Production rate and climate influences on the variability of 10Be deposition simulated by ECHAM5-HAM: Globally, in Greenland, and in Antarctica


  • U. Heikkilä,

    Corresponding author
    1. Australian Nuclear Science and Technology Organisation (ANSTO), Lucas Heights, NSW, Australia
    • Corresponding author: U. Heikkilä, Australian Nuclear Science and Technology Organisation (ANSTO), Locked bag 2001, Kirrawee DC, 2232 NSW, Australia. (

    Search for more papers by this author
  • A. M. Smith

    1. Australian Nuclear Science and Technology Organisation (ANSTO), Lucas Heights, NSW, Australia
    Search for more papers by this author


[1] Ice core concentrations of 10Be are used as a proxy for solar activity, but they might be affected by atmospheric transport and deposition and their changes. During the Holocene, the influence is likely to be small, but during glacials it has to be accounted for. First, the climate influence has to be understood during the present climate. This study uses an ECHAM5-HAM 30-year climatological simulation of 10Be to investigate the production and climate-related influences on 10Be deposition with focus on Greenland and Antarctica. We examine the climate modes driving snow accumulation and hence potentially 10Be deposition over a climatologically relevant period. The North Atlantic Oscillation (NAO) is found to be the main driver of changes in precipitation and 10Be deposition in Greenland, in agreement with previous studies. In Antarctica, the picture is more complex as precipitation and 10Be deposition are only weakly correlated with the Southern Annular Mode (SAM), El Niño-Southern Oscillation (ENSO), or Zonal Wave 3 pattern (ZW3). The results suggest that on seasonal scale, 10Be deposition is linked with both precipitation rate and tropopause height, mainly due to the similar seasonal cycle. However, the correlation with tropopause height persists on the annual time scale. All in all, 10Be variability in Antarctica is an interplay of several processes whose contribution varies in time and space. When interpreting 10Be ice core records for solar activity, the time scale is essentially important. On seasonal scale, the 10Be signal is dominated by weather influences, but on multiannual scales, the production rate is the main driver. On multidecadal scale, large long-term trends in climatic factors have the potential to distort the signal again as is seen in 10Be records during glacials. This study shows how climate modes connect to 10Be variability and how this connection could be used to correct for the climate impact. The established connections during present climatic conditions can be used as a basis to investigate these connections during glacial climate in a glacial model simulation.

1 Introduction

[2] Cosmogenic beryllium-10 (10Be, half-life 1.36 × 106 years) is a widely used proxy for past solar activity as well as past geomagnetic field strength [e.g., Beer et al., 1990; Foukal et al., 2006; Korte and Muscheler, 2012; Muscheler et al., 2007; Raisbeck et al., 1990, 2006; Steinhilber et al., 2008; Vonmoos et al., 2006; Wagner et al., 2000]. 10Be records covering a few hundred years and more are most commonly obtained from polar ice sheets. The regular layering of the ice allows for creating an ice chronology, and hence time series of 10Be concentrations can be retrieved. The source of cosmogenic 10Be is in the upper atmosphere where it is produced by secondary particles of cosmic rays colliding with atmospheric nitrogen and oxygen atoms [e.g., Lal and Peters, 1967; Masarik and Beer, 1999; Webber and Higbie, 2003; Kovaltsov and Usoskin, 2010]. The intensity of cosmic rays in the atmosphere is modulated by the solar activity and the strength of the geomagnetic field so that the amount of 10Be produced reflects these changes. Due to the dipole geomagnetic field, the modulation of 10Be production is strongly latitude dependent. Hence, the amplitude of the temporal variability of the 10Be snow concentrations depends on the atmospheric source region of 10Be. This is determined by the atmospheric pathways 10Be takes from its source to polar regions. Another process which influences the 10Be snow concentrations is its removal from the atmosphere either by wet or dry deposition or sedimentation. The atmospheric transport path and the removal processes are sometimes referred to as the “climate impact” on 10Be as opposed to the production influence.

[3] The time scales related to atmospheric processes are often seasonal or shorter. Trends and longer time scale variability are possible as well, but their amplitude is usually smaller than the seasonal variability. Long 10Be ice core records typically have a temporal resolution of a few years, which smooths out the seasonal cycle. During a relatively stable climatic period, such as the Holocene, the impact of seasonal variability is smaller than production variability in most records. This is confirmed by the common solar signal found in cosmogenic radionuclide ice core records from Greenland and Antarctica [Beer et al., 1992; McCracken et al. 2004] as well as between 10Be and the production rate of another cosmogenic radionuclide 14C inferred from tree ring concentrations [e.g., Beer et al., 2011; Knudsen et al., 2009; Muscheler et al., 2000, 2004]. Moreover, the global carbon cycle, which is also influenced by climate, has additional confounding effect upon the archived 14C. Yet all records reveal the same periods of solar activity, even during the last glacial [e.g., Beer et al., 2002].

[4] The assumption of a relatively stable climate (in terms of not distorting the 10Be production signal) is only valid during the Holocene. During glacial periods, snow accumulation has exhibited large variations [e.g., Johnsen et al., 1992; Jouzel et al., 2001] which transfer into 10Be snow concentrations [Beer et al. 2011] and prohibit the use of 10Be as solar activity proxy beyond the Holocene without applying some kind of a correction. Unfortunately, the relationship between 10Be deposition and snow accumulation changes is not linear, and 10Be records cannot as a general rule be corrected for the climate impact by simply subtracting the snow accumulation signal. Furthermore, the snow accumulation is reconstructed and subject to dating errors if the seasonal cycle is not resolved, which hampers the correction of 10Be for snow accumulation changes. The climatic factors driving the 10Be deposition and its relation to precipitation have to be understood in order to filter out the “climate impact” from the 10Be records and to extend the solar activity reconstruction into the glacial time. First of all, a connection between production rate and climate variability has to be established during the present when the climate is relatively stable, a relatively large number of high-resolution observations of beryllium isotopes exist, and the state of atmosphere is quite well known from reanalysis. Once these relationships have been understood in the present, their temporal and spatial scales under a changing climate can be studied.

