Upon closer inspection, the classical view of the synchronous relationship between tropospheric methane mixing ratio and Greenland temperature observed in ice samples reveals clearly discernable variations in the magnitude of this response during the Late Pleistocene (<50 kyr BP). During the Holocene this relationship appears to decouple, indicating that other factors have modulated the methane budget in the past 10 kyr BP. The δ13CH4 and δD-CH4 of tropospheric methane recorded in ice samples provide a useful constraint on the palaeomethane budget estimations. Anticipated changes in palaeoenvironmental conditions are recorded as changes in the isotope signals of the methane precursors, which are then translated into past global δ13CH4 and δD-CH4 signatures. We present the first methane budgets for the late glacial period that are constrained by dual stable isotopes. The overall isotope variations indicate that the Younger Dryas (YD) and Preindustrial Holocene have methane that is 13C- and 2H-enriched, relative to Modern. The shift is small for δ13CH4 (approx. 1‰) but greater for δD-CH4 (approx. 9‰). The YD δ13CH4–δD-CH4 record shows a remarkable relationship between them from 12.15 to 11.52 kyr BP. The corresponding C- and H-isotope mass balances possibly indicate fluctuating emissions of thermogenic gas. This δ13CH4–δD-CH4 relationship breaks down during the YD–Preboreal transition. In both age cases, catastrophic releases of hydrates with Archaeal isotope signatures can be ruled out. Thermogenic clathrate releases are possible during the YD period, but so are conventional natural gas seepages.
The world is entering an unprecedented period in our history where changes to our climate are being induced by our release of radiatively active greenhouse gases (GHGs) into the atmosphere. Anthropogenic emissions of GHGs, such as CO2, CH4 and N2O, continue to increase these tropospheric gas burdens. For example, the 5 Gt of methane in the atmosphere today raises the global temperature by approximately 1.3 K (Donner & Ramanathan 1980), and the change in forcing by methane since 1750 has been 0.55 W m−2 (Minschwaner et al. 1998; Jain et al. 2000). Our knowledge that these changes are occurring are founded on meticulous time-series trace gas measurements, such as the NOAA CMDL CCGG cooperative air sampling network (e.g. Conway et al. 1994; Dlugokencky et al. 2005), or the visionary Mona Lau dataset (Keeling & Whorf 2005).
Today, the global methane mixing ratio, [CH4], is 1780 ppbv (CMDL NOAA data; Dlugokencky et al. 1994, 2003) and is currently rising at a rate of 4±4 ppbv yr−1 (Dlugokencky et al. 1998, 2005; Bousquet et al. 2006). Using bottom-up approaches, i.e. wide-ranging field measurements of CH4 emissions intensities from the significant sources and relevant CH4 sinks, we are able to generate reasonable models to describe today's atmospheric CH4 budget. Although sizable uncertainties remain in accurately characterizing the magnitude and changes of the major types of source fluxes, the budgets can be constrained by knowledge of the tropospheric CH4 burden, the present day stable carbon and hydrogen isotope ratios (δ13CH4 and δD-CH4, respectively) and the radiogenic 14CH4. The increase in present day [CH4] over the natural background is largely the result of anthropogenic releases of gas from sources such as rice paddies, landfills, livestock, biomass burning and natural gas usage. The associated rise in global temperature causes serious feedbacks that exacerbate the issue. These may also involve the elevation of CH4 emissions from natural sources, including enhanced methanogenesis from tropical and boreal wetlands, CH4 release by retreating permafrost, and the potential destabilization of marine and terrestrial CH4 clathrates (gas hydrates). A key component to our understanding of climate variability is to differentiate natural and anthropogenic methane emissions and sinks for present and past periods e.g. Prather et al. (2001) and Wunsch (2006).
The current methane concentration [CH4] represents an approximately 2.5 times increase in tropospheric methane in comparison with times before intensive human interference i.e. the Late Preindustrial Holocene (PIH; 2000 years BP), when [CH4] was approximately 700 ppbv (e.g. Etheridge et al. 1998; Ferretti et al. 2005). Further back in time, during glacial periods, [CH4] was up to five-times less than at present, i.e. [CH4] approximately 320–400 ppbv. We know this from detailed tropospheric gas histories chronologically recorded back to ca 650 kyr BP in the occluded bubbles of the Antarctic ice sheets (Spahni et al. 2005) and 110 kyr BP in the Greenland ice sheets (Brook et al. 1996, 2000). These time-series reveal that tropospheric methane concentrations systematically rose and fell within the range of approximately 350–750 ppbv. Over orbital cycles, the variations of [CH4] in the ice track those of other indicators, such as [CO2] and temperature (Petit et al. 1999). However, a detailed view shows that [CH4] is very closely correlated with Greenland temperature, which is antiphased with Antarctic climate throughout the last glacial period. For the end of the last ice age, the gas stratigraphy of the Greenland GISP2 cores, as established with thermally fractionated nitrogen and noble gas isotopes, suggests that the rapid rises in Greenland temperature lead the corresponding rises in [CH4] by up to 50 years (Severinghaus et al. 1998; Severinghaus & Brook 1999). In comparison, the rises in [CO2], as recorded in Antarctic ice, lag Antarctic temperature by ca 800 years e.g. Caillon et al. (2003). Similarly, for this period, Sowers & Bender (1995) reported that atmospheric [CO2] and [CH4] rose 2000–3000 years prior to the warming in the Northern Hemisphere.
Figure 1 is a traditional time-series representation of the δ18Oice and [CH4] for the past 50 kyr of the Greenland GISP2 ice core datasets. These GISP2 δ18Oice measurements are an indicator of local temperature and were reported by Grootes et al. (1993), Meese et al. (1994), Steig et al. (1994), Stuiver et al. (1995) and Grootes & Stuiver (1997). The GISP2 methane concentrations and gas ages are from Brook et al. (1996, 2000). Figure 1 also shows the Heinrich (H1-5) and Dansgaard–Oeschger (1–13) events observed in the GISP2 core. All data are hosted by ncdc.noaa.gov/pub/data/paleo/icecore/greenland/summit/gisp2. The relationships between δ18Oice and [CH4] are generally inferred only through visual inspection. This mode of comparison is largely necessary due to the uncertainty in aligning the gas age scales with the ice age.
To further elucidate these anticipated relationships, figure 2 plots the Greenland palaeotemperature against [CH4] for the youngest 50 kyr in the GISP2 core. The palaeotemperatures for GISP2 are calculated using transfer functions from corresponding GISP2 δ18Oice measurements as reported by Cuffey & Clow (1997) and Alley (2000, 2004). Figure 2 is constructed by harmonizing the ice and gas ages, i.e. by matching depths (nearest ice age temperature to [CH4] gas age). In addition to sampling and analytical uncertainties, the GISP2 data in figure 2 are subject to (i) factors other than temperature that affect δ18Oice, (ii) errors in the calculation of temperature from δ18Oice measurements (oxygen isotope temperature sensitivity (α), i.e. Δδ18Oice/ΔT), and (iii) errors in the absolute and relative ice and gas ages.
