4. D. S. Broomhead and G. P. King, Phys. D20, 217 14. P. S. Marcus, J. Fluid Mech. 146, 65 (1984).(1986). 15. K. T. Coughlin and P. S. Marcus, ibid. 234, 1

5. M. P. Chauve and P. LeGal, ibid. 58, 407 (1992). (1992); ibid., p. 19.6. J. L. Lumley, in (3), pp. 105-122. 16. R. P. Tagg, H. L. Swinney, Phys. Rev. Lett.7. N. Aubry et al., J. Fluid Mech. 192, 115 (1988). 66, 1161 (1991).8. A. Glezer, Z. Kadioglu, A. J. Pearlstein, Phys. 17. Y. Takeda, K. Kobayashi, W. E. Fischer, Exp.

Fluids A 1, 1363 (1989). Fluids 9, 317 (1990).9. Y. Takeda, Nucl. Eng. Des. 126, 277 (1990). 18. Y. Takeda, W. E. Fischer, K. Kobayashi, T.

10. R. C. DiPrima and H. L. Swinney, in Hydrodynamic Takada, ibid. 13, 199 (1992).Instabilities and the Transition to Turbulence, H. L. 19. Y. Takeda, W. E. Fischer, J. Sakakibara, K. Ohm-Swinney and J. P. Gollub, Eds. (Springer-Verlag, ura, Phys. Rev. E47, 4130 (1993).Berlin, 1985), pp. 139-180. 20. Y. Takeda, W. E. Fischer, J. Sakakibara, Phys.

11. P. R. Fenstermacher, H. L. Swinney, J. P. Gollub, Rev. Lett. 70, 3569 (1993).J. Fluid Mech. 94, 103 (1979). 21. D. Andereck, S. S. Liu, H. L. Swinney, J. Fluid

12. M. Gorman and H. L. Swinney, ibid. 117, 123 Mech. 164, 155 (1986).(1982). 22. R. Tagg et al., Phys. Rev. A 39, 3734 (1989).

13. L. H. Zhang and H. L. Swinney, Phys. Rev. A 31,1006 (1985). 3 August 1993; accepted 30 November 1993

Constraints on Transport and Kinetics inHydrothermal Systems from Zoned Garnet Crystals

Bjern Jamtveit* and Richard L. HervigZonation of oxygen isotope ratios, fluorine, and rare earth element abundances acrossgarnet crystals from the Permian Oslo Rift reflect temporal variation of the hydrothermalsystem in which the garnets grew. A sharp rimward decrease in the 180/160 ratio (of 5 permil) across the interface between aluminum-rich garnet cores and iron-rich rims indicatesinflux of meteoric fluids to a system initially dominated by magmatic fluids. This influx mayrecord the transition from ductile to brittle deformation of the hydrothermally altered rocks.In contrast, fluorine and light rare earth element concentrations increase at the core-riminterface. These data may reflect enhanced advective transport and notable kinetic controlon trace element uptake by the garnets during brittle deformation.

Hydrothermal systems play a key role inthe chemical differentiation of the Earth'scrust. The understanding of such systemsrequires integrated studies of active as wellas ancient systems. Ancient systems con-tain the integrated results of hydrothermalactivity occurring at depths that are inac-cessible to direct observations. The mainproblem when studying ancient systems isto determine their temporal evolution.Analysis of intracrystalline zonation of hy-drothermal minerals may provide a more orless continuous record of the processes tak-ing place during the evolution of ancienthydrothermal systems (I). Such data areneeded to test and complement results ob-tained from theoretical models of hydro-thermal system dynamics. Here, we presention microprobe analyses (IMPA) of zonedhydrothermal garnets from the PermianOslo Rift, Norway, which provide informa-tion on the sources of hydrothermal fluids,their compositional evolution, and the ex-tent and mechanism of mass transfer be-tween fluids and the rock matrix.

