Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection Reconstructions of past variations of total solar irradiance

Figure 6.5:
Reconstructions of total solar irradiance (TSI) by Lean et al. (1995, solid red curve), Hoyt and Schatten (1993, data updated by the authors to 1999, solid black curve), Solanki and Fligge (1998, dotted blue curves), and Lockwood and Stamper (1999, heavy dash-dot green curve); the grey curve shows group sunspot numbers (Hoyt and Schatten, 1998) scaled to Nimbus-7 observations for 1979 to 1993.

As direct measurements of TSI are only available over the past two decades it is necessary to use other proxy measures of solar output to deduce variations at earlier dates. In the simplest type of reconstruction a proxy measure, such as sunspot number (Stevens and North, 1996) or solar diameter (Nesme-Ribes et al., 1993), is calibrated against recent TSI measurements and extrapolated backwards using a linear relationship. The various proxies vary markedly as indicators of solar activity. For example, over the past century, sunspot number and 10.7cm flux showed highest values at the solar maximum of 1958, whereas the aa index, which gives a measure of the magnitude of the solar magnetic field at the Earth, peaked during 1990. This is because whereas sunspot numbers return to essentially zero at each solar minimum the aa index shows 11-year cycles imposed on a longer-term modulation (Lean and Rind, 1998). Other terrestrially based indicators of solar activity recorded by cosmogenic isotopes in tree-rings and ice-cores also show longer term modulation. However, direct solar proxies other than the sunspot number cover too short a period to reliably detect such a trend. Thus, it is not clear which proxy, if any, can be satisfactorily used to indicate past values of TSI.

A more fundamental approach recognises that solar radiative output is determined by a balance between increases due to the development of faculae and decreases due to the presence of sunspots. Longer-term changes are also speculated to be occurring in the quiet Sun against which these variable active regions are set. The sunspot darkening depends on the area of the solar disc covered by the sunspots while the facular brightening has been related to a variety of indices. These include sunspot number (Lean et al., 1995), emission of singly ionised calcium (Ca II at 393.4 nm) (Lean et al., 1992), solar cycle length, solar cycle decay rate, solar rotation rate and various empirical combinations of all of these (Hoyt and Schatten, 1993; Solanki and Fligge, 1998).

In addition to estimates of the impact of active regions on TSI potential contributions due to the variation in brightness of the quiet Sun have been estimated (Lean et al., 1992; White et al., 1992). These were largely based on observations of the behaviour of Sun-like stars (Baliunas and Jastrow, 1990) and the assumption that during the Maunder Minimum (an extended period during the late 17th century during which no sunspots were observed) the Sun was in a non-cycling state. The various reconstructions vary widely in the values deduced for TSI during the Maunder Minimum relative to the present. Mendoza (1997) has pointed out that uncertainties in the assumptions made about the state of the Sun during that period could imply a range of between 1 and 15 Wm-2 reduction in TSI less than present mean values although most estimates lie in the 3 to 5.5 Wm-2 range. Figure 6.5 shows group sunspot numbers from 1610 to 1996 (Hoyt and Schatten, 1998) together with five TSI reconstructions. The sunspot numbers (grey curve, scaled to correspond to Nimbus-7 TSI observations for 1979 to 1993) show little long-term trend. Lean et al. (1995, solid red curve) determine long-term variability from sunspot cycle amplitude; Hoyt and Schatten (1993, black solid curve) use mainly the length of the sunspot cycle; the two Solanki and Fligge (1998) blue curves (dotted) are similar in derivation to the Lean et al. and Hoyt and Schatten methods. Lockwood and Stamper (1999, heavy dash-dot green curve) use an entirely different approach, based not on sunspot numbers but on the aa geomagnetic index, and predict somewhat larger variation over individual cycles but less on the longer term. Clearly, even disregarding the shifts due to absolute scaling, there are large differences between the TSI reconstructions. Thus knowledge of solar radiative forcing is uncertain, even over the 20th century and certainly over longer periods.