[5] A number of observational studies have examined 10Be and 7Be and climate. 7Be is produced in a very similar way as 10Be but has a short half-life of 53.2 days. Unfortunately, these records often cover a few years only, but some longer records exist. Two 8 year records of 10Be and 7Be in precipitation in Switzerland reveal a seasonal cycle with a large amplitude [Heikkilä et al. 2008c]. The maximum in 10Be and 7Be deposition and rainwater concentration occurs during summer but in 10Be/ 7Be ratios, it occurs in late winter and spring, reflecting stratosphere-troposphere exchange. A 25-year 10Be and 7Be surface air filter record from coastal Antarctica exhibits a summer (DJF) maximum as well, but the 10Be/ 7Be ratio peaks during late summer [Elsässer et al. 2011]. 10Be measured in Greenland [Heikkilä et al. 2008b] and Antarctic [Pedro et al. 2011b] snow pits suggests that on seasonal time scales the variability is driven by climatic factors, such as precipitation and atmospheric circulation, but on annual time scales the production rate is the dominant signal. Baroni et al. [2011] used two 60-year Antarctic 10Be records from Vostok and Concordia to investigate the influence of volcanic eruptions and production rate on 10Be and found enhanced concentrations after the major volcanic eruptions. Multiyear 10Be and 7Be measurements in air filters in Scandinavia were found to correlate with the sunspot number [Aldahan et al. 2008]. Cannizzaro et al. [2004] investigated a 21-year 7Be surface air concentration record and also found strong correlation with sunspots. Koch and Mann [1996] used a series of 19 7Be worldwide surface air measurements covering 22 years and found the sunspot as well as the ENSO signals in the data. The use of 7Be is slightly complicated by its decay which shifts the timing of maximum concentrations depending on the age of 7Be [Heikkilä et al. 2008c].

[6] In this study, we investigate the atmospheric drivers of 10Be deposition using a climatological (30-year) simulation of 10Be in the atmosphere. The model employed is the ECHAM5 global circulation model coupled with the aerosol module HAM. The focus regions are Greenland and Antarctica, where most ice cores originate from. We seek to establish a connection between different climatic phenomena and 10Be deposition and to investigate the extent they may distort 10Be production signal. To our knowledge, this is the first climatological type of investigation concentrating on connections between atmospheric and 10Be variability. Previous efforts to simulate the atmospheric transport and deposition using the GISS-ModelE [Field et al. 2006] and ECHAM5-HAM [Heikkilä et al., 2008a, 2008b; 2009] have investigated an atmospheric mean state only or the 21st century climate, trends, and grand solar minima [Field and Schmidt, 2009].

2 Data and Methods

[7] The model employed for this study is the ECHAM5-HAM (hereafter ECHAM). ECHAM5 is a fifth-generation atmospheric global circulation model (GCM) developed at the Max Planck Institute for Meteorology, Hamburg, evolving originally from the European Centre of Medium Range Weather Forecasts (ECWMF) spectral weather prediction model. It solves the prognostic equations for vorticity, divergence, surface pressure, and temperature, expressed in terms of spherical harmonics with a triangular truncation. Nonlinear processes and physical parameterizations are solved on a Gaussian grid. A complete description of the ECHAM5 GCM, including the treatment of greenhouse gases and ozone, is given in Roeckner et al. [ 2003]. The additional aerosol module HAM includes the microphysical processes, the emission and deposition of aerosols, a sulfur chemistry scheme, and the radiative property scheme of the aerosols [Stier et al., 2005]. We used present-day aerosol emissions from the AEROCOM emission inventory representative for the year 2000, described in Dentener et al. [ 2006]. The 10Be production used was adapted from Masarik and Beer [1999] and interpolated to the correct latitude, altitude, and solar activity parameter Φ using the 10Be production function calculated by Masarik and Beer [1999]. The Φ values, with monthly resolution, were taken from Usoskin et al. [2005], available at: Removal of 10Be from the atmosphere follows closely the removal of sulfate aerosol (see Heikkilä et al. [2008a] for details).

[8] For this study, a middle-atmospheric model version with a horizontal resolution of T42 (2.8  ×  2.8°) with 39 vertical levels up to 0.01 hPa ( ∼ 80 km) was used. The run was allowed to spin up for five years to let 10Be reach equilibrium, and the years 1977–2006 were used for the analysis. We use monthly mean model output for this study. The run was forced with prescribed observational monthly mean sea surface temperatures and sea ice cover obtained from the international model intercomparison AMIP2 project [Gates, 1992]. The capability of the ECHAM-HAM model to reproduce the observed aerosol distribution in the atmosphere has been evaluated by Stier et al. [ 2005]. We refer to Hagemann et al. [2006] for a detailed discussion on ECHAM5-HAM's simulation of the hydrological cycle. Details on how 10Be and 7Be are introduced into ECHAM5-HAM as well as their validation with all observations found in literature are given in Heikkilä et al. [2008a] and in a more recent study including some new observations and the validation of the influence of model resolution on the results [Heikkilä and Smith, 2012]. Validation of the climate processes investigated within this study is presented in section 3. The data used for validation is the NCEP reanalysis [Kalnay et al., 1996]. All 10Be data analyzed in this study are results from this ECHAM5-HAM simulation.

