Solar irradiance during the last 1200 years based on cosmogenic nuclides

Based on a quantitative study of the common fluctuations of 14C and 10Be production rates, we have derived a time series of the solar magnetic variability over the last 1200 years. This record is converted into irradiance variations by linear scaling based on previous studies of sun-like stars and of the sun’s behavior over the last few centuries. The new solar irradiance record exhibits low values during the well-known solar minima centered at about 1900, 1810 (Dalton) and 1690 ad(Maunder). Further back in time, a rather long period between 1450 and 1750 ad is characterized by low irradiance values. A shorter period is centered at about 1200 ad, with irradiance slightly higher or similar to present day values. It is tempting to correlate these periods with the so-called “little ice age” and “medieval warm period” respectively An accurate quantification of the climatic impact of this new irradiance record requires the use of coupled atmosphere−ocean general circulation models (GCMs). Nevertheless, our record is already compatible with a global cooling of about 0.5-1°C during the “little ice age”, and with a general cooling trend during the past millenium followed by global warming during the 20th century (Mann et al., 1999).


Solar irradiance variations
above the atmosphere with spacecraft and data are thus limited to the last 2 decades. This is A positive relationship is now well established rather unfortunate, as a direct solar influence has between the 11-year cycle of solar magnetic activbeen invoked to explain slow variations of several ity and the output of the sun (Willson and Hudson, climatic parameters such as sea level atmospheric 1988; Willson, 1997). Over the last two 11-year pressure (Kelly, 1977), sea surface temperatures cycles (#21 and #22), the amplitude of these flucta- (Reid, 1987), equatorial wind patterns (Labitzke tions has remained of the order of 0.1% for the and Van Loon, 1988), land air temperatures (Friistotal radiative output. Willson (1997) even pro-Christensen andLassen, 1991) and strength of posed that the irradiance increased by 0.04% cyclogenesis (Tinsley, 1994). Several authors even between the last 2 minima of solar activity, proposed that the solar variability is responsible although this fact has been questioned recently for a significant part of the high-frequency com- (Frö hlich and Lean, 1998). ponent of the natural variability of climate (Eddy, Such small fluctuations can only be measured 1976; Lean and Rind, 1994;Lean et al., 1995;Crowley and Kim, 1996;Mann et al., 1998Mann et al., , 1999 and should thus be taken into account when century (Hansen and Lacis, 1990;Kelly and (1995) proposed that the irradiance record could be divided into 2 superimposed components: an Wigley; 1992).
11-year cycle based on the parameterization of sun-Astrophysicists have documented simple relaspot darkening and facular brightening (Lean et al., tionships between magnetic activity and irradiance 1992), and a slowly-varying background derived of the sun (Foukal and Lean, 1990; separately from studies of sun-like stars (Baliunas 1992), and a linear relationship also characterizes and Jastrow, 1990). More recently, Solanki and sun-like stars (Zhang et al., 1994). These studies Fligge (1998) reconstructed a solar irradiance have led several workers to use indices of the record back to 1874  by using different relationmagnetic variability of the sun as proxy surrogates ships for estimating the variations due to solar for irradiance fluctuations. For the last 4 centuries, active regions (quadratic calibration between the it has been possible to make use of the record of SSN and the spacecraft TSI record over a decade) sunspots observed with telescopes. Most attention and to the long-term component (linear calibrations has been focused on solar activity minima centred between brightness and chromospheric emission or at about 1900, 1810 and 1690 . The latter is the length of the activity cycle). well-known Maunder minimum (1645-1715 ) As illustrated in Fig. 1, these reconstructions during which almost no sunspots were observed exhibit similar fluctuations, but they differ signiand which coincides with the coldest temperatures ficantly in amplitude. The main differences are of the so-called ''little ice age' (Eddy, 1976 2. Indirect proxies of the solar irradiance 0.3% (Hoyt and Schatten 1993), 0.4% (Zhang et al., 1994;Solanki and Fligge, 1998), 0.5% A major problem has been the scaling between (Nesme-Ribes et al., 1993), 0.54% (Cliver et al., the sunspot record and the total solar irradiance 1998), 0.65% (Reid, 1997) and 1% (Reid, 1991). (TSI). The yearly average sunspot number (SSN) The estimate by Lean et al. (1995) is often used exhibits minimum values which are all similar as a conservative view of TSI changes during the (Hoyt et al., 1994) and thus a simple linear rela-Maunder Minimum. Nevertheless, it should be tionship between SSN and TSI does not allow for noted that in the absence of any long-term trend, long-term variations. Such crude scaling would be the TSI reduction during the Maunder Minimum inconsistent with long-term fluctuations which would be of the same 0.1% magnitude as observed have been inferred to exist in the solar magnetic accurately during the 11-year cycle. record (Feynman and Crooker, 1978; There are only few reliable data on the solar 1992; Solanki and Fligge, 1998). In particular, the variability before the invention of the telescope, fact that recent minima of the solar cycle and the which allowed precise sunspot counting (1610 ). Maunder minimum are both characterized by a Silverman (1992) compiled the observations of near absence of sunspots does not necessarily aurora at low and mid-latitudes with the naked mean that their irradiance was the same as shown eye. This time series reaches the year 1500  and by the bimodal distribution of the magnetic activclearly shows the most prominent solar minima. ity of sun-like stars (Baliunas and Jastrow, 1990; However, this record is biased by the number and Lockwood et al., 1992). quality of observations which increases exponenti-Several authors have proposed TSI reconstrucally with time. This makes it difficult to convert tions for recent centuries by using different types of this indirect record of the solar magnetic variabilinformation on the solar variability: the envelope of ity into terms of irradiance changes. the SSN 11-year cycle (Reid, 1991), the length and decay rate of the solar cycle (Hoyt and Schatten, 1993), the structure and decay rate of individual 3. Solar modulation of cosmonuclide sunspots (Hoyt and Schatten 1993), the mean level production of SSN (Hoyt and Schatten, 1993;Zhang et al., 1994;Reid, 1997), the solar rotation and the solar As a proxy for the solar magnetic variability, it diameter (Nesme-Ribes et al., 1993), and the geo-is also possible to use the high-frequency component of the variations in production of cosmogenic magnetic aa index (Cliver et al., 1998). Lean et al.
