GeophysicalMonograph Series

Including
IUGG Volumes
Maurice Ewing Volumes
Mineral Physics
 GEOPHYSICAL MONOGRAPH SERIES

GeophysicalMonograph Volumes 21 Quantitative Modeling of Magnetospheric

Processes W. P. Olson (Ed.)
1 Antarctica in the International Geophysical 22 Derivation, Meaning, and Use of Geomagnetic
Year A. P. Crary,L. M. Gould,E. O. Hulburt, Indices P. N. Mayaud
HughOdishaw,andWaldoE. Smith(Eds.) 23 The Tectonicand GeologicEvolutionof
2 Geophysicsand the IGY Hugh Odishawand Southeast Asian Seas and Islands Dennis E.
StanleyRuttenberg
(Eds.) Hayes(Ed.)
3 AtmosphericChemistryof Chlorine and Sulfur 24 Mechanical Behavior of Crustal Rocks: The
Compounds JamesP. Lodge,Jr. (Ed.) Handin Volume N. L. Carter, M. Friedman,
4 ContemporaryGeodesy CharlesA. Whittenand I. M. Logan,andD. W. Stearns
(Eds.)
Kenneth H. Drummond (Eds.)
25 Physicsof AuroralArc Formation S.-I. Akasofu
5 Physicsof Precipitation HelmutWeickmann and J. R. Kan (Eds.)
(Ed.)
26 Heterogeneous
Atmospheric
Chemistry David
6 The Crust of the Pacific Basin Gordon A.
R. Schryer(Ed.)
Macdonald and Hisashi Kuno (Eds.) 27 The Tectonicand GeologicEvolutionof
7 Antarctica Research: The Matthew Fontaine Southeast Asian Seas and Islands: Part
Maury Memorial Symposium H. Wexler,M. J. 2 DennisE. Hayes(Ed.)
Rubin,andJ. E. Caskey,Jr. (Eds.) 28 MagnetosphericCurrents Thomas
A. Potemra
8 Terrestrial Heat Flow William H. K. Lee (Ed.) (Ed.)
9 Gravity Anomalies:UnsurveyedAreas Hyman 29 Climate Processesand Climate Sensitivity
Orlin (Ed.) (Maurice Ewing Volume 5) JamesE. Hansenand
10 The Earth Beneath the Continents: A Volume of Taro Takahashi (Eds.)
GeophysicalStudies in Honor of Merle A. 30 Magnetic Reconnectionin Spaceand Laboratory
Tuve JohnS. Steinhartand T. Jefferson
Smith Plasmas Edward W. Hones, Jr. (Ed.)
(Eds.) 31 Point Defects in Minerals (Mineral Physics
11 IsotopeTechniquesin the Hydrologic Volume 1) Robert N. Schock (Ed.)
Cycle GlennE. Stout(Ed.) 32 The Carbon Cycle and Atmospheric CO2:
12 The Crust and Upper Mantle of the Pacific Natural Variations Archean to Present E.T.
Area LeonKnopoff,CharlesL. Drake,and Sundquistand W. S. Broecker
(Eds.)
PembrokeJ. Hart (Eds.) 33 Greenland Ice Core: Geophysics,Geochemistry,
13 The Earth's Crust and Upper Mantle Pembroke and the Environment C. C. Langway,Jr., H.
jr. Hart (Ed.) Oeschger,and W. Dansgaard(Eds.)
14 The Structureand Physical Propertiesof the 34 CollisionlessShocksin the Heliosphere: A
Earth's Crust JohnG. Heacock(Ed.) Tutorial Review Robert G. Stone and Bruce T.
15 The Use of Artificial Satellites for Tsurutani (Eds.)
Geodesy SorenW. Henricksen,
ArmandoMancini, 35 CollisionlessShocksin the Heliosphere:
and Bernard H. Chovitz (Eds.) Reviews of Current Research Bruce T. Tsurutani
16 Flow and Fracture of Rocks H.C. Heard, I. Y. and Robert G. Stone (Eds.)
Borg,N. L. Carter,andC. B. Raleigh(Eds.) 36 Mineral and Rock Deformation: Laboratory
17 Man-Made Lakes: Their Problems and Studies--The Paterson Volume B. E. Hobbs and
Environmental Effects William C. Ackermann, H. C. Heard (Eds.)
GilbertF. White,andE. B. Worthington
(Eds.) 37 Earthquake SourceMechanics (Maurice Ewing
18 The Upper Atmosphere in Motion: A Selection Volume 6) ShamitaDas, JohnBoatwright,and
of PapersWith Annotation C. O. Hinesand Christopher
H. Scholz(Eds.)
Colleagues 38 Ion Acceleration in the Magnetosphere and
19 The Geophysicsof the Pacific Ocean Basin and Ionosphere TomChang(Ed.)
Its Margin: A Volume in Honor of George P. 39 High PressureResearchin Mineral Physics
Woollard George H. Sutton,Murli H. Manghnani, (Mineral PhysicsVolume 2) Murli H.
andRalphMoberly(Eds.) ManghnaniandYasuhiko Syono(Eds.)
20 The Earth'sCrust: Its Nature and Physical 40 Gondwana Six: Structure, Tectonics, and
Properties JohnG. Heacock(Ed.) Geophysics Ga• D. McKenzie
41 Gondwana Six: Stratigraphy, Sedimentology, 3 Origin and Evolution of SedimentaryBasins
and Paleontoloty Garry D. McKenzie(Ed.) and Their Energy and Mineral Resources
42 Flow and Transport Through Unsaturated Raymond A. Price(Ed.)
Fractured Rock Daniel D. Evansand Thomas]. 4 Slow Deformation and Transmission of Stress
Nicholson (Eds. ) in the Earth Steven C. Cohen and Petr Vant'•ek
43 Seamounts, Islands, and Atolls Barbara H. (Eds.)
Keating,PatriciaFryer,RodeyBatiza,andGeorgeW. 5 Deep Structure and Past Kinematics of Accreted
Boehlert (Eds.) Terranes JohnW. Hillhouse(Ed.)
44 Modeling Magnetospheric Plasma T. E. Moore, 6 Properties and Processesof Earth's Lower
]. H. Waite, Jr. (Eds.) Crust RobertF. Mereu,StephanMueller and David
45 Perovskite: A Structure of Great Interest to M. Fountain (Eds.)
Geophysicsand Materials Science Alexandra
Navrotskyand Donald]. Weidner(Eds.)
46 Structureand Dynamics of Earth's Deep Interior Maurice Ewing Volumes
(IUGG Volume 1) D. E. SmylieandRaymond
Hide (Eds.) 1 Island Arcs, Deep Sea Trenches, and Back-Arc
Basins Manik Talwani and Walter C. Pitman III
47 HydrogeologicalRegimes and Their Subsurface (Eds.)
Thermal Effects (IUGG Volume 2) Alan E.
2 Deep Drilling Results in the Atlantic Ocean:
Beck,GrantGarvinandLajosStegena (Eds.)
Ocean Crust Manik Talwani,Christopher G.
48 Origin and Evolution of SedimentaryBasins
Harrison,and DennisE. Hayes(Eds.)
and Their Energy and Mineral Resources(IUGG
3 Deep Drilling Results in the Atlantic Ocean:
Volume 3) Raymond A. Price(Ed.) Continental Margins and Paleoenvironment
49 Slow Deformation and Transmission of Stress
Manik Talwani,WilliamHay, and WilliamB. F.
in the Earth (IUGG Volume 4) Steven C. Cohen
Ryan(Eds.)
and Petr Vanfc•ek(Eds.)
4 Earthquake Prediction--An International
50 Deep Structure and Past Kinematics of Accreted Review David W. Simpson and Paul G. Richards
Terranes (IUGG Volume 5) JohnW. Hillhouse (Eds.)
(Ed.) 5 Climate Processes and Climate
51 Properties and Processesof Earth's Lower Crust
Sensitivity JamesE. Hansenand Taro Takahashi
(IUGG Volume 6) RobertF. Merev, Stephan (Eds.)
Mueller and David M. Fountain (Eds.)
6 Earthquake SourceMechanics ShamitaDas,
JohnBoatwright,
andChristopher
H. Scholz(Eds.)
IUGG Volumes

1 Structureand Dynamics of Earth's Deep

Mineral PhysicsVolumes
Interior D. E. SmylieandRaymondHide (Eds.)
2 HydrogeologicalRegimes and Their Subsurface 1 Point Defects in Minerals Robert N. Schock(Ed.)
Thermal Effects Alan E. Beck, Grant Garvin and 2 High PressureResearchin Mineral Physics
LajosStegena
(Eds.) Murli H. Manghnaniand Yasuhiko
Syono
 GeophysicalMonograph 52
IUGG Volume 7

Understanding Climate Change

A. Berger
R. E. Dickinson
John W. Kidson
Editors

• American
Geophysical
Union
International
Union
ofGeodesy
and
Geophysical
Monograph/IUGG
Series

Library of CongressCataloging-in-PublicationData
Understandingclimatechange.

(Geophysicalmonograph;52/IUGG series;7)
1. Climaticchanges•Congresses.
I. Berger,A. (Andr(•).
II. Dickinson, Robert E. (Robert Earl).
III. Kidson, John W. IV. Series.
QC981.8.C5U48 1989 551.6 89-6746
ISBN 0-87590-457-2

Copyright1989by the AmericanGeophysical

Union, 2000FloridaAvenue,
NW, Washington,DC 20009

Figures,tables,and shortexcerptsmay be reprintedin scientific

booksand
journalsif the sourceis properlycited.

Authorizationto photocopyitemsfor internalor personaluse, or the

internalor personaluse of specificclients,is grantedby the American
Geophysical Unionforlibraries
andotherusersregisteredwiththeCopyright
ClearanceCenter(CCC) Transactional ReportingService,providedthat the
basefee of $1.00per copy,plus$0.10per pageis paid directlyto CCC, 21
CongressStreet,Salem, MA 01970.0065-8448/89/$01.
+ .10.
Thisconsentdoesnotextendto otherkindsof copying,suchascopyingfor
creatingnew collectiveworks or for resale.The reproductionof multiple
copiesand the useof full articlesor the useof extracts,includingfiguresand
tables,for commercialpurposesrequirespermissionfrom AGU.
Printed in the United States of
 CONTENTS

Preface

A. Berger,R. E. Dickinson
andJohnW. Kidson ix

WORLD CLIMATE RESEARCH PROGRAMME

The World Climate ResearchProgramme

Gordon A. McBean 3

II. PALEOCLIMATES AND ICE

2. Long-Term Climatic and Environmental Recordsfrom Antarctic Ice

C. Lorius,J-M Barnola,M. Legrand,J. R. Petit, D. Raynaud,C. Ritz, N. Barkov,Y. S.
Korotkevich,
V. N. Petrov,C. Genthon,
J. Jouzel,V. M. Kotlyakov, F. Yiou,andG. Raisbeck 11
3. The Role of Land Ice and Snow in Climate
Michael H. Kuhn 17

III. VOLCANOES AND CLIMATE

, PetrologicEvidenceof Volatile EmissionsFrom Major Historic and Pre-HistoricVolcanic

Eruptions
JulieM. PalaisandHaraldurSigurdsson 31

IV. BIOGEOCHEMICAL CYCLES, LAND HYDROLOGY, LAND SURFACE PROCESSES

AND CLIMATE

5. Uptake by the Atlantic Ocean of ExcessAtmospheric Carbon Dioxide and Radiocarbon

BertBolin,AndersBj6rkstr6mand BerrienMoore 57
6. African Drought: Characteristics,Causal Theories and Global Teleconnections
Sharon E. Nicholson 79

7. Sensitivity of Climate Model to Hydrology

DuzhengYe 101
8. Stability of Tree/GrassVegetation Systems
PeterS. Eagleson 109

V. TROPICAL OCEAN AND GLOBAL ATMOSPHERE

, Toga and Atmospheric Processes

Kevin Trenberth 117
10. Toga Real Time Oceanographyin the Pacific
DavidHalpern 127
viii CONTENTS

VI. MODELLING CLIMATE, PAST, PRESENT AND FUTURE

11. Aeronomy and Paleoclimate

J.-C. G•rard 139
12. Studies of Cretaceous Climate
Eric J. Barron 149
13. Simulationsof the LastGlacialMaximumwith an AtmosphericGeneralCirculationModel
IncludingPaleoclimaticTracerCycles
Sylvie
Joussaume,JeanJouzel
andRobertSadourny 159
14. ProgressandFutureDevelopments in ModellingtheClimateSystem
with General
Circulation Models
P. R. Rowntree 163
15. QuantitativeAnalysis of Feedbacksin Climate Model Simulations
MichaelE. Schlesinger

The principalaim of this symposiumwas to de-
scribethe contributionswhich are made by eachof
the disciplinesrepresentedin the IUGG to the study
of climate change. In order to present a balanced
program,the Symposiumwas composedof invited
reviewsbut otherviewpointswere put forwarddur-
ing generaldiscussion.The themes coveredreflect
the interests of the seven IUGG Associations
includevolcanism;biogeochemistry; land hydrology;
modelingclimate,past and present;cryosphere;pa-
leoclimates;land-surfaceprocesses;tropical oceans
and the globalatmosphere;cloudsand atmospheric
radiation;aeronomyand planetaryatmospheres; and
modelingfuture climatechanges.
This symposiumattractedenoughpapersfor 3 full
days plus an overflowin posterform. G.A. McBean
opened the sessionwith an overview of the main
objectives of the World Climate Research Pro-
gramme.This programis conceptuallyorganizedon
the basisof three streamsdirectedtowards monthly
and seasonal climateprediction,interannualvariabili-
ty and climatepredictabilityout to decades.Interrela-
tionshipsbetweenWCRP and other geophysicalpro-
gramswere alsonoted.
The impactof the time dependentglobaldistribu-
tion of aerosolsafter the Agung (1963) and E1 Chi-
chon volcanic eruptions was investigatedagainst
natural variability and model assumptions.H. Si-
gurdssonstressedthe importanceof the type and
massyield of volatilesrelease'dduring an eruption.
Petrologicestimatesof volatile degassingshow, for
example, that some eruptionshave yielded of the
orderof10løto10l• kgofsulfur,chlorine
andfluorine
compoundsto the atmosphere,with highest yield
from Tamborain 1815,Laki in 1783, Eldgja in 934AD
and Katmai in 1912.
The role of the circulationand biogeochemistryof
the AtlanticOceanin the determinationof the precise
airbornefractionof anthropogenicemissionof CO2to
the atmospherewas underlined by B. Bolin. The
processesaffectingthe oceanfor the explanationof
the ice age CO2 changes,like the biotic effect on
upper oceanchemistry,was discussedby J.R. Togg-
weiler. In this respect, the presenceof unutilized
nutrients in high latitudes and the production of
refractorydissolvedorganiccompoundsby marine
organismsseem to be two sensitivepoints in the
ocean'sbiologicalsystem.
S.E. Nicholson examined the current state
PREFACE
knowledgeof Africanclimatevariabilityand drought.
Paleo-, historical and recent data conclude that sur-
facefeedbackis likely linked to droughtin the Sahel,
while sea-surfacetemperatureanomaliesappearto be
moreclearlylinked with variabilityin equatorialand
southernregionsthan in the Sahel.The crucialroles
of soil moistureand vegetationin the hydrological
and processeswere discussedby Duzheng Ye through
studiesof climate models sensitivity. Deforestation
and large-scaleirrigation have profound effectson
the climate,the influenceof the hydrologicaldistur-
bancesbeing not necssarilylimited to the disturbed
region.
P.R. Rowntreereviewedthe presentcapabilityof 3-
D modelsof the generalcirculationof the atmosphere
and ocean to simulate existing and past climates.
Theseexperimentscall for consideration of the real-
ism of simulation of air-sea interactions, of cloud
opticalproperties,of land-surface processes, of snow
accumulation on icecapsand of sea-icedistributions.
The leadingcandidatesfor explainingwarm global
Cretaceousclimate(which are the very differentcon-
tinental geometry, the high eustaticsea level, the
higherCO2atmosphericconcentration and the great-
er role of the oceanin polewardheat transport)were
tested by E.J. Barron using a variety of climate
models. Water isotopesand desert dust particles
were includedby $. Joussaume in the LMD general
circulationmodel to better simulate the last glacial
maximum and to help transferingpaleo-datainto
atmosphericparameters.
The role of land icein climatewas presentedby M.
Kuhn who showedthat one third of this century'ssea
level rise are causedby melting mountain glaciers
whereaspolaricedid not givean appreciate contribu-
tion. The non linear W.R. Peltier's model, which
incorporates
the physicsof the isostaticsinkingof the
Earth under the weight of the ice, was shown to
explain
thedominant
105yearcyclein thePleistocene
ice volume, as a subharmonic resonant relaxation
oscillationwhich characterizes
its responseto realistic
astronomicalforcing.
Paleoclimaticdata are now recognizedto play an
importantrole in climaticresearchby providing a
multiple opportunityto test the sensitivityof our
climate models and so to better understand the
behaviorof the climatesystem. J.Cl. Duplessyhas
reviewed 3 of the major recent developmentsin
of paleoceanography: (i) the deepwater circulation
ix
x PREFACE
now be reconstructed, in particularduring the last
glacialmaximumand the peak of the lastinterglacial;
(ii) carbonisotoperecordsof planktonicand benthic
foraminiferashow that over the last glacial-intergla-
cial cyclethe atmosphericCO2 varied in phasewith
the global climate, in agreement with the ice core
data;(iii) acceleratormassspectrometerhasprovided
a detailedchronologyof the retreatof the polarfront
during the last deglaciation. Variations in climate
parametersover the last 150,000 years as deduced
from the Vostok ice core were given by C1. Lorius.
They concernAntarctic surface temperature, conti-
nental and marine aerosolsand atmosphericCO2.
Spectralanalysisof the isotopetemperatureand CO2
profiles show clearly peaks around the orbital fre-
quencies. Besides a long term trend, rapid CO2
variations are also recorded in the ice cores, and data
from Siple(Antarctica)has allowed to reconstructthe
atmospheric CO2 increases since 1800, as clearly
shownby H. Oeschger.
Understandingthe land-surfaceprocesses is one of
the high priorityin climaticresearch.The firstresults,
reviewed by J.C1.Andr(•, of the HAPEX-MOBILHY
programshow that variabilitywithin the experimen-
talzoneof 104km2isdueto differences
in vegetation
cover and crop development, and that the water
budget locally balancesaccuratelyat the monthly
time scale. The climate-soil-vegetationinteractions
were alsoinvestigatedby P.S. Eaglesonusingsimpli-
fied physicalformulations.
Variationsin tropicalsea-surfacetemperatures,of
which the El Nifio phenomenonin the Pacificis the
best example, and the associatedchangesin the
global atmosphericcirculation were discussedby
K.E. Trenberth, J.M. Wallace and D. Halpern. The
historical record shows that ENSO events have cer-
tain featuresthat evolvein a systematicfashion,but
that each event has also its individual character. The
southern oscillation (worldwide) and the E1 Nifio
(moreregional)phenomenaare undoubtedlyrelated
but they do not exhibit a one-to-one relationship.
This is why, although the mechanismsresponsible
for theselargefluctuationsin the climatesystemare
beginningto be understoodin generalterms, there
remainmany fundamentalquestionsthat will haveto
be answeredbeforetheir amplitudeand durationcan
be forecasted.Given the impact of large scaleSST
anomalies,reliableestimatesof their evolutionup to
several months in adance is a goal of the TOGA
program. But achievingthis goal requiresincreased
understandingof the rolesof smallscalenear surface
mixingprocesses,of upper ocean3-D circulationand
tranport field and of the air-seafluxes in changing
SST.All the datarelatedto theseprocesses mustthus
be recorded(mainly from satellites)at the sametime
that realisticoceangeneralcirculationmodelsdevel-
oped.
The Earth'sradiationbudgetplaysa crucialrole in
understandingthe Earth's climate and potential
changescausedby externaland internalforcings.The
Earth RadiationsBudgetExperimentwill thus be of
primaryimportancein particularfor validatingradia-
tive transfer and GCM's, as emphasizedby B.R.
Barkstrom, but also for cloud studies. These are the
primary objectiveof the InternationalSatelliteCloud
ClimatologyProject describedby E. Raschkewho
reviewed also some other national and international
researchprojectsrelated to clouds.
The uncertain feedback effects of cloud-radiative
interactionswas also claimedby V. Ramanathanas
being one of the largestsourceof errors in model
estimatesof the tracegasimpactson climate.Raman-
athan showedthat the non-CO2trace gasesin the
atmosphereare now addingto the greenhouseeffect
by an amount comparableto the effect of CO2 in-
creaseand that the tracegas warming will become
large enoughto rise abovethe backgroundclimate
"noise"beforethe end of this century.The signifi-
canceof chemical-radiative-dynamicalinteractionsto
climateresponsehas alsobeen underlined.
Solar-terrestrial and aeronomic interactions with
the Earth'sglobalclimateplayeda key role through-
out the evolutionof our planet.J.C1.G(•rarddemon-
strated that the apparentstrongerresponseof the
precambrianatmosphereto solar cycleactivity was
possiblythe consequence of a lower 02 atmospheric
contentand a lessdevelopedstratosphere.G.E. Hunt
showedthat the fundamentalprocesses which gov-
ern the wide range of atmosphericmotions in the
planetaryatmospheresare often analogousto pro-
cessescontrollingthe circulationof the Earth'satmo-
sphereand oceansand thereforethe climate of our
planet.In relationto the impactsof the stratospheric
ozone to the surface climate and the recent Antarctic
ozone "hole", C. Leovy reviewed the modes of
transportof tracegasesinto the lower stratosphere
and upward into the photochemicallymost active
regionsnear 40 km.
The last sessionwas devoted to modeling our
futureclimate.S. Manabediscussed
one interesting
feature of the southern ocean circulation in relation
with the air-sea-ice interactions and with the inter-
hemisphericasymmetryin the transientresponseof
climate. A hierarchyof numericalclimatemodelsof
varying complexityare now available to study a
problemwhich may affectglobalclimate:W. Wash-
ington has illustrated their potential to simulate
changesin the climatesystemwhich may resultfrom
increasein CO2and/orin othertracegases.FinallyM.
Schlesinger analyzedthe feedbackswhich amplify
the direct radiative forcing of the increasedCO2
concentration into three different kinds of models:
EBMs, RCMs and GCMs. He concluded that such an
analysisis a usefulmethodof modelintercomparison
that providesinsighton the causesof the differences
in the models'simulatedCO2-inducedwarming.
All the sessionswereverywell attendedexceptthe
poster session.The nature of the papers and the
disciplineof the attendeesclearlydemonstratethat
the study of climaticchangesis a

one involvingall branchesof geophysics and others,
like chemistryand biology, in particular. It remains
thus one of the most relevant examplesfor global
changesstudieswithin an InternationalGeosphere-
BiosphereProgram.As is customary,these papers
were all reviewedby the authors'peersin order to
ensuretheir strictadherenceto the high standardsof
internationalresearchpublishing.The IUGG editors
expresstheir appreciationto all the scientistswho
gavetime and energyas refereesfor this expression
PREFACE xi
of the importantwork done throughoutthe Interna-
tional Union of Geodesyand Geophysics.
A. Berger is from the Universit(• Catholique de
Louvain, the Institute of Astronomyand Geophysics
G. Lemaitre in Louvain-la-Neuve, Belgium; R. E.
Dickinsonis from National Center for Atmospheric
Researchin Boulder, Colorado and John W. Kidson is
from New Zealand MeteorologicalService,Welling-
ton, New
Section I

WORLD CLIMATE
RESEARCH
Geophysical Monograph Series Understanding Climate Change Vol. 52

THE WORLD CLIMATE RESEARCH PROGRAMME

Gordon A. McBean

CanadianClimate Centre, Institute of Ocean Sciences,Sidney, B.C., Canada

and
Vice-Chairman
WMO/ICSU Joint ScientificCommittee for the WCRP

Introduction coupled, global climate system (atmosphere,

oceans, lithosphere and cryosphere) which varies
In 1979, the Global Atmospheric Research on essentially all time and space scales. At the
Programme (GARP), jointly sponsored by the beginning of GARP, the range of deterministic
International Council of Scientific Unions and weather prediction was 1-2 days; by the early
the World Meteorological Organization, culminated 1980's, improvements in observing systems and
with the Global Weather Experiment. In the same forecast models had extended this to 5-7 days and
year, the First World Climate Conference was held forecasts up to two weeks are deemed possible.
in Geneva and focussed the scientific community's This provides a shorter time scale limit for the
attention on the importance of the variability climate program. In terms of impact on national
and possible change of the earth's climate. The and international policies, the upper limit has
result was the establishment of the World Climate been set at several decades. Thus, the WCRP is
Programme, which has components for applications, concerned with time scales of weeks up to one
data, impacts and research. century. To focus the activities more sharply
The objectives of the WMO/ICSU World Climate this wide range has been divided into three
Research Programme (WCRP) are to determinine: streams (Houghton and Morel, 1984).
- to what extent climate can be predicted; and The First Stream is concerned with extended-
- the extent of man's influence on climate.
range (monthly to seasonal) weather forecasting.
To achieve these objectives, it is necessary In certain instances, such as blocking
(k•O, 1984): situations, the atmosphere appears to have a
(a) to improve our knowledge of global and longer period of predictability. Further, the
regional climates, their temporal influence of surface boundary conditions, such as
variations, and our understanding of the sea surface temperature and soil moisture
responsible mechanisms; anomalies, provide a basis for extended
(b) to assess the evidence for significant prediction. To take advantage of this situation
trends in global and regional climates; we need to improve' the internal dynamics and
(c) to develop and improve physical-mathematical thermodynamics of atmospheric models; our
models capable of simulating and assessing capability of observing surface boundary
the predictability of the climate system conditions; and modelling of their interactions
over a range of time and space scales; and with the atmosphere. If we consider the earth as
(d) to investigate the sensitivity of climate to four interacting components (Figure 1), then the
possible natural and man-made stimuli and to First Stream of the WCRP considers the atmosphere
estimate the changes in climate likely to and its direct, short-time scale interactions
result from specific disturbing influences. with the other components.
The next major time scale of change, the
Strategy and Structure of the WCRP Second Stream, is interannual variability for
which the strongest signal appears to be the
Whereas GARP focussed primarily on atmospheric interaction of tropical oceans with the global
dynamics, the WCRPmust consider the fully- atmosphere. Secondary effects may occur due to
variations in sea-ice boundaries. Initial
studies have shown that a global atmospheric
Copyright 1989 by model interacting with the tropical oceans will
International Union of Geodesy and Geophysics provide useful predictions of climate variability
and American Geophysical Union. on interannual timescales. Thus, Second Stream
Geophysical Monograph Series Understanding Climate Change Vol. 52

4 THE WORLD CLIMATE RESEARCH PROGRAMME

S S
T T
R R
E E
A A
M M

i II-- STREAM
2
STRE
AM :5
Fig. 1. Schematic box diagram of the earth sciences showing how the WCRP's three stream
strategy includes study of increasingly larger fractions of the earth. The main WCRP
activities are indicated on the boxes. See the text for further explanation.

research must include the tropical oceans to a WCRP Activities

depth of a few hundred metres.
In the Third Stream, for periods of decades Numer$cal •xperimentat$on
and longer, we must deal with interactions of the
atmosphere with the global ocean and sea ice, and Numerical experimentation, including model
impacts of changing atmospheric concentrations of development and data assimilation, is being used
radiatively-important gases. Some aspects of the to test physical concepts and to identify
oceans and major parts of the cryosphere and the modelling and observing system deficiencies. The
lithosphere will remain beyond the scope of the numerical experimentation project is closely
WCRP. linked with numerical weather prediction
It is important to stress that the strategy of activities. The areas of study include:
the WCRP is a building block approach; everything systematic errors and climate drift in
required for the First Stream is required for the atmospheric circulation simulations; estimation
Second, and so on. of ocean-atmosphere fluxes; sensitivity of the
Geophysical Monograph Series Understanding Climate Change Vol. 52

MCBEAN 5

atmospheric circulation to tropical sea-surface Global Observing Systems

temperature anomalies; and formulation of cloud
and radiation feedback processes. The Working The global observations needed for the WCRP
Group on Numerical Experimentation is developing must come from an enhanced World Weather Watch
a strategy of model development for refining (WWW) and utilize space-based observing
estimates of the climatic impact of radiatively- techniques whenever possible. The WCRP is
active gases. cooperating with the WWW,the World Climate Data
Programme, and others to improve the data bases
Clouds and Radiation for improving our knowledge of regional and
global climates and their variability. In the
Diabatic heating becomes more important as the last section of this report, still another
length of the time integration increases. The initiative, a Global Energy and Water Cycle
WCRPhas placed high priority on the study of Experiment (GEWEX), built upon the next-
clouds and radiative transfer. The International generation of earth observing satellite system,
Satellite Cloud Climatology Project (ISCCP) was will be discussed.
started in 1982 to provide consistent global data
sets of cloud types, distribution and optical Tropical Oceans - Global Atmosphere - TOGA
properties. Radiances from all operational
meteorological satellites, except INSAT, are now The Tropical Ocean - Global Atmosphere
being routinely collected and delivered to a Programme (WMO, 1985) has process and modelling
Global Processing Centre. The Project also studies and an enhanced observation programme
provides for periodic satellite-to-satellite including numerous tropical sea level stations
radiance intercalibration. Another major and XBT lines, plus seven data management centres
activity has been the organization of inter- (level III atmospheric; sea surface temperature;
comparisons of radiation codes used in climate precipitation' marine climatology' tropical
models. The Earth Radiation Budget Experiment upper-air winds; sea level; ocean sub-surface).
has provided the first calibrated and global set TOGA started in 1985 and will continue until
of moderately high-resolution radiation flux 1995. There are still some important data gaps,
data. Oneof the ERBE
instrume.n•s
has measured particularly over the Indian Ocean. Promising
the solar constant as 1364.8 Wm , with an results are now being obtained from uncoupled
accuracy of better than 0.2 %. A strategy for tropical ocean models with realistic specified
on-going measurement of global radiation fluxes interface conditions and there is considerable
is being developed. optimism in the TOGA community of developing
fully-coupled models. The TOGA Scientific
Land-Surface Processes and Climate
Steering Group has recently agreed that the main
thrusts of TOGA should be:
Several studies (see Mintz, 1984) have shown Thrust I: Development of an operational
that changes in soil moisture and evaporation capability for dynamical prediction
rates can have significant impact on long-range of the coupled tropical ocean/global
forecasts over periods of one month or more. The atmosphere system beginning with the
interaction of the atmosphere with the land current state, that is, prediction
surfaces also has obvious important economic of time-averaged anomalies up to
implications for agriculture and forestry. several months in advance.
However, the extreme heterogenetty of the land Thrust II: Exploration of the predictability of
surfaces has made parametertzatton of their the longer time-scale climate
interactions with the atmosphere very difficult variations of the tropical
to model. To study albedo, evapotransptration, ocean/global atmosphere system on
changes in soil moisture, runoff, etc., which are time scales of one to several years
very complicated processes, the WCRPhas and understanding the mechanisms and
initiated a series of Hydrological-Atmospheric processes underlying the
Experiments (HAPEX). The first, HAPEX-MOBILHY predictability.
(Andre et al., 1986), was conducted in a 100-km In view of the global impact of E1 Ntno and
square area of southwest France in 1986. It is other tropical ocean phenomena, there would be
hoped that similar experiments will be conducted great benefits in being able to predict the
in seasonally snow-covered, boreal forest development and evolution of an E1 Nino.
regions, in the tropical rain forests and in
other representative land-surface areas. The World Ocean Circulation Experiment - WOCE
WCRP is also cooperating with the International
Satellite Land-Surface Climatology Project It is clear that decadal and longer variations
(ISLSCP, co-sponsored by IAMAP, COSPARand UNEP). in climate depend critically on the behaviour of
A series of satellite pixel-scale experiments are the global oceans. We can now see new space-
underway. The First ISLSCP Field Experiment based (altimeters and scatterometers) and in situ
(FIFE) was conducted in a 10-km square area of observing techniques to study the global oceans
the central United States in 1987. as was never feasible before. At the same time,
Geophysical Monograph Series Understanding Climate Change Vol. 52

6 THE WORLD CLIMATE RESEARCH PROGRAMME

advances in computing power and analytical transports of mass, heat and ice through Fram
techniques are making it possible to consider Strait will provide a constraint on Arctic Ocean
comprehensive global-ocean models. The World models. The air-ice-sea component of the
Ocean Circulation Experiment (WMO, 1986) has two Greenland Sea Progrmmme will provide data for
basic goals: development and validation of sea ice models.
I To develop models useful for predicting Ways of measuring the vertical oceanic heat flux
climate change and to collect the data are being explored.
necessary to test them; and Although Arctic sea ice variations are
II To determine the representativeness of reasonably well documented by existing
specific WOCEdata sets for the long-term observations, the situation in the Antarctic is
behaviour of the ocean, and to find methods much more difficult. In cooperation with SCOR
for determining long-term changes in the ocean and SCAR, an Antarctic sea ice project is being
circulation. developed to coordinate ship-based hydrographic
The WOCEscientific plan identifies three main surveys, deploy suitably instrumented data buoys
within the Antarctic sea ice zone and examine the
thrusts or 'Core Projects":
(1) The Global Description - essentially to formation of deep ocean and bottom water. These
raise the level of description of the global activities will constitute a part of or will be
oceans to that presently available for the closely co-ordinated with the WOCESouthern Ocean
North Atlantic, with the addition of the Project.
measurement of sea level and wind stress by The optical properties of clouds and their
satellite-borne systems. effect on radiative fluxes are very important to
(2) The Southern Ocean Experiment - concerned the heat balance of the polar regions. The ISCCP
with the Antarctic Circumpolar Current and (see (ii) above) has a special project on
its interaction with the oceans to the developing algorithms for detection of polar
north. clouds from space. Improvements in modelling the
albedo of sea ice and snow are also essential.
(3) The Gyre Dynamics Experiment - designed to
support the development of eddy-resolving The important hydrological and snow-albedo
oceanic circulation models, with priority feedback processes at high latitudes are to be
included in a boreal forest HAPEX of the Land-
given to studying small-scale dynamical
processes that may need to be parameterized Surface Processes and Climate Progrmmme(iii).
in global ocean models.
The intensive WOCE observations will be Global Energ•y and Water Cycle Experiment - GEWEX
concentrated within a five-year period in the
early 1990s to coincide with the time- frame of The distribution, duration and amounts of
the altimetry and scatterometer satellites. Some precipitation constitute the most significant
manifestations of climate for human activities
ship-borne surveys are already underway.
andare sensitive indicators of climate change.
Polar Regions Programme Alterations of precipitation by a CO -induced
climate change may have more socioeconomic impact
While the tropical regions are the main than a temperature rise. However, at present,
recipients of solar radiation, the polar regions the representation of the water cycle in global
are the main sink of radiative energy. To fully weather and climate models is still too crude to
understand the global climate we must examine allow useful projections of future conditions.
both. Variations in the extent and thickness of Looking ahead to the next decade, it is
sea ice in the polar regions (Untersteiner, 1984) expected that there will be increased computer
have an impact on both interannual and longer capability to deal with the finer-mesh models
time-scale variability. The uptake of that will be required to deal adequately with
atmospheric carbon dioxide into the oceans occurs clouds and precipitation. The WCRP's and others'
principally at high latitudes and so does the progrmmmes on clouds and radiation, numerical
formation of deep ocean water. The Sea Ice and experimentation, land- surface processes and other
Climate Progrmmme is the main WCRPactivity in areas will have greatly improved our
the polar regions but other studies are also understanding of the processes of the global
relevant and hence included here. Towards the water cycle.
objective of developing fully interactive There is now growing concern about major
atmosphere-ice-ocean models, sea ice models will global changes. Water and energy cycles are
be tested with time-dependent atmospheric forcing critical and are the basis on which all other
fields (from observations) and specified oceanic global cycles must rely.
heat fluxes for their ability to reproduce the In the period 1995 and beyond, it is expected
time-dependent distribution of sea ice over that new space platforms will offer greatly
periods of several years. The performance of increased opportunities for new remote sensing
atmospheric models in polar regions is being instruments and earth observation. They will
evaluated. The WCRP cooperates with the Arctic make possible the deployment of much larger and
Ocean Sciences Board in its study of the higher-power instruments, such as active remote-
Greenland Sea. The determination of the sounding radars and lidars.
Geophysical Monograph Series Understanding Climate Change Vol. 52

MCBEAN 7

TIME SCHEDULE OF MAJOR WCRP ACTIVITIES

1985 1990 1995

TOGA , I

I WOCE !

ERS-I,2
NSCAT
TOPEX/POSEIDON

AIR- SEA FLUXES

ISCCP - --

G
RADIATION PROGRAMME-•- E
W
SEA ICE E
X
PLSPC/ISLSCP
WINDSAT

',[•1 $ c RAINSAT
M F 0 H
0 I V I ß

B F I N
I E E A
L T
H
Y

AIgorithm development
Model development,
NUMERICAL EXPERIMENTATION

Global Observations
Fig. 2. Time schedule of major WCRP activities. See text for further explanation

The Joint Scientific Committee is now solid) and energy within the global atmosphere
formulating a strategy for a Global Energy and and on the underlying surface, which may occur
Water Cycle Experiment, GEWEX, which would start naturally or through the influence of man's
in the period 1995+ and run about five years. activity.
The proposed objectives for GEWEXare:
I To describe and understand the transport of Greenhouse Gases Programme
water (vapour, liquid and solid) and energy in
the global atmosphere and at the underlying It is evident that changing concentrations of
surface; and greenhouse gases (e.g., water vapour, carbon
II To develop methods of predicting changes in dioxide, ozone, methane, chloro-fluoro-carbons)
the distribution of water (vapour, liquid and will have a major impact on the earth's climate.
Geophysical Monograph Series Understanding Climate Change Vol. 52

8 THE WORLD CLIMATE RESEARCH PROGRAMME

The WCRPis already developing models of of IOC/SCOR and their working groups and
radiative fluxes and their effect on the Prof. P. Morel, Director/WCRP have all made
atmosphere. A new major programme on the water special contributions.
cycle is being envisaged. What is still needed
is the development of comprehensive global models References
of the dynamical, physical, chemical and
biochemical interactions which control the large- Andre, J.C., J.P. Goutorbe and A. Perrier, 1986:
scale distribution of greenhouse gases and HAPEX-MOBILHY, A hydrologic atmospheric pilot
related chemical species. A specific programme experiment for the study of water budget and
towards this objective is under development. evaporation flux at the climatic scale. Bull,
Amer, Meteor, Soc., 67, 138-144.
Summary Houghton, J.T., and P. Morel, 1984: The World
Climate Research Programme. In The Global
The World Climate Research Programme is a Climate (J.T. Houghton, ed.), Cambridge Univ.
multi-component, endcavour to address the large- Press, 1-11.
scale variability of the global climate system. Mintz, Y,. 1984: The sensitivity of numerically
Several activities are underway (Figure 2)(e.g., simulated climates to land-surface boundary
TOGA, HAPEX-MOBILHY, ISCCP) and several more are conditions. In The Global Climate (J.T.
soon to start (e.g., WOCE). The WCRPis looking Houghton, ed.), Cambridge Univ. Press, 79-105.
ahead to a global water cycle experiment to start Untersteiner, N., 1984: The cryosphere. In The
before the end of the century. The WCRPcalls Global Climate (J.T. Houghton, ed.), Cambridge
upon scientists from many disciplines and Univ. Press, 121-140.
nationalities to work together towards WMO, 1984: Scientific Plan for the World Climate
understanding and prediction of the global Research Programme. WCRP-2, September. WMO/ICSU
climate. Only with their assistance can the (available from WMO, Geneva).
programme be a success. WMO, 1985: Scientific Plan for the Tropical Ocean
and Global Atmosphere Programme. WCRP-3,
September. WMO/ICSU (available from WMO,
Acknowledmments. _
The WCRP has been Geneva).
developed through the efforts of many scientists. WMO, 1986: Scientific Plan for the World Ocean
The members of the Joint Scientific Committee, Circulation Experiment. WCRP-6, July, WMO/ICSU
the Committee on Climate Changes and the Oceans (available from WMO, Geneva).
Section II

PALEOCLIMATES AND
Geophysical Monograph Series Understanding Climate Change Vol. 52

LONG-TERM CLIMATIC AND ENVIRONMENTAL RECORDS FROM ANTARCTIC ICE

•C. Lorius, •J-M. Barnola,•M. Legrand, •J. R. Petit, •D. Raynaud, •C. Ritz,
2N. Barkov, 2Y. S. Korotkevich, •V. N. Petrov, 3C. Genthon, 3j. Jouzel,
4v. M. Kotlyakov, SF.Yiou, and SG. Raisbeck.

Abstract. Various records obtained from the greater exposure of continental shelves. On the
Vostok (East Antarctica) ice core allow recons- other hand there is no indication of a long term
truction of temperature, accumulation (precipita- relationship between volcanism and climate.
tion), aerosol loading and atmospheric CO concen-
tration histories over the last climatic cycle Introduction
(160 000 years). The results agree with those pre-
viously obtained from two other deep Antarctic ice Data on environmental and climatic changes
cores going back to the Last Glacial Maximum. having affected our planet in the past can be
The Vostok isotope-based temperature and CO obtained from various marine and terrestrial sedi-
records show a large 100 ky signal with changes ments and from ice cores. Although ice-core stu-
of the order of 10øC and 70 ppmv respectively. dies have some inherent limitations (small number
They are closely associated and show periodicities of available records, relative shortness of the
characteristic of the earth orbital parameters. period covered, difficulty of accurate dating},
These features suggest a fundamental link between they also offer certain advantages and unique pos-
the climate system and the carbon cycle and point sibilities. In particular, they can provide high
out the possible role of CO , in addition to inso- resolution records and access to the most impor-
lation inputs, in accountin•for the observed
tem- rant climatic parameters
(temperature,precipita-
perature history. tion, relative humidity and wind strength) as well
The accumulation (precipitation} record appears as the past atmospheric composition including
to be governed by temperature with values during trace gases and aerosols of various origins.
the coldest stages reduced to about 50 % of the This paper focuses mainly on the isotope,
current rate. Ice deposited during these coldest aerosol and CO data from the Vostok ice core over
stages is also characterized by high concentra- the last 160 k• and their climatic interpreta-
tions of marine and terrestrial aerosols ; these tion. For the Last Glacial Maximum, these results
peaks likely reflect strengthened sources and essentially confirm those already obtained from
meridional transport during full glacial condi- the two other Antarctic deep ice cores (Byrd and
tions, linked to higher wind speeds, more exten- DomeC} as shortly discussed for each of these pa-
sive arid areas on surrounding continents and the rameters. Over the full glacial-interglacial
cycle, the Vostok isotope-based temperature
history is discussed in terms of spectral charac-
• teristics and in relation to CO , orbital forcing,
Laboratoire
de l'Environnement,
de Glaciologie
B.P.
96
et de G•ophysique
38•02 St Martin
precipitationandaerosolloading.
d'H&res
2
cedex (France). The Vostok Ice Core
Arctic and Antarctic Research Institute,
BeriDgaStreet 38, 199236Leningrad
(USSR). The Soviet Antarctic station of Vostok is
3Laboratoire
deG&ochimie
Isotopique-LODYC
(UA located
in EastAntarctica
(78ø28'S
and106ø•8'E)
CNRS1206) CEA/IRDI/DESICP/DPC
91191Giœsur at an elevationof 3•90m, with a mean
annualtem-
Yvettecedex(France). peratureof - 55ø C anda current snowaccumula-
Institute of Geography,
Academy
of Scienceof tion of about2.3 g cm
-2 yr-1 Aninitial dril- .

the USSR, 29 Staronometny,

Moscow
109107(USSR). ling went downto 950 m in successivesteps
5Laboratoire
Ren8Bernas,
C. S. N. S. M.91406 between
1970and1974.In 1980,thedrillingof a
0rsay (France). seconddeephole wasstarted. Data are nowavaila-
ble on an almost continuous basis for the isotopic
(deuterium) composition of the ice and on a dis-
Copyright 1989 by continuous basis for impurities in ice and CO
International Union of Geodesyand Geophysics contained in air bubbles downto a depth of •
and American Geophysical Union. 2083 m.
Geophysical Monograph Series Understanding Climate Change Vol. 52

12 LONG-TERM RECORDS FROM ANTARCTIC ICE

Establishing a reliable time scale is the first Depth (m)

important step of data interpretation. For Vostok, 500 •ooo •5oo 2000
this was established through a glaciological I I I I
3'0 5B ?.• 106
approach using a two-dimensional ice flow model,
well suited to the Vostok area, and accounting for
the change in snow accumulation with time [Lorius
C O E FG,
et al., 1985] ; this accumulation aspect is fully
discussed in the following section• A key feature
of the dating strategy is independence with ,

respect to other paleo-series. This dating puts

the 2083 m level at ~ 160 ky (with an accuracy of ß

about 10-15 ky).

!
The Climatic Record
8,e0 %0
The existence of a relationship between the -!
2
isotope content of snow (deuterium or oxygen 18)
and the temperature of the site is well documented oAT
over polar areas. Such a relationship results from
the fractionation processes which take place
.2øC
during the atmospheric water cycle and is the -4
c
basic tool for reconstructing past climatic condi-
'-,6
tions from ice cores. Our confidence in such tem-
perature reconstruction for central East Antarctic -.6
deep ice cores relies on : i) a particularly well
-10
obeyed linear relationship observed between the
annual averages of the surface temperature and of
the snow isotope content in this part of
Antarctica [Lorius and Merlivat, 1977], ii} the
good agreement between observed slopes and the
value derived from a one-dimensional isotope model
[Jouzel and Merlivat, 1984]. For the deuterium
content, upon which the following Vostok climatic : ß

interpretation is based, its value is 6 ø/oo/øC

(expressed in 6D per mill with respect to S.M.O.W, 0 50 lO0 iS0Age (k yr)
the Standard Mean Ocean Water).
Fig. 1. a) Vostok isotope profile (deuterium con-
The deuterium profile is given in figure la
tent in ø/oo versus S.M.O.W with successive clima-
with respect to depth (upper scale) and time
tic stages A to H as defined by Lorius et al.
(lower scale) with individual values representing
averages on 1 to 2 m ice increments. Stages from A
[1985] ; b) Marine6•80 recordof Martinsonet al.
[1986] ; c) Smoothed Vostok isotope temperature
to H designate successive warm and cold periods.
record expressed in øC as a difference with res-
Stage A corresponds to the present Holocene inter-
pect to current surface temperature value (from
glacial ; stages B to F cover the last glacial
with C and E being slightly warmer interstadials.
Jouzel et al.,1987] ß d) Vostok•øBe concentration
(from Raisbeck et al., 1987). Note inverted scale
Stage G characterizes the previous interglacial
to facilitate com parison with climatic records.
while H is the last part of the previous ice age.
The upper scale gives the depth of the Vostok
Using a gradient of 6 ø/oo/'C the deuterium re-
ice core ; the lower scale indicates the age
cord was interpreted in terms of temperature
(ky BP) of the various records.
after correcting for the isotopic changes of sea
water (fig. lb). No corrections were applied for
the possible influence of ice origin and ice cap
thickness changes. The isotope temperature curve al., 1979] records over the last deglaciation and
is given in figure lc as a difference with respect the Holocene periods. The temperature shift asso-
to the present day value, after smoothing to ciated with this climatic transition is around 9øC
filter out the very high frequency oscillations. both at Dome C and Vostok and its shape is
The record obtained shows that the shift associa- remarkably similar for the two cores. A quantita-
ted with the last deglaciation was around 9øC. tive interpretation of the Byrd profile would lead
This compares
quite well with the 11øC change to a quite comparable
value as the associated•0
derived solely from the crystal size variation shift is very similar in the three cores. This
with depth [Petit et al., 1987], providing inde- supports the large geographical significance of
pendent support for our temperature interpreta- the Vostok isotope temperature profile.
tion. The main features of this record are in fact
There is a very good agreement between Vostok, even of global significance, at least qualitative-
Byrd [Johnsen et al., 1972] and Dome C [Lorius et ly. as suggested by the comparison with the marine
Geophysical Monograph Series Understanding Climate Change Vol. 52

LORIUS ET AL. 15

8•s0 record (fig lb) of Martinson et al. [1987] both may be affected by atmospheric
circulation
which is thought to essentially represent global changes, the overall consistency between the two
continental ice volume changes. The two profiles methods supports the estimate of past accumulation
correspond very closely down to 110 ky BP. However changes and also indirectly supports the deuterium
such consistency is no longer observed for the temperature interpretation.
earlier part of the record (before 110 ky BP} as
there is a major difference in the duration of the CO and Climate
last interglacial estimated respectively from the 2
ice and marine sediments. This likely results from Atmospheric CO data obtained recently both
relative uncertainties in both chronologiesand from Greenlandand2Antarcticice cores have shown
does not affect the above conclusion on the repre- in particular that concentrations during the Last
sentativeness of the Vostok record. Glacial Maximum were lower than Holocene values by
Aside from the large ~ 100 ky signal, visual 25-30 % and that the last deglaciation was charac-
inspection of the temperature curve (fig. lc) also terized by a concentration increase from about
clearly shows a ~ 40 ky oscillation with four well 190-200 to 270-280 ppmv [De!mas et al., 1980 ;
marked temperature minima roughly in phase with Neftel et al., 1982].
the past total insolation at the Vostok latitude The Vostok ice core has made it possible to
(governed by the obliquity cycle with a period of extend the record of past atmospheric CO right
41 ky). Thereare also similarities betweenthis throughthe glacial-interglacial cycle [Ba•nola et
Vostok record and the 65øN July insolation changes al., 1987]. Measurements performed at 66 different
which play a key role in the Milankovitch theory depth levels are separated by time intervals
of the ice ages (this July 65 ø N insolation is ranging from about 2 to 4.5 ky. Due to the gradual
largely influenced by precessional changes with enclosure of atmospheric air in ice, air extracted
periodicities of 19 and 23 ky}. Spectral analysis from the core is younger than the age of the snow
[Jouzel et al., 1987] confirms these visual featu- deposit and this difference (ranging from 2.5 to
res showing that aside from the ~ 100 ky glacial- 4.3 ky depending upon climatic conditions) has
interglacial oscillation, the Vostok temperature been taken into account in establishing the CO
record is dominated
by a strong ~ 40 ky signal. It time scale. The best estimatesof the CO concen5
also suggests a lesser influence by a component trations are plotted (fig. 2a) togethe• with the
slightly greater than 20 ky. The two frequency associated uncertainty bands. The CO concentra-
bands can be associated with the obliquity and tion exhibits twovery large change•betweentwo
precession cycles respectively, thus supporting levels centred around 190-200 and 260-280 ppmv
the role of astronomical forcing in determining with the low and high values associated with full
the late Pleistocene climate [Berger, in press], glacial and interglacial conditions respectively.
already convincingly demonstrated on the basis of The low values previously recorded in other ice
deep sea core records [Hays et al., 1976]. cores [Delmas et al., 1980 ; Neftel et al., 1982]
Beyond this temperature record, Vostok data are fully confirmed. The high level is comparable
allow evaluation of past accumulation changes. As with the so-called "pre-industrial" CO2 concentra-
noted above knowledge of this parameter is requi- tion which prevailed about 200 years ugo.
red for dating ; also, in a more global perspec- Indeed there ms a remarkable correlation (r =
tive growth and retreat of ice sheets largely ß 79) between the Vostok CO and temperature
depend on accumulation
changesoccurringover the records (figures 2a and2b}. Althoughsome diffe-
glacial-interglacial time scale. One approach (the fences are observed between the two records, such
one used for dating} assumes [Robin, 1977] that as the absence of a low CO value associated with
the precipitation rate is governedby the amount the rather cold stage •round 110 ky BP and the
of water vapor circulating above the inversion existence of a CO2 -temperature lag when going from
layer, itself controlled by the saturation vapor warm to cold periods, these results nevertheless
pressure, i.e. by the temperature [Lorius et al., provide the first direct evidence of a close asso-
1985). From the Vostok temperature data we deduce ciation between atmospheric CO and climatic (tem-
that snowaccumulationwas quite similar during perature) changeson a glacial-interglacial time
the Holocene and the previous Interglacial periods scale.
and reduced to about 50 % of the modern value Although there are too few CO data to study
during the coldest periods. Another approach the phaserelation with the isotOpe-temperature
et al.f0 1981
[Raisbeck ; Yiou et al ., 1985] is record , such a close association suggests that CO
based on the Be profile (fig. ld). At Vostok, mayhave, throughits radiative effect and th•
concentrations of this cosmogenic isotope were associated feedbacks, participated in the glacial-
quite similar during interglacials. The glacial- interglacial change. Indeed a simple linear mul-
interglacial concentration changes have been tivariate analysis suggests that CO changes may
interpreted as reflecting lower precipitation have accountedfor morethan 50 % of2 the Vostok
rates during cold periods. Indeed, apart from two temperature variability, the remaining part being
beryllium peaks [Raisbeck et al., 1987], there is associated with orbital forcing [Genthon et al.,
& goodcorrelation betweenthe •recipitation rate 1987]. Qualitatively these results agreewith the
estimatedassuming a constant•VBedeposition flux findings of Broccoli andManabe[1987] who found
and that derived from temperature change. Although that the low CO concentration was the main cause
Geophysical Monograph Series Understanding Climate Change Vol. 52

LONG-TERM RECORDS FROM ANTARCTIC ICE

C02 tions [Wanget al., 1986] or, as recently sugges-

ppmv ted by Charlson et al. [1987], in the amount of
A H • cloud condensation nuclei which in the marine
• atmosphere are probably mainly formedfrom the
oxidation products of the dimethylsulfide emitted
• by the ocean.
2 2• An important spectral characteristic of the CO
record is that, in addition to the 100 ky signal]
AT0 • it suggests
a varianceconcentration
around20 ky
ø C o2 • [Barnola et al., 1987]. Beyond significant impli-
t• cations related to the understanding of modifica-
-6 tions of the carbon cycle, this suggests that
astronomical forcing is involved in both tempera-
ture and CO changes, although the role of this
process is complex and not yet understood.
Nevertheless the close correlation between CO
t• and temperaturerecords and their spectral charac•
AI teristics supports the idea that climatic changes
•ng.g-• couldbe triggered by an insolation input, with
the relatively weak orbital forcing strongly
0 amplified by possibly orbitally induced CO
changes. In particular these results suggestthat
the 100 ky glacial-interglacial oscillation may
be related to the observed large CO variations
rather than to postulated non-linearitzes in ice
sheet growth and decay forced by insolation
0 changes. This interpretation can only be consi-
dered as tentative until backed by a clear
ng.g-• __ understanding of the physical mechanisms involved.
S04 But in any case the results obtainedfrom the
ng.g-• Vostok ice core convincingly indicate that there
is an interactive link between orbital forcing,
CO and climate, supporting Sundquist [1987] in
Age(kyr) hi• suggestion
that further progresswill require
climate and the carbon cycle to be treated as part
Fig. 2. Vostok ice core records; a) CO_
l. concentra- of the same global system rather than as separate
tions (ppmv)with envelope of uncertaznty (from entities.
Barnola et al., 1987) ; b) Smoothed Vostok isotope
temperature record (as in Fig. lc) ; c) Aluminium
content (from Legrand et al., 1988) ; d) Marine The Aerosol Record
Sodium content (from Legrand et al., 1988) ; e)
Sulfate content (from Legrand et al., 1988). The basic assumption for reconstructing past
aerosol data from ice cores is that the concentra-
tion of impurities in the snow layers is directly
of Last Glacial Maximum conditions in the southern related to the concentration in the atmosphere. A
hemisphere. A more quantitative agreement would common characteristic of the Byrd [Cragin et al.,
likely be obtained by taking into account all the 1977 ; Palais and Legrand, 1985], DomeC [Petit et
potential climatic feedbacks possibly combined al., 1981 ; Thompsonand Mosley-Thompson, 1981]
with a polar amplification (see Genthon et al., and Vostok [De Angelis et al., 1987 ; Legrand et
1987). al.,1988] cores is a much larger impurity con-
Our statistical approach [Genthon et al., 1987] centration during the Last Glacial Maximum than
is of course very simplistic. First non-linear in- during the Holocene. These changes, affecting both
teractions certainly exist between the different the continental and to a lesser extent the marine
components of the climatic system [Imbrie and contributions, cannot be explained by accumulation
Imbrie, 1980 ; Le Treut and Ghil, 1983 ; Saltzman, variations.
1987]. Second the role of the ocean, most likely For the input of dust mainly of continental
important in linking Northern and Southern hemi- origin the LGM/Holocene ratio is ~ 8 for Byrd (a
sphere climates, is not specifically taken into core showing many ash layers which may have a
account, and neither are changes of sea ice extent local volcanic origin} ~ 27 for Dome C and up to
which are also believed to have an important 30 for Vostok. Marine aerosol concentrations were
climatic role. It would also be useful to explore also higher for the LGM with LGM/Holocene ratios
a wider range of possible forcings such as changes of about 2 to 3 for Byrd and ~ 5 both at Dome C
in the atmospheric optical depth resulting from and Vostok. Besides a possible effect connected
variations in the aerosol loading [Rasool and with a lower accumulation these high glacial con-
Schneider, 1971] and in other trace gas concentra- centration values have been interpreted as resul-
Geophysical Monograph Series Understanding Climate Change Vol. 52

LORIUS ET AL. 15

ting from strengthened sources and meridional Acknowledgments. This work was supported in
transport linked to higher wind speeds (likely France by several institutions (Terres Australes
induced by higher temperature gradients with lati- et Antarctiques Fran•aises, CNRS-Institut National
rude}, more extensive arid areas over the surroun- des Sciences de l'Univers and C.E.A} and in the
ding continents and the greater exposure of conti- USSR by the Arctic and Antarctic Research Insti-
nental shelves due to a lower sea level [Petit et tute. It is based on a drilling programme
al., 1981]. More generally marine and terrestrial performed by Soviet Antarctic Expeditions. We are
aerosol concentrations measured in ice are also grateful to the NSF-Division of Polar
strongly dependent upon climatic conditions of Programs who provided logistic support.
global {source strength and atmospheric transport
efficiency}, regional {sea ice extent) and local References
{rate of snow accumulation) concern. Changes in
continental and marine inputs and in all major De Angelis, M. , N.I. Barkov and V.N. Petrov,
ions is now documented over the full glacial- Aerosol concentrations over the Last Climatic
interglacial cycle [De Angelis et al., 1987 ß Cycle (160 ky) from an Antarctic ice core.
Legrand et al.' 1988]. As an illustration, the Nature, 325, 318-321, 1987.
Vostok aluminium record, an indicator of continen- Barnola J.M., D. Raynaud, Y. S. Korotkevich and
tal aerosol, is given in figure 2c as evaluated by C. Lorius, Vostok ice core : a 160,000 year
Legrand et al., in press. This record is characte- record of atmospheric CO , Nature, 329, 408-414,
rized by sharp increasesduring cold stages B, D 1987. 2
and H and also by a slight increasing trend in the Berger A., Milankovitch theory and climate, Review
A1 background over the last glacial period which of Geophysics, in press.
may be explained by the emergence of large parts Broccoli, A.J. and S. Manabe, The influence of
of the continental shelves due to sea level continental ice, atmospheric CO2 and land
lowering [De Angelis et al., 1987] and also con- albedo on the climate of the Last Glacial
nected to the snow accumulation trend. As for the Maximum,Climate Dynamics, •, 87-99, 1987.
LGM, the sharp increase of continental input at Charlson, R.J., J.E.Lovelock, M. 0. Andrae and
the end of the previous ice age and during the S.G. Warren, Oceanic phytoplankton, atmospheric
cold stage D may be explained by expansion of arid sulfur cloud albedo and climate, Nature, 326,
areas, higher emissions due to higher wind speed 655-661, 1987.
over continents and more efficient atmospheric Cragin, J.H., M. M. Herron, C. C. Jr. Langway,
transport. There is not yet a clear explanation of G. Klouda, Interhemispheric comparison of chan-
the absence of large concentration values during ges in the composition of atmospheric precipita-
stage F, but this suggests that global climatic tion during the late cenozoic era. in Dunbar
conditions may have been different from full Maxwell J. (ed.), Polar oceans, Proceedings of
glacial during this stage at the beginning of the the polar oceans conference, Montreal, 1974,
last glacial period. Calgary, Arctic Institute of North America, 617-
The marine aerosol record (as illustrated by 63•, 1977.
the Na marine concentration profile in fig. 2d) Delmas, R.J., J. M. Ascencio and M. Legrand, Polar
shows also higher values during cold climatic con- ice evidence that atmospheric CO , 20,000 yr BP
ditions superimposed
on a well marked increasing was 50 % of the present, Nature• 284, 155-157,
trend for the concentrations during the last 1980.
glacial period. In this case we observe a moderate Genthon C., J.M. Barnola, D. Raynaud, C. Lorius,
increase of marine deposits during stage F, possi- J. Jouzel, N.I. Barkov, Y. S. Korotkevich and
bly related to climatic changes around Antarctica V.M. Kotlyakov, Vostok ice core : the climate
as depicted in the Vostok temperature record. In response to CO and orbital forcing changes
comparison,the S0,-- profile (fig. 2e) exhibits over the last •limatic cycle (160,000 years),
rather weakvariations (up to a factor 3). Diffe- Nature, 329, 414-418, 1987.
rent sources contribute to the observed inputs ; Hays, J.D., J. Imbrie and N.J. Shackleton, Varia-
continental dust and sea salt inputs are higher tions in the Earth orbit : pacemaker of the ice
during cold climatic stages but the acidic-gas ages, Science, 194, 1121-1132, 1976.
(H_S0j) derived contribution remains relatively Imbrie J and J Z Imbrie Modelling the climatic
stable over the whole climatic cycle, indicating response to orbital variations, Science, 207,
the absence of a long-term relationship between 243-253, 1980.
volcanism and climate. More generally a comprehen- Johnsen, S. J., W. Dansgaard, H.B. Clausen and
sive study of all major ions present in the ice C.C. Langway, Oxygen isotope profiles through
points out significant changes in the aerosol the Antarctic and Greenland ice sheets, Nature,
loading associated with climate which may have 2•5, 429-434, 1972.
some impact on the radiation balance of the earth Jouzel,J. and L. Merlivat, Deuterium and oxygen 18
atmosphere system. The mineral acid contribution in precipitation. Modelling of the isotopic
represents a large part (over 50 %) of ice impuri- effect during snow formation, J. Geophys. Res.,
ties deposited during interglacials. For glacial 89, 117•9-11757, 198•.
ice the inputs of marine and terrestrial salts Jouzel C., C. Lorius, J. R. Petit, C. Genthon,
become preponderant, representing up to 75 % of N.I. Barkov, V.M. Kotlyakov and V. N. Petroy,
total soluble impurities [Legrand et al., 1988]. Vostok ice core : a continuous isotope tempera-
Geophysical Monograph Series Understanding Climate Change Vol. 52

16 LONG-TERM RECORDS FROM ANTARCTIC ICE

ture record over the last climatic cycle sol content from East Antarctic ice core samples
(160,000 years), Nature, 329, 403-408, 1987. and past wind strength, Nature, 293, 391-394,
Legrand M., C. Lorius, N.I. Barkov, V. N. Petrov, 1981.
Vostok (Antarctica) ice core ß atmospheric Petit, J. R., P. Duval and C. Lorius, Long-term
chemistry changes over the last climatic cycle climatic changes indicated by crystal growth in
(160,000 yr), Atmospheric Environment, 22, 2, polar ice, Nature, 326, 62-64, 1987.
317-331, 1988. Raisbeck, G.M., F. Yiou, M. Fruneau, J.M.
Le Treut, H. and M. Ghil, Orbital forcing, cli- Loiseaux, M.
•0
Lieuvin, J.C. Ravel and C. Lorius
matic interactions and glaciation cycles, J. Cosmogenic Be concentrations in Antarctic ice
Geophys. Res., 88, 5167-5190, 1983. during the past 30 000 years, Nature, 292, 825-
Lorius, C. and L. Merlivat, Distribution of mean 826, 1981.
surface stable isotope values in East Antarctica Raisbeck G.M.,= F. Yiou, D. Bourles, C. Lorius,
observed changeswith depth in a coastal area. J. Jouzel, andN. I. Barko•d Evidence for two
In ß Isotopes and Impurities in Snow and Ice. intervals of Enhanced Be Deposition in
Proc. Grenoble Symp. August-September 1975, IAHS Antarctic Ice during the Last Glacial period,
Nø 118, 127-137, 1977. Nature, •26, 273-277 ,1987.
Lorius, C., L. Merlivat, J. Jouzel and M. Pourchet Rasool, S.I. and S. H. Schneider, Atmospheric
A 30,000 yr isotope climatic record from carbon dioxide and aerosols : effects of large
Antarctic ice, Nature, 280, 644-48, 1979. increases on global climate, Science, 173, 138-
Lorius, C., J. Jouzel, C. Ritz, L. Merlivat, 141, 1971.
N.I. Barkov, Y.S. Korotkevich and V.M. Kotlyakov Robin, G. de Q., Ice cores and climatic changes,
A 150, 000 year climaticrecordfromAntarctic Land.,B 280, 143-•6 • ,
Phil.Trans.R.Soc. •977.
ice, Nature, •16, 591-596, 1985. Sa[[•[ • C--•r•o••6xide andthe 81 0 record
Martinson, D. G., N. G. Pisias, J. D. Hays, J. of late-quaternary climatic change : a global
Imbrie, T. C. Moore and N.J. Shackleton, Age model, Climate Dynamics, [, 77-85, 1987.
dating and the orbital theory of the ice ages ß Sundquist, E., Ice core links CO to climate,
development
of a high resolution 0 to 300 000 y. Nature, •29, 389-390, 1987. •
chronostratigraphy, Quaternary Research, 27, 1, Thompson, L. and E. Mosley-Thompson, Microparticle
1-30, 1987. concentration variations linked with climatic
Neftel, A., H. 0eschger, J. Schwander, B. Stauffer change : evidence for polar ice cores, Science,
and R. Zumbrunn, Ice core sample measurements 212, 4496, 812-815, 1981.
give atmospheric CO content during the past Wang, W.C., D. J. Wuebbles, V.M. Washington, R. G.
40,000 years, Nature,2295, 220-223, 1982. Isaac and G. Mollnat, Trace gases and other po-
Palais, J.M. and M. Legrand, Soluble impurities tential perturbations to global climate, Reviews
in the Byrd station ice core, Antarctica ß of Geophysics, 24, 110-140, 1986.
their origin and sources, J. Geophys. Res.,90, Yiou, F., G. Raisbeck, D. Bourles, C. Lorius and
1143-1154,1985. Barkov N.I., •øBe at VostokAntarctica during
Petit, J.R., M. Briat and A. Royer, Ice age aero- the last climatic cycle, Nature, 316,616-617,1985.
Geophysical Monograph Series Understanding Climate Change Vol. 52

THE ROLE OF LAND ICE AND SNOW IN CLIMATE

Michael H. Kuhn

Institut ftir Meteorologieund Geophysik,Universit•itInnsbruck

A-6020 Innsbruck, Austria

Introduction An Inventory of Ice and Snow

Some of the most distinct climatic Table 1 contains a summary of global

contrasts like winter and summer or ice ice masses and snow covered areas. The
ages and interglacials are associated mass in long term storage is roughly
with changes in snow and ice at the 30-1018 kg and if it were melted and
earth's surface. The presence of ice on evenly distributed over today's oceans
land has effects of a magnitude similar would stand 83 m high. It would take
to or greater than those of vegetation or roughly 1025 J to melt it, which, under
soil moisture as far as the boundary con- the hypothetical assumption that this
ditions for atmospheric circulation are melting would be accomplished in 10 000
concerned. In the hydrological cycle snow years, would require a heat flux density
has a remarkable seasonal storing capaci- of 0.06 W m-2 when averaged globally, or
ty. Ice sheets can increase surface ele- about 2 W m-2 when averaged over the land
vation by kilometers, their development area permanently ice covered today.
leads to sea level changes and on a 100 Of the global average annual •recipi-
000 years time scale they despress the tation of nearly 1 m (1000 kg m-z) 6 per
earth's crust and trigger lithospheric cent (60 kg m-2) fall as snow. A corre-
feedbacks. spondingly much higher amount of water is
Apart from their direct and indirect taken into seasonal storage in those
climatic influence glaciers and ice areas actually affected by snow fall. A
sheets are of interest to climatologists spring snow pack of 1 m water equivalent,
both in view of predicting future and un- which is not uncommon in midlatitude
riddling past climatic situations: their mountains, requires 64 W m-2 in order to
slow motion at meters to kilometers per melt in a period of two months, while on
year, long turnover times and large ther- a global, annual average solid precipi-
mal inertia make long term forecasts in tation turns over 0.6 W m-2
the cryosphere more meaningful than in Both sea ice and seasonal snow cover
the atmosphere. On the other hand, mo- extent fluctuate from year to year. Figu-
raines and other geomorphic traces of re 1 shows a 1973-85 mean northern hemi-
past ice extent have significantly promo- sphere snow cover of 32-106 km2 with ma-
ted paleoclimatic reconstruction. Final- xima ranging from 42 to 52- 106 km2. Note
ly, the polar ice sheets contain ice that that November values are not well corre-
originated in the atmosphere hundreds of lated with the seasonal maxima.
thousands of years ago. With the appro-
priate technology ice cores are now being Boundary Conditions
used as archives containing information
on temperatures, precipitation, dust and As a boundary of atmospheric circula-
trace gas content of the atmosphere as tion and of energy and mass exchange bet-
far back as 150 000 years. ween air and soil, snow and ice covered
surfaces have the following characteri-
stics that distinguish them from bare
soil:
•pyvight 1989 by - High albedo
Internagional Union of •odesy and Geophysics - Potential evaporation
a• Amevio• Geophysioal Union. - Stable surface boundary layers
17
Geophysical Monograph Series Understanding Climate Change Vol. 52

18 LAND ICE AND CLIMATE

TABLE 1. Ice on Land

Area Volume Mass Mean thickness

106km2 106km
3 1018kg km
Northern hemisphere
Greenland 1.7 2.7 2.4 1.6
Arctic islands 0.35 0.2 0.2 0.6
Mountain glaciers 0.2 0.03 0.03 0.2

Southern hemisphere
Mountain glaciers 0.003 0.01 0.01 0.3
Antarctica including 13.6 30.1 27.1 2.2
ice shelves

Earth 15.9 33.0 29.7 2.1

Permanently snow or 16 11% of continents

ice covered land

Maximum simultaneous 53 35 % of continents

cover on land
( Northern hemisphere
winter)

- Strong thermal insulation, suppression In these examples, the turbulent flux

of soil heat flux. of sensible heat is directed towards the
All of these properties make snow and surface, supplying energy for snow melt
ice colder than bare soil under compara- (a and b) together with positive net
ble external conditions. Their combined radiation. In examples c and d, sensible
climatic effect is: heat replaces the energy loss of net
- local temperature inversions radiation. In all four examples energy
- baroclinic boundary layers and associa- fluxes into the snow and evaporation from
ted small scale circulation the snow surface play a subordinate role.
- large scale differential heating and The formulation chosen here applies to
associated synoptic effects. the snow surface only. Since the snow
pack is translucent to shortwave radia-
The Surface Enerqy Balance of Snow tion and permeable to both air and liquid
and Ice

We shall treat the surface energy 64. I06kmr

balance in the conventional way, i.e. ex- 56
press the distribution of energy gained 48
from net radiation R to sensible heat
4O
flux H, latent heat flux of evaporation
LE and heat fluxes associated with tempe- 32
rature changes S or melting M in the snow 24
and underlying ground.
16

R = H + LE + S + M (1) 8

0
where the individual terms are most con-
1973 75 77 79 81 83 1985
venientl¾ taken as energy flux densities
(W m-2). Figure 1. Time series of Northern Hemi-
Table 2 presents examplesfor melting sphere snowcover area, in 106 km2. The
snow and for Antarctic mid-winter condi- dots mark the November snow cover for
tions. each year. After Ropelewski (1986).
Geophysical Monograph Series Understanding Climate Change Vol. 52

KUHN 19

TABLE 2. Examples of Snow Surface Energy Balances. R Net

Radiation, H Sensible Heat, LE Latent Heat of Evaporation, and
M, of Melting, S Heat Flux into Snow, all Expressed as Flux
Densities in W m-2.