The factors affecting δ18Oice, such as seasonality and proximity of precipitation, sea-surface conditions and atmospheric circulation, have been discussed in detail by numerous authors, including Paterson (1994), Cuffey et al. (1995), Boyle (1997), Fawcett et al. (1997), Jouzel et al. (1997), Krinner et al. (1997), Werner et al. (2001) and Masson-Delmotte et al. (2005). The α value has been shown to vary over time. The present day α value is estimated at 0.67‰ K−1 (Johnsen et al. 1989), while estimates of the Holocene α value range from 0.53 to 0.67‰ K−1 (Cuffey et al. 1994, 1995). An α value of 0.3‰ K−1 for the Younger Dryas (YD) was reported by Cuffey et al. (1995) and Johnsen et al. (1995) and this was later substantiated with gas isotope ages by Severinghaus et al. (1998), Grachev & Severinghaus (2005) and Huber et al. (2006). Using the work of Schwander et al. (1997), Lang et al. (1999) reported an α value of 0.4–0.5‰ K−1 for the Dansgaard–Oeschger events between 20 and 40 kyr BP. The accuracy of the α value for a chosen age translates directly into the accuracy of the temperature. To test this, figure 3 is a plot of GISP2 δ18Oice against [CH4], constructed by using interpolated δ18Oice ice ages to fit the [CH4]. Figure 3 produces essentially the same plot as in figure 2. This points out that the changes observed in temperature and methane are much larger than any error associated with the chosen α value.
Uncertainty created by errors in the absolute ice ages (estimated as approx. 2% by Severinghaus & Brook (1999)) does not affect the temperature–methane relationships because the error is applicable to both the ice and the gas ages. Estimating the relative difference between the ice and gas ages is however a source of error of potentially ≤100 years. This is related to the confidence in determining the closure age (Δage=gas age−ice age difference; Brook et al. 2000). It is also due to the fact that the ice taken for the gas measurements is not identical in age to ice samples taken for δ18Oice. The uncertainty increases somewhat with depth, but with a few exceptions is ≤100 years. We tested the robustness of the temperature–methane and δ18Oice–[CH4] relationships observed in figures 2 and 3 by purposefully misaligning (shifting) temperature and δ18Oice ice ages by up to 187 years too young and 168 years too old relative to the interpolated ice ages. Figure 4 shows that the patterns seen in figures 2 and 3 are largely maintained despite the age displacement, indicating that over century-range age intervals, the age matching is not critical.
Acknowledging age-scale uncertainties and sensitivities, we feel it is valuable to make the following observations. In general, figure 2 demonstrates the well-established trend towards increasing methane mixing ratios with increasing local temperatures. The linear regression through the entire dataset (line 3), although potentially a spurious correlation, reveals an approximately 20 ppbv increase in [CH4] with each 1°C rise. In addition, figure 2 indicates some other interesting features. We divided the 50 kyr BP dataset into several time blocks, reflecting various time periods of interest. For example, the Holocene data (133–9 646 years BP) are clearly segregated from the rest of the data. These warmer Holocene data (approx. −30°C) have no apparent relationship to temperature, despite an approximately 200 ppbv range in [CH4]. Excluding the Holocene data, the [CH4]–temperature correlation (line 2) drops to an approximate 1°C rise with a 30 ppbv increase in [CH4]. In addition, the older data (38 913—49 925 years BP) indicate a tight, steeper trajectory (line 1) of approximately 1°C rise with a 10 ppbv increase in [CH4]. Four examples of shorter transitions are given in figure 2 for (i) 11 407–11 589 years BP i.e. the YD–Preboreal period (YD–PB), discussed in more detail in this paper, (ii) 12 500–14 538 years BP, (iii) 14 486–16 767 years BP, and (iv) 38 166–38 619 years BP. These all occupy distinct regions of the [CH4]–temperature diagram. Similarly, the δ18Oice–[CH4] relationships show the age segmentation as shown in figure 3.
The correlation between [CH4] and δ18Oice, and hence the derived Greenland temperatures, given by figure 3, indicate a teleconnection between the climate in Greenland and that of methane producing regions, including those controlled by tropical monsoon (Wang et al. 2001; Hughen et al. 2004). The important implication of figures 2 and 3 is that this teleconnection is not constant and is subject to subtle variations even over relatively short geological time spans. These are probably governed by different climate conditions within and since the last ice age. The point is that caution is required when reconstructing latitudinal climate connections from the Greenland [CH4]–temperature relationship. However, the variations in the [CH4]–temperature relationship may also provide clues for understanding the interdependence of North Atlantic and tropical climate.
The variations in [CH4] observed in figures 2 and 3 beg the question as to their cause. Are different intensities of the source fluxes (even those that are intermittent) and/or of the sinks changing the atmospheric methane burden on these time scales? Or, are the same sources and/or sinks changing their intensities with time? To address these questions, we use the combination of methane source and sink flux estimates together with their corresponding carbon and hydrogen stable isotope ratios (δ13CH4 and δD-CH4). This paper examines atmospheric methane budgets for four time periods, namely (i) Modern, (ii) PIH (approx. 300–2000 years BP), (iii) YD–PB transition (approx. 11.4–12.2 years BP), and (iv) Late Glacial Maximum (LGM; approx. 18 000 years BP). In particular, the objective is to combine the first reported δ13CH4 and δD-CH4 measurements in ice for palaeoatmospheres.
2. Source–sink components of Modern global methane inventory
Various approaches are used to estimate the Modern fluxes of methane to the atmosphere. These include inverse modelling, e.g. Hein et al. (1997), Houweling et al. (2000), Law & Vohralik (2001), Enting (2002), Butler et al. (2004), Michalak et al. (2005) and Chen & Prinn (2006), satellite measurements, e.g. SCIAMACHY (Frankenberg et al. 2005, 2006) and bottom-up mass balance approaches, e.g. Matthews & Fung (1987), Cicerone & Oremland (1988), Stevens & Engelkemeir (1988), Aselmann & Crutzen (1989), Houghton et al. (1990), Fung et al. (1991), Whiticar (1993), Lelieveld et al. (1998), Lassey et al. (2000) and Kaplan et al. (2006).
