GEOPHYSICAL RESEARCH LETTERS, VOL. 24, NO. 3, PAGES 273-276, FEBRUARY 1, 1997
Cooltropicaltemperatures
shifttheglobal(•lSO-T
relationship:An explanationfor the icecore
51sO- boreholethermometry conflict?
Edward A. Boyle
Departmentof Earth,Atmospheric,andPlanetarySciences,Massachusetts
Instituteof
Technology,Cambridge
Abstract.
The discrepancybetween central Greenland
Several explanations have been offered for this
boreholetemperaturesandthe isotopiccompositionof Last discrepancy [Charles et al., 1994; Cuffey et al., 1995;
Glacial Maximum ice can be explainedby a shift in the Severinghaus and Brook, 1996]. These argumentshave
•5180-Trelationship
forthehydrological
cyclelinkedto been summarized by [Jouzel et al., in press] as
cooler tropical temperatures.This concept is illustrated (1) changes in the origin of precipitation (e.g:, moisture
using a simple Rayleigh distillation model. An estimate originates
froma seawater
source
witha different
fil80 or
for o•=A•5180/AT(LGM-Holocene)
of-0.37 %o/øCis temperature),(2) microphysicaland atmosphericprocesses
determinedwith a simplegraphicaltechnique.
(with the atmosphericmoisture path dependingon small
scale details of atmospherictransport), (3)seasonality of
the precipitation (e.g., if less snow accumulatesin winter
Introduction
The internal temperatureof the central Greenland ice
sheetimplies that the Last Glacial Maximum (LGM) was
15-20øC colder than the Holocene [Cuffey et al., 1995;
Johnsen et al., 1995]. This estimate doubles that inferred
than in summer), and
(4) cloud/surfacetemperaturedifferences(e.g., changesin
the cloud inversionheight).
Plausible as all of these suggestions are, all raise a
fundamentaldifficulty for the ice core isotopethermometer:
fromtheoxygenisotopic
composition
of theicecalibrated we cannot as yet specify exactly how these processes
from the modernspatial•80-surface temperaturedifferedin the past,andtheir existencemustbe inferredad-
relationship.The boreholethermometryresult has been hoc from borehole-thermometry
calibrationsof the isotope
supportedby an estimateof firn thickness(thicker firn thermometer.For example, if moisturesourceschange,we
implies colder surface temperatures) derived from a would have to infer these changesfrom a highly resolved
comparison
of •15N2 thermaldiffusionfractionationatmospheric general circulation model (present low-
relativeto •18 0 of the ice, as seenat abruptclimate resolution models being insufficient to document the
transitions such as the end of the Younger Dryas
sourcesof Greenlandprecipitation [Jouzelet al., in press]
[Severinghauset al. ,1996]. Theseresultsimply that the with accurate boundary condition reconstructions (sea
oxygen isotopethermometergives the wrong answerin surface temperature and ice sheet topography. If these
Greenlandduringthepastseveraltensof thousands
of years explanations for the discrepancy are correct, then the
if themodern
spatial
•180_T calibration
isused.
isotope thermometer is unreliable wherever in time or
The inconsistencyof thermal and isotopicestimatesis spaceit cannotbe calibratedby boreholethermometry.
puzzling
because
the•180_surfac
etemperature
relationship The purpose of this paper is to argue for a simpler
has survivedmany testsin the modernworld and recent global-scaleorigin for the isotope-boreholethermometry
past.Dansgaard[1974] established
a consistent
empirical discrepancy.
I propose
thatthe slopeof the fi180_T
relationshipbetweenannualmeansurfacetemperature
and relationship for polar snow remains constantat all times,
the isotopiccompositionof precipitation.More detailed but that the interceptvariesin a simplefashiondetermined
globaldata and an atmospheric
generalcirculationmodel by changes in tropical sea surface temperatures.If this
witha hydrological
•180 component
support
a simple explanationis correct,the ice core thermometeris reliable
global•180-Trelationship
[Jouzel
etal., 1994].In central whenevertropicalseasurfacetemperatures
canbe specified.
Greenland,
temperature
and•180 timehistories
confirm
the temporal applicability of a simple Dansgaard-like Discussion and Simple Rayleigh Model
spatialrelationship:
the•180-T slopes
determined
for
seasonalcyclesof the pastfew years[Shumanet al., 1995]
andfor centennial/millennialtemporalvariationsassociated
with the "little ice age" and "medieval warm period"
[Cuffey et al., 1994] were about 0.5-0.6 %o/øC, only
slightlylower than seenin the modernspatialcalibration.