Figure 6.6: Global, annual mean radiative forcings (Wm-2) due to a number of agents for the period from pre-industrial (1750) to present (late 1990s; about 2000) (numerical values are also listed in Table 6.11). For detailed explanations see Section 6.13. The height of the rectangular bar denotes a central or best estimate value while its absence denotes no best estimate is possible. The vertical line about the rectangular bar with “x” delimiters indicates an estimate of the uncertainty range, guided by the spread in the published values of the forcing and physical understanding. A vertical line without a rectangular bar and with “o” delimiters denotes a forcing for which no central estimate can be given owing to large uncertainties. The uncertainty range specified here has no statistical basis and therefore differs from the use of the term elsewhere in this document. A “level of scientific understanding” (LOSU) index is accorded to each forcing, with H, M, L and VL denoting high, medium, low and very low levels, respectively. This represents our subjective judgement about the reliability of the forcing estimate, involving factors such as the assumptions necessary to evaluate the forcing, the degree of our knowledge of the physical/chemical mechanisms determining the forcing, and the uncertainties surrounding the quantitative estimate of the forcing (see Table 6.12). The well-mixed greenhouse gases are grouped together into a single rectangular bar with the individual mean contributions due to CO2, CH4, N2O, and halocarbons (see Tables 6.1 and 6.11) shown; halocarbons refers to all halogen-containing compounds listed in Table 6.1. “FF” denotes fossil fuel burning while “BB” denotes biomass burning aerosol. Fossil fuel burning is separated into the “black carbon” (bc) and “organic carbon” (oc) components with its separate best estimate and range. The sign of the effects due to mineral dust is itself an uncertainty. The indirect forcing due to tropospheric aerosols is poorly understood. The same is true for the forcing due to aviation via their effects on contrails and cirrus clouds. Only the first type of indirect effect due to aerosols as applicable in the context of liquid clouds is considered here. The second type of effect is conceptually important but there exists very little confidence in the simulated quantitative estimates. The forcing associated with stratospheric aerosols from volcanic eruptions is highly variable over the period and is not considered for this plot (however, see Figure 6.8d). All the forcings shown have distinct spatial and seasonal features (Figure 6.7) such that the global, annual means appearing on this plot do not yield a complete picture of the radiative perturbation. They are only intended to give, in a relative sense, a first-order perspective on a global, annual mean scale, and cannot be readily employed to obtain the climate response to the total natural and/or anthropogenic forcings. As in the SAR, it is emphasised that the positive and negative global mean forcings cannot be added up and viewed a priori as providing offsets in terms of the complete global climate impact.

An alternative approach which has been used to reconstruct TSI (Reid, 1997; Soon et al., 1996) is to assume that time variations in global surface temperature are due to a combination of the effects of solar variability and enhanced greenhouse gas concentrations and to find that combination of these two forcings which best combine to simulate surface temperature measurements. However, these authors did not take natural climatic variability into account and a TSI series derived by such methods could not be used as an independent measure of radiative forcing of climate.

The estimate for solar radiative forcing since 1750 of 0.3 Wm-2, shown in Figure 6.6, is based on the values in Figure 6.5 (taking the 11-year cycle minimum values in 1744 and 1996). Clearly the starting date of 1750 (chosen for the date of the pre-industrial atmosphere in Figure 6.6) is crucial here: a choice of 1700 would give values about twice as large; a choice of 1776 would give smaller values (particularly using the Hoyt and Schatten series). The range of 0.1 to 0.5 Wm-2 given in Figure 6.6 is based on the variability of the series, the differences between the reconstructions and uncertainties concerning stratospheric adjustment (see Section However, because of the large uncertainty in the absolute value of TSI and the reconstruction methods our assessment of the “level of scientific understanding” is “very low”.