[9] The climatic factors driving 10Be deposition are the atmospheric transport path of 10Be from source to archive and the removal of 10Be by wet or dry deposition (friction-based deposition occurring at surface) or sedimentation (gravitational settling occurring anywhere in the atmosphere) [Heikkilä et al. 2011]. Wet deposition is by far the most dominant form of deposition, exceeding 90% in global mean [Heikkilä et al. 2008a], being > 90% also in Greenland and on the Antarctic coast. In central Antarctica, the fraction reduces to < 40% based on observations [Raisbeck and Yiou, 1985]. The model predicts a fraction of < 50% which is slightly lower than observations. The atmospheric pathway is influenced by distribution of air mass and the wet deposition by precipitation rate. Various modes can be identified in atmospheric circulation variability, such as the North Atlantic Oscillation (NAO), the El Niño-Southern Oscillation (ENSO), and the Southern Annular Mode (SAM), all acting in different parts of the world. Because most of 10Be is produced in the stratosphere where residence time is long, the stratospheric air concentrations are higher by several factors than the tropospheric ones. The location and strength of the Stratosphere-Troposphere Exchange (STE) has the largest influence on 10Be air concentrations in the troposphere, especially in the midlatitudes [Field et al. 2006; Heikkilä et al. 2011]. In this study, we concentrate on the phenomena with a potential impact on Greenland and Antarctica and study correlations between climate indices and monthly mean 10Be deposition flux, precipitation rate, and other relevant variables. Details on the calculation of the indices is given in the corresponding subsections.

[10] We base this analysis on grid-point correlation between different variables to assess the temporal and spatial distribution of the connections between atmospheric drivers and 10Be deposition. Empirical orthogonal function (EOF) analysis is another useful method to investigate spatial relationships in temporal variability. It allows even for a quantification of the contribution of each component to total variability. The EOF analysis turned out less suitable for our analysis as there are many competing factors driving the variability (production rate and various climate-related processes), leading to meaninglessly flat EOF patterns.

3 Results

3.1 10Be Deposition Variability

[11] First the typical amplitude of variability modeled on monthly scale is presented (Figure 1), both precipitation rate (top of figure) and 10Be deposition flux (bottom). The 30-year mean (left) as well as the relative variability (standard deviation divided by the mean) are depicted. Similar patterns in both precipitation rate and 10Be deposition can be identified. Dry areas have limited deposition of 10Be. The wettest areas in the tropics experience less 10Be deposition because the 10Be production rate is low. The deposition peaks at midlatitudes due to the stratosphere-troposphere exchange injecting stratospheric air rich in 10Be into the troposphere, where it is efficiently scavenged at midlatitude storm tracks with their associated high precipitation rate. In Greenland, the 10Be deposition is higher than in Antarctica due to the generally higher precipitation rate. The relative variability is largest where precipitation rate is lowest. 10Be deposition varies less in dry areas than precipitation rate. The variability is lowest at midlatitudes with their constantly high precipitation rates. The variability is slightly larger over the continent of Greenland than in surrounding northern polar regions. In Antarctica, the variability is similar to Greenland, with slightly higher variability in the dry interior.

Figure 1.

ECHAM-modeled precipitation rate (top) and 10Be deposition flux (bottom) 1977–2006: mean (left) and relative variability expressed as standard deviation divided by the mean (std/mean), calculated on monthly basis (right).

3.2 Production Rate Versus Climate Variability

[12] Atmospheric processes which influence 10Be deposition typically have a time scale ranging from hours to days to seasonal. On the other hand, temporal resolution of long ice core data is typically in the order of several years rather than seasonal. In the following, we investigate how the factors influencing the 10Be variability are affected when filtered to different time scales. When used for reconstructing solar activity, the climate impact on 10Be deposition is assumed to be constant. In order to investigate this, we show correlations between the modeled 10Be deposition flux and the solar modulation function Φ which modulates the 10Be production rate (Figure 2, [Usoskin et al. 2005]). Both monthly mean and 25-month running mean correlations are shown. Note that a negative correlation is expected because cosmic ray intensity in the atmosphere and hence 10Be production rate are anticorrelated with solar activity. Solar activity exhibits three circa 11-year cycles during the period investigated. It is hence autocorrelated, which reduces the degrees of freedom. The same autocorrelation is expected from 10Be deposition flux as it reflects production variability. The correlation between precipitation, presenting climate influence, and 10Be deposition is also shown. Correlations are only shown where they are significantly different from zero between 95% confidence intervals. No adjustment of confidence intervals was made for the smoothing. On a monthly scale, the correlation with Φ is low, between 0 and − 0.2 in most areas, insignificant in the tropics where the solar modulation on production is low and precipitation rate high, and slightly higher (between − 0.2 and − 0.4) in the midlatitudes where STE is strongest. The 25-month running mean data presents correlations from − 0.6 to − 0.8 from 40° to 90° but somewhat lower ones from 40°S to 40°N (from − 0.2 to − 0.6). Also areas in Greenland and Antarctica exhibit lower, but still significant, correlation.

Figure 2.

