sign in sign up
<<
Home , Ice core

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 14, NO. 2, PAGES 559-572, JUNE 2000

On the origin and timing of rapid changesin atmospheric during the

EdwardJ. Brook,• SusanHarder, • Jeff Sevennghaus, ' • Eric J. Ste•g,. 3and Cara M. Sucher•'5

Abstract. We presenthigh resolution records of atmosphericmethane from the GISP2 (GreenlandIce SheetProject 2) corefor fourrapid transitions that occurred during the past50 ka: theend of theYounger Dryas at 11.8ka, thebeginning of theBolling-Aller0d period at 14.8ka, thebeginning of interstadial8 at 38.2 ka, andthe beginning of intersradial12 at 45.5 ka. Duringthese events, concentrations increased by 200-300ppb over periodsof 100-300years, significantly more slowly than associated temperature and snow accumulationchanges recorded in the ice corerecord. We suggestthat the slowerrise in methane concentrationmay reflect the timescale of terrestrialecosystem response to rapidclimate change. We find noevidence for rapid,massive methane emissions that might be associatedwith large- scaledecomposition of methanehydrates in sediments.With additionalresults from the Taylor DomeIce Core() we alsoreconstruct changes in the interpolarmethane gradient (an indicatorof thegeographical distribution of methanesources) associated with someof therapid changesin atmosphericmethane. The resultsindicate that the rise in methaneat thebeginning of theB011ing-Aller0d period and the later rise at theend of theYounger Dryas were driven by increasesin bothtropical and methane sources. During the (a 1.3 ka cold periodduring the last deglaciation)the relativecontribution from borealsources was reduced relativeto the early andmiddle periods.

1. Introduction associatedwith orbitalforcing. A 20,000 periodicityin the Vostok record was attributed to the effects of solar insolation Studies of ice cores from and Antarctica reveal that variationson the tropical monsooncycle [Chappellaz et al., significant changes in atmospheric methane concentration 1990; Petit-Maireet al., 1991], or alternatively,to temperature occurredwith a range of periodsover the past400,000+ and precipitationvariations in high latitude regions [Rasmussen and Khalil, 1984; Khalil and Rasmussen, 1987; [Crowley,1991]. More recently,higher-resolution records have Raynaud et al., 1988; Stauffer et al., 1988; Chappellazet al., becomeavailable from coresthrough the rapidly accumulating 1990, 1993a, 1997; Etheridge et al., 1992; Jouzel et al., 1993; Greenland . Methane records from the Greenland Ice Blunier et al., 1993, 1995; Brook et al., 1996a, 1999; Petit et al., SheetProject 2 (GISP2) andGreenland Ice CoreProgram (GRIP) 1999]. Wetlandsare the major naturalmethane source [Fung et ice cores support the general patterns revealed at Vostok al. , 1991; Chappellaz et al., 1993b; Hein et al., 1997], and [Chappellazet al., 1993a;Brook et al., 1996a]. They alsoreveal researchin modem suggeststhat methane emissions that atmospheric methane levels varied significantly on depend on temperature,hydrologic balance, and possibly net millennial timescalesduring the last glacial period. The ecosystemproduction [Whiting and Chanton, 1993; Bubier and millennialvariations were coeval with millennial-scalewarmings Moore, 1994; Schlesinger, 1996]. Warm, wet, and highly and coolings(interstadial events) inferred from oxygenisotope productive conditions are associatedwith higher methane recordsfrom central Greenlandice cores,proxies for surface emissions. As a result, ice core methane records have been temperature[Johnsen et al., 1993; Grootes et al., 1993]. The ice interpreted as indicators of terrestrial climatic conditions on a core methanerecord therefore indicates a close relationship variety of timescales. betweeninterstadial climate and changesin terrestrialmethane Long records from the Vostok ice core indicate that emissions[Chappellaz et al., 1993a]. This conclusionaugments atmospheric methane concentrations varied on timescales a largebody of evidencesuggesting a correlationbetween rapid in Greenland ice core records and terrestrial •Departmentof , Washington State University, Vancouver. climatechange in a variety of locationsaround the world [e.g., 2ScrippsInstitution of Oceanography,University of California,San Grimm et al., 1993; Broecker et al., 1998; Gasse and Van Diego, La Jolla. Campo;1998, Grigg and Whitlock,1988; Allen et al., 1999]. 3Departmentof Earthand EnvironmentalScience, University of Pennsylvania,Philadelphia. Gaps in our understanding of millennial-scale shifts in nGraduateSchool of Oceanography,University of RhodeIsland, methaneconcentration include incomplete knowledge of (1) the Narragansett,Rhode Island. detailed timing and magnitude of the rapid concentration SNowat NOAA Office of Global Programs,Silver Spring, Maryland. changes,(2) the possiblerole of massive releasesof clathrate- boundmethane from marinesediments and permafrost,and (3) Copyright2000 by the AmericanGeophysical Union. the methanesource locations and how they changedwith time. Paper number 1999GB001182. In this paper we examine these issuesusing methanerecords 0886-6236/00/1999GB001182512.00 covering the past 50,000 years from two ice cores, the GISP2