R = H + LE + S + M

a) Equilibrium line at Hinter- 66 = -32 + 3 + 0 + 95

eisferner, 2960 m,
15 July - 18 August 1971
b) Bad Lake, Saskatchewan, 14= -8+ 1 +2+19
27-30 March 1976
c) Plateau Station (80øS, 40øE, -15 = -13 + 0 -2
3625 m), June + July 1967
d) Maudheim (75•S, 11•E, sea level), -23 = -15 -4 -4
June + July 1951

Sources- b) adapted from Male and Gray, 1981

d) Liljequist, 1957

water, energy and mass is being exchanged IO0-

internally to the effect that the size of o/o
snow grains increases with age. _

Albedo
80-
27.3.1978
Increasing grain size, liquid water -

content and impurities lead to a lowering

of snow albedo. As the snow metamorphism
progresses faster at higher temperatures, •O-
an absorption-temperature feedback will
accelerate the ripening of the snow pack. -
Reflection from snow does not take place
in a specular fashion at the surface but 40
mainly by multiple scattering within the
snow. The emerging flux shows selective .
absorption by ice and liquid H20 bands
and displays anisotropic spatial INNSBRUCK
distribution. •O-
Albedo changes accompanying the decay
of an alpine snow pack were monitored -
with a spectral radiometer as shown in
Figure 2. The curves show the change from 0 I I I i I !

high visible to low near infrared albedo 1550nm

associated with absorption bands in the
vicinity of 1, 1.3 and 1.6 •m. As the Figure 2. Spectral albedo of a decaying
grain size increases from less than 1 mm alpine snow field near Innsbruck at
(27.3.) to nearly 5 mm (14.6.) diameter, 1900 m. Grain diameters increased from
near infrared albedo drops to values com- less than 1 to nearly 5 mm from March
parable to that of rock. to June.
In the dry snow facies of polar ice
caps the scattering geometry at low solar
incidence is far more important than me- rements of bidirectionl reflectance (Kuhn
tamorphism and grain size. Figure 3 shows 1985) of antarctic and alpine snow show
the increase of albedo with sinking sun reflectance enhancement by a factor of
at four antarctic stations, Plateau and two, approximately, in the solar meri-
Charcot being situated in the dry snow dian.
facies while the other two are coastal Low sun and high albedo make measure-
stations with seasonal melting episodes. ments and modelling of polar radiation
The curves in fig. 3 are to be un- fluxes highly susceptible to errors: at
derstood as hemispherical integrals over 10 degrees solar elevation and an albedo
an anisotropic reflectance field. Measu- of 0.8, the value of the direct component
Geophysical Monograph Series Understanding Climate Change Vol. 52

20 LAND ICE AND CLIMATE

ß1.0 are sufficiently known. For given densi-

ß Plateau Feb. 66 ty, heat capacity and conductivity, the
oCharcot thermal diffusivity and the downward pro-
ß Maudheim Dec./Ja gression of a harmonic temperature change
ßMirnyy , .9 at the surface can be computed. Table 3
summarizes relevant values for snow, ice,
sand and rock. The depth Da at which the
amplitude of the annual temperature wave
.8 is damped to 1/e of its surface value il-
o lustrates the insulating properties of
ß the material, low density snow being the
• best insulator of the examples given in
.7 < Table 3. Higher frequency fluctuations
penetrate less efficiently, the daily
wave• for instance, to a depth Dd =
0 cosz 365-•/2 Da = 0.05 Da. Dry winter snow,
Figure 3. Increase of hemispherical albedo which generally is of density less than
with increasingsolar zenith distanceZ, 300kgm-3 thusis a veryefficient insu-
for four antarctic stations (Kuhnet al., lator betweensoil and atmosphere. Com-
1977). pared to wet soil, rock or ice, low
density snow effectively reduces subsur-
face heat storage. This situation can be
of absorbed shortwave radiation is in er- grossly summarized by saying that the dry
ror by 10 per cent if either albedo is
wrong by ñ 0.02 or the level of the in-
strument is off by ñ 1 deg.
Nonetheless,the annual, energyweigh- .8 -
ted albedo of dry snow stations, which
ranges from 0.82 to 0.85 with a modal va-
lue of 0.84 seems to be a reliable upper
value (Kuhn et al. 1977, Koerner 1980). A
The associated planetary albedo was com-
puted as 0.74 over the Antarctic Interior
(Kuhn 1975). .6
For obvious reasons, there are no ot-
her fix points in the albedo scale. "Wet"
snow, whatever that is, will range widely
about 0.7, and alpine summer snow (with a
density of 500-600 kg m-3, grain diame-
ter 3-5 mm) has an albedo typically bet-
ween 0.5 and 0.6 depending on dust conta- .4
mination. Clean, snow free glacier ice
has albedo values near 0.4 while dust or
debris-covered ice may go down to 0.15. B
Vague as these figures may seem, they
are easier to handle in a global model
than the spatially and temporally highly
variable albedo of midlatitude model grid
points, representing an area, say, 100 by .2
100 km. The work of Robinson and Kukla
(1986) from which Figure 4 is taken, con-
centrated on late winter development of
albedo of Eastern US surfaces. Figure 4
shows farmland albedo changing from dry 14 18 22 25 1 3 24
snow values in February to a snow free
Feb. Mar.
0.2, while forest albedo declines from
0.2 to 0.1 and a variety of surfaces ran- Figure 4. Albedo of major surface elements
ge inbetween. in southeastern New York and northern New
Jersey under a variety of snow cover con-
Thermal Insulation by the Snow Cover ditions: (A) dark soiled farmland, (B) de-
ciduous forest, (C) mixed coniferous fo-
The equations of molecular heat con- rest and (F) shrubby grassland. ( From
duction and ensuing temperature changes Robinson and Kukla, 1984).
Geophysical Monograph Series Understanding Climate Change Vol. 52

KUHN 21

TABLE 3. Penetration of Annual Temperature Wave Damping of

Amplitude to 0.37 of its Surface Value at Depth Da

Dens ity Conductivity Di ffusivity D

-3 -1 -1 2 -1 a
kgm Wm K m s m

Snow
-6
50 0.007 0.07- 10 0.8
100 0.03 0.15 1.2
300 0.27 0.45 2.1

Ice

917 2.5 1.28 3.6

Sand

dry 0.29 0.22 1.5

40 % water 2.2 0.76 2.7

Rock

granite 3.7 1.9 4.4

snow energy balance is an unbuffered, simple, analytical terms, the synoptic

quasi-instant conversion of sensible heat and planetary wave scale effects of snow
gain into radiative heat loss (see Table cover have to be simulated in general
2, c and d) and vice versa. circulation models. In the following, two
To give an arbitrary example, I found examples of large scale simulations will
the February heat content of an alpine be presented.
snow pack at 3000 m elevation to be only In Figure 5 Dewey (1986) demonstrates
15 MJ m-2 less than that of a zero-degree that in winters with extensive snowcover
snow pack of same mass, a difference that the number of cyclones increases over the
can be restored by the energy gain of on- southern part of the US. Increasing cyc-
ly two summer days. lonic activity accompanies the southward
extension of the snow cover. In Eastern
Stability of Cryospheric Boundary Layers Canada in turn, cyclones become less fre-
quent when the margin of the snow cover
The thermal decoupling of cold surfa- moves south.
ces from the warmer atmosphere is enhan- A far reaching experiment was recently
ced by a temperature-stability feedback. carried out by Barnett et al. (1988) who
Consider the midwinter situation of an- investigated the effect of Eurasian snow
tarctic energy budgets where, approxima- cover on global climate, especially on
tely R = H (see Table 2), net radiation the Indian monsoon. When doubling the Eu-
loss being a function of surface tempera- rasian snow cover in their simulation
ture To, and H being determined by the they found that it retarded the warming
free atmosphere - surface temperature of the Asian land masses, thereby increa-
difference Ta-T o as well as by a stabili- sing surface pressure by up to 8 hPa,
ty dependent a transfer coefficient OH. cooling the surface to 200 hPa layer by
1.7 to 3.6 K. Since ocean temperatures
R = L+ - oTo4 = oH (TO - Ta) (2) changedlittle, this caused a weaker me-
ridional temperature gradient and a wea-
A disturbance in the radiative forcing ker subtropical easterly jet. The results
(where L+ is the longwave downward flux) of these simulations, especially the re-
will lead to a change in T O that includes duction of Indian precipitation by about
a gain of several per cent from the sta- 300 mm, agree well with what one observes
bility dependence of o H. in poor monsoon years. Reduction of Eura-
sian snow cover to half yielded similar
Synoptic and Large Scale Effects results with opposite sign.
The authors found teleconnections via
While local modification of the surfa- Rossby waves to the pressure field in the
ce energy budget can be expressed in West Pacific and in Western Tropical
Geophysical Monograph Series Understanding Climate Change Vol. 52

22 L•IND ICE •IND CLIM•ITE

lands bear ice covers underlines the im-

portance of land surface elevation for
ice sheet development.

Glacier Mass Balance

The mass balance of an existing gla-

cier or ice sheet is dominated by atmo-
spheric processes such as solid precipi-
tation P, redistribution by snow drift D,
deposition or erosion of snow by avalan-
ches A, Evaporation E, melt M and calving
C. P, D, C and A are strongly influenced
by topographic features, E and M only to
the extent that surface aspect governs
the energy budget. The mass budget of an
entire ice body is

B = P + D + A + E + M + C (3)
90 '70

• I•
where P, D, A
(accumulation)
(ablation).
and are
C, usually
E, M positive
negative

Specific mass balance terms are refer-

Figure 5. The difference in the total num- red to unit horizontal surface area and
ber of cyclones ( as calculated for 5 o customarily denotedby lower case let-
squares) between winters with extensive ters. The line connectingall points whe-
snow cover and winter with reduced snow re the annual value b = 0 is called the
cover. Positive values indicate that there equilibrium line. Further details are gi-
is greater cyclonic activity during winters ven in an earlier paper (Kuhn 1981).
with extensive snow cover. (From Dewey, A stationary or steady state glacier
1987). keeps its shape by annually transporting
net specific mass gain from the accumula-
tion area across the equilibrium line to
Africa. The similarity with winter sea- the ablation area. By dividing annual net
sons of ENSOevents lead them to remark mass gain in the accumulation area by the
that "snow fall perturbations may act as cross-sectional area underneath the equi-
trigger for some ENSOevents. They cannot librium line, a characteristic velocity
directly force an ENSOsince their cha- of the ice flow is derived which has ty-
racteristic time scale is about one sea- pical values of 10 to 102 m yr -1 in moun-
son ...". tain glaciers and 103 to 104 m yr -1 in
While the two reports quoted above polar outlet glaciers. Considering the
treat seasonal snow cover, the growth of sizes of glaciers and ice sheets one
ice sheets to heights above 3000 m adds a finds mean residence times of ice of less
third dimension to the forcing of atmo- than 1000 years in mountain glaciers and
spheric motion by snow and ice. By inser- of 10 000 years in polar ice caps, with
ting ice age (18 ky BP) topography into a extremes probably reaching 200 000 years.
simple two-layer model Lindemann and From the areas and volumes given in
Oerlemans (1987) found that orographic Table 1, the typical thickness of gla-
forcing was more important than thermal ciers has the order of 100 m, that of ice
forcing. Their results are reproduced in sheets is in the km-range. Comparing the
Fig. 6 showing how the Laurentide and life times and dimensions of these ice
Fennoscandian ice sheets produce a wave bodies to those of seasonal snow one
number three anomaly in the 500 hPa finds a remarkable gap in the size and
height. Note that the Greenland and Ural age spectrum. This impression is confir-
ice sheets do not have any significant med by glacier inventories and by
influence on this pattern. The negative individual observations.
anomalyover the eastern part of the lar- While the seasonal snowcover may be
ge ice sheets promotessnowaccumulation gradually dissolved in smaller patches
so that large ice sheets develope further and vanish in spring as is evident in Fi-
by a planetary wave feedback. gure 4, the majority of glaciers and ice
The fact that in the wave number three sheets have a sharp boundary. In the time
pattern of Fig. 6 a third, large ice domaine this is paralleled by the rapid
sheet is missing while Greenland, the transitions from glacials to interglaci-
Ural Mountains and central Asian high- als or the switching back and forth at
Geophysical Monograph Series Understanding Climate Change Vol. 52

KUHN 2•

Figure 6. Perturbation of 500 hPa height in gpdam (heavy lines) due to

topographic forcing by the 18 000 years BP ice sheets (stippled areas).
From Lindeman and Oerlemans (1987).

higher frequency between less long-lived than 4 K, about 0.4 K of this being due
stable states of the climate system. to surface albedo changes.
An obvious requirement for such sharp As was mentioned, planetary albedo is
or rapid transitions is the existence of higher than surface albedo over most of
feedbacks involving snow and ice, and it the globe but lower (about 0.75) than
is helped by the potential of ice to suf- surface albedo (0.84) over the dry snow
fer catastrophic, dynamic changes. facies of the Antarctic Plateau. This
masking of surface contrasts by atmo-
Feedbacks and Instabilities spheric backscattering and absorption
decreases the efficiency of the albedo
The Albedo-Temperature Feedback feedback. Shine et al. (1984), among
others, have pointed out that cloudiness
We have mentioned so far two is higher over snow free than over snow
feedbacks, one involving surface layer covered surfaces, a fact that further
instability and one involving orographic contributes to reduce the surface albedo
forcing of planetary waves. The one most feedback.
often quoted in connection with snow and
ice is the albedo-temperature feedback. Topographic Feedbacks Promotinq Ice
With rising global temperature, planetary Sheet Development
albedo is diminished by the melting of
ice thus adding to the original forcing. It would be more precise to speak of
This feedback was the central theme of an energy budget-albedo feedback rather
the classical papers of Sellers (1969) than of a temperature effect. However,
and Budyko (1969). both temperature and energy balance dis-
Hansen et al. (1984) have modelled the play a strong negative correlation with
effects of doubling CO2 or increasing the surface elevation, both dT/dz and db/dz
solar constant by 2 % and found that an are negative as is change of T and b with
original temperature change of about 1.2 latitude • . It was long known to alpine
K (simila• in both cases) was amplified glaciologists that a glacier with a flat
by various feedbacks to a total of more part at the altitude of its equilibrium
Geophysical Monograph Series Understanding Climate Change Vol. 52

2• LAND ICE AND CLIMATE

line is most sensitive to climate chan-

ges, for instance to a temperature
change, since 290

dB d(bS) dz _ dS db dz •
d• = dz dT- (b(z) • + S(z) •) d-• •-
A
(4) •
- 280
S being the glacier surface area up to •
altitude z. A flat part means a large va- ß
lue of dS/dz and the value of db/dz is E
characteristic
1984), while
of a
dT/dz
given climate
is roughly
(Kuhn •
constant.
This shows that dS/dz determines the sen- 270
sitivity of glacier mass balance to cli-
matic changes on a local scale. Consi- 1

dering the global scale one finds 1320 1340 13'60 13'80 14'00
d:B
= dB d__z
• (5) Solar
constant
(W/m
2)
dT dz de dT Figure 7. Mean annual Northern Hemisphere
surface air temperature plotted as a func-
For a given meridional temperature gra- tion of solar constant. (A) with realistic
dient dT/d½ the continental slope dz/d½ topography, (B) topography has been elimi-
is thus the parameter that determines nated. From Birchfield and Wertman, 1983.
mass balance sensitivity to a change in
global temperature. Birchfield and Wert-
man (1983) noted that "The Himalaya- librium line, since occasional, extreme
Alpine Belt, the Tibet Plateau, and the summers may waste the ice accumulated
Colorado Plateau appear in the zonally under nearly average conditions.
averaged elevation of the continents as a The Tibetan Plateau is in a similar
single plateau at an altitude of more situation with respect to present equi-
than 1600 m centered between 30 ø and librium lines and may well have experien-
40øN". Any climatic forcing that moves ced a rapid glaciation in the quaternary.
the snow line (equilibrium line) south- Kuhle (1986) believes to have found mor-
ward will thus be enhanced when a further phological evidence for a substantial
southward shift means an increase of sur- pleistocene ice cover in spite of the low
face above sea level. This effect is il- latitude position of the plateau, his as-
lustrated in Figure 7 which shows Nort- sociated climatic speculations, however,
hem Hemisphere temperatures following a have not been confirmed.
change in solar constant for sea level
and for a realistic topography. Reversible and Irreversible Chanqes
Koerner (1980) has given an interes-
ting contribution to the problem of the Annual snow cover extent, equilibrium
initiation of the Labrador-Ungava Ice line lowering and other processes invol-
Sheet. In defeat of the earlier hypothe- viQg snow masses of the order of 103 kg
sis that this ice sheet had been initi- m-2 have an almost instantaneous adjust-
ated by the buildup and subsequent spread ment and are reversible on a time scale
of mountain glaciers, Koerner points at of years. A mass balance disturbance is
the situation of the Baffin Island Plate- transmitted downglacier at a speed se-
au just below today's equilibrium line. veral times that of mean mass flow so
In the terms used above this plateau me- that most ice bodies are continuously in
ans high climatic sensitivity due to high a transient state, lagging behind smaller
values of dS/dz. climatic changes.
Barry et al. (1975) found Little Ice A developing ice sheet soon reaches a
Age conditions in Labrador approaching size large enough to modify atmospheric
ice sheet initiation and stated that temperatures or circulation. For example,
(today) "the climatic changes required to Braithwaite (1977) found surface air
initiate the necessary snow line lowering temperatures over Ellesmere Island gla-
may involve only a minor summer cooling". ciers 1 to 2 degrees lower than over ad-
Koerner (1980) stresses the point that a jacent bare ground. This cooling effect
reduction of annual mass balance variance is magnified when an ice sheet builds up
is just as important for ice sheet ini- to several hundred meters thickness at-
tiation as a lowering of the mean equi- raining a surface climate significantly
Geophysical Monograph Series Understanding Climate Change Vol. 52

KUHN 25

different from that governing its initial sea level changes or by thinning or
growth. When sea level temperature or sea thickening of the ice shelf. When ice at
level energy balance have returned to the old grounding line becomes afloat due
their original, no-snow values, the top to either sea level rise or ice sheet
of the new ice sheet is still in a healt- thinning the grounding line retreats. Ma-
hy glacial condition. ny marine ice sheets are terminating on
This, by definition, is an unstable sills, bedrock sloping downward inland
reaction to climatic forcing. It is not due to a combination of glacial erosion
entirely irreversible, however, and on a and isostatic depression (Fastook 1984).
large enough time scale the ice sheet Under such conditions the grounding line
surface temperature will follow a hy- retreats into deeper water, greatly am-
steresis loop around its original point plifying the response to the climatic
when forced by a negative and subsequent signal and possibly becoming unstable as
positive sea level temperature excursion. was recently the case with Columbia Gla-
cier in Alaska (Meier et al. 1985).
Surges, a Periodic Instability With long distance grounding line re-
treat the ice sheet surface becomes
Glacier surges are a dynamical insta- steeper and ice outflow is accelerated,
bility that is due to internal causes, the ice sheet becomes thinner and requi-
not triggered by climatic forcing. Only a res further grounding line retreat. Nume-
minor fraction of glaciers are capable of rous speculations have been made about a
surging and few have been observed, one possible catastrophic drainage or collap-
outstanding example being the 1982/83 se of the West Anatarctic Ice Sheet. Af-
surge of Variegated Glacier (Kamb et al. ter detailed consideration of all dyn-
1985). When basal water pressure comes amical processes involved Bentley (1984)
near ice overburden pressure, basal sli- comes to the conclusion that, if it
ding greatly accelerates producing ice drains, it will take at least 500 years
velocities of up to 50 m per day. Under for doing so.
usual conditions, higher water pressure In a classical paper Wilson (1964)
enlarges the subglacial tunnel system discusses global cooling due to the al-
which controls water outflow. Only at bedo increase by ice bergs spreading over
high enough speeds frictional melting the Southern Ocean following an Antarctic
enables a subglacial cavity network surge.
without major tunnels to be established
that in turn promotes further pressure to Snow and Ice in the Hydrological Cycle
build up. The conditions suitable for
this second mode of subglacial water flow Apart from the various ways in which
seem to be fulfilled cyclically with pe- ice and snow cover modifies the climate
riods of several decades. West Greenland via the surface energy budget, some cli-
outlet glaciers are believed to be perma- matically relevant processes involve mass
nently in the surging mode. transport or mass storage in the
Since surges are not climate-related cryosphere. These range from the daily
their moraines should not be used indis- freeze and thaw cycle to glacial sea le-
criminantly in paleoclimatic reconstruc- vel changes.
tions. The role of surges in the climate
system lies in their potential of drai- The Seasonal Cycle
ning large ice basins in extraordinarily
short time. Figure 1 shows that an appreciable
amount of water is seasonally stored on
The Groundinq Line Instability and the the land surface. Compared to no-snow
Possible Drainaqe of West Antarctica conditions on the same soil this means
delayed runoff and prolonged evaporation
A survey of climate-related dynamical at the potential rate not only from the
behavior of terrestrial and marine ice snow but also from the soil moisture that
sheets was given by Bentley (1984). He is charged by melt water. The hydrologic
shows that changes in the surface mass effect is not restricted to the local
balance (precipitation or melting) are scale: many semi-arid areas receive an
not nearly as effective in accelerating important fraction of their water balance
ice flow as are shifts in the grounding from ice and snow melt in distant moun-
line position. At the grounding line the tain. This is particularly important in
hydrostatic pressure of the ice load summer-dry regions with sufficient winter
equals that of the sea water, an equi- precipitation that would otherwise drain
librium that may be disturbed either by unused.
Geophysical Monograph Series Understanding Climate Change Vol. 52

26 LAND ICE AND CLIMATE

While the peak of annual runoff occurs . I

3800
during the spring in seasonally snow co-
vered lowland it is farther shifted to
July and August in glacierized mountain
regions. Figure 8 illustrated this effect 3000
with data from an alpine basin (•tztal,
47øN, 11øE).

hA J F M A M J J A S O N D
mm

4o0 83 %
1500
glacierized
200

2640m
L-'t___
200 44%
Om
1905m
• o - 1o o 1o 2oøc
0

Fig. 9. Monthly mean tempertures of Austrian

2o0 11%
• stations (1851-1950) as function of evaluation.
1160m • Isothermal winter conditions between 1000 and
! ß
•
or
1500 m are due to inversions in valleys and basins.

10 I '0
I 925m •c •• sensitivity of this effect to climatic

0 '
! i i •:.• change
•o
is highest
area with altitude
wherethe change
is most rapid
of
and whe-
Fig. 8. Monthly
runoffexpressed
in mm re temperature changesmost slowly with
(kgm-2 mo
-1)for fouralpinegauging
stations
that altitude.
are successively lower and drain less glacierized
catchment
areas in the basinof 0tztaler Ache. Sea Level Chanqes

Difficult as the determination of

A Possible Runoff Chanqe in a Warminq recent sea level changes may be on a glo-
Climate bal scale, there is evidence that since
1900 sea level has risen about 80 mm
To a useful approximation, the limit (Gornitz, in press). Of this rise more
of snowfall is determined by the 0øC-iso - than half is caused by thermal expansion
therm. Following a climatic rise of mean of the oceans, the rest being attributed
temperatures, then, considerable frac- to ice melt.
tions of a region may lose their seasonal From a survey of the restricted number
snow cover and consequently change their of mass balance measurements of suffi-
annual river runoff characteristics. cient length Meier (1984) found that
Figure 9 gives an idea of the possible about 30 mm could be attributed to this
change in the Alps. Mean monthly tempera- century's melting of extrapolar glaciers
tures in Austria presently favor a seaso- alone. The possible future contribution
nal snow cover over the entire country, of these glaciers to a four-degree global
frequent valley inversions cause nearly warming is expected to be limited to an-
isothermal conditions from 1000 to 1500 m other 120 mm (Kuhn, in press).
elevation. A five-degree warming is equi- The additional contributions to sea
valent to replacing the present -5øC iso- level rise from Greenland and Antarctica
therm (heavy line in Figure 9) by the new are still largely unknown. Whether South
0øC isotherm so that the lower limit of Greenland surface melt, wasting of an-
the winter snow cover would then be at tarctic ice shelves from below and sto-
1600 m. Similar as in equation (4) the rage of possibly increased snow fall in
Geophysical Monograph Series Understanding Climate Change Vol. 52

KUHN 27

polar regions are all insignificant com- feedback effects at the beginning of an
pared to glacier melt or whether they ice age, Quaternary Research, 13, 153-
simply have cancelled in this century is 159, 1980.
still a matter of stimulating discus- Kuhle, M., Die Vergletscherung Tibets und
sions. die Entstehung von Eiszeiten, Spektrum
der Wissenschaften, 1986 (9), 42-54,
References 1986.
Kuhn, M., Climate and glaciers. Procee-
Barnett, T. P., L. D'•menil, U. Schlese, dings of the Canberra Symposium on sea
and E. Roeckner, The effect of Eurasian level, ice and climatic change, Inter-
snow cover on global climate, Science, national Association of Hydrological
239, 504-507, 1988. Sciences Publication, 131, 3-20, 1981.
Barry, R. G., J. T. Andrews, and M. A. Kuhn, M., Mass budget imbalances as cri-
Mahaffy, Continental ice sheets: condi- terion for a climatic classification of
tions for growth, Science, 190, 979- glaciers, Geografiska Annaler, 66 A
981, 1975. (3), 229-238, 1984.
Bentley, C. R., Some aspects of the cryo- Kuhn, M., The bidirectional reflectance
sphere and its role in climatic change, of polar and alpine snow surfaces, An-
AGU Geophysical Monograph, 29, Maurice nals of Glacioloqy, •, 164-167, 1985.
Ewing Volume •, 207-220, 1984. Kuhn, M., A. J. Riordan, and I. A.
Birchfield, G., and J. Wertman, Topogra Wagner, The climate of Plateau Station,
phy, albedo-temperature feedback, and In The Climate of the Arctic, (G.
climate sensitivity, Science, 219, 284- Weller, and S. A. Bowling, eds.). AMS,
285, 1983. Fairbanks, 255-267, 1975.
Braithwaite, R. J., Air temperature and Kuhn, M., L. S. Kundla, and L. A.
glacier ablation, a parametric appro- Stroschein, The radiation budget at
ach, Ph.D. Thesis, McGill University, Plateau Station 1966/67, AGU, Antarctic
Montreal, 146 pp., 1977. Research Series, 25, 41-73, 1977.
Budyko, M. I., The effect of solar radia- Kuhn, M., Possible future contributions
tion variations on the climate of the to sea level change from small gla-
Earth, Tellus, XXI (5), 611-619, 1969. ciers. Paper given at the International
Dewey, K. F., The relationship between Workshop on the Effects of Climatic
snow cover and atmospheric thermal and Change on Sea Level, Severe Tropical
circulation anomalies, Snow Watch 1985, Storms and their Associated Impacts.
Glacioloqical Data, Report GD-18, World Norwich, UK, 1-4 Sep. 1987 (in press).
Data Center A for Glaciology, Boulder, Lindeman, M., and J. Oerlemans, Northern
37-53, 1986. Hemisphere ice sheets and planetary
Fastook, J. L., West Antarctica, the sea- waves- a strong feedback mechanism,
level controlled marine instability: Journal of Climatoloqy, •, 109-117,
past and future, AGU Geophysical Mono- 1987.
graph, 29, Maurice Ewinq Volume •, 275- Liljequist, G. H., Energy Exchange of an
287, 1984. Antarctic snow-field, Norweqian-
Gornitz, V., Mean sea level changes in British-Swedish Antarctic Expedition,
the recent past. Paper given at the In- 1949-52, Scientific Results, Vol. II,
ternational Workshop on the Effects of Part 1 D, Norsk Polarinstitutt, Oslo,
Climatic Change on Sea Level, Severe 1957.
Tropical Storms and their Associated Male, D. H., and D. M. Gray, Snowcover
Impacts. Norwich, UK, 1-4 Sep. 1987 (in ablation and runoff, In Handbook of
press). Snow (D. M. Gray, and D. H. Male,
Hansen, J., A. Lacis, D. Rind, G. Russel, eds.), Pergamon Press, 360-436, 1981.
P. Stone, I. Fung, R.Ruedy, and J. Meier, M. F., Contribution of small gla-
Lerner, Climate sensitivity: Analysis ciers to global sea level rise.
of feedback mechanisms, AGU Geophysical Science, 226, 1418-1421, 1984.
Monograph, 29, Maurice Ewinq Volume •, Meier, M. F., L. A. Rasmussen, and D. S.
130-163, 1984. Miller, Columbia Glacier in 1984:
Kamb, B., C. F. Raymond, W. D. Harrison, Disintegration underway. U.S. Geolo-
H. Engelhardt, K. A. Echelmeyer, N. gical Survey Open-File Report, 85-81,
Humphrey, M. M. Brugman, and T. 17pp, 1985.
Pfeffer, Glacier surge mechanism: 1982- Moser, H., H. Escher-Vetter,
H. Oerter,
1983 surge of Variegated Glacier, O. Reinwarth, D.
AbfluBZunke,
in und
Alaska, Science, 227, 469-479, 1985. von Gletschern, GSF-Bericht 41/86 (ISSN
Koerner, R. M., Instantaneous glacieriza- 0721-1694), Gesellschaft f•r Strahlen-
tion, the rate of albedo change, and und Umweltforschung M•nchen, 1986.
Geophysical Monograph Series Understanding Climate Change Vol. 52

28 LAND ICE AND CLIMATE

Robinson, D. A., and G. Kukla, Albedo of earth-atmosphere system, Journal of Ap-

a dissipating snow cover, Journal of plied Meteorology, •, 392-400, 1969.
Climate and Applied Meteoroloqy, 23, Shine, K. P., A. Henderson-Sellers, and
1626-1634, 1984. R.G. Barry, Albedo-climate feedback:
Ropelewski, C. F., Snow cover in real ti- the importance of cloud and cryospheric
me climate monitoring, Snow Watch 1985, variability, In: A. Berger, and C.
Glacioloqical Data, Report GD-18, World Nicolis (eds.), New Perspectives in
Data Center A for Glaciology, Boulder, Climate Modellinq, 135-155, 1984.
105-108, 1986. Wilson, A. T., Origin of ice ages: an ice
Sellers, W. D., A global climatic model shelf theory for Pleistocene glacia-
based on the energy balance of the tion, Nature, 201, 147-149, 1964.
Section III

VOLCANOES AND
Geophysical Monograph Series Understanding Climate Change Vol. 52

PETROLOGIC EVIDENCE OF VOLATILE EMISSIONS FROM MAJOR HISTORIC AND

PRE-HISTORIC VOLCANIC ERUPTIONS

JulieM. Palais*and Haraldur Sigurdsson

GraduateSchoolof Oceanography,Universityof RhodeIsland, Kingston,R.I. 02881

Abstract.Estimatesof volcanicvolatilecomposition
and
massreleaseof sulfur, chlorine and fluorine to the
atmosphere fromtwelvelargeRecentandQuaternary
volcaniceruptions hasbeenmadeon basisof pre-eruption
volatilecomposition, asdetermined by electronmicroprobe
in glassinclusionstrappedin phenocrysts in tephra.These
estimates extendourknowledgeof atmospheric loadingby
volcanicgasesto includeeventsmorethantwo ordersof
magnitudelargerthanrecenteruptions observedwith modem
methods. Resultsfor severaleventsagreewell withother
independent estimates,basedonicecoresandatmospheric
studies.The resultsshow,thatyield of sulfur,chlorineand
fluorineto theaunosphere is notonlydependent on total
eruptedmass,butlargelydetermined by thecomposition
theerupting magmas. Thusvolcanicvolatileyieldfrom
high-silicaor rhyoliticexplosiveeruptions is oneor two
ordersof magnitude lowerthanduringeruptionof equal
mass
of basaltic
ortrachytic
ma,g•as.Sulfuryieldduring
individualeventsis up to 3x10lø g, suchasin thecaseof the
basalticfissureeruptionsof Laki andEldgjain Iceland.
However,in certaintrachyticeruptions theyieldof halogens
mayexceedsulfuroutput,suchasduringthegreat1815
Tamboraeruptionin Indonesia,whenchlorineandfluorine
yieldtotheatmosphere
isestimated
as2x1014
and1.7x10
TM
g, respectively.
Petrologicestimates of sulfuryieldcorrelate
closelywithnorthernhemisphere annualtemperature
anomalies observed followingtheeruptions, in agreement
with thefindingsof Devineet al [1984]on a smallerdata
base.
Introduction
Volcanicactivityis oftencitedasoneof themajorcauses
of climatecooling,ontimescales rangingfroma few years
to millionsof years[Lamb, 1970;KennettandThunell,
1975,1977;Porter,1981;Rampinoet al., 1985].While
*Present
address:
Glacier
Research
Group,
University
of
New Hampshire,Durham,N.H. 03824
Copyright 1989 by
International Union of Geodesy and Geophysics
and American Geophysical Union.
earlywork in thisfield attributedtheclimaticeffectsof
volcaniceruptionsto dustveilsof volcanicash•..amb, 1970;
1971], more recent studieshave shownthat acidic volcanic
aerosols aremoreimportantin producinga climaticimpact,
bexause of theirlongeratmospheric residencetime [Pollacket
al., 1976;RampinoandSelf, 1982, 1984;Devineet al.,
1984;Rampinoet al., 1985].In fact,smallsulfur-rich
eruptions(bothexplosiveandnon-explosive) mayproduce
similaror evengreateratmosp•c-c•c effects,as
meas• by thestratospheric opticalthickness and
deviations in meanhemispheric temperature, aslarge,
explosivesilicicsulfur-poor ash-producing eruptions[Serfet
al., 1981;Sigurdsson,1982;RampinoandSerf,1982, 1984;
of Roseet al., 1983;Devineet al., 1984;Sigurdsson et al.,
1985;Rampinoet al., 1985].It is nowapparent thatthe
potentialclimaticimpactof a volcaniceruptionis notonly
governed by theintensityor magnitude of theeruption,but
probablymoreimporumtly by thechemicalcomposition of
themagma,i.e. theconcentration andtypeof volatile
comlxments, whicharedegassed duringeruptionandmay
generatea volcanicaerosol.
The determination of pre-emptionvolatileconcentrationof
themagmaandtheyieldof volcanicgases to theatmosphere
duringeruption is thereforeof majorinterest
in thestudyof
thepossiblerelationship betweenvolcanicaerosols and
climate.Thepotentialof trappedglassinclusions in
phenocrysts asrecorders of pre-emptionvolatilecontentof
magmaswasfirstrecognizedby Anderson[ 1974],who
appliedthisme• in estimating thevolcanicvolatile
contributionto thesulfurandchlorinebudgetof theoceans.
The me• wasalsoappliedto the 1976Mt. St. Augustine
eruptionby Johnston[ 1980], who demonstrated
the
potentially
greatcontribution
of volcaniceruptions
to the
chlorinebudgetof theEarth'sstratosphere.
Thesestudies
pavedtheway for thepetrologicestimates
of volcanic
degassing
duringearlierhistoricandpre-historic
eruptions.
Petrologicestimatesof volatileemissionsfrom a number
of recenthistoricandseveralQuatenmryeruptions were
madeby Sigurdsson [1982], Devineet al.[1984], and
Sigurdsson et al. [1985] (Table 1). Devineeta!. [1984]
showedthat,in general,basalticmagmasreleaseaboutan
orderof magnitude moresulfurthansiliciceruptions of
similarmagnitude.Theseauthorsalsofoundthatsurface
temperature de• afterfourhistoricvolcaniceruptions
(Laki 1783,Tambora1815,Krakatau1883,Agung1963)
werepositivelycorrelatedwith the estimatedmassof sulfur
Geophysical Monograph Series Understanding Climate Change Vol. 52

VOLCANIC VOLATILE EMISSIONS

TABLE 1. EstimatedMassof VolcanicVolatilesReleasedin Major HistoricandPre-HistoricEruptions

(Devine et al., 1984; Sigurdsson
et al., 1985)

Eruption Age TotalErupted

Mass H2SO4 HC1 HF TotalAcids
Kg metrictons metrictons metrictons metrictons
RoseauTuff 28,000BP. 7.5x 1013 n.d. 1.5x 107 n.a. 1.15x 107
Minoan 3,500BP. 6.3x 1013 3.86x 106 n.d. n.a. 3.86x 106
Hekla-3 2,800BP. 5.5.x1012 4.84x 105 2.55x 105 n.a. 7.39x 105
Hekla-
1 1104AD. 1.3x 1012 2.24x 105 n.d. n.a. 2.24x 105
Katla 1357AD. 4.2x 1011 7.22x 105 n.d. n.a. 7.22x 105
Laki 1783AD. 3.4x 1013 9.03x 107 1.60x 106 n.d. 9.19x 107
Tambora 1815AD. 2.4x 1014 5.24x 107 2.16x 108 1.26x 108 3.94x 108
Krakatau 1883AD. 2.7x 1013 2.94x 106 3.75x 106 n.d. 6.69x 106
Agung 1963AD. 2.4x 1012 2.84x 106 1.53x 106 8.0x 105 5.17x 106
Surtsey 1963AD. 2.8x 1012 3.52x 106 1.44x 105 n.a. 3.66x 106
Heimaey 1973AD. 3.6x 1011 4.68x 105 n.d. n.a. 4.68x 105
Soufriere 1979AD. 4.0x 1010 1.53x 104 3.50x 104 n.d. 5.03x 104
Mt.St.Helens 1980AD. 6.5x 1011 7.90x 104 3.52x 104 5.3x 104 1.67x 105
Krafla 1981AD. 5.6x 1010 1.01x 105 3.77x 103 n.a. 1.05x 105
n.a.= not analyzed;n.d.= not detected

releasedby theeruptions,continningthesuggestion that majordementanalyses

havebeendiscussed
in detailby
sulfateaerosols
fromvolcaniceruptions havea greater DevineandSigurdsson[1983] andDevineet al. [1984].
climaticimpactthansilicatedust Temperaturedecreases Sodiumandpotassium areanalyzedseparately usingthe
werefoundto be relatedto theestimates
of sulfurreleasedby decaycurvemethodof NielsenandSigurdsson [1981]. For
a powerfunction,wherethepowerto whichthe sulfur sulfurandchlorineanalyseswe usea traceelementprogram
massesare raisedis 0.345 (r=0.971) [Devine et al., 1984]. whichmeasures countson thebackground, at a fixedoffset,
In thisstudywe haveappliedthe sametechniques asthose for 100seconds on eachsideof thepeak(whichis
usedby Devine et al. [ 1984] to examinetwelveother determined duringstandardization),
andtwiceonthepeakfor
importantvolcanicevents.Theseincludethe35,000yr. B.P. leo seconds each. Operating
conditions for thetracedement
Campman eruption(PhlegreanFields,Italy), the 130A.D. analyses(S andC1)includea 1 to 5 gm beamdiameter,15
Taupoeruption(NorthIsland,New Zealand),the 1400yr. kv accelerating
voltage,1.65kv detectorvoltage,anda
B.P. Rabauleruption(Papua,New Guinea), the 934 A.D. 2.5x10
-8amp
beam
current.
Wealso
use
pulse
height
Eldgjaeruption(Iceland),the 1362Ora•fajokulleruption analysisto filter outbackgroundradiation.
(Iceland), the-1500 A.D. and-1800 A.D. Mount St. Fluorineanalyses wereperformedwitha similarprocedure
Helenseruptions(Wash.,U.S.A.), the 1835Coseguina asthatemployedby the traceelementroutineusedfor the S
eruption(Nicaragua),the 1886Taraweraeruption(North andC1analyses. Stepscanswereperformed on thestandard
Island,New Zealand),the 1902SantaMaria eruption (KE-12, seeTable2) aswell asfor eachsampleto determine
(Guatemala),the 1912Katmaieruption(Alaska,U.S.A.), theshapeof thefluorinepeakandthebestoffseton eachside
andthe 1956Bezymiannyeruption(Kamchatka,U.S.S.R.). of thepeakto measurebackground counts.We measured the
Eacheruptionis discussedin detailbelow,followinga brief countratetwiceon thelow sideof thepeak(300 stepsoff the
description
of theanalyticalandstatistical
methodswe have peak, leo s each),twice on thepeak(leo s each)andonce
used to obtain our results.
on thehighsideof thepeak(1eo s), at between50 to 90
stepsoff thepeak.Thisprocedureenabledusto betterdefine
AnalyticalMethodsandSignificance
Tests thebackground levelssothattheycouldbe subtractedfrom
theaveragemeasuredpeakrate.
Chemicalanalyses wereobtainedusinga JEOLJXA-50A Table2 givestheresultsof testanalyses thatwemadeon
electronmicroprobe.The procedures
thatwe followedfor soda-limeandnaturalglassstandards (NBS-610/620and
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON •

TABLE 2. AccuracyandPrecisionin MicroprobeAnalysis Finally,we assumethattime wereno sulfur- or

of Sulfur, Chlorine and Fluorine. chlodne-•g mineralsin themagmathatcouldhave
decomposed and releasedvolatilesin theprocess.As
discussedby Devineet al. [1984], whenall possibleerrors
Standard NBS-620
(1) NBS-610
(2) KE-12(3) dueto invalidassumptions areconsidered together,the
Sulfur ppm 1154 565 tendencyis to underestimatethevolatileyield by thismethod,
1 Std. Dev. 67 (5 %) 55 (10 %) andthusthepetrologleestimates areminimumestimates.
AcceptedValue 1155 500 A number of measurementsof S, CI and F concentration
No. of Analyses 15 231 (ppm)aremadeon eachsample(inclusionsandmatrix
Chlorineppm 3228 glass),andtheaverageandstandard
deviationarecalculated.
1 Std. Dev. 164(5 %) In order to take into account the standard deviation of the
AcceptedValue 3300 analyses,we haveassessed thestatistical
significanceof the
No. of Analyses 20 inclusionmatrixglassdifferenceby student's
T-test.
F ppm 4338
1 Std. Dev. 548 (13%) Results
AcceptedValue 4400
No. of Analyses 9 Dataon thepetrologic estimates of volatileemissions for
thetwelveeruptionsstudiedarepresented below.The results
(1) Averageof testanalysesof NBS-620, NationalBureauof of ourglassinclusionanalyses (majorandtraceelements)
Standards(NBS), StandardReferenceMaterial 620 andthecalculations of themassof volafilesdegassed for
soda-limeglass;acceptedvalueis averageof two bulk eacheruptionare summarized in Table3 (seeAppendix1 for
analysesdoneby theLeco(TM) inductiontitrationmethod; sampleinformation). Pertinentinformation oneacheruption
analyzedby W. Stockwell[LecoCorp,St. JosephMI, is presented,includingobservations on the effectsof the
quotedin Devine et al., 1984]. eruptionsfromoptical,climatological andotheratmospheric
(2) Averageof testanalysesof NBS-610, soda-limeglass; studies.ProxydatafromGreenlandandAntarcticice core
acceptedvalue from NBS (1970). studiesis also discussedin orderto assess thereliabilityof
(3) Averageof testanalysesof KE- 12,pantellerificobsidian thedifferentmethods(e.g.petrologic,atmospheric, ice core)
from Ebuma,Kenya; collectedby D.K. Bailey and R. for estimatingthe•nt of volafilesdegassed from volcanic
Macdonald;acceptedC1analysisby S.A.Malik andD.A. eruptions.
Bungard[quotedin Devineet al., 1984]. Standarddeviation
is one standarddeviation,numberin parentheses is BezymiannyEruption, 1956 A.D.
coefficient of variation.
The March 30, 1956eruptionof Bezymiannyvolcano
(Kamchatka Peninsula,U.S.S.R.) is bestknown for the
giganticcataclysmic
explosionthataccompanied theevent.
Beforetherenewedactivity,whichbeganwith a periodof
KE-12) to checkour traceelementstandardizations.We volcanicean•uakes on September 29, 1955,Bezymianny
estimatea detectionlimit of 50 ppmanda precisionof 5 to volcanowasgenerallyregardedasextinct[Gorshkov,1959].
10 % for S and CI. The detection limit for fluorine is Volcanicearthquakes continueduntil October21, 1955 when
estimated300 ppmanda precisionof 5 to 10 %. thefirstgasoutburstanda seriesof vulcanjan-typeash
Calculation of the massof sulfur, chlorine and fluorine eruptionstookplace.Explosionscontinueduntil March 30,
emittedto theatmosphere is madeby measuring thecontent 1956,whenthe "giganticparoxysmal explosion"occurred
of theseelementsin glassymeltinclusions whichhavebeen [Gorshkov, 1959].
trappedin magmaphenocrysts (p•-erupfionvolatile Theeruption
produced
a directed
blast,inclinedat 30-40ø
content),andthatin matrixglasses (degassed magma).The to thehorizon,whichremovedtheupper200 m of the
minimummassesof S, CI andF thathavebeendegassed in volcano,producing a craterabout2 km wideandan eruption
volcaniceruptions aredetermined by subtracting theaverage columnupto 45 km height.Followingtheexplosions two
matrixglassconcentration fromtheinclusionmeanandthe pyroclasticflowsandseverallaharswereproduced.
differenceis thenscaledto themassof magn• erupted. BetweenApril andlateautumnof 1956,two endogenous
A numberof assumptions madein thisapproach put domes were extrudedin the new crater of the volcano,
certainlimitations on theaccuracy of theestimates [Devineet accompanied
by weakto moderateexplosions[Gorshkov,
al., 1984]. First,we assumethatonly minorcrystallization 1959].
takesplacebetweenthetimethatthemeltinclusions are Locallytheashfallwasin a narrowzone(10(0150 km
trappedandthetime of eruption.Second,we assumethat wide)to thenorthandeast,upto 400 km fromBezymianny.
little or no degassingof themagninoccursbeforetrappingof In addition,a strongmistof sulfurousgaseswasreportedto
themelt inclusionsandthatall melt inclusionsaretrapped thewestandsouthof thevolcano,althoughno ashfallwas
moreor lesssimultaneously. A thirdassumption requiresus reportedin thatregion[Gorsb.kov, 1959]. However,the
to infer thatthecrystalcontentof theeruptedmaterialwas atmospheric effectsof theeruptionwerevery widespread and
relativelylow andthusthattheentirevolumeof erupted a stratosphericdustlayerwasreportedoverGreatBritainon
materialwaseffectivelyliquid,sothatwe canscalethemelt April 3-4, 1956 [Bull andJames,1956].
inclusion-matrixglassdifferenceto thetotalmassof erupted We haveanalyzedthemajorandtraceelement
material.Fourth,we assumethatall gasesreleasedin the composition of glassinclusionsin plagioclase
crystalsand
eruptionwerederivedfromtheeruptedmaterialandthat matrixglassin tephracollected80 km northof thevolcano
intrusivemagmacontributed littleto thevolatileemissions. (columns41 to 44; table3). Two compositionallydistinct
Geophysical Monograph Series Understanding Climate Change Vol. 52

54 VOLCANIC VOLATILE EMISSIONS

TABLE 3. Electron microprobe analyses of glass inclusions and matrix glasses from volcanic eruptions,
and estimates of sulfur, chlorine and fluorine yield to the atmosphere.

Eruption Eldgja 934 AD Oraefajokull 1362 AD Taupo-I-Iatepe131 AD Taupo-Rotongaio

Anal. No. 1 2 3 4 5 6 7 8
Glass Inclusions Matrix Inclusions Matrix Inclusions Matrix Inclusions Matrix

SiO2 47.12(144) 46.06(62) 71.74(94) 73.07(94) 71.81(57) 74.92(96) 72.17(143) 74.85(90)

TiO2 3.91(43) 4.77(26) 0.16(2) 0.27(2) 0.22(4) 0.23(8) 0.29(8) 0.30(3)
A1203 13.86(181) 13.22(19) 12.95(38) 12.9(29) 12.34(36) 13.01(23) 13.02(136) 12.79(16)
FeO* 14.57(183) 15.4(32) 3.40(16) 3.11(25) 2.04(27) 1.77(28) 2.22(61) 1.80(19)
MnO 0.26(6) 0.22(5) 0.10(7) 0.02(2) 0.02(4) 0.01(2) 0.07(6) 0.01(2)
MgO 5.14(54) 5.41(4) 0 0.01(1) 0.10(2) 0.12(1) 0.16(28) 0.56(119)
CaO 9.8(43) 10.17(15) 0.93(6) 1.01(7) 1.33(5) 1.49(7) 1.32(10) 1.16(52)
Na2Ot 2.42(51) 2.91(12) 5.74(59) 5.71(63) 4.75(0) 4.76(0) 4.02(0) 4.02(0)
K20 0.59(30) 0.72(3) 3.60(15) 3.61(16) 2.59(0) 2.69(0) 2.93(0) 2.93(0)
P205 0.5(27) 0.43(5) 0.13(8) 0 0.06(7) 0.12(3) 0.17(10) 0.11(6)
Total 98.17(66) 99.31(90) 98.75(101) 99.71(116) 95.24(89) 99.00(102) 96.40(156) 98.50(107)
No. Anal. 6 8 8 4 16 6 9 6
Mineral Pyroxene Pyroxene Pyroxene Pyroxene
S(ppm) 1266(319) 178(75) 77(21) 50(8) 42(22) 48(28) 44(22) 38(25)
No. Anal. 1 1 17 9 5 4 8 8 5
Cl(ppm) 482( 131) 331 (26) 2084(63) 1944(136) 1729(60) 1734(66) 1773(192) 1303(213)
No. Anal. 12 6 7 5 10 6 7 7
F(ppm) 764(583) 534(171) 1589(521) 1111(407) 327(101) 322(44) 406(181) 280(105)
No. Anal. 5 5 5 4 8 5 8 4

EruL• Volumo
(km3DRE): 9 2 1.4 0.7
Yield
(metrictons):
H2SO
4 8.4x 107 3.8x 105 0 3 x 104
HCI 3.9 x 106 6.6x 105 0 7.8x 105
HF 6.2 x 106 2.4x 106 1.7x 104 2.2x 105
Total Acids
(metric
tons): 9.4 x 107 3.4x 106 1.7x 104 1x 106
Yield (grams):
Sulfur 2.7 x 1013 1.2x 1011 0 9.7 x 109
1088+_98
ppmõ 27+Sppm 6+14ppm
(99.9%) (97.7%) (85.5%)
Chlorine 3.8 x 1012 6.4x 1011 0 7.8x1011
151+39ppmõ 140+_.65
ppm 470+_108ppm
(98.4%) (94.9%) (99.8%)
Fluorine 5.8 x 1012 2.2x1012 1.6x 1010 2x1011
230+_272ppmõ 478+309ppm 5+41ppm 126+_83ppm
(51.8%) (78.5%) (0%) (74.2%)

See
Appendix
forsample
identification.
Values
inparentheses
represent
thestandard
deviation
interms
ofleast
units
cited
forthe
valueto theirimmediateleft, thus47.12(144)indicates
a standard
deviationof 1.44wt. %. na. indicatesnotanalysed.* All iron
calculatedasFeO. tall analyses correctedfor sodiumlossby decaycurvemethodof NielsenandSigtmtsson [1981].õ Volatile
differencebetweenmeltinclusion meanandmauixglassmeanand•t confidencein parentheses.n.d.= notdetected.

groupsof inclusionsandcmxcspondingma• glasses were Rhyoliticglassinclusions

havemuchlowersulfurcontent
discovered.Daciticinclusions
andmatrixglasscontainfrom thanthedaciticinclusions,
or an averageof 49+ 23 ppmS.
64to66 % SiO2 andin therhyolitic
group
of glasscs
SiO2 TheseresultsshowthatBezymiannyvolcanoerupteda
ran•:•.from75.5to76.5%.In addition
tothemajor
element mixedmaglmin 1956,composed mainlyof dadticma•
variationbetweenthe•etwo groups,the sulfurcontentof the with minoramountsof rhyolite.Gorshkov[ 1959]reports
inclusionsis alsodramaticallydifferent.Inclusions
in the whole-rock
SiO2 of about58 to 60 % forejectafromthe
1ow-s'• grouphave514:!:128ppmsulfurandthe 1956eruption.Recently,however,the tephrabeingemitted
co-e••g d•cificmauixglasshas47+ 41 ppmsulfur. fromBezymianny
hasbecome
morea•id(61to62% SiO2;
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 55

TABLE 3 cont. Electronmicroprobe

analysesof glassinclusions
and matrixglassesfrom volcaniceruptions,
and estimatesof sulfur, chlorine and fluorine yield to •e atmosphere.

Eruption Taupo-Plinian Katmai 1912 Early Katmai-Mid Katmai-Late

Anal. No. 9 10 11 12 13 14 15 16
Glass Inclusions Matrix Inclusions Matrix Inclusions Matrix Inclusions Matrix

SiO2 71.15(81) 72.92(55) 67.16(212) 75.40(194) 68.06(143) 74.92(87) 70.83('111) 75.39(146)

TiO2 0.24(5) 0.2(4) 0.56(14) 0.39(12) 0.6(10) 0.28(8) 0.4(9) 0.33(4)
A1203 12.09(32) 12.75(28) 14.12(109) 13.14(53) 14.27(68) 12.47(44) 12.04(130) 12.06(45)
FeO* 2.23(19) 1.88(14) 3.37(89) 1.83(51) 3.02(34) 1.64(17) 2.24(45) 1.59(31)
MnO 0.14(5) 0.12(8) 0.11(6) 0.12(6) 0.11(6) 0.09(3) 0.12(7) 0.06(5)
MgO 0.08(1) 0.09(2) 0.82(40) 0.39(18) 0.71(18) 0.28(5) 0.34(8) 0.25(3)
CaO 1.36(9) 1.33(8) 3.14(77) 1.38(75) 2.81(47) 1.40(15) 1.83(29) 1.30(17)
Na2Ot 4.01(42) 4.01(42) 4.16(55) 4.29(57) 4.16(40) 4.2(49) 4.93(30) 4.37(61)
K20 2.51(12) 2.51(12) 2.44(12) 2.48(28) 2.68(21) 2.69(27) 2.85(9) 2.85(9)
P205 0.04(3) 0.22(5) 0.25(62) 0 0.24(10) 0 0.13(5) 0.14(11)
Total 93.85(56) 96.03(67) 96.13(228) 99.42(77) 96.66(92) 97.97(117) 95.71(79) 98.34(1!9)
No. Anal. 6 8 11 6 9 7 6 10
Mineral Pyroxene Pyroxene Plagioclase Pyroxene
S(ppm) 46(19) 38(4) 359(176) 53(19) 116(72) 92(54) 93(55) 51(18)
No. Anal. 6 3 6 7 9 4 8 9
Cl(ppm) 1750(102) 1656(59) 1739(142) 1847(73) 1712(166) 1671(32) 1959(469) 1808(89)
No. Anal. 9 7 10 10 6 5 9 12
F(ppm) 338(92) 259(39) 382(167) 399(70) 416(182) 4 28(90) 422(314) 413(86)
No. Anal. 7 3 5 8 4 5 8 8

Em_p• Volume
(km3DRE): 5.1 3 3 3
Yield (metrictons):
H2SO
4 2.9x 105 6.5 x 106 5.1x 105 8.9x 105
HCI 1.1x 106 0 2.9x 105 1.1x 106
HF 9.9x 105 0 0 6.6x 104
Total Acids
(metric
tons): 2.4x 106 6.5x 106 8 x 105 2x106
Yield(grams):
Sulfur 9.4 x 1010 2.1x 1012 1.7x 1011 2.9 x 1011
8+__8
ppm 306+_72
ppm 24_+36
ppm 42+_20ppm
(45.5%) (99.8%) (26.6%) (94.8%)
Chlorine 1.1x 1012 0 2.8 x 1011 1 x 1012
94+_41ppm 41+._69ppm 151+158ppm
(94.6%) (25.2%) (67.8%)
Fluorine 9.3 x 1011 0 0 6.2 x 1010
79_+41 ppm 9+_115ppm
(77.6%) (o%)

SEANBull.1986,11:4,p. 20) indicating

thepresence
of a Katmai Eruption, 1912 A.D.
moreevolved
batchof magma under thevolcano,
whichmay
havethesamesource
astherhyolitic componenterupted
in TheKaunaiexplosive
eruptionin AlaskaonJune6 to 8,
1956.Thelow-silicadaciteglassinclusions andmatrix 1912,wasthemostvolqminous
eruptionof thiscentury,
glassesaremorerepresentative of the•ifion of the prodt•cing
about15krna(D.R.E.)of ma• of whichabout
mainBezymianny tephra,andwe havebasedourestimates
of volatilereleasein theeruption
onsulfur,chlorineand 9 km:'wereemittedduringtheplinianphase[Curtis,1968;
Hildreth,1983].Theeruptionis bestknownfor the
fluorine
datafromthect•cite
glass
inclusions
(table3).We pyroclastic
anddebrisflowsin theValleyof TenThousand
estimatea total of 7x 10ø metrictonsof acidswere releasedin
Smokes. In thisstudywearemainlyconcerned withthe
theBczymianny
eruption,
ofwhich
52%wasH2SO
4 and tephra
deposits
formedin theplinianphases of activity.
48 % was HCI.
Historical
• of threedistinct
periods
of ashfallat
Geophysical Monograph Series Understanding Climate Change Vol. 52

36 VOLCANIC VOLATILE EMISSIONS

TABLE 3 cont. Electron microprobe analyses of glass inclusions and matrix glassesfrom volcanic
eruptions,and estimatesof sulfur, chlorine and fluorine yield to the atmosphere.

Eruption Kammi-Late Katmai-Total Rabaul Plinian RabaulIgnimbrite

Anal. No. 17 18 19 20 21 22 23 24
Gla3s Inclusions Matrix Inclusions Inclusions Matrix Inclusions Matrix

SiO2 72.92(154)75.39(146) 65.79(126) 65.48(164) 68.22(119) 66.93(34) 67.35(43)

TiO2 0.27(9) 0.33(4) 0.81(8) 0.89(6) 0.78(11) 0.77(3) 0.80(12)
A1203 11.72(23) 12.06(45) 14.77(47) 15.39(42) 15.42(50) 15.02(33) 14.95(43)
FeO* 1.56(42) 1.59(31) 3.73(49) 3.87(61) 3.51(62) 2.87(69) 3.69(43)
MnO 0.09(5) 0.06(5) 0.04(5) 0.07(6) 0.08(10) 0.11(9) 0.03(2)
MgO 0.20(8) 0.25(3) 1.01(16) 1.00(12) 1.09(13) 0.87(9) 1.05(7)
CaO 1.28(14) 1.30(17) 2.77(25) 3.04(56) 2.87(35) 2.53(18) 3.07(21)
Na2Ot 3.5(37) 4.37(61) 4.89(17) 4.62(3) 4.76(14) 4.6(21) 4.81(40)
K20 2.91(15) 2.85(9) 3.23(8) 2.96(14) 3.11(14) 3.32(11) 3.11(37)
P205 0 0.14(11) 0.19(7) 0.27(12) 0.28(14) 0.31(6) 0.17(6)
Total 94.45(159)98.34(119) 97.23(109) 97.59(93) 100.14(141• 97.33(98) 99.03(91)
No. Anal. 9 10 11 4 7 2 3

Mineral Plagioclase Plagioclase Pyroxene Plagioclase

S(ppm) 48(33) 51(18) 314(83) 615(222) 279(113) 218(41) 570(205)
No. Anal. 9 9 7 6 23 3 5

Cl(ppm) 2221(585) 1808(89) 2932(368) 2274(555) 2678(292) n.a. n.a.

No. Anal. 11 12 10 8 11
F(ppm) 218(138) 413(86) 243(129) n.a. 264(92) n.a. n.a.

No. Anal. 5 8 6 11

Em•ned
Volume
0unøD•): 3 9 0.6 0.6 3.5
Yield (metrictons):
H2SO4 0 7.9x 106 1.5x 105 1.5x 106 0
HCI 2.9x 106 3.2x 106 3.8x 105 0 n.a.
HI: 0 6.6x 104 0 n.a. n.a.
Tolal Acids
(metric
tons): 2.9x 106 1.1x 107 5.3x 105 1.5x 106 0
Yield (grams):
Sulfur 0 5 x 1010 4.8x 1011 0
35+39ppm 336+_94ppm
(0%) (99.9%)
Chlorine 2.9x 1012 3.7x 1011 0 n.a.
413+_178ppm 254+146ppm
(97.4%) (90.6%)
Fluorine 0 n.•. n.•. n.•.

Kodiak, Alaska,describedby Griggs[ 1922] andCurtis Hildreth[1986] haveprovidedestimates of theeruptionrate

[ 1968],havebeencorrelatedwith thepliniantephralayersA, andcolumnheightfor differentphasesof theeruption.
C-D and F-G [Hildreth, 1983; Fiersteinand Hildreth, 1986]. Columnheightswerecalculated usingbothvolumetric
Major elementanalyses of glassandphenocrysts from eruptionrateestimates andmaximumlithicvs.areaisopleth
pumicecollectednearNovarupta(thesourceventof all the mapsdetermined from studiesof theplinianfall deposits.
ejecta)showthattheinitialfall unit(A) wasrhyoliticand Thesecalculations suggest
maximumcolumnheightsfor the
contains1-2 % phenocrysts. The laterfall units(C-D and initial "A" rhyolificphaseof 28 to 30 km. For thedacitic
F-G), on the other hand,were found to be > 98 % dacitic eruptions,columnheightsof 18 to 25 km (phaseC-D) and
andcontain30-45 % crystals.The sequence of eventswhich 15to 20 km (phaseF-G) werecalculated. Isopachmapsof
producedthe Katmai(Novarupta)deposits, describedin theplinianfall tephrastudiedby FiefsteinandHildreth
detailby Hildreth[ 1983],suggestthattheeruptionwasthe [1986]suggest thatstrongwindsoccurred duringsome
resultof magma-mixingandfractionation in a shallow,zoned phasesof theeruptionandmayexplaintheobserved areavs.
magmachamber. thicknessandaxisof dispersal paRems. Tephracollectedin
Recentstudieson ejectadispersalanddynamicsof the abyssalpistoncoressoutheast
of Kodiakislandandfrom a
1912plinianeruptionsat Novaruptaby Fiersteinand siteon Kodiakisland[Federman,1984]confinnthe
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 37

TABLE 3 cont. Electron microprobe analysesof glass inclusions and matrix glassesfrom volcanic
eruptions, and estimates of sulfur, chlorine and fluorine yield to the atmosphere.

