Changes in surface solar radiation in Northeastern Spain over the past six centuries recorded by tree-ring δ13C

Although solar radiation at the surface plays a determinant role in carbon discrimination in tree rings, stable carbon isotope chronologies (δ13C) have often been interpreted as a temperature proxy due to the co-variability of temperature and surface solar radiation. Furthermore, even when surface solar radiation is assumed to be the main driver of 13C discrimination in tree rings, δ13C records have been calibrated against sunshine duration or cloud cover series for which longer observational records exists. In this study, we use different instrumental and satellite data over northeast Spain (southern Europe) to identify the main driver of tree-ring 13C discrimination in this region. Special attention is paid to periods in which the co-variability of those climate variables may have been weaker, such as years after large volcanic eruptions. The analysis identified surface solar radiation as the main driver of tree-ring δ13C changes in this region, although the influence of other climatic factors may not be negligible. Accordingly, we suggest that a reconstruction of SSR over the last 600 years is possible. The relation between multidecadal variations of an independent temperature reconstruction and surface solar radiation in this region shows no clear sign, and warmer (colder) periods may be accompanied by both higher and lower surface solar radiation. However, our reconstructed records of surface solar radiation reveals a sunnier Little Ice Age in agreement with other δ13C tree-ring series used to reconstruct sunshine duration in central and northern Europe.

cloud cover series for which longer observational records exists. In this study, we use different instrumental and satellite data over northeast Spain (southern Europe) to identify the main driver of tree-ring 13 C discrimination in this region. Special attention is paid to periods in which the co-variability of those climate variables may have been weaker, such as years after large volcanic eruptions. The analysis identified surface solar radiation as the main driver of tree-ring δ 13 C changes in this region, although the influence of other climatic factors may not be negligible. Accordingly, we suggest that a reconstruction of SSR over the last 600 years is possible. The relation between multidecadal variations of an independent temperature reconstruction and surface solar radiation in this region shows no clear sign, and warmer (colder) periods may be accompanied by both higher and lower surface solar radiation.
However, our reconstructed records of surface solar radiation reveals a sunnier Little Ice Age in agreement with other δ 13 C tree-ring series used to reconstruct sunshine duration in central and northern Europe.

Introduction
Surface solar radiation (SSR or global radiation) may change as a response to a perturbation of the climate system, for instance, due to anthropogenic greenhouse gases, as the amount of cloud cover or the radiative properties of clouds may respond to changes in atmospheric temperatures (Stevens and Bony 2013). Climate models still show significant discrepancies in the simulation of the response of clouds to climate change (Stephens et al. 2005;Boucher et al. 2013). The uncertainties in the simulation of cloudiness and SSR are still the single most important reasons for the large spread in the climate sensitivity among climate models (Dessler 2010;Flato et al. 2013).
It is known that SSR has not been constant through time and since the mid-20 th century it has experimented a widespread declining phase (dimming period) followed by an increasing phase (brightening period) (Stanhill and Cohen 2001;Wild et al. 2005;Wild 2012). Among the possible causes of these changes in SSR are anthropogenic and natural aerosols, aerosolcloud interactions and variation in cloudiness induced by the internal variability of the climate system (Wild 2009 and references therein). These changes in SSR should have an effect on surface temperatures (Wild et al. 2007;Wang and Dickinson 2013). On the other hand, changes in surface temperatures may affect SSR through induced changes in cloudiness (Dessler 2010). Part of the difficulties in disentangling the mutual interaction between variations in SSR and temperature are due to the short length of the existing instrumental SSR series (Wang and Dickinson 2013).
The use of related variables for which longer observational series exists (Román et al. 2014;Wild et al. 2007;Makowski et al. 2009;Wang and Dickinson 2013) have allowed the extension of the SSR inferences back a few decades, but not before the 20 th century in most of the cases. The relations between the changes in temperature and the associated changes in SSR at the timescales involved in climate change could be better assessed if the period of analysis could be extended back in time covering climatic periods where temperatures may have considerably differed from the present climate, such as the Little Ice Age (LIA) or be roughly similar to 20 th century temperatures as in the Medieval Climate Anomaly (MCA).