559 560 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

core from centralGreenland and the Taylor Dome core from the wasnormally analyzed twice. The loop pressuresfor the first and Ross Sea sector of Antarctica. We use the results to constrain the secondmeasurements were 40-50 and 15-20 torr, respectively. rate and magnitudeof methaneconcentration change associated Methane concentrationswere quantified by measuringpeak with rapid climate shifts during the last glacial period and areasusing ELAB (OMS Tech, Miami, Florida) chromatographic deglaciation.We alsodetermine the interpolarmethane gradient, softwareand computerinterface. Our working standardwas a a possibleproxy for the geographicaldistribution of methane high-pressurecylinder of syntheticair preparedby Scott-Marin sources[Rasmussen and Khalil, 1984], at selectedtime intervals, SpecialtyGases. This standardhas a methaneconcentration of andexamine the implicationsof changesin the gradient. 962+ 6 ppb(mean + 2 ti/nesSE, whereSE is standarderror), determinedby comparisonto two similar cylinders that were 2. Samples and Analytical Procedures calibrated at the NOAA/Climate Monitoring and Diagnostics Laboratory,which have methaneconcentrations of 1488+3 ppb 2.1. Ice Core Samples and 1824+3 ppb. We calibratedthe instrumentdaily with six to eightinjections of the workingstandard at a rangeof sampleloop The GISP2 ice core was recovered in 1993 at 72ø36'N, pressuresbetween 10 and 60 torr. The averagedaily precisionof 38ø30'Win centralGreenland. We have extendedour original these runs of the calibration standard was + 9 ppb (2 x SE, GISP2 data set [Brooket al., 1996a]by analyzingsamples from n-347). 129 additionaldepths in the upper2438 m, which representsthe Contaminationin our sampleprocessing was evaluatedin two past 50 ka. We focused on abrupt transitionsin the record, ways. First, we ran a total analytical blank, as follows. An ice includingthe terminationof the YoungerDryas periodat •-11.8 sample analyzed the previous day was melted and refrozen 3 ka, the beginningof the B011ing-Aller0dperiod at --14.5 ka, and in a vacuum vessel. After each refreezing step any the onsetsof interstadials8 and 12, rapid warming eventsthat remainingair in the vesselwas pumpedout. The ice samplewas occurredat-- 38.2 and45.5 ka in the GISP2 isotoperecord. The removedfrom the vessel, and the edgeswere trimmed with the GISP2 methane data set over the 0-50 ka interval now includes band saw normally used to cut ice samples. The ice cube was 289 sampledepths. The Taylor Dome ice core was recoveredin then placedin a secondvessel, and the ambientair was pumped 1994 at 77ø48'S,159ø53'E on Taylor Dome, a smallice domein out while maintainingthe vesselat -30øC. The vesselwas then the Ross Sea sector of Antarctica. The Taylor Dome cooled to -80øC. One hundred torr of standard air was introduced paleoclimaterecord extends to greaterthan 130 ka at a depthof into the vacuum line and sample vessel, the vessel valve was 554 m [Grootesand Steig, 1994;Steig, 1996; Steiget al., 2000] closed, and the remainder of the line was evacuated. One hundred but we focushere on methanemeasurements of 137 samplesin torr in the samplevessel resulted in a pressureof 40-50 torr in the upper 445 m (0-50 ka), with high resolutionmeasurements the GC sampleloop when the gaswas expanded,a value similar over the time intervals listed above. Data are available from the to that obtainedwhen analyzingmost ice samples. This "blank National Oceanic and Atmospheric Administration (NOAA) sample"was then melted and refrozen. We analyzedthe air in GeophysicalData Center(http://www.ngdc.noaa.gov/) or from the headspace and determined the blank as the difference the authors. between the expected value of the standardand the measured value. Blanksmeasured in this way over a 27 monthtime period 2.2. Analytical Methods had a meanof 16 + 4 ppb (mean + 2 x SE, n-63). Ice samplesweighing--35 g were placedin cubicalstainless In an attempt to locate the source of contamination, we steel vacuum vessels with volumes of- 40 cm 3. The vessels performedan additionalblank experiment. After analysisof an have Nupro SS-4H stainlesssteel bellows valves and were sealed ice was completedand while the sample was still with gold o-rings. The vesselswere attachedto a vacuumline frozen, we evacuatedthe sample containerand introduced 100 with stainlesssteel Swagelock fittings with Teflonferrules. They torr of standard in the line and vessel. The vessels were sealed were maintained at -25 ø to -30øC in an alcohol bath cooled with for 5 min., and the line was evacuated. The gas in the vessels a cryogenicprobe chiller (FTS SystemsMFC-100). Ambientair was then analyzedin the normalfashion. This blank test should was removedby pumpingwith a mechanicalvacuum pump for detect contamination due only to the vacuum line and sample 30 min. Ice sampleswere isolated from the pumpand allowedto vesseland exclude contaminationspecifically due to the melting melt in a warm waterbath for 30 minutes. The sampleswere and refreezingstep. The meanof blanksperformed this way with refrozenby loweringthe vesselsslowly into a -80øC alcoholbath different sample vesselsand samplesover a one month period overa 30 min. timeperiod. This procedureexpels the air trapped was 3+4 ppb (mean+ 2 x SE, n=20), suggestingthat most of the in the ice into the evacuatedvessel headspace. The vesselswere contamination in our procedure is due to the melting and attached to a welded stainless steel vacuum line constructed with refreezingstep. We are not certainif this contaminationis due to Nupro SS-4H valves. The line served as the inlet to a Hewlett- small vacuum leaks or chemical reactions between the water and PackardModel 5890 SeriesII gaschromotagraph (GC) equipped the stainlesssteel vacuumvessel. Fuchs et al. [1993] suggested with a six-portgas sampling valve, a 10 cm3 sampleloop, a that the compressionof stainlesssteel on stainlesssteel might Poropak Q column, and a flame ionization detector. The GC producemethane. Our resultssuggest that the mechanicalaction sampleloop andvacuum line wereevacuated to a pressureof - of the stainless steel bellows and the contact of the stainless steel 10'3 torr with a mechanicalvacuum pump. The headspace gas stem tip with the valve body are not major sources of wasintroduced to thesample loop, and the pressure in thesample contaminationin our analyses. loop was measured with an MKS Model 122A Baratron To generatethe recordspresented here we analyzed at least CapacitanceManometer fixed to the outlet port of the gas two separateice samplesat each sampledepth and analyzedthe samplingvalve. The gasextracted from individualice samples headspacegas for each twice (averagingthe resultsfor the two BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD 561 injections). If analysesof duplicate ice samplesdisagreed by matrix only. Below we describethe correlationprocedure and more than 5%, a third replicatewas analyzed. For 859 duplicate the approach taken to account for this effect. pairs (not all measuredfor this study) meeting the 5% criteria, analyzed between 1993 and 1997 in the URI laboratory, the 3.2. GISP2 Gas Age Timescale and the Gas Age- average deviation from the mean value was 1.2%. We Difference occasionallyalso rejecteda result if the pressurein the sample The depthat whichair is trappedin polarice (the closeoff loop was significantly lower than usually observed,which we depth)is a functionof temperatureand ice accumulationrate, and attribute to incomplete melting or gas loss prior to analysis. is typically50-100 m belowthe ice surface. Abovethat depth, These analysesgenerally resultedin anomalouslyhigh methane air is transportedin the by convectionand . Gases concentrations. in polar ice are thereforeyounger than the surroundingice. In FigureI we presentmethane results for GISP2and Taylor Dome 3. Methane Records as a function of gas age. In this sectionand section3.3 we explainhow thesetimescales were derived. 3.1. Gas Age and Ice Age Timescales For the GISP2 recordwe usedthe gasage timescaledescribed To compare methane records from different ice cores, a byBrook et al. [19•16a].This timescale is based on the GISP2 commontemporal framework is necessary.The methanerecords layer countingtimescale [Meese et al., 1994], which provides themselvesprovide one way of creatingthat frameworkbecause agesfor the ice matrix as a functionof depth. GISP2 gas ages the atmospheric mixing time (--1 year) is short enough that were calculated us;ng the steady state Herron and Langway interannual variations in methane concentrationare essentially [1980] densificationmodel [Brook et al., 1996a]. This model synchronousin both hemispheres.The isotopiccomposition of allows us to calculate the depth at which air bubbles were atmosphericO2 (fi•8Oatm)trapped in polarice can be usedin a isolatedfrom the atmosphereand the ageof the air bubblesat that similarway, becausefi18Oat mvaried significantly over the past point,as a functionof temperatureand accumulation rate. This 50,000years, primarily owing to changesin the•80 of seawater value is referredto as the gas age-iceage difference,or "A age." [Bender et al., 1994]. The gas age timescaleis obtainedby subtractingA age, which Our strategy was to use the GISP2 timescale, determined varieswith time, from the ice age timescale.The GISP2 gasage scaleused here agreeswell ( + 100 years)with a later GISP2 gas primarily by countingannual layers for the last 50 ka [Meeseet al., 1994] as a primary timescale. We then correlatedmethane age timescaleof Schwanderet al. [1997], whichincorporates gas and•80•tm variations in GISP2with similarvariations in Taylor age-iceage differencescalculated from a dynamicdensification Dome to place GISP2 and Taylor Dome on a commontemporal model [Barnola et al., 1991]. framework. Doing so is complicated by the fact that gases trappedin polar ice are youngerthan the surroundingice matrix, 3.3. Correlation of Taylor Dome and GISP2 Timescales while the layer counting timescale provides ages for the ice For Taylor Dome we createda correlationgas age timescale by visually matchingcommon inflection pointsin the GISP2 and

-32 Taylor Dome methane and •l•o•t • time series, each -34 _L.,.z...... 8 12 • • 34 5 6 7 10 correspondingto a timewhen methane or $•Oatmwas changing O -38 rapidly, and interpolatingbetween those points [Sucher, 1997; 7_• -40 •: Steig et al., 1998]. Becausethe atmospheremixes in about a ß•-36-42 • YD2• 700800 _• year, and the lifetimes of methaneand oxygenin the atmosphere 600 • are longer,inflections in the methaneor •5•8Oa,mrecords should 500400 • occur at essentiallythe same time in both polar regions.Using 800 300 cr methanefor correlationis complicatedslightly by the existence

600 of an interpolar methane gradient. We accounted for this 500 complicationby making the a priori assumptionthat the gradient 400 700t ' •' was linearly proportionalto the methaneconcentration over the 300 time periodof interest. We assumedthat therewas no gradient when methane concentrations were at their lowest levels and that -40-42 •,.O-' -44 the ratio of Greenland/Antarctic concentrations was 1.10 at the maximum concentrationobserved. The latter figure is basedon 0 10 20 30 40 50 earlierwork [Rasmussenand Khalil, 1984] and the assumptionof Age (ka) no gradient at times of low concentrationsis equivalent to assumingno borealmethane sources at thosetimes. Figure 1. Methaneand records from (A) GISP2 To correlatethe recordswe createda gradient-adjustedTaylor and (B) Taylor Dome ice cores. Isotopedata for GISP2 are 2 m Dome methane time series using these assumptions. The averagesfrom Grooteset al. [1993], and data for Taylor Dome are 0.5 m averages from Steig et al. [2000]. Numbered resultingcontrol points are listed in Table 1. Belowwe calculate interstadialsin the Greenland record are indicated in Figure 1. interpolar methane gradients with our data by averaging Timescales for GISP2 methane and isotope records are from concentrationsfrom both recordsover time periodsduring which Brook et al. [1996a]. Timescalesfor Taylor Dome methaneand methanelevels were relatively stable. The assumptionmade isotoperecords are described in the text. aboveabout the interpolarmethane gradient does not influence 562 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