6.11.2 Mechanisms for Amplification of Solar Forcing Solar ultraviolet variation

The Sun emits radiation over the entire electromagnetic spectrum but with most energy in the near ultraviolet, visible, and near infrared regions: 80% of TSI lies between 400 and 1,600 nm. The variations in TSI, discussed above, of a few tenths of one per cent thus represent the integrated change across ultraviolet, visible, and near infrared wavelengths. Most of this radiation passes through the atmosphere unhindered to the tropopause but shorter wavelengths are absorbed in the middle atmosphere where they result in O3 formation and local heating. In the ultraviolet the amplitude of variability is much higher. Since 1978 (see Cebula et al. (1998) for a review) satellite data have shown variations over the 27-day solar rotation period of e.g., 6% at 200 nm and 2.5% at 250 nm. Problems with sensor drift and calibration preclude direct detection of variability on the 11-year time-scale at wavelengths longer than about 250 nm. Instead ultraviolet irradiance cycles are deduced from observations by scaling the 27-day variations to selected solar activity indices and assuming that the same scaling applies over longer time-scales (Lean et al., 1997). With the launch of the SOLSTICE (Rottman et al., 1993) and SUSIM (Brueckner et al., 1993) instruments on UARS in 1991, measurements have now been made from near solar maximum to solar minimum with long-term precisions that approach 1% at some wavelengths (Rottman et al., 1993; Floyd et al., 1998). Careful cross-calibration of NOAA-11 SBUV/2 with Shuttle SBUV observations (Cebula et al., 1998) have also produced spectral variations 1989 to 1994, but also with uncertainties of a few per cent, which exceeds the actual irradiance variability at the longer ultraviolet wavelengths. Comparison of the SOLSTICE, SUSIM, and SBUV/2 data show reasonable agreement during their 2 to 3 year overlap period (DeLand and Cebula, 1998) and suggest a decline of about 7% at 200 to 208 nm and of about 3.5% at 250 nm from solar maximum in 1989 to near solar minimum in 1994. These estimates agree well with the modelled scalings deduced from the 27-day variations (Lean et al., 1997).

Variations in stratospheric composition and thermal structure resulting from ultraviolet irradiance changes may have an impact on tropospheric climate. The first mechanism whereby this might happen is through changes in radiative forcing (Haigh, 1994). Thus, in addition to a direct increase in downward short-wave irradiance at the tropopause, higher solar activity can cause an increase in downward infrared flux by heating the stratosphere and also radiative forcing due to O3 changes. However, the sign of the O3 effect is not well established. Haigh (1994) found that O3 increases reduced the solar radiative forcing by about 0.1 Wm-2 at solar maximum, Wuebbles et al. (1998) computed a value of -0.13 Wm-2 due to O3 increases since the Maunder Minimum and Myhre et al. (1998a) about -0.02 Wm-2 from minimum to maximum of the solar cycle. Hansen et al. (1997a) showed an additional forcing of about +0.05 Wm-2 from minimum to maximum of a solar cycle due to O3 increases and lower strato-spheric warming. Haigh (1999) and Larkin et al. (2000) suggest that the O3 effect has little effect on radiative forcing at the tropopause but significant effect on where the additional radiation is absorbed (more within the troposphere rather than at the surface). These disparities may represent the different approaches used. Haigh (1994), Wuebbles et al. (1998), and Myhre et al. (1998a) calculated the O3 response using two-dimensional chemical-transport models in which temperature changes are estimated using the fixed dynamic heating approximation. The Hansen et al. (1997a) value was deduced from studies with a simplified GCM of sensitivity to slab O3 changes (of unspecified cause) and the assumption that the solar-induced O3 change is all within the 10 to 150 hPa region. Haigh (1999) and Larkin et al. (2000) specified solar irradiance and O3 changes and calculated the stratospheric temperature response in GCMs. The negative radiative forcing values probably correspond to O3 change profiles which peak at higher altitudes, and thus have less impact on lower stratospheric temperature and long-wave radiative forcing, although the different methods for calculating temperature change may also be important.