Map of correlation coefficients between 10Be deposition flux and the solar modulation function Φ: monthly mean (top left) and 25-month running mean (top right). Map of correlation coefficients between 10Be deposition flux and precipitation rate: monthly mean (bottom left) and 25-month running mean 1977–2006 (bottom right). The seasonal cycle has not been removed from the monthly means. Only correlations significant at the 95% level are shown.

[13] Cross-correlation between precipitation rate and 10Be deposition on monthly and 25-month running mean scales shows an opposite pattern. The correlation is high ( > 0.8) in the tropics and areas in the subtropics, between 0.6 and 0.8 at high latitudes, and relatively low (0.3 to 0.4) over the Southern Ocean, Northwest Pacific, and Sahara. When filtered with 25 month running mean, which effectively removes the seasonal cycle, the correlation becomes weaker, remaining strong only over the tropical Pacific ocean. These findings suggest that on seasonal time scales climate variability, mainly precipitation, dominates the 10Be deposition variability. When the seasonal cycle is averaged out, the correlation of production rate is largest with deposition variability.

[14] 10Be surface air concentrations measured in air filters are also useful tracers when short time scales are investigated. We now assess whether the 10Be surface air concentrations reflect the production variability. Figure 3 illustrates the correlation map between 10Be surface air concentration and Φ. Similarly to 10Be deposition, the correlation is low on monthly time scale but very high ( < − 0.7) with 25-month running mean. Solely in the northern very high latitudes ( > 80°N) and at a spot in the tropical Pacific, the correlation drops to (0.3–0.4). The correlation with air concentration is likely to be higher than that with deposition because air concentrations are a result of model tracer advection which is not parameterized, unlike deposition. Surface air concentrations hence prove to be useful proxies for solar activity as well as deposition flux when the seasonal cycle has been averaged out.

Figure 3.

Map of correlation coefficients between 10Be surface air concentrations and Φ: monthly mean (top) and 25-month running mean 1977–2006 (bottom).

3.3 Factors Driving 10Be Deposition in Greenland

3.3.1 North Atlantic Oscillation—NAO

[15] NAO influence has been detected in precipitation and hence snow accumulation variability in Greenland with typically a positive (negative) correlation on the east (west) coast and no significant correlation in central Greenland [e.g., Appenzeller et al., 1998a; 1998b; Sodemann et al., 2008]. Therefore, we start by investigating if 10Be deposition follows the precipitation variability driven by the NAO. We calculate the station-based NAO index (normalized SLP difference between Ponta Delgada, Azores and Stykkisholmur, Iceland) from the ECHAM-modeled monthly mean sea level pressure (SLP) data. As the correlation with NAO is strongest during winter [Hurrell, 1995], the December-January-February-March (DJFM) means are used. In order to validate the model-based index, we repeat the calculations using the NCEP reanalysis.

[16] Figure 4 (top left) shows the grid-point correlation between precipitation and the calculated NAO index in the North Atlantic region. Correlation is shown only where it is significant above 95% level. The correlation is strongest over the North Atlantic. There is significant negative correlation over western Greenland and a smaller area with positive correlation in eastern Greenland, in agreement with Appenzeller et al. [1998a]. The correlation calculated using NCEP data is very similarly distributed but slightly reduced in comparison with ECHAM. Correlation with 10Be deposition has a very similar pattern with precipitation, although the correlation is weaker. Figure 4 also shows the histogram of the monthly mean NAO indices based on ECHAM and NCEP. Both distributions look similar. The indices compared here use standardized units so that the mean and standard deviation are forced equal. The similar form of the distribution can be used as validation of the ECHAM's ability to realistically reproduce the distribution of the index. We also performed an EOF analysis of the SLP and found the first EOF to be the typical NAO positive pattern with a high over the Azores and a low over Iceland in both ECHAM and NCEP data (not shown). The leading principal component (PC) correlated strongly (r = 0.76) with the station-based NAO index. These results suggest that the ECHAM model is able to capture the SLP and precipitation variability reasonably well. NAO seems to be the main driver of the 10Be deposition on interannual scale and changes in the 10Be deposition follow NAO strength.

Figure 4.

Map of correlation coefficients between the NAO index (DJFM mean 1977–2006) calculated from the ECHAM-modeled and NCEP data with the corresponding precipitation rate and 10Be deposition flux. Only correlations significant at the 95% level are shown. A histogram of the NAO indices calculated from monthly mean ECHAM and NCEP data is also shown.

[17] The similar spatial correlation between NAO and precipitation or 10Be deposition seems to confirm that precipitation dominates 10Be variability on short time scales. The direct correlation between precipitation and 10Be deposition simulated by ECHAM is shown in Figure 2. The correlation in Greenland on monthly scale is strong and significant. This seems to be supported by observations elsewhere: positive correlation between precipitation and 10Be deposition measured in precipitation [Heikkilä et al. 2008c] has been found on a monthly scale. Unfortunately, the number of long time series of 10Be observations in precipitation is very limited, and a better spacial coverage is required to confirm the picture.

