SUPPLEMENTARY INFORMATION DOI: 10.1038/NGEO1922 Independent variations of CH4 emissions and isotopic composition over the past 160,000 years Authors: Lars Möller, Todd Sowers, Michael Bock, Renato Spahni, Melanie Behrens, Jochen Schmitt, 5 Heinrich Miller, Hubertus Fischer* *Correspondence and requests for materials should be addressed to H.F. 2. Supplementary Methods and technical descriptions 10 2.1 δ 13CH4 measurements performed at Alfred Wegener Institute (AWI) EDML δ13CH4 measurements were performed using a continuous flow gas chromatography combustion isotope ratio mass spectrometry system (GC/C/IRMS) with a preceeding purge and trap extraction and pre-concentration setup. A few modifications have been made compared to the published experimental and data processing procedure1,2. Sample sizes between 150-200 g of ice, 15 equivalent to 15-20 ml of air at standard temperature and pressure (STP) were used after scraping off 1-2 mm at the surface with a microtome knife to minimize the risk of contamination e.g. by drill fluid. Sealed with a copper gasket in an ultra-high-vacuum stainless steel to glass sample vessel with an inner volume of 350 ml, the samples were melted after evacuation and the air stripped out of the water and vessel head space with a helium carrier gas stream (150 mL/min) for 90 minutes. 20 The bulk of the entrained water vapor was removed by a cooled Nafion membrane and most of the easily condensable gases like CO2 and N2O are trapped in a ⅛ inch steel trap at -196 °C (liquid nitrogen, LN2). While methane from the sample air was subsequently pre-concentrated on a ⅛” steel tubing filled with an adsorbent (Hayesep D 80/100 mesh, Supelco), bulk components of the air 1 NATURE GEOSCIENCE | www.nature.com/naturegeoscience © 2013 Macmillan Publishers Limited. All rights reserved. were vented at -140 °C. The retained sample methane and residual air components were focussed 25 on a cryofocus trap and separated on a 30 m GC column (30 °C, CarbonPLOT, Agilent, Böblingen, Germany). The methane was then quantitatively combusted to CO2 in a 940 °C oxidation furnace (Ni/CU/Pt wires in Al2O3 tubing, ThermoFinnigan, Bremen, Germany) and admitted to the IRMS (Isoprime, Elementar, Germany) using an open split. 30 A pure CO2 working standard (ISO-TOP, Air Liquide, Germany) was admitted to the IRMS during each acquisition. Moreover, a pure methane standard (99.995 vol.-% purity; AirLiquide, Germany) admitted to the GC stream was used to monitor fractionation in the GC or the combustion oven. 20 ml STP synthetic atmospheric air standard injections from two different standard bottles (“SynthAir” with 1000 ppb CH4 (99.995 vol.-% purity), 250 ppm ISOTOP CO2 and 250 ppb N2O 35 (99.999 vol.-% purity); “Crystal-Mix” with 1081 ± 22 ppb CH4 4.5, 249.9 ± 5.0 ppm CO2 and 259 ± 26 ppb; both AirLiquide, Germany) were used to correct for long-term systematic shifts and internal calibration. The mean δ13CH4 value for “SynthAir” was -40.93 ± 0.10 ‰ (n=448) and -49.66 ± 0.11 ‰ (n=136) for “Crystal-Mix” after drift correction. The absolute accuracy of the system was determined by 16 measurements of a modern air sample from Neumayer Station, Antarctica. The 40 target value (δ13CH4: -46.97 ± 0.04 ‰ versus V-PDB (n=7)) was determined by off-line sample preparation and dual-inlet IRMS on a MAT252 mass spectrometer (ThermoFinnigan) at the Institut für Umweltphysik, Heidelberg3. A calculated machine offset of 0.02 ‰ of our GC/C/IRMS determined values after correction for Kr interference (see below) relative to the Heidelberg reference value was used to correct the EDML δ13CH4 data of this study. Accordingly, our data are 45 reported relative to the V-PDB scale as defined by the Heidelberg measurements4. The EDML δ13CH4 time series includes 129 data points and 32 replicates from different depth intervals with a mean reproducibility of the replicates of ± 0.18 ‰. A total of 19 outliers, either caused by machine instabilities or other experimental problems, were excluded from the dataset and © 2013 Macmillan Publishers Limited. All rights reserved. 50 remeasured. Processing of the chromatogram for each measurement has been performed using a self developed, fully automated peak detection, integration and referencing script written in the Python programming language (www.python.org). This script allows uniform and comprehensible background and peak detection, genuine automation for post-processing (like e.g. long term trend corrections) and data archiving. Using this script we were able to reproduce published δ13CH4 data 55 for termination I2, which were obtained with the software provided with the MS (Isoprime, Elementar, Germany) with an offset of 0.10 ‰ (1σ: 0.17€‰, n=34) without and 0.17 ‰ (1σ: 0.15€‰) with correction for Kr interference (see below). Note that both offsets are smaller than the measurement uncertainty of this earlier data set (0.3 ‰, 1σ). This re-evaluated record for termination I was used in Figure 1 and 2 in the main text, and is also shown in supplementary 60 Figure S1 for the comparison with the unmodified, uncorrected data of our previous publication. 2.2 δ 13CH4 measurements performed at Pennsylvania State University (PSU) For the Vostok ice core sample measurements at Penn State University, two different extraction methods were employed. A dry extraction system5 was used to liberate gas from ice samples where N2O measurements were needed6,7. Trapped gases were liberated from 1-1.5 kg ice samples using a 65 “cheese grater” oscillator with immediate cryogenic freezing of liberated air into a 35 ml stainless steel sample tube immersed in liquid Helium. Once the air was transferred, the sample tube was isolated, removed from the Helium Dewar and equilibrated at room temperature before CH4 and N2O analyses were performed using standard GC techniques. The sample tube was then attached to the PreCon device where the isotopic composition of N2O and CH4 were determined8,9. For those 70 samples where we did not need N2O data, a wet extraction technique was used to liberate the occluded air. Ice samples weighing 500-700 g were placed into a stainless steel extraction cylinder and sealed with a copper gasket. After evacuation, the ice was allowed to melt for 40 min in 50 °C water before being placed into a liquid nitrogen Dewar for 40 min to refreeze the meltwater. The headspace was then flushed with He (40 ml/min) through a H2O trap (-110 °C) with the CH4 © 2013 Macmillan Publishers Limited. All rights reserved. 75 ultimately trapped on a Hayesep D trap at -130 °C. After 40 min of flushing, the Hayesep D trap was isolated and attached to the PreCon for CH4 isotopic analyses10. Both extraction systems were routinely checked for contamination/fractionation using standard air samples of varying sizes to mimic the amount of CH4 we extract from the ice core samples (1-3 nmol of CH4). The air standards were introduced over the residual crushed ice or degassed refrozen meltwater and 80 processed as if they were a real ice sample. For the dry extraction runs, the average of eight separate standard runs was -47.30 ± 0.31 ‰ (1σ). The average value is 0.17 ‰ lower than the assigned δ13CH4 value for the air tank (-47.13 ‰ VPDB) but within the uncertainty associated with the method itself. For the wet extraction system, we processed 15 standard air samples through degassed water samples. The average δ13CH4 value for these runs was -47.10 ± 0.34 ‰. These 85 results are close to the assigned value for the standard (-47.13 ‰) and 0.2 ‰ higher than the results from the dry extraction system. To account for the δ13CH4 difference between the two extraction systems, we add 0.2 ‰ to all the dry extraction data to be consistent with the wet extraction data and the assigned value for the air standard. We estimate the overall uncertainty based on the replicate analyses of the standard air samples to be 0.3 ‰ (1σ). 