By how much does changing radiation from the sun influence the earth’s climate, presently and in the recent past, compared with other natural and anthropogenic processes? Current knowledge of the amplitudes and timescales of solar radiative output variability needed to address this question is described from contemporary solar monitoring and historical reconstructions. The 17-yr observational database of space-based solar monitoring exhibits an 11-yr irradiance cycle with amplitude of about 0.1%. Larger amplitude solar total radiative output changes—of 0.24% relative to present levels—are estimated for the seventeenth-century Maunder Minimum by parameterizing the variability mechanisms identified for the 11-yr cycle, using proxies of solar and stellar variability. The 11- and 22-yr periods evident in solar activity proxies appear in many climate and paleoclimate records, and some solar and climate time series correlate strongly over multidecadal and centennial timescales. These statistical relationships suggest a response of the climate system to the changing sun. The correlation of reconstructed solar irradiance and Northern Hemisphere (NH) surface temperature anomalies is 0.86 in the pre-industrial period from 1610 to 1800, implying a predominant solar influence. Extending this correlation to the present suggests that solar forcing may have contributed about half of the observed 0.55°C surface warming since 1900 and one-third of the warming since 1970. Climate model simulations using irradiance reconstructions provide a tool with which to identify potential physical mechanism for these implied connections. An equilibrium simulation by the Goddard Institute for Space Studies GCM predicts an NH surface temperature change of 0.51°C for a 0.25% solar irradiance reduction, in general agreement with the preindustrial parameterization. But attributing a significant fraction of recent climate warming to solar forcing presents serious ambiguities about the impact of increasing greenhouse gas concentrations whose radiative forcing has been significantly larger than solar forcing over this time period. Present inability to adequately specify climate forcing by changing solar radiation has implications for policy making regarding anthropogenic global change, which must be detected against natural climate variability.
Earth’s climate fluctuates (Bradley 1991; Hartmann 1994). Volcanic eruptions and the sun’s activity are potential causes of natural climate change (Hansen and Lacis 1990), as are internal oscillations and couplings between the ocean and the atmosphere (Rind and Overpeck 1993; Crowley and Kim 1993; Mehta and Delworth 1995). Industrial activity and associated trace gas releases may also be altering the earth’s climate (Houghton et al. 1992; Houghton et al. 1995). Necessary for reliable detection of suspected anthropogenic climate change is proper specification of natural climate fluctuations that are occurring independently of human activities.
Concern abounds that the 0.55°C increase in the global surface temperature of the earth during the past century (Parker et al. 1994) may signify the climate system’s response to anthropogenic influences that have escalated during the industrial epoch (Kuo et al. 1990;Hansen et al. 1993; Visser and Molenaar 1995). Instrumental records, shown in Fig. 1, document this temperature increase globally (land plus ocean) and in both the Northern and Southern Hemispheres (NH and SH; Houghton et al. 1992). The temporal structure of the observed warming is clearly more complicated than the simple upward trend expected from monotonically increasing annual greenhouse gas concentrations alone (e.g., Lau and Weng 1995). The warming is also geographically complex (Parker et al. 1994)—significant variance exists at multiannual and decadal timescales with different amplitudes at different locations, sometimes masking the upward trend (Allen and Smith 1994;Mann and Park 1994).
It is a challenging task to identify the natural and anthropogenic causes of twentieth century climate change with the scientific certainty needed for environmental policy making. A recent paleoreconstruction of NH summer surface temperatures (Bradley and Jones 1993), also shown in Fig. 1, suggests that the twentieth century warming may actually be part of a longer-term, larger amplitude warming that commenced in the seventeenth century, prior to the industrial epoch. During a period from roughly 1450 to 1850—loosely called the Little Ice Age—surface temperatures were at times cooler by as much as 0.7°C than present, especially in northern Europe (Bradley and Jones 1993; Damon and Jirikowic 1994) (e.g., near 1600 in Fig. 1). Rather than a response to anthropogenic influences, the warming since the Little Ice Age might signify the climate’s recovery to warmer temperatures as may have been present in the twelfth century so-called Medieval Climatic Maximum (Hughes and Diaz 1994) that preceded it.
Since 1850, industrially produced concentrations of greenhouse gases—CO2, CH4, N2O, chloroflurocarbons (CFCs)—and of tropospheric sulfate aerosols, have increased (Houghton et al. 1995). The overall activity level of the sun has risen, too (NRC 1994). Earth’s surface is warmed both by increased greenhouse gas concentrations and the enhanced solar radiation speculated to accompany the sun’s increased activity, since both input additional energy in the climate system. In contrast, increasing atmospheric aerosol concentrations are expected to cool the earth’s surface by reflecting more of the sun’s radiant energy back to space. Whereas surface cooling is expected from tropospheric industrial aerosols, which have increased in the past century (Penner et al. 1994; Penner et al. 1995), the amount of aerosols injected into the stratosphere by volcanoes decreased for much of the twentieth century (Lamb 1977; Sato et al. 1993; Robock and Free 1995), contributing to surface warming by allowing sunlight to warm the earth’s surface, unobstructed by atmospheric aerosol scattering, reflection, or absorption.
Changing atmospheric ozone concentrations further complicates the interpretation of recent climate change. Surface cooling is expected in the past few decades from the depletion of stratospheric ozone by CFCs since ozone is a greenhouse gas (Schwarzkopf and Ramaswamy 1993) as well as Earth’s biological ultraviolet shield (de Gruijl 1995). At the same time, anthropogenic tropospheric ozone increases contribute to surface warming (Lacis et al. 1990). Since changing solar UV radiation modulates ozone concentrations (Hood and McCormack 1992; Hood et al. 1993; Chandra and McPeters 1994; Haigh 1994; Hood 1997), the ozone influence on climate exhibits both natural and anthropogenic components. Surface albedo is changing too, in response to altered land use patterns (Hannah et al. 1994).