CampanianPlinian TarawetsPlinian
25 26 27 28 29 30 31 32
Inclusions Matrix Inclusions Matrix Inclusions Matrix Inclusions Matrix

SiO2 59.88(137) 60.43(93) 58.47(64) 52.95(59) 51.79(138) 53.74(70) 52.32(128)

TiO2 0.28(6) 0.36(7) 0.35(5) 0.91(7) 0.92(40) 1.08(33) 0.76(23)
A1203 19.4(63) 19.31(30) 18.63(49) 19.15(68) 20.65(2.52) 20.86(206) 18.96(413)
FeO* 2.88(60) 3.21(11) 2.79(35) 9.56(85) 7.38(116) 5.35(167) 7.53(242)
MnO 0.19(7) 0.12(8) 0 0.21(3) 0.18(4) 0.10(7) 0.19(8)
MgO 0.49(13) 0.49(10) 0.58(10) 3.14(36) 2.96(132) 2.04(123) 4.23(168)
CaO 1.91(28) 2.04(22) 2.17(19) 7.41(22) 11.98(1.89) 12.91(181) 11.52(101)
Na2Ot 4.04(51) 4.13(82) 3.39(24) 0.79(55) 3.05(73) 3.92(105) 3.29(77)
K20 7.68(93) 6.55(70) 8.69(87) 0.46(13) 0.61(25) 0.51(13) 0.59(20)
P205 ß07(4) 0.15(8) 0.18(7) 0.13(10) 0.18(11) 0.05(5) 0.14(11)
Total 96.82(148) 96.79(127) 95.25(300) 94.72(151)99.69(189) 100.58(49) 99.71(160)
No. Anal. 15 6 8 7 6 3 8
Mineral Pyx/Plag? Pyx/Plag? Pyroxene Pyroxene
S(ppm) 313(109) 380(176) 565(189) 263(217) 1888(594) 74(27) 801(201) 47(12)
No. Anal. 11 8 11 6 5 7 3 7
Cl(ppm) 3649(407) 5500(1736) 3588(101) 6539(900) 1286(301) 418(195) 1857(797) 377(228)
No. Anal. 9 6 10 9 2 7 3 5
F(ppm) 1408(230) 3553(389) 1634(102) 2735(405) 772(37) 728(138) 1848(700) 691(241)
No. Anal. 6 5 7 6 4 4 4 4

EruptedVolume
(km3DRE): 23 23 0.35 0.35
Yield (metrictons):
H2SO4 0 5.5x 107 5.4x 106 2.3x 106
HC1 0 0 8.7x 105 1.5x 106
HI: 0 0 4.6x 104 1.2x 106
Tolal Acids
(metrictons): 0 5.5x 107 6.4x 106 5 x 106
Yield (grams):
Sulfur 0 1.8x 1013 1.78x 1012 7.39x 1011
302+105ppm 1814+266ppm 754+116pp
(98.7%) (99.9%) (99.9%)
Chlorine 8.5x 1011 1.5x 1012
868+225ppm 1480•_71ppm
(99.6%) (98.2%)
Fluorine 4.3 x 1010 1.1x 1012
44+_71ppm 1157+370ppm
(42.7%) (95.9%)

conclusions
of Hildreth[ 1983] andsuggestthe zonationin theearly,middleandlatephasesof theeruption.Volatile
themagmachamber mayhavebeenthedominantfactor emission
fromtheearlyphase,oftheKatmaieruption
was
responsible
for thecompositional
variabilityin theKatmai dominatedby sulfur(6.45x10
v metrictonsH2SO4).The
deposits. eruptionappearsto havebecomeprogressively
enrichedin
We have determinedthe sulfur, chlorine and fluorine
contentof glassinclusionsin plagioclaseandpyroxene halogens
(mainly
C1),
e.specially
inlatestages,
whenlx106
phenocrysts andof matrixglasses in threesamples from (pyroxene)
to2.92x10
ø (plagioclase)
metrictonsof HC1
early,middleandlatestages of tephrafall deposition wereemittedto theatmosphere. Lyonset al. [in prep.]have
(columns11 to 19, table3). We havedividedthe volumeof identifiedtheKatmaieruptionin an ice corefrom Dye 3,
theentire
plinian
falldeposit
(9km3D.R.E.)
equally
among Greenlandon thebasisof peaksin excesschlorideand
thethreephasesof theeruptionin orderto estimatevolatile sulfate.They alsofind a patternof increasedsulfate
mass.Thereforethereportedtotalvolatilesreleasedfromthe followedby chlorideduringthecourseof theKatmai
Katmaieruption(column19) is the sumof thesamplesfrom eruption.
Geophysical Monograph Series Understanding Climate Change Vol. 52

50 VOLCANIC VOLATILE EMISSIONS

TABLE 3 cont. Electron microprobe analyses of glass inclusions and matrix glasses from volcanic
eruptions, and estimates of sulfur, chlorine and fluorine yield to the atmosphere.

Mt. St. Helens -1530 Mt.St. Helms -1800 Coseguina1835

33 34 35 36 37 38
Inclusions Matrix Inclusions Matrix Inclusions Matrix

SiO2 71.17(162) 73.19(76) 69.57(158) 71.85(86) 59.39(211) 61.46(504)

TiO2 0.03(6) 0 0.05(5) 0.07(9) 0.97(8) 0.64(49)
A1203 14.33(64) 14.45(32) 15.19(164) 14.75(54) 16.10(53) 17.63(656)
FeO* 1.35(31) 1.37(11) 1.96(54) 1.98(37) 7.99(110) 6.16(411)
MnO 0.02(3) 0.01(3) 0.03(3) 0.04(4) 0.12(9) 0.23(20)
MgO 0.24(7) 0.30(8) 0.56(17) 0.47(14) 2.55(73) 1.36(96)
CaO 1.53(11) 1.65(7) 1.99(34) 2.34(42) 4.99(97) 5.68(294)
Na2Ot 4.07(38) 4.07(38) 4.85(80) 4.85(80) 4.13(34) 4.57(0)
K20 2.23(24) 2.23(24) 2.05(16) 2.05(16) 2.09(19) 1.71(0)
P205 0.10(11) 0 0.12(13) 0.01(2) 0.27(10) 0.20(17)
Total 95.07(190) 97.27(104) 96.37(102) 98.41(115) 98.6(85) 99.64(77)
No. Anal. 9 6 7 7 21 4
Mineral Plag Plag Plag/Pyrox
S(ppm) 71 38(15) 48(24) 46(22) 355(63) 357(81)
No. Anal. 1 5 7 9 15 4
Cl(ppm) 767 732(107) 1665(540) 1217(51) 1870(215) 1408(288)
No. Anal. 1 10 9 8 15 10

F(ppm) n.a. n.a. 0 0

No. Anal.

Eru_lT,
ed Volume
(km3DRE): 1 0.25 10
Yield (metrictons):
H2SO4 2.3x 105 3.5x 103 0
HCI 8.3x 104 2.7x 105 1.2x 107
n.a. n.a. 0
Total Acids
(metrictons): 3.2x 105 2.7x 105 1.2x 107
Yield(grams):
Sulfur 7.6x 1010 1.2x 109 0
33+7ppm 2+12ppm
(99.9%) (0%)
Chlorine 8.1x 1010 2.6x 1011 1.2x 1013
35+34ppm 448+181
ppm 462+107
ppm
(0%) (96.2%) (99.9%)
Huorine n.a. n.a. 0

SeetheAppendix
forsample
identification.
Valuesinparentheses
represent
thestandard
deviation
intermsof leastunitscitedfor
thevalue
totheir
immediate
left,thus
47.12(144)
indicates
astandard
deviation
of1.44wt.%.n.a.indicates
notanalysed.
*All
ironcalculated
asFeO.tall analyses
corrected
forsodium
lossbydecaycurvemethod
of NielsenandSigurdsson
[1981].õ
Volatiledifference
betweenmeltinclusion
meanandmatrixglassmeanandpercentconfidence
in parentheses.
n.d.= notdetected.

Althoughfluorinewasreportedto havebeena major SantaMaria Eruption,1902AJ).

component of thefumarolicemanations in theValley of Ten
ThousandSmokes[Zies, 1929], our resultssuggestthat The plinianeruptionof SantaMaria volcano,Guatemala
muchof thefluorinein themagmawasretainedin thematrix in October1902wasoneof thelargestof thiscentury,
glass,andthusprobablydid notcontributegreatlyto the lasting18 to 20 hoursandproducing aneruptioncolumn
volcanicvolatileinputto theatmosphere duringtheeruption. estimatedto havereachedwell intothe stratosphere,
28 km
Lyonset al. [in prep.],however,find fluoridein theice at
the time of theKatmaieruption,oneof only two volcanic high[Williams
andSelf,19õ3].Theeruption
isestimated
to
events(theotherbeingHekla, 1947) to exhibit fluoridein
haveproducedabout20 km• of whitedaciticpumice(8.5
ice coreaciditylayersin theperiodfrom 1869 to 1984. km3 D.R.E.)
containing
20-30%crystals;
mainly
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 39

TABLE 3 cont. Electron microprobeanalysesof glass inclusionsand matrix glassesfrom

volcaniceruptions,and estimatesof sulfur, chlorineand fluorine yield to the atmosphere.

Eruption Santa
Maria1902 Bezymianny
1956 Bezymianny
1956
Anal. No. 39 40 41 42 43 44
Glass Inclusions Matrix Inclusions Matrix Inclusions Matrix

SiO2 69.72(228) 73.55(93) 63.99(44) 66.06(41) 76.39(1.07) 75.43(231)

TiO2 0.24(5) 0.27(2) 0.84(8) 0.71(7) 0.19(5) 0.11(6)
A1203 13.83(56) 13.86(29) 14.54(66) 15.23(20) 10.66(26) 12.34(151)
FeO* 1.61 (50) 1.35(11) 5.11(52) 4.76(26) 1.89(24) 0.19(10)
MnO 0.10(11) 0.07(5) 0.21(6) 0.24(6) 0.11(3) 0.08(4)
MgO 0.55(38) 0.35(7) 1.43(25) 1.20(11) 0.40(10) 0.06(14)
CaO 1.66(25) 1.65(10) 3.26(41) 3.48(15) 0.77(12) 0.71(60)
Na2Ot 5.16(84) 5.60(3) 3.78(56) 4.34(0) 3.78(54) 4.97(0)
K20 2.67(26) 2.71(1) 3.25(25) 3.42(0) 3.67(75) 4.86(2.26)
P205 0.24(19) 0.01(3) 0.20(8) 0.07(5) 0.01(1) 0.02(1)
Total 95.77(220) 99.42(112) 96.62((89) 99.51(45) 97.86(115) 98.78(63)
No. Anal. 22 19 9 5 11 5
Mineral Pyroxene PI agioclase Plagioclase
S(ppm) 198(89) 108(48) 514(128) 47(41) 49(23) 27(14)
No. Anal. 17 12 10 4 6 3
Cl(ppm) 1388(119) 1212(64) 3984(495) 2721(342) n.a. n.a.
No. Anal. 7 7 8 5
F(ppm) 0 0 0 0 0 0
No. Anal.

EruptedVolume
0cm3DRE): 8.5 1.0 1.0
Yield (metrictons):
H2SO4 5.6x 106 3.7x 106
1.8x 105
HCI 3.7x 106 3.4x 106 n.a.
I-IF 0 0 0
Total Acids
(metric tons): 1.2x 107 7.06x 106 1.8x 105
Yield (grams):
Sulfur 1.84x 1012 1.2x 1012 5.7x 1010
90+_26ppm 467::L-45
ppm 22:!:12plmi
(99.9%) (99.9%) (79%)
Chlorine 3.59x 1012 3.28x 1012 n•.
176+__51
ppm 1263:!:232
ppm
(99.2%) (99.9%)
Fluorine o o o

plagioclase,
hornblende, orthopyroxene,
clinopyroxene and H2SO
4(5.63x106)
and
39%(3.68x106)
ofwhich
was
titanomagnetite[WilliamsandSelf, 1983;Rose,in press]. HCI. LegrandandDe• [ 1987]reporta smallw.ak of
Tracesof darkgray,scoriaceous high-A1basaltoccurbothas sulfin-icacidin theice at Dome C, Antarcticawhichthey
discrete
blebswithinthedaciticpumiceandasisolated attribute
totheeruptio•
ofSanta
Mariaandcalculate
global
fragments scattered
throughout thedeposit.
Mediumgray acid fallout of 2.2x10 t metric tons. No evidence of the Santa
mixedpumices arealsoa minorcomponent of thedeposit Mariaeruptionhasyetbeenrepon• in iceco•s from
andareevidencethattheeruptionmayhavebeentriggeredby Gre•nlmxl,buta smallpeakin theCreteicecoreat thistime
theinjection
of a hightempexaturemaficmagmaintoan intervalmay•spond to theevent[Hammeret al. 1980].
alreadyvapor-saturateddaciticmagmabody[Sparkset al.,
1977;Williams, 1979; Rose,in press].
Ourresultsof themajorandtraceelementcompositions TaraweraEruption,1886 A D.
of glassinclusions
andmatrixglassfromthe1902Santa
Mariaeruptionaregivenin columns 39 and40, respectively, The largestandmostdeslxuctive volcaniceruptionin New
ofTable3.Weesl•ma•thattheSanta
Mariaeruption
emitled Zealandin historicaltimeswasthebasalticplinianfiss•
a total of 9.31x10 ø metric tons of acids, 61% of which was eruptionof Tarawerain 1886.Theactivitywasmainlyalong
Geophysical Monograph Series Understanding Climate Change Vol. 52

40 VOLCANIC VOLATILE EMISSIONS

a4Innlong
fissure,
which
erupted
atotal
of0.7km3of thelackof outcropsof the 1835deposit[Williams,1952].
high-al•nina
basaltic
magma,
witha mass
eruption
rateof Original
estimatesforthevoll•me
ofthe1835eruption
1.Sx10ø kg/s[W•er et al., 1984].The pliniancolumn products
rangefrom150lqm • (D.R.E.)[Penck,
quoted
in
heightis estimated to havebeen28 km to 34 km, basedon Williams, 1952] to 50 km• (D.R.E.) [Reclus, 1891].
calculations of Walker et al. [1984] andCareyandSparks Williams
[195•]concluded
thatthetotalvolume
didnot
[1986],respectively. The powerfulnatureof theTarawera exceed10 km• (D.R.E.), althoughthisis only an orderof
eruptionis thoughtto bedueto theinteraction of thebasaltic magnitude estimate.
magmawithgroundwater froma nearbygeothermal field Ourresultsof themajorandtraceelementcompositions of
[Walkeret al., 1984].Violentphreaficexplosions associated glassinclusions andmatrixglassfrom the 1835Coseguina
withtheTaraweraeruptionsuggest thata significantfraction eruptionaregivenin columns37 and38, respectively, of
of thetotalvolafilesmayhavebeenderivedfromground Table3. AlthoughtheCoseguina eruptionmayhavebeen
water.Calculations of thegasexit velocityby Walkeret al. responsiblefor a largeemissionof fine ashto the
[ 1984] leadto an estimateof 1.5 to 3.0 % for the total atmosphere, our resultssuggestthattheaerosolcloudfrom
volatilecontentof theeruptedgas-clast mixture,including thiseruption
mayhavecontained
smalli}mounts
of sulfurbut
anygroundwater whichmayhaveinteracted withthe wascomposed
mostlyof HCI (1.23x10• metrictons).There
eruptingmagma. is no evidence from Cneenland ice cores of sulfuric acid
Petrologicstudyof glassinclusions in pyroxenes from deposition
from theCoseguina eruption.Legrandand
twotephrasamples fromtheeruption(early,TW-1 andlate, Delmas[ 1987]reporta smallsulfi•c acidpeakin theice at
TW-2 sta__geejecta)suggest a totalvolatileyieldof ab9ut DomeC, Antarctica,
atthetimeof theCo•guinaeruption
1.15x107
metric
tons
•asacids)
(early
ejecta-
6.5x10
ømetric andestimatea globalacidyield of 2.3x10• metrictons,only
tons;lateejecta-5x10u metrictons).Whilebothphases of a factorof 2 greaterthanour petrologicestimate.It shouldbe
theeruption appearto haveemittedsignificant quanfites
of noted,however,thatthereis a discrepancy in the
sulfur,chlorineandfluorine,theearlyejectawereenrichedin composition of the acids; our petrologicestimatesuggesting
sulfurrelativeto late-stageproducts, whichwer• richerin a volcanicvolatileemissioncomposed dominantlyof HC1,
halogens.The inclusions in theearlyeruptedtephracrw-1) whereasice corestudiessuggestan aerosolcomposed
containthehighestconcentration of sulfur(1888+594;n=5) entirelyof H2SO4.
of anynaturalglassthatwe haveanalyzedsofar.
Furthermore, according to majorelementanalyses, the Mount St. Helens Eruptions 1500 A1). and 1800 A1).
magmain theearlystageof theeruptioncontained about4.9
% totalvolatiles,mostof whichwerelost on eruption.The Sulfur emission from the 1980 Mount St. Helens
magmaeruptedin late-stage activitywasvirtuallydevoidof eruptionwaslow [Sigurdsson, 1982],asis typicalof
volatiles.
explosiveeruptionsof mostdacitemagnms[Devineet al.,
Becauseof its locationin mid-latituderegionof the 1984]. In orderto characterize
timherthevolatiledegassing
southernhemisphere, it is unlikelythattheTarawent of thisvolcano,we havestudieddacitetephrafall fromtwo
eruptionwouldberecorded in Cneenland icecoresor thatit previouseruptionsof Mount St. Helens.They arethe "T"
wouldhavean effecton northernhemisphere climate. tephrafall from theGoatRockseruptiveperiod(ca. 1800
LegrandandDelmas[ 1987],however,reporta peakof A.D.) andthe "W" tephrafrom theKalamaeruptiveperiod
sulfuric acid in an ice core from Dome C, Antarcticawhich [ca. 1530 A.D. Mullineaux and Crandell, 1981; Hoblitt et
theyattributeto theTaraweraeruption.
No estimate
of global al., 198ff, Mullineaux,1986].Thevolumeof eruptedtephra
acid fallout is feasible on basis of the ice core as the Tarawere
peakoverlapswiththatof the 1883Krakataueruption,but hasnotbeenaccurately
•termined,butweestimate
volumes
of about 0.25 and 1 km•(DRE) for the "T" and"W" tephras,
LegraMandl)elmas[ 1987]estimate
a H2SO4 deposition respectively,
on thebasisof publishedfield observations
fluxof3.7kg/kra
2fortheTarawera
eruption. [Crandell et al., 1975; Crandell and Mullineaux, 1978;
Mullineaux, 1986].
CoseguinaEruption,1835 A 1). Ourresults
indicate
thethe1800A.D.emotion
ofthe"T"
tephrahada negligiblesulfuryield (1.11Jx10:'
g S; basedon
The 1835eruptionof Coseguina volcanohasbeen 2_+12 ppmS inclusion-matrixglassdifference)butemitted
regardedasoneof thelargestandmostviolent volcanic substantial
quantifies
ofchlorine
(2.58x1011
gC1).
Ourdata
eruptions
occun/ngin historictimesin theAmericas. indicatethatthe 1530A.D. "W" eruptionmay haveemitted
Coseguinais locatedon the southernendof theGulf of asmuch
as7.6x1010
gSandnegligible
C1(inclusion-matrix
Fonsecain the northwestcomerof Nicaragua.The eruption glassdifferenceonly 35 +34 ppmC1).As with the"T"
waslocallyof greatsignificance[Williams,1952],with f'me eruption,we did notmeasurefluorinein thesamples of the
ashfall nearthe volcanoand as far as Jamaica(1300 km to
"W" tephra, however,basedon the analysesof the 1980
the east),anddarknesslastedlocally for 2 to 3 days. tephra,the 1530A.D. eruptiveeventmay have producedas
Composition of theejectais andesitic,
withpredominantly
(~ 66%) paleto darkbrownglass,containingminutegrains much
as2x1011
gF.
of augite,ironoxide,andabundant plagioclasemicrolites Theseresultsarevery similarto thoseof Sigtl.
rdssonet al.
[Wffi•s, 1952].Discretecrystalsof plagioclase, augite, [1985] whofoundS, C! andF yieldsof 2.5x10 10g,
hypersthene andmagnetite arealsocommonin the 1835 3.4x1010
gand5x1010
g,respectively,
forthe1980
Mount
ejecta,alongwith lithiefragments fromtheoriginalsummit St. Helenseruptiveevent.Theseresultsconrumourdataon
cone. Taupo,reportedbelow,andthefindingsof others[Devineet
Oneof themajordifficultiesin tryingto reconstruct
an al., 1984;Sigurdssonet al., 1985]thatdaciticmagmasare
accuratescenarioof theeruptionandtotaleruptedvolumeis generallysulfur-poor
andoftenhalogen-rich.
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON

Or•efajokullEruption1362A•D. theDye 3 corefrom Greenland.Herron'sresultssuggesta

ten-fold increase in sulfate and a hundred-fold increase in
The Ora:fajokulleruptionin Icelandin 1362A.D. wasa fluorideconcentration overbackgroundvaluesin theice
Orrhyolitic
explosive
eruption,
which
ejected
about
2 associatedwith theeruption.Chlorideconcentrations were
D.R.E. of magmaandlaidwasteto severaldistrictsin only slightlyelevatedin theEldgjalayer,consistent with the
southernIceland [Thorarinsson,1958]. We have made petrologicresults.Hammer[ 1983]notedthattheincreasein
estimatesof thevolatiledegassing
duringthiseruption,on fluorideconcentration,associated with theEldgjaeruption
the basis of the data in columns 3 and 4 of Table 3. Our beganalmosta full yearpriorto thesulfateincrease.This is
results
indicate
a totalof 3.4x106
metric
tonsofacid,of consistentwith theidea thatfluorine(asH• may havebeen
adheringto fine graineddustassociated with theeruption
which11% wasH2SO4, 20 % wasHC1and69 % was [Oskarsson,1980]. Falloutof thisfine grainedtephra,
HF. Of all theeruptionsthatwe havestudiedsofar, the perhaps
dueto particleaggregation,
couldthenexplainthe
Ora•fajokull eruption
emittedthelargestproportionof deposition
of theHF priorto theH2SO4, whichwouldhave
halogensto theatmosphere, or slightlymorethanthe
Tamboraeruption,in which87 % of thetotalgasemissions a longeratmospheric residencetime.Few studieshave
werehalogens[55 % HCI and32 % HF; Sigurdsson et al., examinedwhetherthereis any fme grainedtephraassociated
1985;seetable1]. The Oraefajokullvolcanoemitted with volcanicaciditylayersin Greenlandice, however,
comendite magrm[Carmichael, 1967],butcomendite and becauseof thedifficultiesof workingwith smallamountsof
otherperalkaline magmas aregenerally enriched
in halogens fine-grainedtephra[Hammer,1984].
[BaileyandMacdonald,1970].However,because of the
Rabaul Eruption, 1400 yearsB J•.
large
difference
inthemassofthese
twoeruptions
(2.4x10
TM
kgvs.4.6x1012
kg),theTambora
eruption
emitted
nearly
2 Rabaulvolcano,on New Britianislandin PapuaNew
ordersof magnitude
moreHC1andHF thantheOr•efajokull Guinea,containsan 8x14 km caldera,whichhasbeenhighly
eruption. activein thepast.The volcanoshowedevidenceof renewed
Ice coreevidenceof aerosoldepositionfrom the
activityin 1983,whena dramaticshort-term
increasein
Ora•fajokull
eruptionis unclear.
An aciditylayerappearsin seismicityandgrounddeformation tookplacewithinthe
boththeCreteandDye 3 Greenlandicecoresat thislevel, caldera.The mostrecenteruptionoccurred
in 1937to 1943,
but it is unsure whether these acidities are due to a volcanic
eventor an artifactof surfacemeltingon theice sheet butthelastmajor
eruption
is14Cdated
1400
yr.B.P.This
[Hammer, 1984]. Until anothercoreis examinedin this eventproduceda wide-spread
lglinian
dacite
falldeposit,
with
interval,we haveno way of comparingthepetrologic a volumeof 1.66kmax(0.6
kmO,DRE). Followingtheplinian
estimatewith any proxydata. phase,a further8 km• (3.5 km• DRE) wereeruptedduring
theignimbritephase[Heming,1974;Walkeret al., 1981].
Eldgja Eruption, 934 A.D. We havestudiedthe volatilecontentof glassinclusions
andmatrixglassin samples of theplinianandignimbrite
The basalticfissureeruptionof Eldgjain Icelandin 934 phases of the 1400yr. B.P.Rabauleruption.Theresultsare
A.D. wasthe secondlargesthistoriclavaeruptionon Earth givenin columns 20 to 24 of Table3. Glassinclusions in
(sec•ond
onlytoLaki1783),withanestimated
volume
of9 pyroxene phenocrysts fromtheplinianphasearehigherin
kma (DRE) of magma[Larsen,1979].Theeruptioninvolved sulfurandslightlylowerin chlorinecontentthaninclusions
vigorousandhighfire-fountains
whichformedlargetephra in plagioclase.Thismaybetheresultof degassing afterthe
deposits
alongmarginsof theeruptivefissure.Ourresults, pyroxene crystallized,
justpriorto theformation of the
shownin columns1 and2 of Table 3, indicatea volatileyield plagioclase. We haveusedthepyroxene data(336+94ppm
of2.7x1013
S,3.8x1012
gC1and5.8x1012
gF, S) to obtaina maximumpetrologic estimatefor sulfur
corresponding
toamassof9.39x
107metric
tons
ofacids degassing fromthiseruption. Ourresultssuggest thatthe
(H2SO4+HCl+HF).
Bycomparison,
Hammer
etal.[1980] Rabaulrflinianeruptionhada maximumsulfuryieldof
estimate
1.65x108
metric
tons
global
acid
fallout
fromthis 4.8x10
TMg(-1.5x106
metric
tons
H2SO4).
The difference between the chlorine concentration in the
eruption,
onbasisof D.C. electrical
conductivity
measurements of Greenland ice cores. We consider these plagioclase
inclusions
(2932+368ppm)andthematrixglass
estimates
in verygoodagreement
whenassumptions
and (2678+292
ppm}is254+146
ppmC1.Thisgives
a C1yield
uncertainties in both measurementsare considered. The ofabout
3.7x10'
1gC1'•3.75x105
metric
tons
HC1).Our
factor
oftwodifference
between
these
estimates
may!•e fluorineanalyses onglassinclusionsin plagioclase
(243+
relatedto the chlorinecontent.Hammer [ 1980] estimatesthat 129ppm)andassociated matrixglass(264+ 92 ppm)
at least65 % of the acidfalloutwas HC1asopposedto only indicateno fluorinedegassingfrom thiseruption.No
4 % in our petrologic
estimate,
wheresulfurdominates (89 fluorineanalyses weremadeon inclusions in pyroxene.
%). This discrepancy
may bedueto reactionsbetween Therefore,if we combineresultsfromtheplagioclase and
volcanicandotheraerosolsduringatmospheric
transport,as pyroxene inclusions,weestimate thattheplinianphaseof the
H2SO4 reacts
withNaC1fromseasalttoproduce
HC1and Rabauleruption
produced• maximum volcanic
yolatile
yield
(asacids)
of about1.9x10
u metric
tons(1.5x10
ø metric
tons
Na2SO4 [LegrandandDelmas,
1984;PalaisandLegrand, H2SO4,
3.75x103
metric
tons
HC1).
1985, Chuanet al., 1986]causingan overestimate
of We havealsoanalyzedsamplesfromthe 1400yr B.P.
volcanic C1 in the ice cores.
Rabaulignimbritephase.Ouranalyses indicatenosignificant
Herron[1982]studied theconcentration
of anions(SO4=, degassing
of sulfurin theignimbrite
phaseof theeruption.
NO3-,C1-andF-)inicefromtheEldgjaeruption
interval
in However,a higherconcentration
of sulfurin thematrixglass
Geophysical Monograph Series Understanding Climate Change Vol. 52

42 VOLCANIC VOLATILE EMISSIONS

(570 + 205 ppm) thanin the inclusions(218 _.+

41 ppm) A total
of > 35km3D.R.E. ofrhyolite
magma erupted
suggeststhatsulfurwastakenup by thematfix glassby during
this'•ventincludes
alltephmfalldeposits
(8.2km3
metasomatism afteremplacementof theignimbrite,aswe D.R.E.),theignimbrite
units
(10.5km3D.R.E.),
havefoundin the Campanianignimbrite(seebelow). A co-ignimbrite
ash(--7km3D.R.E.)_andtheprimary
material
maximumestimateof volatiledegassing from theentire nowunder LakeTaupo [8to20km3 D.R.E.;Wilson and
Rabauleruptioncanbe madeif we assumethattheentire Walker, 1985]. Tephradispersal
from theeruptionindicates
volumeof the1400yr B.,P.Rabauleruption
(plinian
plus a columnheightfor theplinianphaseof no lessthan50 km
ignimbritephase-4.1 kin:' D.R.E.) degassedwith the same [Walker, 1980; CareyandSparks,1986].
yield astheplinianphase.This would give a maximum Wehave
studied
samples
fromtheHatepe
•linian
pumice,
volcanic
volatile
30eld
(according too.•rpetrologic
estimate) theRotongaio phreatoplinian ashandtheTaupoplinian
of about1.26x10
t metrictons(lx10t metrictonsH2SO4, pumicephases of theeruption.Theconcentrations of S, C1
2.56x106
metric
tons
HC1). andF in glassinclusions andmatrixglasses fromall three
Our analysesof the Rabaul 1400 yr B.P. eruptiveevent samples areverysimilar(columns 5 to 10of Table3). We
indicate
a minimum
totalvolatileyield(H2SO4 + HC1)of haveusedthevolumesestimatedby WilsonandWalker
[ 1985]to calculatethetotalvolatileemissionfor thethree
1.9x106
metric
tons
andamaximum
yield
(iftheentire phasesof theeruption.Examinationof Table 3 revealsvery
volumeof theeruptionis assumedto havedegassed
in the
similarbutlow (closeto ourdetectionlimit of 50 ppm)
same
manner
astheplinian
phase
alone)
ofabout
1.3x
107 concentrations of sulfurin inclusionsand matrixglasses
metrictons.It hasbeenproposedthatthiseruptionproduced from the threephasesof the eruption.The inclusion-matrix
a wide-spread"dryfog" andotheratmosphericphenomena glassdifferencefor sulfuris not significantandfor all intents
whichaffected
Europe
andtheMidq•eEastin 536A.D.,with andpurposeswe estimatethattherewasnegligiblesulfur
an estimatedaerosolmassof 3 x 10ø metric tons [Stothers,
emissionfrom theTaupoeruption.Sinceour analyses reveal
1984]. This estimatewas madefrom directancient very similarS, C1andF concentrations in the inclusionsand
observations of theopticaldepthperturbationof the matrixglasses of all threephases,we canextrapolate our
stratosphereandby scalingtherelativesignalstrengths for analysesto the totalvolumeof the eruption.Our results
theEldgjaandTamborneruptionsto the536 A.D. event demonstrate thattheTaupoeruptioninvolvedvery
usingice coredataof Herronet al. [ 1982] andHammeret al. sulfur-poormagmaandcausednegligibleemissionof sulfur
[ 1980]. As we discussbelow,ice coreestimates of global to theatmosphere, desp,
ite thelargevolumeof magmathat
acidfalloutfromtropicalandhighnorthernlatitudeeruptions waserupted
(> 35km3D.R.E.).
cannotbecompared, sincethelattermaybe significantly On the other hand, emissionsof chlorine and fluorine
overestimated in Greenland ice cores.
Our resultsthereforecall intoquestiontheestimateof
weresignificant.
Scalingtheaverageinclusion-matrix
glass
Stothers[ 1984]. Althoughthe increasesin anion differences
forchlorine
(-200ppm)andfluorine
(-10• ppm)
concentrationobserved by Herron[ 1982]in theDye 3 core to theestimatedtotalvolumefor thee.•pfion (-35 lcrn._
•)
(535 A.D.) andthestrongaciditysignalobserved by gives
atotal
chlorine
yield
of1.5x1013
g and8x1012
g
fluorine.The valuesquotedin table3 for theyieldof C1and
Hammer
etal. [1980]at,about540+ 10A.D.(globalacid F arescaledonlyto thevolumeof magmaeruptedin the
falloutestimateof 7x10t metrictons)maybetheresultof the
phasesof the eruptionthatwe studiedandare thusminimum
sameeventdiscussed by Stothers[ 1984], we believehis estimates while the above estimate is a maximum value.
aerosolloadingis overestimated by oneor two ordersof
magnitude.Futureice corestudiesmayrevealadditional Judgingfromthelow oxidetotalsof theglassinclusions
informationconcerning theproportions of acidsin the comparedto thematrixglass,we estimatethatthe
aerosollayerproducedby thiseruption.Our resultsindicate pre-emptiontotalvolatilecontentof theTaupomagmawas
thatapproximately 80 % of theRabaulvolcanicvolatile between3 to 6 wt. % andthetotalvolatilelossoneruption
release
wasH2SO4 andtheremainder
waslargelyHC1. wasabout2 to4 wt.%; H20 wasprobablya major
componentof the totalvolatiles.
Our resultsare similarto thoseof Dunbaret al. [ 1985]
Taupo Eruption, 131 A.D. who foundchlorinecontents of meltinclusionsin rhyolitic
obsidianclastsin Taupotephrarangingfrom0.18-0.24wt.
The Taupoeruptionin New Zealandwasoneof the % andsulfurcontents belowtheirdetectionlimit of 200 ppm.
largestknownexplosiveeruptionswithinthelast7000 years They estimatedanoriginalmagmaticwatercontentof 2.5 to
andfallsin a classof itsown asan "ultraplinian"
eruption, 4 wt. %. TheseresultsdemonstratethattheTaupoeruption
because of thehighmasseruptionrate,highdegreeof involvedrelativelysulfur-poor
magma,andcausedonly
fragmentation andconsequently thewidedispersal of the minor emissionof sulfurto theatmosphere. A significant
tephra
[Walker,
1980].
Thestatistical
meanof2214Cage degassingof bothchlorineandfluorineoccurred,however,
detemxinations,
yields1819+ 17 yrs.B.P. or 131A.D. for andif oneusesthetotalvolumeof theeruptionandthe
thedateoftheeruption
[Healy,1964]•
Twoplinianpumice averageC1andF yieldestimates, the totalvolatileemissions
falls(Hatepe
plinian
pu[nice
(1.4kmøD.R.E.)andTaupo fromtheeruption
isestimated
as2.35x107
metric
tons
acids.
plinianpumice(5.1 km• D.R.E.)) weregenerated,
aswell as It hasbeenproposedthattheTaupoeruptionwas
three
phrea•tomagmatic
ashfalls(initial
phreatomagmatic
•sh responsiblefor theatmospheric
effects(redcolorationof the
(0.005kma D.R.E.), I-Iatepephreatoplinian
ash(1._0kma sun and moon) observedin China and Rome in 186 A.D.
D.R.E.)andRotongaio
phreatoplinian
ash(0.7km3 [Wilson et al., 1980]. This date is, however, close to 3
D.R.E.)) andseveralignimbriteflow units(earlyintraplinian standard
deviations
outside
themean
ofthe14Cageforthe
ignimbrite
flow
units
(0.5km
3D.R.E.)
and main
Taupo eruption.Furthermore,our resultssuggestthatthevolcanic
ignimbfite(10 kin:' D.R.E.), [WilsonandWalker, 1985]). volatilemass(especially
H2SO4)wastoolowtohavesuch
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 45

farranging
effects,
andthatthe14Cageof131A.D.is
probablya morereliableagefor theeruption.Nevertheless,
we do haveevidencethat significantamountsof HCI and HF
wereinjectedinto thestratosphere
(eruptioncolumnheight >
50 km). Little is known aboutthe climatic effectsof the
halogenacids,whichare not knownto form atmospheric
aerosols,
but theymay havehada majoreffecton the
chemistryof the stratosphere
(e.g.ozone depletion)at that
time [Crutzen,1974; Mankin andCoffey, 1984].
Evidencefor distantdeposition
fromtheTaupoeruption
hasbeenfound in Greenlandice cores. Hammer [ 1984]
documenteda "dust"layerin theDye 3 corefrom South
Greenland, which is dated at about 174 to 175 A.D. The
layeris composed of silicateparticlesof "highSi content"
(precisecompositionunknown)andis not associated with
highacidity.In Antarctica,closerto thesourceof theTaupo
eruption,a layer at 104.2 to 104.5m depthin the Dome C ice
core(estimatedage .--2000 yr B.P.) wasfoundto have
elevatedlevelsof D.C. electricalconductivity, sulfate,and
microparticles[Maccagnanet al., 1981;Benoistet al., 1982].
However,major elementanalysesof glassshardsfoundin
thislayer(SiO2 = 66.54+1.51%)by deAngeliset a1.[1985]
do not supportthe conclusionthatthematerialis from the
Taupoeruption.Therefore,confirmationthattheTaupo
eruptionproducedeitheracidfalloutor ashdeposition in
Antarcticamustawaitfutureicecorestudies. Ourpetrologic
datasuggest,however,thattheeventmaybeonlybarely
detectable
in polarice coreson thebasisof acidityandD.C.
conductivitymeasuremenu
alonebecauseof thelow volcanic
volatileyield.
CampanianEruption, 35,000 yearsBJ•.
discussedabove, we have found a similar enrichment of
halogensin matrix glassrelativeto glassinclusionsin
samplesof the 1400 yr. B.P. Rabaulignimbrite.These
resultssuggestthattheenrichmentof halogensin thematrix
glassmay be relatedto eithera syn-eruptiveor
post-deposifional
processes in the ignimbrite.
Discussion
The resultsof petrologicestimatesof degassing
from the
twelveeruptionspresentedin thispaperandthe seventeen
eventsstudiedby Devineet al. [ 1984] canbe comparedwith
volcanic aerosol mass estimates based on ice core and aerosol
opticaldepthstudiesfor theseevents.Thesepetrologic
studiesnow providethedatabaserequiredto re-evaluatethe
potentialclimaticimpactof volcanicdegassing fromlarge
eruptionsand a furthertestof the claim of Devine et al.
[ 1984] andRampinoandSelf [ 1984]of a directrelationship
betweenmeannorthernhemispheresurfacetemperature
declineaftervolcaniceruptionsandthe massof sulfur
releasedby the eruption.
Aerosol Estimates
The massof totalacidsreleasedby theeruptionswhichwe
studied(thosefor whichaerosoldataalsoexist)areplotted
versusthetotalaerosolmassin figure 1, calculatedfrom
globalpeakopticaldepthestimates[Deirmendjian,1973;
Stothers,1984] anddirectmeasurements usingairbornelidar
andparticlecounters[McCormicket al., 1981;Kent, 1982;
McCormick,1982]. For mostof theeruptionsthe agreement

A largeexplosiveeruptionin thePhlegreanFieldsin Italy

about
35,000
yrs.B.P.produced
atotal
of80km3 of 1015
magmaandled to calderacollapse[Barberiet al., 1978;Rosi
et al., 1983].The Campanianeruptionis thelargesteruption 10TM-
documented in theMediterraneanregionduringthelate
Pleistocene. The productsof theeruptionincludethe
Campanian ignimbritedepositon landanda majortephrafall 1013-
A KK
layer whichextendsoverthe easternMediterranean region,
beyon•Crete.Thetephra
falldeposit
alonehasa volme of 1012
73kmø[about
23km3 D.R.E.,
Cornell
etal.,1983]. MSH
We havestudiedtwo samplesof tephrafrom a 50 cm
thickpumicefall from theLavorateQuarry, 50 km eastof 1011
1:1
thePhlegreanFieldscaldera,wherethetephrafall depositis
overlainby incipientlyweldedpyroclasticflow. Our results 1010
(columns25 to 28; table3) showthattheearliestphaseof the
eruption
mayhave
emitted
2x1013
gofS.Sulfur
degassing 109 i i i i i
in thesecondhalf of theeruptionwasapparentlynegligible;
however,thismay bedueto metasomatic alterationof the 109 1010 10TM 1012 1013 10TM 1015
uppertephrafall unit, asdiscussedbelow. Petrologic
Estimate
of TotalAcids(g)
We find higherconcentrationsof chlorineandfluorinein
thematrixglassesthanin theglassinclusions of tephrafall Fig. 1. Plotof totalacids(g) estimatedby petrologic
samples,indicatingno apparentdegassing of chlorineor rneth• versusatnx)spheric estimatesof aerosolmass(g)
fluorineduringtheeruption.We believe,however,thatC1 calculatedfromopticaldepth•on anddirect
andF contentin thematrixglassof thesesamplesmay have measurement for eruptionsreportedin thispaper.Key: T -
beenaffectedby post-depositionalhydrothermal alterationor Tambora, SM- SantaMaria, K- Katmai, KK- Krakatau, R-
metasomafism associatedwithcoolingof theoverlying Rabaul,A- Agung,F- Fuego,MSH- Mount St. Helens,S-
ignimbfitedeposit.Otherstudiesof the S,C1andF contentof Soufriere.Linearregression equation(y = 184.5065*
glassinclusionsandmatrixglassin samplesof the x•0.8372; R= 0.96) doesnot includeR and S (see
Campanianignimbrite [Devineet al., 1984;M. Rosi,pers. discussionin text). Line c,om•spondingto 1:1relationship
comm. 1986] reveal similar results.Furthermore,as alsoshownfor com•.
Geophysical Monograph Series Understanding Climate Change Vol. 52

44 VOLCANIC VOLATILE EMISSIONS

betweenourpetrologicestimates andtheaerosol [1980],asno hydrochloric acidor hydrofluoricacid(HX)

measurements is good, exceptSoufflereandRabaul. wereassociated with anyof theeruptionsthattheystudied
In caseof the Rabauleruption,the aerosolloading [Legrand,1985].
(3x1014 g)estimated byStothers [1984a] from observations The figureshowsthattheice coreestimateis greaterthan
in the literature at the time, is for an event which occurred thepetrologic estimatefor manylow-sulfuryielderuptions
around536 A.D. andmay not correspond to the Rabaul (Krakatau,Agung,Katmai,Coseguina andSantaMaria). In
eruption.Becauseof thepoorcorrelationbetweenour thecaseof eruptionswith high sulfuryield (e.g.Laki,
petrologic estimate oftotal acids (1.48x1012 g)andthetotal EldgjaandTambora),thecorrespondance is good.The
aerosolestimateof Stothers[ 1984a],we suspectthatthe petrologicmethodtendsto underestimate theamountof
"mysterycloud"eventof A.D. 536 describedby Stothers releasedvolatilespeciesin general,andthereforeourresults
[ 1984a]is probablynot due to the 1400 yr. B.P. Rabaul indicate minimum values [Devine et al., 1984].
eruption,butderivedfrom anotherevent. The poorcorrespondance betweenthepetrologicestimate
Sigurdsson et al. [ 1985] havediscussed thediscrepancy for theRabauleruptionandtheicecoreestimateof global
betweenthepetrologicestimateandthevolcanicaerosol acidfalloutattributedby Stothersto thiseruption,indicates
estimate [McCormick et al., 1981] for the Soufflere 1979 that these estimates are for two different events. We therefore
eruption.They concludedthatsincethesatellite-based attributetheice coreaciditypeakin 540 + 10 A.D. [Hammer
aerosolextinctionmeasurements probablyincludedonly et al. 1980] andthe 536 A.D. atmosphericevent[Stothers,
abouthalf of theeruptedmass(onlytwo of theeighteruption 1984]to anothereruptionof unknownorigin,with volatile
plumes),thetwo estimatesareonly off by anorderof outputovertwo ordersof magnitudelargerthantheRabaul
event.
magnitude.In addition,a comparison of sulfurmassflux
emission[measuredby COSPEC;Hoff andGallant, 1980] Our estimates
of sulfuricacidyield areplottedin figures
andpetrologicestimates of sulfuryieldpermassof erupted 3a and3b versustheestimateof sulfuricaciddeposition flux
magmagivesvery similarresults,supporting thevalidityof (kg/km2;
thetime-integrated
deposition
ofacid
which
takes
thepetrologicmethodfor the studyof volatilerelease into account the mean accumulation rate at the ice core site
[Sigurdsson et al., 1985]. andtheacidconcentration in thepeak)in Greenland
[Hammeret al., 1980] and Antarctica[Legrand,1985;
Ice Core Estimates
LegrandandDelmas,1987],respectively.
Therelationship
betweenourestimatesof sulfuricacidyield (g) andth_e
ice
coreestimates
ofsulfuric
aciddeposition
flux(kg/km
2)on
Petrologicestimates
of volcanicdegassing
arecompared theice sheets(GreenlandandAntarctica)(Figures3a and3b)
in figure2 withicecoreestimates
of totalacids(H2SO4 + is verygood,withfew exceptions.
Theorderof magnitude
HX) from severaleruptions[Hammeret al., 1980; Legrand difference
inthesulfuric
aciddeposition
flux(kg/km
2)in
andDelmas,1987].The estimatesof LegrandandDelmas Greenland versus Antarctica is due to a combination of
[1987]represent
onlyH2SO4;however,theyare factors,includingtheorderof magnitude
differencein
comparablewith the totalacidestimates
of Hammeret al. accumulationrates between the two sites, and differences in
theamountof drydeposition and/ordirectstratospheric
injectionof volcanicdebris[LegrandandDelmas,1987].
In Figure3a, theGreenlandice coreestimatesfor theLaki
1015
(64.08ø N) andEldgja(63.75ø N) eruptions
appear
to be
overestimated,
comparedthepetrologicestimates,
considering
thegoodagreementfor theothereruptions
(Tambora,Agung,KrakatauandKatmai).The apparent
1014 overestimate
from ice coresis probablydueto thehigh
northern latitude location of these volcanoes, their relative
proximityto theGreenlandice sheet(- 1500km) andthe
natureof theatmospheric transportinvolved(tropospheric
vs. stratosphefic).The Antarcticice coredeposition flux for
1013= theAgungeruptionmay be anoverestimate (figure3b),
whereasthedepositionflux for the SantaMaria eruptionmay
be underestimated, comparedto theexcellentcorrelation of
petrologicandice coreestimates for theotherthreeeruptions

1012 ,
1:1 (Tarawera,KrakatauandTambora).The discrepancy
Agungeruptionis readilyaccountedfor by theobserved
two-thirds/one-thirddistribution
for the

1012 1013 1014
PetrologicEstimateof Total Acids(g)
Fig. 2. Plotof totalacids(g) estimatedby petrologic
methodversusglobalacidfallout(g) estimatedfrom
Greenland[Hammeret al., 1980] andAntarctic[Legrandand
Delmas, 1987] ice core studies.Key - sameasin Fig. 1
includingE- Eldgja,C- Coseguina.Linearregression
equation(y = 1.366e+8*x•).4111, R--0.92) calculatedfor
all datapoints.Line corresponding to 1:1relationship
also
shownfor comparison.
of the 1963Agungaerosol
101s
betweenthe Southernand NorthernHemispheres,
respectively[Castlemanet al., 1974;Delmaset al., 1985;
LegrandandDelmas,1987].The slightunderestimate of the
depositionflux in AntarcitcafromtheSantaMaria eruptionis
likelydueto thenorthern latitudelocation(14ø N) of the
volcano.
Halogens
The resultsshowvery significantreleasesof HC1andHF
gasesto theatmospherein manyeruptions.Of thetwenty-six
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 45

2OO
andchemicalmodelssuggestthatHC1andHF will not form
liquidaerosols
undernormalstratospheric
conditions
[Miller,
A 1983; Solomonand Garcia, 1984].
Studiesof theatmospherefollowingrecenteruptions
shows
highlysignificant
increase
of chlorine.Thusspectroscopic
observationsof the total columnamountof HC1, made
shortlyafterthe 1982E1Chichoneruption,showan increase
100
of the stratospheric
HC1burdenof 40 per centovera large
partof the globe,with an estimatedtotal increaseof about
4x1010
gHC1,
orabout
9percent
ofthetotal
global
stratosphericHC1burden[Mankin andCoffey, 1984]. The
hydrogenchloridemay in partbe derivedfrom releaseof
gaseous HC1duringbreakdownof haliteparticlesobserved
in the aerosolcloud from the volcano [Wood et al, 1985], as
o well asfromdirectdegassing of themagma.The atmospheric
10•2 10'•3 10'14 1015estimateof Cadle et al. [ 1977] of the massof HC1 released
bytheAgtmg 1963 eruption (1.2x1012 g)isvirtually
Petrologic
Estimateof TotalAcids(g) identical
toourpetrologic
estimate
(1.42x
1012g),whereas
the estimateof Stolarskiand Buffer [1978] for the Krakatau
20
1883eruptionis oneorderof magnitudelower thanthe
petrologicvalue.StolarskiandCicerone[1974] were the f'u'st
to considerthatchlorinefromvolcanoesmightperturbthe
stratosphericozonelayer.Basedon very conservative
estimatesof chlorinedegassing,the stratospheric
photochemistry modelof Stolarskiand Buffer[ 1978] for the
Krakatau1883eruptionestimatesabout7 per centozone
columndepletion.Outputof chlorinefrom the 1815 Tambora
eruptionis aboutthreeordersof magnitudehigheron basis
Agungß of petrologicdata,andmodelingof theimpacton theozone
layerof suchlargevolcanicchlorineemissions is urgently
needed.
Santa Maria
Tarawera

TemperatureData
0 !

10•2 10•3 10TM Devineet al. [ 1984]evaluated thepossibleeffectof

Petrologic
Estimateof SulfuricAcid (g) volcaniceruptions onclimateandproposed a relationship
between themassyieldof sulfurto theatmosphere froman
Fig. 3. (A): Plotof totalacids(g) estimated by petrologic eruption(asestimated by thepetrologic
method)andthe
methodversusdeposition flux of totalacidsestimatedby observed decrease in meannorthern hemisphere surface
Hammer et al. [ 1980] from the Crete ice core, Greenland. landmass temperature in theoneto threeyearsfollowingthe
Key sameasin Figs.1 and2. Linearregression
equation(y eruption,onbasisof published temperaturedata[Rampino
= 0.0015' x^0.3183, R = 0.80) calculatedwithout the Laki andSelf, 1982, 1984].Devineet al.[1984]foundthatthe
(L) andEldgja(E) points(seetext). meansurfacetemperature decreasewasrelatedto theestimate
(B): Plotof sulfuricacid(g) estimatedbypetrologicmethod of sulfuryield by a powerfunction(r= 0.971), with the
vs.depositionflux of sulfuricaicdby LegrandandDelmas powerto whichthesulfurmassis raisedbeingequalto
0.345.
[1987]fromtheDomeC icecore,Antarctica.Key sameas
in Figs.1 and2. Linearregression equation(y - 4.0828+ We havecombined newdataonsulfuryieldpresented in
2.429e^-13x,R = 0.99) calculated withoutSantaMaria (SM) thispaperanddatafromtheliterature for theFuegoeruption
andAgung(A) points(seetext). withdatabaseof Devineet al. [1984]in figure4a. The
resulting
equation
describingtherelationship between
estimated sulfur mass and observed surface landmass
temperature decrease is similarto thepreviousone[Devineet
eventsstudiedsofar by thepetrologic
method,thirteen al. 1984], with thepowerto whichthe sulfurmassis raised
emittedsulfur-dominated volcanicgases,sevenevents beingequalto 0.3076 andr=-0.92.The totalyield of volcanic
emittedhalogen-dominated gasclouds,whilesulfurand volatilecomponents asdetermined by thepetrologicmethod
halogens wereemittedin roughlyequalproportions in the (sulfur,chlorineandfluorine)for theseeruptions, calculated
remainingsixevents.Thepetrologic studiesindicatethatthe asacids,is plottedversustheobserved northernhemisphere
halogen outputmaybetrulyexceptionally largein some temperature decrease in figure4b. The relationshipis alsoa
events,suchasduringthegreat1815Tamboraeruptionin powerfunction,with thepowerto whichthe totalmassof
Indonesia,whenchlorineandfluorineyieldto the acidsaxeraisedbeingequalto 0.274 andr= 0.87.
atmosphere
isestimated
as2x10
TMand1.7x10
TMg, Althoughtheseresultsappearto confu'mtherelationship
respectively.However,it shouldnot be assumedthat these proposed by Devineet al. [ 1984],it shouldbe emphasised
halogen
gasesformaerosols
in thestratosphere,
asphysical thatthesampleis limitedandthetemperature deviations listed
Geophysical Monograph Series Understanding Climate Change Vol. 52

46 VOLCANIC VOLATILE EMISSIONS

101 andAgung)wererecentlyexaminedby Angelland

Korshover[ 1985].Theycalculatedtemperature deviations
A usingdatasetsof GrovemanandLandsberg[ 1979;priorto
1881] of Joneset al. [ 1982; 1881 through1981]. The
Laki
temperaturedeviationscalculatedby AngellandKorshover
100 [ 1985] are basedon the differencebetweenthe average
Santa
M•ari•* temperature
deviations
forthefiveyearsimmotiately
before
andafterthevolcaniceruptions.RampinoandSelf [ 1984],
Fuego_•.•Krakatau on the otherhand,subtractedthe lowesttemperatureanomaly
Agung
:'-•-Katm• in the threeyearsaftertheeruptionfrom themean
temperature of theyearbeforetheeruption.As statedby
• 10_1
- Angell andKorshover[ 1985] "...therelatively-longinterval
of five yearsbeforeandaftertheeruption...maybe toolong,
basedon theusualstratosphericlifetimeof volcanicdust
[e.g.,Oliver, 1976].Kelly andSear[1984], for example,
10-2 i i i
provideevidencefor abruptloweringof Northern
1010 1011 1012 1013 1014 Hemispheresurfaceair temperatures two to threemonths
aftereruptionsin thepastonehundredyeardataset,anda
PetrologicEstimateof SulfurYield
returnto pre-emption temperaturelevelduringthethirdyear
followingeruption.We suspectthatthefive-yearaveraging
101
of AngellandKorshover[ 1985]may thereforebe toolongto
discernanyeffectfroma short-livedvolcanicaerosolcloud
B and havethereforeusedthe databaseof Rampinoand Self
L [ 1984],basedon three-yearaverages, to calculatethe
volcano-related NorthernHemispheretemperature anomalies
100-
F SM/••• listed in table 4.
As discussedby Handler[ 1984]andAngelland
KK/• '•C T Korshover[ 1985], warmingof the tropical(andperhapsthe
extratropical)
troposphereassociated with E1Nino events
10-1-
•
MSH
BßK mightmaskthecoolingassociated
timingof theE1Ninocoincides
Sucha phenomenon
with aneruptionif the
with a volcaniceruption.
wouldcomplicate therelationships
between
volcanic
aerosols
(H2SO4 orHC1)andtemperature
decreasesdiscussed
above.Relatedto thisis the possibility
10-2 i i i i thatvolcanicaerosolcloudsmayeventrigger(althoughnot
10l0 1011 1012 1013 1014 1015 actuallycause)anEl Ninoevent.Handler[ 1984]and
Schattenet al. [ 1984] haveproposedmodelsof sucha
PetrologicEstimateofToralAcids(g)
triggeringmechanism, involvingfeedbacks
affectingthe
Fig. 4. (A): Plot of sulfuryield (g) estimatedby petrologic dynamicsof theatmosphericcirculation(shiftin theITCZ
methodvs. estimatednorthernhemispheretemperature and/ordecreases
in thepoleto equatortemperature gradient
decrease.Linearregressionequation(y = whichwouldaffectzonalwind velocitiesandatmospheric
5.89le-5*x•0.3076, R=0.92) calculatedwith all points. See motions).It is notewonhythatthelasttwo anomalous E1
text for further discussion. Nino events(thosethathavenot followedthenormalpattern)
(B): Plot of totalacids(g) estimatedby petrologicmethod occurredin 1963and 1982-83,or coincidingwith the Agung
vs. estimatednorthernhemispheretemperaturedecrease. andE1Chichoneruptions. The strongest
E1Nino eventin
Linearregression equation(y = 9.220e-5*x•0.2739,R = recenttimesoccurredin 1940-41 andwas apparentlyalsoan
0.87). calculatedwith all points.Seetextfor further anomalousevent,similarto the 1982-83event.Althoughno
discussion. knownsignificant volcaniceruptionsoccurredat thistime,
Lyonset al. [in prep.]havedocumented a 1941volcanic
event on the basis of increased excess sulfate and chloride in
an ice corefrom Dye 3, Greenland.
in table4 are all associated
with largeerrors.This is The occun'ence of frosttingsin treesexhibitsa good
especially
maefor thetwoearliesteruptions
(Laki-1783and relationshipwith thetimingof knownvolcaniceventsandE1
Tambora-1815),as temperature
recordsarelesscomplete Ninos,includinga frosteventin 1941,for whichthereis no
priorto about1880[GrovemanandLandsberg, 1979;Jones known volcanicevent [LaMarche and Hirschboeck,1984]. A
et al., 1982;Angell andKorshover,1985].Furthermore,the majorvolcaniceruptionmay haveeasilygoneunreported
northernhemisphere temperature
deviationsfor theTambora duringthe secondworldwar, asthenumberof volcanoes
andLaki eruptions werecalculated froma differentdataset reportedactivein thewesternPacificandIndonesia in the
andby a differentmethodthanfor theotherevents[Rampino years1941to 1945droppedby one-thirdfrom the
and Sell 1984]. preceeding five years[S'lmikinet al. 1981]. It is thuslikely
The surfacetemperature
changesfollowingsixmajor thata significantvolcaniceruptionwasresponsible for the
volcaniceventsbetween1780 and 1980 (Laki-Asama, volcanicaerosolin 1941andmay havebeenassociated with
Tambora,Coseguina,
Krakatau,SantaMaria-Pelee-Soufriere the frosttingsandE1Nino event.
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 47

TABLE 4. Estimatesof totalvolcanicaerosolsfrom petrologic, In general,the yield estimatesof sulfur,chlorineand

atmosphericandice coretechniques,and estimatedtemperature fluorineemissionson basisof thepetrologicmethodcompare
decreases
afterthe eruptions. well with othertechniques (atmosphericandice core
methods)for estimatingvolcanicvolatileemissions. The
EruptionPetrologic 1 Atmospheric 2 IceCore 3 AøC 4 correspondance betweenthepeu'ologicandatmospheric
grams grams grams aerosolestimatesis goodwhile ice coreestimatesseemto
compare
bestforthelargest
eruptions
()1013g).Thedata
Agung 5.17x 1012 3to9x1012 2x1013 0.3 presentedheresupporttheobservationof Devineet al.
(3x1013) [ 1984]of a relationshipbetweenthe sulfuryield of a number
Krakatau 6.69x 1012 3 to9x 1013 5.5x 1013 0.3 of majorhistoriceruptionsanda northernhemisphere
( 3.8x1013) temperature decrease.A similarcorrelationis seenfor total
Tambora 3.94x 1014 2 x 10TM 1.5x 10TM 0.7 acidsvenustemperature decrease, althoughtherelationship
(1.5x10
TM) is not as good.
Laid 9.19x 1013 1 x 10TM 1.0 The chemicalcomposition of volcanicvolatiles,i.e. the
1.7x 10TM proportionsof sulfur,chlorineandfluorine,differs
Eldgja 9.39x 1013 systematically from volcanoto volcano,andin general
Katmai 1.11x 1013 1.34x1013 3x1013 0.2 reflectsthe chemistryof the eruptingmagmasandthe
Rabaul 1.48x 1012 3 x 10TM 4.5x 1013(5) solubilityof sulfurandhalogensin thesemelts.Thus
StHelens 1.67x 1011 3 x 1011 <0.1 eruptionof high-silicamagmas,suchasrhyolitesand
Coseguina
1.23x1013 (2.3x1013)0.4 dacites,generallyresultsin relativelylow yield of sulfurand
Tarawera 1.13x1013 halogensto the atmosphere, in spiteof the highmass
Santa
Maria9.3lx1012 <2x1013 (2.2x1013)0.4 eruptionrateandlargetotaleruptedmasscharacteristic of
Bezymianny
7.06x1012 0.3 manysuchevents.
Thustheveryviolen,•
131A.D.Taupo
Fuego 4.95x10123 to 6x1012 0.4 eruptionin New Zealandemitted35 km• of silicicmagmaup
Oraefajokull
3.39x1012 to 51 km height,butproduceda negligibleyield of S, C1and
Taupo 2.41x1012 F. We observe,for example,thatthe atmospheric yield of
Soufriere 5.03x1010 2.3x109 H2SO4 fromeruptions
of silicicmagmasistypically100to
500ppmpereruptedmass,whereas theH2SO4 yieldof
1DatafromDevine
etal.[1984],
Sigurdsson
etal.[1985], basalticeruptionsto the atmosphere is in therangeof 1000 to
Rose et al[ 1982] and this work. 5000 ppm.Their explosivenature,alongwith the low
2DatafromDeirmendjian
[1973],
Volz[1975],
Stothers contentof associated acidicvolatiles,suggests thatsilicic
[1984a; 1984b], Kent [1982], McCormick et al [1981; explosiveeruptionsmay be moreeasilyrecognizedin ice
1982], RampinoandSelf [ 1984], Cadleet al [ 1977]. coresby thepresenceof fine silicateash,ratherthanby the
3Data
from
Hammer
etal[1980]
(H2SO4);
number
in presenceof aciditypeaks.
At theotherextremearefissureeruptionsof basaltic
parentheses
fromLegrand
andDelmas[1987](H2SO4 magmas,suchasfrom EldgjaandTaraweraandthe 1783
nly). Laki eruption,which injectlarge amountsof sulfur,chlorine
Reported
Northern
Hemisphere
surface
temperature
change andfluorineinto the atmosphere. While mostfissure
assocaited
with the eruption.Data from Sigurdsson[ 1982], eruptionsprobablyproducetropospheric aerosolclouds,
RampinoandSelf [ 1982],Robock[1981] andJoneset al theireruptioncloudsmayreachthe stratosphere undercertain
1981]. circumstances[Devine et al., 1984; Wolff et al., 1984;
Unknowneruptionreco•ed asan aciditylayerin Walker et al., 1984;CareyandSparks,1986;Stotherset al.
GreenlandDye-3 core540_+10AD. PossibleRabauleruption 1986].
(1400 yrsBP.; Hammer[ 1980]. A thirdcompositional typeof volcanicvolatilesare
exhibitedduringeruptionof trachyteandtrachyandesite
magmas,suchasthe greatTamboraeruptionof 1815.Such
Conclusions magmastypicallyhavevery highyield of halogensto the
atmosphere,in additionto sulfur.The relationshipbetween
The petrologicestimateof volcanicvolatileemissions magmacomposition andthepetrolo•c estimates of volcanic
from majorvolcaniceruptionsis a viablemethodof yield
ofH2SO
4,HC1
and
HFperkm3ofmagma
erupted
are
determining themassyield to theatmosphere of components shownin figures5a, 5b and5c andlistedin table5. In
suchas sulfur,chlorineandfluorine.As thisapproachonly general,thereis a fairly systematic
decrease
in sulfur,
requiressamplingandstudyof the geologialdepositfrom the chlorine
and
fluorine
yield
perkm3with
increasing
SiO2of
eruption,it canbe appliedto theestimationof volatileyield
fromanyvolcaniceventin thegeologicrecord.This is of themagma.Most of thebasalticeruptions(Eldgja,Surtsey,
particularvalue,astherecurrence intervalof trulylarge KraJl•
a, Laki), however,havea quite•low yield of HC1per
volcaniceventsis severalhundredyears,andsuchevents km6ofmagma
(~ 10:•tons
HCl/km:•).
Theonlyexcep•tion
is
havenotyet takenplacesincetherigorousmonitoringof the theTarawerabasalticeruption,with a yieldof 3.4x10ø tons
atmosphere began.The methodprovidesonly a minimum HC1/km
3magma.
These
results
confmn
theobservation
of
estimate,buttheresultshavehelpedus milne our Devineet al [ 1984]thateruptionsof basalticmagmarelease
undentanding of a numberof majoreruptionsfor whichlittle an orderof magnitudemore sulfur,chlorineandfluorinethan
other information exits. eruptionsof silicicmagma.
Geophysical Monograph Series Understanding Climate Change Vol. 52

45 VOLCANIC VOLATILE EMISSIONS

108 TABLE5. Eruptions studied

bythepetrologicomethod
and
theiryieldof acids(H2SO4, HC1,HF) perkm• of magma
A erupted[Devineet al., 1984andthiswork].Eruptions
are
107- listedin orderof increasing
H2SO4 yield.
EruptionH2SO4/km3 HC1/lcm
3 HF/km
3
metric tons metric tons metric tons
106-
Tarawera 1.10x 107 3.38x 106 1.79x 106
Eldgja 9.31x 106 4.34x 105 6.89x 106
105- Laki 7.17x 106 1.27x 105 0
Krafla 5.05x 106 1.89x 105 n.a.
Soufriere 4.82x 106 2.32x 107 0
Kada 4.81x 106 0 n.a.
3.70x 106 3.36x 106 0
ß ! ß

40 50 80
Bezymianny
Heimaey 3.61x 106 0 0
SiO2 Magnin Surtsey 3.52x 106 1.44x 105 n.a.
Agung 3.16x 106 1.70x 106 8.89x 105
Rabaul 2.47x 106 0 n.a.