90 2.3 Correction of δ 13CH4 data due to Kr interference during IRMS measurements The chromatographic separation of CH4 and the noble gas krypton (Kr) imposes special demands on a setup used to separate air components, owing to the very similar physico-chemical properties of these compounds. Kr has previously not been accounted for in δ13CH4 studies, as none of the multiple stable isotopes of Kr are close to the mass to charge ratio (m/z) 44, 45 and 46 considered in 95 CO2 based δ13CH4 measurements. Recently Schmitt et al. (2013) demonstrated11, that the doubly charged isotope 86Kr2+ does in fact interfere with the δ13C measurement of CH4, if Kr enters the ion source of the IRMS. After a thorough review of the raw data for the ice core δ13CH4 time series, we were able to identify irregularities in the raw chromatograms at the peak flanks of the CH4-derived CO2 for the © 2013 Macmillan Publishers Limited. All rights reserved. 100 measurements performed with the AWI instrument. The Kr peak causes anomalies in the ratio of m/z = 45 to m/z = 44 and the m/z = 46 to m/z = 44 ratio which generates higher δ13CH4 values the more Kr contributes to the total peak areas (hereafter referred to as Kr effect). While the symptoms are distinct due to instrumental differences of the two setups, we also found the PSU system to be affected by the Kr interference. 105 Because polar ice core samples are not easily replaced, we were unable to repeat the measurements of the EDML and Vostok δ13CH4 time series with enhanced instrumental setups. Instead we corrected the δ13CH4 values for this Kr effect as described in detail in Schmitt et al. (2013)11, which we briefly summarize below. 110 A posteriori correction of the Kr effect We applied the following strategies to correct PSU and AWI δ13CH4 measurements for the Kr effect and account for relative laboratory offsets with respect to the VPDB scale. AWI EDML measurements are corrected for Kr individually by a method applied to the respective raw data chromatograms. The method uses the visible anomalies seen in the m/z ratios 45/44 and m/z 46/44 115 and subtracts the derived interference from the raw data time series11. Afterwards, the chromatograms are reprocessed and the isotopic composition of the CH4 peak calculated. We refer to these values as the Kr-corrected δ13CH4 values. In contrast, δ13CH4 values obtained without the subtraction algorithm are referred to as original δ13CH4 values. The difference between original and corrected values are termed Kr correction values or Δδ13CKr. For the EDML time series Δδ13CKr 120 range between 0.4 ‰ for interstadial (medium CH4 mixing ratios) to 0.8 ‰ during glacial (MIS 2) and stadial conditions (lower CH4 levels) (compare also Supplementary Figure S1). The results are further calibrated internally and are tied to the VPDB scale as outlined in section 1.1. Unfortunately, an analogous direct approach could not be applied to the Vostok ice core δ13CH4 acquisitions performed at the PSU, as the raw chromatograms were not stored after processing the 125 δ13C data. Instead, we had to choose an indirect way to correct for the Kr effect in the Vostok © 2013 Macmillan Publishers Limited. All rights reserved. δ13CH4 time series. For the purpose of this correction the atmospheric krypton mixing ratio can safely be considered constant over time11,12. If instrumental conditions are uniform, the Kr effect scales only with the atmospheric concentration of methane, it is thus directly proportional to the Kr/CH4 ratio11. We 130 inferred the linear relationship between Kr effect and 1/CH4 from a series of air samples with differing methane mixing ratios but similar Kr levels. The three ambient air samples were retrieved at Niwot Ridge preserve, Colorado, US in 2007, and were part of the „2007 - IPY International Ice Core Gas Intercalibration Exercise“ launched by Todd Sowers (Penn State University) and Ed Brook (Oregon State University). The samples of recent ambient air were diluted with ultra pure air 135 (free of CH4) in order to simulate the full spectrum of atmospheric variability from present day, preindustrial to glacial conditions. Note, that the dilution did not affect the noble gas concentration in the cylinders. The raw data chromatograms of “IPY” air samples performed with the PSU instrument were treated in the same manner as the EDML ice core samples, with the routine adapted to the specific 140 characteristics of the PSU setup. Based on these PSU measurements we obtained Kr correction values of 0.12 ‰, 0.26 ‰ and 0.66 ‰ for cylinder “CA03560”, ”CC71560” and “CA01179”, respectively. To a first approximation, Δδ13CKr scales with the inverse of the CH4 mixing ratios of 1852 ppb, 906 ppb and 365 ppb (illustrated in Figure S2), yielding a dependence of Δδ13CKr to CH4 for the PSU measurements. Supplementary table 1 provides a detailed compilation of results for the 145 “IPY” air samples for both the PSU and the AWI instruments. When the linear Δδ13CKr - 1/CH4 relationship was applied to the CH4 mixing ratios13 of the Vostok ice core, we derived Δδ13CKr values in the range of 0.4 ‰ for interglacial conditions (e.g. MIS 5.5) to 0.7 ‰ and 0.8 ‰ for the glacial stages MIS 2 and MIS 6, respectively (see Figure S1). Note that 150 while these corrections on the PSU and AWI data of several tenth of a permille are significant, they are still small compared to the atmospheric changes of several permille observed in our ice core © 2013 Macmillan Publishers Limited. All rights reserved. data on glacial/interglacial time scales, and do not influence our interpretations. 2.4 AWI and PSU laboratory inter-calibration In order to minimize offsets between the absolute standardization of both laboratory setups at the 155 AWI and the PSU, we performed inter-calibration measurements on three IPY air samples and on ice core samples from the WAIS Divide ice core WDC05 A (79°27.70S 112°7.510W; 1.759 masl.). We applied the CH4 dependence of Δδ13CKr to account for the Kr effect encountered in the ice core and air sample measurements using the PSU instrument, and the raw data correction procedure described for the EDML ice samples in section 1.3 for the IPY air and WDC05 A measurements 160 performed with the AWI setup. The results for both systems, presented in more detail in Supplementary table 1 and 2, lead to an AWI-PSU laboratory offset of 0.09 ‰ with respect to δ13CH4. As a final adjustment, the Vostok δ13CH4 time series is hence shifted by 0.09 ‰ towards lower values to account for the differences in the absolute standardization of both laboratories. Note that this inter-laboratory offset is significantly smaller than the measurement uncertainty of 0.3 ‰, 165 showing that the AWI and PSU data sets are fully compatible after correction of the Kr effect. 