LETTER doi:10.1038/nature09426 An influence of solar spectral variations on radiative forcing of climate Joanna D. Haigh1, Ann R. Winning1, Ralf Toumi1 & Jerald W. Harder2 The thermal structure and composition of the atmosphere is deter- mined fundamentally by the incoming solar irradiance. Radiation at ultraviolet wavelengths dissociates atmospheric molecules, ini- tiating chains of chemical reactions���specifically those producing stratospheric ozone���and providing the major source of heating for the middle atmosphere, while radiation at visible and near- infrared wavelengths mainly reaches and warms the lower atmo- sphere and the Earth���s surface1. Thus the spectral composition of solar radiation is crucial in determining atmospheric structure, as well as surface temperature, and it follows that the response of the atmosphere to variations in solar irradiance depends on the spec- trum2. Daily measurements of the solar spectrum between 0.2 mm and 2.4 mm, made by the Spectral Irradiance Monitor (SIM) instru- ment on the Solar Radiation and Climate Experiment (SORCE) satellite3 since April 2004, have revealed4 that over this declining phase of the solar cycle there was a four to six times larger decline in ultraviolet than would have been predicted on the basis of our previous understanding. This reduction was partially compensated in the total solar output by an increase in radiation at visible wave- lengths. Here we show that these spectral changes appear to have led to a significant decline from 2004 to 2007 in stratospheric ozone below an altitude of 45 km, with an increase above this altitude. Our results, simulated with a radiative-photochemical model, are consistent with contemporaneous measurements of ozone from the Aura-MLS satellite, although the short time period makes precise attribution to solar effects difficult. We also show, using the SIM data, that solar radiative forcing of surface climate is out of phase with solar activity. Currently there is insufficient observational evidence to validate the spectral variations observed by SIM, or to fully characterize other solar cycles, but our findings raise the possibility that the effects of solar variability on temper- ature throughout the atmosphere may be contrary to current expectations. The peak of the most recent ���11-year��� solar cycle (identified as num- ber 23) occurred 2000���2002, and from then until about December 2009 the Sun���s activity declined. Figure 1 shows the difference between 2004 and 2007 in solar spectral irradiance measured by SIM. This is quite unlike that predicted by multi-component empirical models, based on activity indicators such as sunspot number and area, as exemplified by that of Lean5 (also shown in Fig. 1). The SIM data indicate a decline in ultraviolet from 2004 to 2007 that is a factor of 4 to 6 larger than in the Lean data and an increase in visible radiation, compared with a small decline in the Lean data. Other empirical mod- els6,7 show larger-amplitude variationsinthenear-ultravioletthan does the Lean model but none reflect the behaviour apparent in the SIM data. Also shown in Fig. 1, for wavelengths 116���290 nm, are independent measurements made by the Solar Stellar Irradiance Comparison Experiment (SOLSTICE) instrument on SORCE. The data from SIM and SOLSTICE both indicate substantially more ultraviolet variability thandoestheLeanmodel.SIMcalibration,and instrumentcomparisons, are discussed in detail in ref. 8. To investigate how these very different spectral changes might affect the stratosphere, experiments have been carried out using a two- dimensional (latitude-height) radiative-chemical-transport model of the atmosphere9. This model includes detailed representations of photo- chemistry and radiative transfer and has been used in many studies involving radiation-chemistry interactions10,11. (See Supplementary Information for further details.) This type of model produces realistic simulations of the upper stratosphere (above about 25 km) but is less reliable at lower altitudes where photochemical time constants are longer and a more accurate representation of transport processes is required. The results below come from four model runs using solar spectra derived from the SIM measurements (with SOLSTICE data for wavelengths less than 200 nm) and those produced by the Lean model, each for both 2004 and 2007. In Fig. 2 we present latitude���height maps of the difference between 2004 and 2007 in December ozone concentrations. The Lean spectral data produce a broad structure of ozone concentrations greater in 2004 than in 2007, with maximum values of around 0.8% near 40km, whereas the SIM data produce a peak enhancement of over 2% in low latitudes around 35 km, along with significant reductions above 45km. The predicted temperature differences (Supplementary Fig. 1) are also very different, with the Lean data set showing temperatures 0.3���0.4 K greater in2004 than in 2007at thetopofthe model domain, whereas the SIM data set produces a peak warming of 1.8 K at the summer polar stratopause. These temperature differences are qualitatively similar to, but about 50% larger than, those estimated by ref. 12 with an idealized forcing in a full climate model, possibly owing to the broader spectral 1 Blackett Laboratory, Imperial College London, London SW7 2AZ, UK. 2 Laboratory for Atmospheric and Space Physics, University of Colorado, 1234 Innovation Drive, Boulder, Colorado 80303-7814, USA. 600 500 SIM Lean model Difference in spectral irradiance at 242 (mW m ���2 nm ���1 ) Difference in spectral irradiance at 242 nm (mW m ���2 nm ���1 ) SOLSTICE 400 Wavelength (nm) 300 200 242 4 2 0 ���2 6 10nm 5 0 ���5 15 700 Figure 1 | Difference in solar spectrum between April 2004 and November 2007. The difference (2004���2007) in solar spectral irradiance (W m22 nm21) derived from SIM data4 (in blue), SOLSTICE data8 (in red) and from the Lean model5 (in black). Different scales are used for values at wavelengths less and more than 242 nm (see left and right axes respectively). 6 9 6 | N A T U R E | V O L 4 6 7 | 7 O C T O B E R 2 0 1 0 Macmillan Publishers Limited. All rights reserved ��2010

