The ice-core record of the carbon-13 content of atmospheric methane (δ13CH4) has largely been used to constrain past changes in methane sources. The aim of this paper is to explore, for the first time, the contribution that changes in the strength of a minor methane sink―oxidation by atomic chlorine in the marine boundary layer (ClMBL)―could make to changes in δ13CH4 on glacial-interglacial timescales. Combining wind and temperature data from a variety of general circulation models with a simple formulation for the concentration of ClMBL, we find that changes in the strength of this sink, driven solely by changes in the atmospheric circulation, could have been responsible for changes in δ13CH4 of the order of 10% of the glacial-interglacial difference observed. We thus highlight the need to quantify past changes in the strength of this sink, including those relating to changes in the sea-ice source of sea salt aerosol.
Methane (CH4) is an important atmospheric constituent on account of its potency as a greenhouse gas and its strong influence on the tropospheric oxidizing capacity. We know from the polar-ice record that between the last glacial maximum (LGM; 21 kyr before present (BP)) and the pre-industrial Holocene (PIH; 1 kyr BP) its concentration, [CH4], rose from around 360 ppbv to about 700 ppbv [e.g., Loulergue et al., 2008], but how much of this change was source-driven, and how much was sink-driven, remains uncertain [see, e.g., Valdes et al., 2005; Kaplan et al., 2006; Fischer et al., 2008]. This study focuses on the 12/13C-isotopic composition of CH4 trapped in polar ice, δ13CH4, which provides a complementary constraint on the CH4 budget and past changes therein [see, e.g., Ferretti et al., 2005; Fischer et al., 2008]. In the nomenclature of Schaefer and Whiticar , δ13CH4 can be expressed as the sum of the average isotopic composition of CH4 sources, δ13CE, and the average influence, by way of isotopic fractionation, of CH4 sinks, ɛWT (equation (1)). δ13CE can be broken down into the strength of each source, Ei, and its isotopic composition, (equation (2)). Similarly, ɛWT can be broken down into the fraction of CH4 removed by each sink, Fj, and the fractionation coefficient associated with it, αj (equation (3)). Here, we explore the influence that CH4-oxidation by atomic chlorine in the marine boundary layer (ClMBL) has on δ13CH4 (encapsulated by the term, (αCl-1).FCl, in equation (3)), specifically the contribution that changes in the strength of this sink could make to glacial-interglacial changes in δ13CH4.
The main source of ClMBL is sea salt aerosol (SSA), which is produced by the action of the wind on wave crests, and from which BrCl and Cl2 (that photolyse to give ClMBL) are liberated [see, e.g., Vogt et al., 1996; Platt et al., 2004]. The abundance of SSA in pseudo steady-state is determined by the rate of SSA production and the rate of SSA removal (e.g., by wet and dry deposition). The abundance of ClMBL derived from SSA then depends on the abundance of SSA, the acidity of the atmosphere, as BrCl and Cl2 are believed to be liberated from SSA that has been acidified by the products of dimethyl sulfide (DMS) oxidation [Vogt et al., 1996; Platt et al., 2004], and the intensity of radiation of the wavelengths required to photolyse BrCl and Cl2. However, as a first step in exploring the potential for ClMBL to contribute to glacial-interglacial changes in δ13CH4, we explore the influence of the circulation alone, on the grounds that the production of SSA is highly sensitive to the wind speed [see, e.g., Monahan et al., 1986; Andreas, 1998; Witek et al., 2007], and several lines of paleodata, including the ice-core records of dust and sea salt [e.g., Thompson and Mosley-Thompson, 1981; Petit et al., 1981; Hansson, 1994; Röthlisberger et al., 2002] (and the recent review by Fischer et al. ), could indicate changes in this at the LGM.
We do so via a number of simple calculations employing a variety of model simulations of the PIH and LGM circulations. It is important we explore a variety of simulations, as there has been little consensus regarding the changes at the LGM, particularly in the region of the southern hemisphere westerlies (SHW), with estimates ranging from a 40% reduction in surface wind speeds [Kim et al., 2003] to a 25% increase in surface wind stress (implying a 12% increase in wind speeds) [Shin et al., 2003] in this region. Much of the literature has focused on changes in the strength (and position) of the SHW, owing to the bearing these could have on glacial-interglacial changes in CO2 [see, e.g., Toggweiler, 1999; Toggweiler and Russell, 2008]. As the SHW exhibit some of the highest surface wind speeds globally and cover a broad swath of the Southern Ocean (see Figure 1), changes in their strength could also have bearing on the global strength of the ClMBL sink, and hence δ13CH4. By employing a variety of simulations, we probe the range of influences circulation-driven changes in the strength of this sink could have had on δ13CH4.