Table 1. Gas Age Control Points for the Taylor Dome 4. Results and Discussion Timescale for 50-0 ka: Correlation with GISP2 Atmospheric Methaneand •j18Oat mRecords. 4.1. Rapid MethaneVariations During the Past50,000 Years Age basedon gas Correlated 4.1.1. Patterns of variation. In Figure 1 we presentthe Depth,m Correlation,kabp Variable GISP2 and Taylor Dome methanerecords for the past50,000 yearsas well as isotopicrecords (•'8Oic e) from the GISP2 59.0 0.0 O-agedepth in f'nm [Grooteset al., 1993] andTaylor Dome [Steiget al., 1998;2000] 285.5 5.9 methane 334.5 8.3 methane ice cores.The generalmethane trends over this time periodare 339.0 8.8 methane known from previouswork [Chappellaz et al., 1990, 1993a, 341.6 8.9 methane 1997; Blunier et al., 1995; Brook et al., 1996a]. Between 50 and 356.2 10.9 •18Oatm 26 ka, several oscillations in methane concentration occurred 358.3 l 1.3 methane synchronouslyor near-synchronouslywith •80,ce variations 362.1 11.7 methane observed in Greenland ice cores. Methane remained low (-400 370.2 12.4 fi18Oatm 370.6 12.8 methane ppb) between26 and 17 ka, beganrising slowly at 17 ka, rose 371.5 12.9 methane rapidlyat the beginningof the B011ing-Aller0dperiod (starting at 377.5 14.9 methane 14.8 ka), and remainedat interglaciallevels of 600-700 ppb for 379.5 17.4 methane the next -2000 years. The methane oscillation during the 384.2 26.3 fi18Oatm YoungerDryas [Chappellazet al., 1990] is also observedin both 398.0 34.2 methane 401.0 35.3 methane of our records,as is the prominentmethane minimum duringthe 404.0 36.3 methane mid-Holocene [Blunier et al., 1995]. In general, where they 411.2 38.5 methane overlap,our resultsagree well with thoseof Chappellazet al. 461.7 56.7 fi18Oatm [1997], after correctingfor a known interlaboratoryoffset of 2 % due to differencesin standardgases [Sowers et al., 1997]. Methanedata from Brook et al. [1996a] and this work; •j18Oatmfrom 4.1.2. Timing and rate of millennial-scale methane change. Sucher [ 1997]. The YoungerDryas data (Figure 2) show that the rapid methane rise at the end of the Younger Dryas in the GISP2 record occurredwithin _+ 100 yearsof the associated•80,c • rise,but those calculationsbecause we correlate only at times of rapidly exact phasingcan not be determinedfrom thesedata becauseof changingmethane concentrations and do not calculategradients uncertaintiesin gas age-ice age differences. Further work by for those times. Severinghauset al. [1998] independentlyinferred the timing of The uncertainty of the resulting Taylor Dome gas age thesurface warming from anomalies in the•SN of N2 in ice core timescalerelative to the GISP2 gas age timescaleis a functionof air. These anomaliesare causedby thermalisotopic fractionation the resolution of the records and varies with time. We estimate in the firn due to the temperaturegradient between the warm the uncertainty as one-half the sample resolution of the Taylor surfaceand the coolerclose-off depth. Severinghauset al. [ 1998] Dome record (which has lower sampleresolution than GISP2); _+ showed that the methane increase at the end of the Younger 100 yearsfor 10-15 ka, _+1000 yearsbetween 15 and 35 ka, and Dryas began 0-30 years after the temperaturerise indicatedby _+500 yearsbetween 35 and 50 ka. Later than 8 ka we have very b•80,ce. little age control in this timescale. Becauseaccumulation rates We now also have high resolution methane data for the probably varied between the control points and time periods Bolling onsetand the onsetsof Interstadials8 and 12 (Figure2). betweencontrol points are large for some intervals,uncertainties In Figure 2 it appearsthat the start of the methanerise at the at times betweenthe matchpoints are harderto quantify. At this Bolling transitionwas coincident,within --100-200 years, with level the accuracy of the correlation between Taylor Dome and the majorshift in fi•80,•e at 14.8 ka. The smallapparent of GISP2 is not central to the interpretation of our results and is methanerelative to the major isotopicshift is ambiguousgiven discussedfurther by Steiget al. [ 1998]. uncertainty in the gas age-ice age difference and difficulty in The ice age timescalefor Taylor Dome (used to plot isotope identifyingthe start of thewarming trend. New fi•SNresults from records in Figures 1 and 3) is based on the gas age timescale the B011ing-AllerOdtransition indicate that the methane rise described above. To determine ice ages we calculated A age began-20-30 yearsafter the temperaturerise [Severinghausand usingthe Herron and Langway [1980] model, with temperatures Brook, 1999]. At the beginning of interstadial 8 methane inferred from the oxygen isotope record and accumulationrates apparentlylead the changein b•80,•eby -100 years(Figure 2), inferredfrom measurementsof løBe [Steig,1996, Steig et al., again within the uncertaintyin the gas age-iceage difference. At 1998]. We addedA age to the gas agesto computeice age as a the beginning of interstadial 12 the apparentlead of methane function of depth. Uncertainties in the Taylor Dome ice age versusb•80,• (-300 years)is largerthan our estimateof Aage timescaleare larger than thosefor the gas age timescalebecause uncertainty. Clarification of the phasing of temperatureand they includeuncertainty in A age due to imperfectknowledge of methane at these transitions awaits further measurementsof b•SN past firn conditions (i.e., temperature, accumulation rate, of N 2, which will provideprecise information about their relative densificationhistory). However, in this paper we consideronly timing. Nonetheless,at all four transitionswe studiedin detail it the gas records,for which theseuncertainties are not critical. The appearsthat the rise in methane took place over one to three interpretation of the timing of events in the Taylor Dome centuries, slower than associatedchanges in climate inferred palcotemperature record and additional aspects of ice age fromb•80,•e records and records of snowaccumulation rate (R. timescalesare discussedelsewhere [Sucher, 1997' Steig et al., Alley, personalcommunication, 1999). This result is discussed 1998, 2000]. in section 4.1.3. BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD 563

800 -32 A B c

500-38 400 -40

300 -42 , I ' I ' I ' I ' iiiijllll 11.0 12.0 13.0 14.614.815.0 37.0 38.0 39.045.0 46.0 47.0 Age (ka) Age (ka) Age (ka) Age (ka) Figure 2. GISP2 methaneand isotope data for (A) YoungerDryas, (B) Bollingtransition, (C) Interstadial8, and (D) interstadial12. Methanedata are plotted as filled circles; •5•8Oice data are plotted as thin solid line. Apparent /lagsbetween methane and the isotoperecords are within the uncertaintyof the gasage timescale for GISP2, with the exceptionof Interstadial12 (Figure2d; seetext).

4.1.3. Submillennial methane variations. Our results also et al., 1995]. These have been correlatedwith the first phaseof indicate that between 20 and 10 ka, smaller scale methane the , the Older Dryas, and the InterAllerCd Cold variations occurred coincident with similar features in the ice Period (IACP) (Figure 3). In addition, there is a core•Sz8Oic • records (Figure 3). In theGISP2 •5•8Oice record, three smallnegative •18Oic eoscillation at -1 1.5ka (Figure3) in thepre- brief cold phaseshave been noted during this time period[Stuiver borealperiod (Preboreal Oscillation or PBO) that hasbeen noted

Older Younger Dryas Dryas Oldest PBO IACP , , Dryas -32 I ß . .