Climate responds differently to individual forcings—greenhouse gases, aerosols, ozone, solar variability—because the forcings have distinct regional and altitude distributions, different temporal histories, illustrated in Fig. 2, and different magnitudes, summarized in Fig. 3. Of the various natural and anthropogenic climate forcings, only that by greenhouse gases (particularly CO2) is reasonably well specified over the past 150 yr. Sometimes construed from this lack of knowledge of other influences is the assumption that none but greenhouse gases are significant for recent climate change. However, assuming climate sensitivity κ is in the accepted range 0.3–1.0°C per W m−2, an interpretation of surface warming over the past century based solely on greenhouse gas forcing is inconsistent with observations. A temperature increase ΔT = κΔF in the range 0.7°C–2.4°C is expected from a greenhouse gas forcing of ΔF = 2.4 W m−2 over the past 150 yr, in excess of the actual observed 0.55°C warming. Clearly, present knowledge of climate sensitivity is inadequate for translating greenhouse gas radiative forcing to surface warming. This implies that other forcings, though poorly specified, may be important for interpreting recent climate change. Even the speculated net climate forcing of 1.2 W m−2 shown in Fig. 3 for the past 150 yr predicts an average surface warming somewhat higher than the actual observations, except for low climate sensitivity.
Similarities of surface temperature and solar variability records over decadal timescales, evident in Fig. 2, point to solar forcing as a contributor of recent climate change. Surface temperatures since 1850 are shown in Fig. 4 to correlate well with solar activity (Reid 1991;Friis-Christensen and Lassen 1991; Hoyt and Schatten 1993), with correlation coefficients as high (0.7) as the correlation between surface temperatures and greenhouse gas concentrations (Kelly and Wigley 1992). However, accounting for the entire surface warming by direct solar radiative forcing alone is unlikely: greenhouse gases have increased over this time period, and simulations of a solar warming scenario with simple climate–ocean energy balance models require solar irradiance variations larger than the 0.1% 11-yr irradiance cycle evident in the contemporary solar monitoring database or climate sensitivities in excess of present estimates (Schlesinger and Ramankutty 1992; Kelly and Wigley 1992).
Understanding solar influences on climate requires improved specification of both the amplitudes and timescales of solar radiative output changes on climatological timescales and the climate sensitivity to small insolation changes. Space-based solar monitoring has documented unequivocally the existence of an 11-yr cycle in the primary energy provided from the sun to the earth (its total radiative output). The possibility of larger amplitude changes over longer timescales that might physically account for significant climate change cannot be dismissed. Knowledge of climate response to the sun’s changing solar radiation is rudimentary, encompassed in simple processes that fail to explain climate change observed on timescales from the past century to the 100 000-yr Milankovitch forcing (Rind et al. 1989;Phillips and Held 1994). Present inability to quantify climate forcing by changing solar radiation, whether negligible or significant, is a source of uncertainty that impacts policy making regarding global climate change (George C. Marshall Institute 1989; Houghton et al. 1995).
2. Solar radiative output variability
The sun, whose surface temperature is near 6000 K, provides electromagnetic, particle, and plasma energy to the earth at levels summarized in Table 1. Electromagnetic radiation is by far the largest solar energy source for the earth and the most important for its climate. When the solar spectrum shown in Fig. 5a is integrated over all wavelengths, the total radiative output from the sun at the earth is 1366 W m−2 (with a measurement uncertainty of ±3 W m−2).
There is no doubt that the sun is an active star. Sunspots exhibit a prominent 11-yr cycle, first reported by Schwabe in 1843, and also shorter (e.g., 27-day rotational) and longer (e.g., 88-yr Gleisburg) cycles. Widely adopted as a proxy of solar activity in many geophysical investigations, the sunspot number is in fact but one record, albeit the longest, of myriad other variable solar magnetic phenomena and their modulation of radiation, particle, and plasma energy outputs. Both the total radiative output and the spectral shape of the radiation change during the sun’s activity cycle, as shown by the 11-yr variability amplitudes in Fig. 5b, with the shortest wavelengths varying by many orders of magnitude more than the visible radiation.
Radiation at near-UV, visible, and near-IR wavelengths comprises the bulk of the total radiative output;the sun emits 48% of its radiation at wavelengths between 400 and 800 nm. This radiation is directly available to the earth’s surface and troposphere (see Fig. 5a spectrum at 0 km). Shorter wavelength UV radiation and solar energetic particles deposit their energy in the earth’s atmosphere, mainly above the troposphere. Atmospheric gases—O2, N2, O, and O3—are strong absorbers of UV radiation. Solar radiative energy at wavelengths increasingly shorter than 300 nm varies with increasingly larger amplitudes and is deposited increasingly higher above the earth’s surfaces, at altitudes of unit optical depth (the altitude at which vertically incident sunlight is attenuated to 1/e of its flux on top of the earth’s atmosphere) shown in Fig. 5c.
a. Contemporary observations
During the past 17 yr, solar monitors on earth-orbiting spacecraft have detected the sun’s changing levels of total (spectrally integrated) and UV spectral radiation throughout the 11-yr activity cycle (Rottman 1988; Willson and Hudson 1991; Hoyt et al. 1992; London et al. 1993; Lee et al. 1995; Chandra et al. 1995; Woods et al. 1996; Lean et al. 1997). The data in Fig. 6 provide irrefutable evidence of the 11-yr total irradiance cycle—the solar “constant” is not, in fact, constant. When solar activity is high, as indicated by the sunspot number, so too are the total and UV radiative outputs from the sun (see, e.g., reviews by Lean 1991, 1997). Additional radiative output variations have been detected on timescales of minutes, days, months, and years, superimposed on the overall 11-yr cycle (Hudson 1988; Fröhlich et al. 1991; Fröhlich et al. 1997). Evident in Fig. 6, for example, are excursions of a few tenths percent associated with the sun’s 27-day rotation, superimposed on the more slowly varying 0.1% 11-yr irradiance cycle. The present database is yet too short to resolve the amplitudes of longer period irradiance changes that may be occurring as well (Lee et al. 1995).