2.5 Firn column corrections applied to the δ 13CH4 ice core records During the enclosure of air bubbles in the Antarctic ice sheet, the methane molecules like other air components encounter diffusion processes in the open pore space of the firn column, which also affect their isotopic composition archived after bubble close-off. In the diffusive firn zone14 170 methane is subject to gravitational settling, that enriches the heavier isotope at the bottom of the firn column15. In addition, strong concentration gradients caused by rapid atmospheric methane concentration changes, induce diffusive fluxes that lead to isotopic fractionation16. Finally, thermal diffusion corrections are required when large temperature gradients exist in the firn layer17. Temperature variations at Kohnen Station (EDML core site) and Vostok have been slow during the 175 last glacial cycle. Hence, the firn column down to the close-off depth was essentially in thermal equilibrium and thermal diffusion effects are negligible for the datasets presented here. © 2013 Macmillan Publishers Limited. All rights reserved. In order to quantify the order of magnitude of the respective diffusion effects, we used a firn diffusion model18 with a parameter set adapted to the glacial EDML core characteristics to calculate the combined effects of diffusive fractionation due to gravitation and concentration changes at the 180 surface (see Figure S3). A methane pulse of 200 ppb with an initial rate of increase of 4 ppb/yr and a respective decline of 0.25 ppb/yr (grey line, subfigure a)) was prescribed to mimic the most vigorous natural methane rises throughout the glacial cycle (e.g. at the end of terminations or into DO 21). The δ13CH4 signature of methane at the surface is prescribed at -45 ‰. Accordingly, all changes to this value recorded at the model bubble close-off on the bottom of the firn column (blue 185 line, subfigure b)) are solely due to the two diffusion processes. To account for uncertainties concerning the two most important physical parameters of firnification, site temperature and accumulation rate, we illustrate the range of model results for the fractionation of δ13CH4 with a minimal (-52.14 °C, 2.978 cm water equivalent per year (w.e./yr)), maximal (-46.64 °C, 5.075 cm w.e./yr) and a best guess scenario (-49.52 °C, 3.859 cm w.e./yr). This range of site temperatures 190 and snow accumulation rates is a realistic representation of possible glacial conditions at the core site. The initial δ13CH4 value, after steady state is reached in the model, is -44.57 ‰. This offset of 0.43 ‰ relative to the atmospheric value is due to the gravitational settling that is established in the 195 firn column. It is also on very good agreement with measured δ15N2 for the EDML core for glacial conditions19. The consecutive prescribed methane rise at the surface causes an additional shift in δ13CH4 after bubble close-off in the range of -0.73 ‰ for coldest temperatures and lowest snow accumulation and up to -1.05 ‰ for maximal temperature and accumulation rate, while the best guess scenario amounts to -0.85 ‰. However, the effect is short-lived and decreases to levels below 200 our experimental uncertainty in less than 150 years after the initial methane increase. After about 500 years the δ13CH4 value is essentially back at its starting value before the CH4 increase. During the slower decline of methane concentration back to the base level of 350 ppb, the highest observed © 2013 Macmillan Publishers Limited. All rights reserved. diffusional fractionation of -0.24 ‰ does not even exceed the measurement uncertainty. In conclusion, only those data points of our record may be affected that fall within the relatively short 205 time window during major methane concentration increases. Accordingly, we did not correct our data for these diffusion effects, but have to keep in mind that individual samples that coincide with the rapid CH4 changes may be biased by a few tenth of a permille towards lower (more negative) δ13CH4 values. 210 Accordingly, all our δ13CH4 data are solely corrected for gravitational fractionation. Vostok samples have been corrected with interpolated δ15N2 data according to published procedures14. No δ15N2 record covering the whole time interval of our δ13CH4 data is available for the EDML ice core to date. However, δ15N2 data over the last glacial/interglacial transition vary only between 0.4 to 0.45 ‰. Thus, we shifted EDML values by a constant offset of 0.41 ‰ to higher values. This 215 0.41 ‰ shift does reflect expected values for glacial conditions very well19,20, and is also in line with model studies21. The error introduced by this constant correction is 0.05 ‰ at the most, and thus negligible compared to our overall analytical uncertainty of 0.3 ‰. 2.6 dD(CH4) measurements performed at University of Bern δD(CH4) measurements were performed using a purge and trap extraction coupled to a gas 220 chromatography pyrolysis isotope ratio mass spectrometer (GC/P/IRMS) system as described in detail in Bock et al. (2010)22 with some improvements which will be published elsewhere. External precision of the presented data is about 2.5 ‰ (1σ) based on standard air measurements of corresponding size. In Figure 3 of the main text the error bars represent the standard deviation of standard air measurements (1σ: 1.8 ‰ to 2.9 ‰) used to calibrate the corresponding samples. All 225 δD(CH4) values are given with respect to the international Vienna Standard Mean Ocean Water (VSMOW) scale. No corrections (e.g. gravitational enrichment) have been applied, as these corrections would only be of minor importance regarding the analytical uncertainty, and do not © 2013 Macmillan Publishers Limited. All rights reserved. affect our conclusions. Note, that no Kr interference occurs for our δD(CH4) system. 2.7 Age scales 230 Ice core records If not stated otherwise, all gas ages in this document are reported according to a chronology based on the new synchronization effort for EPICA and various other ice core chronologies by LemieuxDudon et al. (2010)23, hereafter denoted as “Unified” age scale. Where applicable, a direct age calculation was performed by linear interpolation of the depth-age relationship provided by 235 Lemieux-Dudon et al. (2010)23. Other records required additional conversion steps according to the procedures described below. Vostok δ 13CH4 and CO2 records For Vostok we adopted the published Vostok depth to EDC3 gas age relationship of the Vostok 240 CO2 record and interpolated corresponding EDC3 ages for the Vostok δ13CH4 data. However, we observed an evident misalignment of the fast methane concentration rises at Dansgaard/Oeschger (DO) event 24 compared to EDC CH4 data26. As part of the focus of our work is directed at relative timings of δ13CH4 and CH4 rises, we performed a manual methane synchronization to account for this offset. Therefore, we picked five tie-points in between DO event 21 and the Termination 2 245 methane rise at peak flanks (Supplementary table 3) and interpolated the included data points linearly to the EDC3 age scale27. The EDC3 ages including the adjusted section between 83.6 kyr BP and 128.9 kyr BP were then converted to the target “Unified” age model. The adjustments were also applied to the Vostok CO2 record used in this work26,28. Note that the applied adjustments do not affect the conclusions in this publication on the decoupling of CH4 and δ13CH4 or on the 250 coupling of δ13CH4 and atmospheric CO2. Byrd atmospheric CO2 © 2013 Macmillan Publishers Limited. All rights reserved. No official age conversion to the “Unified” age scale exists for the Byrd ice core. Hence, we performed another methane synchronization between the Byrd29 and EDML30 CH4 records in order 255 to obtain corresponding EDML depths. Those were then converted to “Unified” ages. We are aware of the limitations of the methane synchronization approach31, especially if the resolution differences of the compared CH4 datasets are large, or at times were CH4 variations are low and, thus, tie-points are scarce. However, the very good temporal agreement of the Byrd32 and the EDML CO2 data33 show that the relative CH4 synchronization error is small and does not affect the conclusions in our 260 paper. The list of manual tie-points applied in the synchronization are presented in Supplementary table 4. Relative sea-level No direct age conversion to the “Unified” target age scale could be applied to the relative sea-level 265 (rsl.) data without invoking an ad-hoc phase relationship between the sea level and the ice core records. Therefore, we used the sea level record of Rohling et al. (2009)24 on the speleothem synchronized age-scale provided by Grant et al. (2012)25 without any modifications. 