For the purposes of this paper, we do not rederive the Modern budgets, rather we use them to guide us in predicting past budgets. Considerable uncertainty exists in estimating the magnitude of the major source fluxes. To illustrate this, Keppler et al. (2006) recently reported aerobic methane production (AMP) from plants, a previously unknown source. Their scaled estimate for the global emission of methane from this source was 62–236 Tg yr−1, i.e. potentially the largest single source of atmospheric methane. Subsequently, Houweling et al. (2006) reduced this to between 85 and 125 Tg yr−1, i.e. similar to methane flux strength estimated for wetlands (approx. 140 Tg yr−1; Kaplan 2002). To balance the accomodate in the global methane source budgets while incorporating AMP, it is necessary to decrease other sources, particularly wetlands. Otherwise, these are most likely to be redundantly, i.e. double accounted, with AMP.
Difficulties in assigning accurate mean δ13CH4 and δD-CH4 values for the different sources also remain (note, all stable isotope ratios are reported as per mil relative to the Vienna Pee Dee Belemnite (VPDB) and Vienna standard mean ocean water (VSMOW) standards). For example, estimates of the mean δ13CH4 for termite emissions vary between −41 and −82‰ (Tyler et al. 1988; Sugimoto et al. 1998; Tayasu 1998). The extreme variability largely reflects the taxonomy and diet of the termites (Boutton et al. 1983) and the degree of oxidation (Zimmerman et al. 1982). The challenges of using the correct mean δD-CH4 values are exemplified by biomass burning. Early work by Whalen et al. (1990), Whalen (1993) and Whiticar (1993) suggested δD-CH4 for biomass burning from −90 to −300‰. More recently, Snover et al. (2000) reported a mean value of −210‰, which contrasts with the estimated global δD-CH4 value of −169‰ calculated by Yamada et al. (2006). We use a δD-CH4 value of −225‰ for biomass burning in this paper.
Despite the problems in attributing fluxes or isotope signatures to the major sources, we can use the Modern values to provide us with a framework/baseline for the expected relative changes in the three past ages that we examine here. In addition, we can incorporate the sink terms for the different ages to estimate the tropospheric budget. Figures 5 and 6 show, for present day, the flux strengths of the primary sources of tropospheric methane and their respective carbon and hydrogen isotope ratios. Rice paddies are the single largest source at 110 Tg yr−1, but the combination of tropical and boreal wetland fluxes exceed this with 115 Tg yr−1 (table 1). Although uncertain, we assign a value of only 40 Tg yr−1 for AMP, and have decremented the wetland contribution accordingly. We do not assign a flux for the Modern geological source (natural seepages of thermogenic methane; table 1), but we have accommodated this and the corresponding anthropogenic emissions under natural gas and coal emissions (45 and 35 Tg yr−1, respectively).
The δ13CH4 values for the various sources range from approximately −25 to −65‰. The range in δD-CH4 is −140‰ (coal) to −390‰ (termites). Using the values in table 1, the carbon and hydrogen isotope ratio mass balance (FiTδiT) for the source inputs can be calculated by the simple linear mixing equation,(2.1)where Fix are the source fluxes (Tg yr−1), and the isotope ratio of the individual sources is δix. A carbon isotope mass balance from the sources is calculated by equation (2.1) to have a weighted mean δ13CH4 of −54.2‰ (termed δCin in table 1) as shown in figure 2. The corresponding mean δD-CH4 of the sources (δDin) is −294‰ (table 1; figure 6).
Four primary sinks of atmospheric methane are considered, i.e. (i) tropospheric abstraction with hydroxyl radicals (OHċ), e.g. Saueressig et al. (2001), (ii) soil uptake, e.g. Ridgwell et al. (1999), (iii) stratospheric removal, e.g. Rice et al. (2003), and (iv) reaction of CH4 and Clċ in the marine boundary layer (MBL), e.g. Allan et al. (2005). Each sink is associated with an isotope effect that leads to isotopic fractionation . The combined sink isotope fractionation can be recast in equation (2.2) as the isotope offset , i.e.(2.2)where are the sink fluxes (Tg yr−1). The first three sinks combine to give an effective weighted total carbon isotope offset of 7.4‰, which is used in this paper. We currently do not account for the MBL sink, but considering the work of Allan et al. (2005), future budgets may need to do so. The associated hydrogen isotope offset caused by the tropospheric methane sink is 200‰.
The calculated atmospheric isotope value (δT) is calculated using both the source and the sink terms in the net carbon and hydrogen isotope ratio mass balance (FTδT) according to equation (2.3)(2.3)
Using the bottom-up estimates of δin, the mean source input (δ13CH4, −54.2‰; δD-CH4, −295‰), and the sink isotope shifts , the calculated atmospheric δ13CH4 is −46.8‰ and the net δD-CH4 is −95‰ (table 1; figures 5 and 6). The calculated atmospheric values can be compared with the current, global measured atmospheric measured values for δ13CH4 of −47.1‰ (Miller et al. 2002) and δD-CH4 of −86‰ (Whalen 1993; Quay et al. 1999; Bergamaschi et al. 2000). The difference between the measured atmospheric and calculated atmospheric estimates is relatively small for δ13CH4 (0.3‰), but larger (9‰) for δD-CH4. This illustrates the ongoing challenge in assigning the correct fluxes and/or isotope ratios. If we anchor the mass balance with the current measured atmospheric δ13CH4 and δD-CH4 values of −47.1 and −86‰ and incorporate the sink isotope shifts , we can back-calculate the isotope ratio of the global expected mean input of methane to the troposphere. In this case, the expected mean input δ13CH4 is −54.5‰ and δD-CH4 is −286‰ (table 1), compared with δCin of −54.2‰ and δDin of −295‰. The discrepancies point to the deficiencies we have in describing the source and sink fluxes. Presently, we do not have an a priori way to improve the budgets, and the discrepancies remain.
It is interesting to note in figure 5 that many of the sources have carbon isotope ratios approximately −60‰, i.e. close to δCin. This is expected and indicative of the dominant Archaeal origin for the majority of the methane. Most of the leverage towards a 13C-richer input to δCin is provided by biomass burning, landfills, coalbed methane and natural gas emissions (Whiticar 1994). All of these are largely anthropogenic-influenced sources. AMP is presently somewhat of a wildcard, as both the source magnitudes and the isotope ratios (especially δD-CH4) need to be better defined.
A different distribution is seen for δD-CH4 in figure 6. The δD-CH4 signatures of the sources congregate into two primary groups i.e. one group with δD-CH4 less than −300‰ and a second with δD-CH4 greater than −250‰. This segregation largely reflects the δDwater associated with the source and the methane formation pathway involved (Whiticar 1999).