Knowledge of the 11-yr irradiance cycle is presently imperfect. Compounded with uncertainties arising from the limited duration of space-based solar monitoring, which barely exceeds one 11-yr cycle, are instrumental uncertainties that cause spurious variability signals in individual solar radiometers. Radiometer sensitivity changes are often pronounced during the first year of space deployment because the solar flux incident on the instrument modifies surface contaminants, altering the spectral absorptivity and reflectivity of the instrument’s optical components (Luther et al. 1986). This may explain the pronounced decrease in 1979 of the Earth Radiation Budget (ERB) measurements made from the Nimbus-7 spacecraft (Hoyt et al. 1992). Likewise, instrumental changes may have caused the distinct upward trend in 1992 at the beginning of the Active Cavity Radiometer Irradiance Monitor (ACRIM) measurements on board the Upper Atmosphere Research Satellite (UARS), which are not replicated by either the simultaneous ERB or Earth Radiation Budget Satellite (ERBS) data, both of which had been operating for much longer. Instrumental factors possibly caused by sensitivity changes related to temperature or aspect drifts are also implicated in the discrepancies between ERB and ERBS data during 1990–92, near the peak of solar cycle 22 (Lee et al. 1995). But despite these individual instrumental discrepancies, the basic features of the 11-yr irradiance cycle and higher frequency variability emerge clearly, by virtue of the multiple, overlapping, cross-calibrated measurements that comprise the extant database.
The sun’s irradiance fluctuates because, as illustrated in the solar images in Fig. 7, radiation sources are not homogeneously distributed on its disk. Magnetic fields erupting from the solar convection zone into the overlying solar atmosphere generate active regions and complexes in which the local radiation is altered relative to the background solar disk. Both dark sunspots and bright plages are evident in Fig. 7. Magnetic activity erupts, evolves and decays at different rates throughout the 11-yr cycle, generating sunspots, plages, and faculae and a bright emission network that continuously modulate solar total and spectral radiative output (see, e.g., Foukal 1990; Lean 1991), demonstrated in Fig. 7 by the irradiance time series.
Dark sunspots on the solar disk reduce total radiative output (e.g., on 10 February 1993, the right-hand image in Fig. 7) because their emission is less than that of the surrounding disk (Hudson et al. 1982). As the sun rotates, the sunspots appear to move across and off the face of the disk projected toward the earth, modulating solar total irradiance by as much as a few tenths percent on timescales of days to weeks. Although sunspots are the prime cause of 27-day rotational modulation, they are not the only source of irradiance variability; if they were, then the sun would be dimmer instead of brighter during times of high activity when sunspots occur much more frequently, which is not the case (see Fig. 6).
Magnetic regions where emission is enhanced (rather than depleted, as in sunspots), called plages and faculae also contribute to rotational modulation, as shown in Fig. 8, because they too are inhomogeneously distributed on the face of the sun. These regions are evident as complexes of bright emission called plages in the Ca solar images in Fig. 7. When viewed in visible emission they are identified as faculae, the photospheric footprints of the chromospheric plages. As well, small-scale elements form a network of bright emission over the solar disk that changes throughout the solar cycle but is relatively well dispersed in heliographic latitude, contributing minimally to rotational modulation but significantly to solar cycle variability. Bright facular and network emission variations more than compensate for sunspot darkening over the longer timescales of the 11-yr cycle (Foukal and Lean 1988; Lean et al. 1998), as shown in Fig. 9. Although faculae have less magnetic flux than in spots, they extend over considerably more of the sun’s disk and persist longer. Changes in global solar structure separate from the magnetic sunspots, faculae, and network are speculated to also affect radiative output (Kuhn and Libbrecht 1991; but see also Solanki and Unruh 1998 for counterarguments). While the actual detailed identification of the bolometric (spectrally integrated) brightness sources of solar radiative output variability is not yet complete, it is now well established that the net result is a solar cycle total irradiance modulation approximately twice that of sunspots.
The shape of the entire solar radiation spectrum also varies with solar activity. The competing effects of dark sunspots and bright faculae, whose bolometric variations are shown in Fig. 9, change as a function of wavelength, with the result that the amplitude of the spectrum variability is wavelength dependent (Fig. 5b). Total radiative output typifies the behavior of radiation at visible wavelengths (where solar spectral flux peaks; see Fig. 5a) for which the 11-yr variation in bright facular emission is about twice the sunspot emission depletion (Fig. 9). Progressing to shorter UV wavelengths, the facular brightening becomes increasingly larger relative to sunspot darkening (Lean 1989; Lean et al. 1997). For example, facular brightness variations control almost entirely the radiation variability at 200 nm (Fig. 7) over both the 27-day and 11-yr cycles, with little detectable sunspot modulation.
b. Historical reconstructions
Lack of direct solar monitoring for all but the last 17 yr, an extremely short epoch climatologically, motivates the reconstruction of historical irradiance variations from proxy records of solar activity that are available over much longer epochs. The reconstructions rely on proper identification of irradiance variability sources in the contemporary solar monitoring database and parameterization of the variability of these sources using the proxy records.
Clearly, the net solar total irradiance shown in Fig. 6b varies approximately in phase with solar activity, shown in Fig. 6a, but the connection of irradiance and sunspot number is not direct. Figure 9 shows that the amplitude of the irradiance variability in each 11-yr cycle depends on the relative strengths of the sunspot and facular irradiance modulation, which can each have different relationships to the sunspot number in different activity cycles. A reconstruction since 1874 of solar total irradiance modulation by the 11-yr cycle in which sunspot darkening and facular brightening are parameterized separately is shown in Fig. 10. Sunspot darkening is calculated directly from white light observations of sunspot areas and locations made primarily by the Greenwhich Observatory (following Foukal 1981). Facular brightening is parameterized from observations by ACRIM on the Solar Maximum Mission (SMM) spacecraft by correlating monthly mean values of the measured irradiance corrected for sunspot darkening (called the residual irradiance) with monthly mean sunspot numbers (following Foukal and Lean 1990). Physically, this correlation reflects the occurrence of large facular complexes in the vicinity of magnetic active regions that also contain groups of sunspots.
In contrast to the sunspot number record of historical solar activity, which has a pronounced 11-yr cycle but no perceptible overall long term trend, shown in Fig. 11 (top panel) is a high resolution time series of solar activity recorded by 10Be cosmogenic isotopes in a Greenland ice core (Beer et al. 1988; Beer et al. 1994). An 11-yr cycle superimposed on a longer-term variability component is evident in 10Be, and also in the geomagnetic solar activity indices since 1874 (Joselyn 1995). Ice core 10Be and tree ring 14C (Stuiver and Braziunas 1993) are formed by galactic cosmic ray ionization of gases in the earth’s atmosphere that are subsequently incorporated in tropospheric processes. Solar activity modulates cosmogenic isotope production because magnetic coupling between the sun and the earth facilitated by the solar wind plasma is more complex when solar activity is high, inhibiting the flow of galactic cosmic rays to the terrestrial environment and reducing the concentrations of cosmogenic isotopes relative to times of low solar activity. Since solar activity modulates the sun’s radiative output, as evidenced during the 11-yr cycle, irradiance changes might also accompany the longer-term solar activity changes recorded by the cosmogenic isotopes (Fig. 11, bottom panel).
