10 Be in late deglacial climate simulated by ECHAM5-HAM – Part 1: Climatological influences on 10 Be deposition

Abstract. Reconstruction of solar irradiance has only been possible for the Holocene so far. During the last deglaciation, two solar proxies (10Be and 14C) deviate strongly, both of them being influenced by climatic changes in a different way. This work addresses the climate influence on 10Be deposition by means of ECHAM5-HAM atmospheric aerosol–climate model simulations, forced by sea surface temperatures and sea ice extent created by the CSIRO Mk3L coupled climate system model. Three time slice simulations were performed during the last deglaciation: 10 000 BP ("10k"), 11 000 BP ("11k") and 12 000 BP ("12k"), each 30 yr long. The same, theoretical, 10Be production rate was used in each simulation to isolate the impact of climate on 10Be deposition. The changes are found to follow roughly the reduction in the greenhouse gas concentrations within the simulations. The 10k and 11k simulations produce a surface cooling which is symmetrically amplified in the 12k simulation. The precipitation rate is only slightly reduced at high latitudes, but there is a northward shift in the polar jet in the Northern Hemisphere, and the stratospheric westerly winds are significantly weakened. These changes occur where the sea ice change is largest in the deglaciation simulations. This leads to a longer residence time of 10Be in the stratosphere by 30 (10k and 11k) to 80 (12k) days, increasing the atmospheric concentrations (25–30% in 10k and 11k and 100% in 12k). Furthermore the shift of westerlies in the troposphere leads to an increase of tropospheric 10Be concentrations, especially at high latitudes. The contribution of dry deposition generally increases, but decreases where sea ice changes are largest. In total, the 10Be deposition rate changes by no more than 20% at mid- to high latitudes, but by up to 50% in the tropics. We conclude that on "long" time scales (a year to a few years), climatic influences on 10Be deposition remain small (less than 50%) even though atmospheric concentrations can vary significantly. Averaged over a longer period, all 10Be produced has to be deposited by mass conservation. This dominates over any climatic influences on 10Be deposition. Snow concentrations, however, do not follow mass conservation and can potentially be impacted more by climate due to precipitation changes. Quantifying the impact of deglacial climate modulation on 10Be in terms of preserving the solar signal locally is analysed in an accompanying paper (Heikkila et al., 10Be in late deglacial climate simulated by ECHAM5-HAM – Part 2: Isolating the solar signal from 10Be deposition).


Geoscientific Instrumentation
Methods and Data Systems The Cryosphere This discussion paper is/has been under review for the journal Climate of the Past (CP). Please refer to the corresponding final paper in CP if available. model simulations, forced by sea surface temperatures and sea ice extent created by the coupled climate system model CSIRO Mk3L. Three time slice simulations were performed during the last deglaciation: 10 000 BP ("10k"), 11 000 BP ("11k") and 12 000 BP ("12k"), each 30 yr long. The same 10 Be production rate was used in each simulation to isolate the impact of climate on 10 Be deposition. The changes are found to follow 10 roughly the reduction in the greenhouse gas concentrations within the simulations. The 10k and 11k simulations produce a surface cooling which is symmetrically amplified in the 12k simulation. The precipitation rate is only slightly reduced at high latitudes, but there is a northward shift in the polar jet in the Northern Hemisphere and the stratospheric westerly winds are significantly weakened. These changes occur where the