108 Campanian2.41x 106 0 0

Katmai 8.72x 105 3.57x 105 7.38x 103
Santa
Maria 6.62x 105 4.33x 105 0
B Tambora 5.99x 105 2.47x 106 1.44x 106
107- HeklaH-1 4.48x 105 0 n.a.
StHelens1980 3.16x 105 1.41x 105 2.12x 105
Krakatau 2.94x 105 3.75x 105 0
106- StHelens1530 2.32x 105 8.26x 104 n.a.
HeklaH-3 2.20x 105 1.16x 105 n.a.
Oraefajokull1.90x 105 3.30x 105 1.18x 106
105- Minoan 1.54x 105 0 n.a.
Taupo Plinian5.63x 104 2.22x 105 1.94x 105
StHelens1800 1.41x 104 1.06x 106 n.a.
104 . • ß I '
Roseau Tuff 0 3.83x 105 n.a.
40 50 60 70 80 Coseguina 0 1.23x 106 0
SiO2 Magma n.a.= not analyzed

Analysesof boththe Campman andRabaulignimbrites

107 haveshownthatmetasomatism of ignimbritesandtephra-fall
buriedunderignimbritesmay leadto anomalously highlevels
of S, C1andF in the matrix glassof thesedeposits.
c Therefore,in theabsenceof unalteredignimbritesamples,
106- theamountof S, C1andF degassed in ignimbrite-forming
eruptions cannotbe directlyestimated.The petrologic
estimateof thevolatileyield from theplinianphaseof the
105- 1400 yr B.P. Rabauleruption,on the otherhand,is shown
to be at leasttwo ordersof magnitudelowerthantheestimate
of atmospheric aerosolmasspresentin theatmosphere in 536
A.D., asdeterminedby Stothers[ 1984]. We therefore
104- concludethatthe eventin 536 A.D. whichperturbedthe
atmospheric opticaldepthwasprobablyunrelatedto the
Rabauleruption.
103 ß , ß , . , Appendix
40 50 60 70 80

SiO2 Magma SampleIdentification

Fig. 5. Plot of sulfuric(A), hydrochl,,oric

(B) and Samplenumbers referto numbers of columns in Table3.
hydrofluoric
(C)acidyield(tons/km
3 magma)
versus
silica 1. ELD-1, averagecomposition of meltinclusiontrappedby
contentof magma.SeeTable5 for additionaldetails. pyroxenephenocrysts in basaltictephrafromEldgja
Geophysical Monograph Series Understanding Climate Change Vol. 52

PALAIS AND SIGURDSSON 49

eruptionin 934 A.D. Eldgjatephrafall collectedby
Thorvaldur Thordarson, 1.5 km SSW of Kambavam.
Sampleno. 160883,sectionV.
2. ELD-1, averagecomposition of matrixglassin basaltic
tephrafromEldgjaeruptionin 934 A.D. Eldgjatephra
fall collectedby ThorvaldurThordarson,1.5 lcmSSW of
Kambavam. Sampleno. 160883,sectionV.
3. O-1362,averagecomposition of meltinclusions trapped
by pyroxenephenocrysts in rhyolitictephrafall from
Oraefajokull,Icelanderuptionin 1362A.D. Collectedby
H. Sigurdsson, 29/7/83 at Fagurholsmyri, Iceland.
. O- 1362,averagecomposition of matrixglassin rhyolitic
tephrafall from eruptionin Oraefajokull,Iceland1362
A.D. Collectedby H. Sigurdsson, 29/7/83 at
Fagurholsmyri,Iceland.
5. TP-4, averagecomposition of melt inclusionstrappedby
pyroxenephenocrysts in rhyolitictephra,130 A.D.
Hatepeplinianphaseof Taupoeruption,NorthIsland,
New Zealand. Collectedby C.J.N. Wilson, locality at
N103/580133.
, TP-4, averagecomposition of matrixglassin rhyolitic
tephra,130 A.D. Hatepeplinianphaseof Taupoeruption,
North Island,New Zealand. Collectedby C.J.N.
Wilson, locality at N 103/580133.
7. TP-5, averagecomposition of meltinclusionstrappedby
pyroxenephenocrysts in rhyolitictephra,130 A.D.
Rotongaiophreatoplinian phaseof Taupoeruption,North
Island,New Zealand.Collectedby C.J.N. Wilson,
localityat N103/580133.
8. TP-5, averagecompositionof matrixglassin rhyolitic
tephra,130 A.D. Rotongaiophreatoplinian phaseof
Taupoerup6on,North Island,New Zealand.Collectedby
C.J.N. Wilson, locality at N103/580133.
9. TP-6, averagecompositionof melt inclusionstrappedby
pyroxenephenocrysts in rhyolitictephra,130 A.D. Taupo
plinianeruption,NorthIsland,New Zealand.Collected
by C.J.N. Wilson, locality at N103/549233.
10. TP-6, averagecompositionof matrixglassin rhyolitic
tephra,130 A.D. Taupoplinianeruption,North Island,
New Zealand.Collectedby C.J.N. Wilson, locality at
N103/549233.
11. K- 192A, averagecomposition of melt inclusionstrapped
by pyroxenephenocrysts in rhyolitictephra(earlytephra,
basal6 cm of a 19 cm thicksection)from earlyphaseof
1912 A.D. Katmai plinianashfall, Alaska,U.S.A.
Samplecollectedby J. Riehle,9/85, 17 km NE of
NovaruptaDome.
12. K-192A, averagecomposition of matrixglassin rhyolitic
tephra(earlytephra,basal6 cm of a 19 cm thick section)
from earlyphaseof 1912A.D. Katmaiplinianashfall,
Alaska,U.S.A. Samplecollectedby J. Riehle,9/85, 17
km NE of NovaruptaDome.
13. K-192B, averagecomposition of melt inclusionstrapped
by plagioclase phenocrystsin rhyolitictephra (middle
tephra,6 cm of a 19 cm thicksection)from middlephase
of 1912 A.D. Katmaiplinianashfall, Alaska,U.S.A.
Samplecollectedby J. Riehle,9/85, 17 km NE of
NovaruptaDome.
14. K-192B, averagecomposition of matrixglassin rhyolitic
tephra (middletephra,6 cm of a 19 cm thicksection)
from middlephaseof 1912A.D. Katmaiplinianfall,
Alaska,U.S.A. Samplecollectedby J. Riehle, 9/85, 17
km NE of NovaruptaDome.
15. K-192C, averagecomposition of melt inclusionstrapped
by pyroxenephenocrysts in rhyolitictephra(latetephra,
top7 cm of a 19 cm thicksection)fromlatephaseof 1912
A.D. Katmaiplinian fall, Alaska,U.S.A. Sample
collectedby J. Riehle,9/85, 17 km NE of NovaruptaDome.
16. K-192C, averagecomposition of matrixglassin rhyolitic
tephra (latetephra,top7 cm of a 19 cm thicksection)
fromlatephaseof 1912A.D. Katmaiplinianfall, Alaska,
U.S.A. Samplecollectedby J. Riehle,9/85, 17 km NE
of NovaruptaDome.
17. K-192C, averagecomposition of meltinclusions trapped
by plagioclasephenocrysts in rhyolitictephra(latetephra,
top7 cm of a 19 cm thicksection)fromlatephaseof 1912
A.D. Katmaiplinian fall, Alaska,U.S.A. Sample
collectedby J. Riehle,9/85, 17 km NE of Novarupta
Dome.
18. K-192C, averagecompostion of matrixglassin rhyolitic
tephra (late tephra,top 7 cm of a 19 cm thick section)
from latephaseof 1912A.D. Katmaiplinianfall, Alaska,
U.S.A. Samplecollectedby J. Riehle,9/85, 17 km NE
of NovaruptaDome.
19. Total Katmai-Sumof thevolatilesreleasedin early,
middleandlatephasetephra
20. Rab-1,averagecompositionof melt inclusionstrapped
by plagioclasephenocrysts in dacitictephrafrom 1400 yr.
B.P. Rabaulplinianfall deposit,Papua,New Guinea.
Samplescourtesyof Dr. P. Lowenstein.
21. Rab-1,averagecompositionof melt inclusionstrapped
by pyroxenephenocrysts in dacitictephrafrom 1400 yr.
B.P. Rabaulplinianfall deposit,Papua,New Guinea.
Samplescourtesyof Dr. P. Lowenstein.
22. Rab-1, averagecompositionof matrix elassin dacitic
tephrafrom 1400yr. B.P. Rabaulplinianfall deposit,
Papua,New Guinea.Samplescourtesyof Dr. P.
Lowenstein.
23. Rab-3,averagecomposition of meltinclusionstrapped
by plagioclase
phenocrysts in dacitictephrafrom 1400
yr. B.P. Rabaulpyroelasticflow deposit,Papua,New
Guinea.Samplescourtesyof Dr. P. Lowenstein.
24. Rab-3,averagecomposition of matrixglassin dacitic
tephrafrom 1400 yr. B.P. Rabaulpyroelasticflow
deposit,Papua,New Guinea.Samplescourtesyof Dr. P.
Lowenstein.
25. IT-2053, averagecomposition of melt inclusionstrapped
by pyroxenephenocrysts in trachytictephrafrom upper
half of plinianfall, Campanianeruption(.--35,000yrs.
B.P.), PhlegreanFields,Italy. Samplescollectedby H.
Sigurdsson (5/84) in quarrynearLavorate,Italy.
26. IT-2053, averagecomposition of matrixglassin trachytic
tephrafrom upperhalf of plinianfall, Campanianeruption
(.--35,000 yrs.B.P.), PhlegreanFields,Italy. Samples
collectedby H. Sigurdsson (5/84) in quarrynear
Lavorate,Italy.
27. IT-2054, averagecomposition of melt inclusionstrapped
by pyroxenephenocrysts in trachytictephrafrom lower
half of plinianfall, Campanianeruption(.--35,000yrs.
B.P.), PhlegreanFields,Italy. Samplescollectedby H.
Sigurdsson (5/84) in quarrynearLavorate,Italy.
28. IT-2054, averagecompositionof matrix glassin trachytic
tephrafrom lowerhalf of plinianfall, Campanianeruption
(.--35,000yrs. B.P.), PhlegreanFields,Italy. Samples
collectedby H. Sigurdsson (5/84) in quarrynear
Lavorate,Italy.
Geophysical Monograph Series Understanding Climate Change
50 VOLCANIC VOLATILE EMISSIONS
29. TW- 1, averagecompositionof melt inclusionstrapped
by pyroxenephenocrysts in basaltictephrafromplinian
fall, 1886 A.D. Taraweraeruption,North Island,New
Zealand.Collectedby C.J.N. Wilson,localityat
N77/984913
30. TW-1, averagecomposition of matrixglassin basaltic
tephrafromplinianfall, 1886A.D. Taraweraeruption,
North Island,New Zealand.Collectedby C.J.N. Wilson,
localityat N77/984913
31. TW-2, averagecomposition of meltinclusionstrapped
by pyroxenephenocrysts in basaltictephrafromplinian
fall, 1886A.D. Taraweraeruption,North Island,New
Zealand.Collectedby C.J.N.Wilson,localityat
N77/973936,exposure1.5-2.0m belowtopof 1886
scoria fall.
32. TW-2, averagecomposition of matrixglassin basaltic
tephrafromplinianfall, 1886A.D. Taraweraeruption,
North Island,New Zealand.Collectedby C.J.N. Wilson,
localityat N77/973936,exposure1.5-2.0m belowtopof
1886 scotia fall.
33. MSH-201, averagecomposition
of melt inclusions
trapped
byplagioclase
phenocrysts
in daciti6tephra,
Mount St. HelensKalamaeruptiveperiod-1530 A.D.
CW" tephra),Washington,U.S.A. Collectedin Bear
Meadow by H. Sigurdsson,8/14/84.
34. MSH-201, averagecomposition of matrixglassin dacitic
tephra,MountSt. HelensKalamaeruptiveperiod-1530
A.D. CW" tephra),Washington,U.S.A. Collectedin
BearMeadowby H. Sigurdsson, 8/14/84.
35. MSH-202, averagecomposition of melt inclusions
trappedby plagioclase phenocrysts in dacitictephra,
MountSt. HelensGoatRockseruptiveperiod-1800
A.D. CT" tephra),Washington,U.S.A. Collectedin Bear
Meadowby H. Sigurdsson, 8/14/84.
36. MSH-202, averagecomposition of matrixglassin dacitic
tephra,Mount St. HelensGoatRockseruptiveperiod
-1800 A.D. CT" tephra),Washington,U.S.A. Collected
in BearMeadowby H. Sigurdsson, 8/14/84.
37. COS-5,averagecomposition of meltinclusions trapped
by phenocrysm in darkgrey-blackscoriafall from the
1835A.D. Coseguina, Nicaraguaeruption.Sample
collectedby SteveSelf on the E-SE flanksof thevolcano.
38. COS-5,averagecomposition of matrixglassin dark
grey-blackscoriafall from the 1835A.D. Coseguina,
Nicaraguaeruption.Samplecollectedby SteveSelf on
the E-SE flanks of the volcano.
39. SM-1902, averagecomposition of meltinclusions
trappedin phenocrystsof pumiceousdacitictephrafrom
the 1902A.D. eruption.Samplecollectedby Bill Roseon
therim of SieteOrejasvolcanoabout6 km NNW of the
SantaMaria vent, Guatemala.
40. SM-1902, averagecomposition of matrixglassof
pumiceous dacitictephrafrom the 1902A.D. eruption.
Samplecollectedby Bill Roseon therim of SieteOrejas
volcanoabout6 km NNW of the SantaMaria vent,
Guatemala.
41. BZ- 1956,averagecomposition of meltinclusions
trappedin phenocrystsin andesitictephrafrom theMarch
30, 1956 eruptionof Bezymiannyvolcano,U.S.S.R.
Samplecollectedby Dr. G.E. Bogoyavlenskaya at a
distanceof about80 km northfrom the volcano(10 km
northfnxn the settlement
of Klyuchi). At thissitethe
tephralayerwasabout2 cm thick.
Vol. 52
42. BZ-1956, averagecompositionof matrix glassfrom
andesitictephrafrom the March 30, 1956eruptionof
Bezymiannyvolcano,U.S.S.R. Samplecollectedby Dr.
G.E. Bogoyavlenskaya at a distanceof about80 km north
from the volcano(10 km northfrom the settlementof
Klyuchi). At thissitethe tephralayer wasabout2 cm
thick.
43. BZ- 1956,averagecompositionof melt inclusions
trappedin phenocrystsin andesitictephrafromtheMarch
30, 1956 eruptionof Bezymiannyvolcano,U.S.S.R.
Samplecollected
by Dr. G.E. Bogoyavlenskaya
at a
distanceof about80 km north from the volcano(10 km
northfromthe settlement
of Klyuchi). At thissitethe
tephralayerwasabout2 cmthick.Samesampleas41/42
butaverageof high-silicainclusions.
44. BZ-1956, averagecomposition of matrixglassfrom
andesitic
tephrafromtheMarch30, 1956eruptionof
Bezymianny volcano,U.S.S.R. Samplecollected by Dr.
G.E. Bogoyavlenskaya at a distanceof about80 km north
from thevolcano(10 km northfrom the settlementof
Klyuchi). At thissitethetephralayerwasabout2 cm
thick.Samesampleas41/42 butaverageof high-silica
matrix glass.
Acknowledgements. We wish to thankD. Browningfor
hishelpin samplepreparationandmicroprobeanalysis.
S.N. Carey wrotethe computerprogramfor determiningthe
statisticalsignificanceof our resttim.We gratefully
acknowledge thefollowingpeoplewho providedsamplesfor
this study:ThorvaldurThordarson(Eldgja), C. J. N.
Wilson (TaupoandTarawera),J. Riehle (Katmai), P.
Lowenstein(Rabaul),S. Self (Coseguina),W.I. RoseJr.
(SantaMafia), andG.E. Bogoyavlenskaya (Bezymianny).
We alsowishto thankS.N. CareyandW. Cornell for
assistance with themicroprobeandfor helpfuldiscussions.
This work wasfundedby the NationalScienceFoundation
grantEAR-8503104.
References
Anderson,A.T., Chlorine,sulfur,and water in magmasand
oceans,Geol. Soc. Am. Bull., 85, 1485-1492, 1974.
Angell, J.K., The closerelationbetweenAntarctic
total-ozonedepletionandcoolingof the Antarcticlow
stratosphere, Geophys.Res. Lett., 13, 1240-1243, 1986.
Angell, J.K., andJ. Korshover,Surfacetemperature
changesfollowingthe six majorvolcanicepisodes
between1780 and 1980, Jour.Climate Appl. Meteorol.,
24, 937-951, 1985.
Bailey, D.K., andR. Macdonald,Fluorineandchlorinein
peralkalineliquidsandtheneedfor magmageneration in
an opensystem,Mineral. Mag., 40, 405-414, 1975.
Barberi, F., F. Innocenti, L. Lirer, R. Munno, and T.
Pescatore,The Campanianignimbrite:A majorprehistoric
eruptionin theNeapolitanarea(Italy), Bull. Volcanol.,
41, 1-22, 1978.
Benoist,J.P., J. Jouzel, C. Lorius, L. Merlivat, and M.
Pourchet,Isotopeclimaticrecordoverthe last2.5 Ka
from Dome C, Antarctica, ice cores, Annals Glaciol,, 3,
17-22, 1982.
Bull G.A. andD.G.James,Dustin the stratosphere
over
Geophysical Monograph Series
westernBritainon April 3 and4, 1956,Meteorol.Mag.
G.B., 85, 1956.
Carey,S.andR.S.J.Sparks,Quantitativemodelsof the
falloutanddispersalof tephrafromvolcaniceruption
columns, Bull, ¾91canol., 48, 109-125, 1986.
Carmichael, I.S.E., The iron-titanium oxides of salic
volcanicrocksandtheirassociated
ferromagnesian
silicates,Contr. Min. Petrol.. 14, 36-64, 1967.
Castleman, A.W., Jr., H.R. Munkelwitz, and B. Manowitz,
Isotopicstudiesof thesulfurcomponent of the
stratospheric aerosollayer,Tellus,26, 222-234, 1974.
Chuan,R. L., J. Palais,W.I. Rose,and P.R. Kyle, Fluxes,
sizes,morphologyandcompositions of particlesin the
Mt. Erebusvolcanicplume,December,1983,J. Atmos.
Chem.,4, 467-477, 1986.
Cornell,W.C., S.N. Carey, and H. Sigurdsson,Computer
simulationof transportanddepositionof the Campanian
Y-5 ash, J, Volcanol. Geotherm. Res., 17, 89-109,
1983.
Crandell, D.R., D.R. Mullineaux, and M. Rubin, Mount St.
Helens volcano: Recent and future behavior, Science,
187,438-441, 1975.
Crandell, D.R. and D.R. Mullineaux, Potential hazards from
futureeruptionsof Mt. St. Helensvolcano,Washington,
U.S. Geol. Surv. Bull., 1383-C, 1-26, 1978.
Crutzen,P., Photochemicalreactionsinitiatedby and
influencingozonein unpollutedtropospheric air, Tellus,
26, 47-57, 1974.
Curtis,G.H., The stratigraphy
of t•e ejectafrom the 1912
eruptionof Mount KatmaiandNovarupta,Alaska,in
$tugliesin Volcanology,editedby R.R. Coats,R.L. Hay,
and C.A. Anderson, Geol. Soc. Am. Mem., 116,
153-210, 1968.
De Angelis,M., L., Fehrenbach,C. J6hanno,andM.
Maurette,Micrometre-sizedvolcanicglassesin polarices
and snows,Nature, 317, 52-54, 1985.
Deirmendjian, D., On volcanicandotherparticulateturbidity
anomalies,Advancesin Geophysics,16, 267-297, 1973.
Devine,J.D., and H. Sigurdsson, The liquid composition
andcrystallization historyof the 1979Soufrieremagma,
St. Vincent, W.I., J. Volcanol. Geotherm. Res.. 16,
1-31, 1983.
Devine,J.D., H. Sigurdsson,
A.N. Davis,andS. Self,
Estimatesof sulfurandchlorineyield to theatmosphere
fromvolcaniceruptions andpotentialclimaticeffects,J.
Geophys.Res.,89, 6309-6325,1984.
Dunbar,N., andP. R. Kyle,H20 andC1contents,
and
temperature of TaupoVolcanicZonerhyoliticmagmas
(abstract),Inter. Volcanol. Congr.,New Zealand,Feb.,
1986.
Federman,A.N., CorrelationandPetroloeicalInterpretation
of AbyssalandTerrestrialTephraLayers,PhD
Dissei'tation,
Oregon
StateUniversity,
241pp.1984.
Fierstein,J. andW. Hildreth,Ejectadispersalanddynamics
of the 1912eruptionsat Novarupta,KatmaiNational
Park, Alaska, Eos Trans. AGU, 67, 1246, 1986.
Gorshkov,G.S., Giganticeruptionof the volcano
Bezymianny,Bull.Volcanol.,20, 77-109, 1959.
Griggs,R.F., The Valley of Ten ThousandSmokes,
National
Geographi•
Society,
340pp.,1922.
Groveman,B.S. and H.S. Landsberg,Reconstruction
of
NorthernHemisphereTemperature:1579-1880,
MeteorologyProgram,Univ. of Maryland.Publ.79-! • 1,
1979.
Understanding Climate Change Vol. 52
PALAIS AND SIGURDSSON 51
Hammer,C.U., Acidityof polarice coresin relationto
absolutedating,pastvolcanism,andradioechos,J.
Glaciol., 25, 359-372, 1980.
Hammer,C.U., Initial directcurrentin the buildupof space
chargesandtheacidityof ice cores,J. Phys.Chem.,87,
4099-4103, 1983.
Hammer, C.U., Tracesof Icelandiceruptionsin the
Greenland ice sheet,Jokull, 34, 51-65, 1984.
Hammer, C.U., H. B. Clausen,and W. Dansgaard,
Greenlandice sheetevidenceof post-glacial
volcanism
anditsclimaticimpact,Nature,288, 230-235, 1980.
Handler,P., Possibleassociation
of stratospheric
aerosols
andE1Nino typeevents,Geophys.Res.Lett., 11,
1121-1124, 1984.
Healy,J., Stratigraphyandchronology of lateQuaternary
volcanicashin Taupo,RotoruaandGisbomedistricts,
Part 1, N,Z, Geol. Surv.Bull., 73, 88pp., 1964.
Heming,R.F., Geologyandpetrologyof Rabaulcaldera,
Papua,New Guinea,Geol.Soc.Am. Bull., 85,
1253-1264, 1974.
Herron,M.M., Impuritysources
of F-, CI-, NO3- and
SO42-
inGreenland
and
Antarctic
precipitation,
J.
Geophys.Res.,87,3052-3060, 1982.
Hildreth,W., The compositionally zonederuptionof 1912
in theValleyof TenThousand Smokes,KatmaiNational
Park, Alaska,J. Volcanol. Geotherm.Res.,18, 1-56,
1983.
Hoblitt, R.P., D.R. Crandell,D.R. Mullineaux, Mount St.
Helenseruptivebehaviorduringthepast1,500yr.,
Geology,8, 555-559, 1980.
Hoff, R.M. and A.J. Gallant, Sulfur dioxide emissionsfrom
La Soufrierevolcano, St. Vincent, West Indies, Science,
209, 923-924, 1980.
Johnston,D.A., Volcanic contributionof chlorineto the
stratosphere:
More significant
to ozonethanpreviously
estimated?,Science,209, 491-493, 1980.
Jones,P.D., T.M.L. Wigley andP.M. Kelly, Variationsin
surfaceair temperatures:Part 1. NorthernHemisphere,
1881-1980, Mon,Wca. Rev., 110, 59-70, 1982.
Kelly,P.M.,andC.B.Sear,Climatic(impact
of explosive
volcaniceruptions,
Nature,311,740-743, 1984.
Kennett, J.P. and R.C.,Thunell, Global increasein
Quaternary
explosivevolcanism,Science,187,497-503,
1975.
Kennett,J.P., andR.C. Thunell,On explosiveCenozoic
volcanismandclimaticimplications,Science,196,
1231-1234, 1977.
Kent, G.S., SAGE measurementsof Mount St. Helens
volcanic aerosols,NASA Conf. Publ., 2240, 109-116,
1982.
LaMarche,V.C., Jr. andK.K. Hirshboeck,Frosttings in
treesasrecordsof majorvolcaniceruptions,Nature,307,
121-126, 1984.
Lamb, H.H., Volcanicdustin the atmosphere,Phil. Trans.
ROy.Soc.Lond..ser.A, 266, 425-533, 1970.
Lamb, H.H., Volcanicactivityandclimate,Paleogeo_m'aph..
Paleoclimatol.. Paleoecol., 10, 203-230, 1971.
Larsen,G. Um aldurEldgjarhrauna,
Natturufraedingurinn,
49, 1-25, 1979.
Legrand,M. andR.J.Delmas,A 220-yearcontinuousrecord
of volcanic
H2SO4 in theAntarctic
icesheet,Nature,
327, 671-676, 1987.
Legrand,M., Chimiedesneigeset glacesAntarctique:Un
Geophysical Monograph Series Understanding Climate Change
52 VOLCANIC VOLATILE EMISSIONS
refletde l'environment,
UniversiteScientifique
et
Medicale de Grenoble, 1985.
Legrand,M.R. and R.J. Delmas, The ionic balanceof
Antarcticsnow:A 10-yeardetailedrecord,Atmos.
Environ,, 18, 1867-1874, 1984.
Maccagnan,M. J.M. Barnola,R. Delmas,andP. Duval,
Staticelectricalconductivity
asanindicatorof thesulfate
contentof polarice cores,Geophys.Res. Lett.,8,
970-972, 1981.
Mankin, W.G. andM.T. Coffey,Incresesstratospheric
hydrogenchloridein theE1Chichoncloud.Science,226,
170-172, 1984.
McCormick, M.P., Ground-basedand airborne
measurements
of Mount St. Helensstratospheric
effluents, NASA Conf. Publ., 2240, 125-130, 1982.
McCormick, M.P., G.S. Kent, G.K. Yue, and D.M.
Cunnold,SAGE measurements of the stratospheric
aerosoldispersionandloadingfrom the Sourfriere
volcano,NASA Tech. Pap., 1922, 1-20, 1981.
Miller, E., Vapor-liquidequilibiraof water-hydrogen
chloride solutionsbelow OøC. J. Chem. Ene. Data. 28.
363-367, 1983.
Mullineaux,D.R., Summaryof pre-1980tephra-falldeposits
eruptedfrom Mount St. Helens,Washingtonstate,USA,
Bull, Volcanol,, 48, 17-26, 1986.
Mullineaux,D.R. andD.R. Crandell,The eruptivehistoryof
Mt. St. Helens,In; P.W. Lipman,D.R. Mullineaux
(eds.)The 1980 eruptionsof Mt. St. Helens,Washington,
U.S. Geol. Surv.Prof. Pap., 1250, 3-15, 1981.
Nielsen,C. H., andH. Sigurdsson,Quantitativemethods
for electronmicroprobeanalysisof sodiumin naturaland
syntheticglasses,Am. Mineral., 66, 547-552, 1981.
Oliver, R.C., On the responseof hemisphericmean
temperature to stratospheric
dust:An empiricalapproach,
Jour.Appl. Meteor., 15, 933-950, 1976.
Oskarsson,N., The interactionbetweenvolcanicgasesand
tephra:Fluorineadheringto tephrafrom the 1979Hekla
eruption,J. Volcanol. Geotherm.Res.,8, 251-266, 1980.
Palais,J.M. andM. Legrand,Solubleimpurities
in theByrd
Stationice core,Antarctica:
Theiroriginandsources,
J.
Geophys.Res., 90, 1143-1154, 1985.
Pollack,J.B.,O.B. Toon,C. Sagan,A. Summers,B.
Baldwin,andW. Van Camp,Volcanicexplosions and
climatechange:A theoreticalassessment,
J. Geophys.
Res,, 81, 1071-1083, 1976.
Porter,S.C., Recentglaciervariationsandvolcanic
eruptions,Nature,291,139-142, 1981.
Rampino,M.R. andS. Self,Historiceruptionsof Tambora
(1815), Krakatau(1883), andAgung(1963), their
stratosphericaerosols,andclimaticimpact,Quat.Res,,
18, 127-143, 1982.
Rampino,M.R. andS. Self, Sulphur-richvolcanic
eruptionsand stratospheric
aerosols,Nature,310,
677-679, 1984a.
Rampino,M.R., andSelf, S., The atmospheric
effectsof E1
Chichon, $•i, Am,, 48, 1984b.
Rampino,M.R., R.B. Stothers,and S. Self, Climatic effects
of volcaniceruptions,Natur?,313, 272, 1985.
Reclus,E., NouvelleG6ographieuniv•rs½11•, 17, 488-489,
Paris, 1891.
Robock,A., The Mount St. Helensvolcaniceruptionof 18
May 1980: Minimal climaticeffect,Science,212,
1383-1384, 1981.
Vol. 52
Rose, W.I., Jr., R.E. Stoiber, and L.L. Malinconico,
Eruptivegascompositions andfluxesof explosive
volcanoes:budgetof S andC1emittedfromFuego
volcanoGuatemala,in Andesites,editedby R.S. Thorpe,
pp. 669-676, JohnWiley, New York, 1982.
Rose, W.I., R.L. Wunderman, M.F. Hoffman, and L. Gale,
A volcanologist'sreviewof atmospheric
hazardsof
volcanicactivity,FuegoandMt. St. Helens,J. Volcanol.
Geotherm. Res., 17, 133-157, 1983.
Rose, W.I., Santa-Maria, Guatemala: Bimodal soda-rich
calcalkalic stratovolcano,J. Volcanol. Geotherm. Res., in
press.
Rosi,M., A. SbranaandC. Principe,The Phlegraeanfields:
Structuralevolution,volcanichistoryanderuptive
mechanisms, J. Volcanol. Geotherm. Res., 17, 273-288,
1983.
Schatten,K.H., H.G. Mayr, I. Harris, and H.A. Taylor,
Jr., A zonallysymmetricmodelfor volcanicinfluence
uponatmosphericcirculation,Geophys.Res.Lett., 11,
303-306, 1984.
Self, S., M.R. Rampino,andJ.J.Barbera,The possible
effects
oflarge
19thand20th
century
volcanic
eruptions
on zonalandhemispheric
surfacetemperatures,
J.
Volcanol. Geotherm.Res., 11, 41-60, 1981.
Sigurdsson,H., Volcanicpollutionandclimate:The 1783
Laki eruption,Eos Trans.AGU, 63,
601-602, 1982.
Sigurdsson,H. J. D. Devine, andA. N. Davis, The
petrologicestimationof volcanicdegassing,Jokull,35,
1-8, 1985.
Sirnkin, T., L. Seibert, L. McClelland, W.G. Melson, D.
Bridge,C.G. Newhall, andJ. Latter,Volcanoesof the
World, Smithsonian Institution, Wash. D.C., 1981.
Solomon, S. and R.R. Garcia, On the distribution of
long-livedtracersof chlorinespeciesin themiddle
atmosphere. J. Geophys.Res. 89, 11633-11644,1984.
Sparks,R.S.J.,H. Sigurdsson, andL. Wilson,Magma
mixing:a mechanismfor triggeringacidexplosive
eruptions,Nature,267, 315-318, 1977.
Stolarski, R.S., and D.M. Butler, Possibleeffects of
volcaniceruptionson stratospheric minorconstituent
chemistry.Pageoph.117,486-497, 1978.
Stolarski,R.S., andR.J. Cicerone,Stratospheric
chlorine:a
possiblesinkfor ozone.Can. J. Chem. 52, 1610-1615,
1974.
Stothers,R.B., Mysterycloudof AD 536, Nature,307,
344-345, 1984a.
Stothers,R.B., The greatTamboraeruptionin 1815 andits
aftermath, Science, 224, 1191-1198, 1984b.
Stothers,R.B., J.A. Wolff, S. Self andM.R. Rampino,
1986:Basalticfissureeruptions,plumeheightsand
atmospericaerosols.Geophy•. Res. Lett. 13, 725-728.
Thompson,L.G., E. Mosley-Thompson, W. Dansgaard,
andP.M. Grootes,The Little Ice Age asrecordedin the
stratigraphy of the tropicalQuelccayaice cap,Science,
234, 361-364, 1986.
Thorarinsson,
S., The Oraefajokulleruptionof 1362, Acta
Naturalia lslandica, 2, 1-101, 1958.
Walker, G.P.L., The Taupopumice:Productof the most
powerfulknown (ultraplinian)eruption?,J. Volcanol.
Geotherm. Res., 8, 69-94, 1980.
Walker, G.P.L., R.F. Herrting,T.J. Sprod,andH.R.
Walker, Latestmajoreruptionsof Rabaulvolcano,In'
Geophysical Monograph Series
Cooke-Ravian Volumeof Volcanological
Papers,
R.W.
Johnson(ed.), Geol. Surv.PapuaNew GuineaMem.,
10, 181-193, 1981.
Walker, G.P.L., S.Self, and L. Wilson, Tarawera 1886,
New Zealand-a basalticplinianfissureeruption,J.
Volcanol. Geotherm. R•,, 21, 61-78, 1984.
Williams,H., The greateruptionof Coseguina,
Nicaragua,
in 1835, Univ. Calif. Publ. Geol. Sci,, 29, 21-46, 1952.
Williams,S.N., The October1902eruptionof SantaMaria
volcano,Guatemala, MS Thesis,Dartmouth College,
Hanover,N.H., 140 pp. (unpubl.),1979.
Williams,S. N., andS. Self,The October1902plinian
eruptionof SantaMaria volcano,Guatemala,J. Volcanol.
Geotherm. Res., 16, 33-56, 1983.
Wilson,C. J. N., N. N. Ambraseys,J. Bradley,andG. P.
Understanding Climate Change Vol. 52
PALAIS AND SIGURDSSON 53
L. Walker, A new datefor theTaupoeruption,New
Zealand, Nature, 288, 252-253, 1980.
Wilson,C. J. N., andG. P. L W',dker,The Taupoeruption,
New Zealand,I. Generalaspects,Phil. Trans.R. Soc.
Lond., 1985.
Wolff, J.A., S. Self, andM.R. Rampino,Basalticfissure
eruptions,fire fountains,andatmosphericaerosols,
E.O.S., 65, 1148-1149.
Woods, D.C., R.L. Chuan and W. I. Rose, Science. 230,
170-172, 1984.
Zies, E.G., The Valley of Ten ThousandSmokes:I The
fumarolicincrustations andthebearingoredeposition. II
The acidgasescontributed to the seaduringvolcanic
activity,Nat. Geog.Soc.Contr.Tech.Pap..Katmai$•r,,
4, 1-79, 1929.
Section IV

BIOGEOCHEMICAL CYCLES,
LAND HYDROLOGY,
LAND SURFACE PROCESSES
AND
Geophysical Monograph Series Understanding Climate Change Vol. 52

UPTAKE BY THE ATLANTIC OCEAN OF EXCESS ATMOSPHERIC

CARBON DIOXIDE AND RADIOCARBON

Bert Bolin and Anders Bj6rkstr6m

Department
of Meteorology,
University
of Stockholm,
Sweden
Berrien Moore

Universityof New Hampshire,Durham,New Hampshire

Abstract. Inverse methods have been
deduce water circulation, spatial patterns of
turbulent exchange and biological activity in
the Atlantic Ocean, by using a set of stationary
tracers and a condition of quasi-geostrophic
flow. The solution yields a direct meridional
circulation cell with descending motion in the
northern Atlantic with an intensity of 20-25
Sverdrup, a reasonable distribution of vertical
turbulent transfer in the uppermost ocean layers
and cont•ratively large rates of detritus
formation,about4.5 PgC yr-1. - Thesolution
is used to compute the invasion of tritium
1955-1983, and the uptake of excess radiocarbon
and carbon dioxide during the period 1760-1983.
A fair agreementbetween•x•u. ted and observed
changes of tritium and -=C is obtained, but
the period of observations is too short to serve
as a conclusive test of the model. - The uptake
of carbon dioxide during the 220 years period
into the Atlantic Ocean is 33 .+. 5 Pg and it is
further found that significant variations of the
uptake fraction of the CO• emissions may have
occurred due to varying r•tes of emissions in
course of time. The conclusion is tentatively
drawn that the ocean and its carbonate system
may not have been the only sink for
anthropogenic emissions of carbon dioxide into
the atmosphere. - Means for how to further
improve the model and its capability to
reproduce the ocean behaviour are discussed.
Introduction
Burning of fossil fuels, deforestation and
changing land use have c2k3ngedthe global carbon
cycle very significantly during the last two
Copyright 1989 by
International Union of Geodesy and Geophysics
and American Geophysical Union.
used to centuries. We see this most clearly in the
steadily increasing 032 concentrationin the
atmosphere, from a most likely value of about
280 ppm in the middle of the 18th century to 344
ppm in 1983 (cf. Siegenthaler and Oeschger 1987;
Figure 1). This means an increase of the amount
of carbon in the form of 03• in the atmosphere
by 135Pg (Pg=1015g). z. .
The em•ssxons, due to
fossil fuel combustion, during this period of
time have been about 180 Pg C (cf. Rotty, 1987),
while estimates of the emissions due to man's
interference with the terrestrial biota
(vegetation and soils) are much more uncertain
and range between 100-200 Pg C (cf. Bolin,
1986). The airborne fraction during this period
thus has been between 36-48% (cf. Table 1). If
we consider the two periods 1760-1957 and
1958-1985 separately we find that we are not
able to tell if there has been a significant
c2k3ngeof the airborne fraction during this
period of time (cf. Table 1).
It has generally been accepted that a major
part of the emissionsof 032 into the atmosphere
has been absorbed by the world oceans.
Insufficient measuremen• are available that can
quantitatively verify this view. It is rather
primarily being based on simple models of the
circulation of carbon in the sea (cf. Bolin,
1986). The most commonly used model is the one
developed by Siegenthaler (1983), the so called
box-diffusion model with polar outcrops. A
simulation of the fossil fuel uptake since 1860
yields an airborne fraction of 60%. Siegenthaler
(1983) argues, however, that the exchange
between the ocean surface layers and the deep
sea should be enhanced as c• with what is
being obtained by determining the model
t• using steady state distribution of
modified model yields a value for the
airborne fraction slightly above 50%. It is,
however, questionable if such a modification of
the box-diffusion model is justified. We need

57
Geophysical Monograph Series Understanding Climate Change Vol. 52

58 UPTAKE BY THE ATLANTIC OCEAN

350 - I I I I I I I I i I i I' I

330-
z

r¾" 310 -
I--
z
ILl --

Z
o
290 -
o

o o

270 , , , , , , , , , , , , ,
1720 1760 1800 1840 1880 1920 1960 2000
YEAR
Fig. 1. AtmosphericpCO
2 1760-1980as determinedby analysis of air bubbles in glacier ice
and since 1957 by direc• measuremen• (Cc•piled by Siegenthaler ar• Oeschger, 1987).

rather more detailed a•d realistic models to be physical oceanographers and about chemical and
able to determine more precisely the uptake biological processes that are of importance for
capability of the oceans. For this purpose we understanding the role of the oceans in the
should combine our knc•ledge about the global carbon cycle as is being studied by
circulation of the • as developed by the chemical c•eanc•3ra••. If using General Ocean

TABLE1. Estimated Changesof the Atmospheric Componentof the Global

Carbon Cycle for the Period 1760-1957 and 1958-1983.

1760-1957 1958-1983 1760-1983

(:L) Increase of CO2 ppm 34 30 64

Increase of (/32, PgC 72 64 136

(2) Fossil fuel emissions, PgC 80 102 182
(3) Terrestrial emissions, PgC 80 20 100
(lower bound)
Airborne fraction (upper bound) 0.45 0.52 0.48

(3) Terrestrial emissions, PgC 130 70 200

(upper bound)
Airborne fraction (lower bound) 0.34 0.37 0.36

(1) Measurementsas reported by Keeling and colleagues (cf. Bolin, 1986) and
Siegenthaler and Oeschger (1987).
(2) Rotty (1981); Rotty and Masters (1985).
(3) According to Bolin (1986). The lower estimate (100 PgC) implies very small
emissionsbefore1860, as direct assessmentsof releases since 1860 hardly are less
than 100 PgC.Thepartitioning of the emissions to the two periods is approximate.
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL. 59

Circulation Models it is most important that the distributions of temperaS, salinity, total
rates of ocean overturning be carefully ssolved inorganic carbon (DIC), alkalinity,
validated by using tracer distributions, since "C, oxygen and phosphorus. Note that we do not
direct current • are hardly adequate consider the possible role of the formation of
for this purpose. It is obvious that dissolved organic carbon in the decomposition
determination of gaseous uptake in which we are process of organic matter. The following
interested is crucially dependent on careful analysis is still of interest as a study of what
validation of model characteristics in this the classical set of processes imply with regard
regard. Maier-Reimer and Hasselmann (1987) have to transfer patterns of carbon in the Atlantic
recently presented studies of the circulation of Ocean.
carbonin the sea (both12Cand14C)by using We divide our domain into eight
general ocean circulation models. It is, quasi-isopycnic layers (Figure 2) and twelve
however, obvious that a more detailed regions (Figure 3), based on our qualitative
discussion of their results is •nee•ed in order knowledge about the key features of water
to ascertain that the response characteristics circulation. Altogether 84 water reservoirs,
of the modelwith re• to •02 uptakeagree boxes, are defined in this way. An unknown
with those of the real ocean. vector x is specified having the 536 components:
In the present paper we shall attempt an
- advective fluxes of water between all
analysis of oceanuptake of (I) 2 using an inverse
methodology (cf. Bolin et al. I987) in which adjacent boxes (184 components)
case we shall employ the continuity equations
for a set of key tr•• (salinity, dissolved - rates of turbulent exchange of matter
inorganic carbon, radiocarbon, alkalinity, between all adjacent boxes (184 ccmpo•ts)
phosphorus, and oxygen). It is, however, also
important to make use of the extensive production or decomposition of organic
hydrographic data that have been collected for matter, one ccmpo•t for each box (84
many years. We accordingly also impose a c•mponents)
constraint of q•asigeostrophic flow. Our aim is
to develop a multiplebox model of the Atlantic - carbonate production or dissolution, one
Ocean that is carefully calibrated with the data component for each box (84 ccmpo•ts).
referred to and that has sufficient spatial
resolution so that the •02 uptakeduring the Assuming that the steady state concentration of
last few hundred years can be computed with the seven tracers can be determined for the 84
reasonable accuracy. boxes we are able to formulate 8 x 84 = 672
The present analysis will be restricted to equations, one for each of the seven tracers and
the Atlantic ocean because of data limitations each box and one additional set of 84 equations
and generalizations of our results to the world expressing water continuity for each box. We
oceans are not yet possible. A few general have adopted the finite difference approximation
observations will however be made questioning that advection between two boxes carries the
the view that the world oceans represent the mean concentration of these in the direction of
only significant sink for excess atmospheric flow and that turbulent transfers matter in the
carbon. We shall also learn that the emplo•t direction of the gradient between them. It is
of inverse methods to develop boxmodels requires further assumed that the change of the
a careful and detailed analysis in order to show horizontal flow as a function of depth
that the results are reliable. Our conclusions approximately agrees with the thermal wind
will thus not be final and firm, but will computed from hydrographic data, which yields
contribute to the fundamental problems we another set of 80 equations. This total set of
encounter when trying to assess the role of the 672 + 80 = 752 equations can be written in the
oceans in the global carbon cycle. form of a matrix equation (for details, see
Bolin et al., 1987)
Model Development
Ax = b (1)
Derivation of a Basic Steady State Pattern of
Circulation and Biochemical Processes where A is a 752 x 536 matrix, determined by the
concentration distributions in the sea, b
Bolin et al. (1987) have presented a general quantifies the water or tracer flux across
method for derivation of the steady state external bo•ies and radioactive decay (in
circulation pattern and biochemical processes by the case of ' •=-),
C and the prescribed vertical
using inverse methods. We ask the question: What change of horizontal advective flow from one box
steady state patterns of water circulation, to another in the equations for
turbulent transfer, new primary production, q•asi-geostrophic flow. We determine the air sea
decomposition of organic detrital matter and exchange by assuming that there is no net flow
dissolution of biogenic carbonates are required of •02 to or from the A•antic Oceanas a whole
to explain the observed q•si-steady and t/%at the inflow of •=C is balanced by
Geophysical Monograph Series Understanding Climate Change Vol. 52

60 UPTAKE BY THE ATLANTIC OCEAN

O O O O O O O O
O O O O O O O
•D O O O O O O
I I I / I

o o o o o o o
•D o o o o o o
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL 61

lO 11

12

Fig. 3. Thedivisionof the Atlantic Oceaninto 12 regionsfor the definition of the

84-box model (cf. Bolin et al., 1987).

radioactive decay within the basin. In reality
there maybe a net through flow which implies an
error in the way we determine rates of air-sea
exchange.Such an error will affect the computed
uptake
of excess
carbon
and14Cin the
atmosphere.In addition the following inequality
- turbulent flux of matter is always in the
direction of the concentration gradient
primary production only occurs in sunlit
surface boxes and deccaposition and
dissolution of detrital matter take place
Since the model is approximate, equation system
(1) is incompatible and a solution for the
vector x is obtained by minimizir• the norm in
the least square sense. The result of such a
procedureis of course dependenton the relative
weighting of the equations in (1). Bolin et al.
(1987) put equal weight on all tracer equations
by normalizing the equations, but up-weighted
the geostrophic equations in order to fulfill
the geostrophic condition to about 15%. It is
clear that the gross features of the circulation
and biochemical processes were reasonably well
reproducedas we knowthem from other studies,
but resolution as well as specification of the
key processes is not adequate to yield features
such as the boundary currents or the coastal
maximaof primary production. This is obviously
a short-com/ng of the model, that has been
at discussed by Bolin et al. (1987).
Schlitzer (1988) in a similar attempt to
inverse methods to deduce the role of ocean
circulation and biological processes for the
qp,mmi-steady
distribution of tracers applied a
more detailed and probably also more accurate
methodof analysis. It is, however, not possible
in that way to arrive at a closed set of
Geophysical Monograph Series Understanding Climate Change Vol. 52

62 UPTAKE BY THE ATLANTIC OCEAN

equations (1), which can be extended to a set of in the atmosph• and the chang• amount of DIC
time dependent equations (1), that in turn can in the surface boxes caused by influx fr•m the
be used for study of transient changes, which is atmosph• •nd will accord•ly vary with time
the prime objective of the present analysis. Similarlyb" in theradiocarbon
equation
will •
Derivation of the Time Dependent Model
••Cf• due
the to thetransfer
atmosphere, whileof
b•b•_produced
• will be given
as a function of time since return flow from the
We shall use the model and the solution x sea to the atmosphere can be neglected.
In order not to introduce artificial sources
derived in the previoussection•o determinehow and sinks in the course of deriving a tz-mmsient
excess concentrations of •0•, •=C and tritium solution we must d• that the solution x is
in the atmo•ere penetrate-into the ocean
assuming that the water motions and biological derived with the condition that the equations
for the particular tracer (n) to be considered
processes,i.e. the vector x, •main unchanged. be well satisfied in steady state, i.e.
•ne injections of tritium and •=C since 1954 are
tknrti•arly interesting, because there are sc•ne
data that shc• the invasion of these tracers Bn •= An x = bn (5)
into and within the Atlantic Ocean. If we are
able to ascertain that the model is able to This condition was not forcefully imposed for
any particular tracer in deriving the solutions
describe these processes adequately, we may also
given by Bolin et al. (1987), but the errors in
deduce with plausible reliability the likely
the incompatible set of equations (1) were
uptake of carbon dioxide, although direct
observations for verification are not available. rather distributed equally between all tracer
equations. We now rather derive a solution x,
It should be kept in mind, however, that the
transient data for the last 20-30 years do not d•ing that the •ts of equations
validate features of the solution that are referring to DIC and •=C be well satisfied i.e.
we u•wei•t them markedly. To the extent small
importantfor studyingthe •02 uptakeduring the errors remain, these will be added to the right
last centuries. We therefore implicitly still
will be relying on the adequacy of the hand side of equation (4) when integrating with
quasi-steady tracer distributions to derive the time making the implicit assumption that they
solution x, when deducing the response of the are due to processes that have not been
accounted for by the model and that these remain
systemfor the time scales r•evant for unchanged, i.e.
redistribution of •02 and •C in the Atlantic
Ocean during the last 100-200 years.
The time dependent form of equation (1) can be dqn/dt
= _ • qn+ bn +•1 (6)
written (Bolin et al. 1983)
whereBn•is the matrixfor the solutionx1 and
• is theXerrorfield in the set of equations
(5)
dq/dt = - Ax + b (2) for the tracer n. This procedure is of course
acceptable only if the solution x1 is reasonably
similar to the one previously derived with equal
weighting for all tracers.
It is useful to rather transform this equation The transient solution for tritium can be
into the equivalent form
obtained simply, s• the water continuity
equation is exactly satisfied and accordingly no
(3)
dq/dt =-Bq + b artificial sources and sinks are introduced.
It should be remarked that the finite
difference formulation of the advective terms as
where B is the matrix of coefficients that
used may imply a potential numerical instability
describes the way q will change and is defined when employing the equation (4) for tz-mmsient
by x. Equation (3) is a block diagonal matrix of cc•putations. The presence of turbulence, on the
the sets of tracer equations for the n tracers other hand, implies a stabilizing effect.
Whether such an instm_bility exists or not can
dqn/dt = -Bnq
n + bn (4) easily be determined from the signs of the
eigenvalues to the equation (4) and has only
appeared in a few experiments and not in any of
whereqn is the 84-component
vector of those described below.
concentrations of tracer n (tritium, Tr, carbon, The fact that the distributions of DIC and
C, or radiocarbon, *); Bnis the (84 x 84) matrix 14Chavebeenchang• duringthe last few
which represents the subset of equations for hundred years implies that we do not have
tracer n andbn correspon•ngly
the flux fromor accurate observations of steady state
to the exterior and, for •=C, radioactive decay. distributions of these tracers as required for
Note that Bn will not vary with time, while the the derivation of the vector x as outlined
vector bc for the DIC equationsis dependenton above. We assess, howler, that the changeshave
the changesof the partial pressureof •02, p, been rather small ( •=C data are available from
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET ALw 63

the time before major injections occttrred i.e. we use the observed concentration at
because of bomb-testing, see below) and we the time of G•0S•2S minus a correction that
therefore adopt the following iterative is •m] to the estimated • between
procedure to derive a proper steady state
soluti•9, x, andthe transient responseof DIC
and •=C due to the perturbations •ced by Thevalue Pc (to)' correspor•ting
to a
man.
state of no bet flux of 00•. between the
We introduce the notations atmosphere and the sea, is-deduced for the
t = tn = 1760, preindustrial time, before field q% (to) thusderived.
whichVanthropogenic influences are assumed to
have been negliable b) Similarlythe change
of A 14Cin the
sea during the period t o to t 1 is
t = t. = 1957, bomb testing had. Qot yet obtained and an estimated pre-industrial
significantly influ• the itc value
distribution
q*e
foimula
(to) is derived
usingthe
t = t•. = 1973, time for the Gq•OSE
obserPations
oftritium
DIC
and •4Cin
the Atlantic Ocean
qe(to) = qo(tl) - (qc(tl) - qo(tl) ) (9)
t = t• = 1983, the time to which transient
computations
wereextended
Wefurtherusethe notations
qno,qnc andqne 4. Thetracerdistributions
qCe (to) andq*e
(t_) together with the other sets of ti•cer
for observed, computed and estimated distributions can be used to derive a second
concentrations. The following iterative
steady state solution x''
procedure is then applied: The procedure described above can of course
1. A preliminary steady state solution x' is be repeated but the two solutions x' and x'' as
d .cedbyusingthe datasets qC
q o(t])
o (t2) and
as wasdoneby Bolin et aI. (1987)
well
that as Pe(to)
further and Po(•o
iterations o)'
notarethat
seem similar
necessary.
although they do not represent a true steady The solutions x' or x'' are obtained by
state. It should be noted that the atmospheric assuming that a steady state prevails. Bolin et
al. (1987) ascertained that no net flux of any
00•.associatedconcentration
Pc (t2) is computed tracer occurs to or from the domain by assigning
in-such a manner that no net aLr-sea exchar•e of
proper boundary conditions for exchange with the
002 betweenthe atmosphere and sea takes place. Pacific and Indian Oceans. In course of a
Pc (t2) differs little from Po (t2) ' transient computation this balance will be
2. Use the solution x' to derive transient disturbed because of chang• tracer
concentrations in those reservoirs from which
solutions during the time period t_ to t• there is outflow of water. In order to assess in
assuming
in this first iteration t•at th• the best possible way the role of the Atlantic
prescribed atmospheric carbon dioxide is given
Ocean for uptake from the atmosphere it seems
reasonable to change the tracer concentrations
Pc(t) = Pc (t2)+ (Po (t) -Do (to)) (7) of the inflowing water in such a manner that the
increased tracer outflow is balanced. It should

using data for Po (t) as given by be recognized, however, that the Atlantic Ocean
Siegenthaler and Oeschger (1987). may well be a pathway for uptake of tracers with
excess atmospheric concentrations and that in
b) •114C
for atmospheric
002,is reality a net flow from the Atlantic Ocean to
adjacent oceans may take place. It is of course
prescribed in accordance with •ata on the
Suess effect during the time period not possible to address such a question with a
t o < t < t 1 and as due to release of treatment of only the Atlantic Ocean as in the
bomb-
produced
14Cfor t 1 < t < t 2 present case.
(see further below).
Data on Tracer Distributions and Their Change
3.a) Thecomputed
valuesof qC
c (t)
during the period t o < t < t 2 are used We shall use the data given by Bolin et al.
to derive a most plausible distribution of (1987) for salinity, alkalinity, phosphorus and
DISat the timet = to, qC
e (to) , oxygen and assume that no changes have occurred
using the formula during the time period considered. We describe
C C C C belowb¾•efly the data for atmosphericpCO2,
qe(to)= qo(t2)- (qe(t2) - qo(t2)) (8) DIC, •=C and tritium that are used.
Geophysical Monograph Series Understanding Climate Change Vol. 52

64 UPTAKE BY THE ATLANTIC OCEAN

Atmospheric
002 andDissolvedInorganicCarbon atmospheresince the mid• o•.the 18th century
in the Sea - led to a decreaseof the •=•/•ZC isotoperatio
in our environment, i.e. /% •C decreased.
The cbar•es of atmosphericCO
2 havebeen Stuiver and Quay.(1981) concludethat
summarized by Siegenthaler and •er (1987) atmospheric • •C decreasedby about 22%•from
ar• their data are shown in Figure 1. We notice an approximately constant value before the
that an incr•__se by about 10 ppm occurred before middle of last century to 1955, when bomb
significant emissionsof .(392due to fossil fuel testing began, cf Figure 4. Druffel and Linick
ccaubustion began in the midc]le of last century. (1978) and Druffel and Suess (1983) 'have shown
The distribution of DIC in the Atlantic Ocean by the analysis of growthr'•g dated corals from
as determined during the G•0SECS program in 1973 the Florida Straits that Zl •=C for ocean surface
suppleted by TIO data frc• 1981 in the water in this region decreased by about 12%.
northat parts of the sea have been used in during this •me ñod.
the •tations to be described (cf Bolin et Bomb-produced
•C injections
intothe
at. 1987). atm•ere began in 1952 but were hardly
significant before 1955. Due to t• major
testing activities 1958-1962 zl '•C for
Radiocarbon,14C atmospheric CO (lower troposphere) almost
doubled,
but• sincethendecreased
to a value
Broecker et at. (1960) have reported about of about 300 ø/oo abovethe pre-bombvalue in
200 -- ts in the Atlantic Ocean during 1980 (cf Figure 5; Bolin, 1986). A transfer into
1957. These data ar• G•DSECS data frcau the the oceanhas also occurred.as is clear from
deeper parts of the ooean have been used to direct observations.A A •4C concentrationof
db•k•ca most plausible pre-bomb about 100 ø/oo wasreachedaround1970 and a
distribution (cf Broecker and Peng, slight decrease seems to have ocoarred since
1982; Bol• et at., 1987). then. The increase of • 14C until 1973 within
The • •C distribution had, however, already the Atlantic Ocean as observed during the
been disturbe•.by humanaction at that time. The G•0SECS expeditions has been evaluated for the
emissions of •C-free carbon dioxide into the 84 boxes and is given in Table 2.

10 r I I I I I i i i i I

l]l] ,, 1[]

-10

-- _

-20

-3 0 I , I I i ! i I I i I

1820 1850 1900 1950

Fig. 4. C•anges
of atmo•ic /%14C1820-1955
as deduced
fromanalysisof woodsamples.
(stuiver ar• Quay, 1981).
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL 65

120

100

80

ß ,
ß

; ß . ß

. ß.
ß%.•' o o ' , ß z,c::.
oN....,....,......
ß

20 • ; o o o %o _
øøø.'.,, ,, o ,,o

= *
.., c
•'
½' 8 a
i o oo
o c ooo
o
o o
o oc• ½
ao o
,•o
ß'" '" _.
1960 1965 1970 1975 1980

Fig. 5. Changes
of Z•14Cin atmospheric
002(filled dots),andin dissolved
inorganic
carbon in ocean surface water (circles), 1954-1980 (Nydal and L6vseth, 1983).

Tritium below, but several others have been performed,

to which •reference will occasionally be made.
Observed changes in the atmosph• have been I Weightingof the equations for • and qW
sum•mrized by Weiss and Roether (1980). (Some u•re assumed to be 50 and of the geostrophic
misprints in tables of that article have been equations 4; rate of air-sea exchange the
corrected after direct personal communication sable everywhere.
with Roether). Penetration of tritium into the II Weighting
of •, q* andgeostrophic
sea, particularly the Atlantic Ocean has been equations 50; air sea exc•e the same
described by '•stlund et al. (1976, 1987) and the ever2• o
tritium penetration depth in the world oceans in III Weighting
of •, q* andgeostrophic
1973 has been determined (Broeck•_r and Peng, equations 50; air-sea exc•e increased by
1982). We have used the GEOSE• data from 1973 70% in re•ions 1,2, 3 and 12 and reduced
to estimate the distribution in the 84 boxes appropriately in. qther regions to maintain
that are defined by our model (see Table 3). We balanoe between C inflow from the
note that in comparison with concentrations atmos•ere and internal radioactive decay.
c•served in the sea 1957 w• may safely assume There are reas• to believe that air-sea
that concentrationsbefo• bombtesting began exc•e is • at higher latitudes as
were negligiable, i.e. q• (t <
boundary
fluxes
o•t•.itium
from
• = 0.The
at•ere,
which determine b *• in equation (2) (in addition
latitudes
exper•
because of str•r•3er winds. The
is intended to test the
to radioactive decay) have been deduced by Weiss sensitivity of the results caused by
and Roether (1980). Their method is to compute uncertainty due to prescribing a constant
an input rate for each latitude and year, rate factor.
essentially as the product of a time factor A few comu•nts are of interest:
with a latitude factor.
- The average rate of air-sea ex_ch•e 9f 002
deducedfor I a•d I• w•_ 20.5 mol m---'yr
-• and-
for III 18.3 mol m-"yr TM based on the initial
data used. The corresponding values for the
estimated (%760•air-sea exc2k•e rates were
1437_m•lm-"yr-' for I a•d II and 13.2 mol
The discussion will primarily be based on m-'yr TM for III. The former values are of course
tltree expert, which are briefly described not correct because the assumption of a steady
Geophysical Monograph Series Understanding Climate Change Vol. 52

66 UPTAKE BY THE ATLANTIC OCEAN

TABLE
2. Changes
of 14C(in o/oo) in Ocean
Boxes1957-1973
Basedon DataFrom
Broecker et al. (1960) and GqHDSECS Data According to '0stlund et at. (1976) and
'0stlund et al. (1987).
Region Western Basin
Layer 1 2 4 6 8 10 12

1 +210 +160 +175

2 +165 +70 +130 +145
3 +100 +65 +15 +25 +75
4 +95 +50 +10 0 0 +10 +70
5 +25 35 +5 0 0 0 +5
6 35 +10 0 0 0 0
7 20 +5 0 0 0 0
8 20 +5 0 0 0 0

Eastern Basin

1 3 5 7 9 11 12

1 +220 +140 +120

2 +190 +40 +115 +135
3 +105 +175 +10 +20 +105
4 +95 +95 +10 0 0 +70
5 +25 +55 +10 0 0 +5
6 +30 +5 0 0 0
7 +5 0 0 0 0
8 0 0 0 0 0

state is not a correct one. There is in reality fossil fuel_•cm•_•stion.An air-sea exchangerate
a net inflc• of CO
2 into the ocean,while an of 14 motm "yr-• therefore see•_ moye
outflowof 14C•_ to compensate
for the appropriate than about 19 mot m-"yr -• as is the
decrease of Zl • •c in the atmosphere because of outcome of the first inversion. There is of

TABLE 3. Concentrations of Tritium in 1973. Box Averages Based on GEOSECSData (TU).

(cf. '0stlund, et at., 1987).
Western basin

Region 1 2 4 6 8 10 12
Layer
1 6.8 2.6 2.2
2 6.5 1.8 1.3 1.5
3 17.5 3.8 0.2 0.2 0.6
4 11.0 11.3 1.0 0.1 0.1 0.2 0.7
5 3.7 7.3 0.8 0.0 0.0 0.3
6 3.9 0.6 0.0 0.1 0.1
7 1.5 0.1 0.1 0.1 0.0
8 2.9 0.2 0.1 0.0 0.0

Eastern basin

Region 1 3 5 7 9 11 12
Layer
1
2 7.2 1.6
3 7.5 5.0 0.6
4 11.0 6.9 0.7 0.0 0.7
5 3.7 4.1 0.3 0.0 0.3
6 2.1 0.2 0.0 0.1
7 0.1 0.0 0.0
8 0.0 0.0 0.0
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL. 67

coursestill c9•siderableuncertainty aboutthis correlation between the solutions in experLment

value since z• •=C fo•.ooean surface waters I and II is about 0.7 and the same is true for
before bombproduced •4C wasemitt• into the anyone of these and the "base case" (Bolin et
atmo•ere were not well determined. The value al., 1987), where all tracer equations had
is co id ly less than the value around 18 weight one, and the geostr•y equations had
mol
•-•-e•abas
given
byBroecker
and
Peng weight four. This is a genuine uncertainty due to
(1982). The discrepancymaydepe• on the inccmlPlete or not representative data sets,
assumption of no net outflow of •=C to adjacent coarse resolution, sensitivity of small scale
oceans. features of the solution to errors in the data,
The assumptionof a steady state distribution as well as inccmlplete inclusion of the relevant
of DIC and •=C in 1760 is not necessarily processes that affect the tracer distributions
correct although more appropriate than in 1955. (this matter will be further discussed below).
Figure 1 shc•s that atmo•eric (/32 We find that the intensity of advective flow
conoen•tions rose by about 10 ppm between 1760 patterns of inversions I and II are about 70%
and 1860, although the emissions due to fossil and 110% respectively of that of the "base case"
fuel cc•bustion were negligiable before the present• by Bolin et al. (1987). The
ddle of the 19th century. Figure 4 shc•s that similarities between the solutions, particularly
C did not either start to decrease until about between II and the "base case", justify that we
1850. The incr•k3se of atmo•eric pO0. must then pursue the analysis, but we shall have to
either be dueto a net input of •J into the explore the uncertainties of our final results
atmosphere due to deforestation and changing due to the uncertainties in the solution x
land use or to an unbalance associat• with
natural oscillations in the carbon cycle, The transient solutions primarily de•nd on
possibly related to natural climatic variations, the advective and turbulent terms, the •=C
e.g. the little ice-age, during the preceding equations to some degree also on the flux of
centuries. We shall also conduct a few detrital matter. The meridional circulation with
sinking motion in the north Atlantic is of
chara•istics of the Atlantic Ocean of particular importance for the rate of transfer
interest in this context. of matter injected into surface boxes to deeper
layers. A comparison of the solutions I and II
Features of the Steady State Solution x . with geostrophic weighting of 4 and 50
respectively shc•s that the former is quite
A few statistical analy• are first of non-geostr•c because of the comparatively
interest. The correlations b•tween the first and larg•Aweights for the tracer equations for DIC
seconditerative solutions x and,¾'' and and •=C. The meridional circulation is also
between the different solutions x are given in cc•m•ratively weak. Solution II, on the other
Table 4. hand, approximates the geostrophic constraints
We notice that the former are all above 0.9, much better and is characterized by a meridional
i.e. the departure of the syste•.from steady circulation with a return flow southward in the
state before injection of bombl•C does not North Atlantic below 1000 m of about 22 Sv, of
critically influence the solution x. Wes•911 which 8 Sverdrup is below 2500 m (see Figure 6).
still use the more appropriate solution x This is similar to the base case presented by
A change
of theweighting,
ontheother•3nd, Bolin et al. (1987). The ocean circulations
influences the solution significantly. The derived by Maier-Reimer and Hasselmann (1987),

! ! !

TABLE4. Correlations BetweenFirst and SecondIteration Solutions x and x for i

= I, II, III and Between Second Iteration Solutions x and x for the Experiments
i, j = I, II, III. Correlations for Subsets of Componentsof theVectors xi and•,
1.e. Advective, Turbulent and Organic Detritus ComponentsAre Also ShOwn.

Total x Adv. Turb. Organic

components components detritus
components
! !!
0.92 0.94 0.90 0.97
x,i x ,I
0.97 0.99 1.00 0.97
x,i I x ,II
0.98 0.98 0.95 0.98
x III x III
';

xi,I X,,II 0.73 0.69 0.78 0.94

0.71 0.68 0.76 0.93
X,, I X ,III
0.99 1.00 0.98 1.00
X II X III
Geophysical Monograph Series Understanding Climate Change Vol. 52

68 UPTAKE BY THE ATLANTIC OCEAN

DEPTH
m

1000

2000

3000

/,000

5000

60"5 30"5 Equ. 30øN 60"N

Fig.6. Meridional
watercitation in theAtlanticOcean_•s
determined
fromx'' for
experiment II. Numbers on streamlines: flc• in 10• ton yr •( 3.2 Sverdrup). The
analysis has not been extended southward of the boundary between boxes 8/9 ar• 10/11 (cf.
Figure 3). The sink at latitude 10øNar• the source at latitude 30øNdue to the cur•
through•
theCaribbean
Sea
ar•Mexican
Gulf
areshown
asshaded
regions
inunits
of1'0
•
ton yr -•. Note the expandedvertical scale abovea depth of 1000 m.

SouthernHemispherepenetrates to about 10øN.

on theother
ciroulatior• ha.,•,shows
ar• acco•ly si•fic•Uy
also less {•• 3. Upwelling takes place preferentially in the
concentrations in the deep sea. In the light of eastern basin ar• downwelling in the western
these comparisons we shall use the solution II one.

ar• the associat_•4_ transient solution as our 4. The northward flow in the surface and
reference case.
intermediatelayers of the westernbasin outside
The follc•ir• features of the reference the no_r• Americancontinent is about 0.7 10•
solution should be noticed (Figure 6): tons yr • ( 23 Sverdrup) and is obviously much
1. The mean meridional circulation with sinking weaker than the Gulf Stream. In reality some
motSon
in regions
1-3, is about0.701015ton recirculation of Gulf Stream water takes place
yr-• ( 23 Sverdrup) in generalagreem• with to the west of the mid-Atlantic ridge ar• can
the contain view. A weak cell ( 0.10 10'• ton partly explain the discrepancy. Sc•e of the
yr-1) in the oppositedirection is foundbetween tracer transfer by this major coastal current
the equator and 30øNbelow about 2.500 m. (as well as others elsewhere) is therefore
2. TWo meridional cells in the thermocline implicitly aco3unted for by the horizontal
region with u•ling in equatorial regio• and
poleward flow in surface layers are superim• 5. The turbulent components are generally
smaller than the advective ones, but do
described in the previous paragraph. The contribute to the solution. Vertical turbulent
northward flow of •ate water in the transfer plays a significant role only between
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET ALA 69

the uppermost boxes and those immediately below as revealed by acomulation of tracer material
thus accounting for exchange between the surface in course of transient invasions from e.g. the
mixed layer ar• the uppermost part of the atmosphere are remarkably similar for different
x solutions. The solution II is, however,
6. Large horizontal turbulent components (along plausibly one that overstimates the rate of
isopycnic surfaces) appear spotwise in our tracer transfer within our ocean domain.
solutions. These are most likely due to the fact
that tracer gradients often are comparatively Invasion of Tritium into the Atlantic Ocean
small along isopycnic surfaces leaving the rate 1955-1973
of turbulent exchange rather indeterminate.
Accordingly they only contribute marginally to Transient computations to determi/• the
the errors. It should be remarked, however, that uptake of bc•b produced tritium during the
this would not neoessarily be so, when computing period 1955-1973 w•re done using the transfer
tracer invasion into the sea that perferentially vectors x'' from exper•ts I and II. The total
takes place along isopycnic surfaces. For this computed uptake tnlrned out to be 25-30% larger
reason the rate of uptake that we will derive
later may be an overestimate. GqDOS•CS cruises in 1973, the difference between
7. The detritus flux s•tling c•_• of the the two experiments being due to somewhat
surfa_c•b•_•esis 4.7 10• g C yr • ( 50 different flc•s to adjacent oceans. Sarmiento
g C m-'yr •) which seems large as comparedwith (1983) found a corresponding discrepancy between
the basic case presented by Bolin et al., (1987) the given inflow acco• to weiss and Roether
and direct obeervations (cf Bx•echer and Peng, (1980) and observed storage of about 20%. The
1982). This may well be due to the comparatively correlations between c•served and compu•
large turbulent components derived (see above) acomulation into the model boxes where uptake
which transfer DIC upwards from the layers of has occurred are 0.91 and 0.88 for the two
high concentrations into the mixed layer, for transient runs I and II. Table 5 compares the
which a oompensat• detritus flux is derived. uptake by region. we note the unrealistically
8. The solution has been obtained without large values in region 1, and also seemingly too
taking use of the ten•pera• data (cf Bolin et large values in regions 6 and 8 (see further
al, 1987). It is, however, possible to compute a comments in section 4.5).
posteriori the heat transfer that the
steady-state solution implies. We find that the The Suess Effect
northwardtransfer (experimentII) from reg$•ns
4 and 5 into regions 2 and 3 is about • 10-• W The emissionsof 14C free carbondioxide due
compared to a range between 0 and 8 10 '= W as to fossil fuel bur•g have changedthe
assessed by Wunsch (1984). Our larger-•alue is distribution of A •C both in the atmosphere and
primarily caused by turbulent transfer due to the sea (the •es•. effect), already before
rather large horizontal turbulent components (cf injection of •C into the atmosphere due to bc•b
6 above). This implies that the •0 uptake that testing. Wehave usedthe data given • Stuiver
wecompute
below
(seesection
4.6)2may
alsobe and Quay (1981) for the changes of zl •=C in the
somewhat too large as the penetration in the atmo•ere since 1800 and assumed a constant
North Atlantic and horizontal transfer southward
is a majorpathwayfor (/)2 invasioninto the
TABLE
(in 5.
TU xU•10
• ton),
ofTritium 1955-1973 byRegion
Atlantic Ocean.
9. There are other details in the solution that Observed According to
deserve further study. It would obviously be of GHDSECSand Computed in Transient Exper•ts I
interest to relax the constraint due to the and II.

prescribed fixed -•alues of the b-vector that the

present method of the solution is based upon. Region Obs. I II
The use of the simplex linear programming method
1 2 14 26
for solution with given ranges of the components
2 45 53 36
of the b vector is an interesting alternative
3 35 37 25
(cf. Wunsch, 1984). Additional knowl•e about
the ocean circulation in terms of further 4 49 52 51
5 12 22 32
constraints could also be added to the given set
6 4 10 12
of inequality constraints (Bolin et al. 1987).
It is clear from the present solution that
8 5 10 7
the •se of inverse methods for deducing
simultaneously the circulation and biochemistry
10 5 7 6
of the ocean which is being considered, requires
11 3 5 5
a careful and detailed data analysis in relation
12 4 2 2
to the matrix inversic• that provides our
Total 164 212 202
solution. As w• shall see in the following
sectio• the /nteqral properties of the solution
Geophysical Monograph Series Understanding Climate Change Vol. 52

70 UPTAKE BY THE ATLANTIC OCEAN

TABLE 6. Decrease of 14C in the Atlantic OceanDue to Fossil Fuel EmissionsUntil

1955, Suess Effect, in o/oo.