In addition to an 11-yr activity cycle, cosmogenic isotope archives exhibit periods near 88 (the Gleisburg cycle), 210, and 2300 yr (Stuiver and Braziunas 1993;Beer et al. 1994). Furthermore, both tree ring 14C and ice core 10Be data in Fig. 11 document a steady increase in solar activity since the seventeenth century Maunder Minimum (McHague and Damon 1991), perhaps reflecting the net response to solar activity modulation at these longer period cycles. A secular decrease in solar radius of the order of 0.1 arcsec, century−1 over the past few centuries (Gilliland 1981; Sofia and Fox 1994; Fiala et al. 1994) may be related to the overall changing solar activity. What amplitude irradiance change might have occurred as well?
Long-term monitoring of ionized Ca emission, a surrogate for magnetic activity in the sun and stars, provides additional evidence that the sun’s radiative output may compose a longer-term variability component in addition to the 11-yr cycle. Bright Ca emission seen in the solar images in Fig. 7 is also detected in sunlike stars, and fluctuations in Ca emission occur in some sunlike stars on decadal timescales analogous to the 11-yr solar cycle (Wilson 1978). Some stars, however, have no apparent cycle during the few decades over which they have been monitored. In these stars, the Ca emission is reduced below that of the cycling stars, suggesting perhaps conditions analogous to the sun’s Maunder Minimum of anomalously low activity (Baliunas and Jastrow 1990).
Quantitative comparison of solar and stellar brightness changes seen in ionized Ca emission provides a tool for estimating amplitudes of longer term solar radiative output variations. Compared with 13 sunlike stars, the range of Ca emission from the sun during its contemporary 11-yr cycle places it in the brightest one-third of the distribution in Fig. 12 (White et al. 1992). The Ca emission variations on the sun closely track the facular brightness variations that control long-term solar radiative output. When the ACRIM total irradiance observations are corrected for sunspot darkening, the residual irradiances correlate strongly with simultaneously measured Ca fluxes (Livingston et al. 1988), and Ca and UV fluxes also correlate highly (White et al. 1990). Extending the linear relationship between residual total irradiance and solar Ca emission demonstrated in Fig. 12 to the lower solar Ca values inferred from stellar monitoring for noncycling stars predicts a solar total irradiance reduction of 0.24% during the Maunder Minimum relative to present-day mean levels (Lean et al. 1992). Accompanying reductions in solar UV irradiance may have exceeded (by about a factor of 2) the 11-yr cycle amplitudes of 3% at 250 nm and 7% at 200 nm (Lean et al. 1995a). The proposed mechanism is depletion of the bright facular network that normally covers the sun, even during times of present-day minima of the 11-yr cycle, and an additional reduction in the emission from the nonnetwork regions to values presently seen in only the 15% darkest regions on the solar disk (White et al. 1992). Other studies also indicate solar irradiance reductions during the Maunder Minimum from 0.2% to 0.6% below present-day values (Hoyt and Schatten 1993; Nesme-Ribes et al. 1993; Zhang et al. 1994). Independent observations of apparent luminosity changes in sunlike stars likewise estimate that larger irradiance changes are possible than evident in the present day solar monitoring database—exceeding by factors of 2 to 5 the 0.1% 11-yr solar total irradiance cycle (Lockwood et al. 1992).
The reconstruction of solar total irradiance over the past four centuries shown in Fig. 13 assumes a 0.24% solar irradiance reduction during the seventeen century relative to present-day levels. In contrast to the irradiance reconstruction shown in Fig. 10, which accounts only for the 11-yr irradiance cycle, the reconstruction in Fig. 13 combines separately determined 11-yr and longer term variability components (Lean et al. 1995b). After 1874, the total irradiance cycles are those shown in Fig. 10; prior to 1874 (when the Greenwhich data needed to calculate sunspot darkening were not recorded) the 11-yr cycle is reconstructed using directly correlated yearly mean total irradiance and group sunspot numbers (see, e.g., Schatten and Orosz 1990). Following the demonstration by Foukal and Lean (1990) that facular brightening tracks the monthly mean sunspot number throughout the 11-yr cycle, on longer timescales the network facular emission is assumed to likewise track the overall level of solar activity, and the shape of these changes is specified by the average amplitude of the group sunspot number (Hoyt et al. 1994) in each cycle. This longer-term component is scaled to cause an increase of approximately 0.2% in total solar irradiance and 0.97% in UV (200–300 nm) irradiance from the Maunder Minimum to the present. Including the 11-yr activity cycle, overall variability from the Maunder Minimum to the present-day mean is thus constrained to agree with a total solar irradiance change of 0.24% (Lean et al. 1992).
The total irradiance reconstruction in Fig. 13 that includes both an 11-yr cycle and a longer-term variability component tracks independent records of solar activity levels inferred from 14C and 10Be cosmogenic isotopes (Stuiver and Braziunas 1993; Beer et al. 1994). However, this reconstruction differs somewhat from that of Hoyt and Schatten (1993), which has a long-term variability component based on the length (rather than amplitude) of the activity cycle. These differences, which cannot be resolved without improved understanding of the solar origins of the variations, reflect the large uncertainties in reconstructing historical solar irradiances from a limited solar monitoring database, with only rudimentary knowledge of the pertinent physical processes.
3. Statistical sun–climate connections
Similarities among various climate and solar activity records (e.g., Figs. 2 and 4) suggest that climate variability in the recent Holocene may be partly attributable to the variable sun. Some climate records have periodicities at 11 and 22 yr that are common also in solar activity proxies. Other climate records appear to correlate well with long-term solar activity on decadal to century timescales. Some of these relationships are summarized below. That not all climate time series exhibit this anecdotal evidence for solar forcing is usually interpreted as evidence to reject the hypothesis of a sun–climate connection, leading to present ambiguity about the physical reality of the effect. Resolving this ambiguity requires proper identification of physical mechanisms to explain the cycles and correlations.