3. Complementary information on past δ 13CH4 changes 270 3.1 DO methane variability and its missing responses in δ 13CH4 Figure S4 illustrates the missing imprint of the rapid atmospheric methane concentration variability on the carbon isotopic signature of methane in conjunction with the six strongest DO warmings throughout the last glacial. Segments of the CH430 and δ13CH4 data sets, centered around the respective methane rise, have been aligned (Figure S4 a)) in order to study the phasing and 275 individual timing of the concentration changes and its counterpart in δ13CH4 (Figure S4 b)). Note that the data points in δ13CH4 closest to the most vigorous methane increases are likely biased by diffusional fractionation in the firn column (see discussion above). This may lead to offsets of δ13CH4 data points close to major CH4 changes. For example, the two negative excursions in © 2013 Macmillan Publishers Limited. All rights reserved. δ13CH4 of less than 1 ‰ at DO 8 and termination I are located so close to the corresponding 280 methane increase, that they can be attributed to the diffusional fractionation effect. We mark the range of possibly affected values in Figure S4 with a box of the width of 150 yr (gray bar), according to the maximum duration in our firnification model exercise, where the effect of the diffusional fractionation exceeded our experimental uncertainty range of 0.3 ‰. We also show the two major methane increases of termination I and II (right pair of panels in Figure S4). There is no 285 apparent imprint of rapid methane variability on its carbon isotopic signature over the respective DO events and deglaciations. 3.2 Proxy evidence for C4 plant expansion Methane production under natural conditions involves the decomposition of organic precursor material that has previously been accumulated by plants through photosynthetic sequestration of 290 CO2. Owing to fundamental physiological differences of the two major photosynthetic pathways, which characteristically discriminate the heavy isotope ¹³C during carbon assimilation, the typical ranges of isotopic signatures imposed on the plant material differ considerably34,35. 85–90 % of terrestrial plant species today, covering the whole spectrum of vegetation from grasses and herbs to shrubs and trees, follow the C3 photosynthetic pathway36. C3 plant biomass is characterized by a 295 depleted δ13C signature (-32 ‰ - -22 ‰), caused by the lower reactivity of 13CO2 with the primary carboxylating enzyme RUBISCO37. C4 vegetation on the other hand, mostly grasses and sedges, are able to pre-concentrate CO2 internally at the cost of reduced quantum yield38. As a consequence, C4 plant carbon fixation fractionates less against ¹³C (~-16 ‰ - -9 ‰). The isotopic composition of the terrestrial biosphere, i.a. the pre-cursor biomass for methanogenesis, is controlled by the primary 300 productivity of an assemblage of plants under local growth conditions, the individual adaptation of its members to this conditions, as well as its tolerance against limitation factors. Plants of both photosynthetic pathways are unequally tolerant to limitations in CO2, light intensity, local temperature, and to moisture and nutrient availability35,39,40. Seasonality of precipitation has an equally significant impact on the local balance between C3 and C4 vegetation40. However, the © 2013 Macmillan Publishers Limited. All rights reserved. 305 relative importance of any of the limitations, especially in terms of a competitive advantage of plant families against others in the struggle for habitats, remains an unresolved and vitally discussed question36,41-45. Accordingly, it is neither physiologically well constrained how strong a C3 to C4 plant shift might have been in tropical regions under generally colder, drier conditions, and low CO2 levels that were 310 characteristic for the glacial period, nor is it extensively documented by the scarce terrestrial proxy evidence from these areas. Such a shift, however, is one of the relevant processes of our hypothesis to explain the observed, very pronounced δ13CH4 changes. Therefore, we will now discuss some of the ecosystem evidence that is available and relevant for our hypothesis. 315 In temperate regions in Northern- and Meso-America46 and the Chinese loess plateau47-49, growing season temperature and the local climatic constellation seems to out-compete the physiological effect of low CO2 level as predominant control upon the C3/C4 ratio. With warm growing seasons in the tropics, however, water insufficiency and low CO2 possess an increased influence as plantgrowth limitation factors and pose high adaptive pressure on prevalent ecosystems50-53. 320 Recent vegetation model experiments indicate high vegetation sensitivity to low atmospheric CO2 levels during glacial periods44,45,54,55. Globally, simulations for glacial climate conditions and typical CO2 concentrations lead to significant retractions of closed canopy forest habitats in favor of open vegetation types43,56,57. Tropical rainforests seem especially affected by the combined effects 325 of increased aridity and low CO2, and large arboreal areas are replaced by open savanna- and shrublike vegetation. Analogue findings are well documented by terrestrial proxy data from tropical regions in Africa58-61, Meso- and South America61-64. These studies also report large proportions of C4 vegetation 330 contributing to the widespread grasslands in equatorial Africa and South America in the LGM. © 2013 Macmillan Publishers Limited. All rights reserved. Glacial-Interglacial differences in δ13C of vascular plant waxes from sediment cores off the East Atlantic coast close to the river mouths of the Congo and Angola basin, indicate 3-4 ‰ shifts towards higher δ13C values and thus, relative increases of C4 abundance in the range of 20-40 %59. Increased C4 contribution has also been inferred from another marine core retrieved at the Guinea 335 Plateau margin recording Sahara/Sahel vegetation60. It also indicates raised aridity, falling temperatures and exceptionally high C4 predominance for the period between 71 and 65 kyr BP, i.e. the MIS5/4 transition, which is also characterized by a strong increase (+4 ‰) in our δ13CH4 record. A comparable study from the Cariaco Basin in the tropical west Atlantic, reported a 4-5 ‰ δ13C 340 decrease in leaf waxes from the LGM to the Preboreal Holocene62. This probably documents a reoccupation of forest vegetation in the peripheral Amazonian lowlands, that potentially retracted under glacial conditions65-69. Moreover, huge land masses the size of Europe from South Thailand to Sumatra, Borneo and Java became exposed in South-East Asia when sea-level fell in glacial periods. This territory, known as Sundaland, was also vastly covered by savanna type vegetation 345 with considerable C4 contribution70,71. A global shift of C3 to C4 plants may not be representative for the conditions encountered in permanent (tropical) wetlands. Intuitionally, one may expect that the missing water limitation in such a wetland would reduce the adaptive pressure on plants in that ecosystem and hence level 350 competitive advantages of one species against another. C3 plants, for example, are not forced into the natural trade-off between necessary stomatal opening for carbon sequestration and excessive water loss. In this light it is yet not fully understood, why large modern wetland ecosystems in tropical East Africa72, South Africa73,74 or areas of the Amazon floodplain75 exhibit clear C4 plant predominance (mostly papyrus). Moreover, there is evidence that C4 dominance in east-equatorial 355 Africa (at least near lake Challa) persisted during both wet and dry phases under glacial conditions76. For the δ13CH4 change observed in our record we speculate that seasonally inundated © 2013 Macmillan Publishers Limited. All rights reserved. wetlands played an increasing role for CH4 emissions during cold climate conditions. These tropical non-permanent wetlands should foster the shift to open grasslands with high C4 contribution, as the higher water use efficiency and productivity of the C4 plants under low CO2 levels in glacial 360 periods should proof advantageous in this environment of seasonally contrasting very dry to very wet conditions. 