we compared quantitatively the variations of the atmospheric 14C/12C ratio, as measured in treerings, with an independent record of the 10Be concentration analysed in well-dated ice from South Pole (Fig. 2a). The 14C and 10Be time series cannot be compared directly, since the fates of these 2 cosmonuclides are different after their production in the atmosphere (see Bard, 1997Bard, , 1998, for recent discussions on this issue). In order to quantify this differential effect, we used a numerical box model which, in particular, takes into account the 14C dispersal in major reservoirs of the carbon cycle. The relative fluctuations of 10Be concentrations measured in the South Pole ice core were used as a model input to compute a synthetic atmospheric 14C/12C record. Fig. 2b illustrates that the modelled 14C/12C variations agree in timing and amplitude with the 14C/12C data measured in tree-rings at a decadal resolution. In particular, it is straightforward to identify, in both records, periods of maximum 14C/12C corresponding to known solar activity minima centred at about 1900, 1830 (Dalton) and 1690  (Maunder). This close agreement confirms the quality of the record of cosmonuclide production The concomitant production variations of 14C details, the y-axis scales of graphs 1a, 1b and 1c are and 10Be can be used as a proxy for the TSI by different (5, 7 and 10 W/m2, respectively). assuming a linear relationship between magnetic activity and irradiance following previous workers (Lean et al., 1992;Zhang et al., 1994; Solanki and nuclides such as 14C and 10Be (Lal and Peters, 1967). Magnetic fields of the solar wind deflect Fligge, 1998). Fig. 3a,b present TSI curves computed by first smoothing the cosmonuclide pro-the primary flux of charged cosmic particles, which leads to a reduction of cosmogenic nuclide produc-duction record and then applying a linear scaling using the TSI values published previously for the tion in the earth's atmosphere. In particular, it has been shown that the 11-year cycle modulates Maunder Minimum (see discussion above). These data are available in numerical form as supplemen-10Be production recorded in well-dated polar ice from Greenland and Antarctica (Beer et al., 1990; tary information and at http://www.cerege.fr/. Smoothing the record is necessary to remove Steig et al., 1996Steig et al., , 1998. Very high cosmonuclide production (30-50% above the modern value) the meteorological ''noise'' which is partly responsible for the decadal variability of the raw 10Be was also confirmed for the Maunder Minimum based on high 14C content in tree rings (Stuiver, record (Fig. 2a). This short-term variability disappears when considering the 14C time series (i.e., 1980; Stuiver and Quay, 1980) and 10Be in polar ice (Raisbeck et al., 1981(Raisbeck et al., , 1990Beer et al., 1988). Fig. 2b), because the carbon cycle acts as a low pass filter (both in nature and in the modelling In a previous contribution (Bard et al., 1997),  (2a) shows the record of the cosmonuclide production (‰ variations around the present value) during the last millenium as reconstructed from the 10Be concentration measurements in South Pole ice (Raisbeck et al., 1990) and the quantitative comparison with the atmospheric 14C/12C record (Bard et al., 1997). Dots show the raw production record and the solid line a smoothed curve obtained with a Gaussian filter. The lower panel (2b) shows with open dots the detrended 14C/12C record measured in tree rings (Stuiver, 1980). The solid curve shows the tropospheric 14C/12C variations simulated by using a 12 box model and the relative production changes shown in the upper panel. The 14C/12C values are represented as ‰ variations (D14C). exercise). Further analysis of 10Be in ice cores from In particular, a higher resolution would be required in order to test the recent claim by Beer different sites in Antarctica will be necessary to distinguish the short-term global production vari-et al. (1998) that the 11-year solar cycle did not completely cease during the Maunder Minimum. ability from the local meteorological fluctuations.