Climate records exhibit a range of periods, some of which are listed in Table 2 (adapted from Burroughs 1992), none of which reflect truly deterministic climate cycles. Presumably arising from nonstationary processes, climate periodicities exhibit variable phase and amplitude, appearing only in some climate proxy records, in certain geographical regions, in some epochs, and not always in phase with their surmised forcing mechanisms. Burroughs (1992) (see also Hoyt and Schatten 1997) interprets the high occurrence of the cycles listed in Table 2 simply as recognition of potential climate variability modes. Recent analyses of the global surface temperature record since 1860 lends some support for this interpretation by identifying climate cycles in the range 2–8 yr and 11–12 yr with confidence limits in excess of 90% (Mann and Park 1994). However, a robust statistical description of climate variability and of the significance of all of the peaks in Table 2, especially those with longer periods, has proven elusive, even when potential physical mechanisms are identified. Rather, processes internal to the climate system likely cause a “substantial share of the variability of climate” (Mitchell 1976) with external climate forcing processes contributing additional variability to this stochastic background.
Periodicities in the range 2–7 yr are attributed with some confidence to internal oscillations in the coupled ocean–atmosphere system; those at 2–3 yr are connected to the quasi-biennial oscillation (QBO) in tropical stratospheric winds, and those from 3–7 yr to the El Niño–Southern Oscillation (ENSO) (see, e.g., Mann and Park 1994; Dunbar et al. 1994). Although decadal and multidecadal periods are common in climate time series, their physical origins are difficult to specify, partly because of the paucity of global coverage by long-term climate records with high (annual) resolution. Periods near 18 and 34 yr are often connected to lunar forcing. Since sunspot numbers exhibit a pronounced 11-yr cycle, periods near 11 and 22 yr are tentatively connected with solar variability, often for lack of another plausible mechanism. But the transitory nature of this period and its absence in some climate records lead Pittock (1979) to dismiss the implied sun–climate connection as unconvincing. Recognition that the solar radiative output varies does imply a potential mechanism for excitation of these periods by solar forcing, a scenario that was harder to formulate when solar irradiance was assumed to be constant. Alternative mechanisms are also postulated for the decadal periodicities, in particular an internal oscillation of the climate system that is present in some climate model simulations in the absence of external forcing (James and James 1989; Mehta and Delworth 1995).
As noted above, spectral analysis of the 14C record indicates apparent fluctuations in periods ranging from less than 100 to several thousand years, including the 88-yr (Gleisberg) cycle, and ∼210 and ∼2300 yr cycles. Hints of these cycles have also been identified in the climate record, for example, the 2500-yr cycle in marine, glacier, and polar ice core records (Denton and Karlen 1973; Dansgaard et al. 1984; Pestiaux et al. 1987;Wigley and Kelly 1990), and the 88- and 200-yr cycles in varved sediments (Halfman and Johnson 1988; Peterson et al. 1991; Anderson 1992). Whether these are truly periodic in nature is doubtful, but they are indicative of a general coincidence of apparently increased 14C production (i.e., lower solar activity) and colder temperatures (de Vries 1958; Eddy 1976), with the Maunder Minimum–Little Ice Age being the most recent example.
Episodes of correlated climate and solar variability occur throughout the Holocene, on timescales ranging from subdecadal to multicentennial.
Tropospheric temperatures during the past few decades appear to be about 0.5° to 1.5°C warmer during times of cycle maxima, notably in the midlatitude Western Hemisphere (Labitzke and van Loon 1993a,b). A 0.15°C increase in land-surface temperatures from 1986 to 1990 has been attributed to increasing solar irradiance from cycle 22 minimum to maximum activity (Ardanuy et al. 1992). Sea surface temperatures bandpassed to isolate the decadal component of their variability exhibit changes of the order of 0.1°C that are highly correlated (correlation >0.9) with the sun’s 11-yr activity cycle in the past four decades (White et al. 1997). Coral records of δ18O infer that the relationship between the sun’s 11-yr cycle and sea surface temperatures extends over the past 400 yr (Dunbar et al. 1994). Furthermore, solar-related sea surface temperature changes may initiate regional precipitation fluctuations (Perry 1994). High resolution ice core records provide further evidence for apparent correlations of climate parameters with the sun’s 11-yr cycle both at midlatitude high-altitude sites, (Thompson et al. 1993) and in the high-latitude Greenland GISP2 (Grootes and Stuivers 1997) ice core. Furthermore, correlations of climate parameters with the 11-yr cycle may be enhanced significantly when the climate data are sorted according to the phase of the QBO, as illustrated in Fig. 14 (Tinsley 1988; Barnett 1989; Labitzke and van Loon 1990).
Figure 4 demonstrates the correlation of globally and hemispherically averaged surface temperatures and solar activity over multidecadal timescales in the past 140 yr, and the data in Fig. 15 (Eddy 1976, 1977) extend this apparent sun–climate correction to the past few centuries, using solar variability inferred from cosmogenic isotopes. During the past 9000 yr, climate minima identified in a composite of glacial advance and retreat records correspond to six out of the seven lowest levels of solar activity represented by peaks in the 14C isotope data (Wigley and Kelly 1990).
Despite their ubiquity, correlations among climate and solar parameters remain tenuous paradigms for assessing the range of natural variability possible for the earth’s climate, against which to gauge the extent of natural versus anthropogenic variability in the recent century. Neither climate nor solar variability are sufficiently well defined, either spatially or temporally, nor their causes adequately understood, to verify that the correlations really arise from climate forcing by changing solar radiation rather than from statistical coincidence (Baldwin and Dunkerton 1989; Salby and Shea 1991). In the recent Holocene, for example, the correlations arise, for the most part, from the coincidence of the Little Ice Age from roughly 1450 to 1850 with the Spörer and Maunder minima in solar activity and of the thirteenth century Medieval Warm Period with the Medieval Maximum of solar activity. Yet neither the Little Ice Age (Bradley and Jones 1993) nor the Medieval Warm Period (Hughes and Diaz 1994) is a quantitatively well-characterized climatic episode; nor are these events prominent in all climate records (Briffa et al. 1990).