3.3 Impact of location and habitat on the isotopic signature of C3 plants Environmental factors as humidity, light availability, CO2 concentration and use of recycled CO2 differ largely between C3 plant habitats like rainforest and savanna77. Dense closed canopy 365 rainforest habitats provide a higher degree of natural protection from wind movement and hence reduced water loss and air mass exchange, but on the other hand increase the competition for light intensity. The low photon flux caused by the light deficit for example diminishes carbon fixation rates, especially in the undergrowth vegetation. Lower vegetation layers furthermore assimilate more respired CO2 that already underwent fractionation78,79. Rainforest habitats have much higher 370 water supply and air moisture levels compared to open vegetation types. Reduced evapotranspirative water loss in closed canopies allows longer periods of opened leaf stomata and, thus, increased CO2 levels in the leafs. Light deficit and high intracellular CO2 levels enhance the discrimination of the heavier 13C isotope during photosynthesis80. In contrast, open shrub, herb and grassy C3 vegetation is, especially in low latitudes, exposed to high levels of direct sunlight and 375 high leaf temperatures. To avoid extensive water loss, the stomatal conductance of these plants is usually highly restricted37, and carbon dioxide limitation in the leaf cells reduces the relative discrimination of the heavier isotope by the enzymes involved in photosynthesis. As a consequence of these effects, C3 rainforest plant material is found to be 3-4 ‰ more depleted than C3 plant material from open savanna77. 380 © 2013 Macmillan Publishers Limited. All rights reserved. 3.4 Role of water table changes Hydrological changes in the past have direct implications on water table heights in wetlands. The soil oxidative layer thickness directly affects the net/gross primary productivity and thus CH4 emission strength of a given wetland system. But an increased/decreased oxidative layer thickness 385 would also induce an enrichment/depletion in the mean δ13CH4 signature of emitted methane as a higher/lower proportion of methane is oxidized on its way to the surface. This parameter is highly dependent on the local hydrology, topography and soil characteristics and poorly constrained spatially and over time. Our record suggests that it is of minor importance for the millenial scale variability of methane over the DO cycles as there is no remarkable imprint on δ13CH4. A general 390 contribution to the glacial-interglacial difference in δ13CH4 due to enhanced global aridity in the course of the glaciation, however, cannot be ruled out. © 2013 Macmillan Publishers Limited. All rights reserved. References 1. Behrens, M.; Schmitt, J.; Richter, K.-U.; Bock, M.; Richter, U. C.; Levin, I. & Fischer, H. A 395 gas chromatography/combustion/isotope ratio mass spectrometry system for high-precision δ13C measurements of atmospheric methane extracted from ice core samples. Rapid Commun Mass Spectrom, 22, 3261-3269 (2008). 2. Fischer, H.; Behrens, M.; Bock, M.; Richter, U.; Schmitt, J.; Loulergue, L.; Chappellaz, J.; Spahni, R.; Blunier, T.; Leuenberger, M. & Stocker, T. F. Changing boreal methane sources 400 and constant biomass burning during the last termination. Nature, 452, 864-867 (2008). 3. Poß, C. Analysis of the variability of the atmospheric methane budget in high polar regions with a regional trajectory model and using measurements of stable isotopes. PhD thesis, University of Heidelberg, Germany, (2003). 4. Calibration of the Heidelberg δ13CH4 measurements was performed using pure CO2 IAEA 405 Standard Reference Materials (RM 8562: δ13C = -3.72 ‰, RM 8564: δ13C = -10.45 ‰, RM 8563: δ13C = -41.59 ‰5). The absolute agreement of the Heidelberg δ13CH4 measurements on the VPDB CO2 scale is estimated to be better than ±0.1 ‰. An intercomparison with three air samples from the southern hemisphere between Heidelberg and NIWA yielded a difference NIWA-Heidelberg δ13CH4 = 0.04 ± 0.04 ‰3. 410 5. Sowers, T. & Jubenville, J. A modified extraction technique for liberating occluded gases from ice cores. J. Geophys. Res., 105, 29155-29164 (2000). 6. Sowers, T.; Alley, R. B. & Jubenville, J. Ice Core Records of Atmospheric N2O Covering the Last 106,000 Years. Science, 301, 945-948 (2003). 7. Miteva, V.; Sowers, T. & Brenchley, J. Production of N2O by Ammonia Oxidizing Bacteria at 415 Subfreezing Temperatures as a Model for Assessing the N2O Anomalies in the Vostok Ice Core. Geomicrobiology Journal, 24, 451-459 (2007). 8. Sowers, T.; Rodebaugh, A.; Yoshida, N. & Toyoda, S. Extending records of the isotopic composition of atmospheric N2O back to 1800 A.D. from air trapped in snow at the South Pole and the Greenland Ice Sheet Project II ice core. Global Biogeochem. Cycles, 16, 1129- 420 (2002). 9. Sowers, T.; Bernard, S.; Aballain, O.; Chappellaz, J.; Barnola, J.-M. & Marik, T. Records of the δ 13C of atmospheric CH4 over the last 2 centuries as recorded in Antarctic snow and ice. Global Biogeochem. Cycles, 19, (2005). 10. Sowers, T. Atmospheric methane isotope records covering the Holocene period. Quaternary © 2013 Macmillan Publishers Limited. All rights reserved. 425 Science Reviews, 29, 213-221 (2010). 11. Schmitt, J.; Seth, B.; Bock, M.; Fischer, H.; Möller, L.; Sapart, C.; Prokopiou, M.; Röckmann, T. & Sowers, T. On the interference of Kr during carbon isotope analysis of methane using continuous-flow combustion–isotope ratio mass spectrometry. Atmos. Meas. Tech., 6, 1425-1445 (2013). 430 12. Headly, M. A. & Severinghaus, J. P. A method to measure Kr/N2 ratios in air bubbles trapped in ice cores and its application in reconstructing past mean ocean temperature. J. Geophys. Res., 112, D19105- (2007). 13. Petit, J. R.; Basile, I.; Leruyuet, A.; Raynaud, D.; Lorius, C.; Jouzel, J.; Stievenard, M.; Lipenkov, V. Y.; Barkov, N. I.; Kudryashov, B. B.; Davis, M.; Saltzman, E. & Kotlyakov, V. 435 Four climate cycles in Vostok ice core. Nature, 387, 359-360 (1997). 14. Sowers, T.; Bender, M.; Raynaud, D. & Korotkevich, Y. S. δ 15N of N2 in Air Trapped in Polar Ice: a Tracer of Gas Transport in the Firn and a Possible Constraint on Ice Age-Gas Age Differences. J. Geophys. Res., 97, 15683-15697 (1992). 15. Schwander, J.; Stauffer, B. & Sigg, A. Air mixing in firn and the age of the air at pore close- 440 off. Annals of Glaciology, 10, 141-145 (1988). 16. Trudinger, C. M.; Enting, L. G.; Etheridge, D. M.; Francey, R. J.; Levchenko, V. A.; Steele, L. P.; Raynaud, D. & Arnaud, L. Modeling air movement and bubble trapping in firn. J. Geophys. Res., 102, (1997). 17. Severinghaus, J. P.; Grachev, A. & Battle, M. Thermal fractionation of air in polar firn by 445 seasonal temperature gradients. Geochem. Geophys. Geosyst., 2, (2001). 18. Schwander, J.; Sowers, T.; Barnola, J.-M.; Blunier, T.; Fuchs, A. & Malaize, B. Age scale of the air in the summit ice: Implication for glacial-interglacial temperature change. J. Geophys. Res., 102, 19483–19493 (1997). 19. Landais, A.; Barnola, J.; Kawamura, K.; Caillon, N.; Delmotte, M.; Van Ommen, T.; 450 Dreyfus, G.; Jouzel, J.; Masson-Delmotte, V.; Minster, B.; Freitag, J.; Leuenberger, M.; Schwander, J.; Huber, C.; Etheridge, D. & Morgan, V. Firn-air δ15N in modern polar sites and glacial-interglacial ice: a model-data mismatch during glacial periods in Antarctica?. Quaternary Science Reviews, 25, 49-62 (2006). 20. Landais, A.; Dreyfus, G.; Capron, E.; Masson-Delmotte, V.; Sanchez-Goñi, M.; Desprat, S.; 455 Hoffmann, G.; Jouzel, J.; Leuenberger, M. & Johnsen, S. What drives the millennial and orbital variations of δ18Oatm? Quaternary Science Reviews, 29, 235-246 (2010). 