The Dole Effect and Its Variations During the Last 130,000 Years As Measured in the Vostok Ice Core

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 8, NO. 3, PAGES 363-376, SEPTEMBER 1994

The Dole effect and its variations during the last 130,000 years as measured in the Vostok ice core

Michael Bender • and Todd Sowers 2 GraduateSchool of Oceanography,University of RhodeIsland, Kingston

Laurent Labeyrie Centredes Faibles Radioactivites, Centre National de la RechercheScientifique, Commissariat a L'EnergieAtomique, Gif-sur- Yvette, France

Abstract. We review the currentunderstanding of the Dole effect (the observeddifference between the •5180of atmospheric0 2 andthat of seawater)and its causes, extend the record of variationsin theDole effectback to 130kyr before present using data on the •5180 of 0 2obtained from studying the Vostok ice core(Sowers et al., 1993), anddiscuss the significanceof temporalvariations. The Dole effectreflects oxygenisotope fractionation during photosynthesis, respiration, and hydrologic processes (evaporation, precipitation,and evapotranspiration). Our bestprediction of the present-dayDole effect,+20.8 %0,is considerablylower thanthe observedvalue, + 23.5 %0,and we discusspossible causes of this discrepancy. During the past 130 kyr, the Dole effecthas been 0.05 %0lower thanthe present value, on average.The standarddeviation of the Dole effect from the meanhas been only _+0.2 %0,and the Dole effect is nearly unchangedbetween glacial maxima and interglacial periods. The smallvariability in theDole effect suggeststhat relative rates of primaryproduction in theland and marine realms have been relatively constant.Most periodicvariability in the Dole effectis in the precessionband, suggesting that changes in thisglobal biogeochemical term reflects variations in low-latitudeland hydrology and productivity or possiblyvariability in low-latitudeoceanic productivity.

Introduction they can potentially be made to do so. Ice core reconstructionsof the concentrationsof bioactivegases in air Our understanding of the response of the biosphere to provide an integratedglobal signal which complementsthe Pleistoceneclimate change is actually quite limited. For the proxy studies cited above. Variations in the concentrationsof terrestrial realm, our most detailed knowledge comes from themost abundant bioactive gases (CO 2, CHn, N20 ) reflect,in extensive studiesof pollen in sedimentswhich record climate part, global scale changes in the marine and terrestrial during and after the last glacial termination [e.g., COHMAP, biospheres.Atmospheric concentrations of CHn havebeen 1988]. Pollen data for earlier times are extremely informative found to vary with a period correspondingto that of the but geographicallyrestricted. In the marine realm, the most precession of the equinoxes [Chappellaz et al., 1990]. ambitious attempts to understand the response of the Concentrationsare higher when temperaturesare warmerand biosphere to climate have focused on studies of primary precipitationis greater. Glacial-interglacialvariations in the productivity in the past inferred from two proxy indicators: concentrationof CO2 havebeen attributed to changesin the organic carbon accumulation rates [e.g., Sarnthein et al., fertility and carbonexport of the upperwater ecosystem [e.g., 1988, 1992] and foraminiferal taxonomy [Mix, 1989a, b]. Mix, 1989a; Knox and McElroy, 1984], althoughvariations Paleoproductivity estimates made from many other proxies in oceancirculation and otherfactors must also have played an have enlivened the discussionwith respectto specific areasof important role (e.g., Boyle, 1988; Broecker, 1989). On a the ocean. shortertimescale, the transientminimum in the CO2 Proxy studiesof the responseof the terrestrial and marine concentrationof air at the end of the last glacial termination biosphere are limited in that they do not generally provide reflects rapid growth of the land biosphere[Neftel et al., information on a global scale, although with enough effort 1988]. Theb•80 of O2 is an additionalvariable which reflects the 1Alsoat Centredes Faibles Radioactivites, CNRS-CEA, Gif-sur- globalresponses of the land and marinebiospheres to climate Yvette, France. change,albeit in a complexmanner. The Dole effectis defined 2Presentlyat Lamont-Doherty Earth Observatory of Columbia asthe difference between the b•80 of atmospheric02 in airand University,Palisades, NY. theb•80 of contemporaneousseawater. The magnitudeof the Dole effect, which today is about +23.5 %o [Kroopnick and Copyright1994 by the AmericanGeophysical Union. Craig, 1972],mainly reflects the isotopic composition of 02 producedby marine and terrestrialphotosynthesis, as well as Paper number94GB00724. the extentto which the heavy isotopeis discriminatedagainst 0886-6236/94/94GB-00724510.00 during respiration. As previously discussed, /5180 of 364 BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE

atmospheric02 variedover glacial/interglacialtimescales in The5180 of 02 producedby photosynthesisis similar to responseto changesin the •5180of seawater[Horibe et al., thatof the sourcewater [e.g., Guy et al., 1993]. The •5180of 1985; Bender et al., 1985, Sowers et al., 1993]. Other factors, 0 2 producedby marineplants today is thus0 %o.The •5180 of whichare of secondaryimportance in controllingthe •5•80of 0 2 producedon the continentshas been estimatedto lie atmospheric0 2, are (1) changesin terrestrialand marine between+ 4 and + 8 %o[Dongmann, 1974; Farquhar et al., fertility, (2) varying isotope fractionation associatedwith the 1993]. These elevated •5180values are the result of elevated hydrologic cycle, and (3) changes in respiratory isotope leaf water•5180 values resulting from evapotranspiration. effects on either a speciesor communitylevel. The•5180 of 02 consumedbyrespiration is ~ 20 %oless than The responsetime of •5•80of 02 is comparableto the thatof the source02 [Guyet al., 1993;Kiddon et al., 1993]. turnovertime of 02 in air with respectto photosynthesisand At isotopicsteady state, the •5•80of 02 moleculesactually respiration,about 1.2 kyr (Table 1). In general,increases in consumedby respirationmust be thesame as the •5•80 of 02 the ratio of terrestrialto marineproduction will causethe •5•80 producedby photosynthesis[Lane and Dole, 1956]. Today of atmospheric02 to rise. Withinthe terrestrial realm, large- thisconstraint is satisfied with •5•80 of atmospheric02 = + scale changesin ecology, precipitation,and humidity can all 23.5 %o. In the eventof deviations,the steadystate value of causevariability in the Dole effect. the Dole effect would be restoredwith an e-foldingtime equal In this paper we present a record of variations in the Dole to theturnover time of 02 in air, ~ 1.2kyr. effect over the last 130 kyr, based on the analysis of air We now recognizethat severaladditional processes affect occludedin the GISP2 and Vostok ice cores. We then explore the•5180 of atmospheric02 [e.g.,Berry, 1992]. Themost the implicationsof this record for changesin the fertility of importantis evapotranspiration,which can causevery large the land and marine biospheresand changesin hydrologic enrichmentsin the •5•80of leaf waterof landplants because of fractionationof oxygen isotopesduring the Late Pleistocene. the preferrentialevaporation of the light isotope[Dongrnann et al., 1974; Farris and Strain, 1978; Forstel, 1978]. Due largely to the meticulouswork of R. D. Guy and colleagues Dole Effect Reconsidered [Guy et al., 1989, 1992], our estimates of isotope fractionationduring photosynthesisand respirationare based The •5•80of surfaceseawater today is, on average,close to on much better data than was available to Lane and Dole zero on the SMOW scale, while that of 02 is constant [1956]. In addition, we now have better estimates of the throughoutthe atmosphereat a value of about+ 23.5 %o. Over relative rates of land and oceanproduction. Finally, we can the time period of interest here, photosynthesis and now begin to assessthe impactof severaladditional processes respiration are the most important reactions producing and on the Dole effect,including isotopic exchange between CO 2 consuming02 . The isotopiccomposition of 02 in air must and02 drivenby photochemicalprocesses in the stratosphere therefore be understood in terms of isotope fractionation (first suggestedby Dole et al. [1954], p. 66), mixing in the associatedwith these reactions [Lane and Dole, 1956; Berry, aphoticzone of theocean and its influence on the •5•80 of 02 1992; and referencestherein]. consumedby respirationin that environment,and equilibrium

Table la. Terrestrial Mass Balance of 02 and •SlsOof 02

Production Production Reference

Grossproduction excluding photorespired 0 2 (GPP) 14.1 Farquhar et al. [ 1993] Gross production including photorespiration(14.1/0.69) 20.4 Farquhar et al. [ 1980]

Process Fraction of respiratory Isotope effect Reference 0 2 consumption

•5180of terrestrialphotosynthetic 02 w. r. t. SMOW 4.4 %o Farquhar et al. [ 1993] Discriminationagainst O 18 during respiration Dark respiration 59% 18.0 Guy et al. [ 1992, 1993] 10% Mehler reaction 15.1 Guy et al. [ 1992] Photorespiration 31% 21.2%o Guy et al. [ 1992] Flux weighted terrestrial respiratoryisotope effect, 18.7%o excluding dark respiration Equilibriumenrichment in •5180of leafwater w. r. t. air +0.7%o Bensonand Krause [ 1984]