There is some indication that this consistency may not
extend to the early Holocene,however [Cuffey and Clow,
in press;Cuffey et al., 1995].
Following the work of Dansgaard [1974], our
understandingof the isotopic compositionof atmospheric
precipitation begins from the concept of progressive
Rayleigh distillation from a tropical source vapor.
Subsequentwork hasestablishedthat the detailsof isotope
fractionation in water vapor and precipitation can be
complex [Jouzel, 1986]. In particular,it is understoodthat
(1) themechanistic
temperature
recorded
by ice•5180is
not surfacetemperaturebut ratherthe temperatureat cloud
height(systematically
offsetto coolertemperatures
thanon
Copyright
1997bytheAmerican
Geophysical
Union.
the ground),and(2)the curvilinear
•5180-Tbehavior
predicted by the Rayleigh model should be modified by
including a kinetic isotope fractionation,which resultsin
Papernumber97GL00081.
0094-8534/97/97GL-00081 $05.00
273
274
BOYLE.' COOL TROPICAL TEMPERATURES SHIFT GLOBAL •5180-T
a linear•5180-Trelationship
(snowcrystal
growth
is not
generally
thermodynamically
reversible,
andlight 160
the fractionof the initial watervapor
remaining(setby the equilibriumvapor
pressureof water,with an exponential
d_ependence
upontemperature)
diffuses slightly faster to the growing ice crystal than
heavy 180[Jouzeland Merlivat, 1989]).With these
modifications,
thespatialvariations
of snow•5180with
R/R i = 180/160
ratiooftheremaining
watervapor
surface temperature are expected to follow a linear
comparedto its initial value
relationship. However, even with these modifications,the
predicted slope is greater than observed in the modern This behavior of this simple model for changinginitial
spatial
calibration
[Jouzel
etal.,in press].
equilibration
temperatures
(figure1) highlights
thekey
For purposes
of conceptual
illustration,
the premise
of theargument
presented
here:the•5180-T
path
behaviorof the simpleunmodified
Rayleighmodelis depends
ontheinitialequilibration
temperature.
All paths
shownfigure 1. It is assumed
thatwatervaporin the startout with the first precipitation
havingthe same
tropicsis fractionated
fromstandard
meanoceanwater isotopic
composition
astheoceansource
water(•5180=
isotopiccomposition
(SMOW, now VSMOW) with 0%orelativeto VSMOW).Proceeding
alongthecooling
isotopic
equilibrium
prevailing
between
liquidwaterand trajectory
fromtheinitialequilibration
temperature,
the
watervapor.Afterinitialequilibration,
thewatervaporis pathsaresimilarbutoffsetrelative
to oneanother
(figure
assumed
to ceaseinteracting
with theocean.Further1). As notedbefore,watervaportransport
andisotope
changes
in theisotopic
composition
of watervapor
resultseparation
ismorecomplicated
thanthismodel
implies.
In
froma progressive
decrease
in atmospheric
watervaporas particular,the empiricaland kinetically-modified
air cools,withthecondensate
immediately
removed,
as relationship
is linearwithT attemperatures
wheresnow
described
bytheequation:
falls (ratherthancurvilinearas impliedby a simple
(1) Rayleigh model).
R =fa(T)-1(
Spatially distributed global data on the mean annual
Ri
where
•5180of snowfollowa relativelysimpleanduniform
temperatureof air mass
Rliquid/Rvapor
isotope
ratiofractionation
(somewhattemperaturedependent)
Initial
equilibration
temperature:
-10-
25ø0
.
-20-
15ø6
.................
;.,•./,•'
•,
-30-
lOOO
..............
.,.•:.-?:
-40-
.¾:.'Z
-5o-
1986].
Although
the •5180-Trelationship
shouldfollowa
-t30-
.,
,.'/;,½
-70-80-90
-60
trajectory. In Greenland, this trajectory has a slope of
-0.7%0 per øC of surfacetemperaturechange.The system
thusoperatesalmostas if a singlewater vaporsourcewas
being progressivelyfractionated.This simplicity derives
from the dominanceof evaporationin tropicaloceans:the
tropicsencompassby far the largestarea of the ocean(the
latitudeband 20øSto 20øN encompasses
37% of the ocean
surface)and the exponentialdependence
of vaporpressure
on temperature (-5% per øC) ensures that tropical
evaporationdominatesthe flux of water vapor into the
atmosphere.This statementshould not be interpretedas
literally indicatingthat water vaporin Greenlandoriginates
from the equator; studies indicate that Greenland
precipitation has a subtropicalorigin [e.g. Koster et al.,
-50
-z•O -50
I
-20
ß
I
-10
ß
I
0
ß
I
10
2'0
Temperature, øC
Figure 1. Simple Rayleigh Distillation Model.