Introduction
Cosmogenic radionuclides, such as 10 Be and 14 C, are commonly used proxies for solar activity. Their production rate in the atmosphere responds to changes in cosmic ray 5 intensity, which again is modulated by solar activity and the strength of the geomagnetic field. Their transport from the source to natural archives is subject to climatological modulation: atmospheric circulation and precipitation in the case of 10 Be and reservoir exchange times in the case of 14 C. Comparison of both radionuclides allows for the separation of the common production, or solar, signal from the climate modulation. 10 Recent results show that during the Holocene the climate impact remains relatively small and the solar signal can be retrieved (e.g. Steinhilber et al., 2012). However, extending the solar forcing function into the last deglaciation becomes challenging as large discrepancies arise between 14 C and 10 Be (e.g. Muscheler et al., 2004). The reason for these differences has to be understood before the solar forcing function can 15 reliably be extended to this period. In order to address the sensitivity of 10 Be deposition onto polar ice sheets to climate changes we perform time slice simulations within the deglacial climate at three different stages: 12 000 BP (years before 1950 CE), 11 000 BP and 10 000 BP. A preindustrial control simulation is also performed for comparison. These simulations are 20 referred to as "ctrl", "12k", "11k" and "10k" throughout the manuscript. The model employed is the ECHAM5-HAM chemistry-transport model. This incorporates the aerosol module HAM, which describes aerosol emission, physical and chemical as well as deposition processes. This model only describes the atmosphere and hence requires the sea surface temperatures and sea ice cover to be described. In this study these vari-Introduction  . ECHAM5-HAM resolves aerosol processes explicitly. This is important in order to obtain a realistic picture of aerosol deposition and its climate modulation, but at the same time it increases the runtime of the model. Such a time-slice approach focusing on periods of special interest, such as the Last Glacial Maximum or the mid-Holocene, is commonly used in model intercomparison experiments (e.g. Braconnot et al., 2012), 5 or when addressing questions such as atmospheric particle transport under a changing climate (e.g. Krinner et al., 2010;Mahowald et al., 2006). The focus of this study is to investigate the impact of the deglacial climate on 10 Be deposition. The reconstructing of the solar signal from the 10 Be deposition rate during this period is addressed in an accompanying paper (Heikkilä et al., 2013). In order to 10 be able to distinguish the imprint of climate only, and also because the actual solar activity and therefore the 10 Be production rate are unknown, we use the same (theoretical) production rate in all simulations. This means that even if the model climate in each time slice simulation is quite different, the global mean 10 Be deposition rate will be equal due to mass conservation. Because of this a direct comparison with mea-15 sured 10 Be concentrations is not possible. In terms of climate these time slice simulations produce typical climatic conditions given the boundary conditions determined by greenhouse gas concentrations (GHGs), orbital parameters and aerosol load. The internal climate variability produced by the model will be representative in a statistical sense under these boundary conditions, although the actual timing and magnitude of 20 individual weather events will not be reproduced. Comparison of 10 Be and 14 C during the Younger Dryas has previously been used to detect changes in North Atlantic deep-water formation and the carbon cycle (Muscheler et al., 2000(Muscheler et al., , 2004: however, both 10 Be and 14 C are influenced differently by changes in climatic conditions which hampers the detection of the common solar signal. To the 25 best of our knowledge this is the first model study to address 10 Be deposition changes due to climate changes beyond the Holocene.

Description of the simulations
Boundary conditions for the ECHAM5-HAM simulations were obtained from simulations of the pre-industrial ctrl, 10k, 11k and 12k climate conducted using the CSIRO Mk3L climate system model. This is a coupled general circulation model, consisting of components that describe the atmosphere, land, sea ice and ocean (Phipps et al., 5 2011(Phipps et al., 5 , 2012. The model was integrated for 5000 yr in each case to let the system reach thermal equilibrium and the last 35 yr were then used to provide boundary conditions for the ECHAM5-HAM simulations of the atmosphere. For each time slice simulation, the CSIRO Mk3L simulations used constant orbital parameters and greenhouse gas concentrations (obtained from Liu et al., 2009, see Table 1). The total solar irradiance was held fixed at 1365 W m −2 in each experiment. Land ice extents were prescribed according to the ICE-5G reconstruction v1.2 (Peltier, 2004). No changes were made to global sea level or to the positions of the coastlines. The ECHAM5-HAM atmospheric model used sea surface temperatures (SST) and sea ice cover generated by the CSIRO Mk3L model as ocean boundary conditions.