3.4 Factors Driving 10Be Deposition in Antarctica

3.4.1 Southern Hemisphere Annular Mode—SAM

[18] The SAM is a zonally symmetric circulation pattern with anomalously low SLP over Antarctica and anomalously high SLP over the Southern Ocean during its positive phase. It has been found to be the leading mode of variability of nearly all atmospheric fields in the Southern Hemisphere [e.g., Rogers and van Loon, 1982] and the major contributor to stable isotope variability in Antarctica [e.g., Noone and Simmonds, 2002]. We calculate the SAM index as the difference in the temporally normalized monthly zonal mean SLP between 65°S and 40°S following Gong and Wang [1999] and Nan and Li [2003]. We analyzed both the de-seasoned (monthly anomalies after removing the seasonal cycle) and normal precipitation and 10Be deposition. In the case of SAM, correlations remained the same regardless of using the SAM index based on absolute SLP or seasonal anomalies, indicating that the anomalies are larger in scale than the seasonal cycle. We performed a consistency check using the EOF analysis of the SLP. The first EOF exhibited the typical SAM positive pattern with a low over Antarctica and a ring of high pressure over the Southern Ocean. The first PC correlates with the SAM index calculated in this analysis with r = 0.86.

[19] Figure 5 illustrates the correlation maps between the SAM index and precipitation for the ECHAM model and the NCEP reanalysis. The correlation is positive over the Southern Ocean between approximately 60°S and 45°S and negative at lower latitudes. Over Antarctica, the correlation is negative but relatively weak and in many places insignificant. This is consistent with the finding of van Ommen and Morgan [2010] that precipitation at the coastal Law Dome station was not mainly driven by SAM. Again, NCEP precipitation correlates weaker than the ECHAM-modeled one, but the distribution is similar. In general, the correlation is much weaker (approximately − 0.3 to 0.3) than in Greenland ( − 0.6 to 0.6). The 10Be deposition correlates weaker than precipitation with the SAM index over the Southern Ocean. In Antarctica, the correlation is stronger than with precipitation, indicating that another atmospheric field driving 10Be deposition also correlates with SAM. This is likely to be tropopause height, which we return to in section 3.4.4. As the indices are calculated from normalized fields, both ECHAM- and NCEP-based indices have the same mean and standard deviation. However, the slightly negative skewness exhibited by the NCEP index is reproduced by ECHAM and indicates that the simulation of the SLP variability over the 30 year period has been successful.

Figure 5.

Map of correlation coefficients between the SAM index (monthly mean 1977–2006) calculated from the ECHAM-modeled and NCEP data with the corresponding precipitation rate and 10Be deposition flux. Only correlations significant at the 95% level are shown. A histogram of the SAM indices calculated from monthly mean ECHAM and NCEP data is also shown.

3.4.2 El Niño—Southern Oscillation—ENSO

[20] The Southern Oscillation index (SOI) is defined as standardized difference between SLP at Tahiti and Darwin. El Niño and its atmospheric component Southern Oscillation (ENSO) are responsible for distributing air masses mainly in the tropics, but a teleconnection between moisture fluxes and precipitation in the Southern Ocean with ENSO has been found [e.g., Yuan and Yonekura, 2011, and references therein]. A connection between precipitation and Southern Oscillation has also been found in Antarctica by a modeling study by Noone and Simmonds [2002]. Therefore, an impact of Southern Oscillation could possibly be detected in the 10Be deposition. Figure 6 presents the grid-point correlations with precipitation and 10Be deposition. No connection is evident from the figure. To investigate whether a common seasonal cycle is influencing the correlation, we repeat the analysis with seasonal cycle removed. De-seasoning precipitation and 10Be deposition revealed no correlation with SOI. Our results do not suggest a large contribution by ENSO in the precipitation or 10Be deposition variability at high latitudes on monthly time scales.

Figure 6.

Map of correlation coefficients between the SO index (monthly mean 1977–2006) calculated from the ECHAM-modeled and NCEP data with the corresponding precipitation rate and 10Be deposition flux. Only correlations significant at the 95% level are shown. A histogram of the SO indices calculated from monthly mean ECHAM and NCEP data is also shown.

3.4.3 Zonal Wave 3 Pattern—ZW3

[21] The ZW3 together with the zonal wave 1 pattern form the asymmetric part of the extratropical circulation in the Southern Hemisphere. The ZW3 is associated with blocking, and its positive phases allow for meridional transport of moisture towards the Antarctic coast [Raphael, 2004]. Its signature has also been found in the sea ice variability around Antarctica [Raphael, 2007]. van Ommen and Morgan [2010] introduced a modified ZW3 index for the Australian sector only and found a seesaw pattern between precipitation at the Law Dome station, Antarctica and on the Australian west coast. Hence, ZW3 has the potential to influence the 10Be deposition variability as well.

[22] The ZW3 index in this study is defined following Raphael [2004]. The index is calculated as a mean of three locations (see [Raphael, 2004]) marked with crosses in Figure 7. It can be calculated from SLP variability, geopotential height variability at 500 mbar or geopotential zonal anomalies at 500 mbar. We tested all three possibilities and found partly varying results. The index calculated from SLP and geopotential anomalies gave the most similar results. Following Raphael [2004], we concentrate our analysis on the index based on geopotential anomaly.

Figure 7.

Map of correlation oefficients between the ZW3 index (monthly mean 1977–2006) calculated from the ECHAM-modeled and NCEP data with the corresponding precipitation rate and 10Be deposition flux. Only correlations significant at the 95% level are shown. A histogram of the ZW3 indices calculated from monthly mean ECHAM and NCEP data is also shown.