21. EPICA community members, One-to-one coupling of glacial climate variability in Greenland and Antarctica. Nature, 444, 195-198 (2006). 22. Bock, M.; Schmitt, J.; Behrens, M.; Möller, L.; Schneider, R.; Sapart, C. & Fischer, H. A © 2013 Macmillan Publishers Limited. All rights reserved. 460 gas chromatography/pyrolysis/isotope ratio mass spectrometry system for high-precision δD measurements of atmospheric methane extracted from ice cores. Rapid Commun. Mass Spectrom., 24, 621-633 (2010). 23. Lemieux-Dudon, B.; Blayo, E.; Petit, J.-R.; Waelbroeck, C.; Svensson, A.; Ritz, C.; Barnola, J.-M.; Narcisi, B. M. & Parrenin, F. Consistent dating for Antarctic and Greenland ice cores. 465 Quaternary Science Reviews, 29, 8-20 (2010). 24. Rohling, E. J.; Grant, K.; Bolshaw, M.; Roberts, A. P.; Siddall, M.; Hemleben, C. & Kucera, M. Antarctic temperature and global sea level closely coupled over the past five glacial cycles. Nature Geosci, 2, 500-504 (2009). 25. Grant, K. M.; Rohling, E. J.; Bar-Matthews, M.; Ayalon, A.; Medina-Elizalde, M.; Ramsey, 470 C. B.; Satow, C. & Roberts, A. P. Rapid coupling between ice volume and polar temperature over the past 150,000 years. Nature, 491, 744-747 (2012). 26. Loulergue, L.; Schilt, A.; Spahni, R.; Masson-Delmotte, V.; Blunier, T.; Lemieux, B.; Barnola, J.-M.; Raynaud, D.; Stocker, T. F. & Chappellaz, J. Orbital and millennial-scale features of atmospheric CH4 over the past 800,000 years.. Nature, 453, 383-386 (2008). 475 27. Parrenin, F.; Barnola, J.-M.; Beer, J.; Blunier, T.; Castellano, E.; Chappellaz, J.; Dreyfus, G.; Fischer, H.; Fujita, S.; Jouzel, J.; Kawamura, K.; Lemieux-Dudon, B.; Loulergue, L.; Masson-Delmotte, V.; Narcisi, B.; Petit, J.-R.; Raisbeck, G.; Raynaud, D.; Ruth, U.; Schwander, J.; Severi, M.; Spahni, R.; Steffensen, J. P.; Svensson, A.; Udisti, R.; Waelbroeck, C. & Wolff, E. The EDC3 chronology for the EPICA Dome C ice core. Clim. Past, 3, 485- 480 497 (2007). 28. Lüthi, D.; Floch, M. L.; Bereiter, B.; Blunier, T.; Barnola, J.-M.; Siegenthaler, U.; Raynaud, D.; Jouzel, J.; Fischer, H.; Kawamura, K. & Stocker, T. F. High-resolution carbon dioxide concentration record 650,000-800,000 years before present.. Nature, 453, 379-382 (2008). 29. Blunier, T. & Brook, E. J. Timing of Millennial-Scale Climate Change in Antarctica and 485 Greenland During the Last Glacial Period. Science, 291, 109-112 (2001). 30. Schilt, A.; Baumgartner, M.; Schwander, J.; Buiron, D.; Capron, E.; Chappellaz, J.; Loulergue, L.; Schüpbach, S.; Spahni, R.; Fischer, H. & Stocker, T. F. Atmospheric nitrous oxide during the last 140,000 years. Earth and Planetary Science Letters, 300, 33-43 (2010). 31. Blunier, T.; Spahni, R.; Barnola, J.-M.; Chappellaz, J.; Loulergue, L. & Schwander, J. 490 Synchronization of ice core records via atmospheric gases. Clim. Past, 3, 325-330 (2007). 32. Ahn, J. & Brook, E. J. Atmospheric CO2 and Climate on Millennial Time Scales During the Last Glacial Period. Science, 322, 83-85 (2008). 33. Bereiter, B.; Lüthi, D.; Siegrist, M.; Schüpbach, S.; Stocker, T. F. & Fischer, H. Mode change of millennial CO2 variability during the last glacial cycle associated with a bipolar © 2013 Macmillan Publishers Limited. All rights reserved. 495 marine carbon seesaw. Proceedings of the National Academy of Sciences, 109, 9755-9760 (2012). 34. O'Leary, M. H. Carbon Isotope Fractionation In Plants. Phytochemistry, 20, 553-567 (1981). 35. Farquhar, G. D.; Ehleringer, J. R. & Hubick, K. T. Carbon Isotope Discrimination and Photosynthesis Rid A-3722-2008. Annual Review of Plant Physiology and Plant Molecular 500 Biology, 40, 503-537 (1989). 36. Gerhart, L. M. & Ward, J. K. Plant responses to low [CO2] of the past. New Phytologist, 188, 674-695 (2010). 37. Farquhar, G. D.; Oleary, M. H. & Berry, J. A. On the Relationship Between Carbon Isotope Discrimination and the Inter-cellular Carbon-dioxide Concentration In Leaves Rid A-3722- 505 2008 Rid B-8211-2009. Australian Journal of Plant Physiology, 9, 121-137 (1982). 38. Ehleringer, J. R. Implications of quantum yield differences on the distributions of C3 and C4 grasses. Oecologia, 31, 255-267 (1978). 39. Cowling, S. A. & Sykes, M. T. Physiological Significance of Low Atmospheric CO2 for Plant/Climate Interactions. Quaternary Research, 52, 237-242 (1999). 510 40. Ehleringer, J. (2005). The Influence of Atmospheric CO2, Temperature, and Water on the Abundance of C3/C4 Taxa. In: Baldwin, I.; Caldwell, M.; Heldmaier, G.; Jackson, R.; Lange, O.; Mooney, H.; Schulze, E.-D.; Sommer, U.; Ehleringer, J.; Denise Dearing, M. & Cerling, T. (Ed.), Ecological Studies, Springer New York. 41. Sankaran, M.; Hanan, N. P.; Scholes, R. J.; Ratnam, J.; Augustine, D. J.; Cade, B. S.; 515 Gignoux, J.; Higgins, S. I.; Le Roux, X.; Ludwig, F.; Ardo, J.; Banyikwa, F.; Bronn, A.; Bucini, G.; Caylor, K. K.; Coughenour, M. B.; Diouf, A.; Ekaya, W.; Feral, C. J.; February, E. C.; Frost, P. G. H.; Hiernaux, P.; Hrabar, H.; Metzger, K. L.; Prins, H. H. T.; Ringrose, S.; Sea, W.; Tews, J.; Worden, J. & Zambatis, N. Determinants of woody cover in African savannas. Nature, 438, 846-849 (2005). 520 42. Edwards, E. J.; Osborne, C. P.; Strömberg, C. A. E.; Smith, S. A. & Consortium, C. 4. G. The Origins of C4 Grasslands: Integrating Evolutionary and Ecosystem Science. Science, 328, 587-591 (2010). 43. Prentice, I. C.; Harrison, S. P. & Bartlein, P. J. Global vegetation and terrestrial carbon cycle changes after the last ice age. New Phytologist, 189, 988-998 (2011). 525 44. Bragg, F. J.; Prentice, I. C.; Harrison, S. P.; Eglinton, G.; Foster, P. N.; Rommerskirchen, F. & Rullkötter, J. Stable isotope and modelling evidence for CO2 as a driver of glacial– interglacial vegetation shifts in southern Africa. Biogeosciences, 10, 2001-2010 (2013). 45. Claussen, M.; Selent, K.; Brovkin, V.; Raddatz, T. & Gayler, V. Impact of CO2 and climate on Last Glacial Maximum vegetation - a factor separation. Biogeosciences Discuss., 9, 15823- © 2013 Macmillan Publishers Limited. All rights reserved. 530 15852 (2012). 46. Huang, Y.; Street-Perrott, F. A.; Metcalfe, S. E.; Brenner, M.; Moreland, M. & Freeman, K. H. Climate Change as the Dominant Control on Glacial-Interglacial Variations in C3 and C4 Plant Abundance. Science, 293, 1647-1651 (2001). 47. Zhang, Z.; Zhao, M.; Lu, H. & Faiia, A. M. Lower temperature as the main cause of C4 535 plant declines during the glacial periods on the Chinese Loess Plateau. Earth and Planetary Science Letters, 214, 467-481 (2003). 48. Vidic, N. J. & Montañez, I. P. Climatically driven glacial-interglacial variations in C3 and C4 plant proportions on the Chinese Loess Plateau. Geology, 32, 337-340 (2004). 49. Wang, G.; Feng, X.; Han, J.; Zhou, L.; Tan, W. & Su, F. Paleovegetation reconstruction 540 using δ13C of Soil Organic Matter. Biogeosciences, 5, 1325-1337 (2008). 50. Ehleringer, J. R.; Cerling, T. E. & Helliker, B. R. C4 photosynthesis, atmospheric CO2, and climate. Oecologia, 112, 285-299 (1997). 51. Street-Perrott, F. A.; Huang, Y.; Perrott, R. A.; Eglinton, G.; Barker, P.; Khelifa, L. B.; Harkness, D. D. & Olago, D. O. Impact of Lower Atmospheric Carbon Dioxide on Tropical 545 Mountain Ecosystems. Science, 278, 1422-1426 (1997). 52. Sage, R. F.; Wedin, D. A. & Li, M. (1999). 10 - The Biogeography of C4 Photosynthesis: Patterns and Controlling Factors. In: Sage, R. F. & Monson, R. K. (Ed.), C4 Plant Biology, Academic Press. 53. Sage, R. F. The evolution of C4 photosynthesis. New Phytologist, 161, 341-370 (2004). 550 54. Harrison, S. P. & Prentice, C. I. Climate and CO2 controls on global vegetation distribution at the last glacial maximum: analysis based on palaeovegetation data, biome modelling and palaeoclimate simulations. Global Change Biology, 9, 983-1004 (2003). 55. Prentice, I. C. & Harrison, S. P. Ecosystem effects of CO2 concentration: evidence from past climates. Clim. Past, 5, 297-307 (2009). 