-34 ß -36 ß -38 8OO • ; ß

-40 : 700 • -42 ,. ß

.. 600 • ; JA.GsP2J ,. 500

ß ß •t•.t'• , 400 -o

. ß c:z 800 v 300 • 700 c- 600 - .c: 500 - o 400 - 300 - - -40 0

: ., !...i.. - -42 2, i ' - -44 ' • "'i *" * 10 12 14 16 18 2O Age (ka) Figure 3. (A) GISP2 and (B) Taylor Dome methane and isotope records for the time period 20-10 kyr and correlationof cold substagesin latestQuaternary with the GISP2 isotoperecord [Stuiver et al., 1995]. The cool periodin the early Preboreal(Preboreal Oscillation or PBO), which is associatedwith a smallminimum in methane, was not discussedby Stuiveret al. [ 1995] but is prominentin a numberof paleoclimaterecords (see text). The fact thatmethane apparently increased before the 818Oice temperature is anartifact of uncertaintyin thegas age-ice age difference(see text for furtherdiscussion). 564 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD in other high-latitudeNorthern Hemisphere paleoclimate records 100years. Additional •5•Soice data increase the resolution to - 40- [e.g., Lehman and Keigwin, 1992). The GISP2 methanerecord 50 yearsover this interval,but do not reveal a similaroscillation containssmall oscillationsthat appear to correspondwith the [M. Stuiver,personal communication, 1998]. However,a similar Oldest Dryas, IACP, and the PBO (Figure 3). The Older Dryas, oscillation is found in an alkenone-basedsea surfacetemperature the smallest and shortestof these four events, is not clearly proxy from a high resolutionmarine sedimentrecord at the evidentin the methanerecord. All of thesefeatures are relatively Bermuda Rise [Sachs and Lehman, 1999], suggestingthat this small, but similar minima that may correspondto the IACP and methaneoscillation may be the resultof a real climateevent. the PBO are possiblypresent in the Taylor Dome methanerecord It is importantto considerthe extent to which these data (Figure 3). The coincidenceof small minima in methanewith representthe true atmospherichistory during these short-term thesesmall isotopictemperature transients in GISP2 suggeststhat events. Two processesaffect the relationshipbetween the true climate may have changedover broad geographicareas during atmosphericmethane and the ice core methanerecord. theseevents, as during the larger interstadialevents. Hughen et First, gasesmix by diffusion in the firn layer. This process al. [1996] made a similar conclusion from analysis of varved smoothesthe atmospherichistory and has been modeled by sedimentsin the Cariaco Basin. They studiedhigh resolution Schwander et al. [1993; 1997], Battle et al. [1996], and records of sediment reflectivity (grey scale) which reflect Trudingeret al. [ 1997]. The age distributionof gasmolecules at variations in the organic carbon content of sedimentsand the the depth of gas trapping can be calculatedwith appropriate relative importanceof dry, upwelling conditionsand wet non- diffusion models. The width of the age distributiondepends on upwelling conditionsin the trade wind belt north of Venezuela. temperature,accumulation rate, and meteorologicalconditions These recordsare interpretedas proxiesof tropicalAtlantic trade that control the thickness of a convective zone in the upper wind strengthand precipitation.Hughen et al. found oscillations several meters of the firn [Schwander et al., 1997]. Under in the gray scale record corresponding to the four events modern conditionsat GISP2, the width of the age distributionat discussedabove, suggesting that the tropicalAtlantic experienced half-heightis - 10 years[Schwander et al., 1997]. greater trade wind activity and lower precipitationduring these Second, the bubble close-off processthat occurs when firn brief cold events. Drier conditionsin the tropics would likely transformsto ice is not instantaneousbut happensover a finite resultin lower methaneemissions from tropicalwetlands. depth interval, potentiallyintroducing additional smoothing of the gas record. A numberof studiesindicate that gas mobility 4.2. Rate and Magnitude of Methane Change During Rapid near the firn-ice transition is low owing to early close off of Climate Transitions wintertimelayers [Etheridgeet al., 1992;Martinerie et al., 1992; Schwander et al., 1993; Battle et al., 1996]. As a result, this Methane variations observed in ice cores on all timescales secondprocess is not expectedto add additionalsmoothing to the (with the exception of the recent industrial period) have a methane record at the GISP2 site [Schwanderet al., 1993; 1997]. maximum range of-400 ppb and have primarily been viewed as We calculatedthe smoothingeffect of firn diffusion using a tracers of temperature and precipitation in wetland regions numerical diffusion model [Severinghaus et al., 1998] that [Chappellazet al., 1990, 1993a;Brook et al., 1996a). Alternative calculates the concentration or isotopic composition of a gas views suggest that rapid, and large, atmospheric methane speciesat the depth of bubble close off. The model is time variations may have been caused by catastrophic release of dependentand includesthe effects of concentrationdiffusion, methane from methane hydratesin marine sedimentsand that these large emissions might have forced or amplified rapid gravitationalsettling, and thermal diffusion. Gas diffusivities climatechange [Nisbet, 1992, Kennettet al., 1996; Thorpeet al., were basedon a density-depthrelationship and empirical density- 1996]. Such eventscould have been rapid enoughthat they have diffusivity relation derived from CO2 measurementsin South not been observedat the resolutionof previousice core methane Pole firn air [Battle et al., 1996]. Methane diffusivity was records[Thorpe et al., 1996]. In this sectionand section4.3 we assumedto be 1.35 times the CO2diffusivity [Schwanderet al., use our results to examine these ideas. 1993]. We predictedthe agedistribution of air trappedduring the The four transitions we examined in detail have similar largeclimate shift at the end of the YoungerDryas (Figure4) by characteristics (Figure 2). Concentrations rose rapidly from fo•'cing the model with a delta function and examining the ""levels of 450-500 ppb to "interstadial"levels of 600-770 resulting time series of concentrationsat the lock-in zone. ppb (Figures 1 and 2). The transitionswere generally smooth, Temperatureand accumulationrate boundaryconditions were but some one-point reversalsare evident during the Younger basedon YoungerDryas conditionsas describedby Severinghaus Dryas and interstadial 8. Once interstadial conditions were et al. [1998]. The predictedage distributionat the baseof the reached, methane concentrationsstayed high for roughly the firn has a width at half-height of- 20 years (Figure 4). This durationof the warmingevent (with the exceptionof interstadial value is a conservative maximum because the model contained 12) (Figures 1 and 2). However, we note that methane and no near-surface convective zone; convection would reduce the isotopic temperaturedo not track each other perfectly. During width of the age distribution. the B011ing-Aller0dperiod, for example, temperaturesinferred In the GISP2 record, methane concentrations rose at the end of from•SOic e fell graduallyover the time period from 14.5to 13.0 the Younger Dryas, the start of the Bolling period, and at the ka, while methaneconcentrations rose gradually. We also note onsets of interstadials8 and 12, over time periods of 100-300 that during interstadial12, approximately300 years after the years (Figures 2 and 3). In contrast,the associatedtemperature beginningof the methanerise (Figure 2), there is a negative shiftsinferred from the •1801c e ice record(Figure 2), areseveral oscillationin methanethat lastedfor - 200 years. This feature decadeslong at most, and records of snow accumulationrate at does not seem to correspondto any significant event in the the GISP2 site also suggestextremely rapid changesin climate GISP2•5•8Oice record (Figure 2), whichhas a resolutionof - 80- for many of these transitions [Alley et al., 1993; R. Alley, BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD 565

0.05 et al., 1994;Allen et al., 1999]. The lag in methaneresponse may alsoreflect changes in the hydrologicbalance in wetlandregions at the beginningof interstadialevents. If climateshifted to wetter 004 conditionsat the beginningof interstadialevents, several decades might have elapsedbefore rising water tables could saturate surfaceor near-surfacesoils and promote anaerobic conditions. 4.3.2. Methane hydrates. We suspect that massive 003 destabilization of methane hydrates in marine sediments or permafrostwould likely leadto morerapid and larger changes in 002 the methane concentration than we observe in our records. For example,Thorpe et al. [1996], in a comprehensivetreatment of this problem,modeled a 4000 Tg (4000 Tg is roughly20 times 001 the pre-industrialmethane budget) burst of methanein the high- latitudeNorthern Hemisphere with a 2-D atmosphericcirculation o and chemistrymodel., They showedthat such an instantaneous release would cause a rapid rise in methane concentration 0 10 20 30 40 50 60 reachingmaximum levels within a few years. In their modelthis Gas Age Relative to Surface (yr) eventproduced atmospheric methane concentrations in excessof 2000 ppb in the NorthernHemisphere and locally in excessof Figure 4. Modeled age distributionof air trappedat the baseof 25,000 ppb immediatelyafter the event. Although such high the firn at the GISP2 site under Younger Dryas conditions. Age concentrationsare not observedin any existing ice core methane distribution was calculated with a tim diffusion model that is records,brief periodsof elevatedmethane concentration might described in the text. not have been recordedin previousrecords given their limited resolution(> 100 years)[Thorpe et al., 1996].

personalcommunication, 1999]. The data for all four transitions show that the time for methane to rise from stadial to interstadial 80O valuesis longerthan the width of the age distribution(-20 years). Therefore the ice core concentrationhistories presented here are 750 true representationsof atmospherichistory, with the exception that changesoccurring over one to two decadesor less are 700 smoothed(Figure 5). This conclusionis verified by the fact that faster,decadal variations of 151•Nof N2 in trappedair in GISP2 65O are preservedin ice of similar age [Severinghauset al., 1998; Severinghausand Brook, 1999]. 600 4.3. Origin of Rapid Changesin Methane? 550 4.3.1. Emission from terrestrial ecosystems. One explanation for the longer timescale for rising methane levels 5OO may be that changesin the temperatureand hydrologiccycle in tropical or temperateregions occurred on a slower timescalethan rapid temperature shifts recorded in high-latitude ice cores. 450 However, some tropical records suggestequally rapid climate -300 -200 -100 0 100 shifts at this time. For example, in their study of the Carioco Basin climate record (discussedin section4.1.3), Hughen et al. Time Before Increase (yr) [ 1996] concludedthat the event correlativeto the Younger Dryas Figure 5. Illustration of the effect of mixing in the firn on ice termination occurred in less than a decade. A secondpossible core methane record using the firn diffusion model of explanationfor the slower methane responseis that the century Severinghauset al. [1998] for conditions at the end of the timescale for the methanerise is related to ecological changes Younger Dryas. "Surface" indicates the hypothetical forced by climate change.Changes in methaneemissions from atmosphericmethane history; "lock in" indicates the methane wetlands during these climate transitionsmay have involved history recorded by the ice core. Smoothing of the ice core changesin primary productivity,species composition, and other methanerecord by diffusiondoes not significantlyinfluence the ecosystemdynamics that influence methanogenesis.The slower ice core recording of the magnitude or apparent speed of timescale for methane increases may represent a lag in the transitionsin atmosphericmethane concentration that take more than - 100 years. Thereforeapparent uniform transitionsof this responseof these processesto climate change [e.g., Prentice, lengthor longerrepresent the true rate of changeof atmospheric 1986; Webb, 1986; Campbell and McAndrews, 1993]. This methane at the surface of the ice core location. The offset hypothesis is supported by terrestrial evidence that indicates betweenthe hypothesizedatmospheric history and that recorded rapid, large changesin vegetationpatterns occurred in a variety by the ice core record is -10-20 years and representsthe time of regions,including the tropics,on centennialtimescales at the necessaryfor an atmosphericsignal to diffuseto the baseof the Younger Dryas, Bolling, and other interstadialtransitions [Lowe firn. 566 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