The combination of tropospheric δ13CH4 and δD-CH4 together with the total net flux i.e. sources − sinks, do not provide sufficient constraints to unambiguously model the tropospheric methane budget (e.g. Whiticar 1993; Lassey et al. 2000; Bréas et al. 2001). However, the isotope signatures of the different sources do occupy specific regions of a δ13CH4 versus δD-CH4 plot, i.e. CD diagram (Whiticar et al. 1986). Figure 7 is the CD diagram for present day methane depicting the fields of the various methane origins, such as thermogenic, geothermal and methanogenic. The latter, i.e. Archaeal methane, is traditionally labelled with the misnomer ‘bacterial’ or ‘biogenic’ methane. The column heights of the methane sources shown in figure 7 are proportional to their flux strengths. Also shown in figure 7 is the integrated mean δ13CH4 and δD-CH4 input values (δCin, δDin) and the current calculated atmospheric δ13CH4 and δD-CH4. The list of methane sources shown along the side of figure 7 include some that are not important in Modern tropospheric methane budgets, but factor into the palaeoatmospheres, as discussed below. Although various sources are scattered across the plot, there is an aggregation of several larger sources in the sector of figure 7 that is associated with methyl-type fermentation by Archaeans.
3. Source components of PIH, YD–PB and LGM global methane inventories
Recognizing the uncertainty in the Modern inventory, estimates in palaeobudgets of tropospheric methane can be made in a relative sense. Essentially, we attempt to determine how the global palaeoenvironment has changed and how these changes translate into shifts in the methane source and sink fluxes and isotope signatures. The influence of these factors on the δ13CH4 of source types for the PIH, YD–PB and LGM is treated in detail by Schaefer & Whiticar (in press). For these timeframes, methane concentration studies have been made on numerous ice cores from Greenland and the Antarctic, including the Vostok and GISP2 cores (e.g. Craig & Chou 1982; Chappellaz et al. 1993a, 1997; Brook et al. 1996, 2000; Etheridge et al. 1998, Petit et al. 1999; Dällenbach et al. 2000; Stauffer et al. 2002). In contrast, only a few isotope measurements have been made on non-Modern methane. These currently include δ13CH4 determinations by Craig et al. (1988), Ferretti et al. (2005), Sowers et al. (2005), Schaefer et al. (2006), Sowers (2006a) and Schaefer & Whiticar (2007). The only values for palaeo-δD-CH4 from ice-core samples are published by Sowers (2006b). As a consequence, the available constraints on the palaeoatmospheric budgets are currently limited to these few δ13CH4 and δD-CH4 measurements.
Certainly changes to the inventory of sources, such as the addition of anthropogenic sources, can differentiate Modern budgets from those in the past. However, there are several ways in which changes in the environment can over time alter not only the fluxes, but also the isotope signatures of sources and sinks of methane. These, in turn, shift the global δ13CH4 and δD-CH4 budget. A primary factor is changes to the isotope signature of organic matter that serves as methane precursors. This includes (i) substrates for methanogenesis e.g. wetlands, termites, etc., (ii) kerogen for natural gas (coal, thermogenic or geothermal gas), or (iii) fuel for biomass burning. Environmentally related changes to the sinks can also exert control on the methane isotope budgets.
Changes to the carbon isotope ratio of organic matter (δ13Corg), which serves as precursor substrates for methanogenesis, can potentially lead to systematic shifts in δ13CH4. These organic compounds are predominantly produced and obtain their signature through plant biosynthesis. The δ13C of plant matter depends directly on atmospheric δ13CO2, (e.g. White et al. 1994; Gröcke 2002), as well as on the photosynthetic pathways (Calvin–Benson cycle and Hatch–Slack cycle, i.e. C3 and C4 photosynthesis, respectively) and various environmental parameters, such as water tensions, temperature and biostressors (e.g. infestations).
Schaefer & Whiticar (in press) studied how fluctuations in atmospheric δ13CO2 for PIH and LGM as recorded in ice cores and plant fossils changed the δ13CH4 of various methane sources. The same work also examines the influence of shifts in C3 and C4 vegetation, as modelled by Collatz et al. (1998), on the carbon isotope signature of various source types. The latter are sensitive to these climatically induced vegetation changes, because C3 and C4 plants carry different isotope signatures of approximately −27 and −15‰, respectively, e.g. Ehleringer et al. (1997), which are passed on to the methane generated from these precursor materials. Detailed calculations of potential changes in source δ13CH4 are presented in Schaefer & Whiticar (in press), and the following is a short summary of the relevant information.
Plant fossils indicate a 1.1‰ 13C-enrichment of atmospheric δ13CO2 during the PIH, relative to the present (Marino et al. 1992), in good agreement with δ13CO2 reconstructions from ice cores by Francey et al. (1999) and Smith et al. (1999). Marino et al. (1992) also reported a relative 0.4‰ 13C-enrichment in δ13CO2 during glacial conditions. Ice-core data reveal a 13C-enrichment of δ13CO2 by 0.9‰ in the LGM and a 0.4‰ 13C- enrichment versus today during the YD cold period. Thus, all things equal, the shift in δ13Corg transfers into a similar shift in the δ13CH4 of methanogenic emissions.
δ13CH4 changes caused by C3–C4 vegetation shifts are difficult to assess, e.g. relative expansion of grasslands (Prentice et al. 1993; Steudler et al. 1996). However, if the amount generated remains constant for each type, then a shift in the proportion of C3 and C4 can transfer into a similar shift in the δ13CH4 of the emissions. Collatz et al. (1998) reported that C4-dominated grasslands increased from 70% at the LGM to 74% in the PIH, then dropped to the present day 57% due to anthropogenic land conversion and rising [CO2]. This is reflected in our methane budgets. Although the C3/C4 is poorly constrained, vegetation models by Kaplan et al. (2004) supported the findings of Collatz et al. (1998). Furthermore, Schaefer (2005) found that even a ±50% change in C3–C4 methane fluxes resulted in an atmospheric δ13CH4 shift of only 0.6‰.
Ruminants will produce methane with a δ13CH4 dependent on the ratio of C3 and C4 in their diet i.e. a 12‰ difference in δ13CH4 for C3 versus C4, as shown by Rust (1981), Metges et al. (1990), Ehleringer & Monson (1993), Levin et al. (1993), Schulze et al. (1998) and Bilek et al. (2001). The proportion of C3 versus C4 plants will also impact the δ13CH4 emitted from wildfires. For example, Chanton et al. (2000) reported for biomass burning δ13CH4 ranges of −26 to −30‰ for C3 forest fires versus −17 to −26‰ for C4 grass fires. Today's average biomass burning δ13CH4 is calculated to be −24.6‰ according to data from Hao & Ward (1993). The C3 versus C4 change with time translates to a wildfire average δ13CH4 of −26.6‰ in the PIH and −26.7‰ in the LGM (Schaefer & Whiticar in press).