Terrestrialrespiratory isotope effect (= 18.7%o-0.7%o) 18.0%o

Terrestrial Dole effect 22.4 BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE 365

Table lb. Marine Mass Balance of O2 and •180 of O2

Production term Production Reference (x1015moles/yr)

Marinegross production (=4 x seasonalnet production) 12 Keelingand Shertz[ 1992] Marinegross production - recycling within the ocean 10.6

Process Isotopeeffect Reference

fi•80 of 02 producedby marinephotosynthesis •180 of marinephotosynthetic 02 0 %o Discriminationagainst 180 during marine respiration Averagemarine eR (applies to euphoticrespiration) 20 %o Kiddon et al. [1993] EffectiveœR for respirationbelow the euphotic zone 12 %o Bender [ 1990] Equilibr.enrichment in •5•80 of O2 in waterw.r.t. air 0.7 %o Benson and Krause [1984] EffectiveeR for marinerespiration* 18.9 %o

Marine Dole effect 18.9

* =0.95x(20 %o-0.7 %o)+0.05x12%o isotopefractionation between 02 in air and02 dissolvedin over a timescale of months. We assume that the average seawater. timescalefor gasequilibration between the euphotic zone and theatmosphere is 30 days,the euphotic zone is 60 m thickand b•so of 0 2 ProducedDuring Photosynthesis its [02] is 250 gmol/L,the ratio of net/totalproduction is In the terrestrial environment, gross production is the 0.25,and net primary production is 8.5 molm '2 yr -• (Table1). productionterm of interestrelative to the Dole effectbecause From these numbers we calculate that gross primary all photosyntheticO2 escapesfrom leavesto label the productionis 34 molesm -2 yr -•. We alsocalculate that 88 % of atmospherebefore any is consumedintracellularly. This is the photosyntheticO2 goesinto the atmosphere,and 12 % is casebecause leaves are open to diffusionof CO2, a gaswhose consumedby respirationbefore it crossesthe air/seainterface atmosphericconcentration is only 0.0016times that of 02. to labelthe atmospheric O2 pool. We thereforediscount gross In the ocean,photosynthesis takes place in the mixedlayer oceanicproduction by 12 % in computingthe •5•sO balance of or in adjacentunderlying waters. Gas in the mixed layer 0 2ß equilibrateswith the atmosphereover a timescaleof weeks, Guyet al. [1993]recently showed that •5•sO of O2 produced while gasin the underlyingseasonal thermocline equilibrates by spinachthylakoids (photosyntheticmembranes) was

Table lc. Calculation of the global Dole effect, model leaf water enrichment, and atmospheric O2 turnover time

Isotope effect

Stratosphericdiminution due to O isotopeexchange of 02 andCO2 0.4 %o Calculated Dole effect* 20.8 %o Observed Dole effect 23.5 %o Leaf waterenrichment giving observed Dole effect,all otherterms constant 8.5 %o

Turnover time

Turnovertime of atmosphericO2'* 1.2 kyr

*=[fractionof productionon landx landDole effect] + [fractionof productionin oceans x marine Dole effect] - stratosphericdiminution **=(atmosphericinventory / [terrestrial production + marine production]) = (3.7x10 •9moles/[20.4 + 10.6]x 10•5 moles/yr] 366 BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE indistinguishable from that of ambient water in which the the surroundingair, non steady state conditions with respect thylakoids were suspended. Guy et al. and earlier workers [e. to the leaf water balance (especially at night, when stomata g., Stevenset al., 1975] found slight enrichments(~1%o) in close), variationsin the kinetic isotopeeffect for the escape the15•80 of 02 producedby phytoplanktonrelative to ambient of leaf water due to the biological variability in the nature of waters. They attributedthese enrichmentsto algal respiration the stomata [Farris and Strain, 1978], and variations in the in the photosynthesizing cultures rather than to •5•80of H20 gasin ambientair. Datain Figure1, on the photosynthetic fractionation. Thus the 15•80 of isotopic composition of leaf water from a forest in Julich, photosynthetic02 has the sameisotopic composition as its Germany,illustrate the expecteddiurnal cycle in the •5•80of sourcewater, or very nearly so. leaf water, together with variations demonstrating the Marine photosynthesis. The 15•80 of surfaceseawater existenceof complex disequilibriumeffects. ranges from about - 1%o to + 1%o on the SMOW scale, At any time when the humidity is < 100%, the ideal value averaging nearly zero [Craig and Gordon, 1965]. Thus the for the isotopiccomposition of leaf water can be modifiedby 15•80of 02 producedby marine photosynthesis is ~0 %o. advectionof water throughthe leaf to the site of evaporation Land photosynthesis. To understand the factors that controlthe 15•80of terrestriallyproduced 02, oneneeds to understandthe factors controlling the 15•80 of leafH20. Leaf water ultimately originates from surface seawater, but its 25 isotopic composition, relative to the source, is fractionated by equilibrium and kinetic effects attendingevaporation and 2O precipitation [Craig and Gordon, 1965]. The 15180of precipitation varies spatially and temporally and averages about - 7 %o. This water is absorbedinto the roots, stems, and 15 branchesof plants without significant fractionation. Water in o o spruce leaves, however,is enrichedin 180 because,during 10 evapotranspiration, the light isotope evaporates more rapidly. Quantitativedescriptions of the leaf water enrichment Iorch assume a two-step processfor evaporation. First, leaf water evaporatesinto saturatedair within the "head space"(stomata) birch of the leaf with a compositionfixed by equilibrium between leaf water and the head spacevapor. Second,leaf water escapes from the stomatainto free air which has a humidity < 100 %. -5 For a simple model in which evapotranspiration is branch simulatedby the steadystate evaporation of water from a sheet of paper, the "leaf water" enrichment is given by the "Craig -10 equation:" air humidity -15 X X ,. •Jleaf-•Jstem = œ* + (1 - h) œk+ h (•Jatm- •Jstem ) (1) 9 15 21 3 9 15 [Craig and Gordon, 1965; Dongmann, 1974; Farris and Strain, 1978],where •5 is the•5•80 of water. The subscriptsleaf, stem, øC and atm refer to leaf water, stem water (i.e., local meteoric % rel humidity source water), and atmospheric water, respectively. h is 3O - 100 relative humidity, œ* is the equilibrium isotope effect for evaporationof water (liquid) to water (gas),and œkis the - 80 kinetic isotopeeffect associatedwith the diffusion of water out of the stomatainto the free atmosphere. œ* is about+9 %0and 2O • is about27 %0[Farquhar et al., 1993; Yakir,D., J. Berry,L. - 60 Giles, and C. B. Osmond, Isotopic heterogeneityof water in transpiring leaves: Identification of the component that - 40 controlsthe •5•80of atmosphericO2 andCO 2, submittedto 10 Plant and Cell Environment, 1994]. In the case where the relative humidity is 100 % and - 20 temperature atmospheric water is in equilibrium with soil water, the leaf water enrichment would be 0 %0 relative to local meteoric 0 i I I I I 0 water. In the case where the humidity is zero, the calculated 9 15 21 3 9 15 leaf water enrichment is -36 %0 relative to local meteoric HOUR water. Otherfactors affecting the •5•80of chloroplastwater, Figure 1. •5•80 of H20 in leaf waterof threetrees, branch andhence the •5•80 of photosyntheticO2, includethe normal water, and air humidity as a functionof time of day for a forest diurnal variation in humidity (which is at a minimum in the in Julich, Germany. Relative humidity and temperatureare afternoon), elevation of temperaturesinside leaves relative to also shown. Redrawn from Forstel [1978]. BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE 367