15
The greenhouse gas concentrations, orbital parameters, orography and land sea mask were set to the same values as in the CSIRO Mk3L simulations according to the PMIP2 protocol (http://pmip2.lsce.ipsl.fr/). The SSTs and sea ice were updated monthly. Aerosol load was set to preindustrial values for the simulations and was taken from the AEROCOM aerosol model-intercomparison experiment B, available at 20 http://nansen.ipsl.jussieu.fr/AEROCOM. The ECHAM5-HAM model was run for 35 yr, from which the first 5 yr were used to spin up the model and discarded. The horizontal resolution used was the T42 spectral resolution, corresponding to ca. 2.8 • or ca. 300 km. The model top was at ca. 30 km, including 31 vertical levels. The model output was analysed monthly. In order to analyse the atmospheric transport path of Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | for a more realistic particle transport towards the poles following Bourgeois and Bey (2011). The 10 Be production rate on the monthly scale was constructed using a lowfrequency solar activity parameter Φ based on a reconstruction of the 14 C production rate, and a modern 11 yr cycle was added on top. This leads to a 10 Be production 5 rate which is ca. 70 % larger than typical values during the last ca. 30 yr (see for ex. Heikkilä and Smith, 2012). Table 1 summarises the global mean temperatures and precipitation rates over each 10 30 yr period. It is clear from the results that the global mean temperature closely follows the GHGs. The GHGs during 10k and 11k are fairly similar, leading to a cooling of ca. 1 • from ctrl. The 12k simulation is cooler, 2.7 • relative to ctrl, due to a larger reduction of the GHGs. These differences are somewhat larger than the 0.6 • global cooling during the Younger Dryas (12.9 to 11.7 ka BP) reported by  son (2010). An explicit comparison is hard because the model simulations present two 30 yr snapshots during this period. Shakun and Carlson (2010) analysed all available proxy data worldwide but large areas, especially oceans, are still uncovered. Because of this their global mean change would be milder if ocean temperatures were given more weight. However, the model simulations presented here have been fully equilibri-20 ated, and hence the magnitude of the cooling might exceed the transient response of the climate system. The precipitation rate varies less between the simulations, however a linear response to cooling, albeit a weak one, is evident (Fig. 1). the stratosphere, and the residence time is of the order of one year in the stratosphere and ca. three weeks in the troposphere (e.g. Heikkilä and Smith, 2012;Koch et al., 2006;Land and Feichter, 2003). The total contents of 10 Be in the stratosphere and the troposphere in these simulations are slightly larger than modern values due to the higher production rate.

5
During the deglaciation the stratospheric fraction of production is slightly increased due to a lower tropopause height. Again, this change is larger during 12k than 10k and 11k. Moreover, the fraction of wet to total deposition is decreased to 87-88 % in all simulations, which is a reduction relative to the typical simulated present-day value of 91-92 % produced by this model (Heikkilä et al., 2009). The stratospheric and 10 tropospheric residence times are both increased, leading to higher atmospheric 10 Be contents. This is likely to be due to circulation changes which will be discussed in the following section.

Changes in atmospheric circulation
We first investigate the general state of climate in the deglaciation simulations com-15 pared with the control climate. Figure 2 shows the zonal mean temperature change relative to ctrl in all simulations. The 10k and 11k simulations show a moderate cooling in the tropical upper troposphere and a warming of the stratosphere. The cooling is slightly increased in the 11k simulation, including a cooler region in the northern high latitude lower levels. This pattern is intensified in the 12k results with the 50-90 • N 20 lower level cooling becoming stronger. A surface cooling and stratospheric warming in response to reduced atmospheric GHGs is consistent with the nature of the observed response to increased GHGs during the 20th century (Trenberth et al., 2007). Zonal mean westerly winds (Fig. 3) show a similar response: a slight weakening of the SH mid-latitude and NH low latitude stratospheric winds and a slight intensification in the 25 troposphere (10k and 11k). In the 12k results, again, these changes are intensified. The NH polar jet is shifted by ca. 5 • towards the north. A similar shift, although weaker, is observed in the SH midlatitude jet. The weakening of the stratospheric winds leads 3687 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | to a slower circulation and hence less efficient particle transport, causing the stratospheric residence time of 10 Be to increase (Table 1). The shift found in the NH tropospheric westerlies can possibly be connected to the position of storm tracks and therefore to precipitation patterns. However, no such shift is obvious in the precipitation rate (Fig. 4), nor in the sea level pressure (not shown).