Inferences from sun–climate correlation studies can depend critically on the type and length of the climate and solar variability records chosen for the study. Although in the past 130 yr the Hoyt and Schatten (1993) irradiance reconstruction correlates well with NH surface temperatures (correlation of 0.8), from 1700 to 1990 its correlation with the Bradley and Jones (1993) NH temperature data drops to 0.5 (Crowley and Kim 1995). A higher correlation of 0.7 is obtained with the Bradley and Jones (1993) NH temperature record using the irradiance reconstruction in Fig. 13 in which the longer-term component is based on the amplitude, rather than length of the 11-yr cycle (Lean et al. 1995b). Yet when annual sunspot numbers (SSN) (which lack the longer-term component evident in cosmogenic isotopes and the irradiance reconstructions, Fig. 13) are used as a proxy for solar activity “the influence of SSN on global temperatures is found to be negligible” (Visser and Molenaar 1995).
Changing strengths of sun–climate correlations in different epochs may reflect real changes in the relative amplitudes of the various climate forcings, for example, prior to and during the industrial epoch. This is demonstrated in Table 3 by the correlation of the decadal means during the past four centuries of data similar to those in Fig. 2 (the sun’s total irradiance, the volcanic dust veil index, CO2 concentrations, and surface temperature anomalies). From 1610 to 1800 the correlation of reconstructed solar irradiance and NH temperature is 0.86, whereas the correlation of NH dust veil index with surface temperature is −0.005, implying a predominant solar influence in this preindustrial period. From 1800 to 2000 the correlation of surface temperature and the NH dust veil index is stronger (−0.51), reflecting extended volcanic activity in the nineteen century. Since 1800, both greenhouse gas concentrations and solar activity have steadily increased (although at different rates) whereas volcanic activity declined in the twentieth century relative to the nineteenth century. From 1800 to 2000, surface temperatures correlate more strongly with CO2 levels than with reconstructed solar irradiance, in contrast to the prior two preindustrial centuries.
Contrary to the implications of Table 3 that solar forcing may account for a significant fraction of recent climate variability, Robock (1979) suggests that in fact volcanic dust forcing produces the best simulation of climate change in the past 400 yr. Furthermore, his study found no evidence for a solar influence on climate during the Maunder Minimum, in part because the reconstructed temperature time series used was less cold during that epoch than is the Bradley and Jones (1993) reconstruction (e.g., Fig. 1). Other studies that utilize the Crête (Greenland) ice core acidity record have likewise inferred a strong volcanism signal in climate variations of the last 1400 yr, a correspondence that, however, is less impressive when background acidity levels of presumably nonvolcanic origin are removed (Crowley et al. 1993). Whereas clusters of intense volcanism might occasionally cause a decadal-scale climate excursion, Crowley and Kim (1995) attribute 30% to 55% of climate variability on decadal–centennial timescales to solar variability.
Assuming that climate forcing by changing solar radiation S is responsible for the surface temperature variability ΔT from 1610 to 1800, a linear relationship of ΔT = −168.666 + 0.1233 × S is deduced from the preindustrial data (update from Lean et al. 1995b) and extended in Fig. 16 to the industrial period from 1800 to 2000. According to the simple preindustrial parameterization, surface temperature changes arising from changing solar radiation may have contributed about half of the 0.55°C warming since 1900, but since 1970 no more than one-third of the 0.35°C warming is attributable to the variable sun.
4. Simulations of climate response to changing solar radiation
Until physical causal mechanisms are identified to explain the apparent associations of solar and climate variability it will be difficult to prove or disprove that these associations arise from climate forcing by changing solar radiation, rather than from other mechanisms such as internal oscillations, or simply from statistical perturbations from a postiori choices. Uncovering potential mechanisms for decadal and centennial climate change requires improved specification of the climate system’s response to individual and combined radiative forcings, and simulations of the expected influence of realistic solar radiation changes over these timescales.
a. Equilibrium simulations
Although the climate system response to radiative forcing likely depends on the strength, history, geographical distribution, and attitudinal localization of the specific forcing, these relationships are poorly quantified. In practical applications, climate sensitivity is specified to be in the range κ = 0.3°–1°C per W m−2, such that an equilibrium temperature change ΔT = κΔF (°C) results from a radiative forcing of ΔF (W m−2). Changes in solar radiation ΔS cause radiative forcing ΔFs = ΔS × 0.7/4 where the factor 0.7 accounts for the reflection back to space of a portion of the incident solar energy (the albedo) and the factor 4 is the spherical average. With this simple prescription, an equilibrium surface temperature change in the range 0.07°–0.24°C is estimated to result from the total irradiance change in solar cycle 21 (ΔS = 1.1 W m−2, Fig. 10) (Wigley and Raper 1990) and a larger temperature change in the range 0.17 to 0.57°C is estimated for the speculated longer-term irradiance change of 0.24% (ΔS = 3.3 W m−2) from the seventeenth century to the present. Consistent with this an equilibrium simulation by the Goddard Institute for Space Studies (GISS) general circulation model (Hansen et al. 1983)—whose sensitivity is in the range 0.7°–1°C per W m−2—estimates a global surface temperature decrease of 0.47°C for a 0.25% solar irradiance decrease (Rind and Overpeck 1993). An additional complicating feature is that κ may be different for decadal and century scale perturbations, with the longer-term forcing exciting more oceanic response and system feedbacks (Hansen et al. 1985).
An important feature of the GISS simulation shown in Fig. 17, however, is that although a 0.25% reduction in solar irradiance causes 0.47°C global cooling, some geographical locations cool and others warm by more than 1°C as a result of dynamical circulation patterns driven by the differential heating of the land and the oceans (Fig. 18, top panel). Although many of the changes in Fig. 17 simulated by the GISS model were of the same order as the model standard deviation, especially at high latitudes, the geographical patterns were verified to some extent by comparing two different time slices of the simulation after the model had reached equilibrium (Rind and Overpeck 1993). The model was integrated for 60 model years using a version that allowed sea surface temperatures to change while keeping ocean heat transports specified at current values (Hansen et al. 1984). The differences in surface air temperatures shown in Fig. 17 were estimated during years 36–45 of the experiment, and a similar time period for the current climate control run, and are generally similar to estimates from later years.