Equilibriumvaporpressures
for water(>0øC) andice
(<0øC)wereusedto computef. Water vaporpressure
estimatedfor range0_<T(øC)<30as:
(3.6923153078749E-07
+ 1.53605633443583E-
09*(T- 15) - 5.41333194861509E-11'(T-15)^2 +
5.3576557376315E- 16*(T- 15)^3- 1.28751659504212E13'(T-15)^4)*exp(0.0602300128260019'(T+273.16)).
Ice vaporpressureestimatedfor -60<T(øC)<0as:
(2.06727283917237E-07+6.40752162250554E-09*(T30)-4.2067291755106 IE-11 *(T-30)^2 5.50575367980902E- 13'(T-30)^3 4.1231149597534E- 13*(T30^4))*exp(0.0602644337097994'(T+273.16)).
Temperature
dependence
of t• afterDansgaard[1964].
similar trajectoryno matterwhat the initial temperatureis,
the absolutepathway dependson the initial temperature.
When the Dansgaardrelationshipwasestablished,
however,
it generallywas believedthat tropicaloceantemperatures
had remainedrelatively stableduring the extremeclimate
cyclesof the Pleistoceneglaciations[e.g. CLIMAP, 1981].
Hence
it was reasonable
to have assumed that the same
15180-T
pathway
applied
tobothmodern
andLGMoceans.
In recentyears, severallines of evidencehave suggested
that LGM tropical oceantemperaturesmay have been as
much as -4-6øC cooler than they are at present.This cool
LGM tropical temperatureinterpretionis basedon (1) the
noble gas contentof tropical aquifers[Stuteet al., 1995;
Stuteet al., 1992],(2) •5180andSr/Cain corals[Becket
al., 1992; Guildersonet al., 1994], and (3) snow lines and
the isotopiccomposition
of low-latitudemountainglaciers
[Porter,1979;Broeckerand Denton,1990;Thompsonet
al., 1995] The validity of this new view is still being
debatedwithinthe paleoclimatological
community,but it
is sufficiently
established
to consider
theconsequences
of
cooltropicsfor the oxygenisotopecomposition
pathway
for globalwatervapor.
BOYLE' COOL TROPICALTEMPERATURESSHIFT GLOBAL •5180-T
The role of moisture source temperature is well
established (Aristarain et al., 1986; Grootes, 1993), and
discussionsof the borehole thermometry/isotopeconflict
275
increased ice volume) increases the shift due to cooler
source
waters.
Theresulting
shiftin the$180_
T trajectory
has significant consequencesfor interpretation of the
alwaysmentionthe temperatureof sourcemoistureregions oxygen isotopic composition of LGM ice. The lowregion
ofthe$180_
T relationship
isexpanded
as a possiblefactor [Cuffey et al., 1995; Jouzel et al., in temperature
press].What is new here is (1) explicit linkage between in figure 2 to show that if (1) central Greenland surface
cooltropics
andtheglobal•5180-Trelationship,
and(2) temperaturescooledfrom about-31øC to -45øC as implied
separating the problem into two separate components by boreholethermometryandnitrogenisotopedata,and(2)
(slope=site temperature change; intercept=source the trajectory followed by global water vapor shifted by
temperaturechange).Here, the changein sourcemoisture -5øC as requiredby coolertropics(dashedline on Fig. 2),
in $180of central
Greenland
iceshould
be
temperatureis proposedas global or large-scaleregional thenthechange
-•-40%o,not-45%øas impliedby the modern•5180-T
rather than as local in origin.
trajectory.Sucha shiftwouldresultin the LGM ice having
LGM-Modern
a •5180about-5%o
relative
to theHolocene.
Usingthis
Comparison
As implied by figure 1, cooling of the initial sourceof
equilibratedwater vapor shifts the trajectoryto parallel
pathsdeterminedby the initial water vapor composition.
Henceit is reasonableto assumethat,if tropicalseasurface
isotopicshiftwith the modernDansgaardrelationship
(with
the same intercept as today), this isotopic shift would
imply a cooling of only -•8øC in surface temperature.
Statedanotherway, this result implies that a comparison
of modern and LGM isotopic values and temperatures
temperatures
did coolby 5øC,thepathway
for •180 in shouldshowan apparent
A•5180/ATrelationship
of
global precipitationwould have shifted to heavier values.
The shift is difficult to specify precisely by simple
arguments because of the complexity of water vapor
transport,howeverthereis little doubtof its signand order
-•0.37 %o/øCcomparedto the modern value of-0.70 %o/øC.