5
There is a general decrease in the simulated precipitation rate in the mid-and high latitudes (both NH and SH), in agreement with the cooling caused by the lower GHG concentrations (see Fig. 1). At low latitudes, however, there is an increase in precipitation of up to 50 %, mostly in dry areas (Sahara, west of the South American and African continents). In order to study the intensity of internal climate variability we performed 10 an EOF analysis for the sea level pressure. The first EOF of SLP (associated with NAO in the North Atlantic sector and SAM south of 20 • in the SH, Fig. 5) does not exhibit significantly larger variability in the deglaciation simulations than in the control, suggesting that the internal climate variability described by these modes did not change significantly.

15
The differences between the simulations seem fairly symmetric except for the shift of the zonal wind in the NH troposphere poleward of 50 • . The reason for this asymmetry might be the larger sea ice content in the deglaciation simulations, especially in the NH (Fig. 6). Sea ice conditions influence atmospheric circulation patterns, such as sea level pressure, strength of the westerlies and position of the storm tracks, in the 20 northern mid-and low latitudes and even globally via teleconnections affecting largescale circulation patterns such as ENSO (e.g. Alexander et al., 2004;Blüthgen et al., 2012;Budikova, 2009;Screen and Simmonds, 2010). However, we find a decrease in geopotential height at the surface where the sea ice increase is largest, both in the NH and the SH, although the anomaly is visible at 500 hPa only in the NH. Thus the 25 1000-500 hPa thickness is reduced in the deglaciation simulations due to increased sea ice. The reduction is largest near the Hudson Bay area where sea ice differences are the most pronounced. This is consistent with Overland and Wang (2010)  tropopause height is seen at 35 • N. This suggests that the anomalous NH temperature, shift in the westerlies and longer atmospheric residence time of 10 Be are resulting from the increased sea ice in the deglaciation simulations.