Equilibrium simulations of climate response to changing solar radiation using the Laboratoire de Meteorologie Dynamique (LMD) atmospheric GCM estimate surface temperature reductions of 1.5°C for the seventeenth century Maunder Minimum, for a 0.4% irradiance decrease—a 320% larger temperature reduction than the GISS simulations for a 60% larger irradiance change—with minimal geographic inhomogeneity (Nesme-Ribes et al. 1993). Perhaps the factor of 3 larger cooling, a result of the high sensitivity of the LMD model, was sufficient to overcome the regional advective patterns initiated by the differential land–ocean heating for less cooling; different geographical locations may exhibit quite different responses to small changes in solar radiative forcing, depending on the magnitude of the forcing. As a result of reduced solar irradiance in the seventeenth century, the GISS and LMD simulations also predicted changes in other climate parameters, including evaporation minus precipitation shown in Fig. 18 (bottom panel).
Simulations of the climate response to changing solar radiation with energy balance models or GCMs typically utilize only the total changes in solar radiative output and ignore the spectral dependence of this variability (Pollack et al. 1979). A related omission in these simulations is that solar modulation of the ozone layer and the middle atmosphere are not accounted for, thus neglecting possible climate forcing by radiative and dynamical coupling of the middle atmosphere with the troposphere (Haigh 1994, 1996). Both total column ozone amount and the altitude distribution are known to be affected by solar UV radiation at wavelengths less than 300 nm that is absorbed in the earth’s middle atmosphere (see Fig. 5) (Hood and McCormack 1992; Reinsel et al. 1994; Chandra and McPeters 1994; Hood 1997; see also NRC 1994).
Ozone concentration changes may, as shown in Fig. 19, either cool or warm the earth’s surface depending on the altitude of the changes as both short wave absorption and infrared forcing are affected (Lacis et al. 1990; Schwarzkopf and Ramaswamy 1993; Haigh 1996). Stratospheric variations in response to changing solar UV radiation and ozone, also shown in Fig. 19, may affect the troposphere and climate by altering tropospheric dynamics, as suggested both by observations and modeling studies. Some results of tropospheric variations associated with the quasi-biennial oscillation and UV variations from the studies of Labitzke and van Loon are given in Fig. 14. A recent set of modeling studies found that UV variations imparted to a global climate–middle atmosphere model in conjunction with the QBO did have significant effects on both the stratosphere and troposphere (Balachandran and Rind 1995;Rind and Balachandran 1995). The mechanisms involved are summarized in Fig. 20. The QBO alters the meridional gradient of the zonal wind in the lower stratosphere, and UV variations alter the vertical gradient of the zonal wind. Both change the refraction properties of planetary waves in the stratosphere, resulting in dynamically induced warm–cool regions in the stratosphere. This thermal response then alters the vertical stability of the troposphere–stratosphere system and in the model affects tropospheric planetary wave generation, especially for the longest planetary waves.
In conjunction with altered planetary wave generation and propagation (hence weather), changes in cloud cover and atmospheric energy transport occur, altering the mean climate state and producing surface air temperature change. Shown in Fig. 21 is the change in surface air temperature between +5% UV and −5% UV radiation in the QBO west phase (top panel) and east phase (middle panel), averaged over 10 Januaries in the model. Distinctive warm and cool regions arise, which vary between the west and east QBO phase (Fig. 21, bottom panel). The model results are somewhat similar to observations, and imply that the sun’s UV radiative output changes may affect surface temperatures by acting through the stratosphere. When averaged over the solar cycle there is no guarantee that these perturbations will cancel, in contrast to assumptions that since the 11-yr activity forcing is cyclic, any effect is assumed to give a net zero forcing over each cycle. This prospective mechanism emphasizes that the potential complexity of the climate system should not be underestimated, and that caution is needed in translating the magnitude of climate forcing into a climate response (Rind and Balachandran 1995).
b. Time-dependent simulations
Equilibrium temperatures are only realized for climate forcings that persist over sufficiently long times relative to the time response of the climate system. Time-dependent simulations using energy balance models suggest that oceanic thermal inertia dampens the 11-yr cyclic solar forcing by roughly 80%: the expected solar-related surface temperature changes are in the range 0.02°–0.03°C, perhaps too small to be detected in the climate record (Wigley and Raper 1990) except with sophisticated statistical tools (North and Kim 1995; Stevens and North 1996). In response to the longer-term 0.36% irradiance increase since 1700 estimated by the Hoyt and Schatten (1993) reconstruction (Fig. 13), a recent energy balance model estimates surface temperature change in the range 0.2°C–0.3°C (Crowley and Kim 1995). Attempts by energy balance models to simulate climate change in the past 140 yr provide circumstantial evidence for forcing by greenhouse gases, sulfate aerosol, and solar variability, with greenhouse gas forcing dominant in the past century. Unambiguous conclusions are prevented in part by uncertainties in the amplitudes of the aerosol and solar irradiance changes (Kelly and Wigley 1992; Schlesinger and Ramankutty 1992).
Very recently, we have used the GISS 8° × 10° GCM to simulate the time-dependent climate response to solar forcing during the past four centuries, using as input the solar irradiance reconstruction in Fig. 13 (solid line). Power near 11, 22, and 60 yr present in the solar forcing time series is ambiguous in the simulated surface temperature response, although enhanced power is evident at the longer periods. Global surface temperature changes associated with the overall 0.24% irradiance increase from the seventeenth century Maunder Minimum to the present are of the order of the equilibrium simulation (i.e., in the range 0.3°–0.5°C). Five individual time-dependent simulations and associated control runs have been performed and the results are presently being analyzed. The runs were made in conjunction with National Oceanic and Atmospheric Administration (NOAA) Climate and Global Change and Paleoclimate programs, as part of the Analysis of Rapid and Recent Climate Change (ARRCC) effort to characterize climate change and responses to natural (solar, volcanic) and anthropogenic (greenhouse gas, aerosol) forcings during the past 400 yr. Planned is a detailed analysis of time series and geographic maps of transient surface temperature, cloud cover, precipitation, and other climate parameters and comparison of the simulations with available data to better assess the reality of solar and other influences on recent climate change, both globally and regionally.
5. Summary and discussion
Radiation from the sun—the earth’s energy source—varies continuously at all wavelengths and on all observed timescales. A change in total radiative output of about 0.1% has been measured between the maximum and minimum of the sun’s recent 11-yr activity cycle. Accompanying changes occur throughout the solar spectrum, with larger cycle amplitudes in radiation at UV wavelengths than in visible emissions. Although not measured directly, total radiative output changes of a few tenths percent are postulated to occur over centennial timescales, based on evidence from cosmogenic isotope proxies of solar activity and activity levels in sunlike stars, both of which exhibit a larger range of variability than yet evident in the present-day sun.