and by back diffusion of leaf water from these sites. The Respiratoryoxygen isotope fractionation (eR) hasrecently isotopic composition of leaf water is thus intermediate been measuredfor the major biochemicalpathways by Guy et betweenthe •5180value predicted by the Craigequation and the al. [1989,1992, 1993]. They determined œRtO be about18 %0 •5180of stemwater [Yakir et al., 1989; Yakir, D., J. Berry, L. for dark respirationby the cyanide-sensitivepathway and 25 Giles, and C. B. Osmond, Isotopic heterogeneityof water in %0 for respiration by the alternative (cyanide-resistant) transpiring leaves: Identification of the component that pathway.Guy et al. [1992]measured œRtO be about21.2 %0for controlsthe •5•80of atmospheric02 andCO 2, submittedto photorespiration,and 15.3 %0 for the Mehler reaction (in vivo Plant and Cell Environment, (1994).; Farquhar et al., 1993]. photo-oxidation).We thencalculate a valueof eR = 18.8%0 Of interesthere is the •80 enrichmentin waterin chloroplasts, for terrestrialrespiration (Table 1). The true value is likely to wherephotosynthetic production of 02 occurs.Farquhar et al. be slightly higher because of the influence of cyanide - [1993] pointed out that this value might be governed by the resistant dark respiration. Craig equation (equation (1)), since chloroplastsare near the Finally, we note that respirationconsumes 02 from the sites of evaporationon the surface of leaves, and presented liquidpool of cells. The •i180in solutionis enrichedby 0.7 %0 strong supportingevidence. Yakir et al. (Yakir, D., J. Berry, with respectto air [Bensonand Krause, 1984], decreasingland L. Giles, and C. B. Osmond,Isotopic heterogeneityof water in fractionation with respect to the atmosphereby this amount. transpiring leaves: Identification of the component that Marine respiration. Kiddonet al. [1993] determinedER controlsthe •5•80of atmospheric02 andCO 2, submittedto values for a range of microorganisms representative of Plant and Cell Environment, 1994) presented data showing common marine autotrophsand heterotrophsin culture. They that the •5•aO of chloroplast water (and hence of measuredœR values ranging from 15 %0to 25 %0and averaging photosynthetic02) wasintermediate between that predicted by closeto + 20 %0. Larger marine animals(as well as land plants the Craig equation and that of stem water. Thus the basic and animals)have lower E R valuesdue to diffusionlimitation controlon the •80 enrichmentof waterin chloroplastsis under of fractionation.In theseorganisms, ER dependson ratesat debate. which light and heavy isotopes can diffuse to the site of With respectto the Dole effect, the value of interest is the respiration as well as on the biochemical fractionation [Lane globallyaveraged •5•80 of photosynthetic02. Thisis equalto and Dole, 1956; Epstein and Zeiri, 1988; Kiddon et al., 1993]. the time-averaged•5•80 of chloroplastwater during the However,these animals are not importantin the global02 photoperiod,weighted for the rate of photosynthetic02 balance.The averagevalue of œRformicroorganisms studied production.•5•80 of leafwater, and hence photosynthetic 02, by Kiddon et al. is close to that observedby Guy et al. [1992] varies over all scales of space and time, making the globally and expected for cyanide-sensitivedark respiration based on averaged value extremely difficult to estimate. Recently, work of Guy et al. [1989]. Nevertheless,for individual species Farquhar et al. [1993] attemptedsuch a calculationin another there are differenceswhich cannotbe readily explained. contextand estimateda value of + 4.4 %0for the productivity- In this paper,we adopta value of 20 %0for the marinevalue weightedaverage •5•aO of leaf waterand hence photosynthetic of œR' This valuedirectly characterizes marine respiration in 0 2 producedon land. We adoptthis value in the calculations the euphotic zone of the ocean in and the upper part of the that follow but note that the + 4.4 %o value falls close to the aphotic zone which equilibrates rapidly (i.e., seasonally)with lower end of the range of possible values suggestedby the atmosphere. We neglect the fact that the marine Dongmann [1974]. respiratory isotope effect in nature must be higher in the light than in the dark (in which condition Kiddon et al. [1993] made all their measurements),since the isotope effect of 21.2 %0 •180 of 0 2 Consumedby Respiration expected for photorespirationis higher than the isotope effect Terrestrial respiration. Following Farquhar et al. adopted here. Photorespiration is thought to be a minor [1980], we assume that the ratio f of photorespirationto pathway in the ocean. We note that respiration by the carboxylation(or total carbon fixation) is given by equation alternative pathway is associatedwith much more extensive (4) in their paper (p. 79): isotopic fractionation than via the cyanide resistant pathway, but we have no way of assessingthe relative rates for the two ½=(Vomax / Vc max ) x (O/C)x (Kc/Ko)x 10-3 (2) pathways.Finally, we notethat the •i•80 of thedissolved 02 pool of seawateris enrichedwith respectto the •i•80 of air by where (Vo max/ Vc max)is the ratio of the maximum 0.7 %0 [Kroopnickand Craig, 1972]. This enrichmentreflects oxygenationvelocity to the maximum carboxylationvelocity thefact that •80•60 is 0.7%0 more soluble in seawaterthan •602 of RuP2 carboxylase-oxygenase,and O is the 02 partial [Bensonand Krause, 1984]. This enrichmentdecreases the •80 pressure(210,000 ppmV). C is the CO2 partialpressure in fractionation relative to the atmosphereby 0.7 to 19.3 %0. plant cells in ppmV, assumedto be equal to 0.7 x partial A small fraction of marine respiration occurs within more pressureof CO2 in air. (K½/Ko)is the ratioof Michaelis- isolated parts of the ocean (i.e., the main thermocline and the Menton constants for carboxylation and oxygenation, deepsea). In theseareas, the 02 concentrationis significantly respectively (460/330). For the preindustrial value of 280 depleted.The residual dissolved 02 is enrichedin 180because partsper million by volume(ppmv) CO2 in air, q•= 0.31. 0 2 usedfor respirationwas depletedin the heavyisotope [e. Following Guy et al. [1992], we assume that the Mehler g., Kroopnick and Craig, 1976; Kroopnick, 1987]. We reaction accounts for 10 % of terrestrial respiration. We assumethat 5% of gross productionis remineralizedin waters assumethat all dark respiration is via the cyanide-sensitive underlyingthe seasonalthermocline with an 02 concentration pathway. The partitioning of respiration is summarizedin equalto 40%of saturationand a •i•80of 02 enrichedby 8 %0 Table 1. with respect to surface seawater. (The percent saturation 368 BENDER ET AL.' DOLE EFFECT IN THE VOSTOK ICE CORE numberis a roughestimate based on hydrographicsections of M. Thiemens(personal communication, 1993) has pointed dissolved0 2 concentration.The /5180 is estimatedfrom the out that this estimatedstratospheric /5180 decrease of 0.4 %0 relationshipbetween the /5180of dissolved0 2 in subsurface shouldbe accompaniedby a/5170decrease of thesame amount. watersand the 0 2undersaturation [Bender, 1990].) The/5180 of These changes would lead to a non-mass-dependent 0 2utilized by respiration inthese waters will be equal to/5180 fractionationof 170(defined as •5170- 0.5 •5180)of 0.2 %0,well (dissolved0 2) -eR, or 8-20 %0= 12 %0. The effective within the limits of detection. Data on the current value and respiratoryisotope effect for seawaterwith respect to the/5180 its changethrough time wouldallow one to determinerelative of ambientdissolved 0 2 is then:0.95 x 19.3%0 + 0.05x 12%0 ratesat which the non-mass-dependentanomaly is producedby = 18.9 %0. photochemicalreactions in the stratosphereand destroyedby photosynthesisand respiration. Effect of Isotopic Exchange Between O2 and CO2 in the Stratosphereon the /5•80 of O2 Summary of the Discussion of the Dole Effect Recently, Gamo et al. [1989], Yunget al. [1991], Thiemens On the basisof the precedingpresentation (summarized in et al. [ 1991], and Wen and Thiemens [ 1993] demonstratedthat Table 1), we define three termsrelevant to the discussionof isotopic exchange between 0 2 and CO2 occursin the the observed Dole effect and its change through time. The