The garnets that we studied are repre-B. Jamtveit, Department of Geology, University ofOslo, Post Office Box 1047 Blindern, N-0316 Oslo,Norway.Richard L. Hervig, Center for Solid State Science,Arizona State University, Tempe, AZ 85287.*To whom correspondence should be addressed.

sentative of a large population of hydrother-mal garnets from layered shale-carbonaterock sequences in the contact aureole of theDrammen granite (1). These rocks wereinfiltrated by fluids during cooling of thegranite. Early infiltration was pervasive andcoeval with more or less ductile deforma-tion, whereas later infiltration was focusedalong faults and fractures formed duringbrittle deformation (1, 2-4). Garnets grewthroughout these events. Mineral equilibriaand fluid inclusion results indicate that thegarnets formed at temperatures of 350°C to400°C. Backscattered electron (BSE) imag-es (Fig. 1) show a sharp transition from arelatively grossular-rich core (gr = 20 to 40mol%) to an andradite-rich rim (gr <5mol%). The rim furthermore contains sev-eral thin layers of more grossular-rich com-position that give rise to the oscillatoryzonation pattern evident in the BSE image(Fig. 1). The composition of the grossular-rich layers was largely determined by thelocal mineral assemblage during periods ofslow fluid influx and low crystal growthrates, whereas the composition of morerapidly grown (5) andradite-rich layers re-flects infiltration of externally derived hy-drothermal fluids with high concentrationsof As and W (1, 6). The andradite-richgarnets are extremely low in high-fieldstrength elements (HFEs) such as Zr, Y,

SCIENCE * VOL. 263 * 28 JANUARY 1994

and Ti. The lack of evidence that localsources of HFEs (notably zircon andsphene) were removed from the systemindicates that the fluids did not reach equi-librium concentrations with respect tothese components during growth of theandradite-rich rims. Thus, models of fluidflow and chemical reaction kinetics in hy-drothermal systems that assume local fluid-rock equilibrium may not always be valid(7). To obtain more detailed informationabout the mass transport, mass transfer,kinetic dispersion, aqueous complexation,and the fluid sources, we analyzed the gar-net shown in Fig. 1 for oxygen isotope,fluorine, and rare earth element (REE)compositions.

The oxygen isotope zonation profile(Fig. 2A) reveals a marked discontinuity atthe core-rim interface, suggesting that thepore fluid had a relatively higher oxygenisotope ratio when the garnet core wasgrowing. Fractionation factors for the gar-net-HO20 system at 350°C, corrected forcompositional variations of the garnet (8,9), indicate that the core and rim precipi-tated from fluids with average 6180 values[relative to standard mean ocean water(SMOW)] of +10 per mil and +6 per mil,respectively. A value of +10 per mil is thatexpected for fluid equilibrated with theDrammen granite at high temperature (4),whereas the value of +6 per mil mayindicate an appreciable meteoric watercomponent in the hydrothermal fluid [theoxygen composition of Permian meteoricwater in this area was probably around - 12per mil (10)].

Additional information on changes inthe hydrothermal system is provided by Fand REE data from the garnet. Consider-able amounts of F may substitute for OH

Fig. 1. Backscattered electron image of adodecahedral garnet crystal selected for ionmicroprobe analysis. The garnet is essentially abinary solution between the end membersgrossular (gr = Ca3Al2Si3012) and andradite(and = Ca3Fe2Si3012). The image shows an-dradite-rich garnet (light color) growing ongrossular-richer core (dark). Oscillatory zona-tion seen in andradite-rich rim reflects differentrelative proportions of major components an-dradite and grossular. Dark spots show thesites of oxygen isotope analyses.

505

.t.Ruuaaon D

ecember 17, 2020

http://science.sciencem

ag.org/D

ownloaded from

that is invariably present in hydrothermalgarnets (11). The F content of our sample(Fig. 2B) increases abruptly at the core-rimboundary and reaches values near 6000parts per million in the rim. The outermostmargin (0 to 300 pum from the edge) is,however, generally poor in F.