555 56. Jolly, D. & Haxeltine, A. Effect of Low Glacial Atmospheric CO2 on Tropical African Montane Vegetation. Science, 276, 786-788 (1997). 57. Woillez, M.-N.; Kageyama, M.; Krinner, G.; de Noblet-Ducoudré, N.; Viovy, N. & Mancip, M. Impact of CO2 and climate on the Last Glacial Maximum vegetation: results from the ORCHIDEE/IPSL models. Clim. Past, 7, 557-577 (2011). 560 58. Wooller, M. J.; Swain, D. L.; Ficken, K. J.; Agnew, A. D. Q.; Street-Perrott, F. A. & Eglinton, G. Late Quaternary vegetation changes around Lake Rutundu, Mount Kenya, East Africa: evidence from grass cuticles, pollen and stable carbon isotopes. J. Quaternary Sci., 18, 3-15 (2003). 59. Rommerskirchen, F.; Eglinton, G.; Dupont, L. & Rullkötter, J. Glacial/interglacial changes © 2013 Macmillan Publishers Limited. All rights reserved. 565 in southern Africa: Compound-specific δ 13C land plant biomarker and pollen records from southeast Atlantic continental margin sediments. Geochem. Geophys. Geosyst., 7, (2006). 60. Castañeda, I. S.; Mulitza, S.; Schefuß, E.; Lopes dos Santos, R. A.; Sinninghe Damsté, J. S. & Schouten, S. Wet phases in the Sahara/Sahel region and human migration patterns in North Africa. Proceedings of the National Academy of Sciences, 106, 20159-20163 (2009). 570 61. Hessler, I.; Dupont, L.; Bonnefille, R.; Behling, H.; González, C.; Helmens, K. F.; Hooghiemstra, H.; Lebamba, J.; Ledru, M.-P.; Lézine, A.-M.; Maley, J.; Marret, F. & Vincens, A. Millennial-scale changes in vegetation records from tropical Africa and South America during the last glacial. Quaternary Science Reviews, 29, 2882-2899 (2010). 62. Hughen, K. A.; Eglinton, T. I.; Xu, L. & Makou, M. Abrupt Tropical Vegetation Response 575 to Rapid Climate Changes. Science, 304, 1955-1959 (2004). 63. Punyasena, S. W.; Mayle, F. E. & McElwain, J. C. Quantitative estimates of glacial and Holocene temperature and precipitation change in lowland Amazonian Bolivia. Geology, 36, 667-670 (2008). 64. Lane, C.; Horn, S.; Mora, C.; Orvis, K. & Finkelstein, D. Sedimentary stable carbon isotope 580 evidence of late Quaternary vegetation and climate change in highland Costa Rica. Journal of Paleolimnology, 45, 323-338 (2011). 65. Mayle, F. E.; Burbridge, R. & Killeen, T. J. Millennial-scale dynamics of southern Amazonian rain forests. Science, 290, 2291-2294 (2000). 66. Turcq, B.; Cordeiro, R. C.; Sifeddine, A.; Simões Filho, F. F.; Albuquerque, A. L. S. & 585 Abrão, J. J. Carbon storage in Amazonia during the Last Glacial Maximum: secondary data and uncertainties. Chemosphere, 49, 821-835 (2002). 67. Mayle, F. E. & Beerling, D. J. Late Quaternary changes in Amazonian ecosystems and their implications for global carbon cycling. Palaeogeography, Palaeoclimatology, Palaeoecology, 214, 11-25 (2004). 590 68. Mayle, F. E.; Beerling, D. J.; Gosling, W. D. & Bush, M. B. Responses of Amazonian ecosystems to climatic and atmospheric carbon dioxide changes since the last glacial maximum. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 359, 499-514 (2004). 69. Anhuf, D.; Ledru, M.-P.; Behling, H.; Da Cruz Jr., F.; Cordeiro, R.; Van der Hammen, T.; 595 Karmann, I.; Marengo, J.; De Oliveira, P.; Pessenda, L.; Siffedine, A.; Albuquerque, A. & Da Silva Dias, P. Paleo-environmental change in Amazonian and African rainforest during the LGM. Palaeogeography, Palaeoclimatology, Palaeoecology, 239, 510-527 (2006). 70. Bird, M. I.; Taylor, D. & Hunt, C. Palaeoenvironments of insular Southeast Asia during the Last Glacial Period: a savanna corridor in Sundaland? Quaternary Science Reviews, 24, 2228- © 2013 Macmillan Publishers Limited. All rights reserved. 600 2242 (2005). 71. Wurster, C. M.; Bird, M. I.; Bull, I. D.; Creed, F.; Bryant, C.; Dungait, J. A. J. & Paz, V. Forest contraction in north equatorial Southeast Asia during the Last Glacial Period. Proceedings of the National Academy of Sciences, 107, 15508-15511 (2010). 72. Jones, M. B. & Muthuri, F. M. Standing biomass and carbon distribution in a papyrus 605 (Cyperus papyrus L.) swamp on Lake Naivasha, Kenya. Journal of Tropical Ecology, 13, 347-356 (1997). 73. Kotze, D. C. & O'Connor, T. G. Vegetation variation within and among palustrine wetlands along an altitudinal gradient in KwaZulu-Natal, South Africa. Plant Ecology, 146, 77-96 (2000). 610 74. Saunders, M.; Jones, M. & Kansiime, F. Carbon and water cycles in tropical papyrus wetlands. Wetlands Ecology and Management, 15, 489-498 (2007). 75. Piedade, M. T. F.; Junk, W. J. & Long, S. P. The Productivity of the C4 Grass Echinochloapolystachya On the Amazon Floodplain Rid A-2488-2008. Ecology, 72, 1456-1463 (1991). 76. Sinninghe Damsté, J. S.; Verschuren, D.; Ossebaar, J.; Blokker, J.; van Houten, R.; van der 615 Meer, M. T.; Plessen, B. & Schouten, S. A 25,000-year record of climate-induced changes in lowland vegetation of eastern equatorial Africa revealed by the stable carbon-isotopic composition of fossil plant leaf waxes. Earth and Planetary Science Letters, 302, 236-246 (2011). 77. Vogts, A.; Moossen, H.; Rommerskirchen, F. & Rullkötter, J. Distribution patterns and 620 stable carbon isotopic composition of alkanes and alkan-1-ols from plant waxes of African rain forest and savanna C3 species. Organic Geochemistry, 40, 1037-1054 (2009). 78. Medina, E. & Minchin, P. Stratification of δ 13C values of leaves in Amazonian rain forests. Oecologia, 45, 377-378 (1980). 79. Buchmann, N.; Guehl, J.-M.; Barigah, T. S. & Ehleringer, J. R. Interseasonal comparison of 625 CO2 concentrations, isotopic composition, and carbon dynamics in an Amazonian rainforest (French Guiana). Oecologia, 110, 120-131 (1997). 80. Ehleringer, J. R.; Field, C. B.; Lin, Z.-F. & Kuo, C.-Y. Leaf carbon isotope and mineral composition in subtropical plants along an irradiance cline. Oecologia, 70, 520-526 (1986). 81. Mitchell, L. E.; Brook, E. J.; Sowers, T.; McConnell, J. R. & Taylor, K. Multidecadal 630 variability of atmospheric methane, 1000-1800 C.E.. J. Geophys. Res., 116, (2011). 82. Mischler, J. A.; Sowers, T. A.; Alley, R. B.; Battle, M.; McConnell, J. R.; Mitchell, L.; Popp, T.; Sofen, E. & Spencer, M. K. Carbon and hydrogen isotopic composition of methane over the last 1000 years. Global Biogeochem. Cycles, 23, (2009). 83. Chappellaz, J.; Barnola, J. M.; Raynaud, D.; Korotkevich, Y. S. & Lorius, C. Ice-core record © 2013 Macmillan Publishers Limited. All rights reserved. 635 of atmospheric methane over the past 160,000 years. Nature, 345, 127-131 (1990). © 2013 Macmillan Publishers Limited. All rights reserved. Figure legends Figure S1 | Magnitude of the Kr effect Δδ 13CKr derived from two distinct methods for the AWI and PSU instrument 640 a) the new δ¹³CH4 time series of Vostok (light blue) and EDML (red), the Holocene GISP II record11 (dark blue, on its original time scale), and the published EDML record2 over termination I, that was processed anew according to the procedures in this publication (purple). The corrected time series are illustrated with thicker lines and large circle markers. The original records before the correction are shown as thin lines with small white markers. 645 b) The range of Δδ13CKr values applied as individual correction to the respective datasets to account for the Kr-bias on δ¹³CH4. The color coding is the same as in a). Note that Δδ13CKr of both, Vostok and GISP II data measured at the PSU, were inferred indirectly from CH4 mixing ratios, while for the two EDML records the correction is based on the Kr-induced anomaly derived from the ioncurrent ratios (see section 1.3 for a detailed description of both approaches). 