5000 latitude Northern Hemisphere at the initiation of interstadial events. However, clathrate bound methane might be released -r 4000 (D from other locations[Thorpe et al., 1996; Kennet et al., 1996]. .- 3000 While our resultsappear to rule out significantclathrate release in "' 2000 the Northern Hemisphere, a similar event in the Southern o • 1000 Hemisphere might not produce as dramatic a concentration history at the GISP2 site due to atmosphericmixing (Figure 6) 0 2000 [Thorpeet al., 1996]. The Taylor Dome methaneresults have a Max. Observed Ice lower resolution than the GISP2 data and cannot be used to CoreMethane ...-'"'"".,, _ 1500 entirely rule out a SouthernHemisphere or perhapsmid-latitude I ,:,.,.,.,...... 1000 clathraterelease. To addressthis issuewe employThorpe et al.'s predicted Southern Hemisphere atmosphericresponse to their 500 ...... -. clathrate release in the Northern Hemisphere (Figure 6) as a

o crude analog for the responseof the high-latitude Northern lOO 80 60 40 20 o Hemisphereto a clathraterelease at high southernlatitudes. For Model Time (yr) simplicity .we assumethat atmospheric mixing and chemical processesare latitudinallysymmetrical. The ice core responseto Figure 6. Model simulationsof instantaneousrelease of methane from clathratesto the atmosphereand the ice core responseto suchan eventwould be an apparentrise in methaneof- 15 year those events. The ice core response was calculated by duration(Figure 6), again,much faster than observedin our data. convolving the age distribution shown on Figure 4 with the As discussedby Chappellazet al. [1997], we cannotexclude a hypotheticalatmospheric . Thin dashedlines are step slower and smaller release of clathrate methane as the source for functions in the atmosphericmethane concentration,from 350 the atmosphericmethane concentration changes we observe. ppb to 1000, 2000, 3000, 4000, and 5000 ppb, from top to However, we can rule out rapid, large changes in methane bottom, respectively, assuming no change in atmospheric concentrationsthat would provide significant radiative climate methane lifetime (10 years) and simple exponential decay of forcing for the rapid changesin temperatureobserved at the initial concentration. Heavy solid line shows results from the beginning of interstadial events. Our data suggestthat the Thorpe et al. [ 1996] model of a clathraterelease event at 66øN in the NorthernHemisphere, thin solid line representsthe response changesin methaneconcentration in the ice core record (e.g., of the high latitude Southern Hemisphere(e.g., the latitude of shiftsfrom 350-400 ppb to 600-700 ppb) closelyrepresent the Taylor Dome) to thatevent [Thorpe et al., 1996]. true atmospherichistory. These changesare generallybelieved to be too small to contributesignificantly to the severaldegrees of warmingthat occurredat the beginningof interstadialevents In Figure 6 we plot hypothetical ice core concentration and the end of the Younger Dryas [Raynaud et al., 1988; histories arising from instantaneousreleases of methane to the Chappellazet al., 1993a]. atmosphere. To calculate these we smoothed hypothetical Finally, our results show that high methane concentrations atmospherichistories by convolution with the predicted age were maintainedthroughout the Preboreal,B½lling-Allered, and distribution shown in Figure 4. This age distribution was interstadial8 periods. In thesethree cases,century-long rises in calculatedfor conditionsat the end of the YoungerDryas period methane concentrationwere followed by periods of elevated and shouldapproximate the age distributionat other transitions. concentrationsthat lasted1000 yearsor longer,roughly the same We first plot five simple concentration histories based on lengthof time as the warmingevent (Figure 2). lnterstadial12 is instantaneous increases of methane levels to values of 1000-5000 an exception, as it consistsof the previously discussedbrief ppb, and subsequentdecrease in concentrationusing the current decreasein concentrationafter the initial warming,a subsequent atmosphericlifetime of l0 years. In reality, the lifetime would return to high concentrations,then a gradual decline. High- probablyincrease with the emissionof this muchmethane, owing resolutiondata for -500 yearsafter all of thesewarmings (Figure to a correspondingdecrease in atmosphericOH, the principle 2) showthat methaneconcentrations for the mostpart remained methane sink. Using the modern lifetime in our sensitivity within an envelopeof - 50 ppb during theseevents (again with calculations decreasesthe length of time that concentrations the exceptionof interstadial12), with temporalresolution as short remain high, providinga conservativecomparison with our data. as 20 yearsfor the YoungerDryas terminationand the beginning This simpleanalysis (Figure 6) showsthat the high-resolutionice of the B½lling-AllerOd.We find no evidenceof periodic,large core record is inconsistentwith the large and rapid change in injectionsof methanefrom clathratedestabilization during these atmosphericmethane concentrations that an instantaneousrelease time periods. would produce. Smoothingin the firn gives an instantaneous 4.4. Interpolar Gradient event an apparentduration of- 20 years,while the data show that methaneconcentrations rose over time periodsof 100-300 years The majority of modernnatural methane emissions originate at the four transitions we examined in detail. in two major latitudebands; a high-latituderegion between 50*- In Figure 6 we also plot the model atmospherichistory for the 70*N (primarily emissionsfrom boreal wetlands) and a low 4000 Tg releaseat high northernlatitudes (66'N) from Thorpeet latitude region in the tropicsof both hemispheres(primarily al., [ 1996], and our predictedice core response.Examination of emissionsfrom tropicalwetlands) [Fung et al., 1991;Hein et al., these results leads to a similar conclusion. The GISP2 methane 1997]. In the pre-industrialatmosphere (250-400 yearsago) the record is not consistent, in magnitude or in speed, with an methaneconcentration in the high-latitudeNorthern Hemisphere instantaneousrelease of large quantitiesof methaneto the high- was 8-10% higherthan in the high-latitudeSouthern Hemisphere BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD 567

[Rassmussenand Khalil, 1984; Nakazawa et al., 1993]. This Antarctic core sites during time periods when methane latitudinaldependence of the concentrationreflects a dominance concentrationswere relatively stable [Chappellaz et al., 1997]. of NorthernHemisphere sources in the preindustrialHolocene Doing so avoidsproblems with inaccuraciesin chronologywhich andthe fact that the atmosphericlifetime of methane(-10 years) would affect IPG most significantly during times of rapidly [Prinn et al., 1995] is of the same order as the mixing time changing concentrations. Time-weighted mean concentrations betweenhemispheres (--.1 year). The concentrationdifference were calculatedfor thesetime periodsfollowing Chappellazet al. betweenthe northernand southernhigh latituderegions can be [1997], and the IPG was calculatedfrom the appropriatepair of reconstructedfor past times from ice cores [Rasmussenand means (equation 1). The time periods examined, mean Khalil, 1984; Brook et al., 1996b; Chappellazet al., 1997] and concentrations, and IPGs are listed in Table 2. For the 5-7 ka has beencalled the interpolargradient (IPG). The IPG has been period during the mid-Holocene, the Taylor Dome timescaleis definedin a variety of different ways [Rasmussenand Khalil, not defined precisely. Therefore rather than using a time- 1984;Nakazawa et al., 1993; Chappellaz,1997]. Here we define weighted mean we chose a single value from both cores, the it as minimumconcentration measured for that period. IPG=(Cs/Cs-l) (1) The IPGs are plotted in Figure 7. Also shown are IPGs from the GRIP (Greenland), Byrd (Antarctica), and D47 (Antarctica) whereCs is the atmosphericconcentration in Greenlandand C s is ice core records,recallSulated from the data of Chappellazet al. the concentrationin Antarctica. Positivevalues indicate higher [1997]. Their results cover the later part of the Holocene and concentrationsin the Northern Hemisphere, negative values overlap our estimatesfor the mid-Holocene and Preboreal. Our indicatehigher concentrations in theSouthern Hemisphere. resultsfor the Preborealperiod are in good agreementwith those Figure 7 shows the methane recordsfor the last 40 ka from of Chappellazet al. [1997]. Our value for the mid-Holoceneis Taylor Dome and GISP2. To calculatethe IPG for pasttimes we very close to that determinedby Chappellaz et al. for the time determined average concentrations at the Greenland and period2.5-5 ka but is higherthan that for the precedingperiod.

700 ß • • -e- GISP2 I ' •' I I TaylorDome I 600

500 .,': .• •... .. , .. ..' •. N ß 400 ..,• .-' ,' '... ß

.

ß 300 .... [ .... I .... [ .... ] .... [ .... [ .... [ .... 0 5 10 15 20 25 30 35 40

14 .... I .... I .... I .... I .... I .... I .... I .... o Chappellazetal. (1997) B /t This Work

:-- 10-

• 8- - MH PB LGM o 6 o -• -- •' ' __B/A a IS•8

o 2 YD

0 5 10 15 20 25 30 35 40 Gas Age (kyr bp)

Figure 7. (A) Methanerecords for the GISP2 andTaylor Dome coresover the last 40 ka. (B) Mean interpolar gradients(IPG) for selectedtime periodsbased on the GISP2 and Taylor Dome data and alsoreproduced from Chappellazet al. [1997]. We defineIPG as [Cs/Cs]-l, whereCs is the NorthernHemisphere (Greenland) methane concentrationand Cs is the SouthernHemisphere (Antarctic) methane concentration. The lengthof the horizontal barsillustrates the time periodover whichthe meanvah]e is calculated,and the verticalrange represents the 2(7 uncertaintyin the mean. See text for details. 568 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

Table 2. WeightedMean MethaneConcentrations for InterpolarGradient Calculation. Time Interval

ka bp Greenland Antarctica IPG, %

Mid-Holocene* 5-7 579+_12 534+_11 8.4_+0.5 Numberof depths 1 1

Preboreal 9.5- l 1.5 720+-4 678+_4 6.2_+0.4 Number of depths 32 23

Younger-Dryas I 1.5-12.5 495+_5 476+_6 3.9-!-_1.0 Number of depths 19 12

B olling-Allemd 13.5-14.8 646+_6 625+_9 3.2+_1.0 Number of depths 41 12

Last Glacial Max 18-26 395+_5 377+_7 4.8+_1.4 Number of depths 15 11

Interstadial 8 37-39 576+_5 533+_ 10 8.0+_2.0 Numberof depths 22 4

The error for the methaneconcentration is I weightedstandard deviation of the mean as definedby Chappellazet aL [1997]. The error for the IPG is the 95% confidenceinterval. *Thesevalues are not time-weighted means but are the minimum values for thisperiod. Uncertainty based on 2% error (generalagreement between replicate ice samples)for eachmeasurement.