Except for the outermost margin, thegarnet rim is enriched in the light REEs(LREEs) relative to the core (Fig. 3) as wellas in As, W, Mo, and Fe and is relativelydepleted in Zr, Y, Ti, and Al (1). Lan-thanum shows the most abrupt increase atthe 900-pm boundary, followed by Ce andthen Nd. The outermost margin is low inREEs as well as in As, W, and Mo (1). Thegrossular-rich garnets, including the garnetcores, are typically depleted in LREE ascompared to the abundance in the host rock,whereas the concentration of Sm and heavi-er REEs is comparable to those of the hostrock (Fig. 4). A small, but significant, pos-itive Eu anomaly is evident in most samples.In contrast, the andradite rims are stronglyenriched in the LREEs. However, there is anotable La depletion relative to the otherLREE in the most LREE-rich samples. Therims are furthermore depleted in the HREEand show a distinct positive Eu anomaly.Although the andradite-rich garnets have awide range in REE concentrations, thechondrite-normalized patterns are broadlysimilar. The outermost garnet margins haveflat or slightly LREE-depleted REE patterns(Fig. 4).

Neodymium isotope data from these gar-nets demonstrate that the elevated LREEconcentrations observed in the garnet rims(Fig. 3) are associated with the supply ofREE from the magmatic system (Fig. 5).

The correlation between the contents of

1u0

8

0a0CO

ECLEL

4,2

0

6000

5000

40003000

2000

10000

.9 ^ A ; ni

Hi

TI 1

..... ........ . . .

500 10U0

Distance from garnet g (m)Fig. 2. (A) Oxygen isotope zonation profile ofgarnet shown in Fig. 1 (19). (B) Fluorine zona-tion profile across garnet shown in Fig. 1 (20).

506

trace components derived from the mag-matic system in the zoned garnets, as well asa marked similarity among the zonationpatterns of different garnet crystals from thesame sample (6), indicate that the zonationpatterns are to a large extent controlled bylarge-scale transport processes. In such acase, the garnet composition may vary as aresult of variations in the composition ofthe bulk pore fluid at the site of crystalgrowth or as a result of a change in themechanism by which trace elements aretaken up by the garnet during crystalgrowth. The local pore fluid composition isaffected by the competition between localbuffering and external controls through in-filtrating fluids. Fluctuations in the concen-tration of some externally derived compo-nent in the pore fluid at the site of crystalgrowth may indicate a change in the advec-tive transport rate or a change in the

50 .. U

40 i20015°0_100

40--

10 2040 08010010

concentration of this component at thefluid inlet. In a flow system with finitekinetic dispersion, advective mass transportmay affect the local pore-fluid compositionto different extents for different compo-nents as a result of differences in the fluid-rock partitioning (12). Thus, rapidly trans-ported components may be completely ex-ternally controlled, whereas others are lo-cally buffered.

The oxygen isotope results imply thatthe isotope signature of the garnet wasexternally controlled during all stages ofgarnet crystallization [as suggested in (4)1.Influx of fluids with an appreciable meteoriccomponent during garnet rim growth wouldimply that the fluid pressure at this stageapproached hydrostatic values and thus thata connected porosity had formed from thesite of crystal growth toward the near-surface environment. The formation of aconnected porosity was probably associatedwith extensive brittle deformation near thegranite roof (2), and an accompanying in-crease in rock permeability may have al-lowed a marked increase in the fluid flowrates. Increased flow rate would result inenhanced advective transport rates and ki-netic dispersion and may have resulted inincreased infiltration control on the pore-fluid composition for a wide range of com-

010.

0.1

Dancefrm gae edge m)Fig. 3. Zonation profiles of La, Ce, Nd, Sm, andEr across the garnet shown in Fig. 1 (21).