We start by assuming that ClMBL removed the same fraction of CH4 in the PIH as it is estimated to remove in the present, and hence was responsible for an equal enrichment in δ13CH4 (relative to δ13CE), namely 2.6(±1.2)‰ [Allan et al., 2007]. We thus equate (αCl-1).FCl, integrated seasonally and globally in the PIH, to 2.6‰; see equation (3) and accompanying text. FCl is estimated to be between 0.03 and 0.04 [Platt et al., 2004; Allan et al., 2007, 2010], hence FCl is small compared to 1-FCl (the fraction of CH4 removed by all other sinks) and a modest change in FCl, of up to say ±50%, will have little effect on FOH, Fsoil etc. It follows that, to a good degree of approximation, an X% increase (decrease) in (αCl-1).FCl will be accompanied by a 0.026X‰ enrichment (depletion) in δ13CH4. Assuming the rate of CH4 removal by each CH4 sink is first order with respect to [CH4], and again as FCl is small compared to 1-FCl, we assume FCl is proportional to the product of [ClMBL] and the rate coefficient for the reaction between ClMBL and CH4, kCl. Accordingly, an X% increase (decrease) in (αCl-1).kCl[ClMBL] will be accompanied by a 0.026X‰ enrichment (depletion) in δ13CH4. We can calculate αCl according to equation (4) [Saueressig et al., 1995], and kCl (molecules−1 cm3 s−1) according to equation (5) [Sander et al., 2003], where T is the temperature (K).
To calculate [ClMBL] (molecules cm−3), we use a modified version (equation (7)) of the simple formulation with which Allan et al.  explored the role of ClMBL in spatial and inter-annual variations in δ13CH4 (equation (6)). Equation (6) expresses [ClMBL] in terms of an average concentration of 18 × 103 molecules cm−3 and a seasonal variation governed by the time of year, t (day number). The tanh(3λ) term, where λ is the latitude (radians), simply ensures the seasonal cycles in the northern and southern hemispheres are six months out of phase. To explore the influence that the wind has on [ClMBL], we add a factor of N.VP, where V is the horizontal wind speed (ms−1), P is the power to which this is raised and N is a normalization factor, of which our results are independent as we are only interested in percentage changes in [ClMBL]. We do so on the basis that Gong et al.  suggest the column loading of SSA is proportional to VP (with P = 1.39 in the North Atlantic, 1.46 in the tropical Pacific, and 1.66 in the South Pacific). Assuming (1) the column loading of ClMBL is proportional to that of SSA, (2) the ClMBL is concentrated in the marine boundary layer (MBL), and (3) the height of the MBL does not change, [ClMBL] should also be proportional to VP. Though Gong et al.  did not comment on it, their plots of loading versus wind speed [Gong et al., 2002, Figure 2] indicate a proportionality to V3.41, or similar, at wind speeds above about 5 ms−1. We therefore employ (globally) P = 1.39 at V ≤ 5 ms−1 and 3.41 at V > 5 ms−1, but also explore the sensitivity of our results to variations in this formulation (see later).
With equations (4), (5), and (7), we can calculate (αCl-1).kCl[ClMBL] as a function of season and location during the PIH and LGM, provided we have the necessary wind and temperature data. Summarized in Table 1, these are taken from simulations with five general circulation models, more information on which can be found in the auxiliary material. Figure 1 illustrates the annual-mean surface wind speeds and temperatures in the PIH, and the changes in these at the LGM, according to each model. It also illustrates the percentage changes we calculate in [ClMBL] at the LGM based on the wind data. Subject to the data from each model, we calculate the seasonally and globally integrated value of (αCl-1).kCl[ClMBL] throughout the MBL (treating areas of sea ice and open ocean alike; discussed later) in both the PIH and LGM, and hence the percentage change in this quantity on switching from PIH to LGM winds and temperatures, which we relate to a per-mil change in δ13CH4.