Comparison here is difficult due to dating uncertainties for time (or if theirtemporal variations are known), a box modelmay Taylor Dome. The resultsshow that the IPG was generallylower be usedto estimatechanges in thesource distribution [Khalil and during the (LGM), B011ing-Aller0d,and Rassmussen,1983; Rassmussenand Khalil, 1984; Chappellazet YoungerDryas, suggestinga dominanceof tropical sources.The al., 1997]. The modeldivides the troposphereinto threeboxes, IPG increasedat the beginningof the Preborealperiod (Figure 7), the North(N; 300-900N),the Tropics(T; 30øS-30øN),and the suggesting that northern sources became more significant South(S; 30ø-90øS).Boxes N andS represent25% eachof the immediately after the end of the Younger Dryas. The value for totaltropospheric volume, and box T represents50% of the interstadial8 (8 + 2%) is similar to that duringthe mid-Holocene, volumeof thetroposphere. The massbalance is describedby suggestingsubstantial Northern Hemispheresources during this event, which occurred during the middle portion of the Marine dC/dt=F-•C (2) IsotopeStage 3 (Table 3). This resultis preliminary owing to the limited data from the Taylor Dome core. Variations in the IPG where C (Tg box-l) is a vectorrepresenting the atmospheric during the Holocene are discussedin detail by Chappellaz et al. methaneconcentration in eachbox and F (Tg yr-•) is a vector [1997]. representingthe methanesource in eachbox. The sink(primarily oxidationby OH in the troposphere)is representedby •C where 4.5. Implications of IPG for Methane Source Distribution

The IPG is controlledby the latitudinaldistribution of methane sources, the latitudinal distribution of methane sinks, and the 3,N+n N -nN/2 0 (3) mixing rateswithin and betweenhemispheres. If the sink distributionand mixing rate are assumedto remainconstant over f•(yr-1)=[-;N3'T +(nN-nS +ns)/2! 2 3, S-ns + n S

Table 3. Methane EmissionsCalculated From Three Box Model Using Taylor Dome and GISP2 Measurements (Table 2)

Time Interval, MethaneEmissions (Tg/yr)* Period ka bp North Tropics South+ Total

Mid-Holocene 5-7 65+ 14 79+-23 15 159 Preboreal 9.5-11.5 64+_5 123+_8 15 202 Younger-Dryas 11.5-12.5 39+_6 86+_11 15 141 Bolling-Allermd 13.5-14.8 43+7 127+_15 15 185 Last Glacial Max 18-26 34+_6 65_+9 12 l I l Interstadial 8 37-39 63+_7 80+-11 15 158

*Theerror for the source estimate is the 95 % confidenceinterval based on propagating errors from Table 2. +Southernsource is fixed (see text). BROOKET AL: RAPID METHANE VARIATIONSDURING THE LAST GLACIAL PERIOD 569

The coefficientfor removal,• (yr4),is definedas lit wheret is B•lling-Aller•d period, despite the -- 200 ppb difference in themethane lifetime (year). The transportcoefficient, n (yr'•), is concentration (Figure 7, Table 3). This suggestsa doubling defined as l/x, where x is the transporttime between boxes. (roughly)of tropicalsource strength (Table 3), with little change Assuming steady state, and with knowledge of t, x, C, and the in boreal sourcestrength. Examination of the data in Figure 7, volumes of the boxes, the latitudinal source distribution can be however,suggests that during the 2 ka period immediatelyprior calculatedfrom (1). We used valuesfor theseparameters from to the B•!!ing increase,the interpolar gradient may have been Chappellazet al. [ 1997] to make our resultscomparable with reversed (e.g., Southern Hemisphere concentration higher). their calculationsfor later time periods. (The valuesare 9 months Although our data are insufficient for a precise gradient for xN and Xs, and 15.6, 6.8, and 22.4 years for tN, ta-and ts, calculationfor this time period,recent work by Dallenbachet al. respectively). [2000] confirms this suggestion and indicates that tropical The methane concentration in box S is considered sourcesincreased by - 40 Tg yr4 at the B•lling transitionand homogeneousand is thereforewell representedby measurements borealsources increased by - 30 Tg syr-•. The totalmethane in the Antarctic core. Box N is not well mixed in the modern sourceimmediately after the B•lling transitionwas - 175-185 Tg atmosphere. For example, in the compilation of Fung et al. yr-z by theircalculations, similar to ourvalue of 185Tg yr'•. [ 1991], the methaneconcentration increases linearly by - 60 ppb A global decreasein methanesource strength occurred during between30øN and 90 ø N. The gradientwithin box N is at least the Younger Dryas, followed by an increase in Northern partly a resultof anthropogenicmethane sources in the Northern Hemisphereand tropical sourcestrength at the beginningof the Hemisphere,and must have been lower in the past. Following Preborealperiod (Figure 7). The large difference in the IPG Chappellazet al. [1997] we assumethat the differencebetween between the B•lling-Aller•d and Preboreal periods is the most the concentrationat 70øN and the box N averageis 7% of the robustof our results,owing to the large numberof data pointsin total interhemisphericdifference. Therefore,to calculateaverage both periods. It requires that tropical methane sourceswere concentrations for box N from Greenland concentration data, we similar in magnitude(Table 3) but that boreal sourceswere about decreasethe Greenlandconcentration by 7% of the GISP2-Taylor 50% higher during the Preboreal period. The results further Dome difference. In reality, the 7% value may have changed suggestthat expansionof Northern Hemispheresources occurred with time, but we are currently unable to constrain this effect at the very beginningof the Preborealperiod (Figure 7). At the accurately, and the correction is small in any case. The start of the Preboreal period substantial portions of eastern concentration for box T is not known but is determined in the Canada and Fennoscandia remained ice covered [Peltier et al., modelby fixingthe southernsource at 15 Tg yr-• (12 Tg yr-• for 1994]. Nonetheless, methane emissions from Northern the LGM) [Chappellaz et al., 1997]. These values may be Hemispheresources must have expandedrapidly at the end of the slightly too high. Fung et al. [ 1991] indicatethat natural sources Younger Dryas, perhapsdue to formation of productivewetlands in this latitude band today sum to -.- 11 Tg/yr, and Hein et al. from high-latituderegions which were already ice free. The new [1997] derivea value of-13 Tg yr'z. However,such source methane sourceareas may have been in regions of Siberia that estimatesare difficult to constructand may have been higher remained ice free during the glaciation or other areas that prior to the industrial revolution. As long as this value is held deglaciatedearly. The methaneresults imply that theseregions constant, relative differences in source distribution are not experiencedwarm and wet conditionsat this time, consistentwith significantly affected, but a real changein the southernsource other evidence [e.g., Ritchie et al., 1983; Crowley, 1991]. In (for examplebetween the Younger Dryas and Preborealperiods) summary,regardless of the specific location of these northern would not be representedin our calculations. For example, sources,the interpolar gradient data (Figure 7; Tables 2 and 3) considering the Younger Dryas-Preboreal transition, using a suggestthat tropicalsources were a significant,though variable, valueof 10 Tg yr4 ratherthan 15 Tg yr4 for thetarget southern componentof the methane budget at all time periods over the sourcechanges the northern source by -5 Tg yr4 andthe tropical past50 ka. sourceby +10 Tg yr'•. Thesechanges are smallrelative to the The simple model used here does not account for possible magnitudeof currentuncertainties (Table 3). changesin atmosphericmixing or chemical removal of methane As we have no direct constraints for a number of the from the atmosphere,both of which may have changedover the parameters used in these calculations (e.g., transport time, past 50 ka. For example, the methane lifetime during the lifetimes, and tropicalconcentration), the model resultscan only preindustrialperiod, and particularly the glacial period, is not be usedfor a first-orderinterpretation of changesin the IPG. We well constrainedand may have been shorterthan assumedin the use this model with the GISP'2 and Taylor Dome methanedata to model [Thompson, 1992]. Also, numerous studies have estimate methane emissions for boxes N and T in the mid- suggestedthat wind speedswere higherin glacialtimes, possibly Holocene,the Preborealperiod, the Younger Dryas, the Bolling- resultingin greaterinterhemispheric mixing. We briefly explore Allerod, the LGM, and for a preliminaryestimate for intersradial the sensitivityof the model to these changes(Figure 8). For a 8 (Table 3). For the Preboreal period our sourceestimates given interpolargradient, invoking faster atmosphericmixing overlapwith thoseof Chappellazet al. [ 1997] (Figure 7) and are (shortertransport time) increasesthe calculatednorthern source not statisticallydifferent (at 95% confidenceusing Studentst and decreases the calculated tropical source (Figure 8). test) from theirs. Our results for the mid-Holocene differ from Decreasingmethane lifetime increasesinferred source strength in those of Chappellazet al. from adjacent time periods but are northernor tropicalboxes (Figure 8), althoughthe effectis larger within the range established by their data. However, as in the tropics due to the shorter lifetime. Consideringthe mentionedabove, the chronologyof the Taylor Dome core is not transitionfrom the Younger Dryas to the Preborealperiod, for adequatelyestablished in this time period. example,a decreasein atmosphericmixing that was unaccounted The geographicdistribution of methaneemissions during the for in the model would result in an overestimate of changesin period we designatedas LGM was similar to that during the northern source strengthand an underestimateof changesin 570 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

A i i ! I i i 400 Northern Box Z • Mid Holocene - 300 o • Preboreal - ::3' ..... Younger Dryas 200 '" --- Bolling-Allerod ...... LGM ,,,,. o 100 c

.