The availability of natural proxy records, such as those traditionally used to reconstruct temperature and precipitation (i.e., tree rings, lake sediments, ice cores), allows the assessment of longer-term changes than would be unfeasible just using observational data. In this context, the autotrophic metabolism of plants that depends on temperature, moisture and the incident sunlight (Farquhar et al. 1982(Farquhar et al. , 1989 points to tree rings as one of the very few terrestrial archives that could potentially be used to assess past changes in sunlight-related variables. Although the stable carbon isotopes in tree rings (δ 13 C) are the outcome of an interplay of several factors, the capacity to encode changes in sunlight has been recently demonstrated for Scandinavia (Young et al. 2010(Young et al. , 2012Gagen et al. 2011;Loader et al. 2013) and in the Alps (Hafner et al. 2014).
Such an approach was possible because plant growth relies on the production of carbohydrates from water and CO 2 in a light-dependent process, in which the rate of photosynthetic fixation depends on light intensity. The discrimination of 13 C in organic matter (i.e, tree rings) reflects the balance between the leaf photosynthetic rate and the stomatal conductance of CO 2 (Farquhar et al. 1982(Farquhar et al. , 1989, which are strongly dependent on environmental variables such as temperature, humidity, and solar radiation. Briefly stated, in water-limited environments discrimination of 13 C is theoretically driven by moisture-induced limitations of stomatal conductance, while under non-limiting moisture conditions 13 C discrimination would mainly be controlled by photosynthetic rate driven by solar radiation (Farquhar et al. 1989;McCarrol and Loader 2004).
Although these physiological processes are theoretically understood, the scarcity and limited length of available instrumental SSR records favour the misinterpretation of tree-ring δ 13 C as temperature proxy (Gagen et al. 2007). This scarcity may also have restricted the option to calibrate δ 13 C chronologies to sunlight variables with longer observational records than SRR, such as sunshine duration hours (SD) or percentage of cloud cover (CC) (Young et al. 2012;Gagen et al. 2011;Loader et al. 2013;Hafner et al. 2014). As a consequence, δ 13 C have been used to reconstruct climate variables which a priori may not be the primary drivers of The reconstruction of indirect drivers of δ 13 C would not be incorrect as long as the relationship between the main and the indirect driver has not changed in time. Specifically, temperature reconstructions based on δ 13 C assume a linear relationship between SSR and temperature over the whole reconstruction period. However, the major modulator of SSR at interannual scales is cloudiness. Cloudiness can contribute either to cooling i.e. low-level clouds promoting higher albedo; or to warming, i.e. high clouds emit less infrared radiation out to the space (e.g., Mace et al. 2006), which can compromise a linear relationship to surface temperature. In addition, the relationship between temperature and cloudiness may depend on which of these two is the driving factor and which is passively responding. At interannual timescales, cloudiness is likely modulating the local temperature, particularly in summertime at mid-latitudes, but at multidecadal timescales, CC and cloud type may respond to large-scale multidecadal temperature changes (Dessler 2010).
Similarly, the δ 13 C-based SD/CC reconstructions, although physically more closely related to SSR than temperature, have additional shortcomings. Both, SSR and the part of the SSR used by plants for photosynthesis (the so-called photosynthetic active radiation, PAR) comprise both direct and diffuse fractions. Diffuse fraction may play a determinant role in sustaining photosynthetic activity when the direct fraction is low (Mercado et al. 2009), which occurs under decreased atmospheric transmittance such as cloudy skies or with increased concentration of atmospheric aerosols (e.g., large volcanic eruptions). SD and CC records do not take into account the diffuse fraction of the SSR. Thus, under non-clear skies SD/CC may differ from SSR (e.g., Sanchez-Romero et al. 2014).
In order to identify the main driver of δ 13 C variations in tree rings, special attention needs to be paid to periods in which these climate variables may have diverged. In this context, the perturbation caused by volcanic eruptions may lead to diverging responses between temperature sensitive and sunlight sensitive tree-ring records (Battipalgia et al. 2007). Large volcanic events may cause a large-scale cooling which is detectable locally in instrumental series and in temperature sensitive tree-ring records, such a tree-ring width or maximum density (D'Arrigo et al. 2013). However, the volcanically induced local cooling may not strongly affect the photosynthetic capacity of the tree, which translates in no or nonsignificant changes in tree-ring δ 13 C (Battipaglia et al. 2007).