AMP has also been shown to depend on whether the plant type is C3 or C4. Keppler et al. (2006) report a shift of approximately 6‰ in AMP δ13CH4 derived from C3 versus C4. This can result in AMP δ13CH4 of −50.5‰ for the PIH and −49.9‰ for the LGM.
Strangely, Tyler et al. (1988) found no diet relationships between C3 and C4 plants for the δ13CH4 produced by termites in various habitats, even within a single species. As this is unlike all the other methanogenic sources, Schaefer & Whiticar (in press) incorporated a C3–C4 response. They calculated a termite-derived δ13CH4 of −62.2‰ for the PIH and −62.7‰ for the LGM, compared to a Modern value of −63.0‰.
It is complicated to quantify production of methane from various wetlands and to assign representative isotope signatures. Methanogenesis and the derived δ13CH4 in wetland systems respond to the influences of
dominance of the methanogenic pathway, i.e. methyl fermentation (MF) or carbonate reduction (CR; Whiticar et al. 1986),
the fraction of precursor material utilized,
These factors suggest that there should be a significant difference in δ13CH4 between glacial and interglacial wetlands. Determining a representative value for the boreal wetlands remains difficult, even for Modern conditions. For example, Bellisario et al. (1999) reported δ13CH4 values between −50‰ and −73‰ for minerotrophic systems, whereas Chanton et al. (2000) and Bowes & Hornibrook (2006) found δ13CH4 values ranging from −70 to −90‰. Such 13C-depleted values are very difficult to reconcile for the PIH, YD or LGM budgets (table 1). Realizing the uncertainty, we use more traditional δ13CH4 values of −62‰ for the Modern, PIH and YD boreal wetlands. To give an indication of the sensitivity of the chosen δ13CH4 for boreal wetlands, values of −70 and −80‰ for the Modern budget would shift the δCin to −47.3 and −48.0‰, respectively. The former is comparable to the measured tropospheric δ13CH4 value of −47.1‰, but the latter would require adjustments to other sources to comply. The PIH and YD budgets can accommodate a boreal wetland δ13CH4 of −70‰ (δCin shifts to −53.5 and −52.8‰, respectively), but the LGM remains poorly unbalanced. Boreal wetland δ13CH4 values of −80‰ for the PIH, YD and LGM do not reconcile well. Schaefer & Whiticar (in press) estimate that LGM wetland δ13CH4 was more 13C-depleted than in the PIH by 0.3–0.9‰, depending on the MF-derived fraction (fMF) relative to CR production. The temperature-dependent carbon isotope fractionation contributes to this 13C-depletion of δ13CH4 in the LGM with 0.3–0.4‰. A shift from MF to CR at lower temperatures could have led to an additional contribution of 0.3–0.5‰. As a result, a δ13CH4 value of −63‰ is used for the boreal wetlands in the LGM budget. Similarly, δ13CH4 of tropical wetlands varied by up to 2.1‰ between the LGM and Modern (table 1). Based on this, Schaefer & Whiticar (in press) suggested that glacial–interglacial temperature changes have an impact on the CH4 isotope mass balance that is detectable in ice-core measurements (table 1).
The above mentioned factors influencing the carbon isotope ratio of the tropospheric methane sources are summarized in table 2. The magnitudes of the relative changes in δ13CH4 for the different sources between the LGM and PIH, and between the PIH and Modern are illustrated in figure 8. Note that sink components are not incorporated into this figure. Clearly the shifts in δ13CH4 are generally greater between the PIH and today, than between the LGM and PIH. In addition, the PIH–Modern comparisons frequently show 12C-enrichments for the present sources with δ13CH4 shifts of 1–4‰, due to human intervention. In contrast, the LGM–PIH show more limited differences, with modest 12C-enrichment during the glacial period. The flux weighted mean (δCin in figure 8) reflects this with the δCin approximately +2‰ for PIH-Modern, yet δCin approximately +0.6‰ for the LGM–PIH.
As mentioned, palaeo-δD-CH4 are currently more speculative and less well constrained. On the other hand, several of the environmental factors that can shift δ13CH4 are not applicable to the δD-CH4 systematics. To a large extent, δD-CH4 is controlled by δDwater. For example, Whiticar et al. (1986), Burke et al. (1988), Sugimoto & Wada (1995), Hornibrook et al. (1997), Waldron et al. (1999) and Whiticar (1999) and others showed that in methanogenesis the δD-CH4 is dependent on the MR or CR pathway, i.e. if the hydrogen is transferred from organic substrates or H2. Ultimately, the δD-CH4 for either pathway is determined by δDwater. Similarly, the δD-CH4 in non-methanogenic sources such as coal, natural gas and probably AMP, have their δD-CH4 set by δDorg, which is also determined by δDwater. The challenge is to determine the appropriate hydrogen isotope effects for the different methane forming processes, and then applying the correct adjustments, i.e. temperature dependence, change in ocean and meteoric δDwater. For the purposes of this paper, the δD-CH4 for each source and time period uses the present day value. For AMP, a δD-CH4 value of −260‰ is currently assigned (Frank Keppler 2006, personal communication). Table 1 lists the representative δD-CH4 values taken from a merger of different estimates for Modern methane sources (e.g. Whiticar 1990, 1993; Snover et al. 2000). These source δD-CH4 values are combined with their δ13CH4 and scaled by their flux magnitude in figure 7. The resultant flux weighted mean tropospheric input δD-CH4 (δDin) is calculated using equation (2.1) and also shown in figure 7. To estimate δDin for the three older time periods, the individual methane source fluxes for the PIH, YD–PB and LGM as derived from Chappellaz et al. (1993b) are used to scale the relative proportions of each methane source. The resulting δDin are listed in table 1.
To compare the shifts in the source δ13CH4, δD-CH4 and flux magnitudes for the Modern, PIH, YD–PB and LGM ages, a series of CD diagrams with flux histograms are plotted in figure 9. Although there are substantial changes in the fluxes, the changes in δ13CH4 and δD-CH4 are subtle. The flux weighted mean tropospheric inputs (δCin and (δDin) are shown in figure 9, but their subtle differences with ages are more clearly illustrated by the expanded mean of source inputs box in figure 10.