Climatological monthly means based on 100-year integrations (& 100 years of monthly means)
3.75°lon × 2.5°lat; 19 levels
10 m winds; 1.5 m temperatures
Climatological monthly means based on 100-year integrations
3.75°lon × 2.5°lat; 19 levels
10 m winds; 997 mb temperatures
Climatological monthly means based on 20-year integrations (& 20 years of daily means)
We use mainly climatological monthly-mean data (based on 100-year integrations, or 20-year integrations in the case of HadAM3), these being arguably the most robust. However, we also repeat our calculations with CCSM3 and HadAM3 data, employing a full 100 years of monthly-mean data and a full 20 years of daily-mean data, respectively, to explore the sensitivity of our results to the degree of temporal averaging; see Table 1. To assess the sensitivity of our results to our formulation for P, we repeat all of these ‘base’ calculations (B) subject to an alternative value of P at V ≤ 5 ms−1 (1.66; S1) and alternative values of V at which we switch from P = 1.39 to P = 3.41 (4 ms−1 in S2 and 6 ms−1 in S3). We also assess the sensitivity of our results to: the seasonality of [ClMBL], by repeating the base calculations with the tanh(3λ)sin[2π(t − 90)/365] term in equation (7) set to zero (S4); and the changes in temperature between the PIH and LGM, by changing the winds whilst keeping the temperatures (PIH) constant (S5).
The results of the base (B) and sensitivity (S1-5) calculations are given in Table 2; the numbers in parentheses correspond to the results obtained when less ‘temporally averaged’ data are employed (see Table 1 and accompanying text). We find that the effect on δ13CH4 of switching from PIH to LGM winds and temperatures depends on which model data we use and the degree to which these are temporally averaged, with the base calculations yielding everything from a depletion of 0.46‰ to an enrichment of 0.14‰.
Changes in δ13CH4 (‰) calculated in the base (B) and sensitivity (S1-5) calculations; the numbers in parentheses correspond to the results obtained when less ‘temporally averaged’ data are employed (see Table 1 and accompanying text for details).
The S1 calculations show that our base results are insensitive to the value of P employed at V ≤ 5 ms−1; we get the same results regardless of whether we employ the lowest value (1.39) or the highest value (1.66) Gong et al.  reported based on calculations in the North Pacific and South Pacific respectively. Furthermore, the S2 and S3 calculations show that our results are reasonably robust to changes in the value of V at which we switch from P = 1.39 to P = 3.41, changing by less than or similar to 10% upon increasing or decreasing this by 1 ms−1.
The effect of removing the [ClMBL] seasonality in the S4 calculations is variable, depending on the model data used and the degree to which these are temporally averaged. Mostly, it has a modest effect (of the order of 10%), however it has a more pronounced effect in the calculations with IPSL-CM4 and HadAM3 climatological monthly-mean data. The change in δ13CH4 we calculate could therefore be sensitive to the assumed [ClMBL] seasonality; we have employed the same [ClMBL] seasonality as Allan et al. , reflecting that of the radiation required to photolyse BrCl and Cl2; see equation (7) and accompanying text.
Finally, based on the S5 calculations, it would appear that the changes in temperature between the PIH and LGM are responsible for a depletion in δ13CH4 of approximately 0.05–0.1‰, depending on the model data employed. The depletion reflects a reduction in the rate of reaction between ClMBL and CH4 due to the reduction in temperatures at the LGM (see Figure 1 and equation (5)), only marginally offset by an increase in the fractionation coefficient associated with this reaction (see equation (4)).
Our calculations suggest circulation-driven changes in the strength of the ClMBL sink could have a small but significant effect on δ13CH4 on glacial-interglacial timescales. Depending on the model data employed, and the degree to which these are temporally averaged, we calculate changes in δ13CH4 ranging from a depletion of 0.46‰ to an enrichment of 0.14‰, the magnitudes of which are of the order of 10% of the 3.5‰ glacial-interglacial difference observed [Fischer et al., 2008]. Factors not explored here, which could have also affected [ClMBL] and hence δ13CH4 on these timescales, include: changes in the lifetime of SSA (e.g., due to changes in precipitation); changes in the acidity of the atmosphere (e.g., due to changes in DMS production linked to changes in biology, such as plankton type and/or abundance); and changes in the intensity of radiation required to photolyse BrCl and Cl2 (e.g., due to changes in stratospheric ozone). δ13CH4 could have also been affected by changes in FCl (and Fsoil) accompanying changes in the amount of CH4 removed by OH, also not explored here. If initially FOH = 0.9, Fsoil = 0.06 and FCl = 0.04, and αOH = 1.0039‰, αsoil = 1.02‰ and αCl = 1.06‰, a 10% increase (decrease) in the amount of CH4 removed by OH would lead to a 0.3‰ depletion (enrichment) in δ13CH4.