Borehole thermometry data imply an optimal model fit
A•5180/AT
of 0.35%o/øC
[Cuffeyet al., 1995],in close
agreementwith thatpredictedhere(perhapsmisleadinglyso
of magnitude.
Here,I simplyassume
thatthe $180_T given the assumptionsusedfor this simpleestimate).
It shouldbe emphasizedagain that this argumentdoes
sloperemainsthe sameat all times,and that the intercept
varies accordingto changesin tropical temperaturesand not assumethat the moisturein Greenlandliterally comes
global isootpiccomposition(figure 2). The-1%o positive from the tropics;rather,the term "cooltropics"is takenas
shift in meanocean•5180 duringthe LGM (dueto implying a global cooling of the oceanand atmosphere
-30
hypothetical
LGM
•5180-T
relationship
shifted
to5øC
cooler
....'"
"'""
..."'"
I
tropical
SST,'.......... ...'"'
1
+17/oo
oceanic
•5180
.......
"'_,•/'/f1%o
ocean
iS
increase
1
II
O•
=0.69
%0per
øC..'""• øCTrop.
,
/./
Tcooling•
I
..."/
-35
/
.........................
/
o•= 0.37%0perøC '.•:--••
......
•
-4o L
[.--" i
•
relationship
i LGM
Modern
itemperature
-45
temperature
I
I
I
-40
-35
-30
Temperature, øC
Figure2.Proposed
LGM-modern
•518C-temperature
offset.
276
BOYLE.:COOLTROPICALTEMPERATURES
SHIFTGLOBAL•5180-T
systemincluding the subtropicalregionswhere Greenland Beck, J. W., R. L. Edwards, E. Ito, F. W. Taylor, J. Recy, F.
Rougerie, P. Joannot,and C. Henin, Sea-surfacetemperature
precipitationmay originate.
Summary and Implications
A shifted•5180-Ttrajectory
is a simpleconsequence
of
cooler tropical temperatures.This shift can accountfor
most of apparent disagreementbetween the borehole
from coral skeletal strontium/calcium ratios, Science, 257,
644-647, 1992.
Broecker, W. S. and G. H. Denton, The role of ocean-
atmosphere reorganizations in glacial cycles, Quat. Sci.
Rev., 9, 305-341, 1990.
Charles, C., D. Rind, J. Jouzel, R. D. Koster and R. G.
Fairbanks, Science, 263, 508-, 1994.
temperature
history(and the apparent
A•5180/AT CL1MAP,
p.m., Seasonal reconstructions of the earth's
surface at the last glacial maximum,Geol. Soc. Am. Map and
Chart Seri, Geol. Soc. Am. Map and Chart Seri, 36, 1981.
parameter)and the LGM-Holocene temperaturechange
inferredfrom•5180usingthemodern
spatial•5180-T
relationship.This interpretationdoesnot requireshiftsin
water vapor sourceregions,and it maintainsthe same
Cuffey, K. M., R. B. Alley, P.M. Grootes,J.
oBolzan
and S.
Anandakrishnan,
Calibration
of the•41
otopic
øO
is
paleothermometerfor for central Greenland, using borehole
A•5180/AT slopeat all times.It thereforerestores temperatures, J. Glaciol., 40, 341-349, 1994.
predictability
to theice coreisotopethermometer:
if we can Cuffey, K. M., G. D. Clow, R. B. Alley, M. Stuiver, E. D.
Waddington and R. W. Saltus, Large Arctic Temperature
specifytropicalsurfacetemperatures
as a functionof time,
Change at the Wisconsin-Holocene Glacial Transition,
then we can determine a surface temperature from a
Science, 270, 455-458,
contemporaneous
measurement
of •180 in ice.Of courseDansgaard, W.,
1995.
Stable isotopes in precipitation, Tellus, 16,
436-468, 1974.Guilderson, T. P., R. G. Fairbanks and J. L.
other factors, including regionally differing water vapor
Rubenstone, Tropical temperature variations since 20,000
sources,may alsoaffect the isotopiccomposition
of LGM
years ago: modulating interhemispheric climate change,
ice from central Greenland; however, the significanceof
Science, 263, 663-665,. 1994.
thesechangesneednot be as largeas suggested
previously. Grootes, P.M.,
Interpreting continental oxygen isotope
records, Geophys. Monograph, 78, 37-46, 78.
This reinterpretation of glacial oxygen isotope values
makes some testablepredictions.If this model is correct, Jouzel, J., Isotopes in cloud physics: multiphase and
multistage condensation processes, Amsterdam, Elsevier
Science,
Handbook
of
Environmental
Isotope
Geochemistry, P. Fritz and J. C. Fontes, Amsterdam,
Elsevier Science, 2, 61-112, 1986.