In this study we use a 600-years long δ 13 C tree-ring chronologyfrom a non-moisture-limited site at the eastern Pre-Pyrenees (Spain, Southwestern Europe)where net primary production is potentially constrained by SSR (Nemani et al. 2003) and take advantage of a dense network of station and satellite-derived SSR, SD, CC, and air temperature records located in the vicinity of the sampling site. Our goal is to empirically identify the main driver of 13 C discrimination in tree rings and for that purpose we also specially focus on periods where large volcanic eruptions occurred. Once SSR is identified as the main driver of δ 13 C, we use the long tree-ring δ 13 C chronology to reconstruct SSR over the last centuries at this site.
Finally we discuss the relation to the historical changes in temperature and the agreements with other δ 13 C records encoding sunlight related signals in Europe.

Site description and chronology development
The study site is an east-facing slope sub-alpine forest of Pinusuncinata Ram. located at 2120 m.a.s.l. in the Cadí-Pedraforca Range (UPF), eastern Pre-Pyrenees (Fig. 1a). Mean annual temperature is 6.1ºC and total annual rainfall is over 1000 mm, with more than 300 mm of precipitation falling evenly in June, July and August (Fig. 1b) due to the advection of humid air masses coming from the Mediterranean Sea (Planells et al. 2006). Low temperatures mark the beginning and end of the growing season and moisture is not a limiting factor for treeradial growth (Fig. 1c). Thus, the determinant control on carbon isotope fractionation is likely to be photosynthetic rate rather than stomatal conductance (McCarrol and Loader 2004).
During summer of 2006, a total amount of 75 cores were taken from living trees using increment borers. The samples were mounted, dried and sanded until individual cells were visible under the stereomicroscope. Cores were visually cross-dated following standard dendrochronological techniques (Stokes and Smiley 1968). Tree-ring widths were measured and quality and correct dating of the resulting series checked with the COFECHA software (Holmes 1983).
For the stable carbon isotope measurements, nine trees were selected and the individual rings separated with a razor-blade under a microscope. Due to the critical size of the tree rings produced by the older trees in the most recent centuries, using the same trees to cover the full period was unfeasible. Thus, four trees were selected to cover the period 1600-1900 and the oldest five trees were chosen to cover the period 1600-backwards (Fig. 2a). The period 1550-1600 was individually measured in every sample to ensure a correct overlap.Similarly, the 20 th century was also individually analyzed. The rest of the chronology was build using a combination of pooled and individual measurements every fifth year in order to meet time and costs constraints usually associated with stable isotope measurements, while allowing annual resolution and an estimation of signal replication.
Diverse studies have shown that pooling the cores can yield similar results to those obtained analyzing individual samples (Treydte et al. 2001;Leavitt and Long, 1984;McCarrol and Loader 2004). The similarity of the results obtained by these two methodological approaches was successfully tested in Dorado Liñán et al. (2011) for the data used in this chronology.
Cellulose was extracted from entire rings (early-and latewood) using standard techniques (Boettger et al. 2007). Carbon isotope analysis was conducted on carbon dioxide resulting from combustion of the samples in an elemental-analyzer and an isotope-ratio massspectrometer (McCarroll and Loader 2004). Isotope values are given as δ 13 C -values calculated from the isotope ratios 13 C/ 12 C (= R) as δ 13 C = (Rsample/Rstandard -1)*1000‰ (referring to the international standard VPDB), and have a long-term estimated methodological error of <0. 2‰ (Boettger et al. 2007).
We applied the atmospheric correction to the δ 13 C series to correct for the decreasing trend of atmospheric CO 2 signature due to the increasing fossil fuel burning depleted in 13 C since the industrialization (see details and values in McCarroll and Loader 2004). The corrected δ 13 C individual series were transformed to z-scores before averaging them into a site chronology The resulting δ 13 C chronology from the Cadí-Pedraforca Range (UPF δ 13 C) displays a robust common signal over the period 1332-2006 CE (Fig. 2a). UPF δ 13 C chronology displays a typical positive co-variability to summer temperature and a negative correlation with summer precipitation (Figure 2b). Such a signal is common even in sites known not to suffer from moisture limitations (e.g., Gagen et al. 2007;Saurer et al 2008). According to moist characteristic of UPF and the lack of a drought signal on tree growth, the negative correlation with precipitation may reflect the relation to other factor inversely related to precipitation such as SD or SSR.