4. Sink components of the PIH, YD–PB and LGM global methane inventories
The sinks of atmospheric methane are mercifully less diverse than the sources. As mentioned above, the four primary are
The magnitude of the isotope effects for each sink carries some uncertainty, but typically the reaction rate constants are faster for 12CH4 than the common isotopologues 13CH4 or 12CH3D e.g. Gordon & Mulac (1975), Davidson et al. (1986, 1987), Alperin et al. (1988), Cantrell et al. (1990), Tyler et al. (1994, 2000), Saueressig et al. (1995, 2001), Reeburgh et al. (1997), Snover & Quay (2000), and Michelsen & Simpson (2001). The result of these sinks is an enrichment of the residual tropospheric methane in 13C and 2H relative to the mean of the source inputs. The magnitudes of the sinks are discussed in detail in Schaefer & Whiticar (in press) and earlier by Staffelbach et al. (1991), and are summarized in table 2. The isotope shifts due to the Modern combined sinks are of 7.4‰ for carbon and of 200‰ for hydrogen. However, different environmental conditions in the PIH, YD–PB and LGM compared with today, lead to different values for . We estimate for the PIH as 7.0‰, YD–PB as 6.4‰ and LGM as 5.7‰ (table 1). Although cannot yet be accurately defined for the palaeo-ages, Sowers (2006b) suggested that it could have changed in the range of +3.4‰ and −1.6‰ between the PIH and LGM. For this paper, we use the Modern of 200‰ for all four ages. We recognize and accept the limitation this places on our calculations.
The effect that these calculated changes in and processes that affect the δ13CH4 of source types have on atmospheric δ13CH4 in the past is shown in figure 11. It is important to note that here these budget reconstructions are not intended or adjusted to match past atmospheric δ13CH4 values that have been measured in ice. The latter would require us to rebalance the source inventories, which we attempt later in the paper. Rather, the values summarized in table 2 and shown in figure 11 illustrate the fact that past environmental changes affected atmospheric δ13CH4 through changes in δ13CH4 of the sources and in sink isotope fractionation. It should be emphasized that the δ13CH4 could be affected even if the source inventory, i.e. the relative magnitude of individual fluxes, had remained the same. However, we recognize that some sources do, in fact, undergo higher changes in their isotopic signature than others. To address the issue that the studied changes affect δ13CH4 in addition to, but not independently of, the source inventory, we based the reconstructions shown in figure 11 on source fluxes derived from Chappellaz et al. (1993b). We then calculated the difference in the isotope mass balance between a scenario with modern source δ13CH4 values and sink fractionation and a scenario where the environmental changes in the latter are included. To illustrate this, figure 11 shows the relative importance of the changes in source and sink on δ13CH4 for the PIH, YD–PB and LGM relative to Modern (0‰ line). An interesting detail is our finding that the environmental changes should lead to a small (approx. 1‰) 13C-depletion in the LGM, while first results (Sowers 2006a; Vas Petrenko 2006, unpublished data) show that atmospheric δ13CH4 is strongly enriched in 13C. Obviously, a higher percentage of 13C-rich emissions overwhelms the changes in δ13CH4 of individual sources and sinks.
5. Constraining the PIH, YD–PB and LGM with δ13CH4 and δD-CH4
Bottom-up source scenarios for the past, based on vegetation models and geological evidence, can be constrained by the isotope mass balances, because a valid scenario must produce a calculated atmospheric isotope signature that matches the ice core results. Despite the fact that for every time period a multitude of scenarios is possible, the isotopic constraints improve the budget reconstructions by discriminating against models that violate the mass balance or exceed the uncertainty of the parameters. Using the parameters derived above, we adjusted the source fluxes of bottom-up source scenarios in order to best match the isotopic mass balance for both δ13CH4 and δD-CH4. The resulting methane budgets for PIH, YD–PB and LGM (table 1) are not unique solutions to the problem, but represent our best guess.
Plugging the δCin and values in table 1 into equation (2.3) results in calculated atmospheric δ13CH4 values of −45.6, −45.8 and −45.6‰ for the PIH, YD–PB and LGM, respectively. These are all 13C-enriched compared with the present day calculated atmospheric δ13CH4 of −46.8‰, by 1–2‰.
Similarly, using δDin and equation (2.3) gives calculated atmospheric δD-CH4 values of −98, −96 and −87‰ for the PIH, YD–PB and LGM. Methane for the palaeo-periods is enriched in 2H relative to today (calculated atmospheric δD-CH4=−95‰) by up to 9‰. These age differences in both calculated atmospheric δ13CH4 and δD-CH4 are demonstrated in figure 10 for the Modern, PIH, YD–PB and LGM.
The differences between each age represent the composite of the chosen source isotope ratios, the fractionation of the sinks and the magnitude of the source fluxes. In the case of the Modern, the accuracy of δin and can be evaluated by comparison with direct measurement of tropospheric air samples. To evaluate the PIH, YD–PB and LGM, we resort to ice samples to find the measured atmospheric isotopic signature for each age (Ferretti et al. 2005; Schaefer et al. 2006; Sowers 2006a,b; Schaefer & Whiticar in press).
6. Reconciliation of LGM budgets
The methane budget for the LGM, as presented in table 1 and figures 9 and 10, was developed using the reconstructions for past source δ13CH4 and sink fractionation from Schaefer & Whiticar (in press) and flux estimates as derived from palaeovegetation models and geological evidence. The resulting calculated atmospheric δD-CH4 of −87‰ agrees well with measured atmospheric δD-CH4 of −87‰ by Sowers (2006b). In contrast, our δ13CH4 budget predicts a calculated atmospheric value of approximately −45.6‰, whereas the first results of measured atmospheric values for the LGM (Sowers 2006a; Vas Petrenko 2006, unpublished data) indicate that the real values may be up to 3–4‰ more 13C-rich than our budget reconstruction in table 1 suggests, i.e. δ13CH4 of −43‰. The reconstructions of source type δ13CH4 and sink fractionation by Schaefer & Whiticar (in press) are already incorporated in the YD and LGM budgets. The overall impact (figure 11) is a small 13C-depletion in glacial and near-glacial conditions and cannot account for or contribute to the observed 13C-enrichment.
In order to resolve the discrepancy between calculated atmospheric and measured atmospheric values, we attempt here to reconcile the budget having a source inventory with the appropriate isotope mass balance. Only two, or possibly three, natural sources (biomass burning, geological and perhaps AMP) are sufficiently 13C-enriched to account for the offset, i.e. to drive the measured atmospheric to −43‰. It should be noted that for every scenario of adjusting the sources for new values of δCin and δDin, it was still not possible to fully reconcile all constraints, namely: (i) measured atmospheric δ13CH4 and δD-CH4, (ii) flux, and (iii) YD budget.
Firstly, biomass burning from wildfires is the most 13C-enriched source. However, its natural emissions account for less than 3% of the total source. In order to balance the budget to the preliminary LGM δ13CH4 data, wildfire emissions would have to be seven times higher (greater than 10% of the total). Although this would hardly break the constraints set by δD-CH4, geological evidence does not support such high emissions from biomass burning at late glacial times (Haberle & Ledru 2001).