1400 ), although the timing and signatures of these two climatic events are still debatable (Bradley and Jones, 1993;Hughes and Diaz, 1994;Dahl-Jensen et al., 1998). Indeed, some researchers have concluded that the ''little ice age'' and/or ''medieval warm period'' are regional, rather than global events (see description and discussion of spatial and temporal details in Jones, 1993 andDiaz, 1994). At first sight, TSI variations would not be a likely mechanism, for it would tend to force global effects. However, general circulation modelling of the atmospheric response to a TSI reduction by 0.25% clearly indicates that not every locale experiences cooling, as advective changes in the atmosphere can dominate the radiative cooling (Lean and Rind, 1994). This regional variability of the response could even be greater by taking into account the advective changes of the ocean.
As shown by many studies (Lean, 1989;Lean et al., 1995), the irradiance variations are not spectrally averaged, and the relative amplitude of the UV variability is an order of magnitude larger than the total irradiance changes. Haigh Fig. 3. TSI reconstructions based on the smoothed (1994,1996) has even shown that the climatic cosmonuclides production record. The upper panel (3a) response is enhanced by a positive feedback due shows the last 4 centuries and can be compared directly to stratospheric ozone: the stratosphere is heated,  Lean et al., 1995Lean et al., ), et al., 1984. In addition, the time series of the dashed lines (0.4%, Zhang et al., 1994; Solanki and cosmonuclide production may provide additional Fligge, 1998), dotted lines (0. 55%, Cliver et al., 1998) climatic information if a direct link between cosmic and thin lines (0.65%, Reid, 1997). The shaded area is rays, droplet nucleation and cloud cover is conbracketed by the low and high TSI estimates based on firmed (Tinsley, 1994; Svensmark and Friisthe factors derived from Reid (1997) and Lean et al. Christensen, 1997). (1995), respectively. This latter curve is highlighted by a thick line as it constitutes a rather plausible alternative It is clear that the exact implications of our for extending the solar irradiance over the last millenium irradiance reconstruction in terms of global and (see Section 2 for details). regional climate must be evaluated by means of GCMs, taking into account these atmospheric feedbacks and also the memory effect of the sea-Over the last 4 centuries, our irradiance reconstruction is in reasonable agreement with inde-surface heating. Such results were obtained by Cubasch et al. (1997) for the last 3 centuries, based pendent reconstructions based on observations of the sun (Fig. 1). The extended TSI record suggests on the TSI record presented in Fig. 1b (Hoyt and Schatten, 1993). For the millenium-long record, it that the solar output was significantly reduced between 1450 and 1850 , but slightly higher or will be interesting to evaluate whether the distinct periods of reduced TSI between 1250 and 1750  similar to the present value during a period centred around 1200 . It could thus be argued that could explain part of the general cooling trend which followed the 14th century (Mann et al., irradiance variations may have contributed to the so-called ''little ice age'' (about 1500 to 1750 ) 1999). Nevertheless, it is already possible to calculate a 1st-order global climatic response which and ''medieval warm period'' (about 900 to strongly depends on the overall amplitude of the large range of values deduced by previous authors for the Maunder Minimum. As described in irradiance variations. For the period between 1400 and 1600 , a midrange total irradiance decrease Section 2, these estimates are not always independent, because some of them are based on stellar of 0.35% corresponds to a decrease of about 4.8 W/m2 in total irradiance (1367×0.35/100). observations, some on other physical arguments and some even from the expected climatic record This leads to 0.84 W/m2 when averaged over the whole earth's surface, assuming an albedo of 0.3 of the earth.
Further improvements in the TSI quantification (4.8×0.7/4). At equilibrium, this would lead to a cooling between 0.2 and 0.8°C in typical GCMs could be expected as progress will be made on physical mechanisms defining the nature of long-characterized by sensitivities between 0.25 and 0.95°C W−1 m2 (Myrhe et al., 1998).
term components of the solar variability. It will also depend on improvements in estimating the correlation between TSI and solar magnetic activ-

Conclusions and perspectives
ity. A direct way of doing this would be to compare 10Be production during recent solar minima with Based on 2 independent but concordant production during the Maunder Minimum in the cosmogenic isotope records, we propose a reliable same core from Antarctica as described by solar index for the last 1200 years. This new record Raisbeck and Yiou (1980). Such a calibration agrees with those previously published for the last cannot be done with 14C both because of the 3 centuries based on direct observations of the strong damping of the 11-year production cycle sun's surface. The main outcome of our study is and the interference of the Suess effect (during the that the solar magnetic activity, and thus the TSI, 20th century the measured atmospheric D14C can was usually lower than present during most of the no longer be used to accurately study the natural last millenium, except during a brief period centred variations, since the 14C atmospheric concentraaround 1200 . This solar variability record is tion has been significantly decreased by the qualitatively similar to the climate reconstruction anthropic combustion of 14C-dead fossil fuels). proposed recently by Mann et al. (1999), suggesting that the 20th century warming followed a millenial-scale cooling trend.
Conversion of the new solar variability record 6. Acknowledgements in terms of TSI is attempted by means of linear scaling factors. As illustrated by the shaded zone We thank J.-J. Motte for drawings. This research is supported by CNRS, IN2P3, CEA and EC in Fig. 3, this simple approach leads to a TSI spread which is a direct consequence of the rather grant MilEClim ENV4-CT97-0659.