[23] The correlation between ECHAM and NCEP precipitation with the ZW3 index as well as the 10Be deposition can be seen in Figure 7. The correlation with precipitation follows the three-point pattern typical for ZW3 but is relatively low. In Antarctica, the correlation is significant in some regions but weak. The NCEP precipitation has areas of higher positive correlation than the ECHAM-modeled one, but it is mostly not significant in Antarctica. De-seasoning precipitation or 10Be deposition changed the correlations little. We also tested picking only years with positive ZW3 but found no improvement. Similarly with the SAM index, the (negative) correlation with 10Be deposition is more significant in Antarctica than with precipitation. Together, the SAM, ENSO, and ZW3 results indicate that in Antarctica the connection between precipitation rate and 10Be deposition variability is weaker than in Greenland. The contribution of dry to total 10Be deposition in the interior of Antarctica can be up to 60%, as estimated from observations [Raisbeck and Yiou, 1985] and predicted by the model [Heikkilä et al. 2011] reducing the correlation. On the coast, however, the simulation shows that the situation is different, and the share of wet deposition still exceeds 80–90% [Heikkilä et al. 2011].

3.4.4 Tropopause Height

[24] The relatively weak link between precipitation and/or 10Be deposition with SAM, SOI, or ZW3 discussed in the previous sections suggests that, unlike in Greenland, 10Be deposition in Antarctica is driven by another process on seasonal scale. In a recent study, Pedro et al. [2011b] found that the seasonal (late summer) maxima observed in 10Be snow concentrations at Law Dome coincided with minimum tropopause height. This minimum was simultaneous with modeled maximum fraction of stratospheric 10Be in deposition. It is confirmed by observed 10Be/ 7Be ratios in air filters at Neumayer, coastal Antarctica, indicating maximum fraction of stratospheric air at the surface in late summer [Elsässer et al. 2011]. The seasonal cycle of tropopause height south of 60°S is opposite to the midlatitudes or northern polar regions. The amplitude of the seasonal cycle is also the largest south of 60°S [Son et al., 2011]. Simultaneous seasonality between stratospheric intrusions and 10Be concentrations has not been found outside Antarctica [Graham et al., 2003; Heikkilä et al., 2008c; Zanis et al., 2003].

[25] We investigate the correlation between tropopause pressure and 10Be deposition. Note that high pressure means low tropopause in terms of altitude. Tropopause is defined in ECHAM following the temperature lapse rate definition by WMO [1957]. Definition of the thermal tropopause can be problematic in Antarctica in winter as the temperature profile is flat. However, Zängl and Hoinka [2001] studied the thermal and dynamical tropopauses and found a reasonably good correlation between the “ECMWF-reanalysis (ERA)” and radiosonde-derived tropopause pressure at high southern latitudes, showing that thermal tropopause captures the tropopause variability reasonably well. The seasonal cycle and the amplitude of the ECHAM tropopause is similar to the NCEP tropopause and agreed generally with observed tropopause pressure by Son et al. [2011].

[26] Figure 8 illustrates the correlation between monthly mean tropopause pressure and 10Be deposition in Antarctica. The correlation is stronger and more significant than correlations found between 10Be deposition and SAM, SOI, or ZW3 in most areas of Antarctica. When tropopause height decreases, 10Be deposition increases. In summer when the tropopause is low, stratospheric air with higher 10Be concentrations may become mixed with lower level air, leading to a higher 10Be content in precipitation [Pedro et al. 2011b]. Our primary assumption that this phenomenon is found in Antarctica only can be tested by looking at the global correlation (Figure 8, bottom panel, correlation with 10Be deposition (left) and precipitation (right)). 10Be deposition typically increases with decreasing tropopause height (positive correlation) in dry areas in Antarctica, Australia, oceans to the west of Australia, Africa, and the Americas as well as in Sahara and the middle East. 10Be deposition increases with increasing tropopause height (negative correlation) roughly in the wetter areas, the tropics, over the Southern Ocean and north of 40°N–50°N. The negative correlation can be explained by maximum tropopause height during summer, with higher temperatures and an increased convective activity leading to enhanced precipitation. A correlation map between tropopause pressure and precipitation is also shown in Figure 8. The distribution of correlations by 10Be deposition and precipitation is very similar, except in Antarctica showing the decoupling of 10Be deposition and precipitation there. Correlation with tropopause height and 10Be/ 7Be in deposition (not shown) which cancels out the precipitation variability is positive everywhere except over the Southern Ocean (40°S to 60°S) and in Northern Hemisphere tropics (0°N to 20°N).

Figure 8.

Map of correlation coefficients (monthly mean 1977–2006) between tropopause pressure and 10Be deposition flux in Antarctica (top left) and globally (bottom left), and between tropopause pressure and precipitation (bottom right). Zonal mean correlation coefficient between 10Be concentrations in air at all levels and tropopause pressure (top right).

[27] The top right subplot of Figure 8 shows the zonal mean correlation between 10Be air concentration and tropopause pressure. The black dashed line illustrates the mean tropopause pressure. Correlations are calculated for each grid box at each level and zonally averaged. It is obvious that 10Be concentrations are significantly correlated with tropopause height in the whole atmosphere. The strong positive correlation at the tropopause indicates that 10Be concentrations follow closely changes in tropopause height. The correlation in the stratosphere seems to reflect the structure of the Brewer-Dobson circulation. In the Northern Hemisphere, the decrease in tropopause height occurs during winter, the season of intensified Brewer-Dobson circulation. Tropical air with low 10Be content is lifted up and pushed towards the stratospheric midlatitudes and high latitudes reducing the 10Be concentration. This seasonal variability is in agreement with estimated transport times into the extratropical stratosphere [e.g., Birner and Bönisch, 2011]. The intensified poleward transport is reflected by the area of negative correlation from approximately 30°N to 90°N. A similar effect can be seen in the Southern Hemisphere stratosphere from 30°S to 60°S. The opposite seasonality of the high-latitude tropopause (60°S–90°S) with maximum height in winter causes the correlation to change sign. Correlation with air concentrations in the troposphere agrees with 10Be deposition flux (bottom left panel) at high northern latitudes. In the tropics the correlation is opposite. High rain events might effectively wash out the 10Be content from air below cloud levels. At southern high latitudes, the correlation is also opposite to deposition flux. In summer, the low tropopause coincides with high precipitation, which might lead to a similar washout effect.

