, the GCM studies demonstrated that dynamical changes induced by solar cycle variations can affect the radiative properties of the troposphere by influencing cloud and snow cover. Furthermore, the effects do not cancel when averaged over the ascending and descending portions of the cycle. This implies that a time-integrated solar cycle forcing of the climate system is possible through its impact on tropospheric dynamics and feedbacks, rather than through direct insolation perturbation. If so, this would likely have a very different climate impact than the forcing associated with increasing greenhouse gases, whose effect on the middle atmosphere and tropospheric dynamics is entirely different.
Insolation Changes Due to Orbital Variations
The study of Hays et al. (1976) showed that the climate record deduced from deep sea sediments varied with periodicities that generally matched those of the Earth's orbital variations, specifically variations in eccentricity, obliquity (axial tilt), and date of perihelion (Figure 2.7). The theory that orbital variations are indeed the pacemakers of the ice ages has become widely accepted. However, the theory does have problems, both from the observational and the modeling perspectives, that are instructive for evaluating solar influences on global change, and which must be also addressed by the USGCRP in the broader context of the Earth System History USGCRP science element.
The need to understand this issue arises not primarily from the need to predict future climate based on the orbital configurations, but rather from the standpoint of what it implies about the sensitivity of the climate system, and about the ability of climate models to simulate climate sensitivity, since the forcing can be quantified. The last ice age was presumably initiated during the time of strongly reduced summer insolation nearly 110,000 years before the present (BP). The reductions projected for the next 10,000 years are extremely small in comparison and, from this perspective, another ice age is unlikely in that time frame. This is illustrated in Figure 2.7.
Are the changes in insolation effected by the Earth's orbital variations sufficient to have initiated ice ages that is, are they the real cause of the glacial/interglacial transitions of the Pleistocene? Melting of the ice sheets occurred 15,000 to 10,000 years BP, coincident with high northern hemisphere summer insolation, in agreement with this hypothesis. However, the southern hemisphere climate also experienced rapid warming in this time interval, when southern hemisphere insolation was at a minimum. The apparent synchronicity of the two hemispheres in their responses to orbital variations, which for the precessional cycle has opposite solar insolation effects in the two hemispheres, has long been a
mystery and raises the question of how much of the climate response is actually associated with orbital forcing.
Spectral analysis of the paleoclimate record shows that the maximum power lies in the approximately 100,000 year period, which is of the same order as the Earth's eccentricity variation. However, the changes in eccentricity, on the order of a few tenths of a percent over the past 5 million years, produce little change in net annual solar radiation, so that any possible effects on the seasonal distribution of radiation must be combined with variations in tilt and precession of the Earth's rotation axis, which are larger. Thus it is surprising that the about 100,000 year period dominates in the climate record. Examples of this mismatch can easily be found: the peak of the last ice age, about 20,000 years BP, coincides with a very weak minimum in Northern Hemisphere summer solar insolation, and the deglaciation Northern Hemisphere summer maximum at about 12,000 years BP is no larger than a similar feature at about 30,000 years BP, which did not lead to complete deglaciation. These facts suggest that processes other than direct solar forcing may be responsible for the observed climate record.
Even the timing of the insolation variations relative to the climatic response has been questioned. Winograd et al. (1988, 1992) analyzed the oxygen-18 variations found in a calcitic vein in the southern Great Basin. The uranium series age dates of the calcite vein indicated that major glacial/interglacial transitions occurred some 10,000 to 20,000 years before the solar insolation variations; for example, the peak interglacial in this record appears at 147,000 ± 3,000 years BP, significantly before the insolation peak. While the relevance of this local record to global temperature and precipitation changes may be in doubt, high sea level stands in the period 135,000 to 140,000 years BP have been found by various researchers (e.g., Moore, 1982). The absolute dating capability associated with the calcite vein is in contrast to the approximate dating techniques associated with the deep sea paleoclimate record, where assumptions about sedimentation rates are fundamental in matching the orbital periodicities.
When the orbital solar insolation variations are incorporated in general circulation climate models, the temperature changes are not sufficient to produce ice sheet growth, especially in regions of low altitude accumulation, as was apparently the case for the Laurentide ice sheet (Rind et al.,
1989; Phillipps and Held, 1994). Either the models are incomplete or orbitally induced solar insolation variations are at best only a catalyst for glacial/interglacial changes. Both of these conclusions have important implications for global change projections: the former implies that contemporary GCMs might not be sufficiently sensitive to solar radiative forcing (whether of orbital or solar activity origin), while the latter emphasizes that it is the climate system feedbacks that are most important in producing climate change, invalidating the use of simple transfer functions between radiative forcing perturbations and climatic responses.