Instrumental data
Different sources of monthly mean SD, CC, SSR, and mean air temperature (T) were considered in this study (Fig. 1a). The records of SSR were extracted from Sanchez-Lorenzo respectively. It is worth mentioning that MXD usually encodes the temperature signal of the full growing season, while δ 13 C tends to encode summer climate signals. Therefore, the temperature and sunlight reconstruction that will be compared do not strictly described the same season.

Data analysis
Previous studies used SD records to calibrate sunlight sensitive tree-ring δ 13 C chronologies because the short length of the common period when using SSR records hinders a splitsample procedure. In our particular case, the longest SSR record available starts in 1968 CE validation. The linear relationship between the instrumental records and the UPF δ 13 C was evaluated by the adjusted R 2 (R 2 adj ) and predicted R 2 (R 2 pred ) derived from every crossvalidation. The autocorrelation of regression residuals required to estimate the significance of the regression coefficients was estimated by the Durbin-Watson test.
The effect of large volcanic eruption in long instrumental series as well as in temperature and sunlight-sensitive tree-ring variables was tested by Superposed Epoch Analysis (SEA) (Panofsky and Brier 1958). The evaluation of the volcanic imprint was done in two steps.
First, the assessment of the large volcanic eruptions in temperature and sunlight-related variables was tested on the long instrumental records from Barcelona BaT and BaCC. For the SEA analysis on these series, eight large volcanic eruptions from 1866 CE to 1995 CE in both Northern and Southern Hemispheres were considered: 1883, 1888, 1902, 1912, 1963, 1980, 1982 and 1991. Secondly, we run SEA analyses on UPF δ 13 C and the available

Driver of δ 13 C variations at UPF
The correlations between stable carbon isotopes and monthly T, SD, CC and SSR identify the summer months June, July and August (JJA) as the dominating climate season for tree growth (Fig. 3, left panel). Higher (lower) summer T, SD and SSR (CC) are linked to a significant (p<0.05) positive (negative) response of tree-ring δ 13 C. The comparison of the different set of JJA instrumental records and the individualδ 13 C series (Fig. 3, right panel) evidences that the different T series have a more similar interannual variability than the records within each set of SD, CC and SSR series. Furthermore, the comparison also discloses disagreements in their trends. While the T records show a common and significant upward trend (p< 0.05) during this period, the records of SD, CC, SSR and δ 13 C series do not exhibit such a marked trend.
Particularly, δ 13 C and the CC records do not show any significant trend, while two out of 10 SSR records (Mallorca and Huesca) and three out of the eight SD records (Mallorca, Madrid, Lleida) display significant trends.
Two major volcanic eruptions took place during the last three decades: El Chichón (1982) and Pinatubo (1991). However, two events are not enough to draw statistically robust conclusions about the impact of these perturbations on instrumental records and δ 13 C. The availability of longer T and CC records from the Barcelona station allows for a longer-term analysis of the effect of large volcanic eruptions on the tree-ring and instrumental variables ( Fig. 4, top). During the period from 1866 CE to 1995 CE, eight large volcanic eruptions with radiative impact on both hemispheres occurred. The SEA revealed a significant negative impact (p<0.01) of volcanic eruptions on T while no significant impact was observed on CC or on δ 13 C in this region (Fig. 4, bottom). The linear regression models between UPF δ 13 C and each of the different collections of observational climate data show higher explained variance (R 2 adj) when using SSR series as predictors than T, SD or CC (Table 1 and Fig. 5). Furthermore, the linear regression performed with SSR series as predictor display similar amounts of explained and predicted variance (R 2 pred), while the low predictive skills of regression using T and SD denotes model overfitting. Although most of the regressions do not show significant autocorrelation of the residuals (Fig. 6), the significant trend detected in the residuals of most of the regression models performed with T further supports the hypothesis that a relevant predictor may be missing in these models.
Among the collections of observational records in the vicinity of the sampling site (Fig. 7), the SSR series from La Molina (50km distance) provide the best fit to δ 13 C (R 2 adj=64.1%) and the highest predictive skills (R 2 pred=58.5%) among all models. Furthermore, the regression residuals for SSR show no significant trend (p=0.15). The T record from the nearby station La Molina also displays a good fit (R 2 adj=61.7%) but the lower predictive skills (R 2 pred=49.9%) denotes once more model overfitting. Regarding the CC and SD, the best fit corresponds to the station of Madrid (600km distance). In the case of SD, the R 2 adj and R 2 pred are slightly higher than those described for the SSR station La Molina. However, the long distance between the station and the sampling site and the fact that none of the closer stations gives similar results points to spurious correlations. When extending the common calibration-verification period back to 1976, the number of records available in the vicinity of the sampling site is dramatically reduced. However, models using SSR records still display better fits than temperature records (see Fig. 1 and Table 1 from supplementary material).