Response of the climate system to radiative forcing, including by the variable sun, is not yet understood with sufficient certainty to unambiguously interpret the earth’s surface warming over the past 140 yr. Presently specified climate sensitivity overpredicts the magnitude, and cannot replicate the shape of the surface warming expected from radiative forcing by greenhouse gases alone. This underscores the need to quantify all other natural and anthropogenic influences, including those of less magnitude than greenhouse gases, since the assumption of similar climate response to forcings of similar magnitude may not be valid.
In the recent past, correlations between solar variability and surface temperatures rival those between surface temperatures and greenhouse gases, and there is evidence of the 11-yr solar activity cycle in a variety of climate data. Neither the amplitude of the climate response to changing solar radiation, nor its temporal or geographic character has yet been established with the certainty needed to either validate or dismiss these observed sun–climate associations. Even if putative long-term changes in solar radiation do not occur with the magnitude of the reconstructed irradiances shown in Fig. 13, a full understanding of alternative physical mechanisms for decadal and centennial climate change is required both in the pre- and postindustrial epochs before the circumstantial evidence for solar forcing can be dismissed with impunity.
A significant limitation in reducing uncertainties in the climate response to the sun’s changing radiation is the lack of reliable observational knowledge of the amplitude of solar irradiance changes other than during one recent 11-yr cycle. Not well recognized, in particular, are limitations of the sunspot number record as a surrogate for irradiance variability. If longer-term irradiance variations are indeed occurring, they are not tracked by the sunspot number, which lacks a long-term trend, reflecting the failure of sunspots to account for solar variability mechanisms beyond those associated with activity complexes on the disk (e.g., the variations in the background facular emission from the network that is a postulated source of irradiance variability beyond the range exhibited by the contemporary sun, White et al. 1992). Since different spectral regions impact different aspect of the climate system, also necessary for proper specification of climate forcing by changing solar radiation is knowledge of the spectral dependence of the radiation, which remains thus far observationally undefined at wavelengths longer than 400 nm.
Definitive answers about long-term solar irradiance variability require continuous, uninterrupted solar monitoring by space-based instruments that are cross-calibrated to maintain adequate long-term precision over many decades. Seventeen years of solar monitoring is too short for anything but speculation about the reality and magnitude of long-term irradiance variability. Unless (until) the reality of longer-term solar irradiance variations is established it will be difficult to evaluate whether the correlations among solar and climate parameters are coincident or causal, or whether simulations of climate change using irradiance reconstruction like those postulated in Fig. 13 are relevant. Proposed reductions in environmental monitoring and the lack of access to space threaten to jeopardize the present solar monitoring database by a data gap (which would preclude cross calibration of existing and future data) or termination of the record; present plans are not yet sufficiently secure to ensure a reliable record of solar irradiance for the foreseeable future. Lacking the requisite long-term solar database, uncertainty about solar influences on climate change may well persist indefinitely. Although inferences have been made about long-term solar variability from comparisons with sunlike stars and cosmogenic isotope records, such circumstantial evidence is incapable of ever providing the certainty needed for policy making regarding the sun’s influence (or lack of) on global climate change (see, e.g., Foukal 1994 for a discussion of ambiguities in the interpretation of sunlike stars).
Assuming the validity of long-term solar irradiance variations deduced from monitoring of solar and stellar variability, seen in Fig. 13 to track the cosmogenic isotope changes, the reconstruction suggests that solar radiation changes may have been the predominant climate forcing during the seventeenth and eighteenth centuries. But according to a simple linear parameterization of surface temperature anomalies and solar irradiance based on this preindustrial relationship, less than one-third of the earth’s surface warming since 1970 is attributable to changing solar radiation. Presumably this reflects the increasing dominance in the past century of anthropogenic climate forcing relative to natural solar-induced variations. Cosmogenic isotope records suggest that contemporary solar activity levels are now approaching historically high levels, last seen in the thirteenth century medieval solar activity maximum (see Figs. 11 and 15). Activity levels of the present-day sun are as high as one-third of the brightest sunlike stars. This evidence from both cosmogenic isotopes and sunlike stars points to the likelihood of future solar activity falling to lower levels, rather than increasing. Extrapolation of periodicities present in cosmogenic isotope data infer that this decrease may commence around 2030 (Jirikowic and Damon 1994), after which time the detectability of greenhouse climate forcing should further improve relative to natural solar induced variability.
Even in the event that solar radiation changes are eventually well specified, additional work is needed to understand the impact of this radiation on the climate system in such a way that decadal and centennial variations are fully understood regionally and globally in different epochs. This will likely require time-dependent simulations over past centuries, in the present, and into the future, with GCMs appropriately coupled to middle-atmosphere models, utilizing realistic spectrally dependent variations with properly parameterized wavelength-dependent impact on the climate system. Importantly, analysis of paleoclimate data from over the globe must be integrated into the assessment of the results of the simulations. Such studies are only just beginning. Sun–climate studies in the future require cross-disciplinary endeavors such as promoted and conceived by PACLIM and ARRCC.
Data used in this paper were kindly provided by Doug Hoyt, Dick Willson, Ray Bradley, Juerg Beer, Minze Stuiver, and Melissa Free. Bill Marquette supplied the BBSO Ca images. The NSO produces the 1083-nm EW data. Dust veil index and CO2 data were obtained from the CDIAC, Oak Ridge, andUARS UV irradiance data from the GSFC DAAC. Matthew O’Donnell of the British Meteorological Office provided the IPCC temperature data. J. Lean appreciates ongoing discussion about solar variability with Peter Foukal, Dick White, Gary Rottman, and Andy Skumanich, and thanks Tim Baumgartner for the opportunity to participate in the 1995 PACLIM. Recent discussions with Warren White and Dan Cayan about sea surface temperatures are happily acknowledged. The Strategic Environmental Research and Development Program (SERDP) and NOAA provided partial funding support.
Corresponding author address: Dr. Judith Lean, Naval Research Laboratory, Code 7673L, Washington, DC 20375.