Lacr PrNd SmEuGdTbDyHoErTmYb Lu

Fig. 4. Chondrite-normalized REE patternsshowing typical garnet core and rim composi-tions (closed circles). Additional IMPA, some ofwhich are depicted here, include Eu and Tb.Distances are as in Fig. 2. ICP-MS analyses ofbulk garnet separates are also given (opensymbols). The ICP-MS analyses were done atthe Department of Geology, University of Bris-tol, according to the methods described byKemp (22).

SCIENCE * VOL. 263 * 28 JANUARY 1994

..........................

on Decem

ber 17, 2020

http://science.sciencemag.org/

Dow

nloaded from

ponents during andradite precipitation.The "magmatic" oxygen isotope compo-

sition of grossular-rich garnet cores coupledwith a Nd isotope composition indicative ofa sedimentary source (Fig. 5) probably in-dicate that the advective transport rate ofREEs is much lower than that of oxygen atmoderate fluid flow rates where the extentof kinetic dispersion is limited. Thus, thegarnet core grew downstream of the Ndisotope front but upstream of the oxygenisotope front. The LREE depletion of thecores shows that the relatively smallerHREE fits better into the eightfold coordi-nation site of the garnet lattice than thelarger LREE (13).

The Nd isotope composition of the rims,in contrast to that of the core, indicates anappreciable input of Nd derived from themagmatic system. Furthermore, there is acontinuous change from sedimentary tomagmatic Nd isotope signatures (Fig. 5),indicating that there are no sharp Nd iso-tope fronts. Thus, we suggest that at thisstage the fluid flow rate is high enough tocause extensive kinetic dispersion. As thekinetic dispersion increases, the advectivetransport rates of oxygen and Nd wouldconverge (12). This explains the infiltra-tion control on oxygen as well as on Ndisotope signatures for the rim. Thus, itseems that the differences between thecore and rim REE compositions were, atleast partly, a result of variable advectivetransport rates. Furthermore, differentialtransport rates of the REE are indicated bythe abrupt increase in La concentrationimmediately at the core-rim boundary;this buildup indicates that La was trans-ported more rapidly than Ce and Nd.Because HREEs are compatible in garnet,the low HREE content of the garnet rimshows that the pore fluid must have been

o

0

-2

-4

-6

-8

-10

-12

Magmatic rocks

%

Metasedimentary rocks

0 20 40 60 80 100Nd (ppm)

Fig. 5. Neodymium concentration (23) versus

ENd values calculated at 280 Ma (estimated ageof the Drammen granite) for bulk garnet sepa-rates. The positive correlation observed indi-cates supply of Nd from the magmatic systemduring garnet growth. The age of the Drammengranite and the ranges in ENd for the Drammengranite and the host metasedimentary rocksare from Trennes and Brandon (14) andJamtveit (24).

depleted in HREE. The most likely expla-nation for this is higher retardation of theHREE as compared with the LREE alongthe flow path, possibly as a result ofpreferential incorporation of HREE intogarnet or other phases upstream of theanalyzed crystal.

The oxygen isotope signature for the rimindicates an appreciable contribution frommeteoric fluids at this stage. However, themeteoric contribution required to lower thehydrothermal fluid composition from amagmatic value of +10 per mil to +6 permil (about 20%) would not cause majordilution of the components derived fromthe magmatic system.

The extreme LREE enrichment ob-served for the garnet rims as well as thepositive Eu anomalies are hard to explain byequilibrium partitioning of REE betweenthe hydrothermal fluids and the garnet lat-tice. In fact, the rim REE patterns arequalitatively similar to REE patterns fromacidic, high-temperature hydrothermal flu-ids (15-17). Therefore, we speculate thatthe rim REE composition may to a largeextent have been controlled by irreversibleincorporation of REE initially adsorbed onthe garnet surface during periods of rapidcrystal growth (5). In such a case, the rimREE pattern would reflect surface-fluid REEpartitioning rather than bulk lattice con-trol. If correct, this model could explain theLa (and Ce) depletion observed for themost LREE-rich garnets as the effect of bulklattice control on the REE uptake superim-posed on the surface control.