Therefore, we conclude that the SSR data are best reflecting the real forcing factor of δ 13 C variability at this site, and consequently interpret the variations of δ 13 C tree-ring chronology spanning the period 1332-2006CE as the results of the changes in sunlight (PyrSSR, hereafter).

Changes in sunlight for the last 600 years
The comparison of this PyrSSR record and the preexisting growing season temperature reconstruction at the Pyrenees PyrT (Fig.8) further illustrates the inconsistency of interpreting For example, both records markedly anticorrelate during periods such as the one spanning from around 1600 to 1800 CE. Increased summer SSR was related to both periods of cooler and warmer growing season temperatures. Specifically, the period from the 14 th to the 16 th century was characterized by generally warmer temperatures but alternating periods of higher or lower SSR. During the second half of the Spörer minimum (first half of the 16 th ) temperatures were less warm than during previous century and the SSR was also lower. From the end of the 16 th century until the end of the 18 th century temperatures were gradually decreasing while SSR was high, except for a reduction during the Maunder minimum which coincides with a period of decreased total solar irradiance (TSI). The lowest temperatures during LIA occurred during the Dalton Minimum and coincide with a minimum in summer SSR associated to a marked decrease in TSI. Despite the reduced SSR duringthe Dalton Minimum, the LIA is generally related to higher SSR. From this period until the second half of the 20 th century, the climate at the Pyrenees was characterized by low SSR and a gradual increase in temperatures. The 20 th century shows a maximum in temperatures during the first half of the century and low SSR, increasing during the last decades in line with the global warming and brightening periods described in the literature (e.g., Wild et al. 2007).
The comparison with the historical summer CC and growing season temperature reconstructions from Scandinavia (ScanCC and ScanT, respectively), also shows a common pattern of cloudiness during the central part of the LIA (Fig.8). ScanCC displays a persistent decrease in summer CC from the beginning of the 17 th century until the end of the 18 th century which is consistent with the sunnier summer period described for the Pyrenees.

Discussion
The correlations of UPF δ 13 C and monthly T, CC, SD and SSR indicate a close link of summer climatic conditions and tree growth, but also reveal high co-variance among all four meteorological variables. The short length of the SSR records, which limits the common period of analysis, and the fact that only two major volcanic eruptions occurred during this period, hinders the unequivocal attribution of changes in δ 13 C to one main climatic driver.
However, linear regression models performed for the different sets of climate variables reveal the better explaining and predictive skills of the SSR models and the larger spatial significance of the relationship between UPF δ 13 C and SSR. In addition, the test on the volcanic imprint on the long instrumental records from Barcelona BaT and BaCC and the UPF δ 13 C identified a significant effect (cooling) attributable to volcanic eruptions in BaT, whereas no clear volcanic signal could be detected in BaCC and UPF δ 13 C, strongly suggesting that temperature changes are not the main driver of δ 13 C variations in tree rings at UPF.
Proxy records encoding temperature signals are expected to display a significant change in the values after large volcanic eruptions as a consequence of the decrease in local temperatures that usually follows these events (D'Arrigo et al., 2013). In contrast, as shown by the pioneer study by Battipaglia et al. (2007), large volcanic eruptions producing a regional to global significant cooling did not lead to a significant reduction of tree photosynthetic rates in Italy. Accordingly, the interpretation of UPF δ 13 C as non-temperature proxy record is further supported by the lack of a significant volcanic imprint over the last six centuries, regardless of the sub-sample of volcanic eruptions considered, while PyrT displays significant decreases in temperatures in the year of the eruption and in a few subsequent years. Thus, major volcanic events during the last six centuries reduced tree growth at the Pyrenees probably by inducing a decrease in temperatures that may have shortened the growing season.
However, neither the reduction of the length of the growing season nor the increased concentration of stratospheric aerosols did affect the δ 13 C record, which we interpret as a lack of influence of volcanic eruptions on summer photosynthetic activity at this site.