stratosphereand discussedits mechanism. O(1D), which "terrestrial Dole effect" is defined as the hypothetical Dole ultimatelyderives from 02, reactswith CO2 to formthe CO 3 effectwhen 02 andCO 2 exchangeisotopically only as the transition state complex. The symmetry properties of this resultof photosynthesisand respirationon land. The "marine complexare different for C1603 , C1602170, and C1602180. The Dole effect" is defined analogously. The "stratospheric differencein symmetryproperties causes •70 and 180 to be diminutionof theDole effect" is thedecrease in the/5180of 02 enrichedin the CO2 that is producedby the decayof CO3. in air due to isotopeexchange between 0 2 and CO2 in the These heavy isotope enrichmentsare anomalousin the sense stratosphere.The variousterms are summarizedin Table 1. thatthey are the samefor 170as for 180 [Wenand Thiemens, The estimatefor marineproduction is calculatedfrom seasonal 1993]. In almost all materials,the 170 enrichmentsare net02 productionas estimated by Keelingand Shertz [1992], roughlyhalf thoseof 180, correspondingto the numberof assumingthat gross0 2 productionis four timesseasonal net massunits by which 180 and 170 are heavierthan the major 0 2 production.Other terms are taken from the literatureor isotope(160). estimated as described above. The magnitudeof the stratosphericCO 2 enrichment,with The calculated value for the Dole effect is then the respect to its troposphericvalue, increaseswith height to productivity-weightedvalues of the terrestrialand marine Dole about 11%0 between 27 and 35 km. The 170 enrichment in effects minus the stratosphericdiminution: + 20.8 %0. This stratosphericCO 2 is anomalousas it is nearlythe same as the value is considerablyless than observed(+23.5 %0). The 180 enrichment rather than half [Thiemens et al., 1991], as differencebetween the expectedvalue and the observedvalue notedabove. Since 0 2 is theultimate source O(/D) reacting reflects errors in our estimatesand, conceivably,unrecognized with CO2, the overallreaction represents isotopic exchange processes.Recognizing the contributionof cyanide-resistant betweenCO 2 and02 in thestratosphere: darkrespiration would also increase the calculatedvalue of the Dole effect. C1602+ 180160 > C180160+ 1602. We believe that the biggestcurrent uncertainty lies in the /5180of leaf water. One can explainthe observedvalue of the Theeffect of thistransfer of 180from 02 to CO2 is to decrease Dole effect by assuminga value of + 8.7 %0,rather than + 4.4 the/5180of atmospheric0 2. Thequestion arises as to whether %0, for the /5180 of leaf water, keeping everything else oxygencompounds other than 02 canbe theultimate source of constant. Observations of the/5180 of water in natural leaves O exchangingwith CO2. The answeris no, becausethe from temperateand tropicalforests during the growingseason concentrationof otherO-bearing compounds (e.g., OH, H20) [Dongmannet al., 1974;Forstel, 1978;Zundel et al., 1978] in the stratosphereis extremely low. are compatiblewith the highervalue, but we shouldnot attach The quantitativeeffect of this exchangeon the /5180of too much importanceto this observationbecause of natural atmospheric0 2 can be estimatedby scalingthe stratospheric variability in the /5180 of leaf water. Nonetheless,we 180 enrichmentof CO2 relativeto the troposphericvalue emphasizeuncertainty in the /5180of leaf waterfor three (taken as 11%0); [Thiemens et al., 1991]) by the ratio of the reasons. First, the leaf water enrichment is the second most CO2 concentrationof air to the 02 concentration(353 ppmv/ importantcause (after respiratoryisotope discrimination) of 210,000 ppmv),by the ratio of the turnovertime of 0 2 in air the stationary180 enrichmentof 0 2 relativeto seawater. to the residence time of the troposphere with respect to Second, the term is highly variable on all space and time exchange with the stratosphere(1.2 kyr / 56 years), and by a scalesand thereforevery difficult to quantify. Third, and most factorof -1 (sincethe effect on the/5180 of 02 is inverseto that important,there is no way to explainthe Dole effect without for CO2). (Theresidence time of 02 in air is derivedin Table1, invokinga terrestrialterm (i.e., leaf waterenrichment + eR) and the time assumed to be required for exchange of the greaterthan the observed atmospheric 0 25180 enrichment of + stratosphereinto the troposphere,56 years, is from Schmidt 23.5 %0. and Khedim [1991].) One then computesthat photochemical In the ensuingdiscussion, we regardthe leaf water/5•80 isotope exchangebetween CO 2 and 0 2 in the stratosphere estimateof Farquhar et al. [1993], 4.4 %0, as a lower limit, lowersthe /5180of atmospheric0 2 by 0.4 %0. Thereis since we doubt that uncertainties in other terms can account for obviously considerableuncertainty in this number. the discrepancyfrom the calculatedDole effectof 20.8 %oand BENDERET AL.' DOLE EFFECTIN THE VOSTOKICE CORE 369 theobserved value. We regardthe leaf water value of + 8.7 %0, as into the timing of the changes. The problem is obviously whichbrings the calculationinto agreementwith observation, mostacute at thosetimes when /518Osw and/518Oat m change as an upper limit, sincepart of the discrepancybetween rapidly. calculation and observation may be due to uncertainties For the seawater curve we use/5180 data on the benthic foram indicatedabove (such as overestimationof marine production Uvigerina senticosa from deep sea sediment core V 19-30 andneglect of thelarge 18 0 fractionationassociated with [Shackleton and Pisias, 1985]. The chronology is the alternative pathway respiration). SPECMAP chronology. We correct for the influence of bottom water temperaturechanges over the last 130 kyr on the foram/5180record as described by Sowerset al. [1993]. The chronology for the GISP2 ice core is based on annual Changesin the fi•sOof atmosphericO2 and the layer counting throughout the interval relevant to this study Dole Effect During the Last 130 kyr [Alley et al., 1993]. Ice age-gasage differences(Aage values) Recordsof the/5180of 02 trappedin bubblesversus depth in are calculatedas describedby Sowerset al. [1993b]. Aage is the Vostok and GISP2 ice cores [Sowerset al., 1991; Sowers et 220 + 50 years for Holocene samplesof GISP2 increasingto al., 1993] allow us to reconstructvariations in the /5180of 630 + 100 years for glacial samples. For the GISP2 samples, atmospheric02 and the Dole effect. For these reconstructions Aage addsonly a very small uncertaintyto the calculatedvalue weneed the following information: (1) dataon the/5180 of 02 of the Dole effect. asa functionof depthin theice cores, (2) dataon the/515N of The chronologywe adopt for Vostok is that of Sowerset al. N2 for eachice core sample to correctthe /5180 of 02 for [1993]. They derived a timescale for the Vostok core by gravitationalfractionation [e.g., Craig et al., 1988]; (3) correlatingtheir/518Oat mrecord with the/518Osw curve (from the estimatesof the age of the gas in each ice core samplewe V 19-30 core) which had been placed into the SPECMAP studied,and (4) a recordof the/5180of seawater(/518Osw) versus timescale. This chronology appearsto be robust between 0- ageduring the relevanttime period. We canthen compute the 20 kyr and 60-130 kyr. It is weak for the interveningperiod Doleeffect as a functionof timeby subtracting/518Osw from becausethe absence of largevariations in /518Oatmand/518Osw /5180of contemporaneousatmospheric 02 (/518Oatm). We must make correlation difficult. This chronology has several other calculatethe Dole effect using compatiblechronologies of weaknesseswith respectto the objective of this paper. First, deep sea sedimentsand ice cores. Relative errorsin the it is derived using the method of Martinson et al. [1982] with chronostratigraphiesintroduce artifacts in the Dole effect eight coefficients for the 133 kyr record. This limited number curvessince one would be subtracting/518Oat mat one time from of coefficientsleads to a "stiff' fit which does not reproduce /518Oswat another. Chronological errors introduce errors into the timing of high-frequency events very well. Second, we the calculatedmagnitude of changesin the Dole effect as well accountfor thelag in theresponse of/518Oat mto/518Osw in a