The positive correlation between the Fand the LREE (notably the La) contentsfurthermore suggests that F may have beenan important complexing agent for REE inthe hydrothermal solution. This is in agree-ment with recent models for REE complex-ation in aqueous solutions at high temper-atures (18), and thus the REE fractionationbetween the garnet surface and the hydro-thermal fluid may have been partly con-trolled by the relative stability of REEfluoro-complexes in the aqueous solution atthe ambient conditions.

The low REE contents of the garnetmargin may reflect a decrease in garnetgrowth rate with an associated decrease inthe interface controls on the REE uptake.A decrease in growth rate during growth ofthe outermost rim has been inferred fromthe morphological evolution of these skarngarnets (5). Alternatively, it may simplyreflect depletion of LREE in the hydrother-mal fluid.

The time series information contained inthe garnet zonation profiles provides a de-tailed record of the geochemical evolution ofthe present hydrothermal system and con-strains the effects of transport and reactionkinetics during fluid-rock interactions.

SCIENCE * VOL. 263 * 28 JANUARY 1994

REFERENCES AND NOTES

1. B. Jamtveit, R. A. Wogelius, D. G. Fraser, Geology21,113(1993).

2. P. M. Ihlen, Sver. Geol. Unders. Ser. C 59, 6(1986).

3. ___ R. Trennes, F. M. Vokes, in MetallizationAssociated with Acid Magmatism, A: M. Evans,Ed. (Wiley, London, 1982), pp. 111-136.

4. B. Jamtveit, H. F. Grorud, K. Bucher-Nurminen,Earth Planet. Sci. Lett. 114, 131 (1992).

5. B. Jamtveit and T. B. Andersen, Phys. Chem.Minerals 19,176 (1992).

6. B. Jamtveit, K. V. Ragnarsdottir, B. J. Wood, inpreparation.

7. A. C. Lasaga and D. M. Rye, Am. J. Sci. 293, 361(1993).

8. U. Lichtenstein and S. Hoernes, Eur. J. Mineral. 4,239 (1992).

9. B. E. Taylor and J. R. O'Neil, Contrib. Mineral.Petrol. 64, 1 (1977).

10. B. Jamtveit and T. Andersen, Econ. Geol., inpress.

11. C. E. Manning and D. K. Bird, Am. Mineral. 75,859 (1990).

12. At local fluid-rock equilibrium, advective (one-dimensional) transport of a tracer occurs as ageochemical front that moves with a velocity (V)approximately given by V= Wpf/l Kdps, where Wisthe fluid flux, pf and ps are the fluid and soliddensities, respectively, and Kd is the solid-fluidpartition coefficient by mass [M. J. Bickle and D.McKenzie, Contrib. Mineral. Petrol. 95, 384(1987)]. Thus, at a given distance from the fluidinlet the concentration of rapidly transported trac-ers may be completely controlled by the fluid flow,whereas the concentration of more slowly movingcomponents may be completely controlled by thelocal mineralogy, depending on the relative posi-tions of the respective geochemical fronts. Incontrast, if the rate of fluid-rock interaction iscontrolled by slow reaction kinetics relative to thefluid flow rate, the geochemical fronts will becomebroader as a result of kinetic dispersion and mayeventually disappear as recognizable fronts. Inthe limiting case of infinite dispersion, any tracerwill travel with the same velocity as the fluid itself[M. J. Bickle, Am. J. Sci. 292, 289 (1992)]. Thus,the relative transport rates of different compo-nents converge with increasing fluid flow velocity.