Region 1 2 4 6 8 10 12
Layer
1 -8 -5 -7
2 -5 -4 -5 -5
3 -5 -4 -2 -2 -4
4 -9 -4 -1 0 -1 -2 -3
5 -8 -4 -1 0 -1 0 0
6 -6 -2 -1 0 0 0 0
7 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0

1 3 5 7 9 11 12

-5 -5 -7
-4 -4 -5 -7
-5 -3 -1 -1 -3
-9 -3 -2 0 0 -2 -3
-8 -2 0 0 0 0 0
-6 -1 0 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0

value before that time to determine the • 14C reference solution II are given in Table 7. We
changes in the Atlantic Ocean during the period
1760-1955 with the aid of the solution II. A The area average increase of fa '•C for all
slow drift of less than 1 ø/oo occurredbefore surface boxes is computed
to have been 132 ø/oo
fossil fuel emissions began (about 1850). The if using the reference solution II as compared
comput• Suesseffect was acco•ly derived as with an observedchangeof 138 ø/oo. Wenote
the Zl •=C change during the period 1850-1955.
The result is shc•n in Table 6. The changes in thattheair
assl/med
to beseaex
14.6_ch_a•ge
-••ate
Mm-Zyr ofCO
on2ah•Cbeen
based
the surface waters vary between-4 and -9 ø/oo, balance between preindustrial inflow from the
excluding the southernmost region for which the atmosphere and radioactive decay. As has already
results may be questioned because of influence been pointed out this is less than the most
from the approximate treatment of exchange with commonly
acceptedvalue of about18 Mm-2yr
-1.
adjacent seas. The average Suess effect in (Broecker and Peng, 1982). The solution with
surface water deducedfor 1955 is -5.5 ø/oo, enhanced
air sea exchangein pola¾.regions(III)
co•paredwith -22 ø/oo for atmosphericcarbon yields almost the samevalue for •qC uptake. In
dioxide. In region 4 a value of -8 ø/oo is the case of a comparatively slow ocean
deduced which is significantly less than the circulation (I) the surface /%14Cvalues
observedvalue of -12 ø/oo as reported by increasemorerapidly 163 ø/oo, presumably
Druffel and Suess (1983). The differences can because of less effective transfer away from the
hardly be explained by the deficiancies of the ocean surface. The increase of surface
solution II as discussed in the previous concentrations depend more on the rate of
section. The question is raised of how transfer into deeper layers of the sea than on
representative the measur•m•n• may be. the precise value for the rate of air-sea
exchange.
Uptakeof Bcmbp•ced 14C The A 14Cchanges
in individualsurface
boxes show a somewhat more irregular pattern
Since1955large amounts
of 14Chavebeen (Figure 7). The computed changes with time in
injected into the atmosphere and a transfer into surface box 4, layer 1, are compared with
the oceans has occurred. G•OS• observations observed changes as deduced by analysis of
reveal the increase and Table 2 show the corals in Florida Straits and Bermuda, Figure 8
estimated •hanges during the period 1955-1973. (Druffel and Suess, 1983). The agreement between
The computed cha• as determined with the values computed with the reference solution (II)
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL. 71

Changes of 14C(o/oo) in Oce•%

Boxes1955-1973
as Deduced
in Exper•t II.

Region
Layer 1 2 4 6 8 10 12

+219 +126 +137

+107 +97 +96 +118
+105 +71 +28 +58 +85
+233 +82 +9 +7 +33 +54 +74
+203 +70 +12 +5 +13 +14 +22
+71 +32 +4 +1 +1 0 0
+5 +6 0 0 0 0 0
+1 +5 0 0 0 0 0

1 3 5 7 9 11 12

1 +125 +132 +190

2 +87 +98 +106 +145
3 +118 +47 +22 +7 +45
4 +233 +64 +25 +2 +5 +46 +74
5 +203 +13 +7 +5 0 +7 +22
6 +71 +2 +1 0 0 0 0
7 +5 0 0 0 0 0 0
8 +1 0 0 0 0 0 0

and observations is excellent but probably penetration proceed somewhatmore quickly

somewhat fortuitous in the light of the vertically in the tropics than observations
irrec3ularities between regions mentioned above.
In case of a slower circulation (I) a - The uptake by regions 2-5 is only about 70%
substantially larger increase was obtained for of what observations show. Uncertainties of
the surfaceboxes,while the total uptakeof 14C the observed pre-bomb concentrations and
by the sea was significantly less (see further therefore of penetration could be the major
below), again •cating that the less efficient cause for this discrepancy.
transfer into the interior of the • lead to - The ratio of the uptake in the western
enhanced concentrations in the surface layer. basin to that in the eastern one, is
Thepenetrationofl4c into the • as considerably larger as deduced from the
computed (Table 7) exhibits both similarities computations (48.0/28.2) than observed
and discretkAncies compared with observations (46.6/38.7).
(Table 2). A short time elapsed between the - The total observed uptake is about 10%
major emissions into the atmosphere (1958-1961) larger than the one •ted which
and observations d•ring G•OSE• cruises in 1973. difference hardly is significant with
We cannot expect that the interior • regard taken to both data and model
circulation has brought the invasion very far uncertainties.
from the regions of entrance. There is, however, - A compari•n betweenmodelcomputationsand
marked differences between uptake by different observed •=C distribution in 1981/82
regions already after this short time, Table 8 according to TI• observations could aid us
shows the uptake inteqrated vertically for each further in our attempts to validate the
region. Lack of observations in some regions results from the transient computations.
(cf. Table 2) prevent us from a comparison of This will be done when a more careful study
the total uptake but those that can be made of the sensitivity of the inverse solution
between model computations and observations for to assumptions of weighting and the
boxes with data are of interest.
- A marked difference between uptake in (see further the concluding section of this
middle latitudes and the tropical belt is paper).
clearly seen in both observed and computed In spite of discrepancies between observed
fields. The model computations let the and computed chanqes the overall fields show
Geophysical Monograph Series Understanding Climate Change Vol. 52

72 UPTAKE BY THE ATLANTIC OCEAN

quite small but systematic. The correlation

between the changes is about 0.9. The surface
boxes, shown as circles, have systematically

i
iiI •-200 somewhat smaller concentrations in exper•t
than in I which implies a larger pOD
II
2 difference
between the atmospheric concentration and the
\ iI mean pOOpof the ocean surface boxes.
B•cozdir•ly the cumulative002 flux fr•m the
atmosphere into the Atlantic Ocean was larger in
exper•t II (35.9 P•3Cin 1983) than in I (30.2
PgC in 1983). This is due to the more intense
ocean circulation in II and accordingly more
rapid transfer of 002 into the deeperstrata of
the ocean. We notice systematically higher
concentrations in II in deeper layers than in
exper•t I. As a matter of fact the rate of
air sea exchange is of less importance for the
2 4 6 8 10 12
uptake than is the rate of ocezunoverturning as
Equ. long as the air sea exchange is rapid enough not
Western basin
to let oceansurface pCO
2 lag too muchbehind
A14C the increasir• atmosphericpO02concentrations.
Figure 10 shows the verticaI distribution of
the uptake in a series of profiles for selected
\ -200
re•ions. The computed changes until 1860, 1955
and 1983 are reproduced. A few particular
features of the penetration into the ocean
(Figure 10) are worth noticing, even though
careful improv•t
of the solution by using independent information
may modify some of these (see further next
section).
- Rapid penetration down to the bottc• of the
ocean occurs in region 2, while it does not
reach layer 8 in re•ion 3 because of upwellir•
from the deepest layer of the eastern basin (cf.
also a similar feature of the water flow
obta• by Schlitzer, 1987).
- The vertical distributions in re•ions 4 and 6
Equ. shows clearly the grad•m] invasion of carbon
Eastern basin from the north into intermediate and deep layers
due to the advective flow, while turu•ent
penetration from above is considerably less
(dashed
•ine) changes
of zI•=Cin surface significant.
boxes as computed with the aid of transient
- The penetration of 002 into the intermediate
solution II during the period 1957-1973, for the and deep waters is remarkably similar whenusing
western (a) and eastern (b) basins. the different vectors x that we have derived,
in spite of significant differences between
these, particularly with regard to the turbulent
sufficient similarities to justify an analysis components. Clearly the small scale features of
of the modeluptake of CO).during an extended the solutions are not of prime importance for
period of time using the •olution II. the in--ted_ uptake.
Figure 11 shows some chara•istic features
Uptakeof 002During_
the •eriod 1760-1983 of the transient solution. It should be
emphasized that the integrations start from a
The annual rate of 002 uptak• has been state of no net transfer of O0 between the
c•,•fced using the solutions I and II. In the atmosphere
andtheseai.e. th2e
arealmean
of
light of the analysis in the previous sections 002 partial pressurein the sea is assumed
to be
we shall adopt the latter as one reference case. eqSal to the atmo•eric pO0•. in 1760. Figure 11
Before discussin• the details of these results a shows how a difference be• these develop in
few other features of the transient solutions course of time and accorditaly the uptake
areof •o /ncreases with time.
Figure 9 is a scatter diagram showing the The transient computations imply that we
diff• of the uptake by the different assume
a net emissionof 002 into the atmo•ere
r•servoirs in the experiments I and II. They are due to man, causing an increase of the
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL• 73

A•C
ø/oo

2OO

100
GEOSEC5

1960 1970 1980

o i i i

/
/

Fig. 8. Computed
changes
of A 14Cin the surfaoelayer of region4 as deduced
with the
transient solutions I, II and III (solid lines) cc• with measured changes as
determined from analysis of corals in Florida Straits (dashed line; Druffel and Suess,
1983). The average value in 1973 as det•_•l•iz•ed during the GI•SE• project is also
•• (*).

atmosphericpCO• and a responseof the Atlantic of the world oceans would behave in an analogous
Ocean in terms Bf a net uptake. We can perhaps way as described by our model of the Atlantic
better visualize the characteristics of what Ocean. Not knowir• the total emissions we
happens by assuming hypothetically that the rest explore how cases in which we assume that the

TABLE
8. Uptakeof Excess
Atmospheric
14Cby RegionDuringthe Period1955-1973.
Observations are from the Gt•SECS Expeditions. •tations are According to the
Transient Experiments II and III.

Tota .inventory Inventory

Region
Obs.
10
«øatoms II III Obs.
10I•-atoms/m
2'
II III

1 1.0 2.8 3.7 0.22 0.62 0.83

2 5.8 5.1 5.6 1.20 1.07 1.16
3 7.2 4.1 4.1 1.56 0.88 0.89
4 16.9 11.8 9.0 1.59 1.11 0.85
5 11.5 5.1 5.0 1.57 0.70 0.68
6 2.8 3.5 3.3 0.43 0.54 0.50
7 2.9 4.6 3.9 0.30 0.46 0.39
8 6.9 9.0 8.4 0.86 1.11 1.05
9 4.1 3.9 3.2 0.63 0.60 0.49
10 4 .5 7 .2 7 .3 0.66 1.07 1.08
11 7.4 3.9 5.0 1.17 0.61 0.80
12 1.7 1.9 2.6 0.15 0.17 0.23

Total 72.7 62.9 61.1

Geophysical Monograph Series Understanding Climate Change Vol. 52

74 UPTAKE BY THE ATLANTIC OCEAN

40
o

•'• o
30-

• 20-

• x
• x
• x x
• 10-

: /x • •x
•XXX
O]
0x
X•x
• I
10 20 3•
DICchange
Experiment
•g. •. ••ec •• oE •• •••L•on •• oE

average air-borne fraction during the period time in cases A and B. During the first 120
1760-1983 were 0.50 (case A) and 0.45 (case B). years ,• decreases from an initial value of 100%
This implies for this period to 35-40% in 1870 at which time the atmospheric
co•trations began to increase more rapidly,
Total emission (Pg) 278 310 but slowed dc•n after the turn of the century.
Fossil fuel emissions (Pg) 179 179 Because of the accelerating increase s•
Biosphere emissions (Pg) 99 131 1950, ,• increased significantly during the
Ratio of the total uptake to last 35 years. Although we cannot trust the
the Atlantic Ocean 3.9 4.8 precise values as shc•n in figure 11, it seems
likely that this general feature of our model is
We note that the ratio of the total ocean one which might well also characterize the
surface to that of the Atlantic Ocean is 4.0 or world oceans (cf also Oeschger and Siegenthaler,
4.5 dependent on whether region 12 is included 1985). Because of the fact that irregular
or not. Acco•ly the uptake by the Atlantic variations in reality are superi•posed on such
Ocean per unit area is larger than that of the slow trends, we usually cannot discover these
rest of the oceans in case A and vice versa in until a decade or two have gone by.
case B. Although the calculations described above are
The curves a in figure 11 shows how the based on an analysis of the Atlantic Ocean it
airborne fraction ( • ) would have varied with seems plausible that the airborne fraction of
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL. 75

DIC pMol/kg
10 20 30
DEPTH
km

10 20 30 10 20 30

1
2

10 20 30 10 20 30

N& ION 5

10 20 30 10 20 30

REGION 6 REGION 7

Fig. 10. A•ation of carbon (in the form of dissolved inorganic carbon in the Atlantic
Ocean) as a result of increasing atmospheric 002 concentrations as shownin Figure 1.
Vertical profiles of concentrations are shown f•r selected regions for 1880, 1955 and 1983
as deduced with the aid of the transient solution II.

man'semissionsof 002 into the atm•ere
increased from 1950 t5 1973 and then decreased,
when the annual increase of 032 emissions by
4.5% per year due to fossil fu•l
1973 decreased to only about 1% per year, which
situation has prevailed until recently.
has 032 emissions to the atmospherebefore 1850
due t5 fossi% fuel burningwere small probably
0.05 PgC yr • or less and insignificant before
combustion in 1830. The total emissions until 1860 can hardly
have been more than 4 PgC (cf Rotty, 1981). The
increaseof the atmospheric
032 content
Geophysical Monograph Series Understanding Climate Change Vol. 52

76 UPTAKE BY THE ATLANTIC OCEAN

50
340

330

320

310

300

0.5

0.4

0.3
6

, 0
1800 1850 1900 1950 2000
Fig. 11. Transient exper•t II illustrating CO
2 uptake by the Atlantic Ocean.
a) Airbornefraction of 002 emissions(in %) de•]h•edby consideringthe response
characteristics of the Atlahtic Ocean being rep•tative of the world oceans as a whole
and assuming that the mean airborne fraction 1760-1983 was 50% (case A; dashed line), 45%
(case B, solid-line (cf text).
b) Atmospheric
002•partial pressureas determined
by Oes•er andSiegenthaler(1987), cf
Figure 1.
c) Atlantic • uptake(PgCyr-1).
d) 002 emissionsandtheir partitioning betw•_nemissionsdue_•ofossil fuel combustion
and deforestation and char•ing land use (bioemissions), PgC yr • as computed for the two
cases A (dashed line), B (solid line).

1760-1880 was about 20 PgC, and the uptake
the Atlantic Ocean about 6 PgC according to our
computations. We assess that the total oceans
will have taken up 20-25 I•. Thus, if the
equilibrium between the atmosphere and the sea
was not markedly disturbed already before 1760,
the total emission to the atmosphere during the
period 1760-1880 were above 35 PgC, of which
accordingly only about 10% was due to fossil
fuel c•bustion. Emissions due to
by and expandingagriculture mayaccoxd•! • on the
average have been about 0.3-0.4 PgCyr -x during
this period.
Figure 11d finally shows the emissions that
we deduce corresponding to a total airborne
fraction of 45% and 50% (solid and dashed lines
respectively). The values until 1950 seem
plausible yielding a total emission due to
deforestation and changing land use
deforestation (bioemissions) of 115 and 97 PgC respectively,
Geophysical Monograph Series Understanding Climate Change Vol. 52

BOLIN ET AL• 77

deduced as the difference between the computed coefficients in the transfer terms could be
total emissions and those due to fossil fuel given values within ranges as determined by
combustion (Rotty, 1981). During the last 25 the errors in the data field. Some of the
years, however, the bioemissions become quite tracer continuity equations in A x = b could
small. This result leads to a suspicion that rather be given as inequality constraints in
other sinks than the oceans may have been the set G x > h. The Simplex method of
present to absorb the larger bioemissions that solution offers another suc•,possibility.
seems plausible. 5. The steady state solution x has • used
Let us finally return to the discussion in to compute the invasion of tritium, •C and
the introduction as summarized in Table 1. We carbon from the atmosphere.
have found that about 35 PgC may have been 6. In spite of the •inty of the basic
emitted into the atmosphere before 1880 due to oft•fer solution x the integra• uptake
deforestation and changning land use. In the C during the years 1955-1973 is quite
light of the assessments by Bolin (1986) 100 PgC stable and agrees reasonably well with
seems minimum of these emissions for the period observations during GEOSECS.
1860-1980. Thus the total emissions from 7. Themoredetailedpatternof 14C-penetration
deforestation and charging land use for the as computed show some differences co•pared
total period is likely to have been well above with observations. A further analysis of
100 PgC, most likely 135 PgC. Total emissions, these descrepancies might yield a more
i.e. including those due to fossil fuel reliable solution. Generally great care must
emissions therefore probably have been larger be exercised in the use of available data
than 300 PgC, i.e. Case B is applicable rather and their uncertainties in order to improve
than A. On the other hand it does not seem sucessively the solutions of the basic
likely that the uptake per unit area by the matrix equation.
Atlantic Ocean is less than that of the rest of
8. The Suess effect in the Atlantic•
1955 being due to the emission of C-free
in
the ocean. Other sinks for ODp.maybe of
importance than those considered here. carbon dioxide by fossil fuel combustion in
the past is obtained with the aid of a
Su•nary of Results and Conclusions transient computation. The computed values,
the averageof which is 5.5 o/oo, are
1. Inverse methods (Bolin et al, 1987) have considerably less than the observed being of
been used to deduce ocean circulations, a the order of 10 o/oo. The discrepancycan
spatial pattern of turbulent exchange and hardly be due to inadequacies of the
biological activity that best satisfy the transient solution and the result warrants
steady state destribution of a set of ocean further analysis.
9. The total uptake of carbon dioxide by the
2. An analysis of a series of such solutions Atlantic Ocean during 1760-1983 is computed
and comparison with some independant data to have been 33+ 5 Pg C, of which about 6
show fair agreement for the large-scale I•g occurred before 1860.
pattern (cf Bolin et al, 1987) but also 10. Considerable temporal clmarges of the uptake
significant uncertainties on the smallest in terms of a percentage of the atmos[i%eric
scale of resolution particularly with regard increase (equivalent to a varying air-borne
to the pattern of turbulent flux. fraction) are found. It seems plausible that
3. A prominent feature of the most plausible the air-borne fraction of the emissions was
solution is clumracterized by a direct mean considerably grea• in the 1950' and 1960 ' s
meridional circulation cell w_•than than during the earlier part of this century
intensity of 0.7 10 • ton yr • and it may have decreased since 1973.
( 23 Sverdrup) with sinking motion in the
north and two coupled circulation cells in References
the termocline region with 0.7
1015ton yr-1 ( 23 Sver•), upwelling Bolin, B., 1986. Howmuch032 will remain in the
in the equatorial regions. Turbulent and atmosphere? In Bolin B., D66s, B. Warrick, R.
detrital transfer seem too high and implies and JRger, J. (Eds) The greenhouse effect,
probably somewhat too rapid tracer transfer climatic chanqe and ecosystems. S03PE Report
in transient computations. 29, pp 93-155. J. Wiley, C•ichester, England.
4. The errors in satisfying the set of tracer Bolin, B., Bjdrkstrdm, A., Ho•, K. and Moore,
equations are due to the incompatibility B., 1987. On inverse methods for combinirg
because of crude spatial resolution, chemi• and physical ooeanogra•ic data: A
approximate description of ch•mti• and steady-state analysis of the Atlantic Ocean.
biological processes and above all Report C•-71, Dept. of Meteorology, Univ. of
inaccurate data fields. It would be Stockholm, 220 p.
desirable to obtain solutions in which both Broecker, W.S., Gerard, R., Ewing, M. and
the b-vector (primarily determined by the Heezen, B.C., 1960. Natural radiocartx• in the
prescribed boundary conditions) and the Atlantic Ocean. J. Ge•ys. Res., 6__5,2903-2931.
Geophysical Monograph Series Understanding Climate Change Vol. 52

78 UPTAKE BY THE ATLANTIC OCEAN

Broecker, W.S. and Peng, T.H., 1982. Tracers in Rotty, R.M. and Masters, C.D. 1985. Carbon
the sea. Lamont-Doherty Geological dioxide from fossil fuel combustion: Trends,
Observatory, Columbia Univ. N.Y. resources, and technological implications. In:
Druffel, E. and Linick, T., 1978. Radian in Trabalka, J.R. (Ed.) Atmospheric carbon
annual coral rings of Florida. Geophys. Res. dioxide and the global carbon cycle. U.S.
Lett. 5, 913-916. Department of Energy, DOE/ER-0239, 63-80.
Druffel, E. and Suess, H., 1983. On the Sarmiento, J.L., 1983. A simulation of bomb
tritium entry into the Atlantic Ocean. J_•.
radiocarbon
record
inbanded
cry¬s:
Exchange Physical Oceanoqr. 13, 1924-1939.
atmosphere
andsurfaceocean.J. Ge•phys. Schlitzer, R. 1987. Renewal rates of east
•s. 88 C2, 1271-1280. lantic deepwater estimatedby inversion of
Maier-Reimer, E. and Hasselman, K., 1987. C data. J. Geophys. Res. 92 C3, 2953-2969.
Transport and storage of CO•.in the ocean - an Schlitzer, R., 1988. Modeling the Nutrient and
inorganic ocean-circulation-carbon cycle Carbon Cycles of the North Atlantic. Part I:
model. Climate Dynamics 2, 63-90. Circulation, Mixing Coefficients, and Heat
Fluxes. J. Geophys. Res. To be published.
in the atmosphere (1962-1980). J. Geophys. Siegenthaler, U. 1983. Uptakeof excessCO
2 by
Res., 8--8, 3621-3642. an outcrop-diffussion model of the ocean. J.
'0stlund, H.G., Dorsey, H.G. and Brescher, R. Geophys. Res. 8--8, C6, 3599-3608.
1976. G•OSECS Atlantic radiocarbon and tritium Siegenthaler, U. and Oesc•er, H. 1987.
results (Miami). Tritium Laboratory Data Biospheric CO•.emissions during the past 200
Report 5, Rosenstiel School of Marine and years reconstSructed by deconvolution of ice
Atmospheric Science, Univ. of M•ami. core data. Tellus 39B, 140-154.
'0stlund, H.G., Craig, H., Broecker, W.S. and Striver, M. and Quay, P.D., 1981. Atmospheric
Spencer, D. 1987. G•OSE• Atlantic, Pacific •=C changesresulting from fossil fuel CO2
and Indian Ocean expeditions. Volume 7: release and cosmic ray flux variability. Earth
Shorebased data and graphics. Superintendent and Planetary Science I•tters 5--3, 349-362.
of Documents, U.S. Government Printing Office, Weiss, W. and Roether, W. 1980. The rates of
Washington D.C. 20402. tritium input to the world oceans. Earth and
Rotty, R.M. 1981. Data for global CO
2 production Planetary Science Letters 4--9, 435-446.
from fossil fuels and cement. In: Bolin, B. Wunsch, C. 1984. An eclectic model of the
(Ed.) Carbon cycle modellinq. J. Wiley and Atlantic Circulation - Part I. The meridional
Sons, Chichester, 121-125. flux of heat. J. Phys. Oceanogr. 1--4,1712-1733.
Geophysical Monograph Series Understanding Climate Change Vol. 52

AFRICAN DROUGHT:
CHARACTERISTICS, CAUSAL THEORIES AND GLOBAL TELECONNECTIONS
Sharon E. Nicholson

Department
of Meteorology,
FloridaStateUniversity,
Tallahassee,
FL 32306

Introduction 50% in a semi-arid region such as the Sahel will

have a similar effect on the landscape. Currently,
Three quarters of the African continent is pre- a widely accepted version of the land-surface feed-
dominantly semi-arid drought-prone land. This fact back hypothesis is that in the Sahel, modification
has been dramatically observed during the last few of the land surface might have sufficient impact on
years, during which only a small strip of equator- atmospheric conditions to intensify and perpetuate
ial rain forest and the hyperarid core of the an existing drought, but not enough to initiate
Sahara desert have not experienced severe drought. one. Here the word "hypothesis" must be empha-
In 1983, one of the worst years, rainfall was below sized.
normal over virtually the whole continent (Fig. 1), Given the complexities of the phenomenon, Afri-
a sizable portion of the global land surface. In can drought is truly an interdisciplinary problem,
the semi-arid Sahel along the Sahara's southern an understanding of which requires a contribution
border, dry conditions have persisted since the of the geophysical and biological sciences. Clear-
late 1960s; annual rainfall in recent years has ly, there are many unanswered guestions. Foremost
been about half that of the 1950s (Tab. 1). among these are the causes of drought, whether it
By someestimates, famine in the Sahel claimed can be predicted, whether it will continue and, for
over 100,000 lives in 1973 alone. Thus, African the Sahel, what the humanrole has been. In this
drought is a geophysical problem of tremendous paper, I will show the extent to which we can cur--
importance. Unfortunately, the phenomenon is not tenfly answer these questions by examining and
well understoodand the causesof drought in most summarizingthe results of numerousinvestigations
of Africa have not been established. On the basis dealing with long-term climatic changein Africa,
of the available evidence, the vast majority of temporal and spatial characteristics of recent
meteorologists concur that the factors which reduce droughts, atmosphericcirculation anomalies of wet
rainfall can be traced to anomalous large-scale and dry years, tropical teleconnections to African
patterns of atmosphericcirculation, which may or rainfall, and land-surface processes.
may not be a result of sea-surface temperature
(SST) variations. This point is important, as it Paleo and Historical Climates of Africa
has often been said that drought in the Sahel has
resulted from human-induced environmentalchanges. Palco-climates and paleo-environmentsof Africa
The latter viewpoint, nowfostered mostly by have been studied extensively. Muchof the rese-
non-scientists, is usually based on the supposition arch is summarizedin reviews and overviews, such
that droughts of this extent and severity are a as those of Butzer et al., 1972• Nicholson and
recent phenomenon. This concept was introduced Flohn, 1980; Street and Grove, 1979; Williams and
into the meteorological communityby Charney Faure, 1980; Street-Perrott and Roberts, 1983;
(]975). His numerical modelling efforts suggested Street-Perrott and Harrison, 1985; Cockcroft ' et
that land surface changes,if extreme, could initi- a]., ]987; and Deaconand Lancaster, ]988. The
ate drought in the Sahel. The physical character-- changeswhich have occurred since the late Pleisto-
istics of the surface, such as albedo and soil cene are schematically generalized in Fig. 2.
moisture, can be altered by suchanthropogenicfac- Toward the end of the late Pleistocene, from
tots as deforestation, irrigation, cultivation, and about 20,000 to 12,000 B.P., the desert expandedto
overgrazing. However,a reduction of rainfall by cover most of the continent (Fig. 2). Humid cli-
mates all but disappeared and tropical forests were
confined to a few highland refuges. Conditions
Copyright 1989 by changed abruptly about 12,000 B.P. and deep lakes
International Union of Geodesy and Geophysics formed in many formerly arid locations. The first
and American Geophysical Union. humid period climaxed about 9,000 years ago and

79
Geophysical Monograph Series Understanding Climate Change Vol. 52

80 AFRICAN DROUGHT' CHARACTERISTICS, CAUSAL THEORIES

............ .•'... '-4; , -•7.

,:.:,•?-.---:::":
ß .. :......,,
::...:•-..
,.:-
-'•.--' ---........:-:::.::.:;......:....
.... .

. ..

: .

....... -..--.*.;.
•.
-.** '-:.'.:
-.;,:
4.•
0 * '•'-•,;.,....;:!•
..-,,• ..'
...........
:'--",•.:
:*-,- .•\. - ':.-•..
.-
- 40 • ";'
.... '""--:"•
':"..
,"":
.....
:-'"'"'"'-%'•....
ß -20 ;:..-'-
• • . '""-?.

...:

'*•,%, ....
:->' -:.;
":::.;.......?:"
....<
..'.
,..

,,,
-.....',::.,::'•;!:'•i;•;:•,:*
.........
L..•;•-:•. '"*":>•:':":•"*
%%, -40 .-'. ,..

,•.... •---;:•;:.". -20

:..............
- ....
. ..: .....
: :.?.,...:--::-...
, ½.,:a;.; ....:.:..""w.'..........;-..:-.'.--%
:::::::::::::::::::::
{•:•*:.2:-,
..... •::::::•.•.•:.•..• -.:..:•
.:..,. • ,,
.

"'.•"::•:i:....:..•.-'::':; ' ' .......

'"'"*
.

ß-:.:',-(-'::..' ':

.
•:.:. • • *'. , .: . .:
;...

"b / ""':;'*
'
., •.;
-- :!•.... ::..
?-:;..::!}? ,.... <"' :
, ,,. •7,:.•"
...
•½;::
*' %:':' - 20 - 40 , '½:*;•::'
'"<,-:.:.--::."

Fig. 1. Schematic of rainfall departures (% above or below normal) for 1983 superim-
posed on 14eteosat view.

apparently encompassed the equatorial regions and lihood that is no longer possible. Rift Valley
the subtropics of both hemispheres. h second humid lakes were several hundred feet deeper than at
period prevailed from about 6,000 to 4,000 B.P. present and expanded to form a huge hydrologic
The desert had all but vanished; lowland marshes system. Lake Chad, almost entirely desiccated in
forned in the northwestern Sahara while Neolithic recent years, was then over a hundred meters deep
nan herded cattle in the central Sahara, surrounded and expanded to ten times its normal twentieth
by fauna grazing on what was apparently a savanna century size.
landscape. Lakes dotted now arid regions of •auri- The onset of more arid conditions occurred about
tania; fish hooks uncovered in archaeological 3,000 years ago and since that time conditions
sites there attest to human occupation and a live- generally resembling those of the current century
Geophysical Monograph Series Understanding Climate Change Vol. 52

NICHOLSON 81

TABLE 1. Mean Annual Rainfall (mm) cators, including rainfall, river flow, lakes,
agriculture and harvests.

1950-59 1970-84 Temporal and Spatial Variability of

Rainfall during the Present Century
B I LMA 20 9
Rainfall fluctuations in Africa show preferred
ATBARA 92 54 spatial and temporal patterns of occurrence. The
dominant spatial modes of variability have been
NOUAKCHOTT 172 51 analyzed by Klaus (1978), Gregory (1982) and Motha
et a]. (1980) for West Africa, by Dyer (]975) for
KHARTOUM 178 116 South Africa, and on a continental scale by Nichol-
son (1980b, 1986a, 1986b). The most important
AGADEZ 210 97 characteristics are the large-scale spatial coher-
ence and the relatively small number of typically
T IMBUKTOO 2 41 147 occurring patterns of rainfall anomalies. The most
common mode is drought or below-average rainfall
NEMA 381 210 throughout the subtropics of Africa but increased
rainfall in equatorial latitudes; this occurs
DAKAR 60 9 308 during 13 years between 1910 and 1973. Another
common pattern, typified by 1983 (Fig. 1), is
BANJUL 1409 791 drought or below-average rainfall continent-wide, a
situation occurring during the two historical dry
periods illustrated in Fig. 3. The remaining major
patterns are the inverse of these: continentally
have prevailed. However,extensive periods of-sig- "wet" conditions or abnormallyhigh rainfall in the
nificantly more humidor more arid climate have subtropics and rainfall deficits in equatorial
occurred. Theseare best illustrated by the last latitudes. This clearly implies that large-scale
two centuries, since a greater availability of data circulation changes, and not regional-scale pheno-
allows for more reliable historical reconstructions mena, impose the dominant control on rainfall vari-
than for earlier periods. Conditions for three ability.
periods commencing in the 1820s are summarized in The large-scale spatial coherence also implies
Fig. 3; details can be found in Nicholson, 1978, that a few key regional time series can serve to
1979a, 1980a, and 1981a. describe rainfall variability over much of the con-
A general desiccation occurred throughout much tinent. Four such series are shown in Fig. 4.
of Africa toward ]800; in the subtropics and equa- That for the Sahel is generally representative of
torial regions subnormal rainfall or drought pre- the area from 10øN to 25øN and extending across
vailed for most of two decades in the 1820s and most of the east-west expanse of Africa. That for
1830s (Fig. 3). The situation was particularly East Africa is broadly representative of the equa-
severe in the Sahel, which had experienced similar toria] zone; and the northern and southern Kala-
persistent droughts in the 1680s and in the 1730s hari series are averages of the subtropical lati-
through 1750s. Lake Chad was almost totally desic- tudes of southern Africa. The droughts of recent
cated, as it has been during recent drought years. years and the wet period in the subtropics in the
After the 1830s, rainfall began to increase and 1950s are clearly apparent.
within decades, conditions considerably wetter than These time series also illustrate an important
those of the current century again prevailed. contrast between the Sahel region and other parts
This relatively humid period persisted from the of Africa. Rainfall variability in the Sahel is
1870s to the mid-1890s. Lake levels from Chad in characterized by long-term persistence of anoma-
the north to Ngami in the south rose dramatically, lies, e.g., the long series of above-average rain-
by several meters in many cases. The high stands fall in the 1950s and early 1960s and the continual
which persisted for tens of years were achieved at sequence of dry years since 1968. In comparison,
most briefly, and generally not at all, during the the remaining series show only high frequency vari-
twentieth century. Near Timbuktoo, wheat produc- ability, the largest continually wet or dry period
tion thrived to such an extent from the continually being five years. The persistence in the Sahel has
good Niger floods that grain was exported to sur- been pointed out by numerous authors (Bunting et
rounding areas. Numerous measurements of rainfall al., 1976; Walker and Rowntree, 1977; Kraus,
and river flow confirm the wetter conditions and 1977a; Nicholson, 1979b, 1982, 1983, 1985; Lamb,
suggest that rainfall in subtropical regions was 1982; and Penneft et al., 1985). This character-
then 25-35% greater than the 20th century normals. istic and its absence in southern Africa, a region
Conditions again changed abruptly around 1895, and where teleconnections with Sahel rainfall are evi-
a continent-wide decrease in rainfall culminated in dent, may be indicative of a more local positive
a long period of severe droughts in the 1910s. The feedback mechanism acting to reinforce rainfall
steady decline is evidenced by a variety of indi- anomalies initiated by large scale forcing. Thus,
Geophysical Monograph Series Understanding Climate Change Vol. 52

82 AFRICAN DROUGHT: CHARACTERISTICS, CAUSAL THEORIES

Fig. 2. African paleoclimates c. 18,000 B.P., 9,000 B.P. and 6,000 B.P. compared to
present conditions (rainfall' shading or plus signs, generally greater than at
present; dots or minus signs, less than at present; hatching, decreasing rainfall;
lakes: ß : high stands, O: intermediate stands, A : low stands; based on Street-
Pertoft et al., Nicholson and Flohn, Cockcroft et al., and others).
ß

the persistence can be used as a point of evi- frequencies. Quasi-periodic fluctuations on these
dence in support of the land-surface/atmosphere time scales are characteristic of rainfall through-
feedback hypothesis previously mentioned and de- out equatorial and southern Africa (see also Tyson
scribed in more detail in Section 6. et al., 1975; Rodhe and Virje, 1976; Dyer and
The temporal characteristics suggested by the Marker, 1978; Tyson, 1980), but are notably absent
time series in Fig. 4 are quantified using spectral elsewhere on the continent (Fig. 6). They also
analysis (Fig. 5). The spectra for East Africa and characterize numerous other atmospheric and oceanic
the Kalahari regions each show three significant phenomena in the tropics. All three are evident in
peaks in the range 2.2-2.4, 3.3-3.8 and 5-6 years the Southern Oscillation; the 5-6 year time scale
(Nicholson and Entekhabi, 1986). In contrast, that is the dominant one for SST variability in the
for the Sahel shows "red noise", which is indic- Atlantic and Indian Oceans (Fig. ?); and the Quasi-
ative of persistence, and little power at higher Biennial Oscillation is a well-known phenomenon.
Geophysical Monograph Series Understanding Climate Change Vol. 52

c. 1820-1840 c. 1870 -1895

tO' O' 10' 20' 20' 40' SO' to' O'
,

0 -+'t- i
+'1-
-I- .-I-

+o+O - _ .--
-I-

.... __
-I-
....

.._

-t-

o•
...

I
I0' O'
I0' 20' i•' 40'
i SO' I0' O' I0' 40' SO'

c. 1895-1920

I- -I- --
-I- --

O' I0' 20 )0 40 [0'

,

Fig. 3. African rainfall anomalies for three historical periods (minus signs denote
evidence of drier conditions; plus signs denote evidence of above-average rainfall;
zeroes, near normal conditions; circled symbols denote regional integrators, such as
lakes or rivers).
Geophysical Monograph Series Understanding Climate Change Vol. 52

84 AFRICAN DROUG•{T' CHARACTERISTICS, CAUSAL THEORIES

SAHEL
Lamb attributed the tower Saber rainfall to
southward displacement of the wind convergence zone
and equatorial trough, thermally forced by the
anomalous SST pattern, but other atmospheric anoma-
lies over the At]antic are also evident (Hasten-
rath, 1984a). Lough (1986), examining approxi-
-5O mately the same area and time period, found little
EAST AFRICA evidence of the consistent relationship between
SSTs and Sahel rainfall which Lamb suggested. She

11'Jl,
im,i also found that SSTs appear to lag Sahel rainfall,
a result which implies that SST patterns do not
- [lLr UJyL' ,-LJ uLr force the rainfall
Foiland et at. (1986a,b)
fluctuations.
found an Atlantic SST
-5O anomaly pattern for Sahel dry years which is mar-
50 KALAHARI NORTH kedly similar to that shown by Lamb; this is sig-
nificant, since only one of the five years in

I Rnr,.['•[]•Rn
I"!•
nnn•,•,..nn•l..•rl
I] Lamb's dry composite was also in their
posite. Foiland et al.
dry com-
also showed that the Atlan-

U' tic anomalies

pattern.
are part of a global SST anomaly
They conclude that the relevant factor is
-5O the SST difference between the two hemispheres;
KALAHARI SOUTH the global inter-hemispheric difference correlates
considerably better with Sahel rainfall than does
the interhemispheric SST difference for the Atlan-
tic alone. These authors use the third EOF of

'il
global SSTs (explaining 4.7% of the variance) to
represent interhemispheric SST differences (Fig.
•0 8). Based on a correlation of .6 between EOF 3 and
I I I i i I I I
Sahe] rainfall (significant at the 99% confidence
1900 1910 t920 1930 19•0 t950 1960 1970 1980
level), an attempt was made to predict rainfall
from global SST anomalies. However, prediction may
Fig. 4. Rainfall fluctuations in four sectors of
be premature since SST changes often lag rainfall
Africa 1901-1984, expressed as a percent departure
(Fig. 8) and there is little correspondence between
from the long-term mean (ENSO years are shaded for
the two series during several periods of signifi-
Sahel and East Africa, ENSO +1 is shaded for the
cantly abnormal rainfall. As an example, the coef-
two Katahari series).
ficients of EOF 3 are similar for the 1910s drought
and the wet 1950s. This situation is well illus-
trated by a scatter diagram of the EOF coefficients
Tropical Teteconnections to versus rainfall (Fig. 9)' despite a general posi-
African Rainfall tive correlation, much scatter is evident except
for the most extreme years. Moreover, the reta-
The correspondence between the dominant time tionship appears to be break down prior to the mid-
scales of rainfall variability and those of other ]940s (Lough, 1986; Foiland et al., 1986a),
tropical parameters suggests possible forcing leading to speculation that it may be the steady
mechanisms of rainfall variability. Here, several downward trend in both Sahel rainfall and inter-
of these are examined, beginning with sea-surface hemispheric SST differences which is principally
temperatures. Most studies relating rainfall to responsible for the statistical correlations.
SSTs have been concerned only with the Sahet and a The mpst consistent result of all of these stu-
number of viewpoints have emerged. Only a handful dies ]s a pattern of anomalously high SSTs in the
of studies have dealt with other locations. In Atlantic south of c. 10'N and tower than normal
many cases, weak or strong statistical associations temperatures to the north during years of Sahel
have been documented but the dynamic mechanisms drought. The authors agree that such conditions
accounting for these associations have not been appear to favor decreased Saber rainfall, but are
established. unlikely to be uniquely the cause of drought.
One of the first studies to examine the rela- Recent papers such as Foiland et al. (1986a,b),
tionship between Saber rainfall and sea-surface Semazzi et at. (1988), Hsiung and Newell (1983) and
temperatures was that of Lamb (1978a,b). Using a Nicholson and Nyenzi (manuscript submitted to
Sahel rainfall index heavily weighted toward the Journal of Climate, 1989) show that the Atlantic
western Saber (west of 0øW), he concluded that anomalies are part of a global pattern. That pat-
Sahelian wet years are associated with abnormally tern in turn, or the atmospheric anomalies produc-
high temperatures throughout most of the tropical ing or responding to it, may be the critical factor
Atlantic, but below average SSTs to the north and in Saber rainfall variability. Numerical simula-
west of the convergence zone over the Atlantic. tions (Palmer, 1986' Semazzi, personal communica-
Geophysical Monograph Series Understanding Climate Change Vol. 52

NICHOLSON 85

PERIOD IN YEARS PERIOD IN YEARS

õ0 I0 ? 5 4 3 2.5 2 I.S SO tO 1' 5 4 • 2.5 2 1.5
1 I I I I I I I I I I I I I I I I I

SPECTRUM SPECTRUM
Sahel East Africa

.03

: .

..
. .
99% cg--
. .

. .
. .

. .

.02 ..•
-::::

95% CL.__
ii•i 2.6-3.8yrs ii

.01

.05 .10 .15 .20 .•5 ß05 .10 .15 .L•O .Z5
FREQUENCY FREQUEI•I•'

PERIOD IN YEARS PERIOD IN YEARS

50 tO ? 5 4 3 2.5 2 1.5 SO tO 7 S 4 $ 2.5
I I I I I I I I I I I I I I I i

SPECTRUM
SPECTRUM
Kalahari South
Kalahari North

.04
:e•2.3 yrs
3.8 yrs
3.6 yrs
:
•: :
.O3
• ::
•! •i 99% CL
• 2.3yrs•i 99%
CL--
4 8 yrs • •
::
.

•-!!- - • - •i • c• _ _
• :• :. ::
.02 :
:
:::
:::
:::
:::
:
:.::
:: - yrs •:
i •i •:: ......

:::: ..... -- : .
:::

..

:. ::::::
......................

:: ::::::
::::::
:::::::::::::.:
:::::: : .....................

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

.0!
...................................
•••z••z•z•z• .... •zz•z
...................................

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

:. ::::::::::::
........................................
........................................... ......................................
........................................... ........................................
...........................................
............................................
.............................................
.............................................
.............................................. :::::::::::::::::::::::::::
............................................... .............................................

.o5 .10 .15 .20 .os .IO .,• .2o

FREQUENCY FREOUENCY

Fig. 5. Spectra of rainfall fluctuations for the four locations in Fig.4.

tion) likewise suggest that global, Pacific Ocean, cially the relationship with the Indian Ocean,
and Indian OceanSSTs all have a greater influence since the systemswhich "fail" in the western Sahel
on the Sahel than Atlantic SSTs do. This finding in dry years develop from pressure perturbations in
is more compatible with Sahelian climatology, espe- the eastern Sahel and traverse most of the east-
Geophysical Monograph Series Understanding Climate Change Vol. 52

86 AFRICAN DROUGHT' CHARACTERISTICS, CAUSALTHEORIES

30

20

lO

2.2
-2.4
-2O

-3O

40

30.

20.

-lO.

5.0- 6.3yrs
-20.

-30.

Fig. 6. Distribution of spectral peaks in rainfall at 2.2-2.4, 3.3-3.8, and 5.0-6.3

years (dark shading, 95% confidence level; light shading, 90% confidence level; dots
indicate regions where peak is evident; based on the 84 regions in Entekhabi and
Nicholson, 1986).

west extent of the continent. These are more wind stress in the equatorial Atlantic. This
likely influenced by factors "upstream" in the implies remote forcing of BBT anomalies via the
Indian Ocean than several thousand kilometers down- Kelvin wave mechanism described by Adamec and
stream in the tropical Atlantic. O'Brien (1978). Surprisingly, they found that
A handful of studies have examined the relation- SST, pressure and wind anomalies accompanying wet
ship between sea-surface temperatures and rainfall conditions during the second rainy season of
variability elsewhere in Africa. Hirst and Basten- October and November were dissimilar to those of
rath (1983a, b) showed that increased rainfall wet Match/April seasons and in some ways (e.g.,
during the March/April rainy season along the high pressure and strong southeast trades) resem-
Angolan coast and in the western equatorial region bled the March/April dry composites.
(Zaire basin) is associated with a weakening of A strong relationship between SSTs and rainfall
the high pressure over the Atlantic and reduced along the Benguela coast, which includes the
strength of the southeasterly trades and westward Angolan sector studies by Hirst and Hastenrath, was
Geophysical Monograph Series Understanding Climate Change Vol. 52

N I CHOLSON

PERIOD (YEARS)
5010754 :•?..5 ?. 1.5 I 0.5
PERIOD (YEARS)
r I I I I i I I I I
I0 5 3, 2.5 1.5 .
.17. .15
ATLANTIC ......

: . INDIAN
SOI
.12

- 4 ;; -
,I t f I
- •0 , -

-•1•.• -
.O6
.:1 • •. ,,• _

.O3
'
;•1 •" •, , ! • -

25 IO 5 3 2 I
I I I '1 I I
.OO
.OO .05 . I0 .15 .20 .25
FREQUENCY (CYCLES/SEASON)
10%- SSTz
Fig. 7b. Spectra of the first principal component
of sea-surface temperatures in the tropical Atlan-
tic and tropical Indian Oceans (from Nicholson and
Nyenzi, 1988, manuscript submitted to the Journal
of Climate).

munication) also shows that a relationship exists

99% between East African rainfall and the Indian Ocean.
5%-
During unusually wet periods of the September-to-
December season, pressure is below normal, cloudi-
ness and SSTs above normal and trade-wind conver-
gence is reduced in the western tropical-Indian
Ocean while the ITCZ is abnormally strong. The

o-(•",", ; ...... ; ........................ ,,

25105 3 2 I .5
YEARS -

Fig. 7a. Spectra

ofSSTfluctuations
alongthe • •I ,i ' ' ,

•.n•u.•
oo•st
o•
t•.
sout•..t.•
•t•.t•
(•0's
20'S,1948-72)
- ø i,,:,IV
andtheTahiti-Darwin
of the Southern Oscillation
pressure
index
(1935-73) ....
/ ,

demonstrated
by
Nicholson
and
Entekhabi
(1987).
and rainfa]]
-,oo
i i
Theyfurther showed
that the varianceo[ both SSTs
years (Pigs. 6 and 7).
is
As described in section
concentrated
3,
on time sca]es of 4-6
' -,
the variance of rainfal] throughout the equatoria]
zone and in much o[ southern Africa is similarly
dominated by 4-6 year quasi-periodicities and is Pig. 8. Time series of the third principal coN, on-
strongly coherent with SST fluctuations on this ent of g]obal sea-surface temperatures and Sahel
time scale (Pig. 10). Cadet (1987, personal com- rain[a]], 1901-1984 (from Polland et al., 1986a).
Geophysical Monograph Series Understanding Climate Change Vol. 52

88 AFRICAN DROUGHT' CHARACTERISTICS, CAUSAL THEORIES

I00
I i I ! I I
(Nicholson, 1980b). Likewise, SSTs may be more
important as high frequency forcing of the inter-
annual variability than as a mechanism producing
the ionget-term rainfall trends in the Sahe].
Evidence for a global teleconnection derives
5O directly from studies relating rainfall to tropical
ß
phenomena such as the Southern Oscillation/E1 Nino
ß ß (ENSO) and the 30-60 day oscillation. Only a few
ß
ß ß studies, however, have evaluated African rainfall
ß ß
ß
in a global text. The two most comprehensive are
ß
ß ,

ßß ß
ß
those of Nicholson and Entekhabi (1986) and Rope-
EOF3 ß ß ß
lewsk] and Halpert (1987). These studies used
basically the same data set but different analysis
ß ß ß

ß ß ßß ß

-5O .
ß

ß
ß

ß ß ß
.
techniques, the former dealing primarily in the
frequency domain and the latter solely in the time
ß ß
ß ß
ß
ß ß

domain. Both studies show that. the regions with the

ß
ß el ß
ß

strongest statistical associations between rainfall

ß ß
and the Southern Oscillation are equatorial East
-I00 -- ß ß _
Africa and southern Africa and that no relationship
ß ß ß
ß
is evident in the Sahel. As shown in Fig. 4, East
Africa generally receives above normal rainfall in
ß eee
ß ß
the ENSO year, usually during the "short rains" of
-150 October and November, while droughts tend to occur
in southern Africa early in the year following
ee
ENSO. The association with droughts for South and
i. I I I I I I southeastern Africa was also shown by Stoeckenius
,50 -I00 -,50 O 50 1OO 150 (1981), Rasmussonand Wallace (1983), Dyer (1979),
RAIN Schul ze (1983), and Lindesay et al., ] 986.
(% DEPARTURE- NORMALIZED)

Fig. 9. Scatter diagram of the data in Fig. 8,

comparing component scores of EOF3 and Sahel rain-
fall.
.68

.76

opposite patterns occur in dry years and in the .51

eastern Indian Ocean. Although Cadet considers .61 .63

only East Africa, the strong te]econnections in
rainfall between East and southern Africa (Nichol-
son, 1986b) and the coherent SST variability
between in the Indian and Atlantic Oceans suggest
that the Indian Ocean would influence rainfall
elsewhere in southern Africa. Again, the relation- .73

ships may well be part of a global anomaly pattern

influencing both oceans and the African continent.
In su•m•ary, a number of studies have examined
the relationship between rainfall variability in
the Sahe] and sea-surface temperatures. Despite
agreement on certain associations, such as the
pattern of SSTs coincident with wet or dry years,
there are numerous points of disagreement among
various authors and several questions remain. .47

Among these are whether SSTs may lag and therefore

possibly respond to rainfall, whether the critical
links are with oceanic sectors in proximity to the
Sahel or with a global pattern of which these are a
part, and whether rainfall and SSTs may respond to
comon atmospheric forcing which may be the more Fig. 10a. Coherence square between sea-surface
direct factor in rainfall variability. Some stu- temperatures along the Benguela coast (10ø-20øS)
dies have also suggested that oceanic influences in the 5.0-6.3 year spectral band for all rain-
may be more significant in the western Sahel, pos- fall regions with discrete spectral peaks in this
sibly providing an explanation for the often dif- band (95% confidence level = .58, 99% confidence
fetenrta] behavior of western and eastern sectors level = .63; from Nicholson and Entekhabi, 1987).
Geophysical Monograph Series Understanding Climate Change Vol. 52

N I CHOLSON 89

40.

30.

20.

10.

O.

-10.

-20.

-30.

-40.

40.

30.

20.

lO.

O.

-10.

-20.

-30.

-40.

Fig. 10b. Coherence-square betweenrainfall and the Southern Oscillation in three

spectral bands for all regions with significant spectral peaks in rainfall in these
bands (shading denotes values exceeding the 95%confidence level, from Nicholson and
Entekhabi, 1986).

NicholsonandEntekhabi(1986) providemore de- years) (Fig. 5) also characterizethe Tahiti-Darwin

tail on the relationshipbetweenENSOand African surfacepressureindexof the SouthernOscillation
rainfall. Thethree spectral peaks consistently (Fig. 7). Thecoherence-square (Fig. 10) indicates
evident in rainfall series (~2.3, ~3.5 and 5-6 that the relationship betweenthe two is strongest
Geophysical Monograph Series Understanding Climate Change Vol. 52

90 AFRICAN DROUGHT: CHARACTERISTICS, CAUSAL THEORIES

SAHEL SAHEL MEAN SAHEL SAHEL

WET WET (1901-73) DRY DRY

'11950
1933 1931 1971
1953 1955 1947 1972
1968
15ø 1956 1957 1973

,oø
5ø ,
AM J J AS 0 A MJ J A S 0 AMJ J AS 0 AMJ JA SO A MJ J ASO'
Fig. 11. Location of the ITCZ over West Africa during wet years and dry years, as
approximated by the "center of gravity" of the rain belt for the months of April to
October (three-year averages plus long-term mean; circled areas are rainfall exceeding
200 mm/month; shading marks latitude of maximumrainfall in August; from Nicholson,
•9S•b).

at 5-6 years in equatorial regions, at ~3.5 years droughts. Modern satellite-derived data sets,
in southern Africa, and at ~2.3 years throughout which are hard to compare with conventionally
the eastern half of the subcontinent. In no case observed data, are available only for recent years;
is the relationship with ENSOas strong as with for the Sahel, this includes only drought years and
Atlantic SSTs along the Benguela coast. This sug- no true "normal" or wet year for comparison. The
gests that the direct link with rainfall in equa- result is that moststudies have either considered
torial and southern Africa is probably sea-surface a very limited numberof years or were severely
temperaturesor atmosphericparametersin the restricted in the upperair parameterswhich could
marine sector, which in turn appearto be modulated be assessed. The results of these studies are
by the ENSO events. briefly summarized.Threeareas, the Sahel, East
A numberof other global studies illustrate the Africa and southernAfrica, are considered separ-
influence of majortropical phenomenon on the Afri- ately becausethe analyseshavetendedto do so and
can sector. The ENSO wind compositesof Arkin because different factors appear to operate in
(1982) and Selkirk (1984) illustrate consistent these locations.
sequencesof anomaliesover certain parts of Anearly hypothesisinterpretedchanges
of Sahel
Africa. Arkin, for example,shows a progression rainfall mainly as a function of the position of
from easterly to westerly 200-mbwind anomalies the Intertropical Convergence Zone(ITCZ) and sub-
over southern Africa during the course of ENSO tropical high (Kraus, 1977a, b; Greenhut, 1977;
years, and a progression from westerly to southerly Beer et al., 1977). The only actual evidence to
anomalies over northern Africa. Selkirk's results support the hypothesis came from Laml) (1978a,b),
likewise emphasize the change of the upper-level whose analysis was limited to conditions over the
zonal flow over Africa in relationship to high and Atlantic Ocean. Later, Newell and Kidson (1984)
low index periods. The 30-60 day oscillation stu- and Nicholson (1981b) showedthat over the African
dies of Madden (1987), Weickmannet al. (1985), continent itself there was no systematic southward
Knutson et al. 1986, Knutson and Weickmann (1987) displacement of the ITCZ. In many dry years, the
show variations in outgoing longwave radiation convergence zone is at least as far south as during
(OLR) and wind over the African sector that are droughts (Fig. 11) and it remains near its mean
coherent components of the global 30-60 day signal. position over the continent during most dry years.
Others (e.g., Namias, 1974; Miles and Foiland,
Atmosphere Circulation Changes 1974; Tanaka et al., 1975; Schupelius, 1976; and
Associated with Rainfall Fluctuations Nicholson, 1979b, 1980b) produced additional evi-
dence refuting the hypothesis.
Although numerousstatistical analyses and model A number of more plausible explanations for
simulations have linked African rainfall to sea- drought in the Sahel have since been proposed.
surface temperature fluctuations and tropical Various investigations have shown that, compared
atmospheric phenomena, few studies have dealt dir- with wet years, drought years in the sub-Saharan
ectly with drought mechanismsin terms of local region are characterized by a weaker 200 mb trop-
atmospheric factors (e.g., circulation changes, ical easterly jet (TEJ), a stronger mid-tropo-
atmospheric moisture, stability). This lack may be spheric (700 mb) African easterly jet (AEJ), weaker
due to the dearth of upper air data over Africa. shear in the 700 mb easterly jet south of the
Sufficient radiosonde and pibal data exist for only equator, enhanced Endicy-type overturnings (and
a few years, most prior to the recent severe weaker Walker-type overturnings), increased geopo-
Geophysical Monograph Series Understanding Climate Change Vol. 52

NICHOLSON

tential of the 700 mb surface, increased vertical 1984), but a dynamical explanation for this associ-
shear over West Africa (thus, enhanced horizontal etlon is not readily apparent. The interaction
temperature gradients), and the virtual disappear- between tropical and extra-tropical systems (Kumar,
ance of the 850 mb trough over West Africa (Kidson, 1978; Riehl, 1979) is likewise a significant
1977; Tanaka et el., 1975; Kanamitsu and Krishna- determinant of the character of the rainy season.
mufti, 1978; Newell and Kidson, 1979, 1984; and As an example, the coincidence of an eastward-
Dennett et al., 1985). Surprisingly, there appears moving westerly (extratropical) trough with a
to be no systematic difference in the depth of the westward-propagating trough in the tropical easter-
monsoon layer, its humidity or advective moisture lies can rapidly transform a stable tropical air
flux over West Africa between wet and dry years in mass into an intense disturbance. Knowledge of the
the Sahel (Lamb, 1983). In many cases, linkages relationship between such systems and interannual
with other tropical weather anomalies are evident variability of rainfall is woefully inadequate. A
(Kanamitsu and Krishnamufti, 1978; Krueger and better picture, however, is starting to emerge.
Winston, 1975; Kraus, 1977a; Fleer, 1981; and Several recent papers have shown the importance of
Hastenrath and Kaczm•rczyk, 1981). diagonal cloud bands (also a system manifested
Even fewer systematic studies of wet and dry by tropical/extra-tropical interactions) in deter-
years have been carried out for other parts of mining the seasonal character of rainfall (Her-
Africa. In East Africa, local forecasters gener- rison, 1983, 1984; Harangozo and Harrison, 1983).
ally associate westerly wind anomalies in the mid- These systems influence most of the southern sub-
or lower troposphere with abnormally high rainfall continent and are clearly linked to large-scale
but that is not always the case (Kiangi and Temu, patterns of tropical flow, as evidenced by their
1984). The two monsoon flows prevailing during relationship to the ENSOphenomenon.
most of the year are generally divergent and a
large-scale diffluence of the NE monsoonover East Land Surface Processes
Africa is readily observable (Flohn, 1964; Anyamba,
1984). Thesuggested causesof rainfall variabil- In the early 1970sit becameapparentthat human
ity are quite varied, as the factors controlling activities weredramaticallyaltering the natural
rainfall are complex anddiversein various parts global landscape.Semi-arid]ands,the predominant
of the region. Moreover, most investigations of African environment, wereshown in manycases to
rainfall variability havebeenconfined to case be undergoing a processof "desertification", the
studies of one or twoyears(e.g., Minja, 1985; result of overcultivation,deforestation,overgra-
Anya.•a, 1983, 1984; Agu•a, 1984; Anya.•a and zing, improper irrigation, slash/burn agriculture
Ogallo, 1986; Ogallo and Anyamba,1986; and Flohn, and other forms of land mismanagement(United
1988). Therefore, few generalizations can be made. Nations, 1977). The desertified landscape, rela-
Factors which have been associated with abnormally tively barren of vegetation with bare soil exposed
wet conditions, such as those of 1961/62 or and subsequently eroded, was illustrated by several
1977/78, include strong low-level westerlies from satellite photos showinga contrast between high
the Atlantic and simultaneous intensification of albedo in altered sectors and low albedo in adja-
the Mascareneand Arabian Highs. This is consis- cent protected areas (e.g., Glantz, 1977). Other
tent with Cadet's finding of a strengthening of physical changes expected to accompanythe removal
ITCZ over the western Indian Oceanduring wet years of vegetation include reduced soil moisture, evapo--
in East Africa. During droughts, such as 1972 or transpiration and surface roughness, higher surface
1983/84, low or mid-level easterly anomalies often temperatures, and increased soil erosion and dust
appear in association with a MascareneHigh which generation (Anthes, 1984). A drought would likely
is unusually intense or more zonally oriented. produce similar land surface changes.
There is also evidence that East Africa is influ- The potential feedbacks on the atmosphere in-
enced by fluctuations in the tropical divergent clude changes of latent and sensible heat transfer,
circulations (Kanamitsu and Krishnamufti, 1978; thermal stability and convergence. The hypothe-
Chen and van Loon, 1987). Anomalous
zonal diver- sized feedback was first modelled by Charney
gent circulations (Walker-type overturnings) may (1975). His initial study involved an albedo
account for the apparent inverse relationship change over the Sahara from 35% to 14%; the result
between the intensity of East African rainfall and was a significant reduction in rainfall over the
the Indian monsoon,as suggested by studies of Sahel and a southwarddisplacement of the rain zone
Stoeckenius (1981), Rasmusson and Carpenter (1982), in the high albedo case. In subsequentGCMsimula-
and Ropelewskiand Halpert (1987). tions (Charneyet al., 1975, 1977) the concept of
Although the synoptic systems of southern Africa reduced evapotranspiration was added and a similar
are fairly well described (Bhalotra, 1973; Acharya reduction in rainfall resulted (Fig. 12). A large
and Bhaskara Rao, 1981), in most cases our under- number of studies have similarly examined land-
standing of mechanismsof interannual rainfall surface feedbackmechanisms, simulating not only
variability in the region can be reduced to a few albedo and evapotranspiration or soil moisture but
generalizations. The importanceof easterly flow also surface roughness. These are beyondthe scope
in the upper-troposphere in producing high rainfall of this paper but summarizedin review articles
has been established (Taljaard, 1981a,b and Tyson, such as Mintz (1984) or Nicholson (1988). Despite
Geophysical Monograph Series Understanding Climate Change Vol. 52

92 AFRICAN DROUGHT' CHARACTERISTICS, CAUSAL THEORIES

SAHEL RAJPUTANA W. GREAT PLAINS

EXCESSIVE EVAPORATION EXCESSIVE EVAPORATIOPJ
EXCESSIVE EVAPORATION
9 9 T ,

8 .