[28] Correlation between two parameters can be caused by a mechanism influencing both parameters the same way without the parameters actually being interconnected. We investigate whether the seasonal cycle in tropopause height influences SAM, SOI, or ZW3 similarly as 10Be deposition. Figure 9 illustrates the monthly mean correlation coefficients. The correlation is significant in their respective influential areas: NAO in the North Atlantic and SOI in the tropics. SAM and ZW3 correlate only weakly with a maximum in Antarctica. Hence, a mutual correlation between climate modes and tropopause height does not seem to explain the similar variability between tropopause height and 10Be deposition.

Figure 9.

Map of correlation coefficients between tropopause height and NAO, SAM, SOI, and ZW3 indices (monthly mean 1977–2006).

[29] To further investigate whether the connection found between 10Be deposition and tropopause height is caused by distribution of air masses, specifically stratosphere-troposphere exchange, we use 300 mbar geopotential height anomalies which reflect cyclonic and blocking-like activity including cutoff lows and tropopause folds, associated with STE [e.g., Land and Feichter, 2003]. A positive correlation is found between 10Be deposition flux anomaly with 300 mbar geopotential height anomalies in Antarctica (not shown). Tropopause height anomalies correlate uniformly with the 300 mbar geopotential anomalies (r < − 0.5) except in the tropics (20°S to 20°N) and southward of 60°S (not shown). However, the correlation with 10Be air concentrations is not significant. Put together, a certain correlation between 10Be variability and tropopause height is found on seasonal scale. This correlation is stronger in Antarctica than the correlation with SOI, SAM, or ZW3. It cannot be completely ruled out that the correlation is caused by a mechanism affecting both parameters in the same way.

[30] A possible reason for the generally lower correlations in Antarctica than in Greenland are the different time scales of the analysis. NAO has the strongest impact on precipitation in winter, and typically the DJFM means are used (see section 3.3). This practically represents an annual resolution. Much of the correlation on monthly scale in Antarctica is caused by the similar seasonal cycle of parameters. If the seasonal cycle is removed, the correlation is reduced to − 0.2 to 0.2 and becomes insignificant in many areas. On the other hand, the correlation persists on longer time scales. Figure 10 presents the correlations between 25-month running mean 10Be deposition flux and SAM, SOI, ZW3, and tropopause pressure. This filtering removes the seasonal cycle. The maps in Figure 10 show that correlation with tropopause height over Antarctica is similarly distributed as the monthly mean one showing that it is not only caused by the common seasonal cycle. The correlation with SAM over the Southern Ocean has changed from strong positive to insignificant but has become stronger negative in most of Antarctica. The correlation with SOI has become significant in larger areas in Antarctica but remains patchy. Similarly to SAM, the negative correlation with ZW3 has become stronger.

Figure 10.

Map of correlation coefficients between 10Be deposition flux and SAM, SOI, ZW3, and tropopause pressure in Antarctica, 25-month running mean (1977–2006).

[31] According to these results 10Be, as well as climate variability in Antarctica, is an interplay of several processes acting at different frequencies, and their mutual contribution changes in time and space. However, these connections can be established during different climate stages, and the spatial and temporal dependencies of 10Be snow concentrations can be investigated.

3.5 The Importance of Time Scales

[32] Time scales have proven to be of essential importance when 10Be data are decomposed into production and climate components. Abreu et al. [2012] found that production rate explains 76% of the observed 10Be variability on multidecadal to multimillennial time scales. In the following, we assess the influence of filtering with different frequencies on the change of correlation between various climate indices and Φ with 10Be deposition. The results are shown at neighboring grid points to selected drill sites in Greenland (Figure 11) and in Antarctica (Figure 12). The filter widths chosen are 3 months, 13 months (roughly 1 year), 25 months (2 years), and 121 months (10 years). For Greenland, correlations are shown between 10Be deposition flux, NAO, precipitation, tropopause pressure, and Φ. The drill sites selected are North GRIP (NGRIP), Milcent, and Dye3, which run from north to south and are located near the center of the continent (see Figure 4). As previously seen, the correlation with precipitation is high on short time scales and decreases on annual scales. However, on decadal time scales, the correlation increases again at NGRIP and Milcent. On such time scales, long-term trends become relevant. Similar behavior can be seen with NAO and tropopause pressure. There are spatial differences between sites. Correlation with production rate constantly increases the more the data are smoothed up to the 10-year limit. Finally, this is constrained by the length of our data (30 years). In case of the 10-year smoothing, the degrees of freedom and thus the statistical significance of the correlation are reduced. However, keeping these limitations in mind, we investigate whether the solar signal is preserved in the data in the presence of a possible long-term trend of a climatic variable. These results agree with the findings by Pedro et al. [2012], who found similar correlation between 10Be snow concentrations and solar activity (neutron counts) and precipitation rate in Greenland and a slightly lower correlation with NAO using annual data. Their correlation with NAO at the more coastal Das2 site was higher than at the sites shown here, in agreement with Figure 4.

Figure 11.

Correlation between 10Be deposition flux and climate parameters and Φ on different time scales using running mean filtering (filter width given on x axis) at grid points closest to selected ice core sites in Greenland (1977–2006).