13. C. Pride and G. K. Muecke, Contrib. Mineral.Petrol. 76, 463 (1981).

14. R. G. Trennes and A. D. Brandon, ibid. 109, 275(1992).

15. A. Michard and F. Albarede, Chem. Geol. 55, 51(1986).

16. A. C. Campbell et al., Nature 335, 514 (1988).17. A. Michard, Geochim. Cosmochim. Acta 53, 745

(1989).18. S. A. Wood, Chem. Geol. 88, 99 (1990).19. Oxygen analyses were done on a Cameca IMS 3f

ion microprobe equipped with a standard manu-facturer-supplied duoplasmatron, an alkali metalion source constructed in-house [R. T. Lareau andP. Williams, in Secondary Ion Mass Spectrometry,SIMS V, A. Benninghoven, R. J. Colton, D. S.Simons, W. Werner, Eds (Springer-Verlag, Berlin,1986), Series in Chemical Physics, vol. 44, pp.149-151], and an auxiliary power supply allowingthe sample potential to vary in the range of 0 to5000 V. A 2-nA Cs+ beam accelerated to +10 kVwas focused onto a 30-p.m spot on the sample.Sample charging was controlled with the use of ahigh-current, high-voltage, four-lens electronflood gun mounted on a 2-3/4-inch Conflat flange[R. L Hervig, P. Williams, R. M. Thomas, S. N.Schauer, I. M. Steele, Int. J. Mass Spectrom. Ion.Proc. 120, 45 (1992)]. The electron gun wasaligned to obtain a maximum count rate for 60secondary ions with initial kinetic energies of 0 to40 eV. Negative secondary ions were detected atlow mass resolution (m/Am -- 500). The 150-p^mtransfer optics were used with a 750-p.m fieldaperture. To remove the 160OD, 160H2,' and 170Hinterferences with 10, we used extreme energyfiltering (EEF). It has recently been shown [S. N.

507

:.::::::::.::::.:.:.:::.::.·:i·::::::.:: ·.·. :.:.S·:·'.;tS""'" ''i''''''iii'

.::.·.:.:...:.:.:.:.:.:...:·ff:.:.:;:;f.

on Decem

ber 17, 2020

http://science.sciencemag.org/

Dow

nloaded from

Schauer and P. Williams, Int. J. Mass. Spectrom.Ion. Proc. 103, 21 (1990)] that by collection ofonly ions ejected from the sample with energies>300 eV, most molecular species, includinghydrides, can be effectively eliminated from thesecondary ion spectrum to the part per millionlevel. We counted ions with >350 ±+ 20 eV,which gave -4 E5 counts per second for 60 onthe garnets in this study. Total analysis time was45 min. Typically, we used 100 cycles, whichresulted in craters 5 to 7 pm deep. The errorexpected from counting statistics ranged from 1to 1.2 per mil (1 SD). Actual errors clusteredbetween 1 and 1.6 per mil, and two analysesshowed larger errors. We converted raw ratios toabsolute values by comparing values with thoseof several natural mineral standards and byassuming a linear relation between instrumentaloxygen isotope fractionation and molecularweight of the sample. Plain horizontal lines giveaverage core and rim compositions, and thedashed horizontal lines on the figure representthe oxygen isotope composition obtained forbulk separates of cores and rims analyzed byconventional techniques. The shaded region de-notes the relatively grossular-rich core.

20. Fluorine microanalyses were done with a 0.5-nAO60- primary beam focused onto a spot 20 pim indiameter. Negative secondary ions with 50 ± 20eV excess kinetic energy were detected. The19F- count rate was normalized to the count ratefor 480, and this ratio was calibrated againstDurango apatite. Previous studies in this lab showsmall matrix effects (-15%) for F over a range ofphases, presumably because F is nearly com-

pletely ionized during sputtering [P. Williams, inPractical Surface Analysis, D. Briggs and M. P.Seah, Eds. (Wiley, New York, 1992), pp. 177-228]. Error of measurements is within the size ofthe circles.