E
z 5
.
% ALB •
z
f-I4%ALBEDO
_(2
k- 4•

a.:3-\,•/35Olo
ALBEDO
I

i i
I 2 3 4 I 2 3 4
NEGLIGIBLE EVAPORATION NEGLIGIBLE EVAPCRATICN NEGLIGIBLE EVAPORATION

14%ALBEDO
' ' '
• 4.l- /14'/o
ALBEDO
-I r- 14%ALBEDO
35%ALBEDO
% ALBEDO

I
2 3 4 1 2 3 4 2 3 4
TIME (weeks)- JULY TIME (weeks) TIME (weeks)

Fig. 12. Weekly average rainfall rates (mm/day) for July in three semi-arid regions
from model simulations using surface albedos of 14% and 35% in the test area (from
Charhey et al., 1977).

considerable differences in mechanisms,locations An important question concerningsensitivity is

considered or nature of the models used, nearly all the source of atmospheric moisture in the Sahel
lead to the conclusionthat large-scale land sur- region. Current model results suggestthat changes
face modification can alter climate and that the of soil moisture and/or evapotranspiration would
Sahel is one of the regions mostsensitive to such have a greater impact than albedo. Unfortunately,
effects. studies of atmosphericmoisture over West Africa
Althoughthere is genera] agreementbetween the have reacheddifferent conclusionsconcerning its
diverse numerical simulations, strong empirical sources (e.g., low-level southwest monsoonflow,
evidence of ]and/atmospherefeedbackis lacking. A Indian Ocean, the Mediterranean) and little is
fair amountexists for regional impact (see Anthes, known about the contribution of local evaporation
1984) and one statistical study (Walsh et al., to the supply (Cadet and Houston, 1984; Cadet and
1985) suggests large-scale influence of soil mois- Nnoli, 1987).
ture on atmospheric variables. The multi-year Despitethe belief in widespread"desertifica-
persistence
of rainfall anomalies
in the Sahel, as tion" over WestAfrica (UnitedNations,1977),
described
in Section3, couldwell be a manifesta-there have beenfewsystematicstudies of land
tion of suchfeedback(Nicholson,1986c). Evenso, surface changeand there has been virtually no
a numberof additional questions must be answered attempt to distinguish betweenhuman and natural
in order to determinewhetheror not land-surface factors in any observedchanges. For West Africa,
change can influence rainfall in the Sahel (or rainfall decreasedby a factor of two during the
elsewhere) and to what extent humanfactors contri- time period over which desertification has been
bute. These include the extent of environmental assessed. The reduced rainfall, rather than human
ch&nge, the sensitivity of the atmosphereto sur- impact on the land, is the likely cause of the
face forcing, and the quantitative physical impact suggested "advancement" of the Sahara. Moreover,
of lmndlc&pechanges. in most cases, broad generalizations have been made
Geophysical Monograph Series Understanding Climate Change Vol. 52

NICHOLSON 93

16øW 14" 12" I0 ø 8" 6ø 4ø 2ø 0 2ø 4" 6" 8ø I0 ø 12" 14" 16øE

I I i I I I I I I I I I I I I I I I I I I I I I I I I I I I

16'

•14 o

12 ø

,oo-
:
111]1
fIlljIll
IIII]]
I ',",'11''''111111
',_'
Iili
11II1
II!IIII
]lilliJllllll ]IIIIIii1111I iO o

Sø

:o,
%
_ ,.o Ili1 !111111
III1["
II1111IIIIII!1IIIII!
IIIIIII!
I Ill......
IIll ........
- , ,,,,,,, "III
4"- IT[, 4 •
ITn •.•-2o

i
2•_ i i i i i i i i i i i i i i i i i i i i i i i
16øW 14ø 12ø I0 ø 8" 6ø 4ø 2ø 0 2ø 4ø 6" 8" I0 ø 12" 14' 16øE

Fig. 13. Map of albedo changes over West Africa, pre-agriculture to present (from
Gornitz, 1985).

from a few local studies (e.g., Ibrahim, 1978; Otterman and Tucker (1985) provides some confirma-
Wendler and Eaton, 1983). Recent papers which have tion of the ear]]er result.
attempted to verify the claims of anthropogenic One of the shortcomings of the feedback hypothe-
desertification (Hellden, 1984; Olsson, 1983, 1985) sis is that the proposed feedback mechanisms do not
could not do so and have also demonstrat,ed the adequately provide a "memory"from one year to the
importance of rainfall fluctuations in altering next. If the ]and surface acts to promote multi-
surface characteristics. Similarly, Gornitz (1985), year drought, the surface characteristics respon-
in a monumental study of anthropogenic vegetation sible for the feedback must be related to rainfall
change over West Africa during the last century, in the previous year. This does not seem to be
has shown that only small changes of surface albedo true for soil moisture, since the soil is dry
have accompanied it (Fig. 13). throughout the the dry season. Nor does albedo
Questions concerning the physical impact of sur- show the anticipated relationship with rainfall of
face change also remain. For example, the assumed the previous year.
increaseof Sahelianalbedoaccompanying droughtor The surface-relatedparameterwhichappears to
a denuded land surface (Charney,1975) cannot be changemost consistently in responseto Sahel
firmly established. The albedo of "protected" rainfall is dust production. Prosperoand Nees
patchesin Niger and two areas of southern Tunisia (1977, 1986) haveshownthat African dust, whichis
(.34, .36 and .26 respectively) is considerably transported across the Atlantic and Caribbean,
higher than in surrounding overgrazedareas (.42, appears at Barbados (WestIndies) in dramatically
1983). Courelet al. (1984), combiningtheir own increased concentration during years of Sahel
albedo measurements with those of Norton et al. drought (Fig. 15). The correlation with Sahel
(1979), haveshown that albedoincreased signill- rainfall in the previousyear is exceedinglyhigh.
cantly as the drought p¾ogressed
from 1967 to 1973, The dust influences atmospheric stability, its
going from -.31 to -.38 in the more northern Sahel presence contributing to the formation of inversion
(Fig. 14). However, as the drought continued in layers above the Sahel/Sahara (Prospero and Carl-
the late 1970s, albedo decreased to values well son, 1972; Carlson and Prospero, 1972). Unfortu-
below those prior to the drought. Surface tempera- nately, the impact of dust on atmospheric dynamics
ture changes associated with decreased vegetation over West Africa has not been extensively modelled
are likewise complex and controversial. Otterman but is potentially quite significant (see Coakley
(1974) reported lower radiative temperatures over and Cess, 1985). This phenomenonhas received more
the overgrazed Sinai than in the Negev, where attention in recent years, in part because of the
grazing is more controlled. This result was chal- French ECLATSexperiment, and a number of studies
]enged by Jackson and Idso (1975), amongothers, as of dust mobilization and transport and its radia-
being a manifestation of the lower emissivity of five effects have appeared (Carlson and Benjamin,
the more denuded surface. A more recent study of 1980; ben Mohamedand Frangi, 1983, 1986; Druil-
Geophysical Monograph Series Understanding Climate Change Vol. 52

94 AFRICAN DROUGHT' CHARACTERISTICS, CAUSAL THEORIES

from regiona] factors. Also, such droughts have

0.4 i i ! i i ! I I i I I I 1
occurred in the historical past, long before the
Ferlo human impact on the ]and became significant. How-
0.3
ever, there is considerable support for the idea
that ]and surface processes can intensify and pro-
long a Sahel drought initiated by anomalous large-
0.2 scale atmospheric forcing. The surface changes are
likely a response to the initial decline in rain-
fa]] (as much as 50%), rather than human activi-
0.1 ! ! I I I I ! I I I I I •' • ties. These surface changes then reinforce the
atmospheric conditions which reduced rainfall, thus
perpetuating the drought. The 20-year duration of
the current episode, a highly unusual occurrence
0.3 .ß C-•• elsewhere, is often cited as evidence of such feed-
back. Nevertheless, the idea that Sahel droughts
0.2 ß ß are self-reinforcing as a result of land-atmosphere
o ß feedback remains to be proven.
A real understanding of the role the land-
surface plays in promoting drought in the Sahel
requires a quantified description of changes of
0.5 I i i I I I I I i I I I i land-surface properties accompanying drought or
human-induced surface modification. The most rele-
vant parameters are surface albedo, temperature,
0.4 vegetation cover and soil moisture (Bolle and
Rasool, 1985), and the related fluxes of latent and
sensible heat. Establishing these on a regional or
0.3 continental scale would be nearly impossible with
conventional observational techniques, but remote-
sensing methodologies are currently being developed
0.2 which will soon permit such large-scale monitoring
of the land surface. These are reviewed by Nichol-
2Ox !6 ø •
son (1989). Recent papers applying these to Africa
0.1-- include studies of albedo (Rockwood and Cox, 1978;
Noffton et al., 1979; Courel et al., 1984' Pinty
I I I I ! I I I I I i ! 1
and Szejwach, 1985; Pinty eta]., 1985; and Pinty
68 70 72 74 76 78 80 and Tanr•, 1986), surface temperature, thermal
inertia, outgoing radiation and evapotranspiration
Year (-1900i
(Duvel and Kandel, 1985; Abdellaoui et al., 1986),
vegetation cover (Tucker eta]., 1983, 1985a,b;
Fig. 14. Albedo history of the western Sahara 45
and Sahel between 1967 and 1979. Curves 1-4 are
integrated albedo values for four strips 2' wide in 40

latitude, from 16-0' W: 1 : 20-22'N, 2 : 18-20'N,

3 : 16-18øN, 4 : 14-16øN (from Courel et al.,
1984).
3 30

• 25
Reprinted by permission from Nature, Vol.307, p.530.
Copyright 1984, MacMillan Magazine Limited.
•z 20

her et al., 1982; Druilhet and Durand, 1984;

Guedalia et al., 1984; Reiff et al., 1986; •,o
d'Almeida, 1986, 1987; Fouquart et al., 1987a, b;
Westphal et al., 1987; and Helgren and Prospero, 5

1987). 0
Currently, few scientists accept the original 165188187188180170171172173174175178177178171
le01811e2183184
hypothesis that human-induced land surface changes
caused the recent drought in the Sahel. Several
points of evidence contradict such a suggestion. Fig. 15. Monthly meantrade wind mineEal .aerosol
For one, strongatmospheric
teleconnections
between concentrationat Barbados(Units 10-Vg/m•; from
the Sahel and other African locations are evident Prospero and Nees, 1986, compare with Figure 4).
and droughts often occur more or less synchronously
throughout the continent. Secondly, the Sahel Reprintedby permissionfrom Nature, Vol.320, p.736.
droughts are simply too large in scale to result Copyright 1986, MacMillan MagazineLimited.
Geophysical Monograph Series Understanding Climate Change Vol. 52

N I CHOLSON 95

McGinnis and Tarpley, 1985), and hydrology (Schnei- size or intensity of rain-bearing disturbances and
der et al., 1985; McCauley et al., 1982). how these are influenced by aerosols. Thus, fore-
casting drought in the Sahel is considerably more
complex than elsewhere in Africa since two very
Implications for Forecasting and different types of forcing must be evaluated. It
Future Research will be quite some time before these issues are
resolved.

To improve our ability to forecast droughts in Summaryand Conclusions

Africa, several lines of investigation must be
pursued. First of all, the links between synoptic During the last two decades, most of the African
or regional scale phenomenon and planetary-scale continent has experienced extensive, severe and
circulation must be better established. This is prolonged droughts. The area most affected has
particularly true for the Sahel and for East been the semi-arid subtropical region of the nor-
Africa, where the factors which modify the rain- them hemisphere. While recent droughts have been
bearing disturbances from year-to-year are poorly the worst of the century, such events are an inher •
understood. This precludes relying solely on numer- ent part of the long-term climatic history of
ical models to evaluate interannual variability of Africa and have occurred in past centuries. Des-
rainfall. A related need is further study of the pite numerous claims that the Sahel drought was
characteristics of atmospheric circulation and anthropogenically induced via human impact on the
weather systems over the African continent during land surface, at present no evidence exists to
wet years and dry years. Requisite to such study support such a hypothesis. A vast number of numer-
is increased acquisition and improved dissemination ical simulations do, however, suggest that land
of rainfall and upper air data. Finally, the tele- surface changes, if sufficiently large and exten-
connections with the global tropics must be further sive, can influence large-scale climate, especially
explored, particularly with quasi-global phenomena in the Sahe]. Currently these are viewed as
like ENSOand with the tropical Atlantic and Indian potential factors in perpetuating rather than ini-
Oceans. Pending the results of such investigations, tiating drought, the ultimate trigger being some
forecasting schemes might be developed. These may change in the large-scale atmospheric dynamics.
soon be feasible for equatorial and southern For the Sahel, little is known about the actual
Africa. causes of drought. Elsewhere in Africa, the synop-
For the Sahel, however, reliable methods to tic situations associated with dry years are fairly
forecast drought cannot be developed until the well known, but larger scale factors controlling
question of land surface feedback is resolved. If these are not thoroughly understood. Research has
it is assumed that large-scale factors initiate a centered on sea-surface temperature fluctuations
drought and that the resultant land surface changes and atmospheric circulation and/or stability chan-
reinforce the initial atmospheric forcing, the ges. The latter might be caused by sea-surface
character of a given rainy season represents the temperature fluctuations but could also occur inde-
interplay of these two types of forcing. Presuma- pendently.
bly, then, a drought would continue until an atmos- In the absence of comprehensive knowledge of
pheric perturbation favoring rainfall occurs and is drought-producing mechanisms over Africa, it is
sufficiently strong to overcome the surface-imposed presently impossible to forecast rainfall f]uctua-
drouGht-promoting feedback. To forecast drought in tions. Thus, the question as to whether the cur-
the Sahel, it must first be definitively estab]i- rent drought situation will persist cannot be
shed whether the land surface does play such a role answered. In view of the linkages with tropical
and, if so, through what mechanism. So far, this phenomenonlike the Southern Oscillation, a poten-
has been attempted primarily through numerical tial exists for short- or long-term forecasting in
models and observational evidence of feedback is some regions if the physical mechanisms producing
lacking. the statistically established associations can be
A major weakness of the theory is that the pro- determined. In such a case, forecasting schemes
posed feedback mechanisms, e.g. soil moisture or might be developed, such as those of Nicholls
vegetation, do not adequately provide the requisite (1983), Shukla and Paolino (1983), Shukla and
"memory" from one year to the next. In order to Moo]ey (1987), and Hastenrath (1987) for monsoon
determine the appropriate variables and resolve the rainfall and Hastenrath (1984b) for Northeast
question of "memory", observational studies of such Brazil.
non-atmospheric factors as soil erosion and struc-
ture or efficiency of plant growth must be carried Acknowled•ments• This paper was presented at
out. The best dependence on antecedent years so the IUGGin Vancouver, AuGust 1987. I would like
far appears to be atmosphericdust (Prospero and to acknowledgethe contributions of the FSU Foun-
Mees, 1986). A numberof studies are assessing the dation and the National Science Foundation (Grant
impact of dust on the atmosphere, but the missing ATM 86 14208) in providing travel funds to attend
]ink to rainfall variability is an understanding of the meeting. Laura Easter, Andrew Lare and Ada
the atmospheric factors (e.g., static or dynamic Malo assisted in the preparation of this manu-
stability) •overning such parameters as the number, script. FiGures were drafted by Dewey Rudd.
Geophysical Monograph Series Understanding Climate Change Vol. 52

96 AFRICAN DROUGHT' CHARACTERISTICS, CAUSALTHEORIES

References ben Mohamed,A. and J.P. Frangi, ]986' Results

from ground-based monitoring of spectral aerosol
optical •hickness and horizontal extinction' some
Abdellaoui,A., F. Becker and E. Olory-Hechinger, specific characteristicsof dustySahelianatmos
1986' Useof Meteosatfor mapping thermalinet- pheres. Journalof ClimateandAppliedMeteoro]
tia and evapotranspirationover a limited region 9qY, 2•, 1807-1815.
of Mall. •Q••. Q• C•g•9 gg• •RP•D• Meteor- Bolle, H.J. andS.I. Rasool, ]985' Developmentof
o•9Gy, •5, 1489-1506. the implementationplan for the International
Acharya,U.R. andN.S. BhaskaraRao, 1981' Meteor- Satellite Land- Surface Climatology Projec•
PlQg3 O• Zambi• P•rt •. Government Printers, (ISLSCP). PhaseI' WorldMeteorologicalOrgani-
Lusaka,49 pp. za•ion, WCP34, Geneva,56 pp.
Adamec,D. andJ.J. O'Brien, 1978' The seasonal Bunting, A.H., M.D. Dennett,J. Elston, andJ.R.
upwelling in the Gulf of Guinea due to remote Milford, 1976' Rainfall trends •n the Wes[
forcing. •Q•Kna• Q• ph••_ Q•ea•Q•a_lq.•i,
•, African Sahel. QuarterlyJournal of the R•yal
1050'ø]060. MeteorologicalSociety, •02, 59.64.
Agumba,F., 1984' Climatologicalfluctuations of Butzer, K.W., G.L. Isaac, J.L. Richardson,and C.
spring-rainsin Kenyain relation to large-scale Washbourn-Kamau, 1972' Radiocarbon dating of
circulations. E•9•dpd A•$•raq• 9f P.•P_e•$Prel East African lake levels. Science,175, 1069-
sented • th__eWMO
Reg•_pD_•_•
S•pD•• C9a•9•7 1076.
9B•e 9n GATE• •_.A•4__•X
ag• •rQpical •tcQrolog_y, Cadet,D.L. and S.H. Houston, 1984' Precipitable
WorldMeteorologicalOrganizationTropical Mete water over Africa and the Eastern/Central Atlan-
orologyProgramme ReportSeries No. 16, Geneva, tic Oceanduring the 1979 Summer.Journalof the
258-262. M•teQrg•pG•qa! Soc• of J•pag, 62, 761-1774•
Anthes,R.A., 1984' Enhancementof convectivepre- Cadet,D.L. and N.O. Nnoli, ]987' Water vapour
cipitation by mesoscalevariations in vegetative transport over Africa and the Atlantic Ocean
coveringin semiaridregions. •ogfnal 9• •im•9 during Summer ]979. Quarterly.Jpg.Eng.1
of the
a• A•p•ie• Mptpq•o•, •, 541-554. R9• •9•99•9•9•a• SQ•_•e•3,1•, 581-602-•
Anyamba, E.K., 1983' On the monthlymeantropo- Car]son,T.N. and J.M. Prospero,1972' Thelarge
spheric circulation and the anomalous
circulation scale movement
of Saharanair outbreaksover the
during the ]961/62floodsin EastAfrica. M.S. northernequatorialAt]antic. Journalof
Thesis, Department
of Meteorology,
University of Me.•p9•919•,11, 283-297. -
Natrobt. Car]son,T.N. and S.G. Benjamin,1980' Radiative
Anyamiga,
E.K., 1984' Someaspectsof the origin of heating rates for Saharandust. Journalof the
rainfall deficiencyin East Africa. E_x_•pn••- •9•pheric Sciences,•7, 193-213. -.........
•r.act$ o.• •pers Prp•9•9• •% th• W•9Reg%ona]CharLey,J.G., i9•5'- Dynamics of desertsanddrou-
Scientific •PD•rPg•9 9n •T• W_.A•__•
a• •EQpical ght in the Sahel. Q•r•_•rl3 •.9•rg• of the Royal
Me•99•plogy,World MeteorologicalOrganization Meteorologkq•.•. •.•i9•t, 1_0_•,
193-202.
TropicalMeteorology Programme ReportSeries No. Charney, J.G., P.H. Stone and W.J. Quirk, 1975'
16, Geneva,110-113. Droughtin the Sahara' A biogeophysical feedback
Anyamba,E.K. andL.J. Ogallo, 1986' Anomalies in mechanism. •_•½Ec•,!.8•, 434-435.
the windfield over Africa duringthe East Afri- Charney,J.G., W.J. Quirk, S.H. Chow andJ. Korn-
can rainy seasonof 1983/84. Lonq--Range
Forecas- field, 1977' a comparative studyof the effects
t•.n• •ese•rch. RpPOr••er•9• •0. •, !, World of albedochange
ondroughtin semiarid regions.
MeteorologicalOrganization,Geneva,128-131. Journal of the AtmosR•c Sciences,•j, 1366-
Arkin, P.A., 1982' The relationship between inter- 1385'•
annualvariability in the 200-mbtropical wind Chen,T..C., and H. VanLoon, 1987' Interannual
field and the SouthernOscillation. •o•.•_•3 variationof the TropicalEasterlyJet. Month]•
WeatherReview,•.•Q, 1393-1404. Weather•9•i9•, 1739-1759.
Bhalotra,Y.P.R., 1973' Disturbances
of the summerCoakley,J.A. andR.D. Cess,1985' Response
of the
seasonaffecting Zambia.TechnicalMemorandum
No. NCARcommunity climate model to the radiative
•, ZambianMeteorologicalDepartment. forcingby the naturally occurringtropospheric
Bhalotra, Y.P.R., 1973' Disturbancesof the sun,
net aerosol. Journalof the Atmospheric
Sciences,•,
season affecting Zambia. Technical Memorandum
No. 1677-1692.
•, Zambian
MeteorologicalDepartment,
7 pp. Cockcroft, M.J., M.J. Wilkinson and P.D.Tyson,
Beer, T., G.K. Greenhut and S.E. Tandoh, 1977' 1987' The application of a present-day climatic
Relations between the Z criterion for the sub- model to the Late Quaternary in Southern Africa.
tropical high, Hadley Cell parameters and rain- •_••q Cha•qe, !•, 161-181.
fall in northern'Ghana. Monthl• •pA•her Review, Courel, M.F., R.S. Kandel and S.I. Rasool, 1984'
1__0•,849-855. Surface albedo and the Sahel drought. Nature,
ben Mohamed, A., and 3.-P. Frangi, 1983' Humidity }.Q7, 528-531.
and turbidity parameters in Sahel' A case study d'Almeida, G., 1986' A Model for Saharan dust
for Niamey (Niger). Journal 9• Climate and •p__- transport. Journal 9• Climate and Applied Met-
plied Meteorolou¾, 22, 1820-1823. eoroloqy, •, 903-916.
Geophysical Monograph Series Understanding Climate Change Vol. 52

NICHOLSON 97

d'Almeida, G., 1987' On the variability of desert sols and vertical radiative flux divergence.
aerosol radiative characteristics. J•lurnal of •9•• 9• Cl•m•9 and •pp.• Mp•9_9•g!sgy, •,
Geophysical Research, 92, 3017 3026. 38-52.
Deacon, J. and N. Lancaster. ]988' Late Quaternar3 Glantz, M.H., ed., 1977' •.9•eJ_t•%q•ODl
Palaeoenvironments of Southern Africa. Clarendon •9•%•! •9g•a•a•q• k• and •_o_u_D• Arid Lands.
P•ess, Oxiurd, 225 pp. Westview Press, Boulder, 346 pp.
Dennett, M.D., J• Elston and J.A. Rodgers, 1985' A Gornitz, V., 1985' A survey of anthropogenic vege-
reappraisal of rainfall trends in the Sahel. tation changes in West Africa during the last
Jou[nal q• •l•matology, •, 353-362. century--climatic implications. Climatic Change,
Druilhe•, A. and P. Durand, 1984' Etude de la 7, 285--326.
couche limite convective Sah•lienne en presence Greenhut, G.K., 1977' A new criterion for locating
de brumes seches/Experience (ECLATS). L_• M6t•_9•- the subtropical high in West Africa. Journal
9•9•9, 51-77. •P•!•9• •e•.9•9.lpg3, •, 727,734.
Druilhet, A., J.P. Frangi and P. Durand, 1982' Gregory, S., 1982' Spatial patterns of Sahelian
Donne•s statistiques sur le bilan d'energie de la annual rainfall. •b•¾es fuer
couche de surface Sah•lienne. La •t•oro19_g•e, •99pbysik und Bioklimatoloq•e, Ser. B., •, 273-
29--30, 227 237. 286.
Duvel, J.P. and R.S. Randel, ]985' Regional-scale Guedalia, D., C. Estournel, and R. Vehil, 1984'
diurnal variations of outgoing infrared radiation Effects of Sahel dust layers upon nocturnal
observed by METEOSAT. Jou•D• 9f •imate and cooling of the atmosphere (ECLATS) experiment.
•pP•e• M9•q•olog3_, •_•, 335-349. •9•r_n_•_[9• Climate and Appli__e_•
Meteorology, •,
Dyer, T.G.J., ]975' The assignment of rainfall 644-650.
stations into homogeneousgroups' An application Harangozo, S. and Harrison, M.S.J., 1983' On the
of principal componentanalysis. Qua•e•y J9u•i use of synoptic data indicating the presence of
na• of tb9 •9yal Meteor9•_qgical Society, 1_0•., cloud bands over southern Africa. South African
]005-1013. Journal of Science, 7•, 413-414.
Dyer, T.G.J., 1979' Rainfall along the east coast Harrison, M.S.J., 1983' The Southern Oscillation,
of southernAfrica, the SouthernOscillation and zonal equatorial circulation cells and South
the latitude of the subtropical high pressure African rainfall. Preprints of the 1st Inter-
belt. Qug[•9[_lyQQ.urng.)
of th9 Royal Meteoroloqœ. national Conference on SouthernHemisphereMet-
•q•l $Qci•3, !QS, 445--452. eorology, Americn Meteorological Society, Boston,
Dyer, T.G.J. and M.E. Marker, 1978' On the varia-
302-305.
tion of rainfall over southwest Africa. Th• •opt• Harrison, M.S.J., 1984' The annual rainfall cycle
•_[•q• G9og•phi•_ •o•a_l, •Q, 143-149. over the central interior of Africa. South
Fleer, H., ]98]' Large-scale tropical rainfall •r•½•n •9grn• of •c%•nqg, •õ, 47-64.
anomalies. Bonner Meteorolog. ische Abhandlun99_gn, Hastenrath, S.L. and E.B. Kaczmarczyk, 1981' On the
•, 114 pp. spectra and coherence of tropical climate anoma-
Flohn, H. 1964' Investigations on the Tropical lies. Tg.•l•, •, 453-462.
Easterly Jet. BonnerMeteorolog_•b••handlun- Hastenrath, S.L., 1984a' Interannualvariability
•en, 2•, 83pp. andannualcycle' mechanisms of circulation and
Flohn, H., 1988' East African rains of 1961/2 and climate in the tropical Atlantic sector. Monthly
the abrupt change of the WhiteNile discharge. W9•9• R9v%DW, •, 1097-1107.
Palaeoecoloq3
of Africa, •, in press. Hastenrath,S., 1984b'Predictability of north-east
Foiland,C.K., D.E.Parker,M.N.WardandA.W.Col- Brazil droughts.N•ure, •Q!, 531-533.
man '
1986a' Sahelrainfall' Northern
Hemisphere
Hastenrath, S., anomalies'
soon rainfall
1987' Predictability of Java
A case study.
mon-
Journal
circulationanomalies
andworldwide
seatempera- 9• •l•m•t••n• •PP•9••9t99_rp1993,
•6, 133--141.
turechanges.•ong-Range
Forecastinq
andClimateHelgren,D.M.andJ.M.Prospero,-1987'
Windveloc-
Research,!, Meteorological Office, Braeknell. fries associated with dust deflation events in
Folland, C.K., T.N. Palmer, andD.E. Parker, 1986b' the westernSahara. Journal of Climate and App-
Sahel rainfall and world-widesea temperatures lied Meteorology, .•, 1147-1151.
1901-85; Observational,modelling,and simula- He•i•en, U., 1984' Droughtimpactmonitoring' a
tion studies. Nature, 32•, 602-607. remotesensingstudyof desertification in Kordo-
Fouquart,Y., B. Bonnel,G. Brogniez, J.C. Burfez, fan, Sudan. LundsUniversitets Naturgeografiska
L. Smith, J.J. Morcrette and A. Cerf, 1987a' Institution, •, 61 pp.
Observationsof Saharan aerosols' Results of Hirst, A.C. and S. Hastenrath, 1983a' Atmos-
ECLATSfield experiment. Part I' Optical thick- phere-oceanmechanisms of climate anomalies in
nessesandaerosolsize distributions. Journalof the Angola-tropicalAtlantic sector. Journal
ClimateandAppliedMeteorology,26, 28-37. •hysical Oce•n_9o•faphl,
•, 1146-1157.
Fouquart, Y., B. Bonnel, M. Chaoui Roquai and R. Hirst, A.C. and S. Hastenrath, 1983b' Diagnostics
Santer, 1987b' Observations of Saharan aerosols' of hydrometeorological anomalies in the Zaire
Results of ECLATS field experiment Part II' (Congo) basin. Quarterly Journal of the Royal
Broadband radiative characteristics of the aero- Meteorological Socie•3, •9•, 881-892.
Geophysical Monograph Series Understanding Climate Change Vol. 52

98 M•!tlC• D]tO•G•T' CfI•CTERISTICS, CAUSAL THEORIES

]{slung, J. and R. Newell, 1983' The principal non- 1986' The Southern Oscillation and South African
seasonal modes of variation of global sea-surface rainfal 1. _so_ _u_t__h
African Jou__r__n_a_l .of Science
temperature. Journal of ._P_h_ysical
Ocea_n_o_g_ra•_ht, 196-198.
1__3,1952-1967. Lough, J.M., 1986' Tropical Atlantic sea-surface
Ibrahim, F., 1978' Anthropogenic causes of deserti- temperatures and rainfall variations in Sub-
fication Jou_r_n_a_l_, saharan
in Western Sudan. Ge_oir_a_ph•i_½a_l_ Africa. Mo__nth!_yWe_ a__th_e_
r_ Re_v._iew,_1_14,
23, 243-254. 561-570.
Jackson, R.D. and S.B. Idson, 1975' Surface albedo Madden, R.A., 1987' Seasonal variations of the
and desertification. Science, 189, 1012-1013. 40-50 day oscillation in the tropics. Journal
Kanamitsu, M. and T.N. Krishnamufti, 1978' Northern the _Atmo_sp•_er.i_c.
Sc_iences, 43, 3138-3158.
summer tropical circulations during drought and McCaul ey, J.F., G.G. Schaber, C.S. Breed, M.J.
normal ra]nfal 1 months. ea_t_her R_e_v_iew, Grolier,
Mpn_.t._h_ly
_W. C.V. Haynes, B. Issawi, C. Elachi, and
_1_06, 331-347. R. Blom, 1982' Subsurface valleys and geoarchae-
Kiangi, P.M.R. and J.J. Temu, 1984' Equatorial ology of the eastern Sahara revealed in shuttle
westerlies in Kenya. Are they all rain laden? radar. Science, _218_,1004-1017.
Ex_t.e_nded a_b__st_ra_c_t_sof kape_r_sp_r_e_s__e__nt_e_d at_ th_e_WM_O_ McGinnis, D.F. and J.D. Tarp]ey, 1985' Vegetation
Reg_ion_alS½ientif_ic c_o.n__fer_egc_e on GATE, •AME_X_a_n_d cover mapping from NOAA/AVHRR. A_dv_a_n__c_e_s i_n_S_pa_ce
T_r_.opicalm_e_t_e__o__ro..lo_G3 in_ A_f_r
i__c
a_, World Meteorolog- Research, 5, 359-369.
ical Organization Tropical Meteorology Programme Miles, M.K. and C.K. Folland, 1974' Changes in the
Report Series No. 16, Geneva, 144-146. latitude of the climatic zones of the Northern
Kidson, J.W., ]977' African rainfall and its rela- Hemisphere. N_a_•_u_r_e_,_2_5_2,616.
tion to the upper air circulation. Minja,
Qu_a_rt_e__r__1I W.E.S., ]985' A comparative investigation
_JQu_rn__a_! of the _Roya_! Met_e_or_p.l_9g_i__c_a_l .S.
oci etj, of weather anomalies over West Africa. M.S.
103, 441-456. Thesis, Department of Meteorology, University of
Klaus, D., 1978' Spatial distribution and periodi- Nairobi.
city of mean annual precipitation south of the Mintz, Y., 1984' The sensitivity of numerically
Sahara. Archives _fu__er M_eteorol_pG•_e_•g_eo•p_blsi_k un_d simulated climates to land-surface boundary con--
Sioklimatolo_g_i_e_ S_e__r_. B_., 2_6, 17-27. ditions. The G]9b_al Cl.ima_te, J.T. Houghton, Ed.,
Knuison, T.R., and K.M. Weickmann, 1987' 30-60 day Cambridge University Press, 70-105.
atmospheric oscillations' composite life cycles Motha, R.P., eta]., 1980' Precipitation patterns
of convection and circulation anomalies. Monthly [n West Africa. Month_ly Weathe_r Rev.ie_w, 1QS_,
1567- 1578.
Weat_h_e_r_R e_v i _e_w_
, _1.1_5_,1407- 1436.
Knutson, T.R., K.M. Weickmann and J.E. Kutzbach, Namias, J., 1974' Suggestions for research leading
1986' Global-scale intraseasonal oscillations of to long-range precipitation forecasting for the
outgoing longwave radiation and 250 n•b zonal wind tropics. Preprints of the Internat_i_ona] Tr_op_ical
during northern hemisphere summer. Month_l_yW_e_az Meteorol_ogy Meeting, Nairobi, American Meteoro-
th_e_• Rev.i_ew, 114. logical Society, 141-144.
Kraus, E.B., 1977a' Subtropical droughts and cross- Newell, R.E. and J.W. Kidson, 1979' The tropo-
equatorial energy transports. MQ_n•hJy W_e__a_ther spheric circulation over Africa. Saharan Dust
Review, 10_5, 1009-1018. (Scope Report 14), C. Moral es, Ed., Wiley and
Kraus, E.B., 1977b' The seasonal excursion of the Sons, ]33-370.
intertropical convergence zone. Monthly th_e_r. Newell,
W_e_a. R.E. and J.W. Kidson, 1984' African mean
Rev.iew, _105, 1009-1018. wind changes in Sahelian wet and dry periods.
Krueger, A. and J.S. Winston, 1975' Large-scale Journal of Climatology, 4, 1-7.
circulation anomalies over the tropics during Nicholls• N., 1983' Predicting Indian monsoon
1971-72. Monthly Wea_ther Re_v_i_gw, _10_3_, 465-473. rainfall from sea surface temperature in the
Kumar, S., 1978' Interaction of upper _westerly Indonesia North Australia area. Natu_•re, 306,
waves with intertropical convergen_ce MoDe and 576-577.

the_if e_f_fect on the weather ov_e_rZanib_ia _during Nicholson, S.E., ]978' Climatic variations in the
the rainy season. Government Printers, Lusaka, Sahel and other African regions during the past
36 pp. five centuries. Journal of Arid Environments, 1,
Lamb, P.J., 1978a' Case studies of tropical Atlan-- 3-24.

tic surface circulation patterns during recent Nicholson, S.E., ]979a' The methodology oi histor
sub-Saharan weather anomalies' 1967 and 1968. ical climate reconstruction and its application
Mon_th_lyWeather Review, _106, 482-491. •o Africa. Journal of African History, 20, No. 1,
I. amb, P.J., 1978b' Large-scale tropical Atlantic 31 49.

surface circulation patterns associated with Sub- Nicholson, S.E., 1979b' Revised rainfall series for
saharan weather anomalies. Tellus, 30, 240-251. the West African subtropics. Monthly weat_her
Lamb, P.J., 1982' Persistence of Sub - saharan R__evi_ew_, _107, 620 623.
drought. Nature, 299, 46-47. Nicholson, S.E., 1980a' Saharan climates in his-
Lamb, P.J., 1983' West African water vapor varia- toric times. Tb_e_Sahara an.d_ the N_i•.9, M.A.J.
tions between recent contrasting Subsaharan rainy Williams and H. Faure, Eds., A.A. Ba]kema, Rot-
seasons. Te_l._lus., 3__5_A_, 198-212. terdam, 173-200.
Lindesay, J.A., M.S.J. Harrison and M.P. Haffner, Nicholson, S.E., ]980b' The nature of rainia]]
Geophysical Monograph Series Understanding Climate Change Vol. 52

N I CHOL$ON •

fluctuations in sub-tropical West Africa. •gD• fication, LundsUniversitets Geografiska Institu-

•t•r Review, •08, 473-487. tion Avhandlingar, XCVIII, 170 pp.
Nicholson, S.E., 1981a' The historical climatology Otterman, J., ]974' Baring high-albedo soils by
of Africa. Clima• and Hi•ry, T.M.L. Wigley, overgrazing' a hypothesized desertification mech-
M.J. Ingram and G. Farmer, Eds., Cambridge Uni- anisms. Science, 186, 53]-533.
vetsfry Press, Cambridge, 249-270. Otterman, J., 1981' Satellite and field studies of
Nicholson, S.E., 1981b' Rainfall and atmospheric man's impact on the surface in arid regions.
circulation patterns during drought periods and Tellus, 33, 68-77.
wetter years in West Africa. Monthly Weath• Otterman, J. and C. J. Tucker, 1985' Satellite
Review, 109, 2]9]-2208. measurements of surface albedo and temperatures
Nicholson, S.E., 1982' T• •a•]• • C•im•tic Per- in semi-desert. Jogrn•l 9f •i• •d •l•ed
spgctive. Club du Sahel, Paris, 80 pp. M_e_•goro•oqy,
•, 228-234.
Nicholson, S.E., 1983' Sub-Saharan rainfall in the Palmer, T.N., ]-986' The influence of the Atlantic,
years 1976 80' evidence of continued drought. Pacific, and Indian Oceans on Sahel rainfall.
Monthly Weather •gvie•, •, 1646-1654. Nature, •22, 25]-253.
Nicholson, S.E., 1985' Sub-Saharan rainfall ]981- Pinty, B., G. Szejwach, and J. Stum, 1985' Surface
1984. Joufq•_• of Climate and A•p•e• Meteo•9•Ogy, albedo over the Sahel from METEOSATradiances.
24, 1388-1391. Jou•gal of C1%•9 •nd ARplied Meteoro.ko__,q/,
•.•,
Nicholson, S.E., ]986a' The spatial coherence of 108--113.
African rainfall anomalies' interhemispheric Pinty, B. and G. Szejwach, ]985' A new technique
teleconnections. Journal q• Climat9 an• •pp!• for inferring surface albedo from satellite
Mg•e9•9•9•l, •_•, 1365-1381. observations. •gur• o• C•• •nd •pplied
Nicholson, S.E., ]986b' The nature of rainfall M•eo•lQgl, •4, 741-750.
variability in Africa south of the equator. Pinty, B. and D. Tanr•, ]987' The relationship
Journal of Climato]pgy, 6, 5]5-530. between incident double-way transmittances' an
Nicholson, S.E., 1986c' African drought' an exam- application for the estimate of surface albedo
pie of the influence of land surface properties from satellites over the African Sahel. Journal
on climate? ISLSCP' Pa•metgr•zation Q• Land-- Of Climate and Applig• •e•_9]o__qJ_, •, 892-896.
Surface Ch•racteri.•t•½sj Use of •atel]•te Dgta Prospero, J.M. and T.N. Carlson, 1972' Vertical
in •.•m•e stud•e•j F•r• •gsu!t• q• !.•CP. and areal distribution of Saharan dust over the
European Space Agency, Brussels, SP-248, 405-410. west equatorial North Atlantic Ocean. Journal of
Nicholson, S.E., 1988' Land surface-atmosphere Geophysical Research, 77, 5255.5265.
interaction' physical processesand surface chan- Prospero,J.M. and R.T. Nees, 1977' Dust concen-
gesandtheir impact. P•oGress
ig Pbt•%•g_lGeo- trations in the atmosphere
of the of the equator--
graphy,12, 36-65. ial NorthAtlantic' possiblerelationshipto the
Nicholson,S.E., 1989' Remotesensing of land Sahe]iandrought. Science,]96, ]]96.]198.
surfaceparametersof relevanceto climatestu Prospero,J.M. and R.T. Nees,1986' Impactof the
dies. •_•grg• •_nPhy.•a_•ggg•pht, in press. NorthAfricandroughtandE1 Ninoonmineraldust
Nicholson,
S.E. andD. Entekhabi,
1986' Thequasi-- in the Barbados
tradewinds. N•qf9, $•Q, 735-
periodic behavior of rainfall variability in 738.
Africa and its relationship to the SouthernRasmusson,
E.M. andT.H. Carpenter,1982' Varia-
Oscillation. Arch•vg• •9• MeteorolqGy_•
G99Pblz tions in tropical seasurface temperatureand
s•cs •nd •pclimat.919g_y•
•er. A., 34, 311-348. surface windfields associatedwith the Southern
Nicholson, S.E. and D. Entekhabi, 1987' Rainfall Oscillation/E1 Nifo. Monthly Weather Review,
variability in equatorial and Southern Africa' 110, 354 384.
relationships with sea-surface temperatures along Rasmusson, and J.M. Wallace, ]983' Meteorological
the southwestern coast of Africa. Journal of aspects of the El Nino/Southern Oscillation. Sci-
Climate and Applied Meteg•olgg3, •6, 561-578. ence, 222, 1195--1202.
Nicholson, S.E. and H. Flohn, 1980' African envir Riehl, H., 1979' Cl•mat9 and Weather in the TrQp-
onmental and climatic changes and the general its. Academic Press, New York, 61] pp.
atmospheric circulation in ]ate P]eisi. ocene and Reiff, J., G.S. Forbes, F.T.M. Spieksma, and J.J.
Holocene. Climatic Change, •, 313-348. Reynders, 1986' African dust. researching north-
Norton, C.C., F.R. Mosher and B. Hinton, ]979' An western Europe' a case study to verify trajectory
investigation of surface albedo variations during calculations. Journal of Climate •n• A•pl•ed
the recent Sahe] drought.. Journal of Applied Meteoro•Q•y,2•, 1543-1567.
Meteoro]g•, 18, 1252 1262. Rockwood, A.A. and S.K. Cux, ]978' Satellite in-
Oga]lo, L.J. and E.K. Anyamba, ]986' Droughts of frared surface albedo over north-western Africa.
tropical central and eastern Africa. •o•_-•gG9 Journal of the Atmospheric Sciences, 3_5, 513-572.
Forecasting Research Report Series No. 6, ), Rodhe, H. and H. Virji, 1976' Trends and period-
World Meteorological Organization, Geneva, 67-72. icities in East African rainfall data. Monthly
O]sson, n., ]983. Desertification or climate. Lund W•a%•r Review, 1O•, 307-315.
Studies in Geog[aphy, Set. A., Physical Geog- Ropelewsk], C.F. and M.S. Ha]pert, 1987' Global
raphy, No. 60, The University of Lund, 36 pp. and regional scale precipitation and temperature
Olsson, n., 1985' An integrated study of desertl- patterns associated with El Nino/Southern Oscil-
Geophysical Monograph Series Understanding Climate Change Vol. 52

1OO AFRICAN DROUGHT' CHARACTERISTICS, CAUSAL THEORIES

lation. MO8•y •9•h• Review, •, 1606-1626. Tanaka, M., B.C. Weare, A.R. Navato and R.E. New-
Schneider, S.R., D.F. McGinnis, Jr. and G. Step- ell, 1975' Recent African rainfall patterns.
hens, 1985' Monitoring Africa's Lake Chad basin •ture, 255, 20]-203.
with LANDSAT and NOAAsatellite daf.a. Inter- Tucker, C.J., C. Vanpraet, E. Boerwinkel and A.
•9•.• J9•g•.• q• Rem9•e Sensing, •, 59-73. Gaston, 1983' Satellite remote sensing of total
Schupelius, G.D., 1976' Monsoon rains over West dry matter production in the Senegalese Sahel.
Africa. T..•J•u•, •8, 533-537. R9mote Sensing of Environment, 13, 461-474.
Schultze, G.C., ]983' 'N Moontlike verband tussen Tucker, C.J., C. Vanpraet, M.J. Sharman and G. Van
die Suidelike Ossilasie/E1Nino verskynsels en Ittersum, 1985a' Satellite remote sensing of
droogtes oor die somerreenva]streke van Suid- total herbaceous biomass production in the Sene--
Afrika. 'N Voorlopige studie. Sgut• •_f•fg8 galese Sahe]. Remote Sensing of Envirqnment, ]7,
,

W•9• B•r9• •9wsle•er, 4•9, 79-84. 233-249.

Selkirk, R., 1984' Seasonally stratified correla- Tucker, C.J., J.R.G. Townsbend and T.E. Golf,
tions of the 200 mb tropical wind field to the 1985b' African land-cover classification using
Southern Oscillation. Jourga_• of Climatoloq•, •, satellite data. Science, 227, 369•375.
365-382. Tyson, P.D., 1980' Temporal and spatial variation
Semazzi, F.H.M., V. Mehta and Y.C. Sud, 1988' An of rainfall anomalies in Africa south of latitude
investigation of the relationship between sub- 22 ø during the period of meteorological record.
Saharan rainfall and global sea surface tempera- Climatic Change, 2, 363-371.
tures. A•m9•phere-Ocean, 2•, 118-138. Tyson, P.D., T.G.J. Dyer and M.N. Mametse, 1975'
Shukla, J. and D.A. Mooley, 1987' Empirical predic- Secular changes in South African rainial]' ]880
tion of the summer monsoon rainfall over India. to 1972. Quarterly J9•fqa_l q• the
•o__nk•ly wga•9• Re•W, •, 695-703. o•og•c•] •9ciety, !0•, 817-833.
Shukla,J. and D.A. Paolino, 1983' TheSouthernTyson,P.D., 1984' Theatmospheric
modulation
of
Oscillation and long.rangeforecasting of the extendedwet and dry spells over SouthAfrica.
summer monsoon rainfall over India. Montini Journal of Climatology, 4, 621-636.
•ath9r Review, 1•, ]830-]837. United Nations, 1977' D9•f••_o_n_
Stoeckenius, T., 1981' Interannual variations of • C•gseguenges. Pergamon, Oxford, 448 pp.
tropical precipitation patterns. Monthl• •e•ther Walker, J. and P.R. Rowntree, 1977' The effect of
Review, !Q9, 1233-1247. soil moisture on circulation and rainfall in a
Street, F.A. and A.T. Grove, 1979' Global maps of tropical model. Q•a•9.•t •q•• • • •93al
lake-level fluctuations since 30,000 BP. Q•ter- •9t•Q..•pgical •Og•gty, •Q3, 29-46.
g•.r• Res9•ch, !2, 83-118. Walsh, J.E., W.J. Jasperson and B. Ross, 1985'
Street-Perrott, F.A. and N. Roberts, 1983' Fluc- Influences of snow cover and soil moisture on
tuations in closed basin lakes as an indicator of monthly air temperature. M•g•b•l Wg•h_e_l
past atmospheric circulation patterns. •riations 113, 756-768.
in the Global Water Budpet. A. Street-Perrott, Weickmann, K.M., G.R. Lussky and J.E. Kutzbach,
M. Beran and R. Ratcliffe, Eds., Reidel, Dor- 1985' Intraseasonal (30-60 day) fluctuations of
drecht, 331-345. outgoing longwave radiation and 250 mb stream
Street-Perrott, F.A. and S.P. Harrison, 1985' Lake function during northern winter. Monthl• Weather
levels and climate reconstruction. Palaeoclimate Review, •3.
and Model•igR, A.D. Hecht, Ed., John Wiley and Wendlet, G. and F. Eaton, 1983' On the desertifi -•
Sons, New York, 291-340. cation in the Sahel zone (Part 1' ground observa-
Taljaard, J.J., 1981a' The anomalous climate and tions). •lim• C•DD•, •, 365-380.
weather systems of January to March, 1974. South Westphal, D.L., O.B. Toon and T.N. Carlson, 1987'
Africa Weather •ure• Technical •a__pD• • .•, A two-dimensional numerical investigation of the
92 pp. synamics and microphysics of Saharan dust storms.
Taljaard, J.J., 1981b' Upper-air circulation, tem- •9•g• Q• g9Qp•3sical Research, •, 3027-3049.
perature and humidity over southern Africa. South Williams, M.A.J. and H. Faure, eds., 1980' The
Africa Weather •gr9• Technical •ap9• •9. •.O, Sahara and •.• •.9. A.A. Balkema, Rotterdam,
94 pp. 607 pp.
Geophysical Monograph Series Understanding Climate Change
SENSITIVITY OF CLIMATE
Duzheng Ye
Instituteof AtmosphericPhysics,AcademiaSinica,Beijing,China
Introduction
The importance of land-surface processes in affecting climate
change has been analyzed and discussedby Namias [1962,1963]. The
physics of the land-surface processesaffect the climate because the
ground hydrology, together with the vegetation and soil, determine
the surface moisture availability which, in turn, controls the partition
between the sensibleand latent heat fluxes [Rowntree, 1984] and also
the transfer of momentum. Further the vegetation cover and the soil
moisture content can determine the ground surface albedo in snowless
conditions. Therefore the heat balance and water balance in the plan-
etary boundary layer are highly influenced by the hydrological pro-
cesses.Through the planetary boundary layer, the influence of the
ground hydrological processes can be felt through the whole
troposphere [Yeh, et al., 1984; Rowntree and Bolton, 1983]. Here a
crucial factor is the soil moisture content. In the region of dry anoma-
lies, the following sequenceof events tends to occur: a decrease of
evaporation, a ground surface warming with an increase of sensible
heat flux, a warming of lower layers of the atmosphere with a decrease
of relative humidity (due to decreaseof evaporation and warming of
lower atmophere), a decrease of precipitation with a cooling of the
atmosphere higher up (due to decreaseof latent hea0, and then a
change of upper air circulation [Rowntree and Bolton, 1983]. It is the
decreaseof precipitation which will cause the initial dry anomaly to
persist. In the region of moist anomalies, the opposite sequenceof
events tends to occur.
Besidessoil moisture, the vegetation is another factor of importance
in the hydrological processes. The role of vegetation is shown
schematically in Fig. l from Dickinson [1984]. Simply speaking this
figure shows that the vegetation foliage can prevent part of the precip-
itation from reaching the ground; part of the intercepted water is
re-evaported and part of it drops to the ground. Part of the water
reaching the ground is partitioned into the soil and the rest appears as
runoff. The water in the soil passesdownward and may be tapped by
the roots from the ground water reservoirswhich will eventually feed
the streamflow as another kind of runoff. At the same time, plants ex-
tract water from the soil through their roots and move it into the at-
mosphereas transpiration. The canopy can drastically alter the energy
balance processesat the surface and so alter the ground evaporation
and snow melting. The plants can also significantly reducethe surface
albedo and thus increasethe solar radiation received at the ground (if
other factors are kept the same). Further the plants will increase the
Copyright 1989 by
International Union of Geodesy and Geophysics
and American Geophysical Union.
lol
Vol. 52
MODEL TO HYDROLOGY
surface roughnesswhich can influence the ground hydrological pro-
cesses.These processeshave been discussedin detail by Dickinson
[1984], Eagleson[1982] and others.With no vegetation,only the water
in a shallow layer of soil is available for evaporation and this can dry
out after a few days of high evaporation typical in the tropics and over
the central continents in summer. With deep rooted vegetation some
evaporation is likely to occur even during a long dry period [Rowntree
and Bolton, 1983]. Only through vegetation can water in the deep soil
layers be connected with the atmosphere and take part in the
hydrological processes.
Becauseof the importance of the hydrological processesin climate,
a series of simulation experiments on the sensitivities of climate to
hydrological processeswere done recently by many researchersas re-
viewed by Mintz [1984] and Rowntree [1984]. This paper will also re-
view briefly the results of some of theseexperiments.But due to limi-
tation of space, only the essential points drawn from
three-dimensional GCM•s will be discussed.
Global Experiments with Prescribed
Soil Moisture
The simplest but still illustrative experiment is a global experiment
with prescribed soil moisture. Using the general circulation model of
NASA Goddard Space Flight Center Laboratory for Atmospheric
Sciences,Shukla and Mintz [1981] made experimentswith the global
soil kept either dry or saturated all the time. Fig.2 gives the distribu-
tion of pricipitation (mm/day) in wet-soil (top) and dry-soil case
(bottom) for July. The most .strikingdifferencebetweenthe two cases
is the sharp reduction of precipitation over the continents of the
Northern Hemisphere. Large parts of thesecontinents are almost void
of rainfall except Southeast Asia, tropical Africa and west coast of
north America. This lack of rainfall indicates that over the
extratropical continents in summer much of the moisture for precipi-
tation is derived from evapotranspiration and the rainfall is mainly of
the convective type which derives its moisture from planetary bounda-
ry layer. The intercomparison of the two casesalso shows that the
rainfall in the ITCZ is not much different, indicating that the moisture
converginginto the 1TCZ comesmainly from the oceans.
In the Southern Hemiphere July is, of course, a winter month. It is
interesting to note that little reduction of precipitation from the wet to
the dry case is found in the winter Southern Hemisphere. This sug-
geststhat in winter the continental precipitation mainly come• from
large-scale lifting condensation and that the source of moisture is
miainly from oceansby advection. Thus the evapotranspirationdoe•
not have too much influenceon the precipitation in winter.
Mintz [1984] further showed that for the extratropical continentz in
•ummer the moisture taken away by evapotranspiration is larger than
that added by precipitation. The deficit of soil moisture in summer
Geophysical Monograph Series Understanding Climate Change
102 SENSITIVITY OF CLIMATE MODEL TO HYDROLOGY
• PRECIPITATION
'•--•
• ALBEDO
DRIP
Jl
EVAPORATION
{IR
DOWN
FLOOD
FLOW
PERCOLATION
• TO
GROUNDWATER
Fig.l Schematicdiagramof the processes
that needto be considered
in a canopymodel of surfaceevapotranspirationand energybalance
[Dickinson, 1984].
must be replenishedby winter precipitation. Thus we can sccthat the
annual cycle of soil moisture content must be connected with the
oceanic processes.Since there is no evaporational cooling of the
ground and since more solar radiation (less precipitation, less
cloudness)is absorbedby the ground in the dry case,the ground tem-
perature will be higher than in the wet case. The big differencein the
ground temperature between the two casescan easily be seenin Fig. 3.
The high ground temperature will obviously lead to a turbulent sensi-
ble heat flux and a large loss of long-wave radiation from the ground
in the dry case.Thus the heat balanceand the moisturebalance(more
water vapor in the wet case)in the boundary layer will be much differ-
ent between the wet and dry case.
Shulka and Mintz•s experiment revealed the importance of the
hydrological processesin the climate. Similar experiments have been
done by other authors, for instance,Suarez and Arakawa [seeMintz•s
review, 1984]. The essentialresultswere similar to those of Shukla and
Mintz, but there were large differences between them in the details.
One reason for the difference can be attributed to the difference in
parameterization schemesin the two models [Mintzc, 1984]. The in-
flueneeof parameterizationschemeshas bccn discussedby many au-
thors [laval, et al., 1984; Dickinson,1984; Cunnington and Rowntree,
1986; Wardlow, 1985]. They all found that the parameterization is
important for the details of the simulations.
Experiments with Interactive Soil Moisture - with A Discussion
of the Dependenceof Hydrological Processes
on Climate Regime
Althogh the global experiments with prescribed soft moisture do
show the importance of hydrology in climate, the constraintsof these
Vol. 52
experimentsare too severe.One step to relax the constraintsis to al-
low the soil moiture to vary with time but on a global scale.Using a
M-21 version of the GFDL spectral model [Manbe, et al. 1979]
Kurbatkin, et al., [1979] make an experiment with variable soil mois-
ture in which the globalcontinentsweredry initially but changewith
time as predictedby the model.The distributionsof the rate of evapo-
ration and the rate of precipitationdifference(arid experiment-con-
trol experiment('naturalr)) are shownin Fig.4. It is seenfrom the fig-
ure that both evaporationand precipitationratesof the globalinitially
dry casewere lessalmost everywhereover the continents.Theseresults
were also obtainedin other GCM cxpciments[e.g. Yamazaki, 1986].
It is worth to note the two differencemapsover most of inner parts of
the summer continents(NH) are nearly equal. This indicatesthat over
the extratropical continentsin summerthe moisture for precipitation
is chiefly derived form evapotranspiration. But at midlatitude over the
cast coast of Asia there is an increaseof precipitation. It is also inter-
csting to note that decreasein precipitation is not limited to the con-
tinents; over the oceans,especiallyover the western tropical Pacific,
the decrease of precipitation is also very pronounced; up to - $
mm/day in large areas. Kurbatkin ctal. [1979] attributed this to
suppressof vertical motion over the oceans. The vertical motion over
the continentsin the initially dry caseis increasedalthoughthe precip-
itation rate is decreaseddue to lack of humidity. The increaseof the
continental vertical motion is probably associatedwith the decreasein
vertical motion over the ocean.
To further relax the constraints of the simulation we discard the
global condition and go t6 regional experiment with time-dependent
soil moisture. Many authors have done it, e.g., Rowntrcc and Bolton
[1983], Sud and Fenncssy[1982, 1984], Wardlow [1985], Yeh, ct al.,
[1984], and many others. In the global experimentsabove we have al-
ready sccn that the climatic responseto uniform global condition is
8CN
60N
40N
20N
0
20S
40S
60S
80N
60N
40N
20N
20S
40S
60S
Fig.2. Contoursof precipitationfrom Shuklaand Mintz [1982].For
the top case,they assumedsoil evaporatedas a saturatedsurface.For
thebottomcase,theyassumed
zeroevaporationfrom thesoil.Precipi-
tation greater than 2 mm / day is shaded.
Geophysical Monograph Series Understanding Climate Change
80N
60N
40N
20N
20S
zo
40S o IO
60S
80N
60N
40N
20N
20S
40S
60S
180 120W 60W
Fig.3. Ground-surface temperature(E) in wet- soil case(top) and dry-soil case(bottom), in experimentofShukla Mintz[1981].
not everywere the same. This means that the influence of hydrological
processesdependson the climate regimes.We shall further discussthis
point by usingYeh, et al./s [1984] results.They irrigated the continents
of 30-60N zone to its saturation (15cm) on July 1 and two other
zones, 0-30N and 15S-15N, to saturation on Jan. 1, and then inte-
grated each case for a period of five months. The influence of the
large-scale irrigation on climate was obtained by analyzing the differ-
ence between the irrigated and the standard run.
Fig.5a-c show the latitude-time distribution of the mean zonal dif-
ference in precipitation (cm/day) between the irrigated and control
experiments. One common feature in all the three figures is the in-
creaseof precipitation in the regions where the soil is instantaneously
saturated with water. However, it is clear that the period of enhanced
precipitation last longest in the 30-60N case (about 5 months) and
shortest in the 15S-15N case (about 3 months). These difference in
precipitation among the three casesare qualitatively similar to the dif-
ference in the persistenceof soil moisture anomaly or enhanced evap-
oration (not shown here). It can also be seen that in the 30-60N case
(Fig.5a) the increase is maximum near the center of the irrigated re-
gion and that it decreasesboth northward and southward. The corre-
sponding distribution for 15S-15N case (Fig.5c) is less symmetric. Its
maximum is displaced into the Southern Hemisphere in January and
shifts equatorward during February and March. For the 0- 30N case
(Fig.5b) there is a notable difference from the other two cases,i.e., the
Vol. 52
YE 103
IS
0 60E 120E 180
intensity of the increase is much smaller than in the other two cases.
Further, there is no well-defined maximum of precipitation increase
within the irrgatcd region; instead, a relatively samall increase takes
place throughout the entire zone. But the intensity of the evaporation
increase for 0-30N is of comparable value with that for 30-60N or
15S-15N cases. This indicates that the major part of water vapor
gained from evaporation in the zone 0-30N is transported out and on-
ly a small part is used to increase the precipitation. Ych, ct al. [1984]
invoked an increasein the descendingbranch of the Hadlcy circula-
tion to explain this phenomenon.
Figs.5a and 5b share another characteristic. Both show two belts of
maximum increase, one to the north and the other to the south of the
irrigated region. In the 30-60N case the northern belt of maximum
increasesis found at 60-65N, lasting about two months. The southern
bclt is in the tropics and persistsfor a much longer period. It is intcr-
csting to note that the northern belts of maximum increase in precipi-
tation in Figs.5a (summer) and 5b (winter) arc near the midlatitude
rainbelts in summer and winter, rcspcctivcly, and that the southern
bclts of maximum increase in precipitation in Figs. 5a (summer) and
5b (winter) arc situated over the ITCZ rainbelt which is north of the
equator and south of the equator, rcspcctivcly. This indicates the
largc-'scalc increase of soil moisture can induce certain physical pro-
cesscswhich tend to enhance the activity of the ITCZ and midlatitude
rainbelt cvcn outside the region of soil moisture increase. Ych, ct al.
Geophysical Monograph Series Understanding Climate Change Vol. 52

104 SENSITIVITY OF CLIMATE MODEL TO HYDROLOGY

90N
90N

60

30
ß
i.... '::!::.:.:.:.:..
•.:..-•:.:.:.:.:.:.:.: .... ........':::::'::
...::i:
...:.:.:.:.:.:. !..:.::i]'.-•.•i::'•/"
.'.'.¾'
'.:.'.'
....
.•....:.:.:.:.:.:.:.:.:.:.:.:
:.:...: 3O

-- EQ

30 6O

..::::iiiiiiiiiiiii!:::"
"'"'i:i:i:i:i:!:i:i:i:i:i:iiii::iiii:".•.•i!iiiii!!iiiiiiii!...
3O

90S 90S
0 90E 180 90W 0

90N 90N

6O 6O
... . ...._:.__
! ß....:.>:.:.:.:.>.'.. . ß======================
/•....:'......
ß :.
....
.• .......:.:.:.:.:.:.:.:
........•.._ ._,.•....•..;.

ß
•i.
:..'..
.'•::iii::::::iii!•i::•i"
ß ....... :"
3O 3O

EQ EQ

3O 6O

6O 3O

,....::::
•"' '""'•i•:.::•:.?•: ••?:::::::::'"'
90S ..:.:.:.:.;.:.1.:.:.;.:.:,:.:1:.:.>;.,.,.,.;i;,;( I i .....
:'.••
::•..
:'":
:'-
.•.
' :'.'
".
[-•----.•.•.-..,,..,..._.. ß.J:.:,:.:.:.:.:,I.;.;.;.;.:.:.:.1,:.:.:.;._,.:,_..1.:.:.:.:.:._..:.
90S
o 9o[ ] 8o 9ow 0

Fig.4. Rate of evaporation difference(a) and rate of precipitationdif- experimentsand are horizontally smoothedto emphasizethe large-
ference (b). Difference (Arid Exp.-Control Exp.) distributions are scalefeatures. Isolines are drawn for -5, -2, -1, 0, 1, 2, 5, mm/day
constructedfrom the July-August seasonalmeans of the two model in both distributions. [after Kurbatkin, ct al., 1979].

[1984] postulated thc following physical reasonsfor thc enhancements:
The strengthening of evaporation due to irrigation will produce a
sheetof moist and cool air near the groud. This sheetwill spread from
the region and supply moisture to the neighboring rainbelts and en-
hancesthe already existing convectiveactivity and precipitation there.
Ych, ct al. [1984] also pointed out that for all thc threccasesthcre is
a deercaseof precipitation just south of thc area of maximum increase
in the tropical rainbits. They speculatedthat this reductionmight be
due to the changeof upward motion in theseregions.Yeh, et al. [1984]
also calculated the total increase of precipation (AP) and the corre-

a b
90øN ; ', , . , 60ON 45øN

75ø 45ø 30ø

60ø 30ø 15ø

45ø 15ø' oo

30ø 0o 15ø

15ø 15ø 30ø

oß 30ø5 45øS
J A S O N J F M A M [ M A M

Fig.5. The latitude-time distributin of thc zonal mcan differcnce of precipitation rate (cm/ day) between thc
perturbedand the normal experiment.(a) the 30 ø N-60 ø N case;(b) the 0-30 ø N case;(c) the 15 ø S-15 ø N
½•e. [after Yeh, Wetheraid and Manabe, 1985].
Geophysical Monograph Series Understanding Climate Change Vol. 52

YE 105

TABLE 1. The integrated increase in the water balance components AE, AP, and AR due to an increase of soil
moisture AWs for three latitude zones averaged over the period where the values of AP were Positive (unit: cm)
and the ratios of water balance components to AWs. (After Yeh, Wetheraid and Manabe, 1984).