Secondly, to satisfy the budget, the geological emissions could be increased threefold over the values listed in table 1, to the point that they would account for 50% of the total source. Luyendyk et al. (2005) suggested that at lower glacial sea level, more gas seeps were exposed and raised geological emissions to up to twice the Modern amount. Two arguments stand against such high geological emissions. For one, they would lead to δD-CH4 values that are 25‰ more depleted in 2H than measured. For another, estimated total flux for the LGM would double, which has to be reconciled with sink rates and known atmospheric [CH4]. An additional consideration is that sea-level controlled geological emissions would also be part of the YD budget. At this time, sea-level rise had reached about 50% of the glacial transition (Fairbanks 1989), suggesting that about half the additional geological flux at the LGM must be included in the YD budget. In fact, each process that could account for 13C-enrichment at the LGM would also have to be accounted for in the YD budget and probably offset its mass balance.
Thirdly, AMP is a recently added source to methane budgets. As such, the range of carbon and hydrogen isotope values and how they could change over time is unknown. Although the δ13CH4 has been measured for some Modern plants, the δD-CH4 has not. A simple change of the AMP δ13CH4 approximately from −49.9 to −40‰ could satisfactorily rebalance the budget, but it is unclear at this point why the AMP δ13CH4 would change.
The requirement to adjust both the YD and the LGM methane budgets rules out several other possible explanations for the LGM 13C-enrichment. Schaefer et al. (2006) list five processes that would lead to 13C-enrichment of natural CH4 budgets. These are higher than expected biomass burning, geological emissions, AMP, revised δ13CH4 of tropical wetlands and the MBL sink. We are not aware of processes that would lead to substantial differences for AMP and the MBL sink between LGM and YD, beyond those that have been discussed by Schaefer & Whiticar (in press) and are incorporated in table 1. In contrast, a revised δ13CH4 of tropical wetlands, in agreement with field measurements by Quay et al. (1991) and others, would affect the LGM budget far more than the YD budget. Dällenbach et al. (2000) used changes in the interpolar gradient in atmospheric [CH4] to show that tropical sources, i.e. foremost wetlands, made up 75% of the LGM budget, but only 55% during the YD.
We therefore present alternative methane budgets for the YD and the LGM in table 3. These mass balances include more 13C-rich δ13CH4 for tropical wetlands and an increase in geological methane emissions that is twice as high for the LGM as for the YD. In addition, the relative contributions of tropical and boreal wetlands for each time period were scaled to reflect the findings of Dällenbach et al. (2000). In combination, these changes lower the LGM δ13CH4 strongly while affecting the YD value to a lesser degree. In order to counteract the effect that these changes have on the δD-CH4 mass balance, we also lowered the relative contribution of AMP in favour of the wetland source, while staying within the uncertainty ranges that follow from the work by Keppler et al. (2006) and Houweling et al. (2006).
The resulting mass balances are somewhat too enriched in 13C for the YD and 2H for the LGM, and too depleted in 13C for the LGM, slightly exceeding the analytical uncertainties (±0.3‰ for δ13CH4). The persistent difficulties in reconstructing mass balance scenarios that satisfy the isotope constraints without contradicting geological evidence or modern observations of methane dynamics indicate substantial gaps in our knowledge of the methane system. It is probable that the LGM 13C-enrichment cannot be explained with an appropriate mix of source fluxes alone, but may lie with an environmental control that has been either underestimated or not included in the reconstructions of Schaefer & Whiticar (in press). It is remarkable that the isotope records of Ferretti et al. (2005) show a strong shift in methane isotopes that is not associated with changing [CH4], while the records of Schaefer et al. (2006) and upon closer inspection also those of Sowers (2006b) show unperturbed isotope values during dramatic [CH4] changes. These findings suggest that mass balances that are solely driven by source inventories are not adequate to describe natural methane isotope dynamics. More insight may be gained from extended palaeorecords that may show correlations of δ13CH4 and/or δD-CH4 with other environmental parameters, such as global ice cover or atmospheric carbon dioxide concentration. Such findings may provide a more complete explanation for the glacial 13C-enrichment.
7. Comparison of δ13CH4 and δD-CH4 for YD–PB
As mentioned, only a few measurements of palaeoatmospheric δ13CH4 and δD-CH4 have been made using firn air and ice samples. The best comparison currently available is for the YD–PB transition, a period with a rapid rise in [CH4] from approximately 450 ppbv to over 850 ppbv. This is possible by combining the Greenland GISP2 ice core δD-CH4 data of Sowers (2006b) with the Greenland Pakitsoq ice sheet δ13CH4 data of Schaefer et al. (2006).
Figure 12 shows the times series for the GISP2 and Pakitsoq ice sheets across the YD–PB transition from ca 12.4 to 11.2 kyr BP. Pakitsoq δ18Oice, [CH4] and δ13CH4 are in figure 12a–c, respectively, and figure 12d is the GISP2 δD-CH4 data. The age scales for all plotted parameters have been adjusted to match the gas age scale of Severinghaus et al. (1998). The latter authors used tree-ring studies (Friedrich et al. 1999) to set the onset of the transition at 11 570 years BP (±10 years). The transition using the GISP2 time scale is ca 11 650 years BP. The difference emphasizes the ongoing challenge for a consistent, reliable chronology. Each data point represents approximately a 30-year interval. The greyed region in figure 12 covers the YD–PB transition (11.57–11.46 kyr BP) defined by Grachev & Severinghaus (2005), Petrenko et al. (2006) and Schaefer et al. (2006) based on the Pakitsoq δ15N, δ18Oice, and [CH4].
The δ13CH4 time series reveals two main features: (i) a 13C-enrichment relative to today of approximately 1‰ (−46 versus −47‰ Modern) and (ii) no obvious carbon isotope change across the YD–PB transition. The shift between Modern and YD–PB is explained by Schaefer et al. (2006) and Schaefer & Whiticar (in press) to be due to environmental changes that shifted both source/sink signals and strengths. The second feature means that the rise in [CH4] is either just a proportional increase in the same sources during the entire interval, or that changes in the sources and/or sinks were somehow well balanced. Similarly, the δD-CH4 data in figure 12 remain approximately constant around −95‰ up to ca 11.5 kyr BP. This is approximately 9‰ depleted in 2H relative to the present day. At 11.49 and 11.08 kyr BP, δD-CH4 drops to approximately −108‰, then −116‰, even more 2H depleted than today. Sowers (2006b) attributes the differences in the YD–PB values from other ages due to a combination of factors, similar to those we postulate for carbon (§3), i.e. C3–C4 changes, temperature dependence of the OH-abstraction KIE and, additionally, changes in δD-CH4 due to changes in δDwater.