500

450

g •O-18atm 400 •' ADole effect ß .• ß Model •O-18atm • 20N insolation

-0.5 0 5 10 15 20 25 30 Age (kyr B. P.) Figure2. Experimentalvalues of/518Osw and/518Oat mfrom the GISP2 ice core versus age. The /518Osw - age curveis fromT. Sowerset al. (manuscriptin preparation, 1994). The/5180atm- age curve is fromSowers et al. [1993]. Alsoplotted are values of ADoleeffect (heavy line), calculated from data interpolated to 0.5-kyr intervals,andvalues/518Oatm predicted bya simplemodel assuming constant biogeochemical forcing(i.e., globalcarbon fluxes, leaf waterfractionation, and photosynthetic and respiratory oxygen isotope fractionation as observed today). 370 BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE way which, ideally, forcesthe Dole effect to be about+ 0.2 %o and URI Vostok 3G samples are systematically higher than relative to the present value during glacial terminations. Our Grenoble Vostok 3G samplesby an average value of + 0.10 calculated value of the Dole effect is larger during the %o. This offset most likely reflects an undetected artifact terminations,possibly because •518Oat m appears to decrease associatedwith sealing air samples in glass flasks, but we morerapidly than 818Osw during these events. Third, Sowers et cannot confidently explain its origin. al.'s approach to deriving a chronology for the Vostok core In thispaper we correctthe 815Nof all GrenobleVostok 3G (by correlating8180 records)minimizes variations in the Dole samplesby + 0.10%o. This correction lowers •518Oatm of these effect throughout the record. Largely because of these samplesby 0.20 %o and brings them into excellent agreement problems,our curve of changesin the Dole effect as a function with•518Oatm values for GISP2. of time is provisional and will evolve as we get new data and improve the chronologyof ice and sedimentcores. Valuesof •518Oatmfor Vostokpresented here are based on Timing and Magnitude of Variations in the Dole valuesof 8180of 0 2 and815N of N2 reportedby Sowerset al. Effect During the Last 130 KYR [1993]. These resultswere from trapped gas samplesfrom the GISP2 Record 3G core which were extractedat the Laboratoirede Glaciologie et Geophysique de l'Environnement, in Grenoble, France, In Figure2 weplot •518Osw , •518Oatm, andADole effect versus sealed in glass ampoules, and analyzed at the University of age for the past 30 kyr, basedon data from the GISP2 ice core. Rhode Island (URI). More recent resultssuggest an analytical 818Oswand •518Oatm areplotted with modern values normalized problem with these results (which we call the "Grenoble to zero. We define ADole effect as the difference between the Vostok3G samples"in theensuing discussion). First, •518Oat m Dole effect at time t in the pastand the Dole effect today: values measured at URI for GISP2 samples (Figure 2) are systematicallylower, by -0.2 %0, than values measuredfor ADoleeffect (t) = A•518Oatm(t)- A818Osw(t). (3) Grenoble3G samples. Second, •518Oatm values of samplesfrom the Vostok 5G core, back to 30 kyr in age, extracted and We alsocalculate and plot the hypothetical value of fi18Oatmif measuredat URI [Malaize, 1993], agree with values from the the biogeochemicalforcing had been constant,that is to say, GISP2 core as opposed to Grenoble Vostok 3G data. An if the Dole effect had varied only becauseof the lag in the exhaustive review of our standard and methods calibrations atmosphericresponse to a changein 818Osw.Finally, we plot revealedno errorsexcept that the 815Nof our standardmay averageJune insolation at 20øN. Thisterm gives an indexof have shifted by 0.04 %0 during the period the GrenobleVostok summertime insolation changesin subtropicallatitudes of the 3Gsamples were analyzed. Such a shiftwould cause •518Oatm (= northern hemisphere. Relative variations of June insolation 8180- 2 x 815N) to be 0.04 %0 light for somesamples and at 20øN are very close to thoseof summertimenorthern heavy by the same amount for others. hemisphereinsolation and inferred variations in low-latitude To address the question of calil>ration empirically, we precipitation as deduced from modeling studies [Prell and prepared and analyzed nine samplesfrom the 3G core at URI Kutzbach, 1987]. We focus on low latitudesbecause many ("URI Vostok 3G samples").In Figure3, 815Nand/5180 are climate recordsfrom this region have a substantialamount of plotted versus depth, down to 700 m, for URI Vostok 3G spectral density in the precessionalfrequencies (19 and 23 samples, Grenoble Vostok 3G samples, and Vostok 5G kyr) which correspondto the periodof dominantvariability in samples(analyzed entirely at URI). 8180 valuesfor all three the Dole effect (see below). data sets are in excellent agreement throughout this interval. Between about 16 and 21 kyr B. P. and again for the last 3 At depths above 650 m, where we have the greatestdensity of kyr,fi18Oat mand 818Osw are nearly constant. At thesetimes of comparablesamples, 815N values for the Vostok5G samples very different climate, the Dole effect is at steady state and

0.8 2.5

0.7

0.6

0.5 ...... !...... f ...... i...... "1 0.4 ...... i...... -•l-ra•'! o 3•, •renobl, I.... 0.3 / ' 3a,URX I / 0.2 • ,[ • •,, ,i • , 0.5 I ' i ; ' ; ..... , ,' , ß 100 200 300 400 500 600 700 0 10• 20• 3•0 4• • •0 700 Depth (meters below surface) Depth (meters heiow surface) Figure3. 815Nof N2 and5180 of 0 2 versusdepth in Vostokice cores for samples from the 3G core extracted in Grenoble,the 3G coreextracted at URI, andthe 5G coreextracted at URI. 8180values agree well but815N values are offset by 0.1%0 betweensamples extracted in Grenobleand samplesextracted at URI. For reference,nominal ice agesare 7.4 kyr at 200 m, 25 kyr at 400 m, and42 kyr at 600 m depth. BENDER ET AL.' DOLE EFFECT IN THE VOSTOK ICE CORE 371

very near to its present value. ADole effect reaches a Sowers et al. ages for Vostok maximumat ~ 13 kyr B. P and a minimumat --8 kyr B. P. Until 52O 0.75 Insolationp._ about11 kyr B. P., •518Oatmdata are very close to thecurve 500- •'- ::. ' denoting constant biogeochemicalfractionation. The ADole effect maximum at 13 kyr BP may largely reflect the 1.2 kyr 480 • • :i • • ; •.25 turnovertime of atmosphericO2 andthe associatedlag in the responseof •518Oat mto changesin •518Osw . On the other hand, the minimum at about 8 kyr B. P. is undoubtedly due to

changing isotope fractionations associated with transient 440-D --0.25 processesof the biosphereand hydrosphere.This minimum in 420 , , •.5 0 50 100 150 ADole effect may be related to the inferred maximum in low- latitude precipitationat ~10 kyr B. P. and the intensification Age (kyr B. P.) of the monsoon at this time [Prell and Kutzbach, 1987], Sowers et al. ages - 3 kyr reforestation of the temperate northern hemisphere and the 0.75 associatedminimum in atmosphericpCO 2 [Neftelet al., 520 Insolation 1988], and/or to transients in oceanic circulation and 500 :: productivityat the end of the last ice age. •480 Vostok Record 460 ADole effect calculated for Vostok is plotted versus age, 440 alongwith averageJune insolation at 20øN,in Figure4a. ADole effect varies with the 23-kyr precessioncycle (Table 420 , , }.5 0 50 100 150 2). This periodicityis reflectedin maxima in the Dole effect at Age (kyr B. P.) about 128 - 133 kyr, 108 kyr, 84- 91 kyr, 58 kyr, 35 kyr, 25 kyr, and 13 kyr B. P. The 25-kyr maximum in ADole effect is Sowers et al. ages + 3 kyr anomalous,because it does not correspondto a maximum in 520 • .75 insolation. However, it would disappearif the age assignedto Insolation the•518Oatm minimum at 30kyr were decreased by about 4 kyr :: to align the minimumwith the •5180 minimumat 26 kyr sw .•480500 [Sowers et al., 1993]. Such a change would also bring the :: •..•i chronologyof Sowers et al. into better agreementwith other 460- chronologies proposed for the Vostok ice core during this time period. At the presenttime, we merely note that sample 440- --0.25 •, Dole resolutionbetween 25 and 47 kyr in the Vostok ice core is not 420 i 0.5 good enough to generate a reliable chronology based on 0 50 100 150 •518Oatmalone. Age (kyr B. P.) Variations of ADole effect, as calculated from the Sowers et Figure 4. ADole effect and integratedJune insolationversus al. [1993] chronology,are highly coherentwith, and nearly in SPECMAPage at 20øN. (a) Sowerset al. [1993] chronology phase with, insolation. The phasing is similar to that observed in the GISP2 record. Maxima in ADole effect occur at for •518Oatm(Vostok) versus age, (b) 3 kyr subtractedfrom times of maximum insolation. Minima in ADole effect lead Sowerset al. age, (c) 3 kyr added to Sowerset al. age. The SPECMAPchronology is usedthroughout for the•518Osw curve. minima in insolation (at 100 kyr, 77 kyr, and 7 kyr) or occur This figure illustratesthe high coherencybetween ADole effect in phase (115 kyr, 46 kyr). At the time of termination II, and insolation for the Sowerset al. chronologybut also shows variations in ADole effect may largely reflect the 1.2-kyr that the calculated values of ADole effect are rather sensitive to turnovertime of O2 in air, asfor GISP2(see above). However, for other events the coincidence of the ADole effect maxima relative errors in the ages of the Vostok and SPECMAP records. with respect to insolation cannot be due to the lag in the atmosphericresponse. The reason for this is that the change in •5:8Oswis too small,and the calculatedatmospheric between ADole effect and insolation (Table 2b). In Figure 4c responseis too rapid, to induce a change of the magnitude we calculate ADole effect after holding the chronology of observed. At times other than the terminations,variability in SPECMAP constantbut increasingall Vostok agesby 3 kyr. ADole effect must either be biogeochemicalor an artifact of This changein the Vostok chronologyadds a great deal of the chronology. noise to the ADole record, reduces the coherency between Unfortunately, errors in the chronology of Vostok with ADole and insolation (Table 2), and changes the phase by respect to SPECMAP can induce very large changes in about130 ø (Table2). The coherentnature of the ADoleeffect apparent values of ADole effect. We illustrate this with and insolationcurves is robust,suggesting that changesin the Figures 4b and 4c. In Figure 4b we plot ADole effect after Dole effect are related to insolation variations. However, we holding the chronologyof SPECMAP constantbut decreasing do not yet understandthe chronologyof Vostokwell enough all Vostok ages by 3 kyr. This changein the chronologyof to determinephase differences between ADole effectand other Vostok reduces noise and greatly increases the coherency climate variables. 372 BENDER ET AL.' DOLE EFFECT IN THE VOSTOK ICE CORE

Table 2a. Spectral Analysis Most marine proxy records show variability in the precessionband [e.g., Imbrie et al., 1992], but thoseshowing dominantvariability in the precessionband are rare duringthe Variable Period,kyr f statistic late Pleistocene due to the predominant influence of the ice