Latitute Zone AWs AE AP AR AE/AWs AP/AWs AR/AWs

30N-60N 11.2 15.3 9.6 3.1 1.37 0.86 0.28
0N-30N 13.9 15.5 3.7 1.2 1.11 0.27 0.09
15S-15N 13.0 11.8 6.6 6.7 0.91 0.51 0.52

sponding changes of evaporation (AE) and runoff (R) for the whole
irrigated zone during the period when the average regional change of
precipitation is positive. The results are shown in Table 1. The total
water (AWs) used to irrigate each zone to saturation is also given in
the table. The values of AWs of the three zonesare comparable as seen
in the table. The ratios AE / AWs, AP / AWs and AR / AWs in Table
1 can be used as a measure of efficiency of irrigation in inducing
changesin the hydrological processes.The larger the ratio AP/AWs,
the more benefit is gained from irrigation. When AE/AWs•
AP/AWs the benefit attains its optimum. In the region where
AP / AWs<<AE / AWs the efficiency of irrigation is least. From Table
I it is seen that the highest efficiency of irrigation is in the zone
30-60N and the least is in zone 0-30N. The physical reason for the ef-
ficiency being least in the 0-30N is that this zone is located under the
descendingbranch of the Hadley cell. Another interestingpoint seen
in the table is that in the 0-30N zone both AP / AWs and AR / AWs
are the lowest among the three zones.This indicatesthat the transport
of moisture obtained from the irrigation out off the irrigation area
should be highest in 0-30N zone. Table 1 also tells that
(AP / AWs)(30-60N) >(AP / AWs)(15S-15N). The reason for this is
that (AR/AWs)(30-60N)<(AR/AWs)(15S-15N). Thus the mois-
ture obtained from irrigation going to runoff is greater in the
15S-15N zone than in the 30-60N zone, making lessmoisture availa-
ble for precipitation in the 15S-15N zone. The above discussionclear-
ly showsthe high dependenceof the influenceof the hydrological pro-
cesseson the climate regimes. Many other authors have obtained simi-
lar resultsbut they cannot be disussedhere due to limited space.
Although the physics of the hydrological processesmainly lies at
the ground surface, the upper air will also be influenced through the
atmosphericboundary layer. To illustrate this in an example, we give
Fig.6 which shows the latitude-height distribution of zonal mean
temperature difference between the irrigated and the standard experi-
ment for the zone 30-60N in the secondmonth of irrigation (August)
of Yeh, et al.'s [1984] experiment. The picture shows negative differ-
ence in the lower layers of the atmosphere above the irrigted zone and
positive difference higher up. This indicates that in the lower atmos-
phere the increaseof latent heat flux by irrigation from the earth de-
creasesthe temperatures while at higher levels the increase of latent
heat release in precipitation increase the temperature. Even two
months after the large- scale irrigation, the maximum decrease of
temperatureis still nearly 4U and the maximum increasein tempera-
ture at high levels 1.8U The correspondingchang in zonal-mean
westly wind is given in Fig.7.
Sensitivity of Climate Models
to Surface Albedo
The presistentdrought in Sahel in recent years aroused great con-
cern among the people of the world including scientists and
politicians. Meteorologists studied the causeof the drought from va-
rious points of view and made simulations by GCM under different
land surfaceforcings.In discussing
the simulationof the Sahelian
drought we should keep in mind that this region is situated roughly in
the descendingbranch of Hadley cell and that its spatial scaleis on the
order of a few thousand kilometers.
Charney [1975] first proposed a model for maintaining the drought
in the Sahel. Charney•sdynamicsdepend upon a feedbackmechanism
involving radiation, subsidenceand albedo. A lack of vegetation gives
a higher surface albedo which leads to a net radiative heat loss at the
top of the atmosphere. This will induce a subsidenceof dry air aloft
with a subsequentreduction of precipitation, thus maintaining the de-
sert. This hypothesis was subsequently tested by a numerical
simulation with a general circulation model [Charney, et al. 1977].
Following Charhey et al. Chervin [1979], Sud and Fennessy [1982],
Planton [1986], Cunnington and Rowntree [1986], Laval [1986], and
Carson and Sangster[1981] using different GCM's obtained similar

95

350

680

F,30

•0

90o• ?5 ø 60 ø 45 ø 30 ø 15 ø 0ø 15ø 30 ø 45 oS

! I
Fig.6. The latitude-height distribution of the zonal mean differenceof air temperature(U) over land between
the perturedand normal experiment,August of the 30 ø N-60 ø N case[after Yeh, Wetheraid and Manabe,
19851.
Geophysical Monograph Series Understanding Climate Change
106 •EN•ITIV1TY OF CLIMATE MODEL TO HYDROLOGY
95
20.•
350
515 ,
680
830
940
9'}0
90•N 75 ø 60 ø
I
Fig.7.Thelatitude-height
distribution
of thedifference
of theAugust z•nal mcanwind(m/s) betweenthe perturbed
30 o N-60 ø N and the standardexperiment.[after Yeh, Wetheraid and Manabe, 1985].
resultsof decreasingprecipitation when increasingthe surfacealbedo.
The resultsof theseexperimentssupportedCharney'shypothesis.
All above experiments undoubtdly indicate that a high albedo is
important in maintaining deserts.However to prove this hypothesisas
the main cause of the Sahelian drought, we need observational evi-
dence of the actual albedo change. Charhey, et al. [1977] also found
that changesin the surfacehydrology can have comparable effectsas
a change in surfacealbedo from about 0.1 to 0.3, which is the value
usedby Charney for explainingthe Saheliandrought. Rowntree and
Sangster[1985] recentlyalso simulatedthe decreaseof sahelianprecip-
itation by only decreasingthe soil mosturethere.
In 1986 Cunnington and Rowntree [1986] went further to test the
influenceof the initial atmosphericmoisture state on the precipitation
and hydrologicalfactors of the Sahara.
To close this section we should say that although the soil moisture
content and the albedo are indeed important factors determining the
Saharan climate we sould be very cautions to use the above experi-
mental resultsto state that we actually explained the recent drought in
the Sahel. There have been many long-term natural variations of the
averagedprecipitationoverAfrica [Nicholson,1985].We mustbe able
to detect the albedo- or soil moisture content- induced effects against
this large background noise.
Sensitivity of Climate M odeIsto Vegetation
As pointed out in the Introduction, vegetation is another important
factor determining the surface moisture availability which controls the
partition between the surface fluxes of latent heat and sensiblehent. In
all the foregoing described experiments evaporation and transpiration
were combined as cvapotranspiration. Dickinson [1985] strongly em-
phasized the need for separation of the roles of vegetation and soils in
the GCM's. A comparatively detailed description of the land cover
and soil has been introduced into NCAR Community Climate Model,
and several sensitivity studies have been completed with major em-
phasis given to the effect of the Amazon deforestation. The prelimina-
ry conclu- sions drawn from comparisons of the control and the
deforested runs for the Amazons region were given by Dickinson
[1985]. They arc: •the control and deforestedruns both agree fairly
Vol. 52
•5" 30 ø 15" 0ø
well with the observed seasonalcycle of rainfall. Changes in rainfall
with deforestation are not obvious because of high natural variability
of model convective rainfall. The deforested case has more runoff, es-
pecially during the dry seasonand lessinterception of rainfall. The pe-
riod during the season of driest soil is widened from one month to
several months in southern Amazonia. The soil and air become warm-
er by several degrees, especiallyduring the dry season;sensibleheat
fluxes are increased and evapotranspiration is reduced by
deforestation.'
With emphasis on tropical deforestation, Henderson-Sellers et al.
[1985] studied the possible climate impacts of land cover transforma-
tions. Hansen et al. [1983] have also, but in a much simpler way, in-
corporated the parameters associatedwith different vegetation types
into their climate model, model II. The vegetation distribution over
the earth were compiled by Mathews [1983]. Using this model Rind
[1982] investigatedthe influenceof vegetation on the hydrological cy-
cle. In this model the function of vegetation is much simplified and the
soil is consideredas composedby two layers, an upper layer of 10 cm
depth and a lower layer down to 4 m depth.
Using this model Rind [1984] studied further the albedo, water
holding capacity and diffusion for the vegetation types in the climate.
The albedo effect is most effective in areas with plentiful moisture as-
sociated with evaporation from the ocean surface and during the sea-
son when large-scale dynamics or convection control the precipita-
tions. The water holding capacity influence dominates where local
evaporation is essential for precipitation. The diffusion experiment al-
tered the rate at which water is made available from deeper levels. Its
main effect was a stretching out of time of the soil moisture loss, with
affected the seasonalphase of the precipitation.
The Dependence on The Spatial
Scale of The Anomaly
So far we have been concerned with very large- scale anomalies, ei-
ther on the global scale or on the continental scale. Certainly the im-
portance of the surface hydrology on climate will depend on the spa-
tial scale of the surface anomaly. It may be estimated relative to the
role of advection. Rowntree [1984] estimated the importance of
Geophysical Monograph Series Understanding Climate Change
advected moisture on continental rainfall under the assumptions of
nondivergent flow of air acrossa land mass with a humidity acquired
while crossingthe land mass, an evaporation rate of 3 mm/day (ap-
propriam for moist tropics), and a mean flow of 5 m / s. Under these
assumptions he showed that for an island of 100km horizontal scale
CLo), the prescribedabvectedmoisture flux convergenceplays a dom-
inant role, and local evaporation a very minor role. However for Lo•
1000 km (e.g., Penisular India), local evaporation is a little more im-
portant while for Lo > 3000 km (e.g., Africa), the contribution from
the horizontal moisture flux convergenceis relatively small. Although
Rowntree'scalculationwas basedon very idealizedassumptions,they
give a qualitative earlmate of the scaleeffect in the influenceof evapo-
ration anomalies.
Concluding Remarks
The above discussionsshow that the hydrological porcessesare very
important in climate. The inclusion of hydrology in GCM's is not on-
ly necessaryfor climate studiesor predictions,it can also improve me-
dium-range weather forecast [Blondin, 1985], a few days forecast
[Tada, 1985] and even very short-range weather forecast [Gadd and
Keers, 1970].
Another potentially possible practical application of the surface
hydrological processeswould be the modification of local weather or
climate. Anthes [1984] made a study showing the possibility of
enhancement of convective precipitation in semi-arid regions by
planting alternating bands of dense vegetation with width of
25-50km. Segal et al., [1983] suggestedto modify the local climate by
filling in the QattaraDeppression
(about 18,000km2 in size)with
Mediterranean water which is about 90km from this deppression.
There are, of course, many other possibleapplications.
Since the two crucial factors in hydrology are the soil moisture con-
tent and the vegetation, we strongly recommend that soil moisture ob-
servations be made by satellite and that the astellite data be validated
by ground-truth observationsand global mapping of vegetation.
References
Anthes, R.A. (1984), Enhancement of convective precipitation by
mesoscale variations in vegetative covering in semiarid regions, J.
Climate & Appl. Met., 23, 541-554.
Blondin, C.A. (1985),Treatement of land-surface properties in
ECMWE model, Proc. ISLSCP Con.•grence,Rome, Italy, 2-6 Dec.
1985.
Carson, D./I. and A.B. Sangster (1981), The influence of land-surface
albedo and soil moisture on general circulation model simulations,
GAPP / WCRP: Research Activities in Atmospherical and Oceanic
Modelling, (Ed. by J.D. Rutherford) Numerical Experomentation
Programme. Rport No.2, PP5.14-5.21.
Charney,J.G., (1975),Dynamics of deserts drought in the Sahel,
Quart. J.R. Met. $oc., 101,193- 202.
Charney, J.G.,W.J. Quirk, S.H. Chow and J. Kornfield (1977), A
comparative study of the effects of albedo change on drought in
semi- arid region, J. Atemos. Sci., 34, 1366-85.
Chervin, R.M. (1979),Respones of the NCAR general circulation
model to changed land surface albedo, Report o.fthe JSC study Con-
.•grenceon Climate Models: Per.•rmance, Intercomparison and Sen-
sitivity Studies. Washington, DC, 3-7, April, 1978, GAPP Publ. Se-
ries, No. 22, Vol. 1 563-81.
Cunnington, W.M. and P.R. Rowntree (1986), Simulations of
Shaharan atmosphere-dependence on moisture and albedo, Quart.
J.R. Met. $oc., 112, 971-999.
Dickinson, R.E. (1984),Modelling evapotranspiration for
Vol. 52
YE 107
three-dimensional global climate models. 58-72, In Climate Pro-
cessesand Climate Sensitivity, Geophys. Mon., 29,/I.E. Hansen, and
T. Takanhashi editors.
Dickinson, R.E. (1985),GCM sensitivity studies- implications for
parameterizations of land surface processes, Proc. ISLSCP. Con-
.•rence, Rome, Italy, 2-6, Dec. 1985, EAS. SP-248, 127- 129.
Eagleson, P.S. (1982), Dynamical hydro-thermal balance at
macroscale, In land Sur•ce Processesin Atmospheric General Circu-
lation, Ed. by P.S. Eagleason, Cambridge University Press.
Gadd, A./I and/I.F. Keers (1970), Surface exchangesof sensitiveand
lantent heat in a 10-level model atmosphere, Quart. J.R. Met. Soc.,
96, 297-308.
Hansen,/I.E., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R.
Ruedy and L. Travis: (1983), Efficient three dimensional global
models for climate studies: Models I and II. Mon. Wea. Rev., 111,
609-662.
Henderson-Sellers, A. and H. Gornitz (1985), Possible climatic im-
pacts of land-cover transformations with particular emphasis on
tropical deforestation, Climate Change, 6, 231- 257.
Kurbatkin, G.P., S. Manabe End D.G. Hahn (1979), The moisture
content of the continents and the intensity of summer monsoon cir-
culation, Meteorlogiya i Gidrologya, 11, 5-11.
Laval, K. (1984), Modeling the impact of soil properties on European
climate. In Current Issues in Climate Research, A. Ghazi and R.
Fantechi, editors. D. Reidel Publishing Company.
Laval. K. (1986), General circulation model experiment with suface
albedo changes, Climate Change, 9, Special issue: Climate and
Desertification, 91-102.
Laval, K., A. Perfie and Y. Serratini (1984), Effect of
parameterization of evapotranspira- tion on climate simulated by a
GCM, In New Perspectivesin Climate Modeling, Ed by A.C. Berger
and C. Nicolis, 223-247.
Mathews, E. (1983), Global vegetation and land use: New high resolu-
tion data basis for climate studies, J. Clim. and A ppl Meteor., 22,
474- 487.
Mintz, Y. (1984), The sensitivity of numerically simulated climates to
land-surface boundary conditions, 79-105, In The Global Climate,
Ed. by/i. T. Houghton.
Namias, /i. (1962), Influences of abnormal surface heat sourcesand
sinks on atmospheric behavior, Proc. Internat. Symp. Num. Wea
Predict. Tokyo, Nov. 7-13. 1960, Met. Soc./iapan, 615-627.
Namias, /i. (1963), Surface-atmosphere interactions as fundamental
causes of drought and other climatic fluctuations, Arid Zone Re-
search Changes of Climate Proc, Rome Symp. UNESCO and
WMO, 345-359.
Nicholson, S.E. (1985), African drought: an example of the influence
of land-surface processeson climate. Proc. ISLSCP Con•rence,
Rome, Italy, Dec. 1985, ESA SP-248,405-410.
Planton, S. (1986): Sensitivity of the annual cycle simulated by a
GCM to a change in land surface albedo, Proc. ISLSCP
Con•rence, Rome, December, 1985, ESA SP-248, 135-142.
Rind, D. (1984), The influenceof vegetationon the hydrologicalcycle
in a global climate model, 73-91, In Climate Processesand Climate
Sensitivity, Geophys. Mon., 29, Ed. by /I.E. Hansen and T.
Takahashi.
Rowntree, P.R. and/I.A. Bolton (1983), Simulation of the atmospher-
ic response to soil moisture anomalies over Europe, Quart. J. R.
Met. $oc., 109, 501-526.
Rowntree, P.R. (1984), Review of general circulation models as a ba-
sisfor predicting the effect of vegegationchange on climate, Met. O,
20, Technical Note. II ? 225.
Rowntree, P.R., W.F. Wilson and A.B. Sangster (1985), Impact of
land surface variation on African rainfall in general circulation,
DCTN 30 Met. O.
Geophysical Monograph Series Understanding Climate Change
108 SENSITIVITY OF CLIMATE MODEL TO HYDROLOGY
Segal, M., R.A. Pielke, and Y. Mahrer (1983), On climatic change due
to deliberate flooding Qatara Deppression (Egypt), Climatic
Change, 5, 73-83.
Shukla, J. and Y. Mintz (1982),Influence of land- surface
evapotranspiration on the earth's climate, Sciece,21:$, 1498-1501.
Sud, Y.C. and M. Fennessy (1982), A study of the influence of surface
albedo on July circulation in semi-arid regions using GLAS GCM,
J. Climatology, 2, 105-125.
Sud, Y.C. and M. Fennessy (1984), Influence of evaporation in
semi-arid regions on the July circulation - a numerical study, J.
Climate, 4, 383-398.
Tada, K. (1985), Land-suface parameterization in JMA operational
Vol. 52
spectral model, Proc. ISLSCP Con.•rence, Rome, Italy, 2-6, Dec.
1985, ESA. SP-248.
Warrilow, D.A (1985), The sensitivityof the UK Meteorological Of-
fice atmospheric general circulation model to recent changes in the
parameterization of hydrology, Proc. ISLSCP Con.Igrence,Rome,
Italy, Dec. 1985, ESA. SP- 248, 143-149.
Yamazaki, K. (1986), The sensitivity experiment to land-suface
boundary conditions with the M.R.I.G.C.M. Proc. ISLSCP Con-
.Igrence,Rome, Italy December,1985, ESA. SP-248, 151-157.
Yeh,T.-C., R.T. WetheraldandS. Manabe(1984),The effectof soil
moisture on the short-term climate and hydrology change-a nu-
merical experiment, Mon. Wea. Rev., 112, 474-490.
Geophysical Monograph Series Understanding Climate Change Vol. 52

STABILITY OF TREE/GRASS VEGETATION SYSTEMS

Peter S. Eagleson

Departmentof Civil Engineering

MassachusettsInstitute of Technology,Cambridge,Massachusetts

Abstract. The average annual water balance of landsurface area in Figure 1. The area of bare
tree/grass vegetation systems is modeled as an soil is indicated by Ms . The area shaded by
interactive competition for water and energy. grass (with the sun directly overhead) is Mg
Ecological optimality hypotheses are introduced and that by the trees is Mw. This is often
which allow specification of the fractional area called the "projectlye foliage cover" or the
covered by woodland canopy and by grass canopy "canopy density". Following the observations of
under conditions of natural equilibrium. Three Sarmiento [1984] we assume no runoff.
equilibrium states are found. Two are monocul- We further assume that the taller canopy
tures, i.e., grassland and closed forest which suppresses the atmospheric water vapor transport
are shown to be unstable with respect to per- capacity ep that is effective for the shorter
turbations of canopy cover. The third, a canopy. Accordingly and using a simple linear
tree/grass mixture, is shown to be stable with relation, we have for the grass
respect to perturbations of canopy cover but
metastable with respect to climate change.
epg= (1 - Mw) epw (1)

Introduction
and for the bare soil

We consider a mixed formation of grasses and

woody plants which is often referred to as eps= (1 -Mg) epg (2)
"savanna" [e.g., Dansereau, 1957]. In the ab-
sence of photographs, an accurate visual image is in which the subscripts w, g and s refer to
evoked by Walter's [1973] description" ... homo- trees, grass and bare soil, respectively, and the
geneous grasslands upon which woody plants are ep'Sare average
annualrates(cmsec-1).
more or less evenly distributed". This "even Actual average evapotranspiration rates eT
distribution" of the woody plants is indicative (cm sec-1) are related to the potential rates
of their control by soil moisture availability ep. For the vegetation components,
and is a key to the conceptualization used here.
The climate of savanna vegetation systems is eT = kvep (3)
characterized by marked seasonality in moisture
availability [Monasterio and Sarmiento, 1975]. where kv (dimensionless) is called a "plant
In a survey of the literature describing savannas coefficient", and for the bare soil [see
in South Africa, West Africa, Sudan and South Eagleson, 1978a]
America, Segarra [1983] found the mean rainy
season length (i.e., fraction of year) ml to 5/2
vary from 0.46 to 0.67. eTs = J eps = JoSo (4)
Model of Tree/Grass Equilibrium where so = time and space average soil moisture
concentration in the upper layer (the hydroclim-
We first seek the roots of the time-averaged atologic state variable), where J (dimensionless)
conservation equation for soil water as described is the evaporation efficiency, and where Jo
in detail by Eagleson and Segarra [1985]. The (cm sec-1)is defined in terms of the properties
average annual water balance is modeled one of the soil, water and climate.
dimensionally as shown schematically for a unit We assume the climate to be seasonal and the
soil to be divided into two layers. The top
Copyright 1989 by layer contains soil moisture only during the
International Union of Geodesy and Geophysics rainy fraction m• of the year and is thus
and American Geophysical Union. assumed to be the sole source of moisture for the

109
Geophysical Monograph Series Understanding Climate Change Vol. 52

110 SAVANNA STABILITY

PA

Mg

•M w • ß Ms ,

epw'kvw

TREES
t
epg-
kvg
j . eps
Rs=O

NO
RUNOFF

_-

Rg=O
NO DEEPPERCOLATION
I

Fig. 1. Annual water balance of tree/grass savanna.

grass. Soil moisture unused by the grass where

percolates from the upper to the lower soil layer

during the rainy season where it provides
fraction • of the moisture transpired annually by
K -- ml kvg
/(mskvw) (7)
the trees. The remaining fraction 1 - • of the
and
trees' annual moisture use comes from the upper
soil layer via shallow roots. We assume the
trees are spaced so that they use all the
percolated water. On an annual basis we express
S _=K(1)/(epwkvg) (8)
this as
With reference to Figure 1 the average annual
water balance can now be written in the
talK(
1)s2 : •Mwm6epwkvw (5)
dimensionless form

in which
G: Mw
+ Mg(1
K
-Mw)+ (1 - Mg)R
So5/2(9)
m6 = fraction of year during which soil
moisture in lower layer can support
where
transpiration by trees-

K(1) = saturated hydraulic conductivity of G -- PA/(m•epwkvw) (10)

upper layer soil, cm sec-1.
and
Dimensionlessly, (5) becomes

Mw : (KS/s) so
5 (6) R = Jo/(epwkvg
) (11)
Geophysical Monograph Series Understanding Climate Change Vol. 52

EAGLESON

x =CONSTANT
G--K=1

/ \
/ \
/ \
/ \
/ \
/
/
dMw /
dt EQUILIBRIA / EQUILIBRIA

o I Mw
/
/
/
/
/
/
/
/
/
/

Fig. 2. Stability of savanna woodland equilibria.

in which PA(Cm) is the average annual precip- a closed forest (Mw = 1, Mg = 0) will use the
itation rate. the same amount of water annually as will grass-
land (Mw= 0, Mr = 1), and G = 1 specifies that PA
Considering
kvgandtheproduct
m6kvw
to be is exactly sufficient for both of these extreme
known parameters of the system there are three
states. An intermediate tree/grass equilibrium
unknowns,So, Mw and Mg but only two equations, is given by (14).
(6) and (9). Invoking the hypothesis [Eagleson,
1978b, 1982] that natural vegetation systems will
attempt to minimize water demand stress, we add Importance of Seasonality
the third equation
Note that as seasonality vanishes (i.e., ml
bSo/•Mg= 0. ( 12) and hence m• approach unity) equations (7), (10)
and (13) give

Eqs.
feasible
(6),
solution
(9) and
only
(12)
for
have a physically
kvg/kvw
+1
and

G=K= 1 (13)
PA/(epwkvg)
+1
for which
which can be satisfied either for Mw = 0 or
for
kvg= kvw
. Since under water-limited
Mw = [-X + (X2 + 1)1/2] (14)
conditions
kvg>kvw[Eagleson
andSegarra,1985],
where the latter condition will occur only with
unlimited moisture when = 1. Under
kvg= kvw
X = R/[2(S/•)]1/2 (15) such conditions the vegetation will probably seek
to optimize biomass rather than soil moisture
[Eagleson, 1982] and trees will replace the
with Mg being arbitrary.
There are thus three equilibrium states of the grass. This situation is outside the range of
savanna system. The condition K = 1 implies that the current analysis.
Geophysical Monograph Series Understanding Climate Change Vol. 52

1 12 SAV•NA S•A•œ•ILI f¾

0.5

K=I

0.4

Mw
0.3

so
G=I (

0.2 /
/ 0.8,.
/
/
/
0.1

I I I I I I I
00 0.2 0.4 0.6 1.0
MgandMw
Fig. 3. Sub-optimal equilibria; R -- 24, S/• = 250, K = 1.

We thus see that seasonality of the moisture order-of-magnitude analysis we assume the
supply is a necessary condition to have an absolute value of the first term on the right
equilibrium tree/grass system. hand side of (16) exceeds that of the second.
The sign of dMw/dt will then be that of dSo/dt.
Stability to Vegetation Perturbations
Shouldwe be at the Mw = 0, Mg= I equilibrium
and somehow
decreaseMgwe replace grass,
We begin by considering perturbations to the transpiring at the potential rate, by bare soil
equilibrium vegetation under the assumption that of lower evaporation efficiency. The soil
these cause no change in the parameters of the moisture so will thus rise making dSo/dt , and
climate and/or soil.
hence dMw/dt positive. Should we be at the
In the phase plane diagram of Figure 2 the Mw = 1, Mg = 0 equilibrium and somehowdecrease
three equilibria in M are located along the Mw we will have moisture loss by bare soil evap-
dMw/dt - 0 axis. The solid and dashed lines oration and from growth of grass thus decreasing
are qualitative representations of the only two t o and making dMw/dt negative. This reasoning
possible forms of dMw/dt - f(Mw). The slope establishes f(Mw) as the solid curve (a) in
of f(M w) at dMw/dt - 0 determines the stability Fig. 2 and the intermediate tree/grass equilib-
of the associated equilibria. Writing the rium is seen to be unconditionally stable to per-
conservation equation for water in the lower soil
turbations in Mw. If fire should decrease Mw
layer where Zw is the water table elevation and from this equilibrium value, Mw would be in a
differentiating with respect to time gives region of positive dMw/dt and the system would
return to equililbrlum. Such restoration would
not follow perturbations of the Mw = I and Mw = 0
=m6epwkvw
dMw
dt
= 5m•
K(1)So4
dsø
dt
- n d2Zw
dt 2
(16) equilibria
equilibria
and they
under this analysis.
are therefore
Slash-and-burn
unstable

agriculture would thus be expected to create

savanna from forest and grassland under the
in which n is the soil porosity. From an appropriate (G = K = l) climatic conditions.
Geophysical Monograph Series Understanding Climate Change Vol. 52

EAGLESO• 1 1•

Stability to Climate Change solar energy there are three equilibrium states:
closed forest, grassland, and a tree-grass
Consider Fig. 3 in which for K = 1 and for mixture. Only the last of these appears stable
representative constant values of R and S/• with respect to perturbation in the vegetation
various solutions of the equilibrium equations components but it is roetastable with respect to
(6) and (9) are presented. The dashed curve climate change.
represents Mw(so) as given by (6) and is
independent of the climate parameter G. The Acknowledgments. This work was supported by
solid curves represent Mg(So;G)as given by (6) the National Science Foundation under Grant No.
and (9). The horizontal line G = 1 is the only ATM-8114723. Prior publication of these results
solid line that also satisfies (12) and thus its has been made first, in a more complete form by
intersection with the dashed line represents the the American Geophysical Union [Eagleson and
intermediate tree/grass equilibrium discussed Segarra, 1985], and second, essentially as
above. presented here by NASA [Eagleson, 1986] .
For G > 1 we see that Mg increases with so
until s o reaches the value (s o = 0.33 for
G = 1.2) at which Mw becomes unity. At this References

point Mg is discontinuous since there, due to

(1), Mg must drop to zero. Wenote therefore Dansereau, P., Biogeograph¾, An Ecological
that as so increases under nonoptimal G > 1, Perspective, Ronald, New York, 1957.
the equilibrium tree density is driven to unity Eagleson, P. S., Climate, soil, and
and the equilibrium grass density to zero. The vegetation; 4. The expected value of annual
observation of savanna for G > 1, K = 1 must evapotranspiration, Water Resources Research,
therefore be due to interference by another agent 14(5), 731-739, 1978a.
such as fire. Eagleson, P.S., Climate, soil, and vegetation;
For G < 1 we see that Mgdecreaseswith 6. Dynamics of the annual water balance, Water
increasing so until the grass disappears. At Resources Research, 14(5), 749-764, 1978b.
Mg = 0, so reaches its maximum
value which is Eagleson, P.S., Ecological optimality in water-
less, as is the associated Mw, than that for limited natural soil-vegetation systems; 1.
optimum G = 1. We should therefore not expect Theory and hypothesis, Water Resources
to find closed forest (Mw = 1) where G < 1 and Research, 18(2), 325-340, 1982.
K • 1. Eagleson, P.S., Stability of tree/grass
We thus conclude that if we accept the vegetation systems, in Climate-Vegetation
ecological optimality criterion of (11) as the Interactions, NASA Conference Publication No.
determinant of tree/grass equilibrium, climates 2440, [C. Rosenzweig and R. Dickinson, Eds.],
with G ) or < 1 cannot have stable tree-grass Greenbelt, Maryland, pp. 149-155, 1986.
savannas. For G > 1 pressure to develop toward Eagleson, P. S., and R. I. Segarra,
increasing so will lead eventually to closed Water-limited equilibrium of savanna vegetation
forest while for G < 1 this pressure will lead to systems, Water Resources Research, 21(10),
low density trees and an absence of grass. The 1483-1493, 1985.
flatness of the driving gradient 58o/• in the Monasterio, M. and G. Sarmiento, A critical
small Mw range of most savannas indicates, consideration of the environmental conditions
however, that there is ample opportunity for associated with the occurrence of savanna
other stress-producing factors (such as ecosystems, in Tro•)ical America, edited by
nutrition, pests, fire, etc. that have been F. B. Golley and E. Medina, Springer-Verlag,
neglected in this water-based analysis) to play a New York, 1975.
stabilizing role. Segarra, R. I., Stability of natural savanna
This suggests the monitoring of savanna tree ecosystems, Civil Engineer thesis, Dept. of
density changes as a visible indicator of Civil Eng., Massachusetts Institute of
climatic change. Technology, Cambridge, MA, 1983.
Sarmiento, G., The Ecology of Neotropical
Conclusions Savannas, translated by O. Solbrig, Harvard
Univ. Press, Cambridge, MA, 1984.
When the tree/grass (i.e., savanna) vegetation Walter, H., Vegetation of the Earth, vol. 15,
system is modeled as a competition for water and Springer-Verlag, New York, 1973.
Section V

TROPICAL OCEAN AND

GLOBAL
Geophysical Monograph Series Understanding Climate Change Vol. 52

TOGA AND ATMOSPHERIC PROCESSES

Kevin Trenberth

National Center for AtmosphericResearch,Boulder,Colorado

Abstract. A brief outlineis givenof the TOGA (TropicalOceans 3} Modeling:Atmospheric, with specifiedseasurfacetemperatures
andGlobalAtmosphere} Programalongwith a moredetaileddiscus- (SSTs);oceanographic, with a specifiedsurfaceatmosphere(or
sion of the relationship of the atmosphericcirculation to sea surface fluxes);andcoupledocean-atmosphere models.
temperatures {SSTs}in the tropics. The best knownphenomenon 4) Prediction: Statistical-dynamical,
andusingmodels.
that is part of TOGA is E1 Nifio-Southern Oscillation(ENSO} and 5) Data Management
this paper focuseson the atmosphericcomponentof ENSO. The sim- Theseaspects
are dealtwith in-depthby WCRP (1985a,b) and
ilaritiesand differencesamongdifferentENSO eventsare reexamined NationalAcademyof Sciences
(1986).
as seen through a Southern Oscillation index and indices of SST in The best known phenomenonrelevant to TOGA is El Niio-South-
the tropical Pacific. Variations from event to event are marked and ernOscillation(ENSO). In this paperthe focuswill be on the atmo-
phaselocking of ENSO eventsto the annual cycle is only weak. The sphericcomponentof ENSO and its link to the tropical SSTs and,
1986-87 ENSO has been especiallyanomalouswith regard to timing. in particular, why convectionoccurswhere it does. Many recent
However, it provides an excellent illustration of the link between at- studieshave usedthe tool of compositingto bring out the common
mosphericconvectionand SSTs. The reasonswhy convectionoccurs features
of pastENSOevents[e.g.Rasmusson
andCarpenter,1982];
where it does and the importance of warm water greater than 28øC [vanLoon,1984];[vanLoonand Shes,1985,1987]. Herewe will
are discussed. It is shown that there is a need to better understand
emphasizethe differencesamongthe eventsand focusespeciallyon
the atmosphere-oceanlinks in the tropics and a need to measuremore recent developments,the 1986-87 ENSO event, which turns out to
accurately and understand changesin SSTs. havesomeuniquefeatures,but which nicely illustratesthe relation-
shipsbetweenSSTs and atmosphericconvection.

Introduction
El Nifio-Southern Oscillation Events

The TOGA Program is an international program under the World

ClimateResearch Program{WCRP) focused on the interannual
vari- ENSO events are made up of componentsfrom both the atmo-
abilityof the TropicalOceansand GlobalAtmosphere {TOGA} as a sphereand the ocean. The SO is the atmosphericcomponentof
coupledclimate system. The goalsof TOGA are: ENSO. It has a time scaleof 2-7 years[Trenberth,1976]and con-
ß To gain a description of the tropical oceansand the global atmo- sistsof a global-scale,predominantlystandingwave with centersof
sphereas a time dependentsystem,in order to determine the ex- action in surfacepressureover Indonesiaand the tropical South P•-
tent to which this systemis predictableon time scalesof monthsto cific,seeFig. 1 [Trenberth
andShes,1987].Withinthislargepattern,
years, and to understandthe mechanismsand processesunderlying there are seasonalteleconnectionsto higher latitudesthat tend to be
its predictability. strongest in winter. These teleconnectionsare thought to arise as
ß To study the feasibility of modeling the coupledocean-atmosphere Rossbywavespropagating out of the tropics combined,especiallyin
system for the purpose of predicting its variations on time scales the Northern Hemisphere,with interactionswith the stationaryplan-
of months to years. etarywavesand 'basicstate' flow[e.g.Branstator,1985].However,
thereis evidenceindicatingstatisticallysignificantalthoughweakre-
ß To provide the scientificbackgroundfor designingan observingand
lations in the summer seasontoo, as shownfor instanceby the cor-
data transmissionsystemfor operational prediction if this capabil-
ity is demonstratedby coupledatmosphere-oceanmodels.
relationmapsof [TrenberthandPaolino,1981].This is importantin
the current context becauseof the very strong anomaliesduring the
Formally, TOGA began in 1985 and consistsof a decade-long pro- northern summer of 1987 associated with the current ENSO.
gram with severalcomponents.
The E1 N/rio is thought of as the oceaniccomponentof ENSO. It
1} Long-termobservations:
To obtainbetterglobalatmospheric
data is manifest as major increasesin SSTs in the tropical Pacific. The
sets and tropical ocean data sets that document the phenomenon
of interest and can be used to force models and to validate models.
terminology is confusing,since E1 Nifio was originally a warm cur-
rent off the coast of Peru which penetrated south about Christmas-
2) Empiricalstudies time (henceNiio, the Christ child). The term has sincebecome
associatedwith the occasionalunusually large warmings 'El Nifio
events', which, in turn, are associatedwith basin scaleanomalously
Copyright 1989 by high SSTs. Strictly speaking,the term is used only to refer to the
International Union of Geodesy and Geophysics original South American coastal phenomenon,but in the ENSO con-
and American Geophysical Union. text it hascometo mean a basin-scalewarmingof the tropical Pacific.

117
Geophysical Monograph Series Understanding Climate Change Vol. 52

TOGA AND ATMOSPHERIC PROCESSES

•o • C3
o • o o o

0 0 0 0 0 0 0 0
C) C) C) C) C) C) C) C)
Geophysical Monograph Series Understanding Climate Change Vol. 52

TRENBERTH 1 19

DARWIN SIP 2 UoN'rH nUNN,NGUSxN tent to which the phenomenonis phaselockedto the annual cycle, as
impliedin manystudiesfollowingthe landmarkpaperby [Rasmusson
,Me .,-'% andCarpenter,1982];and 2} the extentto whichthe eventsaresim-
2.5 toe0 , ilar or differ from one another through their life cycle. Subsequently,
------ Ig'2•
....... 198• s thesepoints are reexamined using SST data.
..,_'
1.5 Figure 2 showsthe three year sequencesof anomaliesof Darwin sea
level pressure,as 3-month running means, during all ENSO events
\ ,o since1939. Each eventis labelledby the centralyear [year 0) and
0.$ ',
extendsfromthe yearbefore(year-1)to the yearfollowing(year+1),
usingthe terminolog7 in Rasmussonand Carpenter. The large events
are shown in the top panel, the smaller events in the central panel,
and the compositeplus and minusone standarddeviation(•r} (ex-
-I.5
ß
cluding1986}is givenbelow. Of note is the lack of significance
of
the compositewhich is nowhere more than 2a from the mean. This
-2.5 I ........... i ........... is due mostly to the inclusion of the smaller events in the central
YEAR -1 YEAR 0 YEAR +1 panel, which are of the order of one •r events, both in the SO and
SST fields. However, another factor is that the phase locking with

3.5
;...........
;...........
the annual cycle is not very strong. Phase locking seemsreasonable
1863
in Fig. 2a, but the weaker events do not follow that pattern and the
2.5 ....... 1063
1986-87 sequenceis the most unusual on record. It did not really

---,0so ,/,,,
•,••/....
.......
begin until late in 1986, six months later than usual, and is the most
• 1.5 extreme case on record in May-July of year +1. It has been associ-
ated with a very poor summer monsoonin India, one of the worst in
several decades.
• 0.5 • /

,,.... Relationships with SSTs

[-o.,
'_
&.
-1.5

-2.5 • ...........
YEAR -]
• ...........
YEAR 0 YEAR +
.%_/
Although many studieshave used atmosphericand oceanicparam-
eter anomaliesin their statistical analyses,the link between the two
is complex. Instead, the simplest link appears to be between the
atmosphericanomaliesand the total SST field, especiallyin models
[Shuklaand Wallace,1983]. This is illustratedin a generalway in
•.5 ; ........... ; .........
Fig. 3 which showsthe annual mean SST and outgoing longwavera-
diation{OLR) fieldsfrom [Shea,1086]and [Janowiak
et al., 1085].
2.5 Low OLR in the tropicsis associatedwith high top clouds,and OLR
is therefore an index of convectionin the tropics. The stippled re-
gionscorrespond
to the highSSTs(waterwarmerthan 28ø0) and
lowOLR (the convergencesones).Thereis a closerelationship
over
the oceanswhich is alsopresent,althoughwith migration north and
,,<%,.,•c,.:.. ..;.;.>:.....>;.....;........:.......
south, in the individual seasons.
• .:•:•:•:•:•:•:•:•.•.•:•:•:.:•:•:•:•:•:•:•'
..:•.• '-,,•?•:•:•:•:•:•:•:•::;•.••: Low OLR valuesalso occur over the warm tropical continents. In
• •.s fact, the absenceof a tropical continent in the Indonesianregion is
.,<:•..{•:•:•:•:•:•:•:•:•:•:•:•:.
one thing that makesthe Pacific unique. The warmestwater on the
globe,averagingover 29øC, occursin the tropicalwesternPacificand
is actuallylocatedslightlysouthof the equator,corresponding to the
northernmost
part of the SouthPacificConvergence
Zone(SPCZ}.
YEAR -1 YEAR 0 YEAR + The 1986-87ENSO has beenrelatively modestby somestandards.
Fi•. 2. Sequencesof anomalies in sea level pressure aZ D•in The SST anomalieshave not been huge, or even closeto the magni-
over aH El N•o events s•ce 1939. Each tudesexperiencedduring 1982-83. However,they havebeen sufficient
central%e• (%e• 0). The l•er events•e to significantlychangethe location of the warmest water.
5, •d •he compositemean (excludh• 198•) plusand m•us one This is illustratedin Fig. 4 whichshowsthe total SST field and the
s•nd•d deviation •e • c. anomaliesfor April 1987. The pattern is typicalof that for January-
July 1987. Positiveanomaliesalongthe equatorialsoneof the Pa-
ThusENSO eventsare thosein whichboth an SO extreme(Dar- cificare only ~IøC and are largestin the east. However,in the Pa-
win pressure
high, Tahiti low} and an EN occurtogether. A point cific the warmest water of ~30øC is shifted from west of the dateline
madeby TrenberthandShea[1987]is that the SO andEN are not to ~170øW. The correspondingOLR fields in Fig. 5 show that the
necessarilylinked on a one-to-onebasis. Abovenormal SSTs can oc- SPCZ is shifted north and east, as is characteristicof ENSO events
cur without an SO swing, suchas during 197g. In particular, South [Trenberth,
1976],andmerges
with the ITCZ (InterTropical
Conver-
American coastalevents are not very closelylinked to the SO except genceZone}near 170øW.The maximumtropicalconvection in the
during the major El Nifios where SSTs in the central Pacific are af- westernPacifichas shifted from 150øE to 170øW, as is shownalso by
fected. This is an aspectillustrated later here. ENSO eventsare also the OLR anomalyfield. The OLR anomalies
exceed45 W/m2 and
referredto as 'Warm Events"by vanLoonandShea[1985,1987]. are highly significant. Note the closelink with the total SST field
WeuseDarwinas an indexof the SO to showtwopoints1} the ex- but the relationshipwith the SST anomaly field is lessclear.
Geophysical Monograph Series Understanding Climate Change Vol. 52

120 TOGA AND ATMOSPHERIC PROCESSES

o 3OE 60E 90E ! I$OE I•0 15011 I•01,1 go• 60• •0•
go

6O

•0N

$05

605

•OS ß
0 30E 60E 90E

50N 5O

40N 4O

•:)N

EQ EO

20S

40S 4O

50S

Fig. a. (i) AnnualmeanSSTfor 1050-79fromShea(1086).SSTsgreaterthan 28øCare stippled.(ii) Annualmean

OLR for 1974-83fromJanowiaket al. (1985).In the tropics,valueslessthan240W/m 2 arestippledandcorrespond
to regions of persistent high cloud.
Geophysical Monograph Series Understanding Climate Change Vol. 52

TRENBERTH 121

z z z o u3 (/3 (/3 Z Z Z o •3 •3 •/3

o o o uJ o o o o o o uJ o o o
LU

o.e,
,• . ,•,.>• •& •, ,., . - o
•
0 LU

I I I •

n
•:::::::I '

•
•::• ,
::::::::. i,

• •.'.

• z z z o • • • z z z o u3 (/3 •3
o o o • o o o o o o b.i o o o
Geophysical Monograph Series Understanding Climate Change Vol. 52

1 22 TOGA AND ATMOSPIIERIC PROCESSES

Z Z Z C• 0'•
0 0 0 La 0 0 0
uJ U'• UJ

z z z o (•1 •1 (•1
o o • I•J o o o
Geophysical Monograph Series Understanding Climate Change Vol. 52

TRENBERTH •2•

The evolution of the SST field in several critical regions is moni- beencomputedrelativeto the overallmeans1970-1986(whichdiffer
tored by the U.S. National MeteorologicalCenter'sClimate Analysis somewhat fromthoseusedin the CAC ClimateDiagnostics Bulletin).
Center(CAC)throughaverages overregions
labeledNino1+2 (0. As in Fig. 2, there is someevidencefor similar evolutionof SST
10øS,90-80øW),Nino3 (5øN-5øS,150.90øW)andNino4 (5øN-5øS, anomaliesin each event but the differencesamong the eventsare also
160øS-150øW),
seeFig. 6. marked. Once again, phaselockingwith the annualcycleis not very
strong. The 1986-87event is quite modestalongthe coastof South
America(Fig. 7). Until the middleof 1987,it alsofeaturedmediocre
anomaliesin the Nifo 3 region,but by July anomalieswereexceeding
those of the 1982-83 event. Although the anomaliesin Nifo 4 are
small(Fig. 8), for 1987they are the largestby far of all the ENSO
11to' 133'w eventsshownand, as seenfrom Fig. 4, they are sufficientto move
160'E I 150'w I
the location of the warmest water into this region. Consequently,it
o'q NINO-4I NINO-3 is the modest anomalies in Nifio 4 which turn out to be critical to
the current event.
This makesa very strong casefor knowingthe details of the SST
patternswell. It is especiallyimportant to know SSTs accurately,
to better than 0.5øC, in regionsof water warmer than 28øC since
Fig. 6. Areasaveragedto produceSST indicesfor the Nifio 1+2 subtle changescan alter the region where the warmest water occurs
region and for the Nifo 3 and Nifio 4 regions. by thousands
of km. Gadgilet al. [1984]notedthat deepconvection
mainly occurswhen SSTs exceed 28øC.

The Importance of the Warmest Water

In both Figs. 7 and 8 the SST anomaliesin thesethree regions
are shownfor all the El Nifio eventssince1970 (whichis the cur- Why is the warmest water, water at temperatures higher than
rent limit to reliablevalues),plusestimatesfor the 1940eventfor 28øC, so important? An alternative questionis why does the at-
Nino1+2 [fromRasmusson
andWallace,1983].Theanomalies
have mosphericconvectionoccur where it does? We have implied from
the abovediscussionthat the atmosphericconvectionpreferentially
tends to occur over the region of warmest water. In other words, the
atmosphericconvergencezone tends to be so located. In order to
Nifio- 1 + 2 Far Eastern Pacific 0-10 ¸ S understandthis, at least qualitatively, there are a number of points
that must be taken into account.
1) Masscontinuityin the atmosphere
impliesthat convection
cannot
occur everywhere. There are preferred scales.
.... 1940
2) Thereis a competition
for organized
convection.
Sincethe warmest
...... 1972 watertendsto be ~29øC (Fig. 2), SSTsgreaterthan about28øC
are neededbefore a region can be a competitor. Moreover, the
.... 1976 region of warmest water is likely to 'win"owing to effects aris-
ing from the Clausius-Clapeyron equation. The latter expresses
---1982
the saturation mixing ratio r, as a nonlinear function of tempera-
1986
/',,,/ , _ ture. For SSTs of 20 and 30øC, at 1000 mb, r, increasesfrom 15 to
28 g/kg. Moreover,the increase
in r, perdegreeincrease
in SST at
72' / 30øC is 1.7 times that at 20øC. The consequencesare illustrated

:1 . • : ' / • l ',! ',1-

I! -
in Fig. 9 and discussed
by Neelinand Held [1987]. Convective
instability is greatly and nonlinearly enhanced over warmer w•-
../ ß .; ! ß ter owing to Clausius-Clapeyron effects on the grossmoist static
stability which leads to deeper convectionand excites a positive
,( 1
feedbackloop that is critical in determining the ultimate outcome
I,-
(Fig. 9).
3} Thereis alsoa questionconcerning
howthe atmosphere
'feels'the
ocean and thus knows where the warmest water is. This is best
ß understoodthrough a perturbation approach. Supposethat the
convergence zone is movedfor somereason,then the low level at-
mosphericflow that providesthe convergencewill immediately re-

ß . • ;-"'. suit in hugesensibleand latent heat fluxesinto the atmospherein

thevicinityof thewarmestwater{Fig. 10}. The resultis that con-

"
øJ ßß vection tends to break out in that region and the convergencezone
moves back over the warm water. Then the wind is reduced over
the warm water and the surfacefluxes actually becomeminiraised,
JAN JUL • JUL JAN
which explainswhy it is not possibleto understand the answerto
YEAR -1 YEAR 0 YEAR this questionfrom a budget study standpoint. In the real atmo-
sphere,there are many transients, ranging over time scalesfrom
Fig. 7. Sequences
of SST anomalies
for the Nifo 1+2 region,for all individualconvection,to easterlywaves,to 40-50 day oscillations.
E1 Nifio events since 1970 plus 1939-41. We suspect,however, that the nature of the transientsin atmo-
Geophysical Monograph Series Understanding Climate Change Vol. 52

124 TOGA AND ATMOSPHERIC PROCESSES

Nifio-3 Eastern Equatorial Pacific Nifio-4 Central Equatorial Pacific

I lllll[ lllll[ll llll llllllllllllll lll I IIIll IIIIIIli illllllillilllilii Ill
i.

;,i-- /. ',•', - -. •
•' b ."/" ß
[- ,-'½
x.,J•.\
['l. ,(...'"•//'."•::]:
:1

• "
•' \.,; •,, _.k-.:"-' i:,-•..
V :/.. ; '.'•/ I
':A••...,
..../..,.. eeeeeß
ßße •

"-:
• '; /?....;
L; '."• \..'/
; : '..
ß':k.. ' /' ½•

Ii lillllllltlllllllllllll
..j
Illllllll :..:.
YEAR -1 YEAR 0 YEAR YEAR -1 YEAR 0 YEAR + 1

Fig. 8. Sequences
of SST anomalies
for Nifio 3 and Nifio 4 regions(seeFig. 6) for all E1 Nifio eventssince1970.

THE ATMOSPHERE FEELS THE SURFACE

WARM WATER THROUGH TRANSIENTS:
Atmospheric moist static energy
Supposethe convergencemoves.

Deeper
•,ction

More More
Moisture Latent Heat
genc Release
Fig. 10. Cartoon illustrating how the atmospherefeels the high
SSTs when the region of convergence,given by the cumuluscloud,
is shifted away from the warm water by transients. The vertical
arrows indicate enhanced sensible and latent heat fluxes into the
Fig. 9. Feedbackcycle showingwhy warm water is important. atmosphereand the thin arrow showsthe atmosphericflow.
Geophysical Monograph Series Understanding Climate Change
sphericgeneralcirculationmodets(GCMs) is rather differentand
GCM transients tend to be dominated by the 'convective adjust-
ment" process. The impressionis that in GCMs, the convection
is tied to the warmest water even more strongly than in the real
atmosphere.
The m•in point to be made here then, is that the responseof the
atmosphereto SSTs is a highly nonlinearprocess.Sincethe area of
fairly warm (>28øC) water is extensivea smaJlchangein SST can
alter the configurationand is capableof producinga major shift in
the locationof the convergence soneswith consequencesfor wherethe
releaseof latent heat occurs. In turn, this changesthe atmospheric
forcingof Rossbywavesand the teleconnections into mid-latitudes.
The needfor accurateSST fields,and the needto understandchanges
in SSTs can be seento be paramount, and thus these are central to
the TOGA program. However, it should be noted that other, in
particular radiative, effectsare atso important.
Finally,sinceatmospheric windsdrivethe oceanand, as discussed Trenberth, K. E., Spatial and temporal variations of the Southern
here,SSTsare critical in determiningthe atmospheric response,it is
essentialto considerthe coupledsystemin order to fully understand Trenberth, K. E. and D. A. P•olino, Jr., Characteristic patterns of
atmosphere-oceaninteractions.
Acknowledgments. I am grateful to Vern Kousky and CAC for
supplyingme with the SST indexvalues. The NationalCenterfor
Atmospheric Research is sponsored by the NationalScience
Founda•
tion.
References
Branstator,G., Analysisof generalcirculationmodelseasurfacetem-
peratureanomalysimulationsusinga linear model. Pt. 1. Forced
solutions. J. Atrnos. Sci., 4œ,2225-2241, 1985.
ClimateDiagnosticsBulletin, Climate AnalysisCenter, NOAA, Wash-
ington, D.C.
Gadgil, S., P. V. Josephand N. V. Joshi,Ocean-atmosphere coupling WCRP, Scientific plan for the Tropical Ocean and Global Atmo-
over monsoonregions. Nature, $1œ,141-143, 1984.
Janowiak,J. G., A. F. Kruger, P. A. Arkin and A. Gruber, Atla.s
of outgoinglong-waveradiationderivedfrom NOAA satellitedata.
NOAA Atlas No. 6, U.S. Dept. of Commerce, 44 pp, 1985.
Vol. 52
TRENBERTH •25
National Academy of Sciences,U.S. participation in the TOGA Pro-
gram: A researchstrategy. National Academy Press, 24 pp, 1986.
Neelin, J. D. and I. M. Held, Modeling tropical convergencebased on
the moist static energy budget, Mort. Wea. Rev., 115, 3-12, 1987.
Rasmusson,E. M. and T. H. Carpenter, Variations in tropical sea
surface temperature and surface wind fields associatedwith the
SouthernOscillation/ElNifio, Mort. Wea. Rev., 110, 354-384,
1982.
Rasmusson,E. M. and J. M. Wallace, Meteorological aspectsof the
El Nifio/SouthernOscillation.Science,œœœ,
1195-1202,1983.
Shea, D. J., Climatological Atlas 1950-79: Surface air temperature
precipitation,
sea-level
pressure,
andsea-surface
temperature{45øS
-
90øN),NCAR Tech.NoteNCAR/TN-I•69-I-STR,1986.
Shukla, J. and J. M. Wallace, Numerical simulation of the atmo-
sphericresponseto equatorial sea surface temperature anomalies.
J. Atmos. Sci., 40, 1(•13.1(•30, 1983.
Oscillation. Quart. J. Ro•. Meteor. Soc., 102, 639-653, 1976.
variability of sea level pressurein the Northern Hemisphere. Mort.
Wea. Rev., 109, 1169-1189, 1981.
Trenberth, K. E. and D. J. Shea, On the evolution of the Southern
Oscillation, Mort., Wea. Rev., 115,3078-3096, 1987.
van Loon, H., 1984: The Southern Oscillation. Part III. Associations
with the trades and with the trough in the westerliesof the South
Pacific Ocean. Mort. Wea. Rev., 11œ,947-954, 1984.
van Loon, H. and D. J. Shea, The Southern Oscillation. Part IV. The
precursorssouth of 15øS to the extremes of the oscillation. Mort.
Wea. Rev., 115, 2063-2074, 1985.
van Loon, H. and D. J. Shea, The Southern Oscillation. Part VI.
Anomalies of sea level pressureon the Southern Hemisphere and of
Pacificsea surfacetemperature during the developmentof a Warm
Event. Mort. Wea. Rev., 115, 370-379, 1987.
sphereProgramme. WCRP No. $, WMO/TD No. 64, 147 pp,
1985a.
WCRP, International conferenceon the TOGA Scientific Program.
WCRPNo. 4, WMO/TD No. 65, 1985b.
Geophysical Monograph Series Understanding Climate Change Vol. 52

TOGA REAL TIME OCEANOGRAPHY IN THE PACIFIC

David Halpern

Earth and SpaceSciencesDivision,Jet PropulsionLaboratory

CaliforniaInstitute of Technology,Pasadena,CA 91109

Abstract. Reliable estimates of the evolution Episodes of high SST in the equatorial Pacific
of large scale sea surface temperature (SST) vat- are strongly correlated with regional and, often-
iations up to several months in advance is a prim- times, global atmospheric circulation changes
ary goal of the 1985-94 Tropical Oceans & Global (Rasmusson and Wallace, 1983). The decade-long
Atmosphere (TOGA) Program. Since the beginning of (1985-94) Tropical Ocean & Global Atmosphere
the TOGA, significant innovative accomplishments (TOGA) program, which is one of six subprograms of
include (1) an increase in the quantity of in situ the World Climate Research Program, was formed
data and efficiencies of data management, (2) focus upon the coupled ocean-atmosphere system.
rapid distribution of real time ocean products, The primary aim of the TOGAprogram is the devel-
(3) effective utilization of global observations opment of an operational capability for dynamical
from satellites, and (4) assimilation of data into prediction up to several months in advance of the
an ocean general circulation model to simulate time averaged (month-to-season) anomalies of the
monthly mean features of upper ocean thermal and coupled tropical ocean-atmosphere system.
flow fields. TOGA accomplishments are demonstrat- Variations of SST in low latitudes are related
ed with a discussion of oceanographic conditions to the responses of upper ocean thermal and flow
during June 1987. fields to large scale changes in surface winds, in
addition to local influences due to air-sea mo-
Introduction mentum and surface heat fluxes. Prediction of the
onset of large scale, long period ocean-atmosphere
Episodes of anomalous warm surface water, which interactions up to several months in advance re-
typically exist for about l-year, occur in the quires knowledge of the evolution of equatorial
equatorial Pacific Ocean at irregular intervals SST variations on time scales of days to a
every 4-7 years. During the extremely warm epi- month. This time scale is short compared to the
sode of 1982-83, the monthly mean equatorial sea one applicable for middle latitudes. Philander
surface temperature (SST) rose above its climato- (1979) reported that basin wide density gradients
logical-mean monthly value by as much as 4øC. in the upper portion of a resting ocean would be
This high SST covered a wide area (perhaps 40 ø established in about a decade in middle latitudes
longitude by 10ø latitude) centered approximately and only weeks near the equator. Remote wind
along the equator, intensified the Hadley circula- effects in the tropical ocean are important on
tion, and displaced the Walker circulation (Ras- time scales of weeks because of the rapid propaga-
musson and Wallace, 1983). The upward, eastward tion of dynamic signals, such as Kelvin wave mo-
slope of the thermocline along the equator de- tion along a narrow equatorial zone (Knox and
creased (Halpern, 1987). Sea level decreased in Halpern, 1982).
the western equatorial Pacific, and increased by At the time of onset of E1 Nino of 1982-83,
more than 20 cm in the eastern region (Wyrtki, surface wind measurements from the World Meteoro-
1984). The strengths of the eastward flowing logical Organization volunteer observing ship
Equatorial Undercurrent and westward flowing South (v.o.s.) network and SST observations from v.o.s.
Equatorial Current were substantially reduced and satellite radiance retrievals were the only
(Halpern et al., 1983; Halpern, 1987). The zone oceanographic parameters recorded throughout the
of intense atmospheric convection, which normally equatorial Pacific and distributed as data or data
resided west of about 160øE, moved eastward and products in (oceanographic) real time. Within the
traversed the entire width of the Pacific (Liu, ocean sciences, real time means approximately 30-
1988). days (sometimes up to 60-days) of measurement.
Expendable bathythermograph (XBT) measurements of
Copyright 1989 by the vertical profile of temperature between the
International Union of Geodesy and Geophysics surface and about 450 m were made from a small
and American Geophysical Union. subset of about 25 ships within the vast v.o.s.

127
Geophysical Monograph Series Understanding Climate Change Vol. 52

128 TOGA REAL TIME OCEANOGRAPHY IN THE PACIFIC

network. but five •ears aao these data were •ot several publications (Table 1). This innovative
distributed in real time. To achieve the TOGA development of real time viewing of oceanic condi-
objective of dynamical prediction up to several tions will be discussed in this paper. The Paci-
months in advance, two activities required devel- f[c Ocean is emphasized because distribution of
opment: (1) an operational capability of record- oceanographic data products in real time has de-
ing, transmitting, and distributing data on sea veloped more rapidly there than in the Atlantic or
level, subsurface thermal and flow fields, and net Indian Oceans. This paper is not an exhaustive
air-sea heat and momentum fluxes; and (2) assimi- treatment of all oceanographic research withi. n the
lation of oceanographic data into an operational TOGA program.
ocean general circulation model (until a reliable
interactive ocean-atmosphere general circulation TOGA Real Time Oceanography
model becomes available) for simulation and pre-
diction of oceanographic conditions. The fullness of TOGA real time oceanography was
It was recognized immediately by TOGA scien- demonstrated at the XIX General Assembly of the
tists that the oceanographic data base required International Union of Geodesy and Geophysics
improvements. One of the perpetual contributions (IUGG) in August 1987 when a preliminary version
of TOGA will be the substantial increases in quan- of this paper was presented. Attention was focus-
tity and quality of the subsurface oceanographic ed upon June 1987 oceanographic conditions because
fields. TOGA fostered an awareness of real time real time data products become available with a
oceanography and of prediction of large scale, delay of about 1-month, making June the period
long period SST variations and related features closest in time to the IUGG Assembly when data
such as thermocline depth, zonal slope of thermo- would be available. Also, the warm episode in the
cline along equator, and anomalous behavior of Pacific, which began in 1986 (Bergman, 1987) was
surface current. This situation was virtually continuing, which provided an opportunity for
nonexistent in oceanography before TOGA. This too discussion of ocean dynamics associated with this
is an everlasting contribution of TOGA. As a moderate E1 Nino.
result of the emphasis upon rapid transmission of The 16 June - 1 July 1987 averaged SST anomaly
TOGA data, a number of monthly publications or pattern (Figure 1) resembled the August - October
bulletins regularly display oceanographic data. anomaly distribution of the composite E1 Nino
Examples are listed in Table 1. described by Rasmusson and Carpenter (1982), but
As an indication of the progress due to TOGA, the 16 June - 1 July 1987 maximum amplitude of 2øC
many oceanographic aspects of the evolution of the was more than 0.5øC larger. Levitus (1987) showed
warm episode of 1986-87 were described every month that the magnit,ide of the SST anomaly was depend-
from its beginning via data products displayed in ent upon the climatology and data analysis tech-

TABLE 1. Regularly occurring bulletins documenting upper tropical ocean and

global/regional surface meteorological conditions, and addresses to secure addtional
information.

Bulletin Address
1. Climate Diagnostic Bulletin Climate Analysis Center/NMC, National Weather
Service, National Oceanic and Atmospheric
Administration, Washington, DC 20233, UNITED
STATES OF AMERICA
2. Climate Monitoring Bulletin - Bureau of Meteorology, National Climate Centre,
Southern Hemisphere P.O. Box 1289K, Melbourne 3001, AUSTRALIA
3. Climate System Monitoring Monthly World Climate Program, WMO Secretariat, Case
Bulletin Postale No. 5, CH-1211 Geneva 20, SWITZERLAND
4. Darwin Tropical Diagnostic Statement Bureau of Meteorology, P.O. Box 735, Darwin,
Northern Territory 5794, AUSTRALIA
5. ERFEN Boletin de Analisis Climatico Comision Permanente del Pacifico Sur, Calle 76
No. 9-88, Apartado 92292, Bogota, COLUMBIA
6. MEDS Realtime Data Monthly Marine Environmental Data Service, 200 Kent
Monitor Report/Drifting Buoys Street, Ottawa, Ontario K1A 0E6, CANADA
7. Monthly Report on Climate System Long-Range Forecast Division, Japan
Meteorological Agency, 1-3-40te-machi,
Chiyoda-ku, Tokyo, JAPAN
8. Oceanographic Monthly Summary National Ocean Service, National Oceanic and
Atmospheric Administration, 5200 Auth Road,
Camp Srings, bid 20746, UNITED STATES OF AMERICA
9. Veille Climatique Satellitaire Centre de Meteorologie Spatiale, B.P. 147,
Lannion 22303, FR•CE
Geophysical Monograph Series Understanding Climate Change Vol. 52

HALPERN 129

140øE 160 ø 180 ø 160 ø 140 ø 120 ø 100 ø 80øW

20ON
| I I J I /I I I I / I I ..l I I 120ON
•0øS • - • • • • • • • • • • • • • • • •0øS
140øE 160 ø 180 ø 160 ø 140 ø 120 ø 100 ø 80øW

Fig. 1. Sea surface temperature (SST) anomaly during 16 June - 1 July 1987. Climatologi-
cal-mean monthly SST was defined by Reynolds (1982). Contour interval is IøC and a dashed
line represents a negative value, which means that SST was lower than normal. The 15-day
SST product represents a blend of in situ, but not drifting buoy SST observations, and
satellite data recorded during the previous 15 days. Redrafted from a chart distributed
by the NOAA Ocean Product Center.

nique used to compute the climatology. Comparing ed with XBTs launched from the v.o.s. network. As
the SST anomaly distributions for the 15-day in- a result of the TOGA program, XBT lines are now
tervals before and after the 16 June - 1 July 1987 more uniformly distributed throughout the tropical
period indicated that the location of the SST Pacific. Nearly 10,000 XBTs are launched annually
anomaly was not constant. in the tropical Pacific. Although there seems to
Near surface currents are recorded by the move- be enough XBT observations to produce an adequate
ments of freely drifting buoys drogued at about 15 realization of the monthly mean thermal field,
m depth and tracked several times each day by the bimonthly data products are generated by Pazan et
Argos data collection and platform location system al. (1987). The depth of the 14øC isotherm, which
on-board NOAA polar orbiting satellites. This usually occurs near the bottom of the thermocline,
technique was used successfully during E1 Nino of is an indicator of thermocline depth. During May
1982-83 (Halpern et al., 1983). During June 1987 - June 1987 the thermocline along the equator was
there were 29 dr•fters afloat •n the 20øN - 20øS shallower (deeper) than a 4-year mean depth, which
Pacific (Figure 2). The comparison between the was computed from data recorded during 1979, 1980,
May 1987 surface current field and the June 1987 1981, and 1984, •n the western (eastern) region
distribution indicated that the near equatorial (Figure 3). Throughout the region 10øN - 10øS
eastward current observed in May between 140øW and westward (eastward) of ]30øW, the thermociine
110øW was now confined to a smaller region near depth was less (greater)than the mean depth.
130øW. The eastward flow in May 1987 may have In the tropical oceans a close relationship
contributed to the development of the SST anomaly exists between sea level and the depth of the
in June by advection of warm water from the west, thermocline. Along the equator, changes in zonal
a situation similar to that found during the 1982- sea level slope mirror, to a large degree, varia-
83 E1 Nino (Halpern, 1987). tions in zonal slope of the thermocline (Wyrtki,
Since about 1978, numerous temperature profiles 1984). Two sources of sea level data existed: in
within the upper 500 m have been routinely measur- situ and satellite data. Tide gauge measurements

140øE 160ø 180 ø 160 ø 140 ø 120 ø 100 ø 80øW

20øS
I '•I I ! I I I I
140øE 160
ø 180
ø 160
ø 1410
ø
I I I I I I\• 20øS
120
ø 100
ø 80øW
Fig. 2. Trajectories of satellite-tracked drifter buoys drogued at about 15 m depth
during June 1987. Redrafted from a chart which appeared in the Climate Diagnostic Bulle-
tin.
Geophysical Monograph Series Understanding Climate Change Vol. 52

150 TOGA REAL TIME OCEANOGRAPHY IN THE PACIFIC

140øE 160• 180 ø 160 ø 140 ø 120 ø 100 ø 80øW in situ data was caused by the absence of station
data. The prominent -20 cm sea level departure at
about 5øN, 135øW was typical of the annual cycle
associated with the North Equatorial Countercur-
rent.