This lack of impact of volcanic eruptions on δ 13 C may seem puzzling at first sight, but in theory the δ 13 C record reflects the PAR and not totally reflects SSR. According to the hypothesis of the diffuse SSR/PAR-compensation proposed by Mercado et al. (2009), the reduction in direct PAR due to increases in clouds or aerosols is compensated by the increase of the diffuse fraction of PAR, which may explain the lack of a significant volcanic imprint in δ 13 C. Although the diffuse SSR (PAR) may not always totally compensate the reduction on the direct SSR (PAR) (e.g., Ogle et al., 2005), this approach provides a useful test-bed to disentangle the role of the diffuse radiation in maintaining the photosynthetic rate under low-transmittance skies. Nonetheless, the hypothesis of the diffuse light compensation may not be the only reason for the maintenance of the photosynthetic rates. Changes in cloudiness or atmospheric aerosol (such as those induced by volcanic eruptions) may not only alter the direct/diffuse fractions, but also the ratio PAR/SSR reaching the surface. The limited availability of direct measurements of PAR causes that PAR records are often estimated as a fix proportion of SSR. However, such a proportion is known to change under lower transmittance conditions since clouds, dust and aerosols shows higher transparency to PAR (0.4-0.7µm) than to other fractions of the SSR spectrum such as the infrared wavebands (0.7 to 1.7 µm) (Papaioannou et al., 1993;Jacovides et al., 2003;Bat-Oyun et al., 2012). Thus, the lack of significant changes on δ 13 C under cloudy and dusty conditions derived from volcanic eruptions may be due to either the increase in the diffuse fraction of PAR, or to the general increase of the PAR/SSR ratio. The increase in summer SSR observed in ScanCC and PyrSSR during the recent decades is in line with the widespread surface brightening observed since the 1980s (Wild, 2009;Wild, 2012), which also has been observed in Spain (Sanchez-Lorenzo et al., 2007, 2013b. However, this brightening period was exceeded by far during the LIA, when both records ScanCC and PyrSSR show sunnier summers than nowadays. These results also agree with those described in Scandinavia (Gagen et al., 2011, Loader et al., 2013 and in the Alps (Hafner et al., 2014). Thus, colder growing season temperatures in Scandinavia, Alps and Pyrenees during LIA were associated to higher summer SSR.
While lower temperatures during the LIA have been associated to the lower TSI and the increased concentration in atmospheric aerosols as a result of periods of volcanism (Crowley, 2000;Miller et al., 2012), the mechanism driving SSR changes are not clear yet. Loader et al. (2013)  conditions due to the dominance of Arctic and maritime air masses. At this point, we can only speculate about the dynamical processes that gave rise to the increase of SRR during the LIA in these three regions. The fact that all of them display a similar signal during a cold periodmaybe indicating an overall reduction of evaporation from the ocean as a result of lower sea-surface-temperatures. The accompanying reduction in summer cloud cover over continental Europe could be the main factor rather than changes in large-scale atmospheric circulation.

Conclusions
The joint analysis of instrumental records of different variables related to incoming sunlight, near-surface temperature and δ 13 C tree-ring chronology located in Northeast Spain (Southern Europe), indicates that SSR plays a major role among the drivers of summer carbon fractionation in tree-rings in this region. Also, the SEA applied to different sets of volcanic eruptions and the comparison between the long δ 13 C chronology and temperature reconstructions from this region, rules out δ 13 C as a temperature proxy. We thus interpret the centennial δ 13 C record as an indicator of past SSR which allowed the reconstruction of incoming sunlight over the last 600 years.
The relationship between past temperature and past SSR at the Pyrenees shows no clear relationship through the 600 years as for example temperature and SSR were positively correlated during the MCA but anticorrelated during the LIA.
Overall, the comparison across the existing tree-ring δ 13 C records encoding sunlight-related signals revealed that the brightening phase since 1980s is not unprecedented in the context of the last centuries and LIA appears as a sunnier period in the different tree-ring δ 13 C records.
Our results show the potential of using volcanic eruptions to discern the δ 13 C chronologies that could potentially be used to extend the geographical coverage of reconstructions of incoming sunlight, contributing to better a understanding of the interaction between past temperatures and SSR on continental scales, a key parameter contributing to global climate sensitivity.