Circulation influence on atmospheric distribution of 10 Be
The changes found in climatic conditions of the deglaciation simulations, the weaken-5 ing of the stratospheric jets, the northward shift in the westerlies and the changes in tropopause height all have the potential to affect particle transport and deposition and hence 10 Be. In the following we investigate the modelled changes in 10 Be between the deglaciation simulations and the control climate. Firstly we investigate whether the general relationship between GHGs, surface temperature and precipitation rate could 10 be used to predict the 10 Be deposition under different climatic conditions. In addition to temperature and precipitation, Fig. 1 shows the global and annual mean 10 Be deposition flux (crosses) as a function of mean surface temperature for each year of each simulation. No relationship between global mean temperature and 10 Be deposition is obvious because the amplitude of the 11 yr cycle dominates the changes in tempera-15 ture or precipitation rate. If averaged over the entire 30 yr period the 10 Be deposition change is zero, as determined by the fact that the production rate is the same. Hence, a simple linear relationship between mean climate and 10 Be deposition can not be assumed. Figure 7 illustrates the mean change in 10 Be deposition flux. Because the production 20 rate of 10 Be was the same in all simulations the global mean change equals zero. 10 Be deposition is not significantly changed at mid to high latitudes in the 10k and 11k simulations. It is increased in the NH over Siberia and Greenland, with the exception of Scandinavia and the Hudson Bay area, where changes in sea ice extent are largest.
In the SH the 10 Be deposition is decreased over the Southern Ocean but generally 25 increased over Antarctica. The change in deposition is larger in the tropical region and its sign varies. Then the difference in deposition in the 12k simulation is similarly distributed but more pronounced than in the 10k and 11k simulations, especially at low 3689 Introduction latitudes. However, the simulated change at 12k becomes negative over North America whereas it is slightly positive at 10k and 11k. Over Greenland the deposition increases less at 12k than at 10k or 11k. The changes in 10 Be deposition partly follow the precipitation change, especially in the tropics. Also the positive change over Greenland and the negative one over Scandi-5 navia and the Hudson Bay area are consistent with the precipitation change. However, the precipitation changes at 12k are larger than at 11k or 10k which is not seen in the 10 Be deposition change, and the strongly negative precipitation change at 12k over North America does not correspond to the 10 Be deposition increase there. Table 1 shows that the global mean fraction of wet to total 10 Be deposition is similar in all sim-10 ulations, however it is reduced from the present day value of 91-92 % (Heikkilä et al., 2009). Figure 8 shows the spatial distribution of the change in wet fraction relative to ctrl. The change is small but roughly follows the precipitation change. In Scandinavia and the Hudson Bay area the wet fraction is increased instead of decreased although the 10 Be deposition is decreased, suggesting a dilution effect. In Greenland, where the 15 deposition change differs from the surrounding areas, no change is apparent in the wet fraction. In total the wet fraction change remains small (up to 15 %) but locally the dry deposition and sedimentation increase by over 50 %. However, wet deposition remains the dominant method of removal of 10 Be from the atmosphere. To summarise, the 10 Be deposition differences globally are due to a combination of precipitation and 20 atmospheric circulation changes, triggered by the increase in sea ice extent.
The circulation changes at mid-and high latitudes largely affect particle transport and dry deposition, which causes the air concentration of 10 Be to increase (Fig. 9).
The largest increase in concentrations is found in the tropical tropopause region, indicating a weaker stratospheric transport. Outside of these regions there is a rather 25 homogeneous increase of ca. 5 % in the 10k and 11k simulations. In the 12k simulation, however, the increase in the tropical tropopause region is much larger at up to 80 % and a significant increase (ca. 40 %) is also apparent in the upper troposphere at high latitudes. The 10 Be concentrations are also increased throughout the strato- sphere. These changes follow the pattern of change of age of air between the simulations, derived from the modelled 10 Be/ 7 Be ratios (Fig. 10). 10 Be and 7 Be are produced in a near-constant ratio (gradually decreasing from 0.3 at ca. 30 km altitude to 0.8 near the surface, Heikkilä et al., 2008) which increases because 7 Be decays with its half-life of 53.2 days. The largely increased age of air in the stratosphere is consistent with the 5 reduced strength of stratospheric zonal winds (Fig. 3). The differences are smallest at high latitudes where the production of new 10 Be and 7 Be decreases the mean ratio. These changes in atmospheric circulation, 10 Be air concentrations and transport path of 10 Be from source to archive can potentially influence the mixture of 10 Be from different atmospheric source regions in deposition. The 10 Be production varies by or-10 ders of magnitude across latitude and altitude. Furthermore, the production is more strongly modulated by varying solar activity at high latitudes and altitudes, where it is highest to begin with. Therefore, a change in source-to-archive atmospheric transport path of 10 Be could lead to a catchment of 10 Be from a production area reflecting solar activity with a different amplitude from the global mean production. Figure 11 illustrates 15 the source-to-archive distribution of 10 Be for each of the simulations. The general distribution in these simulations is very similar to the previously studied present day situations (Heikkilä and Smith, 2012, their Fig. 12): 10 Be produced in the stratosphere is mainly transported into the troposphere via the mid-latitudes and deposited equally between the 0-30 • and 30-60 • latitude bands in each hemisphere. The tropospheric 20 production of 10 Be is mostly deposited locally and the neighbouring latitude bands and rarely makes it to the opposite hemisphere. The figure shows that the differences between the control simulation and the deglaciation simulations are fairly small. In the SH the production and deposition distribution is practically unaffected by climatic changes in these simulations. The circulation change can be seen in latitude bands 30-60 • N 25 and 60-90 • N where the fraction of locally produced tropospheric 10 Be in deposition is reduced by up to 8 % and exchanged within the next latitude band (more of the 10 Be produced at 30-60 • N is deposited at 60-90 • N and vice versa). The reduced deposition and longer residence time allows for more longer-range transport within the tropo- sphere. The stratospheric production of 10 Be (ca. 2/3 of total), which is also the most heavily modulated, is unaffected by the deglacial climate. The fairly large differences in atmospheric circulation found within the 10k, 11k and 12k simulations do not influence 10 Be transport as much as between the control and any of the deglaciation simulations.
These differences are comparable in amplitude with those caused by different model 5 resolutions or realisations in Heikkilä and Smith (2012).