, sheets on glacial - interglacial climate cycles with their ADele effect 22.8 0.93 associated obliquity signal. They are restricted to the 13.2 0.84 equatorial Atlantic [Mclntyre et al., 1989; Molfino and CH4 13.6 0.85 Mclntyre, 1990] and the equatorial Indian Oceans [Clemens 22.3 0.99 and Prell, 1990]. It seems possible but unlikely that 14.5 0.93 variability of primary productivity in these restrictedareas of •518Oatm 41.8 0.92 the ocean would have dominated changesin the Dole effect. 21.6 0.93 We therefore look to changesin the continentalbiosphere to 18.1 0.81 explain variations in the Dole effect. The CH4 concentrationis the simplestindicator of large- Analyseswere run usingthe multi-tapermethod. Results scalevariations in the continentalbiosphere. CH 4 variations are given only for frequencieswhere the f-statisticwas greater during the last 150 kyr B. P., measuredin the Vestok ice core than 0.8. by Chappellaz et al. [1990], vary by a factor of 2, with most variability concentratedin the precessionband. Chappellazet Table 2b. Cross Spectral Analysis al. [1990] linked CH4 maxima to maxima in paleoprecipitationas estimatedby Prell and Kutzbach [1987]. The connection results from the fact that precipitation is Variables Period, kyr Coherency Phase,kyr linked to evaporation at low latitudes and therefore to insolation. Increasedprecipitation stimulates the activity of the biosphere,particularly in wetlandswhich are the major CH4-ADole effect 21.5 0.82 5.1 (CH4 leads) CH4 source. Increased precipitation would also cause the Dole effect to ADele-insolation 22.1 0.92 2 change, becausecontinental productivity would increaseand (insolation leads) the leaf water enrichmentwould vary. The effect of increased ADele-insolation; 20.6 0.96 1 terrestrial production would be to raise the Dole effect by subtract3 kyr from (ADele leads) increasingthe productionrate of •80 - enrichedphotosynthetic SPECMAP gas age O2 comingfrom the landbiosphere. The effectof increased wetness and humidity is to lower the Dole effect by A-insolation; 19.4 0.79 7 add 3 kyr to (insolation leads) diminishing the leaf water enrichment. The phasingbetween gas age insolation and the Dole effect is unclear. Improved chronologiesfor deep-sea sediments,together with modeling studies of terrestrial productivity and leaf water enrichments Crossspectral analyses run usingthe Blackman-Tuckey such as those of Farquhar et al. [1993], will allow us to method. improve our record of past changesin the Dole effect and our Resultsare given only when the coherencywas greater understandingof their biogeochemicalimplications. than 0.8. "Insolation"refers to averageJune insolation at 20ø N. Why is Variability in the Dole Effect so Small? As shown above, ADele effect varies with a period of ~23 Why Does ADele Effect Vary With the Periodicity of Precession? kyr, suggesting a response by the low-latitude biosphere linked to insolation changes induced by precession. This Most variability in insolation at low latitudes is associated periodicity not withstanding, ADele effect has been with precession, while at high latitudes tilt also becomes remarkablyconstant. The mean and standarddeviation, based important. The strong precession signal in ADele effect on Vestok, has been - 0.05 %0 + 0.24 %0, as calculated from suggeststhat variationsin the Dole effect are probably linked our record interpolatedto 0.5 kyr intervals. to low latitude processes. There are strongprecession signals Even more striking is the similarity of the Dole effect in both marine and continental proxy records from low between glacial and interglacial times. The ADele effect latitudes (e.g., Clemens and Prell, 1990; Chappellaz et al., duringthe last glacial maximumwas closeto 0 %0 (Figures2 1990). In principle, we could use the phasing of proxy and 4a). The changein ADele effect betweenthe penultimate signals and ADele effect to test whether certain climate glacial maximum (stage6) and the last interglacial(stage 5e) variahlo.•infl•oneo ADnlo effect. gnr example, if lnw-latit•do wa.q...... al.qn very .qmall:8180 SW decreasedby -1.2 %• durinec:, the .... ocean productivity changeswere driving changesin the Dole termination,and •80 atm decreasedby -1.4 %0 (Figure 4a) effect,we wouldexpect these two recordsto be about180 ø out [Sowers et al., 1991]. There may indeed have been a large of phase (high oceanic productivity leads to lower Dole change in ADele effect during termination II similar to the effect). In practice, we do not know phasingwell enoughto change we observed during termination I (Figure 4a). use this criterion. However,between full glacialand full interglacialclimates the BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE 373 change in the Dole effect was very small, despite the drastic temperatureswould have caused qbto decreaseby a factor of changesin the water cycle and biogeography. Why has the 0.88. This value is calculated taking Q10 for the ratio of Dole effect been so constant? oxidation to carboxylation to be 0.57 [Collatz et al., 1991]. Answering this question is impossible at the present time, Overall, we expect qbto have risen from a calculated but we can probably recognize the key influences. The preindustrialvalue of 0.31-0.42 during the LGM. Other things influence of the land biosphere during glacial times would being equal, this increase in qbwould have caused the Dole have been different from that today in the following ways: effect to rise for two reasons. First, it heralds an increase in 1. During glacial times, climate was colder and drier and terrestrial gross production, which raises the Dole effect pCO2 waslower. Muchof thetemperate and polar regions were because of the leaf water enrichment. Second, there is a coveredby ice. Cold climate, low precipitation,and extensive relatively large isotope effect associated with ice cover would all lower glacial productivity; Meyer [1988] photorespiration. has estimatedthat terrestrialnet primary productionduring the On the marine side, isotope discrimination probably was last glacial maximum (LGM) was ~75 % of the presentvalue. similar during the LGM and today. Changes in the oceanic If everything else were constant, this change would have contributionto the Dole effect would have come from changes lowered the rate of photosynthesisby the land biosphereand in marine production. Oceanic production during the LGM hence the Dole effect. differed from that of today in two ways: 2. Drier climates probably would have caused more 1. A number of studies have examined the fertility of the extensive leaf water fractionation, which would cause the Dole deep oceans during the last glacial maximum. The most effect to rise. ambitious of these are the studiesusing sedimentaryorganic 3. Meyer's estimatesof productionduring the LGM do not carbon accumulationrates [e.g., Sarnthein et al., 1988] and includethe influenceof lower atmosphericCO2 on glacial foraminiferal transfer functions [e.g., Mix, 1989a, b]. The productivity.If productionvaries by CO:ø-3 [Keeling et al., former approach has given contradictory results in different 1989], glacial productivityon land would havedecreased by an oceanic areas (hardly surprising, since there is no reason to additionalfactor of (180 ppmV/280ppmV) ø'3, or 0.88, further suspectthat productivity would change uniformly throughout loweringthe Dole effect. The true effect of lowerCO: wouldbe the ocean), but has generally indicated higher glacial values much less if nutrients ultimately limit production. [Sarnthein et al., 1988]. The same is true of the transfer 4. The lowerCO: concentrationin air duringthe last glacial functionapproach in the Atlantic [Mix, 1989a]. On average, period would have caused the ratio of both approachesindicate higher deep ocean productivity photorespiration/carboxylation (qb) to increase from a during the last glacial maximum. However, we note that calculated preindustrial value of 0.31-0.48 during the LGM, neither approachis free of problems,and resultsfrom these according to (2). On the other hand, the colder glacial two approaches are not mutually consistent, sometimes

Table 3. Literature Estimates of Paleoproductivity During the Last Glacial Maximum

Citation Study region Paleoproductivity

• , i i , i i i i i ii i Sarnthein et al. [1988] Global - 1.25 Marino et al. [ 1992] Global Lower in polar ocean,higher elsewhere

EquatorialPacific Thunnell et al. [ 1992] South China Sea 2 Archer [ 1991] Central equatorialPacific Lyle [ 1988] EquatorialPacific > 1 (-2?) Pedersenet al. [ 1991] Easternequatorial Pacific - 3 Lyle et al. [1992] Central equatorialPacific Glacial upwelling more intense

Oregon Coast Dyrnondet al. [ 1992] 0.5

Atlantic Ocean Mix [1989a] Entire Atlantic Ocean -1.25 Molfino and Mcintyre [1990] EquatorialAtlantic Glacial upwelling more intense Lyle [ 1988] EquatorialAtlantic >1 (-2?)