resolution imposed and the lack of ozone���temperature feedback in that model version. The very different scenarios produced by the two spectral data sets suggest they might be distinguishable in observational records. A mul- tiple regression analysis has been carried out ofdeseasonalized monthly meanozonedatafromthe MicrowaveLimbSounder(MLS)instrument on the Earth Observing System (EOS) Aura satellite. Four regression indices were used: a constant, two orthogonal indices representing the quasi-biennial oscillation (which dominates ozone variability in the tropical stratosphere)13 and a solar index constructed from SIM data integrated over 200���400 nm. Motivated by the model results (Fig. 2), we chose two spatial regions, both spanning the tropics, one at altitude 10���6.8 hPa, where the model predicts the largest difference 2004���2007, and one at 0.68���0.32 hPa, where the model shows largest negative values. Figure 3 shows the raw data and the fits reconstructed from the four regression components it also shows (in red) the derived solar component, which is statistically significant at .95% at the upper levels and .99% at the lower levels (see Supplementary Information). Over the period from the late 1970s to the late 1990s tropical ozone at altitudes 35���50 km decreased by about 9% (ref. 13) in response to increasing concentrations of active chlorine species. Since about 2000, however, the trend in chlorine has reversed and ozone has stopped declining. Stratospheric cooling by greenhouse gases has probably also contributed to the ozone trend reversal by slowing the chemical reac- tions that destroy it14. Over the short period of the present study it is not possible statistically to differentiate these factors from each other, or from any solar influence. Nevertheless, it seems likely that the Sun is important in the apparent decrease in ozone below 45 km from 2004 to 2007. The change in sign near 45 km is also more consistent with the modelled response to the SIM spectral variations than to the Lean spectra. Previous analyses13,15 of the solar signal in ozone, averaged over approximately 2.5 solar cycles (1979 to 2005 or 2003), have not shown this structure. This suggests that the declining phase of solar cycle 23 is behaving differently to previous solar cycles or possibly that the solar cycle exhibits different behaviours during its ascending and descending phases. To understand the different spatial structures, and magnitudes, of the modelled ozone responses we consider photochemical processes. The sharp decrease in ozone above 45 km with the SIM spectra (Fig. 2b) is consistent with it being in photochemical steady state with the dominant sinks, that is, increased levels of HOx and O. These losses are compensated by the greater production of Ox through photodissocia- tion of O2 in theHuggins bandand thisdominates the loss lower down. Furthermore, the ozone decreases produce a self-healing effect whereby more ultraviolet radiation is transmitted to lower levels, resulting in greater O2 photolysis and thus more O3. (See Supplemen- tary Information.) To assess the sensitivity of our results to uncertainty in the measured irradiance values at 200���240 nm (see Fig. 1) we carried out another set of experiments (not shown) in which the switchover from SOLSTICE to SIM was imposed at 240 nm (rather than 200 nm). There are differ- ences in detail in the resulting temperature and ozone fields but the general picture is the same: reduced ozone in the upper stratosphere and mesosphere and a positive peak in the middle stratosphere. Therefore there is uncertainty in the magnitude of the response but this does not affect our conclusions with respect to the impact on the middle atmosphere. It also has little bearing on the radiative forcing estimates now presented. The response of tropospheric and surface climate to variations in solar activity is an important consideration in the attribution of surface temperature trends to human or natural factors. Radiative forcing of climate is defined by the Intergovernmental Panel on Climate Change as the change in net flux at the tropopause, taking into account the effects of any stratospheric adjustment16. It is known that solar radi- ative forcing is modulated by the ozone response to changes in solar 0.6 50 40 30 60 1 3 10 30 20 0 Latitude (��N) Latitude (��N) ���20 ���40 40 0.3 0.6 0.5 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.2 0.3 0.3 0.7 0.7 0.7 0.8 ���1.4 ���1.2 ���1.2 ���1.0 ���0.8 ���0.6 2.0 1.4 20 0 ���20 ���40 40 50 40 30 60 1 3 10 30 0.3 1.6 1.8 1.6 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.60.4 0.2 0.0 0.0 1.8 1.6 1.4 1.2 1.0 0.60.8 0.4 0.2 ���0.2 ���0.4 Approximate altitude (km) Pressure (hPa) Approximate altitude (km) Pressure (hPa) a Lean model b SIM/SOLSTICE data Figure 2 | Modelled difference in ozone between December 2004 and December 2007. Estimates of the percentage difference (2004���2007) in zonal mean ozone concentration (labels on contour lines in per cent) produced by the modelusingsolarspectrafromtheLeanmodel(a)andSIM/SOLSTICEdata(b). a 0.68���0.32 hPa 2006 2005 O 3 anomalies (%) O 3 anomalies (%) 2 0 ���2 4 ���4 2 0 ���2 4 ���4 2007 2006 2005 2007 b 10���6.8 hPa Figure 3 | Time series of AURA-MLS v2.2 ozone concentrations. The data (solid black lines) are percentage anomalies of tropical (22.5 uS���22.5 uN) deseasonalized monthly means from August 2004 to November 2007. The values reconstructed from the 4-component regression model are shown as dashed lines. The solar component of the regression is shown in red. Other components are shown, along with the solar component, in Supplementary Fig. 2. Data were averaged between 0.68 hPa and 0.32 hPa (a) and 10 hPa and 6.8 hPa (b). LETTER RESEARCH 7 O C T O B E R 2 0 1 0 | V O L 4 6 7 | N A T U R E | 6 9 7 Macmillan Publishers Limited. All rights reserved ��2010