It is interesting that all of our calculations based on climatological monthly-mean data—arguably the most robust—suggest that the circulation-driven changes in the ClMBL sink would have led to a depletion in δ13CH4 at the LGM relative to the PIH, primarily due to a reduction in the global abundance of ClMBL. Ice-core records show an increase in sea salt at the LGM, by a factor of 15 in the Arctic and 3 in the Antarctic [see Fischer et al., 2007, and references contained therein], which we would expect to have been accompanied by proportional increases in [ClMBL]. Of course, there could have been more ClMBL in polar regions but less at lower latitudes, yielding an overall reduction. However, our calculations yield percentage increases in [ClMBL] in some regions of the Arctic Ocean approaching, but still short of, the 15-fold increase we would expect, and generally capture less of the 3-fold increase expected in the Southern Ocean; see Figure 1. The calculations based on CCSM3 and HadAM3 data yield increases limited to the regions south of about 50°S and 60°S, respectively, accompanied by decreases to the north of these, whilst the remainder of the calculations predominantly show decreases in the Southern Ocean. This raises the question, what SSA source are we missing or underestimating in our calculations, and what influence does it have on δ13CH4?
One possibility is that the simulations of the LGM circulation simply underestimate the wind speeds at high latitudes. If this were the case, it could call into question the validity of these simulations in other regions too. It certainly seems likely that at least part of the glacial-interglacial difference in sea salt (and dust) was the result of changes in wind speeds governing the strength of sea-salt sources, changes in wind patterns determining the efficiency of transport to Arctic and Antarctic ice-core sites and/or changes in precipitation affecting its atmospheric lifetime [see, e.g., Fischer et al., 2007; Petit and Delmonte, 2009]. However, there is some evidence that sea ice, as opposed to open ocean, is the dominant source of SSA reaching both coastal and continental Antarctic sites [e.g., Wagenbach et al., 1998; Rankin et al., 2002; Wolff et al., 2003, 2006]. In our calculations, we have assumed that sea ice is an equally strong source, showing the same dependence on wind speed. If however, sea ice were a stronger source on a per-unit-area basis, the increase in sea-ice at the LGM could have contributed to the 3-fold increase in sea salt seen in the Antarctic, and perhaps the 15-fold increase seen in the Arctic. A sea-ice driven increase in SSA, and hence ClMBL, at high latitudes would tend to strengthen the ClMBL sink, and hence enrich δ13CH4 at the LGM. However, without knowing quantitatively how the strengths of the sea-ice and open-ocean sources compare, we cannot say what the net effect on δ13CH4 would be if the increase in sea-ice were factored into our calculations.
What we can say is, irrespective of whether the net effect amounts to an enrichment or a depletion in δ13CH4, a change in δ13CH4 due to a change in the strength of the ClMBL sink would have implications for our interpretation of the glacial-interglacial δ13CH4 record, and we have shown that δ13CH4 is affected non-negligibly by circulation-driven changes alone. Fischer et al.  attributed the enrichment in δ13CH4 at the LGM to a near-complete shutdown of boreal wetland sources of relatively 13C-poor CH4, whilst biomass-burning sources of relatively 13C-rich CH4 were little or unchanged relative to the pre-boreal Holocene (10 kyr BP). A global synthesis of charcoal records by Power et al. , however, has since shown that the last glacial period (16–21 kyr BP) was the period of least biomass burning in the last 21 kyr, suggesting we still have some enrichment in δ13CH4 at the LGM to explain. An enrichment due to a strengthening of the ClMBL sink could potentially contribute to this, whilst a depletion due to a weakening of the ClMBL sink would further suggest the explanation offered by Fischer et al.  is incomplete. Based on the results to our calculations, the influence that ClMBL has on δ13CH4 cannot be ignored in future interpretations of the glacial-interglacial δ13CH4 record, and hence further research is needed to quantify past changes in the strength of this sink, including those relating to changes in the sea-ice source of SSA.
This work has been carried out as part of the British Antarctic Survey Polar Science for Planet Earth programme. We gratefully acknowledge the funding of the Natural Environment Research Council. The authors also wish to thank the PMIP2 international modeling groups for providing their data for analysis, and the Laboratoire des Sciences du Climat et de l'Environnement (LSCE) for collecting and archiving the model data. The PMIP2/MOTIF Data Archive is supported by CEA, CNRS, the EU project MOTIF (EVK2-CT-2002-00153) and the Programme National d'Etude de la Dynamique du Climat (PNEDC). The analyses were performed using version 10-13-2006 of the database. More information is available on http://pmip2.lsce.ipsl.fr. Finally, we express our thanks to two anonymous reviewers.