The cross plot of δ13CH4 and δD-CH4 (figure 13) for the Schaefer et al. (2006) and Sowers (2006b) Greenland ice data for the YD reveal a most interesting feature for methane. An increase in 13C-enrichment in the methane corresponds to an increase in 2H-enrichment. There are also chronological sequences embedded in this δ13CH4–δD-CH4 relationship. From 12.15 to 11.83 kyr BP (segment 3 in figure 12, squares in figure 13) methane becomes enriched in 2H and 13C isotopes. This is during relatively isothermal conditions (figure 12a), and with no significant changes in [CH4]. From 11.80 to 11.55 kyr BP (segment 2 in figure 12; circles in figure 13), the direction reverses (depletion in 2H and 13C isotopes), essentially on the same δ13CH4–δD-CH4 trajectory as segment 1. According to the findings of Severinghaus et al. (1998), this decades long time period when [CH4] was stable before the end-of-YD rise corresponds to a time of rapid temperature increase.
At 11.52 kyr BP, the trend reverses back to 2H and 13C enrichment, after the GISP2 temperature maximum has been reached and is starting to cool. By the end of the sequence at 11.49 and 11.08 kyr BP i.e. earliest PB, (segment 1 in figure 12, diamonds in figure 13) the δ13CH4–δD-CH4 relationship observed in the YD breaks down. These two PB data points have δD-CH4 values approximately 15–30‰ lighter than those in the YD trend.
Although speculative, a possible explanation for the δ13CH4–δD-CH4 relationship of segments 1 and 2 is that the magnitude of 2H and 13C-rich methane source fluxes, such as coal-derived or thermogenic methane, or biomass burning are changing with time. For example, using the simple linear mixing equation (equation (2.1)) and varying only the geological flux in table 1 between 20 and 45 Tg yr−1 for the different segments, the calculated atmospheric δ13CH4–δD-CH4 values shift from −46.8‰, −100‰ to −45.2‰, −85‰, respectively. For a more rigorous treatment, the total source fluxes need to be adjusted, but the example points out how fluctuating thermogenic gas fluxes alone could account for the δ13CH4–δD-CH4 relationship. If the decrease in the thermogenic source is related to the synchronous rapid rise in temperature, in this example it is unclear and we do not know of any mechanism to explain a causal relationship.
In the YD–PB transition period (segment 1, greyed in figure 12), the correspondence to thermogenic gas is not observed. In particular the δD-CH4 values are substantially more 2H-depleted than in the other segments, in fact more than in the entire Sowers (2006b) database. The youngest δ13CH4 data from Schaefer et al. (2006) at 11.36 kyr BP was used to pair with the 11.08 kyr BP data point of Sowers (2006b) in figure 13. Even if this δ13CH4 value were incorrect, then it would not alter the fact that the PB δD-CH4 values of segment 1 are substantially more depleted in 2H than in the YD.
The δ13CH4–δD-CH4 signatures in segments 1 to 3 show that a massive Archaeal hydrate release, i.e. ‘Clathrate Gun Hypothesis’ as proposed by Nisbet (1990, 2002) and Kennett et al. (2002) is not responsible for the methane increases at the YD–PB transition period. This non-hydrate source for the methane rise is consistent with the interpretation of Schaefer et al. (2006) and Sowers (2006b), as well as other lines of evidence (Chappellaz et al. 1997; Maslin & Thomas 2003; Maslin et al. 2004). It is feasible that a gradual, non-catastrophic, release of thermogenic hydrate occurred during segment 3, but not for segments 1 or 2. This release of thermogenic gas could also be conventional natural gas seepage, however, it is not clear why this source would vacillate over relatively short geological time periods.
Insights into today's rapid rise in tropospheric methane can be gained by examining analogous increases in the past. The key to this is correctly assessing the magnitude of the methane emission sources and sinks. Stable carbon and hydrogen isotopes of tropospheric methane, as recorded in ice samples, provide a useful constraint on the budget estimations. We are able to incorporate anticipated changes in palaeoenvironmental conditions that translate into changes in the isotope signals of methane precursors, to estimate past global δ13CH4 and δD-CH4 values. The overall δ13CH4 and δD-CH4 variations are relatively small and require careful analytical precision and accuracy to delineate isotopic changes to the budgets. However, such changes between today and the PIH, YD–PB and LGM are discernable. The most obvious is that the palaeomethane is consistently enriched in 13C- and 2H relative to the Modern by approximately 1 to 5‰ for δ13CH4 and 9‰ for δD-CH4. This is a somewhat unexpected result, compared with a priori methane budgets that reflect the Modern anthropogenic releases of methane from fossil fuels and biomass burning. These global δ13CH4–δD-CH4 constraints enable us to more faithfully reconstruct the methane emission and sink profiles. Our initial goal was not to fully, and perhaps artificially, balance the budgets, rather, we used currently available best estimates of the flux intensities and the isotope signatures of the sources and sinks to compare with the measured values over the Modern, PIH, YD and LGM time periods. Our attempt to balance the LGM budget using reasonable changes to flux intensities and δ13CH4 and δD-CH4 values was still unsatisfactory. This points out the discrepancies and uncertainties in atmospheric methane budget calculations and the need for further attention.
Close examination of the relationships between tropospheric methane mixing ratio and temperature reveal strong, but changing correlations for different time periods in the latest Pleistocene. Detailed inspection of the YD δ13CH4–δD-CH4 records show a remarkable relationship between them from 12.15 to 11.52 kyr BP. Based on the isotope mass balances, a plausible explanation is the waxing and waning of thermogenic emissions. The tight δ13CH4–δD-CH4 relationship appears to breakdown during the YD–PB transition, suggesting that factors other than thermogenic natural gas are the cause. In both age cases, catastrophic releases of hydrates with Archaeal isotope signatures can be ruled out. Thermogenic clathrate releases are possible for the YD period, but so are conventional natural gas seepages.
We thank Ed Brook for valuable discussions. We also thank Todd Sowers and an anonymous reviewer for very helpful comments. This work was supported by a fellowship from DAAD (German Academic Exchange Service, H.S.), Petroleum Research Fund of the American Chemical Society (H.S.), Canadian Foundation for Climate and Atmospheric Sciences CFCAS MAMMOTH grant (M.J.W.) and NSERC Discovery grant (M.J.W.).
↵† Present address: Department of Geosciences, Oregon State University, 104 Wilkinson Hall, Corvallis, OR 97331, USA.
One contribution of 18 to a Discussion Meeting Issue ‘Trace gas biogeochemistry and global change’.
- © 2007 The Royal Society