Figure 12.

Correlation between 10Be deposition flux and climate parameters and Φ on different time scales using running mean filtering (filter width given on x axis) at grid points closest to selected ice core sites in Antarctica (1977–2006).

[33] We investigate all climate indices previously discussed (SAM, SOI, ZW3 as well as precipitation, tropopause height, and Φ) at the Antarctic stations. We chose the Law Dome, South Pole, and Epica Dronning Maud Land (EDML) sites (see location in Figure 8). Some differences between the sites can be seen. Precipitation (positive) and SAM (negative) have a strong impact on the variability on all time scales. At the South Pole, the impact is weaker because a large share of 10Be is deposited dry. SOI does not seem to influence the 10Be variability largely except for decadal time scales where it starts to correlate strongly with SAM. Both SOI and SAM exhibit a slight increasing trend over this 30 year period. According to this analysis, ZW3 is not an important factor on 10Be deposition on any scales. However, Pedro et al. [2012] found relatively low but significant correlation with 10Be in Law Dome snow and ZW3 on annual scale. Their correlation with SAM or SOI was not significant. Tropopause pressure correlates strongly on short time scales (seasonal), weaker on scales of a few years but stronger again on decadal scales. At Law Dome, the correlation is weak on short scales but constantly increases up to 10 years.

[34] Correlationwith Φ in Antarctica is similar to Greenland. It is weak on short time scales but gradually grows when the filter width increases up to 10 years. At EDML, the change in sign of the correlation happens already at the 10-year scale.This is due to a decreasing trend in precipitation, suppressing the increase in 10Be deposition due to the drop in solar activity starting in the early 1990s.Such a trend is not seen at other sites. It is known from 10Be ice core records that the EMDL site is responsive to solar activity on longer time scales [Steinhilber et al., 2012]. This example thus demonstrates how large enough changes in climatic variables are capable of distorting the production signal in the record, depending on their amplitude and the relative duration with respect to the length of the record.

4 Summary and Conclusions

[35] We have investigated the production rate and climate drivers of temporal and spatial variability in 10Be deposition globally and with focus on Greenland and Antarctica. The data used is a 30-year climatological simulation of 10Be using the ECHAM5-HAM aerosol climate model. The aim is to identify the climatic factors driving 10Be deposition and how their changes will affect the deposition response to production. An understanding of the connection between climate and 10Be deposition is required to correct for any climate change signals in the 10Be ice core records. Such signals are observed during the glacials. How easily this can be done depends on the number of main drivers and their linearity.

[36] We find strong correlation between 10Be deposition and the North Atlantic oscillation (NAO) in Greenland on monthly and winter-mean time scales. The correlation is positive on the east coast and negative on the west coast, following the patterns found by previous studies investigating NAO and precipitation or δ18O. The correlation is weaker but still significant in central Greenland where most of the ice core drilling takes place.

[37] In Antarctica, the situation is more complicated with several climate modes influencing atmospheric variability over Southern Ocean but none of them fully explaining precipitation or 10Be deposition. We find weak negative correlation with Southern annular mode (SAM), Southern Oscillation Index (SOI), and Zonal Wave 3 (ZW3) with 10Be deposition on seasonal scale, the correlation being stronger with 10Be than with precipitation. This suggests that the connection between precipitation and 10Be deposition is weaker than in Greenland. Rather, air exchange with upper atmospheric levels with higher 10Be content is more important. This again is influenced by tropopause height. We find highest correlations of 10Be deposition with tropopause height on seasonal as well as multiyear time scales. The response is strongest in East Antarctica where the tropopause is the closest to surface due to high topography.

[38] Time scales are the key in correctly interpreting 10Be records as solar activity ( Φ) proxy. On seasonal time scales, the variability is dominated by various climatic factors, mainly precipitation except in Antarctica. When the seasonal cycle is filtered out, the production rate of 10Be, or solar activity, becomes the dominant factor. 10Be surface air concentrations were found to behave similarly, with weak correlation with Φ on sub-annual time scales but a strong one on time scales of a year or more.

[39] When investigating the temporal evolution of correlation between climate and 10Be production rate at grid points closest to selected ice core drilling sites, we found that precipitation is an important factor on short (monthly) and long (decadal) time scales. On time scales from 1 year to 10 years, production increasingly dominates. Given the length of the simulation of this study (30 years), we were only able to assess production impact of the 11-year solar cycle. Longer-term trends in solar activity could not have been detected. Climate modes (NAO in Greenland and SAM, SOI, and tropopause height in Antarctica) seemed to influence 10Be deposition variability strongly on short time scales and on long ones, but the production influence dominated in between. However, some regions deviated from this general behavior.

[40] As the typical temporal resolution of ice cores is in the order of years or more, 10Be records are well suited as solar activity proxies. When signals of climate change are needed to be removed from the records, long-term changes have to be identified in the climate modes (NAO in the case of Greenland and tropopause height and SAM in the case of Antarctica). Shifts in spatial patterns have to be identified as they are likely to influence the affected areas. This study has established an interconnection between climate and production-related variability of 10Be and the importance of its temporal and spacial scales. The connections found here during present climatic conditions serve as a base for future investigations of how their contributions to 10Be variability might change during glacial conditions.


[41] The authors would like to thank Dr. Greg Doherty for helping to install, test, and run the ECHAM5-HAM model. The authors acknowledge the valuable discussions with Dr. Andrew Klekociuk, which helped to improve this manuscript.