Summary and conclusions
This study investigates climatic influences on 10 Be deposition during the last deglacial climate. Three time slice simulations were performed, each of 30 yr duration: 10 000 BP (10k), 11 000 BP (11k) and 12 000 BP (12k). These simulations were compared with a control (ctrl) simulation representing the preindustrial climate. The model employed was the atmospheric ECHAM5-HAM aerosol-climate model, driven by sea surface temperatures and sea ice cover produced by the coupled atmosphere-ocean-sea ice-land surface model CSIRO Mk3L. This paper, presenting the first part of this study, focuses on changes in atmospheric mean circulation and climate and their impact on the at-15 mospheric distribution and deposition of 10 Be. The second part (Heikkilä et al., 2013) investigates the influence of deglacial climate on 10 Be deposition in terms of preserving the solar signal. In order to separate the pure climate modulation of 10 Be, and also because the solar activity during the deglaciation is not known, the 10 Be production was kept the same in all simulations. Hence the results are not suited for comparison 20 with actual observations. Generally the results follow changes in the greenhouse gas concentrations. Surface temperatures are colder and the precipitation rate reduced in the 10k and 11k simulations from ctrl, and these differences are amplified in the 12k simulation. Similarly, stratospheric zonal winds are weakened in the 10k and 11k simulations, and more locally in Scandinavia and the Hudson Bay area, leading to a decreased 10 Be deposition in these areas although the 10 Be deposition is generally decreased at midand high latitudes. Sea ice changes increase the fraction of wet deposition to total as opposed to the surrounding areas where dry deposition is increased due to reduced the precipitation rate. The reduced stratospheric winds lead to a reduced 10 Be trans-5 port, increasing the stratospheric residence time by 30 (10k and 11k) to 80 (12k) days. Tropospheric residence time is also increased by a few days, mainly at mid-to high latitudes due to the circulation changes. Atmospheric concentrations of 10 Be are strongly increased (> 50 %) in the upper troposphere at mid-to high latitudes and in the tropical tropopause region. These circulation changes seem to be attributed to the increase 10 in sea ice extent in the deglaciation simulations, leading to anomalies in geopotential height and shifts in position of the polar jet. In the Southern Hemisphere (SH) the disturbances due to increased sea ice are only evident near the surface. Despite the changes found in atmospheric circulation and the regionally varying response of 10 Be air concentrations the source-to-archive distribution is fairly similar in 15 all simulations. The largest change was found in the NH polar troposphere due to the reduced 10 Be deposition, leading to up to 8 % less deposition of locally produced 10 Be, compensated by an increase of 0-30 • N 10 Be. Most of 10 Be produced in the stratosphere is deposited between 60 • S and 60 • N, uninfluenced by the deglacial climate. Generally the longer residence time in the deglaciation experiments allows for slightly 20 increased long-range tropospheric transport in the NH. Due to the longer stratospheric residence time the mixing of the large production gradients of 10 Be along latitude is more thorough.
These results indicate that 10 Be deposition is mostly determined by mass balance. If the production rate does not change, the global mean deposition change will be zero. 25 Regionally the changes might be significant but our results found only slight evidence for that. The increase in dry deposition by up to 50 % locally did not influence the spacial deposition variability much because its total contribution remains small (12-13 % in all simulations). Varying climatic conditions and atmospheric circulation mainly affect the atmospheric residence time (80 days at most in these experiments) leading to a later deposition and a more thorough mixing of 10 Be. However, averaged over a few years, this effect is smoothed out. In fact the longer residence time in colder climates leads to an increased mixing of 10 Be in the atmosphere and therefore a more homogeneous deposition pattern. Combining all these changes the total deposition 5 change was maximally 50 %. This is in agreement with the results of Alley et al. (1995), who found only moderate changes in 10 Be during the Younger Dryas. Given these results, 10 Be deposition flux would be the optimal proxy for solar activity in an ideal archive because, following mass balance, the atmospheric production of 10 Be will have to be deposited at the surface within a few years. In case of 10 Be snow 10 concentrations the mass balance does not necessarily hold due to potential changes in precipitation rate. In reality, however, the 10 Be deposition flux has to be derived from the reconstructed snow accumulation rate which adds some uncertainty to the record.               Fig. 11. Percentage of 10 Be, produced in a given atmospheric compartment (x-axis, 'STR' indicates the stratosphere), which is deposited at a given latitude band (legend) in the 'ctrl', '10k', '11k' and '12k' simulations. Fig. 11. Percentage of 10 Be, produced in a given atmospheric compartment (x-axis, "STR" indicates the stratosphere), which is deposited at a given latitude band (legend) in the "ctrl", "10k", "11k" and "12k" simulations.