21. Rare earth element analyses by ion microprobewere done with the duoplasmatron source togenerate a -1.5-nA60 primary ion beam at 12.5kV focused onto a spot 20 pm in diameter. Posi-tive secondary ions of 42Ca, '39La, 140Ce, 143Nd,147Sm, and 166Er with 75 ± 20 eV excess energywere detected. Erbium contents were correctedfor oxides of Nd and Sm, and concentrations weredetermined according to the methods describedby E. Zinner and G. Crozaz [Int. J. Mass. Spec-trom. Ion. Proc. 69, 17 (1986)]. Errors (1 SD) ofindividual measurements were determined bycounting statistics. Data tables are available fromthe authors on request.

22. A. J. Kemp, A. J. Lewis, M. R. Palmer, in prepa-ration.

23. Neodymium isotope data were obtained at theMineralogical-Geological Museum in Oslo. Isoto-pic ratios were measured on a VG 354 5-collectorfully automated mass spectrometer. Analyticalmethods follow the procedures described by E.W. Mearns [Lithos 19, 269 (1986)].

24. B. Jamtveit, unpublished results.25. We thank K. V. Ragnarsdottir for help with the ICP

analyses and J. Blundy, K. V. Ragnarsdottir, andR. Wogelius for comments on the manuscript.Supported by the Norwegian Research Council(440.92/023 and 440.93/011).

20 August 1993; accepted 2 December 1993

Glacial-lnterglacial Changes in MoistureSources for Greenland: Influences on

the Ice Core Record of Climate

C. D. Charles, D. Rind, J. Jouzel, R. D. Koster, R. G. FairbanksLarge, abrupt shifts in the l8O/160 ratio found in Greenland ice must reflect real featuresof the climate system variability. These isotopic shifts can be viewed as a result of airtemperature fluctuations, but determination of the cause of the changes-the most crucialissue for future climate concerns-requires a detailed understanding of the controls on

isotopes in precipitation. Results from general circulation model experiments suggest thatthe sources of Greenland precipitation varied with different climate states, allowing dynamicatmospheric mechanisms for influencing the ice core isotope shifts.

A network of ice cores now spans much ofthe Greenland ice sheet. The s180 recordsfrom the ice cores reveal a pattern of pro-nounced, abrupt shifts prior to the Ho-locene period (1). The fact that many shiftscan be traced among different cores makes

C. D. Charles, Scripps Institution of Oceanography,University of California at San Diego, La Jolla, CA92093.D. Rind, Goddard Institute for Space Studies, NewYork, NY 10028.J. Jouzel, Laboratoire de Modelisation du Climat et deI'Environnement, CEA/DSM CE Saclay, 91191 Gif SurYvette, France, and Laboratoire de Glaciologie etGeophysique de I'Environnement, CNRS BP96,38402,Saint Martin d'Heres Cedex, France.R. D. Koster, Hydrological Sciences Branch, NationalAeronautics and Space Administration/GoddardSpace Flight Center, Greenbelt, MD 20904.R. G. Fairbanks, Lamont-Doherty Earth Observatory,Palisades, NY 10964, and Department of GeologicalSciences, Columbia University, New York, NY 10025.

508

it clear that these oscillations are not mere-

ly artifacts of local ice sheet dynamics. Yet,their climatic implications have not beenfully resolved. Comparison with the b180-temperature correlation established frommodern precipitation (2) implies that the8180 shifts, in some cases, reflect tempera-ture fluctuations of 7° to 10°C occurringover the course of less than 20 years (3).Obvious questions arise from this type ofinference: (i) What aspects of the climatesystem are capable of creating such largetemperature fluctuations over such a broadarea and so rapidly? (ii) Can such temper-ature shifts be triggered in the modemrnclimate through anthropogenic influences?