. 0 '• - ,•

.

.

.

. /! ß o -100 - (/) • . (• -200

o -300 [ i i i [ ! ! I ' ' ' ' I ' ' ' ' I ' ' ' ' 0 5 10 15 20 25 TransportTime (months) Removal Time (years)

14.5 15.0 15.5 16.0 16.5 17.0 z I , , , , I , , , , I , , , I , , i I I i i i i 150 = -- Mid Holocene :3' B (1:) •- Preboreal 125 • ...... Younger Dryas ß 100 •o Bolling-Allerod c- ...... LGM 75 '" ...... (1:) 50 -•

,so! o "'"-:""="'•'"•'"•'"•"•"'•"":'"--""-':"'------•--•.2: .• 125 • lOO (D ' o 75 o 50 03 : -• 25' o

o •'''•''"'1 .... • .... I .... / .... 6.4 6.6 6.8 7.0 7.2 Removal Time (years)

Figure 8. Dependenceof modelderived source strength in northernand tropicalboxes on (A) transporttime betweenboxes and (B) methaneremoval time. Model calculationsused to producethese curvesheld methane emissionsin thesouthern box at 15Tg yr-] (12 Tg yr4 for LGM, seetext). Verticallines indicate the values of these parametersused in the three-boxmodel. tropical source strength. An increasein removal time for the higherIPG valuesthan observed. Our resultsand recentrelated same transition would result in an overestimate of changes in work [Severinghauset al., 1998; Severinghausand Brook, 1999], sourcestrength in both boxes,but the effect is more significantin also indicate that at least some of the increases in methane the tropics. How either variable changedin the past is not well concentrationoccurred nearly synchronouslywith the onset of known, but within a reasonablylarge range of mixing times and rapid warming eventsin the GISP2 stableisotope record. The removal times the first-order results of our study are not grossly magnitudeand rate of changeof methaneconcentrations appear affected(Figure 8). to be inconsistentwith a release of large quantitiesof methane from clathrates.

5. Conclusions Acknowledgments. We thank Melissa Swansonand Joe Orchardo for laboratoryassistance, Todd Sowers,Jerome Chappellaz, and Thomas Models constrainedby ice core methanemeasurements in both Blunier for sharing methane data and Richard Thorpe for sharing hemispheresfor the time period 5-40 ka supportthe suggestion atmosphericmodeling results. Commentsof two anonymousreviewers helped clarify many of our arguments. This work was supportedby that changesin both tropical and boreal sourcesare requiredto NOAA Climate and Global Change PostdoctoralFellowships to Ed explain observed rapid changes in methane concentration. Brook, Susan Harder, and Jeff Severinghaus, and grants from the Changesin only Northern Hemisphere sourceswould result in NationalScience Foundation including OPP-9714687 and OPP-972591. BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD 571

References

Allen, J.R.M., et al., Rapidenvironmental changes in southernEurope Maasch, A 50,000 year record of climatic oscillationsfrom Florida duringthe last glacial period, Nature, 400, 740-743, 1999. and its temporal correlationwith the Heinrich events, Science, 261, Alley, R., et al., Abruptincrease in Greenlandsnow accumulationat the 198-200, 1993. end of the YoungerDryas event, Nature, 362,527-529, 1993. Grootes,P.M. and E.J. Steig, Taylor Dome ice core study 1993/1994: An Barnola,J.M., P. Pimienta,D. Raynaud,and Y.S. Korotkevich,CO2- ice core to bedrock, Antarct. J. U.S., 29, 79-81, 1994. climate relationship as deduced from the Vostok ice core: A re- Grootes, P.M., M. Stuiver, J.W.C. White, S. Johnsen, and J. Jouzel, examination based on new measurements and on a re-evaluation of the Comparison of oxygen isotope records from the GISP2 and GRIP air dating,Tellus, Ser. B, 43B, 83-90, 1991. Greenland ice cores, Nature, 366, 552-554, 1993. Bender, M., T. Sowers, M.L. Dickson, J. Orchardo, P. Grootes, P. Hein, R., P.J. Crutzen, and M. Heinmann,An inversemodeling approach Mayewski and D. Meese, Climate connectionsbetween Greenlandand to investigate the global atmospheric methane cycle, Global Antarcticaduring the last 100,000years, Nature, 372, 663-666, 1994. Biogeochem.Cycles, 11, 43-76, 1997. Blunier, T., J.A. Chappellaz,J. Schwander,J.M. Barnola,T. Despens,B. Herron, M. M., and C. C. Langway, Firn densification:An empirical Stauffer, and D. Raynaud, Atmospheric methane record from a model, J. Glaciol., 25, 373-385, 1980. Greenlandice core over the last 1,000 years, Geophys.Res. Lett., 20, Hughen,K.A., J. T. Overpeck,L.C. Peterson,and S. Trumbore,Rapid 2219-2222, 1993. climate changes in the tropical region during the last Blunier, T., J. Chappellaz,J. Schwander,B. Stauffer, and D. Raynaud, deglaciation,Nature, 380• 51-54, 1996. Variations in atmosphericmethane concentration during the Holocene Johnsen,S.J., H. B. Clausen,W. Dansgaard,K. Fuhrer,N. Gundestrup,C. ,Nature, 374, 46-49, 1995. U. Hammer, P. Iverson, J. Jouzel, B. Stauffer, and J. P. Steffensen, Broecker, W. S., D. Peteet, I. Hajdas, J. Lin, and E. Clark, Antiphasing Irregular glacial intersradials recorded in new Greenland ice core, between rainfall in Africa's rift and North America's Great Nature, 359, 311-313, 1993. Basin, Quat. Res., 50, 12-20, 1998. Jouzel,J., et al., Extendingthe Vostokice-core record of paleoclimateto Brook, E., T. Sowers,and J. Orchardo,Rapid variationsin atmospheric the penultimateglacial period, Nature, 364, 407-412, 1993. methane concentrationduring the past 110,000 years, Science,273, Kennett,J.P., L.L. Hendy, and R.J. Behl, Late Quaternaryforaminiferal 1087-1091, 1996a. carbonisotope record in SantaBarbara Basin: implications for rapid Brook, E.J., J.P. Severinghaus,M. Swanson,and P. Grootes,Late glacial climatechange, EOS, Trans.AGU, 77(46), Fall Meeting Supplement, methane variations in Greenland and Antarctic ice cores, EOS, Trans. F294, 1996. AGU, 77(17), SpringMeeting Supplement, S156, 1996b. Khalil, M.A.K. and R.A. Rassmussen,Atmospheric methane trends over Brook, E.J., J. Severinghaus,S. Harder, and M. Bender, Atmospheric the last 10,000 years,Atmos. Environ., 21, 2445-2452, 1987. methane and millennial-scale climate change, in Mechanisms of Lehman, S., and L. Keigwin, Sudden changes on North Atlantic Global Climate Changeat Millennial Timescales,Geophys. Monogr. circulationduring the last deglaciation,Nature, 356, 757-762, 1992. Ser., vol. 112, edited by P. U. Clark, R. S. Webb, and L. Keigwin, pp. Lowe, J., B. Ammann, H. H. Briks, S. Bjorck, G. R. Coope, L C. 165-176, AGU, WashingtonD.C., 1999. Cwynar, J. L. de Beaulieu, R. J. Mort, D. M. Peteet, and M. J. C. Bubier, J., and T.R. Moore, An ecological perspective on methane Walker, Climatic changesin areas adjacent to the North Atlantic emissions from northern wetlands, Trends Ecol. and Evol., 9, 460- during the last glacial-interglacialtransition (14-9 ka); A contribution 465, 1994. to IGCP-253, J. ofQuat. Sci.,9, 185-198, 1994. Campbell, I.D., and J.H. McAndrews, Forest disequilibriumcaused by Martinerie, P., D. Raynaud,D. M. Etheridge,J.-M. Barnola, and D. rapid Little Ice Age cooling,Nature, 366, 336-338, 1993. Mazaudeir,Physical and climatic parameterswhich influencethe air Chappellaz, J., J.M. Barnola, D. Raynaud, Y.S. Korotkevich, and C. contentof polar ice, Earth Planet. Sci. Lett., 112, 1-13, 1992. Lorius, Atmospheric methane record over the last climatic cycle Meese, D., R. Alley, T. Gow, P. Grootes,P. Mayewski, M. Ram, K. revealedby the Vostok ice core,Nature, 345, 127-131, 1990. Taylor, E. Waddington,and G. Zielinski, Preliminarydepth-age scale Chappellaz,J., T. Blunier, D. Raynaud,J.M. Barnola, J. Schwander,and of the GISP2 ice core, CRREL Spec.Rep. 94-1, Cold Reg. Res. and B. Stauffer, Synchronouschanges in atmospheric methane and Eng. Lab., Hanover, N. H., 1994. Greenlandclimate between40 and 8 kyr BP, Nature, 366, 443-445, Nakazawa, T., T. Machida, M. Tanaka, Y. Fujii, S. Aoki, and O. 1993a. Watanabe,Differences of the atmosphericCH 4 concentrationbetween Chappellaz, J.A., I.Y. Fung and A.M. Thompson, The atmospheric the Arctic and the Antarcticregions in pre-industrial/pre-agricultural methaneincrease since the last glacial maximum, 1, Sourceestimates, , Geophys.Res. Lett., 20, 943-946, 1993. Tellus, Ser. B, 45B, 228-241, 1993b. Nisbet,E.G., Sourcesof atmosphericmethane in early postglacialtime, J. Chappellaz, J., T. Blunier, S. Kints, B. Stauffer, and D. Raynaud, Geophys.Res., 97, 12,859-12867, 1992. Variations of the Greenland/Antarctic concentration difference in Peltier,R., Ice agepaleotopography, Science, 265, 195-201, 1994. atmosphericmethane during the last 11,000 years, J. Geophys.Res., Petit, J. R., et al., Climate and atmospherichistory of the past 420,000 102, 15,987-15,997, 1997 years from the Vostok ice core, Antarctica, Nature, 399, 429-436, Crowley, T., Ice-age methanevariations, Nature, 353, 122-123, 1991. 1999. D'allenbach,A., T. Blunier, J. Fltickiger,B. Stauffer, J. Chappellaz,and Petit-Maire, N., M. Fontugne,and C. Rouland,Atmospheric methane D. Raynaud, Changes in the atmospheric CH4 gradient between ratio and environmentalchanges in the Saharaand Sahel during the Greenland and Antarctica during the Last Glacial and transitionto last 130 kyrs, Palaeogeogr.,Palaeoclimatol., and Palaeocol.,86, 197- the Holocene,Geophys. Res. Lett., in press,2000. 204, 1991. Etheridge,D.M., G.I. Pearman,and P.J. Fraser,Changes in tropospheric Prentice,I.C., Vegetationresponses to pastclimatic variation, Vegetatio, methane between 1841 and 1978 from a high accumulationrate 67, 131-141, 1986. Antarctic ice core, Tellus, Set. B, 44B, 282-294, 1992. Prinn, R.G., R.F. Weiss, B.R. Miller, J. Huang, F.N. Alyea, D.M. Fuchs,A., J. Schwander,and B. Stauffer,A new ice mill allowsprecise Cunnold, P.J. Fraser, D.E. Hartley, and P.G. Symonds,Atmospheric concentrationdetermination of methaneand mostprobably also other trends and lifetime of CH3CC13 and global OH concentrations, tracegases in the bubbleair of very smallice samples,J. Glaciol.,39, Science, 269, 187-192, 1995 199-203, 1993. Rasmussen,R. A., andM.A.K. Khalil, Atmosphericmethane in the recent Fung, I., J. John, J. Lerner, E. Matthews, M. Prather, L.P. Steele, and P.J. and ancientatmospheres: concentrations, trends, and interhemispheric Fraser, Three-dimensionalmodel synthesisof the global methane gradient,J. Geophys.Res., 89, 11,599-11,605,1984. cycle, J. Geophys.Res., 96, 13,033-13,065, 1991. Gasse,F. and E. V. Van Campo,A 40,000-yr and diatomrecord Raynaud, D., J. Chappellaz, J.M. Barnola, Y.S. Korotkevich, and C. from Late Tritrivakely,Madagascar, in the southernTropics, Quat. Lorius,Climatic and methane cycle implications of glacial- Res., 49, 299-311, 1998. methanechange in the Vostokice core,Nature, 333, 655-657, 1988. Grigg,L.D., andC. Whitlock,Late-glacial vegetation and climate change Ritchie, J.C., L.C. Cwynar, and R.W. Spear,Evidence from north-west in westernOregon, Quat. Res.,49, 287-298, 1998. Canada for an early Holocene Milankovitch thermal maximum, Grimm, E. C., G. L. Jacobsen,W.A. Watts, B.C.S. Hansen, and K.A. Nature, 305, 326-328, 1983. 572 BROOK ET AL: RAPID METHANE VARIATIONS DURING THE LAST GLACIAL PERIOD