The response of O3 to solar variability is not well established. Two-dimensional models (e.g., Haigh, 1994; Fleming et al., 1995; Wuebbles et al., 1998) predict the largest fractional changes in the middle-upper stratosphere with monotonically decreasing effects towards the tropopause. Multiple regression analysis of satellite data as carried out with SBUV data by McCormack and Hood (1996) and SAGE data by Wang et al. (1996) suggest the largest changes in the upper and lower stratosphere and zero, or even slightly negative, changes in the middle stratosphere. However, as the data are only available over about one and a half solar cycles, have large uncertainties, especially in the lower stratosphere, and may not properly have accounted for the effects of volcanic aerosol (Solomon et al., 1996), the true nature of solar-induced changes in stratospheric O3 remains uncertain.

Chandra et al. (1999) have estimated tropical tropospheric O3 column amounts by taking the difference between TOMS total columns in clear-sky areas and those with deep convective cloud. They deduced O3 changes of the order 10% of the tropo-spheric column over the eleven year cycle occurring out of phase with the solar forcing. This they ascribed to a self-feedback effect on O3 production with enhanced levels of O3 aloft resulting in less ultraviolet reaching the troposphere. However, it is also possible that the O3 changes reflect a response to solar effects in tropospheric dynamics. If the changes are real then solar radiative forcing, as represented in Figure 6.6, should be reduced by approximately 30% due to solar-induced decreases in tropo-spheric O3.

Changes in stratospheric thermal structure may also affect the troposphere through dynamical interactions rather than through radiative forcing. Kodera (1995) suggested that changes in strato-spheric zonal wind structure, brought about by enhanced solar heating, could interact with vertically propagating planetary waves in the winter hemisphere to produce a particular mode of response. This mode, also seen in response to heating in the lower stratosphere caused by injection of volcanic aerosol, shows dipole anomalies in zonal wind structure which propagate down, over the winter period, into the troposphere. Rind and Balachandran (1995) investigated the impact of large increases in solar ultra-violet on the troposphere with a GCM and confirmed that altered refraction characteristics affect wave propagation in winter high latitudes. Solar cycle changes to wave propagation in the middle atmosphere were also investigated by Arnold and Robinson (1998) who used a three-dimensional model of the atmosphere between 10 and 140 km to study the effects of thermospheric heating. They found that non-linear interactions in the winter hemisphere resulted in changes to the stratospheric circulation. It is not clear that the signals discussed above are statistically robust in any of these studies.

Haigh (1996, 1999), Shindell et al. (1999), and Larkin et al. (2000) have introduced realistic changes in ultraviolet and O3 into GCMs and found that the inclusion of the O3 has a significant effect on simulated climate. Haigh (1999) using a GCM with a lid at 10 hPa and few stratospheric levels, showed a pattern of change in zonal mean temperature which was consistent over a range of assumptions concerning the magnitude of the ultraviolet and O3 changes. This pattern consisted of warming in the stratosphere (except in winter high latitudes) and a vertical banding structure in the troposphere due to shifts in the positions of the sub-tropical jets. The predicted changes in 30 hPa geopotential heights were of similar form to those observed by Labitzke and van Loon (1997) but of smaller magnitude (by about a factor 3). Larkin et al. (2000) found very similar patterns of change to those of Haigh (1999) using the same solar irradiance/O3 changes but an entirely different GCM with a lid at 0.1 hPa. Shindell et al. (1999) used a model with a detailed representation of the middle atmosphere and a parametrized photochemical scheme which allows O3 to respond to changes in ultraviolet flux but not to changes in wind fields. They showed larger amplitude changes in 30 hPa heights in the winter hemisphere, more like those of Labitzke and van Loon (1997) but with little summer hemisphere response. These experiments suggest that further work is needed to establish the response of O3 to solar variability and to how this affects climate.

Other reports in this collection