These oscillations in 8180 are usuallyconsidered to be driven primarily by theNorth Atlantic because experiments with a

SCIENCE · VOL. 263 · 28 JANUARY 1994

coupled ocean-atmosphere model suggestthat the North Atlantic thermohaline cir-culation, which delivers heat to the highlatitudes, could alternate between two sta-ble modes of operation (on or off), andaspects of North Atlantic deep-sea sedi-ment records exhibit this type of behavior(4). However, several observations alsosuggest the involvement of other climaticprocesses. In a series of model experimentsthat effectively tested the sensitivity of cli-mate to reduced North Atlantic heat trans-port, this mechanism could explain lessthan one-third the magnitude of air tem-perature change over Greenland inferredfrom the ice core isotopic shifts (5). Recentestimation of the rapidity of the changes inthe ice cores (6) suggests that the variationsoccur more rapidly than do variations inEuropean climate records that are undeni-ably the product of North Atlantic surfaceocean changes (7). New dating of the icecore record at Summit suggests that two ofthe most prominent s'80 shifts occur pre-cisely during catastrophic melting of the icesheets, at about 14,000 and 11,000 yearsago (8). Thus, it seems unlikely that theNorth Atlantic acted alone to shape thedramatic ice core records.

To comprehend the full climatic signif-icance of the 8]80 events, we must estab-lish the driving forces for isotopic variabil-ity in precipitation. Although the fraction-ation of H2180 between water vapor andcondensate is temperature dependent, theamount of H2t8O in precipitation at anygiven location is more directly determinedby the amount of original vapor remainingin the air mass (9). Because colder air holdsless water vapor, this distillation processalso depends on temperature, but the corre-lation between sO80 and temperature variesaccording to the location of the originalmoisture sources (whether it is remote orproximal), the trajectories of the moisture(whether it is advected along a cold or warmpath), and the mixing between sources withdifferent degrees of distillation. Thus, localb180 variability may result from a combina-tion of direct air-temperature effects andcomplicated air mass mixing effects. Theseinfluences must be separated for accuratepaleoclimatic reconstruction. More impor-tantly, regardless of how the modem 8QsO-temperature calibrations apply to the rapidshifts observed in the ice cores (10), thecauses of the shifts cannot be deduced fromempirical extrapolations. One must also un-derstand where and how the characteristicsof moisture sources changed under differentclimate conditions.

One approach is to use a general circu-lation model (GCM) that is capable ofconsidering explicitly the complex varietyof processes that act on the isotopes. Previ-ous GCM studies indicated that although

. "";;'. . :.j::::.1.:..::.::::::.:.:.:.;.5.:...:.: ""''"'"';I, ··i.ii.··;··i;;ii;;··i;;·r.· n..·;).", . . ":O .i;;-----------I-------·- ;..:.:...::.:.:.::::: iiiiiiiij.i.'"'" """"":.:::'.::.::":.:.:.:'.'.:.'.i.:....····;;·;;;···;;;··;;;;·;··;- ;;··;;·;;···;;;;;:·.·:.·.·:.:.·.:.::.::. ... ": .·;:·:.·: i.i:·.·:i·.·:·:'.;·.;·.;.*·;.;;·;.·.;...h. ---·.S;i. ······· ··iZl.iii··i·i···i·i·ii·;· ·i.·

on Decem

ber 17, 2020

http://science.sciencemag.org/

Dow

nloaded from

Constraints on Transport and Kinetics in Hydrothermal Systems from Zoned Garnet CrystalsBjørn Jamtveit and Richard L. Hervig

DOI: 10.1126/science.263.5146.505 (5146), 505-508.263Science 

ARTICLE TOOLS http://science.sciencemag.org/content/263/5146/505

REFERENCES

http://science.sciencemag.org/content/263/5146/505#BIBLThis article cites 21 articles, 5 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

1994 by the American Association for the Advancement of Science

on Decem

ber 17, 2020

http://science.sciencemag.org/

Dow

nloaded from