Sachs,J. P. and S. J. Lehman,Subtropical North Atlantictemperatures Stuiver,M.,P.M. Grootes, andT.F. Braziunas, TheGISP2 õ18 0 climate 60,000 to 30,000 yearsago, Science, 286, 756-759, 1999. record of the past 16,500 years and the role of the sun, ocean and Schlesinger,W.H., Biogeochernistry:an analysisof global change,588 volcanoes,Quat. Res., 44, 341-354, 1995. pp., Academic,San Diego, Calif., 1996. Sucher, C.M., Atmospheric gases in the Taylor Dome ice core: Schwander,J., J.-M. Barnola,C. Andri6, M. Leunberger,A. Ludin, D. Implicationsfor East Antarctic climate change,M.S. thesis,Univ. of Raynaud,and B. Stauffer,The age of the air in the firn and the ice at R.I., Narragansett,1997. Summit,Greenland, J. Geophys.Res., 98, 2831-2838,1993. Thompson, A., The oxidizing capacity of the Earth's atmosphere: Schwander,J., T. Sowers, J.-M. Barnola, T. Blunier, A. Fuchs, and B. Probablepast and futurechanges, Science, 256, 1,157-1,165,1992. Malaize, Age scale of the air in the summitice: implicationfor Thorpe,R.B., K.S. Law, S. Bekki, J.A. Pyle, andE.G. Nisbet,Is methane- glacial-interglacialtemperature change, J. Geophys. Res., 102, driven deglaciationconsistent with the ice core record?,J. Geophys. 19,483-19,493, 1997. Res., 101, 28,627-28,635, 1996. Severinghaus,J. and E. Brook, Abrupt climate changeat the end of the Trudinger, C.M., I.G. Enting, D.M, Etheridge, R.J. Francey, V.A. lastglacial period inferred from trappedair in polarice, Science,286, Levchenko, and L.P. Steele, Modeling air movement and bubble 930-934, 1999. trappingin fire, J. Geophys.Res., 102, 6747-6763, 1997. Severinghaus,J.P., T. Sowers,E.J. Brook, R.B. Alley, and M.L. Bender, Webb, T., Is vegetationin equilibrium with climate? How to interpret Timing of abruptclimate changeat the end of the YoungerDryas late Quaternarypollen data, Vegetatio,67, 75-91, 1986. intervalfrom thermallyfractionated gases in polar ice, Nature, 391, Whiting, G.J., and J.P. Chanton,Primary productioncontrol of methane 141-146, 1998. emission from wetlands, Nature, 364, 794-795, 1993. Sowers,T., et al., An interlaboratorycomparison of techniquesfor extractingand analyzingtrapped gases in ice cores,J. Geophys.Res., 102, 26,527-26,538, 1997. Stauffer,B., E. Lochbronner,H. Oeschger,and J. Schwander,Methane E. Brook and S. Harder, Departmentof Geology, WashingtonState concentrationin the glacial atmospherewas only half that of the University, Vancouver, WA 98686. ([email protected]) preindustrialHolocene, Nature, 332, 812-814, 1988. J. Severinghaus,Scripps Institution of Oceanography,University of Steig, E.J., -10 in the Taylor Dome ice core: applicationsto California, San Diego, La Jolla, CA 92093. Antarctic and ,Ph.D. thesis, Univ. of E. Steig, Departmentof Earth and EnvironmentalScience, University Wash., Seattle, 1996. of Pennsylvania,Philadelphia, PA 19104. Steig, E.J., E.J. Brook, J.W.C. White, C. Sucher, S.J. Lehman, M.L. C. Sucher,NOAA Office of Global Programs,1100 Wayne Avenue, Bender, D.L. Morse, and E.D. Waddington,Synchronous climate Suite 1210, Silver Spring,MD 20910. changesin Antarcticaand the North Atlantic duringthe last glacial- interglacialtransition, Science, 282, 92-95, 1998. Steig,E.J., D.L. Morse,E.D. Waddington,M. Stuiver,P.M. Grootes,and P.A. Mayewski,Wisconsinian and Holoceneclimate history from an ice core at Taylor Dome, western Ross Embayment,Antarctica, (Received June 1, 1999; revised October 26, 1999; Geogr.Ann., in press,2000. acceptedNovember 2, 1999.)