Jouzel, J., R. B. Alley, K. M. Cuffey, W. Dansgaard, P.
Grootes, G. Hoffmann, S. J. Johnsen,et al., Validity of the
temperature reconstruction from ice cores, J. Geophys.
Res., in press, 1996.
all LGM sitesshouldfollowa A•5180/ATrelationship
with the modern slope, only shifted-1%o vertically and
-5øC horizontallyfrom themodernintercept.Thispredicted
trajectorycan be testedby nitrogenisotopeand borehole
thermometrydata at additionalsitesthroughoutthe world.
This new interpretationalso has consequences
for the
interpretation
of otherisotopethermometers
influencedby
Jouzel, J., R. D. Koster, R. J. Suozzo and G. L. Russell, Stable
precipitation,
such
ascontinental
carbonate
•5180,
tree-ring water isotope behavior during the last glacial maximum: a
circulation model analysis, J. Geophys. Res., 99,
cellulose
D/H,andpaleo-aquifer
•5180.
Thenegative
shift general
25791-25801,
in15180
resulting
fromtropical
cooling
andseawater
•5180 Jouzel, J. and L.1994.
Merlivat, Deuterium and oxygen 18 in
is estimatedhereasbeinglargerthanthe measuredpositive
precipitation: modeling of the istopic effects during snow
formation, J. Geophys. Res., 89, 11749-11757, 1984.
al., 1985]. Hence continentalisotopedata shouldhave the Porter, S.C., Quaternary Stratigraphy and chronology of
Mauna Kea, Hawaii: a 380,000-yr record of mid-Pacific
same large artifact as does ice in central Greenland;
volcanism and ice-cap glaciation, Geol Soc. Am. Bull pt. I,
15180
changes
at sitessuchasDevilsHole[Winograd
et
continentalisotopicpaleotemperatures
will underestimate 90, 609-61 l, 1979.
continentalcooling. Compensatingfor this effect would Severinghaus,J.P. and E. J. Brook, Gaseousthermal diffusion
as a gas-phasestratigraphicmarker of abrupt warmings in
also eliminate most of the apparenttiming differenceson
termination II between Devil's Hole record and the marine
record and buttress faith in Milankovitch
and the
ice core climate records, Trans. Am. Geophys. Union,
S157,
1996.
Shuman, C. A., R. B. Alley, S. Anandakrishnan, J. W. C.
SPECMAP chronology.
White, P.M. Grootes and C. R. Stearns, Temperature and
accumulationat the GreenlandSummit: Comparisonof highAcknowledgements:
This idea aroseafter a discussion resolution isotope profiles and satellite passive microwave
brightness temperature trends, J. Geophys. Res., 100,
with Jeff Severinghausconcerninghis nitrogen isotopedata.
9165-9177,
1995.
Jess Adkins, Maureen Raymo, Michael Bender, Jean Jouzel, Stute, M., M. Forster, H. Frischkom, A. Serejo, J. F. Clark, P.
RichardAlley, and Kurt Cuffey offeredconstructivecriticismof
Schlosser,W. S. Broecker, et al., Cooling of tropical Brazil
(5øC) during the last glacial maximum, Science, 269, 379this work. My understandingof relevant ice core data and
383, 1995.
conceptshas beenfosteredover the yearsby discussions
with
Stute, M., P. Schlosser, J. F. Clark and W. S. Broecker,
Richard Alley, Jean Jouzel, Jim White, and Pieter Grootes
Paleotemperaturesin the southwesternUnited States derived
thanks to my participation in the GISP2 project, currently
from noble gases in ground water, Science, 256, 10001003, 1992.
funded by NSF grant EAR-9316207. My paleoclimatology
efforts are supportedby NSF grant OCE9402198and NOAA Henderson,J. Cole-Dai, J. F. Bolzan, et al., Late Glacial Stage
grant NA46GP0282.
References
Aristarain, A.J., J. Jouzel, M. Purchet, Past Antarctic
Peninsular Climate (1850-1980) deduced from an ice core
isotoperecord,ClimaticChange,8, 69-90, 1986.
and Holocene tropical ice core records from Huscaran, Peru,
Science, 269, 46-50, 1995.
E.A. Boyle, Departmentof Earth, Atmospheric,and Planetary
Sciences, Massachusetts Institute of Technology,
Cambridge,MA 02139. (e-mail: [email protected])
(ReceivedNovember6, 1996; acceptedDecember5, 1996)