Southern Ocean

Charles et al. [ 1991] Atlantic and Indian regions - 1 but meridionally variable

, 374 BENDERET AL.: DOLE EFFECTIN THE VOSTOKICE CORE giving downcorevariations of productionthat are very productivityand leaf water fractionationassociated with differentin magnitude[e.g., Sarntheinet al., 1992,Figures 3, variationsin humidityand tropicalprecipitation. 4, and 6], which are uncorrelatedwithin individual cores [Sarntheinet al., 1992, Figure7], or whichare anticorrelated Acknowledgments. Thisresearch was supported by the Officeof overlong depth intervals [Mix, 1989b]. A widerange of other Polar Programs,U.S. NationalScience Foundation, and PNEDC, France. GrahamFarquhar, Jean Jouzel, Dominique Raynaud, Joseph approacheshas been used to estimatepalcoproductivity during Berry,Mark Thiemens,and StevenEmerson discussed ideas in the the lastglacial maximum. Results, summarized in Table3, are paperwith us. M. B. happilyacknowledges generous support from C. unanimousin concludingthat equatorialPacific productivity N. R. S. andC. E. A., France,during his sabbaticalat Centredes Faibles was higherduring the last glacialmaximum. Elsewhere,the Radioactivites. senseof productivityvariations differed from one region to another. However, the general sense is that open ocean References productionwas higherduring the lastglacial maximum. This would causea dropin the Dole effect. Alley, R. B., D. Meese,C. A. Shuman,A. J. Gow, K. Taylor,M. Ram, E. 2. The continental shelves, which are highly productive, Waddington,J. W. C. White and P. Mayewski, Abrupt accumulation account for ~7 % of the area of the sea floor and ~20 % of total increaseat the Younger Dryas terminationin the GISP2 ice core, oceanicproduction [Walsh, 1991]. Duringglacial times, most Nature, 362, 527-529, 1993. of the shelf area would have been exposedand would thusnot Archer, D., EquatorialPacific calcitepreservation cycles: Production or dissolution?,Paleoceanography, 6, 561-571,1991. have contributed to oceanic production. In addition, Bender,M., The õ180 of dissolved O2 in seawater: S unique tracer of productionon the shallowpart of the continentalslope would circulationand respirationin the deep sea, J. Geophys.Res., 95, havebeen suppressed by extensivesea ice coverin the Arctic 22,243-22,252, 1990. and Antarctic and by isostaticdepression of the landward Bender, M., L. Labeyrie, D. Raynaud, and C. Lorius, Isotopic portionof the continentalslope bordering glaciated areas. compositionof atmospheric0 2 in icelinked with deglaciation and Theseprocesses would have cause global ocean productivity to globalprimary productivity, Nature, 318, 349-352, 1985. fall and the Dole effect to rise. Benson, B., and D. Krause, Jr., The concentration and isotopic In summary,we can concludethat the Dole effecthas been fractionationof oxygen dissolvedin fresh water and seawaterin ratherconstant during the last 130 kyr becauseenvironmental equilibrium with the atmosphere,Lirnnol. Oceanogr.,29, 620-632, 1984. changeshave inducedroughly offsetting changes in oxygen Berry, J., Biosphere, atmosphere, ocean interactions: A plant isotopefractionation of 0 2. We canrecognize the nature of physiologist's perspective, in Primary Productivity and thesechanges, but we cannotaccount quantitatively for their BiogeochernicalCycles in the Sea, editedby P. G. Falkowskiand A. varyinginfluence on the Dole effect. D. Woodhead, pp. 441-454, Plenum,New York, 1992. Boyle, E., The role of verticalchemical fractionation in controllingLate Quaternary atmospheric carbon dioxide, J. Geophys.Res., 93, Summary and Conclusions 15,701-15,715, 1988. Broecker, W. S., The causeof the glacial to interglacialatmospheric Our ability to explain the magnitudeof the contemporary CO: change:A polaralkalinity hypothesis, Global Biogeochern. Dole effect is a measureof our understandingof the global Cycles,3, 215-239, 1989. cyclesof oxygenand water. A varietyof recentstudies have Chappellaz,J., J. M. Barnola,D. Raynaud,Y. S. Korotkevich,and C. improved our understandingof many of the principles Lorius, Ice core record of atmospheric methane over the past governing oxygen isotope fractionation during 160,000years, Nature, 345, 127-131, 1990. photosynthesisand respiration,of productionrates on land Charles, C. D., P. N. Froelich, M. A. Zibello, R. A. Mortlock, and J. J. and in the oceans, and of the effect of stratospheric Morley, Biogenic opal in southernocean sedimentsover the last photochemistryonthe •5180 of O2 in air. However,our attempt 450,000 years: Implications for surface water chemistry and to quantitativelyaccount for the Dole effectin termsof these circulation, Paleoceanography,6, 697-728, 1991. principleswas not very successful:We estimatedthe present- Clemens,S., and W. Prell, Late Pleistocenevariability of Arabian sea summermonsoon winds and continentalaridity: Eolian recordsfrom dayDole effectto be + 20.8 %0compared to theobserved value the lithogeniccomponent of deep-seasediments, Paleoceanography, of + 23.5 %0. The agreementis considerablyworse than it 5, 109-145, 1990. might appear given the fact that respiratory isotope COHMAP members, Climatic changes of the last 18,000 years: fractionation alone must account for -20 %0 of the stationary Observationsand model simulations,Science, 241, 1043-1052, 1988. enrichmentof the 180of O2 comparedwith seawater.Our Collatz, G. T., J. T. Ball, C. Grivet, and J. A. Berry, Physiologicaland underestimateof the true magnitudemost likely reflects an environmentalregulation of stomatal conductance,photosynthesis underestimate in the leaf water enrichment, failure to consider and transpiration:A model that includesa laminar boundarylayer, respirationby the alternative(cyanide resistant) pathway, and Agric. and For. Meteorol.,54, 107-136, 1991. perhapsan overestimateof marineproduction. Craig, H. and L. Gordon, Deuterium and oxygen-18 variation in the We haveused data on the•51•O of O2 fromthe Vostok ice ocean and marine atmosphere. paper presentedat Conferenceon Stable Isotopes in Oceanography, International Atomic Energy coreto calculatechanges in the Dole effectduring the last 130 Agency,Pisa, 1965. v,,,. c•,,,- ,-,•,,lt• •h,,,•, that ,,•,.;•t;,,.• ;. the Craig, H., Y. Horibe, and T. A. Sowers, Gravitational separationof been relatively small during most of the last glacial - gasesand isotopesin polarice caps,Science, 242, 1675-1678,1988. interglacialcycle. These small changesare not consistent Dole, M., G. Lane, D. Rudd, and D. A. Zaukelies,Isotopic composition with large glacial increasesin global oceanicproductivity. of atmosphericoxygen and nitrogen,Geochirn. Cosrnochirn. Acta, 6, Changesin the Dole effect vary with the periodof precession, 65-78, 1954. and we concludethat they largely reflect changesin terrestrial Dongmann,G., The contributionof land photosynthesisto the stationary BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE 375