650 Figure S2 | Linear scaling of the size of interference on carbon isotope measurements caused by Kr (Δδ 13CKr) relative to variations in CH4 mixing ratios for the PSU instrument Atmospheric krypton mixing ratios are considered constant over time13.The size of the interference (Kr effect or Δδ13CKr) thus scales with CH4 concentrations. Measurements of three ambient air 655 samples were used to infer a linear relationship between Δδ13CKr and the inverted CH4 mixing ratio for the PSU instrument, that is used to account for the Kr effect of ice core (Vostok, WDC05) and laboratory inter-calibration measurements (see section 1.3 for further details). Results for the IPY cylinder with a mixing ratio representative for glacial conditions (“glacial”, 365 ppb) is shown as a light blue diamond, the one with “preindustrial” values (906 ppb) in orange, and the one 660 representative for modern concentrations (“present-day”, 1852 ppb) in dark red, all illustrated with its corresponding 1σ error range. The linear relationship in red allowed a first order estimate of © 2013 Macmillan Publishers Limited. All rights reserved. Δδ13CKr for ice core samples measured at the PSU lab, based on interpolated CH4 mixing ratios of the respective samples. 665 Figure S3 | Model results to quantify fractionation processes in the firn column due to atmospheric CH4 concentration changes a) Artificial atmospheric methane pulse at the model firn “surface” (grey line), and its corresponding concentration after the “bubble close-off” at the bottom of the firn column at three different accumulation rate and temperature regimens (“maximum” temperature: purple line; 670 “minimum” temperature: red line and “best guess” temperature: light blue line). b) Shifts of the carbon isotope signature of methane caused by the gravitational and diffusional fractionation in the firn column from the constant value at the “surface” (gray dashed line) to the signals enclosed in the model “bubbles” (colored lines) at the firn bottom, according to their respective scenario (color coding similar to a)) 675 Figure S4 | Representation of major Dansgaard-Oeschger methane rises during the last glacial period, and the respective imprint on its carbon isotope signature a) Six strong methane rises (color coding in the figure legends) coeval to major Dansgaard/Oeschger (DO) events (left and middle figure column) and the two ultimate terminations 680 (right column) in a 6,000 year time window aligned and centered around the major methane rise (illustrated by the vertical gray line). All CH4 data from the EDML ice core30 b) Corresponding time windows of the respective δ¹³CH4 data sections from Vostok and EDML (this study, color coding and alignment according to a)). The shaded bar represents the time window, in which data points might be biased by diffusional fractionation beyond our measurement uncertainty. © 2013 Macmillan Publishers Limited. All rights reserved. 685 Tables Supplementary Table 1 | Comparison of δ¹³CH4 results for the measurements of air samples performed at the AWI and PSU laboratories, that were part of the „2007 - IPY International Ice Core Gas Intercalibration Exercise“. Ambient air was diluted to resemble atmospheric methane mixing ratios typical for present day 690 (1852 ppb), pre-industrial (906 ppb) and glacial (365 ppb) conditions. δ¹³CH4 values are reported with respect to VPDB. The first two columns report original measurements, the following two columns the respective values after the correction for Kr interference. Deviations in the carbon isotopic signature caused by the ionized Kr (Δδ13CKr) and its dependency on CH4 levels are summarized in the final column. The results are further used to infer the absolute standardization 695 offset between both laboratories. S a m p le I D | e p o c h δ 1 3 C H 4 (‰ ) 1 σ (‰ ) O rig in a l A W I a n a ly s e s ( 5 /2 0 1 0 ) C A03560 | pre s e nt da y C C 71560 | pre indus tria l C A01179 | g la c ia l P S U a n a ly s e s ( 7 /2 0 0 7 ) C A03560 | pre s e nt da y C C 71560 | pre indus tria l C A01179 | g la c ia l D if f e re n c e ( A W I -‐ P S U ) C A03560 | pre s e nt da y C C 71560 | pre indus tria l C A01179 | g la c ia l A v g . la b o f f s e t S t . d e v . ( 1 σ) δ 1 3 C H 4 (‰ ) 1σ (‰ ) Δδ 1 3 C K r (‰ ) C o rre c t e d f o r K r -‐47.14 -‐47.07 -‐46.25 0.05 0.09 0.11 -‐47.33 -‐47.40 -‐46.97 0.06 0.12 0.13 0.19 0.33 0.72 -‐47.08 -‐47.15 -‐46.86 0.16 0.10 0.06 -‐47.20 -‐47.41 -‐47.52 0.16 0.10 0.06 0.12 0.26 0.66 -‐0.06 0.08 0.61 0 .2 1 0 .3 5 -‐0.13 0.01 0.55 0 .1 4 0 .3 6 Supplementary Table 2 | δ¹³CH4 values of WDC05 A ice core material measured with the AWI and PSU instruments to further test the alignment of both systems. 700 No adjustments have been applied to correct for gravitational settling. Values denoted as “original” © 2013 Macmillan Publishers Limited. All rights reserved. are inferred according to the standard routines in the respective laboratories, before the correction for Krypton interference. Acquisitions with the AWI instrument were treated similar to EDML ice and IPY air samples correcting the chromatograms for the Kr interference, PSU measurements using the CH4 dependent Δδ13CKr to account for Kr. WDC05A [CH4] values and approximate age of 705 the samples (black diamond) were interpolated from data of Mitchell et al.,201181. Note, that the two PSU measurements (marked by asterisks) are not from the same depth as the AWI sample, but δ¹³CH4 variability (1σ) is less than 0.3 ‰ over the depth interval 161.5 m (1593 AD) to 173.4 m (1540 AD) and thus in the the order of our analytical uncertainty82. S a m p le Lab ID 710 WD C 05A* WD C 05A WD C 05A* PS U AWI PS U D e p t h A g e ♦ (m ) (AD ) 164.96 166.78 169.80 1571 1564 1551 δ 1 3 C H 4 1 σ (‰ ) (‰ ) δ 1 3 C H 4 1 σ (‰ ) (‰ ) Δδ 1 3 C K r (‰ ) n O rig in a l -‐47.94 -‐ -‐47.53 0.03 -‐47.53 -‐ C o rre c te d f o r K r -‐48.28 -‐ -‐47.81 0.04 -‐47.87 -‐ 0.34 0.28 0.34 -‐ 2 -‐ Supplementary Table 3 | Manual CH4 tie-points used for the adjustment of the misaligned Vostok section in between 129-84 kyr BP28. The synchronization was performed with EDC26 and Vostok83 atmospheric methane records. The EDC3 ages of the five listed Vostok sample depths were adjusted according to the corresponding 715 EDC tie points. All Vostok records presented in this study are dated according to this adjustments. t ie -‐p o in t E D C d e p t h (m ) E D C 3 a g e ( y e a r s B P ) E D C C H ₄ (ppb) V o s t o k d e p t h (m ) V o s t o k C H ₄ (ppb) 1 2 3 4 5 1 24 1 .67 1 36 7 .89 1 47 3 .40 1 54 3 .59 1 72 3 .46 8 36 2 7 9 58 6 6 1 06 7 81 1 15 0 81 1 28 8 71 5 00 .7 4 70 .2 5 10 .1 4 80 .2 5 59 .9 1 26 6 .83 1 44 0 .34 1 53 6 .00 1 63 5 .97 1 88 1 .99 5 00 .1 4 70 .0 5 10 .8 4 80 .4 5 60 .1 Supplementary Table 4 | Methane synchronization tie-points used for dating the Byrd ice core © 2013 Macmillan Publishers Limited. All rights reserved. records shown in this work. The age conversion to the “Unified” age scale23 is based on 720 EDML30 and Byrd29 atmospheric methane records. t ie -‐p o in t 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 E D ML d e p t h “ U n if ie d ” a g e E D ML C H ₄ B y rd d e p t h B y rd C H ₄ (m ) ( y e a rs B P ) (ppb) (m ) (ppb) 717.02 829.64 1070.95 1154.20 1173.61 1233.17 1260.48 1286.47 1337.80 1365.07 1403.97 1436.97 1489.88 1601.65 1627.36 1666.48 1680.64 1688.09 1760.37 1860.22 1914.24 1949.26 2023.03 2065.88 11599 14543 23231 27748 28810 32339 33728 35417 38258 39433 41378 43074 46719 53264 54732 57333 58176 58586 63448 71706 75872 78706 85207 89241 609.35 529.66 383.60 392.50 419.31 435.59 456.60 467.63 488.11 423.15 443.35 452.35 469.20 491.10 493.91 532.10 538.28 497.84 468.08 470.32 460.21 480.43 550.63 493.35 1071.76 1195.84 1446.29 1498.37 1526.10 1595.76 1617.35 1654.37 1716.45 1744.35 1780.49 1806.94 1863.57 1960.44 1973.59 2000.30 2011.48 2017.44 2062.90 2082.65 2100.52 2111.69 2133.58 2139.75 574.28 543.30 361.95 392.42 417.05 443.86 417.67 449.47 490.03 415.49 422.27 441.63 461.97 488.11 491.04 523.93 533.62 463.90 476.49 460.38 455.18 494.55 546.72 489.08 © 2013 Macmillan Publishers Limited. All rights reserved.