enrichmentof •80 in theatmosphere, Radiat. Environ. Biophys., 11, Lane, G., and M. Dole, Fractionation of oxygen isotopes during 219-225, 1974. respiration,Science, 123, 574-576, 1956. Dongmann, G., H. W. Nurnberg, H. Forstel and K. Wagener, On the Lyle, M., Climatically forced organic carbonburial in equatorial Enrichmentof H2•80 in theleaves of transpiringplants, Radiat. Atlantic and Pacific Oceans,Nature, 335, 529-532, 1988. Environ. Biophys.,11, 41-52, 1974. Lyle, M. W., F. G. Prahl, and M. A. Sparrow,Upwelling and Dymond, J., E. Suess,and M. Lyle, Barium in deep-seasediment: A productivity changesinferred from a temperaturerecord in the geochemicalproxy for paleoproductivity,Paleoceanography, 7, 163- centralequatorial Pacific, Nature, 355, 812-815, 1992. 181, 1992. Malaize, Bruno, thesis. Epstein,S., and L. Zeiri, Oxygenand carbonisotopic composition of Marino, B, M. McElroy, R. Salawitch,and W. Spaulding,Glacial-to- gasesrespired by humans,Proc. Natl. Acad. Sci. U.S. A., 85, 1727- interglacial variations in the carbon isotopic composition of 1731, 1988. atmosphericCO2, Nature, 357, 461-466, 1992. Farquhar,G. D., S. yon Caemmerer,and J. A. Berry, A biochemical Martinson, D. G., W. Menke, and P. Stoffa, An inverse approach to modelof photosyntheticCO2 assimilationin leavesof C3 species, signalcorrelation, J. Geophys.Res., 87, 4807-4818, 1982. Planta, 149, 78-90, 1980. Mcintyre, A., W. Ruddiman,K. Karlin, and A. Mix, Surfacewater Farquhar,G. D., J. Lloyd, J. A. Taylor, L. Flanagan,J.P. Syvertsen,K. responseof the equatorial Atlantic Ocean to orbital forcing, T. Hubick, S. Wong, and J. R. Ehleringer,Vegetation effects on the Paleoceanography,4, 19-55,1989. isotopecomposition of oxygen in atmosphericCO2, Nature, 363, 439- Meyer, M. K., Net primaryproductivity estimates for the last 18,000 443, 1993. yearsevaluated from simulationsby a globalclimate model, M. S. Farris,F. andB. R. Strain,The effects of water-stressonleaf H2180 dissertation,Univ. of Wis. Madison, 1988. enrichment,Rad. and Environm.Biophys., 15, 167-202, 1978. Mix, A., Productivityof the PleistoceneAtlantic oceanestimated from Forstel,H., The enrichmentof •80 in leaf water undernatural foraminiferalspecies: Implications for paleo-pCO2, Nature,337, conditions,Radiat. Environ. Biophys.,15, 323-344, 1978. 541-544, 1989a. Gamo, T., M. Tsutsumi, H. Sakai, T. Nakazawa, M. Tanaka, H. Honda, Mix, A., Pleistocenepaleoproductivity: Evidence from organiccarbon H. Kubo, and T. Itoh, Carbonand oxygenisotopic ratios of carbon and foraminiferalspecies, in Productivityof the Ocean:Present and dioxide of a stratosphericprofile over Japan,Tellus, 4lB, 127-133, Past, edited by W. H. Berger, V. Smetacek,and G. Wefer, John 1989. Wiley, New York, 1989b. Guy, R. D., J. A. Berry, M. L. Fogel and T. C. Hoering,Differential Molfino, B. and A. Mcintyre, Precessional forcing of nutricline fractionationof oxygenisotopes by cyanide-resistantand cyanide- dynamicsin the equatorialAtlantic, Science, 249, 766-769,1990. sensitiverespiration in plants,Planta, 177, 483-491, 1989. Neftel,A., H. Oeschger,T. Staffelbach,and B. Stauffer,CO 2 recordin Guy, R. D., J. A. Berry, M. L. Fogel,D. H. Turpinand H. G. Weger, the Byrd ice core 50,000-5,000 years BP, Nature, 331, 609-611, Fractionationof the stableisotopes of oxygenduring respiration by 1988. plants-Thebasis of a new techniqueto estimatepartitioning to the Pedersen,T. F., B. Nielsen, and M. Pickering,Timing of late Quaternary alternativepath, in Biochemicaland PhysiologicalAspects of Plant productivity pulses in the Panama Basin and implications for Respiration,edited by H. Lambersand L. H. W. van der Plas,pp. atmosphericCO2, Paleoceanography, 6, 657-677, 1991. 443-453, SPB AcademicPublishing, The Hague,Netherlands, 1992. Prell, W. L. and J. E. Kutzbach, Monsoon variability over the past Guy, R. D., M. L. Fogel,and J. A. Berry,Photosynthetic fractionation of 150,000years, J. Geophys.Res., 92, 8411-8425,1987. stableisotopes, Plant Physiol,101, 37-47, 1993. Sarnthein,M., K. Winn, J.-C. Duplessy,and M. R. Fontugne,Global Horibe,Y., K. Shigehara,and C. C. LangwayJr., Chemicaland isotopic variations of surface ocean productivity in low and mid latitudes: compositionof air in a Greenlandice core,Earth Planet.Sci. Lett., 73, 207-210, 1985. influenceon CO2 reservoirs of the deepocean and atmosphere duringthe last 21,000 years,Paleoceanography, 3, 361-399, 1988. Imbrie, J., et al., On the structureand origin of major glaciationcycles, Sarnthein, M., U. Pflaumann, R. Ross, R. Tiedemann, and K. Winn, 1, Linearresponses to MilankovitchForcing, Paleoceanography, 7, Transfer functions to reconstruct ocean paleoproductivity: a 701-738, 1992. Keeling,C. D., R. B. Bacastow,A. R. Carter,S.C. Piper,T. P. Whorf, comparison,Geol. Soc. London, 63, 411-427, 1992. M. Heimann, W. Mook, and H. Roeloffzen, A three-dimensional Schmidt, U., and A. Khedim, In situ measurementsof carbon dioxide in model of atmosphericCO 2 transportbased on observedwinds: 1. the winter Arctic vortex and at midlatitudes: An indicator of the Analysisof observationaldata, in Aspectsof ClimateVariability in the 'age' of stratosphericair, Geophys.Res. Lett., 18, 763-766, 1991. Pacific and the WesternAmericas, Geophys. Monogr. Ser. Vol. 55, Shackleton,N.J., and N. G. Pisias,Atmospheric carbon dioxide, orbital editedby D. H. Peterson,pp. 165-236,American Geophysical Union, forcing,and climate, in TheCarbon Cycle and Atmospheric C02: Washington,D.C., 1989. Natural VariationsArchean to Present,Geophys. Monogr. Ser., Vol. Keeling, R., and S. Shertz, Seasonaland interannualvariations in 32, editedby E. T. Sundquistand W. S. Broecker,pp. 303-317, AGU, atmosphericoxygen and implicationsfor the global carboncycle, Washington,D.C., 1985. Nature, 358, 723-727, 1992. Sowers, T., M. Bender, D. Raynaud, Y. S. Korotkevich, and J. Kiddon, J., M. Bender, J. Orchardo, J. Goldman, D. Caron, and M. Orchardo,The •5180 of atmospheric0 2 from air inclusions in the Dennett, Isotopic fractionation of oxygen by respiring marine Vostokice core:Timing of CO2 andice volumechanges during the organisms,Global Biogeochem. Cycles, 7, 679-694, 1993. penultimatedeglaciation, Paleoceanography, 6, 679-696, 1991. Knox,F., andM. McElroy,Changes in atmosphericCO2; influence of Sowers, T., M. Bender, L. Labeyrie, D. Martinson, J. Jouzel, D. marinebiota at highlatitudes, J. Geophys.Res., 89, 4629-2637,1984. Raynaud,and Y. S. Korotkevich, 135,000 year Vostok - SPECMAP Kroopnick,P.M., Oxygen18 in dissolvedoxygen, in GeosecsAtlantic, commontemporal framework, Paleoceanography, 8, 737-766, 1993. Pacific, and Indian Ocean Expeditions:Shorebased Data and Stevens,C., R. Schultz, C. Van Baalen, and P. Parker, Oxygen isotope Graphics,Vol. 7, editedby H. G. Ostlund,H. Craig,W. S. Broecker, fractionationduring photosynthesisin a blue-greenand a greenalga, andD. Spencer,pp. 3 and27-182, U.S. GovernmentPrinting Office, Plant Physiol.,56, 126-129, 1975. Washington,D.C., 1987. Thiemens, M., T. Jackson, K. Mauersberger, B. Schueler, and J. Kroopnick,P., andH. Craig,Atmospheric oxygen: Isotopic composition Morton,Oxygen isotope fractionation in stratospheric CO2, Geophys. andsolubility fractionation, Science, 175, 54-55,1972. Res. Lett., 18, 669-672, 1991. Kroopnick,P., andH. Craig,Oxygen isotope fractionation in dissolved Thunell, R. C., M. Qingmin, S. Calvert and T. F. Pedersen,Glacial- oxygenin the deepsea, Earth Planet. Sci. Lett., 32,375-388, 1976. Holocene biogenic sedimentationpatterns in the south China Sea: 376 BENDER ET AL.: DOLE EFFECT IN THE VOSTOK ICE CORE

Productivityvariations and surface water pCO 2, Paleoceanography, Zundel,G., W. Miekeley,B. M. Grisi,and H. Forstel,The H2•80 7, 143-162, 1992. enrichmentin the leaf waterof tropictrees: comparison of species Walsh, J. J., Importance of continental margins in the marine from the tropicalrain forestand the semi-aridregion in Brazil, biogeochemicalcycling of carbonand nitrogen,Nature, 350, 53-55, Radiat.Environ. Biophys., 15, 203-212, 1978. 1991. Wen,J., andM. H. Thiemens,First multi-isotope study of theO(•D) + CO2 exchange and stratospheric consequences. J. Geophys. Res., 98, M. Bender and T. Sowers,Graduate School of Oceanography, 12,801-12,808, 1993. Universityof RhodeIsland, Kingston, R102881. Yakir, D., M. DeNiro, and P. Rundel, Isotopicinhomogeneity of leaf L. Labeyrie,Centre des Faibles Radioactivites, CNRS-CEA, Domaine water: Evidence and implicationsfor the use of isotopic signals du CNRS, 91190 Gif-sur-Yvette, France. transducedby plants, Geochim. Cosmochim.Acta, 53, 2769-2773, 1989. Yung, Y., W. De Mone, and J.P. Pinto, Isotopicexchange between carbondioxide and ozone via O (•D) in thestratosphere, Geophys. (ReceivedJuly 22, 1993; revisedFebruary 7, 1994; Res. Lett., 18, 13-16, 1991. acceptedMarch 11, 1994.)