Much of the structure portrayed by the Geosat

data was simulated by Leetmaa (personal communica-
tion) with a wind-driven ocean general circulation
model. This primitive equation model (Philander
and Seigel, 1985) has been compared on several
occasions with tropical surface and subsurface
ocean observations with satisfactory results
(Richardson and Philander, 1987; Philander and
140øE 160 ø 180 ø 160 ø 140 ø 120 ø 100 ø 80ow
Seigel, 1985; Garzoli and Philander, 1985). The
Fig. 3. Normalized depth of the 14øC isotherm modeled surface dynamic height anomaly for June
during May - June 1987. Anomaly was defined as 1987 (Figure 4C) was defined as the difference
departure from average monthly mean distribution between the June 1987 simulation forced with June
computed during 1979, 1980, 1981, and 1984. 1987 surface winds determined from v.o.s. wind
Dashed contours represent negative values, which data and the climatological-mean monthly simula-
mean that the 14øC isotherm was shallower than tion forced with Hellerman and Rosenstein's (1983)
normal. Redrafted from a chart which appeared winds for June. The model result and Geosat data
in the Climate System Monitoring Monthly Bulletin. showed a similar pattern of sea surface height
variation in the North Equatorial Countercurrent
where heights were less than the mean (although
showed that the June 1987 sea level deviation from different mean values were used in Figures 4B and
the 7-year (1975-81) mean sea level was below 4C). The model simulation of sea level and in
average in the western Pacific with a-27 cm mini- situ sea level measurements (Figures 4A and 4C)
mum in the Solomon Islands and was above average had excellent correspondence in the southwest
ia the eastern Pacific (Figure 4A). The previous Pacific with mimimum sea level near 10øS, 160øE.
month's distribution was very similar. Along the The in situ tide guage network did not contain any
equator, the June 1987 sea level deviation repre- spatial structure of sea level eastward of 140øW
sented a decrease in the zonal pressure gradient, (Figure 4A) because of the absence of data, in
a decrease of near surface westward flow, and a contrast to the modeled surface dynamic height
reduction in the Equatorial Undercurrent. A sec- (Figure 4C), which contained fewer occurrences of
ond sea level data set is produced by the U.S. alternating positive and negative departures from
Navy's Geosat mission, which measures sea surface mean sea level than the Geosat data (Figure 4B).
topography with a root-mean-square error in the The magnitude of the differences between Figures
western equatorial Pacific approaching 4 cm for 2 ø 4B and 4C were about as large as the monthly mean
latitude by 8 ø longitude regions and time scales values. Because the pattern of monthly mean dif-
longer than about 1-2 weeks (Cheney et al., 1986; ferences varied from month-to-month, it is not
Wyrtki, 1987). The 1 ,June 1987 sea level depart- clear which of these two sea level products is
ures from a 1-year (April 1985-March 1986) annual more representative of the natural state. The
mean distribution based upon Geosat data (Figure reasonable agreement between model simulated sur-
4B) has features both different and similar to the face dynamic height, Geosat data, and in situ sea
sea level pattern estimated from in situ data level measurements represented an advance towards
(Figure 4A). Along the equator, the reduced zonal the TOGA objective of developing an operational
slope of sea level was produced in both data capability for dynamical ocean prediction.
sets. In the central Pacific the positions of the One reason for the success of ocean
general
zero isoline were very different: essentially a c[rc,•latio• models in tropical regions is that the
north-south trend for the in s[tu data and an thermal and flow variations are not very intense
east-west trend for the satellite data. This was and essentially linear. Current and temperature
due in part to the nearly total lack of tide guage v•r[ations in low latitudes are more strongly
data from the central and easter,] Pacific in con- related to wind forcing than to instabilities of
trast to the regularly spaced altimeter data. the flow field, which co,nmonly occur in middle
Within the approximate 5øN - 10øN North Equatorial latitudes. In tropical regions, most of the meso-
Countercurrent, sea level departures from annual scale variability is associated with long (approx-
mean sea surface topograph•es were not the same. imately 1000-km wave length) linear or weakly
Dissimilar definitions of long term sea surface nonlinear wavelike features (Legeckis, 1977; Hah-
topography used in Figures 4A and 4• contributed sen and Paul, 1984; Halpern et al., 1988), in
to the contrasting patterns, and the tide guage contrast to the intense smallscale eddies and
data should be analyzed relative to an April 1985- rings prevalent throughout the mid-latitude
March 1986 mean value. In the eastern and central ocean. Essential ocean dynamical features in low
regions, the much reduced spatial structure of the latitudes, except for microscale mixing processes,
Geophysical Monograph Series Understanding Climate Change Vol. 52

H•LPERN 131

(A)
140" E 160•' 180 •' 160" 140 120 100' 80 'W

• •. -10 -5
'• . -20 -15- _/ / _ /

I I I I I I I I I I I• I 20•S

(B)
20ON [ I I I I I I I I I I I I I 20ON

0 0

0ø --• -""-•0--- -- ....... 4

0o

20•S I 20øS

(c)
20øN I i I ! I 20ON

-12 ..... ' 4

0o 0o

20øS 20øS
140øE 160ø 180ø 160ø 140ø 120ø 100ø 80øW

Fig. 4. (A) Sea level during June 1987 estimated from tide guage measurementsrecorded at
island and coastal stations. This map represents departure of sea level from 7-year (1975
- 81) mean. Solid dots represent tn situ sea level stations. Dashed lines represent
negative values, which mean that sea level was lower than the mean. Redrafted from a
chart appearing in the Climate Diagnostic Bulletin. (B) Sea level anomaly during 1 June
1987 estimated from Geosat sea surface topography data using daily data, a 15-day decorre-
lation time, and a grid of 8ø longitude by 2ø latitude. Anomaly is defined as difference
from 1-year (April 1985 - March 1986) mean. Dashed contours represent negative values.
Redrafted from a chart which appeared in the Climate Diagnostic Bulletin. (C) Surface
dynamic height anomaly during June 1987 estimated from an ocean general circulation mod-
el. Anomaly defined as difference between monthly mean distribution and climatological-
mean distribution determined from Hellerman and Rosenstein's (1983) wind field. Dashed
contours represent negative values. Redrafted from a diagram kindly provided by Dr. Ants
Leetmaa (Climate Analysis Center).

are resolvable by primitive equation models run on cline depth and of the zonal pressure gradient
currently available supercomputers. Equatorial force in the upper ocean. Simulations of the
wave-like structures were simulated by Philander monthly difference of the depth of the 20øC iso-
et al. (1986) with a 33-km latitudinal by 100-km therm relative to the climatological-mean monthly
longitudinal model geometry near the equator. distribution computed with the Hellerman and Ros-
The depth of the 20øC isotherm along the equat- enstein (1983) wind field were computed by Leetmaa
or, which usually occurs in the middle of the (personal communication). Figure 5 indicated that
thermocline, is a suitable indicator of thermo- maximum departures from the mean may have been
Geophysical Monograph Series Understanding Climate Change Vol. 52

132 TOGA REAL TIME OCEANOGRAPHY

IN THE PACIFIC

140OE 160ø 180 ø 160 ø 140ø 120 •' 100ø 80•W April 1987 the westerly wind anomaly was confined
o I i i i ! I I I / I I I i I -o to the 170øE to 170øW region, but in May the west-
1986[x •0_• • • •- %• • L I•10 -N
1986
erly wind anomaly increased in strength and longi-
tudinal coverage from 160% to 140øW (Figure
•,r"•J•/'•
D , '-'--:'C--'
30' ,- -D
6B). In June the westerly wind anomaly was limit-
-3 ed to the region from the date line to 100øWand
19871./ •0 -•10••.•,,•• '• 1987
- F the easterlies in the far western Pacific were
-M stronger than normal. If this situation had con-
" 30 20 ttnued and expanded towards the east, it would
-A signal the demise of E1 Nino conditions: this did
-M not occur and SST continued to be warm.
It is tempting to associate April and May 1987
surface wind patterns with the June 1987 oceano-
J-•{)
N,
-20
•k._ .,... •,5\ •('/..---/._..30
-20 •"•
•'•• - AJ graphic conditions.
that reduced
Conventional wisdom dictated
easterlies in May would produce in
S June a decreasedeast-west tilt of the thermocline
along the equator (Figures 3 and 5), elevated and
'-'l I I I I I I I I I I I I I I I o depressed sea level in the eastern and western
140øE 160ø 180ø 160ø 140ø 120ø 100ø 80øW
equatorial regions (Figure 4), respectively, and a
reduction in the strength of the South Equatorial
Fig. 5. Time - longitude section of the model Current along the equator (Figure 2). The east-
simulation of monthly mean anomaly of the 20øC ward direction of the buoy drifting near 0 ø, 130øW
isotherm depth. Anomaly defined as difference was anomalous because the South Equatorial Current
between monthly mean depth of 20øC isotherm and is usually well developed in June (Halpern,
climatological-mean monthly depth determined from 1987). This eastward motion was indicative of
a wind-driven ocean model using Hellerman and advection of warmer water from the west, which
Rosenstein's (1983) wind field. Dashed contours contributes significantly to the generation and
represent negative values, which mean that the maintenance of the anomalousSST pattern (Figure
20øC isotherm was shallower than normal. Re- 1). The reduced equatorial upwelling caused by
drafted from a chart which appeared in the Climate the westerly wind anomaly also contributed toward
SystemMonitoring Monthly Bulletin. the anomaly of high SST, though no measurements
exist to quantify this assertion.

reached in June 1987; subsequent model runs for Discussion

several months after June 1987 verified this con-
clusion. In June 1987, the thermoclinewas deeper The net heat flux betweenthe oceanand atmos-
east of 130øW(i. e., about 30 m deeper between phere is a significant componentof SST varia-
105øWand 95øW) and shallower to the west, with tions, but currently there is no readily access-
anomalousnegative values (i. e., shallower) of ible and reliable product. Monthly meannet sur-
more than 20 m west of 175øW. This model result face heat flux variations estimated from v.o.s.
is sensitive to the surface winds: a 20% change in data (e. g., wind, air and sea surface temper•-
the surface
windstresswill produce
a thermoclinetures,humidity,
cl•udamount,
etc.) areuncertain --

depthdifference of about25 m (Harrisonet al., to •30 to •50 Wm ß A monthlymeannet surface

1988),whichis about50%of theabsolute
range heatfluxuncertainty
of 40Wm
-2 corresponds
to a
displayedin Figure3. monthlymeanSSTvariation of 1.2•C for an iso-
In the equatorial zone, the east-west surface thermal mixedlayer of 20 m thickness.
wind stress componentis very important in the Muchprogressis being madetowardsdevelopment
generationof thermaland flow variations (Gill, of satellite analysesof the constituentsof the
1982). Manysurface wind productsexist at pre- net surface heat flux. For instance, the surface
sent, but intercomparisontests revealed large latent heat flux, whichis a large contributorof
differences (Halpern and Harrison, 1982), primar- air-sea heat flux, can be estimatedfromsatellite
ily becauseof the scarcity of in situ surface measurements
of surfacewindspeedandtotal water
wind data. As a result of the TOGAprogram, the content in the atmosphere(Liu, 1988). Short wave
numberof island and mooredbuoystations report- radiation flux, whichis also a large component
of
ing wind data in real time wasincreaseddramatic- the net surface heat flux variation betweenclear
ally. At Christmas
Island (2øN,157•W),the April and cloudyregions, seemsto be measurable
from
1987 zonal surface wind componentwas essentially space (Gautier, 1988), and there is expectation
the sameas the 25-yearaverageApril windcomput- that longwaveradiative flux mightalso be deter-
ed by Wyrtki andMeyers(1976). However, in May mlnedfromsatellite data (Frouinet al., 1988).
the zonal componentbecamea westerly anomaly(i. Variability in the tropical oceanis primarily
e., a reductionin speedof the normallyoccurring atmosphericallyforced, yet little is knownabout
easterlyor westward
wind)of 4-5 m s-1 andthe the structureof the surfacewindfield. The
wind even blew towards the east (Figure 6A). In sensitivity of monthly meanSST simulations to the
Geophysical Monograph Series Understanding Climate Change Vol. 52

HALPERN 133

(A)

t I I I I !
MHR HPR MHY JUN JUL
1987

(B)
140oE 160 180 160• 140• 120 100• 80•W
I I •1 7;rl / I I ! / • 7 I I I I I I I 20ON
/ , • , . • • • • • • • • - • t • • • •• ..... t_ L

20øS F:< i
140øE 160 180 160 TM,'- : ' ' 120 o "' ': -
•8. 6. (A)Time
varla[ions
offsurface
zonat
windcomponent
(• s-t) measured
at Christmas
Island (20N• I57"W)• which 1s one of several opera[ional• rear time ?OOAequa[oriat •acif-
tc 1sland and mooredbuoy stations. Dashed t•nes represent 25-year (t947-72)
cal-mean monthly zonat componentwind determinedby Wyr[k[ and Meyers (t976). D[asram
kindly producedby •aut Frei•a8 (Pacific Marine Envlronmen[atLaboratory). (B) Surface
wlndstressanomaly
durin8May1987. A windstressvectoris de•t½ed_½s
the w•ndcompo-
dynecm assumln8
a dra• coefficiento• t.4xtO• andair densityo• t.2 k8 m- ß Monthly
meananomal• deffinedas departure from 23-year (t96t -83) mon[htymeandls[rlbu[ton.
wes[erl• wlnd anomalymeansa reduction &n the speedo• the normally occurtin8 easterly
westward wind. Adapted •rom a chart which appeared in the Cttma[e Olasnos[[c Butte[tn.

monthlyaveragedsurface wind was demonstratedby than the total numberof v.o.s. surface wind ob-
Harrison et al. (1988). A 20% error in wind servations measuredthroughouthistory. Unfortu-
stress, which is about the minimum
error expected nately, these satellite scatterometers do not
with presenttechnology,
wasequivalent
to a 2øC yield accura_•e
estimates
of the windspeed
below
uncertainty in $ST. Large wind speed variations about 3 m s ß A new, innovative satellite scat-
occur on short time scales of 1- to 5-days (Figure terometer designed to measurethe tropical surface
6A). Th•
0.5 ms- , accuracy
isof
which monthly
mean
necessary
for v.o.s.studies,
climatedatato wind
•20%field
for with
wind a root-mean-square
lessthan3 ms-1accuracy
speeds of
is current-
is doubtful, except along a few well-traveled ly being considered for developmentby NASA.
shipping routes becauseof inadequatesampling A major obstacle remainingin the development
(Halpern, 1988). Rosati and Miyakoda(1988) ira- of ocean modelsfor simulation of SST is the par-
provedthe SSTsimulationby usingwindswith a 1- ameterizationof horizontal and vertical turbulent
day Nyquist period, which substantially reduced mixingsin the upper ocean. Philanderand Seigel
the underestimation of the net surface heat flux. (1985) used a vertical mixing parameterization
In the early 1990s, every day there will be based upon the Richardsonnumber. Rosati and
morethan 50,000 surfacewindvectorswith a reso- Miyakoda(1988) demonstratedthat a turbulence
lution of nearly 50-kmavailable from the National closure schemewas more appropriate than constant
Aeronautic and SpaceAdministration's (NASA)scat- eddy viscosity. Evaluating different mixing par-
terometer (called NSCAT),whichmaybe launchedin ameterizations is exceedinglydifficult becauseof
1993 on Japan's ADEOSsatellite, and from the the scarcity of suitable in situ measurements.
EuropeanSpaceAgency's(ESA)scatterometersched- Peters et al. (1988) reported substantial dtffer-
uled to be launchedin 1991 on the ERS-1 satel- ences betweenPhilander and Seigel's (1985) Rich-
lite. The number of satellite surface wind vec- ardson number formulation and Ln situ turbulent
tots to be recorded in one monthwill be greater mixing data recorded at 0ø, 140øW. However, the
Geophysical Monograph Series Understanding Climate Change Vol. 52

154 TOGA REAL TIME OCEANOGRAPHY IN THE PACIFIC

in situ mixing observations should not be applied tory, California Institute of Technology, under
beyond the time and place of their measurement contract with the National Aeronautics and Space
until more is learned about the causes of their Administration.
variability because dissipation rates were 100
times larger at night than during daytime. References

Conclusion Bergman, K. H. (1987) The global climate of Sep-

tember - November 1986: a moderate ENSO devel-
TOGA ocean studies are the latest beneficiaries ops in the equatorial Pacific. Monthly Weather
of numerous multidisciplinary tropical oceanogra- Review, 115, 2524-2542.
phy projects conducted during the previous 15 Cheney, R., B. Douglas, R. Agreen, L. Miller, D.
years. Since the beginning of the TOGA decade in Milbert and D. Porter (1986) The Geosat alti-
January 1985, significant innovative accomplish- meter: a milestone in satellite oceanography.
ments include (1) increase in the quantity of in Transactions American Geophysical Union, 67,
situ data and efficiencies of data management, (2) 1354-1355.
real time ocean products and their rapid distribu- Frouin, R., C. Gautier and J.J. Morcerette (1988)
tion, (3) effective utilization of global observa- Downward long wave irradiance at the ocean
tions from satellites to estimate sea surface surface from satellite data: methodology and in
topographyand latent heat flux, and (4) numerical situ validation. J•urnal of Geophysical Re-
model simulation of oceanographic features. search, 93, 597-620.
The possibility of the TOGA program achieving Garzoli, S. and S. G. H. Philander (1985) Valida-
measurable success in the prediction of time aver- tion of an equatorial Atlantic simulation model
aged anomalies up to several months in advance using inverted echo sounder data. Journal of
rests with effective utilization of satellite Geophysical Research, 90__,9199-9201.
observations and assimilation of observations into Gautier, C. (1988) Surface solar irradiance in the
coupled ocean-atmosphere models Presently, the central Pacific at 10øS 140øW during Tropic
differences between model simulation and measure- Heat: in situ and satellite measurements.
ments were equivalent to uncertainties caused by Journal of Climate and Applied Meteorology, in
inadequate wind field and net surface heat flux. press.
A 20% error in wind stress produces uncertainties Gill, A.E. (1982) Atmosphere-Ocean Dynamics.
of 2•C in SST and 25 m in the depth of the thermo- Academic Press, New York, 662 pp.
cline along the equator. •ssimilation of XBT Halpern, D. (1988) On the accuracy of monthly mean
measurements partiallyA compens•Ses
adequate wind data. 40 W m
for the
erro•
lack of
in the net
wind speeds over the equatorial Pacific. Jour-
nal of Atmospheric and Oceanic Technology, 5,
surface heat flux yields a 1.2•C SST error for a 362-367.
20 m thick mixed layer. Halpern, D. (1987) Observations of annual and E1
The future challenge is to use satellite data, Nino thermal and flow variations at 0 ø, 110•W
in situ observations, and general circulation and 0 ø, 95øW during 1980-1985. Journal of
models synergistically. Perhaps our colnplete Geophysical Research, 92, 8197-8212.
understanding of the physics of tropical SST vat- Halpern, D. and D. E. Harrison (1982) Intercompar-
iations awaits analyses of model results and a few ison of tropical Pacific mean November 1979
specialized observations. The TOGA decade, which surface wind fields. Report 82-1, Department
has already altered perceptœons of tropical oceaa- of Meteorology and Physœcal Oceaaography,
ography and ocean-atmosphere interactions, will Massachusetts Institute of Technology, Cam-
undoubtedly have an everlasting impact upon global bridge, 40 pp.
geosciences. Halpern, D., S. P. Hayes, A. Leetmaa, D. V. Hansen
and S. G. H. Philander (1983) Oceanographic
Acknowledgements. This paper is dedicated to observations of the 1982 warming of the tropi-
the memory of Adrian Gill ia recognition of hœs cal eastern Pacific. Science, 221, 1173-1175.
essential contributions to the development of the Halpern, D., R. A. Knox and D. S. Luther (1988)
TOGA program. I am grateful to many colleagues Observations of 20-day period meridional cur-
who regularly send their data product in advance rent oscillatœons in the upper ocean aloag the
of publication in the bulletins: R. Cheney and L. Pacific equator. Journal of Physical
Miller (Figure 4B); P. Freitag (Figure 6A); D. Oceanography,in press.
Hansen (Figure 2); A. Leetmaa (Figures 4C and 5); Hansen, D. V. and C. A. Paul (1984) Genesis and
J. O'Brien (Figure 6B); D. Rao (Figure l); K. effects of long waves in the equatorial Pacif-
Wyrtki (Figure 4A). This manuscript has benefited œc. Journal of Geophysical Research, 89,
from comments from many TOGA colleagues, in part- 10431-10440.
[cular M. Cane, R. Cheney, J.-R. Donguy, J. Flet- Harrison, D. E., W.•. Kessler and B. S. Giese
chef, C. Gautier, R. Lambert, A. Leetmaa, G. Mey- (1988) Ocean c•rculation model hindcasts of
ers, K. Takeuchi, Y. Tourre, K. Trenberth, and K. the 1982-83 E1 Nino: thermal variability along
Wyrtki. The research described œn this paper was the ship of opportunity tracks. Journal of
performed, in part, by the Jet Propulsion Labora- Physical Oceano•rap.h•, in press.
Geophysical Monograph Series Understanding Climate Change Vol. 52

HALPERN 135

Hellerman, S. and M. Rosenstein (1983) Normal Philander, S. G. H., W. J. Hurlin and R. C. Pacan-
monthly wind stress over the world ocean with owski (1986) Properties of long equatorial
error estimates. Journal of Physical Oceano- waves in models of the seasonal cycle in the
•raphy, 13, 1093-1104. tropical Atlantic and Pacific Oceans. Journal
Knox, R. A. and D. Halpern (1982) Long range Kel- of Geophysical Research, 91, 14207-14211.
vin wave propagation of transport variations in Rasmusson, E. M. and T. H. Carpenter (1982) Varia-
Pacific Ocean equatorial currents. Journal of tions in tropical sea surface temperature and
Marine Research, 40 Supplement, 329-339. surface wind fields associated with the South-
Legeckis, R. (197Y) Long waves in the eastern ern Oscillation/E1 Nino. Monthly Weather
ß

equatorial Pacific Ocean: a view from a geo- Review, 110, 354-384.

stationary satellite. Science, 197, 1197-1181. Rasmusson, E. M. and J. M. Wallace (1983) Meteoro-
Levitus, S. (1987) A comparison of the annual logical aspects of the E1Nino/Southern Oscil-
cycle of two sea surface temperature climat- lation. Science, 222, 1195-1202.
ologies of the world ocean. Journal of Ph¾si- Reynolds, R. W. (1982) A monthly averaged climat-
cal Oceanosr.•phy, 17, 197-214. ology of sea surface temperature. NOAA Tech-
Liu, W. T. (1988) Moisture and latent heat flux nical Report NWS 31, National Oceanic and At-
variabilities in the tropical Pacific derived mospheric Administration, Rockville, Maryland,
from satellite data. Journal of Geophysical 35 pp.
Research, 93, 6749-6760. Richardson, P. L. and S. G. H. Philander (1987)
Pazan, S. E., W. B. White and Y. He (1987) Annual The seasonal variations of surface currents in
report on tropical Pacific subsurface thermal the tropical Atlantic Ocean: a comparison of
structure - 1985. SIO Reference Number 87.1, ship drift data with results from a general
Scripps Institution of Oceanography, La Jolla, circulation model. Journal of Geophysical
California, 66 pp. Research, 92, 715-724.
Peters, H., M. C. Gregg and J. M. Toole (1988) On Rosati, A. and K. Miyakoda (1988) A GCMfor upper
the parameterization of equatorial turbu- ocean simulation. Journal of Physical Ocean-
lence. Journal of Geophysical Research, 93, osraphy, in press.
1199-1218. Wyrtki, K. (1984) The slope of sea level along the
Philander, S. G. H. (1979) Variability of the equator during the 1982/83 E1 Nino. Journal,,

tropical oceans. Dynamics of Atmospheres and of Geophysical Research, 89, 10419-10424.

Oceans. 3, 191-208. Wyrtki, K. (1987) Comparing Geosat altimetry and
Philander, S. G. H. and A.D. Seigel (1985) Simu- sea level. Transactions American Geophysical
lation of E1 Nino of 1982-83. In: Proceedings Union, 68•, 731.
of the 16th International Liese Colloquiumon Wyrtki, K. and G. Meyers (1976) The trade wind
Ocean H•drodynamics, edited by J. Nihoul, field over the Pacific Ocean. Journal of
Elsevier, New York, 517-541. Applied Meteorolo$¾, 15, 698-704.
Section VI

MODELLING CLIMATE,
PAST, PRESENT AND
Geophysical Monograph Series Understanding Climate Change Vol. 52

AERONOMY AND PALEOCLIMATE

J.-C. G•rard

Institutd'Astrophysique,Universit•de LiC•ge
B-4200Ougr•e-Li&ge,Belgium

Abstract. Solar-Terrestrial and aeronomic interactions ity) and in spectralcomposition

(solarirradiance).Fol-
with the Earth's globalclimate playeda key role through- lowing the sun formation 4.6 Ga ago and the T-Tauri
out the evolution of the planet. The evolving composition phasewhereboth the solarluminosityand the ultravio-
of the terrestrial atmospheredue to geochemicaland bio- let radiation increasedsubstantiallyduringapproximately
sphericinteractions strongly controledthe paleoclimate. 10 Ma, the solar luminosityincreasedsteadily. It rose
An example is the apparent absenceof glaciation in a pe- from a value about 215-30%less than the present one as
riod when the solar luminosity was substantially smaller the sun evolvedon the main sequenceof the Herzprung-
than at the presenttime (youngsunluminosityparadox). Russel diagram. This nearly linear increaseis an un-
The most convincingexplanationproposedsofar assumes escapableconsequence of the progressiveconversion of
the existenceof larger amountsof infrared radiatively ac-hydrogeninto heliumin classicaltheoriesof the stellar
tivesgases(CO2, CH4 ) in the ancientatmosphere. interiors.If the atmosphericcompositionwasunchanged
The vertical thermal structure of the atmosphere was duringthis period,energy-balance models(North et al.,
notably different from the contemporaryone, due to the 1981)and radiative-convective models(Wangand Stone,
lower abundanceof oxygenand ozonebeforethe develop- 1980; G•rard and Franõois,1988) predict that, as a con-
ment of green-plantphotosynthesis.For 02 mixing ratios sequenceof the ice-albedofeedback,the Earth's climate
less10-4 time the presentatmosphericlevel, the presence wouldadopt a stableconditionwherethe planetis totally
of l 1-year and longer periodicitieshas been observedin ice-coveredif the solar constant is decreasedby only a few
precambrian and more recent varve formations. It has percent. However,geologicalevidenceindicatesthat no
been interpreted as an indication of a stronger response world-wideglaciationoccureduntil 2.3 Ga ago, at the
of the ancient atmosphereto solar cycle activity. This time of the Huronianglaciation(Crowley,1983). More-
increasedsensitivity was possibly the consequenceof a over, modelssuggestthat global glaciationsof this type
lower 02 atmosphericmixing ratio and a less developed would be irreversible and luminosities higher than the
stratosphere. presentone would be requiredto unfreezethe system.
Introduction Changesin the continentalmassand geographicdis-
tribution probablytook place and possiblycounteracted
The evolution of the global climate systemduring the in part the lower solar luminosity(Endal and Schatten,
past 4.15billion years was determinedby both endogenic 1982; Schattenand Endal, 1982). However,variationsin
and exogenicforcingswhich are still controlingtoday's the abundanceof infrared-activegaseshas beensuggested
climate. Amongst the external factors, the dominant one as the most likely mechanismto maintain the global tem-
is the solarenergyinput and its temporalvariations. The perature above the freezing point. It is noxvgenerally
solar radiation reachingthe planer's atmosphereis likely believed that carbon dioxide probably played the major
to havevaried in time both in total power(solarluminos- role in the thermostatic control of the Earth's climate.
In the presentatmosphere,ozoneplaysa dominantef-
fect in the determination of the thermal structure of the
Copyright 1989 by
International Union of Geodesy and Geophysics stratosphereand a secondaryrole in the greenhouseheat-
and American Geophysical Union. ing of the mesosphereand the troposphere.The strato-

139
Geophysical Monograph Series Understanding Climate Change Vol. 52

•40 AERONOMY AND PALEOCLIMATE

sphericrisein temperatureis associated with the absorp- tally frozen solutiondependson severalfactorssuchas the
tion of solarradiation shortwardof 300 nm. In the prim- global albedo-temperaturefunction and the tropospheric
itive atmosphere,the only sourceof oxygen•vasthe pho- lapserate (figure1). Saganand mullen(1972)initially
todissociation
of watervapor(Kastinget al., 1980). This suggestedthat large amountsof NH3, an active green-
mechanism wasshownto yield only very small(10-8) of housegas, would be able to elevate the surfacetempera-
the presentatmosphericlevel or PAL) amountsof free ture and compensatefor the past lower solar luminosity.
oxygenand negligiblequantitiesof ozone. Later, as pho- However,due to the short lifetime of ammonium(Kast-
tosynthetic activity began to develop at the ocean sur- ing, 1982) in the primitive atmosphere,this moleculewas
face about 3.5 Ma ago. (Schopf,1983), the oxidationof abandonedand CO2 appearedas the most likely infrared
the reducing seawaterprobably acted as an efficient sink absorber. Hart (1978) calculateda scenariofor the CO2
for atmosphericoxygen and preventedthe accumulation atmosphericlevel which is compatiblewith the absenceof
of 02 in the atmosphere. The simultaneouspresenceof global glaciation in the Precambrian. His work was based
ferrous and ferric iron observed in the banded iron forma- on a simplifiedglobal climatic model coupledto a geo-
tions(BIFs) is usuallyinterpretedasan indicationof the chemicalapproximation.This CO2 historywasfrequently
existence of smallamounts(< 10-6 PAL) of atmospheric used in conjonction with radiative-convectivemodels to
02 duringthe precipitationof the oxidizediron (Holland, test the validity of the thermostatic effect of this con-
1984, Fran(•oisand G•rard, 1986). Photooxidationof Fe stituent (Owen et al., 1979; Kasting et al., 1984; Kuhn
2+ionsmayalsohavebeena significant source of a abiotic and Kasting,1983;Kiehl and Dickinson,1987). The re-
formationof the BIFs (Bratermanet al., 1983;Fran(•ois, sults of Kasting et al. (1984) and Kiehl and Dickinson
1987 ). More recently,oxidationof continentalcrust and (1987)(figure2) showthat the combination of increasing
solarluminosityand decreasingCO2 mixing ratio is able
formation of red rocks contributed to limit the level of
oxygenin the atmosphere.Roughly400 Ma ago,the oxy- to maintain the global temperature of the planet within 4-
gen reachedapproximatelyits presentatmosphericmix- 15K. However,the treatmentby Owenet al. (1979)yields
ingratio(Cloud,1983).Therefore, mostof its an averagesurfacetemperature 3.15Ga ago approximately
throughout
hasvariedcontinuously 7 K warmer than Kiehl and Dickinson who found that the
history,the Earth's atmosphere
in composition and thermal structure. The abundance pressurescaling of the mean band halfwidth of the CO2
of otherminorconstituents
(N20, NO2,CH4,NH3) has bands at 961 and 1064 cm -1 accounts for most of the
probablyalso changedsubstantiallyduring the Earth's temperature difference with Owen et al.'s calculations.
historybut they appearto haveplayeda lesscrucialrole Adopting the samescenario,we usethe one-dimensio-
in the climatic evolutionof the planet. nal radiative-convectiveglobalmodeldescribedby Gdrard
Finally,if evidenceof a controlof solarcycleactivity and Francois(1988) and Francois(1988) to calculatethe
overthe climatehas remainedelusivein the corntempo- evolutionof the mean surfacetemperature. Briefly, this
rary atmosphere,precambriansedimentaryrockshavere- model solvesthe thermodynamicequationusinga for-
vealeda clear signatureof the ancientsolar variations. It ward time-marchingmethod. The absorptionof solar ra-
is therefore important to examine whether the different diation by CO2,H20, O3 and cloudsis calculatedusing
atmospheric
composition whichprevailedin the pastmay the Lacisand Hansen(1974) formulation,includingthe
havefavoreda strongerresponse
of the globalclimateto effects of Rayleigh scattering and surfacealbedo. The
solar forcing. cloud coveris characterizedby a singlelayer with fixed
Carbon Dioxide and Precambrian Climate altitudeand opticaldepth. Rossowet al. (1982)demon-
strated the important potential role of cloud feedbackon
The responseof the surfacetemperatureto variations the stabilizationof the Earth's climate duringits evolu-
in solar luminosity was examined in details with one- tion. In particular, they demonstratedthat the strong
dimensionalenergybalancemodels(North et al., 1981; negative cloud feedback present in their model is able
Endal and Schatten,1982). More recently,Gdrardand to partly compensatethe lower ancient sun luminosity.
Francois(1988) useda radiative-convective modelto test The greenhouseeffect due to infrared absorptionby O3,
the sensitivity of the climate responseto decreasesof N20 and CH4 is treated following the method by
the solarluminosity.In particular,they shoxved that, in manathan(1976) and Donner and Ramanathan(1980).
agreementwith energy-balance models,irreversibleglobal The contribution of water vapor is calculated from the
glaciations•vouldbe predictedif the solarconstantdrop- parameterization
givenby Sasamori(1968). The varia-
ped by a fexvpercent belo•vits presentvalue. The exact tion of the CO2 absorptancewith the CO2 columngiven
position of the discontinuityfrom the unfrozen to the to- by Kasting et al. 1984) is adopted. This expressionwas
Geophysical Monograph Series Understanding Climate Change Vol. 52

G•3RARD 141

6O

20

ß
ß ß

ß ß
ß ß

-2O

-6O

-100 I I I I I I I I I
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
RELATIVESOLARLUMINOSITY(L/L
O)
Fig. 1. Globalmeansurface temperature asa function
of theluminosity
relativeto thepresent one
for differentcases.In thefirstcase,the tropospheric
temperaturegradientis fixedto 6.5 K/km (full
line),whilein thesecond
curve
it isfixedtothemoistadiabatic
value(dashed
line).Forbothcurves,
the temperature dependence
of the surface
albedois adopted
fromGdrardandFrancois (1988)and
correspondsto an icealbedoof 0.5. In the third case(dash-dotted
line),a moistadiabaticgradient
is usedandthe albedo-temperature relationfromWangand Stone(1980)is adopted.The locations
of the jumpsto and from the frozensolutionsare alsoindicated.

obtainedby parameterizingresultsof laboratory absorp-the water vapor feedback, account for the differencesin
tion measurements. Whenever the temperature gradient ZXT8 . Overall, all models lead to the conclusionthat it
is possibleto find a plausible CO2 scenarioable to coun-
tends to exceed the moist adiabatic lapse rate, a con-
vectiveadjustementis appliedto restorethe convective teract the effect of the reduced past solar luminosity and
profile. This convective
adjustement is madefollowing thus to avoid global glaciation. Yet, it is important to
the methodof Manabe and Wetheraid(1967). Our cal- stressthat, at this point, the only detailed geochemical
culatedsurfacetemperatureT• alsoremainwithin5 K of model extendsback in time only 100 Ma ago (Lasaga
the value calculatedfor presentconditions,but the vari- et al., 1985). Theseglobalcarbonate-silicate geochemical
ationsare slightlylargerthan thoseobtainedin previous cyclecalculationspredict a significantincreasein the past
studiesIt is likely that differences
in the radiativecode, CO2 atmosphericcontent and a parallel rise in the global
and in the convectiveadjustmentscheme,amplifiedby surfacetemperature reaching 8 K, 100 Ma ago. Devel-
Geophysical Monograph Series Understanding Climate Change Vol. 52

142 AERONOMY AND PALEOCLIlvlATE

2O j 1 above and the diffusive-chemical model is similar to that

ß Kicht and Dickinson (1987)
a Owen et at.(1979)
15 a Kasting et at.(198Z,)
+ This study
10
0 1 2 3
TIME BEFORE PRESENT(Byr)
Fig. 2. Changein meansurfacetemperaturefrompresent cover characteristics.However, the absolutevalue of Ts is
conditions calculated with various models for the time
period 0 to 3.5 Ma beforepresentusingthe C02 and solar
luminosityscenariofrom Hart (1978).
opment of suchnumericalgeochemicalmodelsextending
further back in time and linking C02, paleoclimateand
tectonic activity is unfortunately limited by the lack of
geochemicalconstrains.
The Climatology of the Oxygen-PoorAtmosphere
In the present atmosphere, ozone plays a dominant
role in the determination of the thermal
stratosphereand a secondaryrole in the mesosphereand
troposphere. This peak in temperature is associatedwith
the absorption of solar radiation shortward of 300 nm.
Presently, the contribution of ozone to trapping of ther-
mal infrared radiation amounts to 2.3 W.m -2 out of a
total of 148 Wm-2 (Dickinsonand Cicerone,1986). It is
believed that, in the ancient atmosphere, containing no
or trace amounts of oxygen, the thermal structure was
significantlydifferent from the contemporaryone. The
climatologyof thesephasesof the Earth evolutionwhen
the 02 level was substantially less than presently is dis-
cussed below.
The model used to investigate this question is a cou-
pled radiative-convectivemodel coupled to a diffusive-
photochemicalcalculation. The first aspectwaspresented
describedby G•rard and Fran•;ois(1987) and Fran<;ois
and G•rard (1988). The continuityand turbulent flux
equationsare solvedfor O, 02, O(1D), 03, H, OH, H02,
H202, N20, NO, NO2, NO3, N205, HN02, HN03, N204
, and CO. Short-livedspecies(H2CO, CH3, CH302, H3
CO, CH300H, HCO) are assumedto be in photochem-
ical equilibrium. The vertical mixing ratio distributions
of C02, CH4 and H2 observedin the presentatmosphere
are adopted.The H20 profileis consistentwith the calcu-
lated verticaltemperaturedistribution. To accountfor the
feedbackbetween the water vapor and ozonedistributions
and the thermal structure, numerical integration is alter-
nated between radiative and photochemicalcalculations
until convergenceis achieved.
The effect of the 02 rise during the Earth's evolution
on the ozonecolumnand on the averagesurfacetempera-
ture is shownin figure3, ignoringthe variationof the solar
luminosity. Our model calculationsyield a surfacetem-
4
peratureTs slightlylargerthan the observedmeanglobal
temperature, due to the particular choiceof the cloud
not essentialhere sincewe are only interestedin the effect
of 02 variations on the mean surfacetemperature changes.
When 02 is decreasedfor 1 to 10-1 PAL, this model pre-
dicts a 3-4 K drop of Ts. This changeis the result of
two combinedeffects. First, as 02 is decreased,the ozone
distribution is reajusted vertically and the ozonecolumn
is slightlyincreased.Consequently, the efficiencyof the
greenhouse is larger and the surfacetemperaturetends
to increase. Simultaneously,the visible albedo decreases
due to changesin the ozone and water vapor concentra-
tion. Another sourceof temperature changesamounting
to about 5 K is due to the total pressuredrop accompa-
nyingthe 02 decrease.This is the resultof the reduction
structure of the of the line broadeningof H20 and C02 infrared bands.
The relative importanceof thesecompetingeffectsis de-
pendenton the assumptionconcerningthe tropospheric
lapserate. In this case,the useof a moistadiabaticlapse
rate stabilizesthe surfacetemperature and causesthe
changeto be smallerthan in other studies(Levineand
Boughnet,1979).
When the 02 level is further decreased,the reduction
of the ozone column and its vertical redistribution causes
the temperature drop. This reduction is particularly ef-
ficient between 10 -3 and 10 -4 PAL. This is due to the
localization of the ozonepeak between 10 and 20 km for
this O2 level. Indeed, sensitivitystudies(Wang et al.,
1980;Fram;ois,1988) haveshownthat the sensitivityof
the climate to ozoneperturbationsmaximizeswhen the
perturbations are located in this altitude range. A T,
Geophysical Monograph Series Understanding Climate Change
lO20 I I I i
1018
1016 282
__//•////, This study• 280
• -'-'--KQsting
.... Levine
et
al. •278-
(1981)
lO1•.
10-•' 10-2
0 2 LEVEL ( PAL)
Fig. 3. Left ß dependenceof the ozone vertical column upon the oxygen level at the surface. Com-
parison is made with previousstudies. Right ß calculated variation of the global surfacetemperature
with the 02 atmosphericlevel.
drop by another 0.9 K is predicted when the 02 level in
the atmosphereis set to a vanishinglysmall value. Con-
sequently,a global temperature decreaseof 8.(3 K is pre-
dicted if the oxygen was removed from the atmosphere,
all other conditionsbeing held constant.
The vertical distribution of the global mean temper-
ature calculation with this model levels of 02 from 10-4
to 1 PAL are shown in figure 4. The progressivedevel-
opment of the stratosphereis observedas the oxygenand
the ozone abundances are increased. At 0.1 PAL, the
stratospherictemperature peak is lower in magnitude and
altitude. This change is due to the decreaseof ozone in
the upper stratosphere. By contrast, a small increase of
the temperature is obtained in the region near 20 km as
a result of the slightly enhancedozone concentrationsin
the lowerstratosphere.At 10-2 PAL of 02, no strat0-
spheric maximum is predicted by the model due to the
low stratosphericozone density. As the 02 level is further
decreasedto 10-a PAL, an inflexionin the temperature
near 20 km is the only effect of vanishinglysmall 03 con-
centrations.Finally, at 10-4 PAL, the role of ozonein the
thermal structure of the atmospherebecomesnegligible.
It is also noted that, in spite of the important changes
in the vertical temperature distribution, the top of the
convectiveregion is located in each casebetween 10 and
14 km.
Vol. 52
GI•RARD 143
I ! I
288
• 286
• 28•,
I I I 1
1 10-•. 10-2
02 LEVEL (PAL)
Responseof the Precambrian Atmosphere
to Solar Cycle Activity
Analysisof periodicitiesobservedby Williams (1981)
and Williams and Sonett (1985) in the annual deposits
(varves)of a precambrianlakein SouthAustraliastrongly
suggesta solar control of the mean annual temperature
by solar activity. These (380 Ma old deposits were found
on the Elatina formation in the Flinders Range, North
of Adelaide. They were formed during the Marinoan
glaciation, one of the series of glacial. episodeswhich
characterizethe end of the Precambrian era. The pale-
ographic setting of the site indicates a marked seasonal,
arid, periglacial climate. These circumstancesimply a
strong seasonalcontrol of the glacier meltwater discharge
into the periglacial lake. Laminae characteristic of dis-
tal clasticvarveswere obtained, avering a total period of
19, 000 va (varve years). Non-randomcyclicitiesin the
thickness of the annual varves were found in the sam-
ple showing characteristic periods of 8 to 1(3va, with a
mean period of 12 years. Recent re-analysis by Sonett
and Williams (1987) indicatethat the actual averagepe-
riod closeto 11.7 va. Longerperiodsof 22 and a 314 years
were also obtained in the Fourier analysis of the time se-
ries. Comparisonwith modern analoguesindicatesthat
the Elatina varve thickness may be interpreted as an in-
Geophysical Monograph Series Understanding Climate Change Vol. 52

•44 AERONOMY AND PALEOCLIMATE

•o
•..•
•'•. I
I !

/ x PAL PAL

\
\
\
l
/
.-.30 /

ß..10 PAl
..
'x

._-L' ' /

10 PAL' -.. \t
' ' 'x,\

o I I
120 lZ.O 160 180 200 220 2/,0 260

TEMPERATURE (K)

Fig. 4. Mean vertical atmospherictemperature distribution calculated with the coupledradiative-

convective photochemical model for different mixing ratios of 02 at the Earth's surface.

dicator of the mean annual temperature. Therefore, the
Elatina series can be consideredas a proxy of the time-
variation of the mean annual temperature.
A weak solar influenceon glacial climate was also dis-
coveredin the analysis of a 236-year varve sequencefrom
Skilak Lake, Alaska (Sonett and Williams, 1985) . The
geographiclocation of the lake, the climatic environment
and seasonalityof meltwaters make it a good analogy
with what appear to have been the conditions prevail-
ing at the time of the Elatina deposition. However, the
Elatina periglacial environment was dryer and probably
free of glacial ice whereas Skilak Lake is in a relatively
humid periglacial climate with permanent glaciers as a
source of meltwater. A correlation study between the
varve sequencethicknessand the sunspotindicesbetween
years 1700 to 1930 reveals a good correlation between
the two series with common periods of 11 and 22 years.
By contrast, analysis of three time seriesof varves from
the EoceneGreen River formation (Crowleyet al., 1986)
showedonly a ;veak 11-yearsignalduring a restricted por-
tion of the 7496-yearsamplecoredepositedapproximately
50 million years ago.
Finally, 2 Ga-old varve recordsfrom the BIFs found
in the Harmersleybasinin Australia, showclearevidence
of a 23.3 yr cyclewhich may be associatedwith the Hale
(22-year)solarcycle.However,WalkerandZahnle(1986)
interpreted this periodicity as reflecting the climatic in-
fluence of the lunar nodal tide, which is weakly present
in modern climate records with a period of 18.6 years.
In this theory, it would arise from the precessionof the
Moon'sorbital plane occuringwith a longerperiod at the
Precambrian as a result of the smaller Moon-Earth dis-
tance in the past.
The clarity of the solar-cyclesignaturein the Elatina
Formation and the remarkablelength of the recordmake
this set of data unique at the present time. Compared
to the generalabsenceof solar signalin publishedvarve
series,it is legitimateto wonderwhetherspecificenviron-
mental conditionsprevailednear the end of the Precam-
brian era which couldexplain the specificityof the Elatina
responseto the solar cycle forcing.
It is striking that the Elatina varveswereformeddur-
ing a period characterizedby a seriesof glacialagesoccur-
ring between0.9 and 0.6 Ma ago: a mostunusualclimatic
Geophysical Monograph Series Understanding Climate Change
event. The distribution of continental masses is difficult
to reconstruct for this remote period. However, paleo-
magnetic studies indicate that the three major phasesof
extensive glaciations occured at low latitudes. Indeed,
most regionsof the Earth, with the possibleexception of
Antarctica, containing Precambrian rocks show evidence
for glaciation during this phase of the Earth's history
(Crowley,1983,Christie-Blick,1982). More specifically,
a ond model, the solar cycle modulation extends to 300 nm.
recent study of paleomagneticmeasurementsof the rema-
nent magnetization of the Elatina Formation was made
by Embletonand Williams (1986). The resultsclearly
demonstrate that the deposition was made at a paleo-
magnetic latitude of 5 degrees,in agreementwith previous
determinationsfrom other paleomagneticAustralian late
Precambrianrocks(Huimin and Wenzhi, 1985). Thus,
the occurenceof ice sheetsand periglacial climate near
sealevel at low latitudes, consideredtogether with the ap-
parent absenceof glaciation during the late Precambrian
is in itself, as a major enigma. The climatic seasonality
responsiblefor the deposition of varves is also difficult to
explain if the region was located at low latitudes at this
period.
The periodicities associated with the Elatina varves
have been interpreted as being of solar origin since the
11-yr and longer periods known to occur in the mod-
ern sun were found in the sediments. Could the differ-
ent atmospheric conditions at the late Precambrian be
responsiblefor the larger climatic responseto the solar
forcing? A possiblemechanismlinking the global climate
and the solar activity cycle was analyzed by G•rard and
Francois(1987). In this scenario,the lowerozonelevelin
the atmosphere680 Ma ago and its different vertical dis-
tribution could possiblyhave made the tropospheremore
responsiveto the 11-yr modulation of the UV solar irra-
diance. Indeed, fluctuations of the temperatures associ-
ated with the 27-day and 11-yr cycles are observedto-
day in the stratosphere.(Keating et al. 1986; Chandra,
1984). It may be expectedthat in an atmospherepoorer
in ozone and with an ozone peak closerto the ground,
the responseof the surfacetemperature would be ampli-
fied compared to the modern atmosphere. This idea was
quantitalively examinedusing the coupledchemical-R-C
model described above. Since the Elatina varves were de-
posited during a periglacialperiod, the model cloud layer
is fixed at low altitude with an optical thicknessr = 10
to produce surfacetemperaturesin the range 268-275 K,
about 15 to 20 K lower than for the present atmosphere.
The responseof the global surfacetemperature to an
imposed variation of the solar UV irradiance associated
ßrith the 11-yr solar cycle is strongly dependenton the
sp•tral distribution of the modulation. Therefore, two
Vol. 52
GERARD
models of UV solar cycle modulation were tested. The
first one (caseA), assumedno variability above200 nm
and a max/min irradianceratio varying linearly from 2.6
at 120 nm to 1 at 200 nm. The mean solar irradiance
distributionreportedby Brasseurand Simon(1981) was
adopted. In this case,the calculated surface temperature
variation remains negligiblysmall ( 0.03 K). In the sec-
In this case(caseB) the surfacetemperatureresponseis
muchlarger and exceeds0.5 K at 3 x 10 -3 PAL of 02.
A significant responseis only obtained in case A if the
variability ratio max/min is substantiallyincreased.
Figure 5 illustrates the dependenceof Ts and the ozone
vertical column on the 02 level for both cases. At low
02 levels, the peak of ozone is located near 10 km, inde-
pendently of the oxygen concentration. The magnitude
of the temperature increasesin parallel to the ozone col-
umn since the "greenhouse"fluctuation is progressively
enhanced.However,for 02 > 10-3 PAL, the ozonemax-
imum and the altitude of the maximum ozone variation
shift toward higher altitudes. This altitude changeof the
solar cycle responsetends to reduce the amplitude of the
surfacetemperature responseuntil, at still higher levels,
the sign of the Ts modulation itself is inverted. This de-
pendenceon 02 is complex and dependson the thermal
structure as well as the altitude of the ultraviolet energy
deposition. Numerical tests also indicate that these re-
suits are nearly unaffectedby an increaseof the CO2 par-
tial pressure. By contrast, it is found that during cold
climatic conditions,correspondingto a dryer troposphere
and to a larger ozone troposphericcontent, the solar ul-
traviolet radiation penetrateslessdeepin the atmosphere.
Both effects contribute to increase the variation of the
ozone column during a solar cycle. The model indicates
that the temperature variation responseincreasesby a
factor of 3.5 when the mean surfacetemperature drops
from 18ø C to 0ø C. This factor may explain, in part, the
specificity of the Elatina periglacial environment to the
solar cycle activity.
Summary
Energy-balanceand radiative-convectivemodels show
that the past reducedsolarluminosity shouldhave gener-
ated global irreversible glaciationsas a result of the ice-
albedofeedback. Geologicand paleontologicevidencein-
dicatesthe absenceof a totally frozen Earth at any stage
of the planer'sevolution. Increasedlevelsof active green-
house gases such as CO2 appear as the most plausible
compensatingfactor to maintain the global climate within
the limited range by palcoclimates. Numerical models
Geophysical Monograph Series Understanding Climate Change Vol. 52

146 AERONOMY AND PALEOCLIMATE

I I

mean altitude cloud layer

and modern climate.
'E
--- low altitude cloud layer u 12
and cold climate.

0.5 _.•.•. ß\
...... low attitude cloud layer.
cold climate and solar % ! \,,.....
ß
"'•
\\
•\
variabilityfor),< 200nm
( enhanced
20-fold
) •10 / //• -
LU
•8 i • :. •
',: • •

ß
o •,
o
LL,I
z
o 2 -
N
O

m -0.5 I ! o I I
11j
3 162 fi• 1 163 •2 16• 1
OXYGEN I_EYEL. (F'•L.) OXYGEN L.EYEL. (F'•L.)

Fig. 15.Variationsin surfacetemperature(left) and verticalozonecolumn(right) as a functionof the

oxygenlevel in the atmospherefrom 10-3 to 1 PAL. The dashedline is obtainedfor a singlecloud
layer located at 7150mbar and a mean surfacetemperature between- 5 and + 2øC. The secondcurve
(dotted line) is for a solarUV variabilityconfinedto wavelengths
below200 nm (seetext).

show that it is possibleto find atmospheric carbon diox- References

ide evolution able to keep the global surfacetemperature

within a few degreesof its contemporaryvalue. The accu-
mulation of photosyntheticoxygenin the atmospherewas
another important sourceof climatic forcing. Its pres-
ence and the formation of the ozone layer significantly
perturbed the thermal structure of the atmosphere and
contributed to rise the surface temperature. The lower
ozone abundance of the Precambrian atmosphere may
have been an important factor to increasethe climate re-
sponse to solar cycle activity observed in the varves of
the Elatina Formation. Numerical simulations
that the difference in solar irradiance between
imum and minimum conditionscould have generatedsur-
face temperature variations substantially larger than in
the present atmosphere.
Acknowledgment,.I am especiallygrateful to my col-
league Louis Fran(;oisfor valuable discussionsand critical
reading of the manuscript. This researchwas supported
by the Belgian National Foundationof ScientificResearch
(FNRS).
Brasseur G. and P.C. Simon, Stratospheric chemical and
thermal responseto long-term variability in solar UV ir-
radiance, J. Geoph. Res., 86, 7343-7362, 1981.
Braterman P.S., A.G. Cairns-Smith and R.W. Sloper, Photo-
oxidationof hydratedFe2+ - significance
for bandediron
formations, Nature, 303, 163-164,1963.
Chandra S., An assessment of possibleozone-solarcyclerela-
tionship inferred from NIMBUS 4 BUV Data, J. Geophys.
Res., 89, 1373-1379 ,1984.
indicate Christie-Blick N., Pre-Pleistoceneglaciation on earth: Im-
solar max- plications for climatic history of Mars, Icarus, 50 , 423-
443, 1982.
Cloud P., Aspectsof Proterozoicbiogeology,Geol. Soc. Am.
Mere., 161, 245-251, 1983.
Crowley K.D., C.E. Duchon and J. Rhi, Climate record in
varved sedimentsof the Eocene Green River Formation,
J. Geophys.Res., 91, 8637-8647, 1986.
Crowley T.J., The geologicrecord of climate change,Rev.
Geophys.and SpacePhys., œ1,828-877, 1983.
DickinsonR.E. and R.J. Cicerone,Future global warming
atmospherictrace gases,Nature, 319, 109-115, 1986.
Geophysical Monograph Series Understanding Climate Change
Donner L. and V. Ramanathan, Methane and nitrous oxide:
their effects on the terrestrial climate, J. Atmos. Sci.,
37,119-124, 1980.
Embleton B.J.J. and Williams G.E., Low palaeolatitude of
depositionfor late precambrianperiglacialvarvites in South
Australia: implications for palaeoclimatology,Earth Planet.
Sci. Letters, 79,419-430, 1986.
Endal A.D. and K.H. Schatten, The faint young sun-climate
paradox: continental influences. J. Geophys. Res., 87,
7295-7302, 1982.
Fran(;oisL.M., Reducingpowerof ferrousiron in the Archean
Ocean, 1. Role of FeOH+ photooxidation,Paleoceanogra-
phy, 2 , 395-408, 1987.
Fran(;oisL.M., The effectof cloudparametrization and tem-
perature profile on the climate sensitivity to ozonepertur-
bations, J. A tmos. Res., 21, 305-321, 1988.
Fran(;oisL.M. and J.C. Gdrard, Reducingpower of ferrous
iron in the Archean Ocean, 1. Contribution of photosyn-
thetic oxygen, Paleoceanography, 1, 355-368, 1986.
Fran(;ois L.M. and J.C. Gdrard, Ozone, climate and bio-
spheric environment in the ancient oxygen-poor atmo-
sphere, Planet. SpaceSci., in press,1988.
Gdrard J.C. and Fran(;oisL.M., A model of solar-cycleef-
fects on palaeoclimate and its implictions for the Elatina
Formation, Nature, $2t7, 577-580, 1987.
Gdrard J.C. and L.M. Francois, A sensitivity study of the
effects of solar luminosity changeson the Earth's global
temperature, Ann. Geophys.,t7, 101-112, 1988.
Hart M.H., The evolution of the atmosphere of the Earth,
Icarus, 33, 23-39, 1978.
Holland H.D., The chemicalevolutionof the atmosphereand
oceans Princeton University Press, Princeton, 1984.
Huimin Z. and Z. X'VenzhiZ., Paleomagnetic data, late pre-
cambrian magnetostratigraphy and tectonic evolution of
eastern China, Precambrian Res., 29, 65-75, 1985.
Kasting J.F., Stability of ammonia in the primitive terres-
trial atmosphere, J. Geophys. Res., 87,3091-3098,1982.
Kasting J.F., Theoretical constraintson oxygen and carbon
dioxide concentrations in the Precambrian atmosphere,
Precambrian Res., $•, 205-229, 1987.
Kasting J.F. and T.M. Donahue, The evolution of atmo-
spheric ozone, J. Geophys. Res., 85, 3255-3263, 1980.
Kasting J.F., J.B. Pollack, T.P. Ackerman, Responseof the
Earth's atmosphere to increasesin solar flux and impli-
cations for loss of water from Venus, Icarus, 57, 335-355,
1984.
Keating G.M., M.C. Pitts, G. Brasseurand A. De Rudder,
Responseof middle atmosphereto short-term solar ultra-
violet variations, 1, observations,J. Geophys. Res., 92,
889-902, 1986.
Kiehl J.T. and R.E. Dickinson, A study of the radiative
Vol. 52
Gi•RARD •47
effects of enhanced atmospheric CO2 and CH4 on early
earth surface temperatures, J. Geophys. Res., 92, 2991-
2998, 1987.
Kuhn W.R. and J.F. Kasting, Effectsof increasedCO2 con-
centrations on surface temperature of the early earth.,
Nature, $01, 53-55, 1983.
Lacis A.A., J.E. Hansen, A parameterization for the ab-
sorptionof solarradiation in the Earth's atmosphere,J.
Atmos. Sci., $1, 118-133. 1974.
Lasaga A.C., R.A. Berner and R.M. Garrels R.M., An im-
proved geochemicalmodel of atmospheric CO2 fluctua-
tions over the past 100 million years, in The carbon cy-
cle and atmospheric C02' natural variations Arcbean to
present, E.T. Sundquist and ]V.S. Broeckereds., pp. 397-
411, GeophysicalMonograph 32, American Geophysical
Union, •Vashington, 1985.
Levine J.S. and R.E. Boughner, The effect of paleoatmo-
spheric ozoneon surfacetemperature, Icarus, $9, 310-314,
1979.
Levine J.S., T.R. Augustsson,R.E. Boughner, hi. Natara-
jan and L.J. Sacks, Comets and the photochemistry of
the paleoatmosphere, in Comets and the origin of life.,
edited by C. Ponnamperuma, pp. 161-190, D. Reidel,
Dordrecht, 1981.
Manabe S. and R.T. Wetheraid R.T., Thermal equilibrium
of the atmosphere xvith a given distribution of relative
humidity, J. Atmos. Sci., 2•, 241-251 ,1967.
North G.R., R.F. Cahalan and J.A. Coakley Jr., Energy
balance climate models, Rev. Geophys. Space Phys., 19,
91-121, 1981.
Owen T., R.D. Cess and V. Ramanathan, Enhanced CO2
greenhouseto compensatefor reduced solar luminosity
on early earth, Nature, 277, 640-642, 1979.
Ramanathan V., Radiative Transfert within the earth's tro-
posphereand Stratosphere' a simplified radiative-convective
model, J. Atmos. Sci., $$, 1330-1346, 1976.
Rossow \V.B., Henderson-Sellers A. and Weinreich S.K.,
Cloud feedbackß a stabilizing effect for the early Earth?,
Science, 217, 1245-1247, 1982.
Sagan C. and G. lXlullen, Earth and •lars' Evolution of
atmospheresand surface temperatures, Science. 177, 52-
56, 1972.
Sasamori T., The radiative cooling calculation for applica-
tion to general circulation experiments, J. Appl. Meteo-
rol., 7, 721-729, 1968.
Schatten K.H. and A.S. Endal, The faint youngsun - climate
paradox' volcanic influences, Geophys. Res. Left., 9,
1309-1311, 1982.
Schopf J.W., Earth's earliest biosphere:its origin and evo-
lution, Princeton University Press, Princeton, N.J., 1983.
Soneft C.P. and G.E. Williams, Solar aflqnityof sedimentary
Geophysical Monograph Series Understanding Climate Change Vol. 52

a48 AERONOMY AND PALEOCLIMATE

cyclesin the late precambrian elatina formation, Aust. J.
Phys. 38, 1027-1043, 1985.
Soneft C.P. and G.E. Williams, Frequency modulation and
stochasticvariability of the Elatina varve record: a proxy
for solar cyclicity?, Solar Phys., 110,397-410, 1987.
Walker J.C.G. and K.J. Zahnle, Lunar nodal tide and dis-
tance to the Moon during the precambrian,Nature, 320,
600-602, 1986.
'•VangW.C. and P.H. Stone,Effect of ice-albedofeedbackon
globalsensitivityin a one-dimensionalradiative-convective Australia, Nature, 318, 523-527, 1985.
climate model, J. Atmos. Sci., 37, 545-552, 1980.
Wang W.C., J.P. Pinto and Y. L. Yung, Climatic effectsdue
to halogenatedcomponentsin the Earth's atmosphere,J.
Atmos. Sci., 37, 333-338, 1980.
Williams G.E., Sunspot periods in the late precambrian
glacial climate and solar-planetaryrelations, Nature, 291,
624-628, 1981.
Williams G.E. and C.P. Soneft, Solar signaturein sedimen-
tary cyclesfrom the late precambrian Elatina Formation,
Geophysical Monograph Series Understanding Climate Change Vol. 52

STUDIES OF CRETACEOUS CLIMATE

Eric J. Barron

Earth SystemScienceCenter, Penn StateUniversity,

UniversityPark, Pennsylvania

Abstract. The warm Cretaceous climate is Introduction

characterized by (1) warm oceanic deep water,
>12øC, (2) tropical surface ocean temperatures Paleoclimatology has been revolutionized by
similar or slightly warmer than today's values, the theory of plate tectonics and the development
(3) warm poles as evidenced by abundant high of methods to reconstruct global geography.
latitude floras and faunas and no direct evidence First, paleogeography provides the framework for
for permanent ice, (4) continental interiors with interpreting environmental data. Without this
a smaller degree of seasonality, (5) a globally framework climatic reconstruction on long time
averaged surface temperature 6-12øC higher than scales is untenable. Second, plate tectonics and
the present day value, and (6) unusually abundant a constantly changing geography provide a
coals, evaporites, bauxites, black shales and spectrum of reasonably well-defined causes for
other rock types. climatic change. A series of papers ascribing
A significant role by external forcing factors climate change to changes in geography, in
is required to explain the Cretaceous warmth. particular explaining the contrasts between major
The leading candidates for explaining warm global glacial episodes and warm, apparently ice-free
temperatures are the very different Cretaceous climates (Crowell and Frakes, 1970; Luyendyk et
continental geometry, the decrease in continental al., 1972; Frakes and Kemp, 1972, 1973; Hays and
area due to high eustatic sea level, higher Pitman, 1973; Berggren and Hollister, 1974;
atmospheric CO concentrations in the Cretaceous Kennett, 1977; Donn and Shaw, 1977; Tarling,
atmosphere,
an• amplificationof thesefactors 1978; Beaty, 1978) appeared in concert with the
through a greater role of the ocean in poleward first publications of past continental
heat transport. reconstructions and major efforts to determine
A variety of climate models, including Energy the evolution of the major ocean basins.
Balance models, mean annual atmospheric General Each of these contributions focussed on one of
Circulation Models (GCM's), annually varying a few simple, but compelling arguments on how
atmospheric GCMs with simple coupled oceans and geography might alter the global climate: (1)
ocean GCMs have been utilized to study Cretaceous land provides a surface for the accumulation of
climate. Among many different insights from high albedo snow, and consequently greater land
model simulations, several are notable: (1) area at the pole promotes glaciation, (2) the
multiple forcing factors including a higher distribution of land can greatly influence, by
atmospheric CO_ level are required to achieve barriers or gateways, the poleward heat transport
Cretaceous warmth if the spectrum of simulations by the oceans or (3) because of the different
are reasonable estimators of climate sensitivity, thermal and albedo properties of land and sea,
(2) a dramatic change in the role of the ocean land-sea distribution and total land area will
would create additional problems rather than modulate climate.
solutions to the question of Cretaceous warmth, Paleoclimatology is also experiencing great
(3) some permanent ice in Antarctica may be advances through the development and application
easier to justify than the assumption that the of comprehensive models of the climate system. A
globe was entirely ice free and (4) in focusing hierarchy of climate models have been applied to
attention solely on temperature considerable test specifically whether paleogeography explains
climate variation has been ignored which will the major changes in climate found in the
provide important constraints on understanding geologic record. The majority of these
the record of Cretaceous climate. applications have considered the Cretaceous time
period (~65 to 140 million years ago) because of
Copyright 1989 by the record of warmth and large differences in
International Union of Geodesy and Geophysics global geography (e.g. Barron, 1983). The first
and American Geophysical Union. efforts to explore geography as an explanation of

149
Geophysical Monograph Series Understanding Climate Change Vol. 52

150 STUDIES OF CRETACEOUS CLIMATE

MERCATOR AT 1OO MILLION YEARS

60 ,

3,0
0

30

60

Fig. ]. A reconstruction of mid-Cretaceous geography extending to 70øN and S (after

Barron et al., 1981b). Present continental outlines and the 2000m isobath are included
with mid-Cretaceous continental area shaded. Thirty degree paleolatitude lines are
indicated.

Cretaceous warmth (Barron et al., 1981a; Barron Climate Models and Cretaceous Temperatures:
and Washington, 1984) demonstrated that a The Major Discrepancies
hierarchy of models fail to achieve Cretaceous
warmth given paleogeography as the primary The initial efforts to determine the role of
forcing factor. Cretaceous geography (Figure 1) in explaining
However, these contributions did not refute Cretaceous temperatures (Figure 2) focussed on
hypotheses that geography explained Cretaceous experiments with an Energy Balance Model
warmth. Rather, they left as an open question (Barron et al., 1981a) and with a mean annual
whether the discrepancy between the model General Circulation Model (Barron and
experiments and climatic reconstructions could Washington, 1984).
be explained because of (1) model limitations, Barron et al. (1981a) utilized a model based
(2) the need for climatic forcing factors in on a zonally averaged energy balance of the
addition to geography and/or (3) the need to vertically integrated earth-atmosphere system
reinterpret Cretaceous paleoclimatic data. computed for the annual cycle. The energy
This contribution summarizes a series of transported poleward was approximated as a
subsequent efforts to demonstrate or eliminate diffusion process. The zonally averaged land
one or more of these points as an explanation distribution was derived from Figure 1. Two
for the discrepancy between model results and conclusions were apparent. First, without
climatic data. At the end of this effort the major prescribed changes in cloud cover or some
same three questions are pertinent, however the additional net global heating, the Cretaceous
evidence is clearer: (t) geography is a planetary temperature could not be achieved (a
substantial climatic forcing factor but (2) 6-12øC increase in comparison with the present
geography is unlikely to be the sole day). The EBM achieved a 1.6øK warming as a
explanation for Cretaceous climate. function of geography alone. Second, in each
Geophysical Monograph Series Understanding Climate Change Vol. 52

MID-CRETACEOUS TEMPERATURES Cloud-climate feedback must also be considered

310 I i i i I I as a potential limitation of any current
climate model experiment. The efforts to
resolve the role of model limitations in

290 explaining the discrepancy between model

results and Cretaceous temperature data were
oximum
based on an investigation of the role of oceans
and the seasonal cycle in CCM simulations.
270 Minimum
Presenl The Role of the Oceans in
Explaining Cretaceous Temperatures
250
A greater role of •he oceans in poleward heat
transport as a mechanism for explaining
230 I Cretaceous polar warmth receives considerable
90 60 30 0 30 60 90
qualitative support. First, in order to
LATITUDE explain warmer poles with a reduced temperature
gradient a greater portion of the heat
Fig. 2. Estimates of mid-Cretaceous mean annual
transport must have been accomplished by the
temperatures (o Kelvin) (maximum and minimum
oceans, since with such conditions the role of
interpretations) based on data described in
the atmosphere must be reduced. Second, the
Barron (1981a), and given in reference to present
presence or absence of land or land barriers,
day values.
particularly at high latitudes should have a
marked effect on oceanic poleward heat
transport and therefore presumably on the total
sensitivity experiment, the greatest response poleward heat transport of the ocean-atmosphere
in temperature occurred in high latitudes, but
a similar sign (less magnitude) change occurred
as well in the tropics. Evidently to achieve SURFACE TEMPERATURES'
sufficiently warm poles the model would PRESENT DAY AND CRETACEOUS
generate tropical temperatures which would 320
exceed observations and some life limits. _

Barron and Washington (1984) completed a 310

similar experiment utilizing a version of the
Community Climate Model (CCM), a spectral
3OO
General Circulation Model of the atmosphere at
the National Center for Atmospheric Research.
The associated global grid has 40 latitudes and
48 longitudes and the model uses nine levels in
the vertical.
and Washington
The model version
(1984) utilized
used by Barron
mean annual
280
insolation
surface
and an energy balance
temperatures
ocean.
were computed based on the
Sea
270
surface energy balance and the ocean model 26O
lacks heat transport or heat storage. The
Cretaceous geography experiment resulted in a 250
globally averaged surface temperature warming
of 4.8øK (Figure 3). The same two conclusions 240
derived from the EBM studies above applied in
this case, although the CCM had a greater 230 / , i I • i I I I I I I I I I I I I
sensitivity to geography than did the EBM 90 60 30 0 - 30 - 60 - 90
experiment.
LATITUDE
What explains the discrepancy between the
climate models and the Cretaceous temperature Fig. 3. Zonally-averaged surface temperatures
data? From the perspective of climate models, (o Kelvin) derived from mean-annual simulations
the limitations of the above experiments are using the NCAR Community Climate Model (CCM), a
clear, lack of a realistic ocean in both cases, spectral General Circulation Model of the
a seasonal cycle in the EBM but highly Atmosphere. Present-day simulated surface
parameterized poleward heat transport and more temperatures are compared with values generated
realistic poleward heat transport but a lack of from a simulation with mid-Cretaceous geography
a seasonal cycle in the CCM experiment. (after Barron and Washington, 1984).
Geophysical Monograph Series Understanding Climate Change Vol. 52

152 STUDIES OF CRETACEOUS CLIMATE

system. In fact, the role of ocean heat of a simple thermodynamic equation for heat
transport in climatic change is rather poorly storage in an ocean layer 50 meters thick (see
known (see review by Covey and Barron, 1988). Washington and Meehl, 1984 for additional
However, for the specific case of the Cretaceous discussion). The heat flux into the ocean is
a series of sensitivity experiments with the CCM given by the surface energy balance.
were completed by Schneider et al. (1985) to The results of this experiment are in contrast
evaluate whether a greater role by the ocean with the results for the mean annual simulation
could explain the model-observation discrepancies with Cretaceous geography described by Barron and
cited above. Washington (1984). First, the globally averaged
The Schneider et al. (1985) experiments did surface temperature warming due to the change in
not utilize a dynamic ocean circulation model geography and removal of permanent ice caps is
coupled to the CCM. Instead, extreme near 3.0øC in the annual cycle experiment
assumptions on the efficiency of ocean heat compared to 4.8øC in the mean annual experiment.
transport and how these assumptions could alter Interestingly, using the same two models, the
sea surface temperature were used to specify sensitivity to a doubling in atmospheric CO
sea surface temperatures in CCM experiments. (Washington
andMeehl,1984)wasgreaterin2the
The experiments were not designed to be annual cycle experiment (3.5øC) than in the mean
realistic, but rather to evaluate the impact of annual experiment (1.3øC). The greater
a greater role by the ocean and reduced sensitivity of the model to higher CO• levels
equator-to-pole temperature gradients on the evidently occurred because of the effect of
climate. In particular, a greater role by the incorporating the mixed layer on the latitude of
oceans and warmer polar oceans (even if set at the sea ice margin.
20øC) did not prove sufficient to maintain high The primary question to be addressed is why
latitude continental regions above freezing in the annual cycle experiment with Cretaceous
winter. This result is in apparent contrast to geography •llustrated less sensitivity than the
data of extreme warmth and lack of winter mean annual model. The seasonal cycle
freezing at high latitudes. In fact, experiment resulted in a greater discrepancy
experiments with an assumed or implied very between model results and observations. A
efficient ocean heat transport resulted in series of sensitivity experiments demonstrated
substantially colder continental interiors in which aspect of Cretaceous geography resulted
winter. This result occurred because of a in the warming in the mean annual experiments
decreased vigor of the atmospheric circulation, (Barron and Washington, 1984). Reduced
which in turn provides insufficient northern hemisphere land area at high latitudes
ocean-to-land atmospheric heat transport to in comparison with the present day explained
mitigate mid-winter continental radiative the majority of the Cretaceous northern
cooling. hemisphere warming illustrated in Figure 3.
If the above results are valid, a greater role Given mean annual insolation and an equilibrium
by the oceans is unlikely to solve the problems simulation for an ocean without heat capacity,
presented by comparing Cretaceous climate model the differences between land and ocean surface
experiments w•th Cretaceous temperature data. characteristics dominate the model-derived
northern hemisphere warming. Surface
The Role of the Seasonal Cycle in temperature isotherms are displaced poleward
Explaining Cretaceous Temperatures over land and ocean due to a warming related to
the decrease in total land area.
The role of the annual cycle of insolation was The seasonal simulation included ocean heat
not addressed in the first general circulation storage. In this case the difference in land and
model experiments •o investigate geography as an ocean surface characteristics is much less
explanation of Cretaceous wa•nth. Crowley et al. significant than is the thermal inertia of the
(1986) emphasize that, in particular, summer system. Summer heat storage and winter heat
temperatures may be a critical parameter in loss result in northern hemisphere ocean mixed
initiating glaciation. Depressed summer layer temperatures which are not substantially
temperatures on high latitude continents may different from the present day control
allow winter snow to remain throughout the year. experiment. The increased area of ocean
Of further importance, few geologic indicators resulted in a zonally averaged temperature at
respond to mean annual conditions. Seasonal mid to higher latitudes in the northern
i.nfo•at•on is essential to compare model results hemisphere which is cooler in summer and warmer
with observations and to ascertain the real in winter. This model response is directly
nature of any discrepancy. comparable to the change in land area (Figure
A full annual cycle simulation was performed 4) or the degree of continental flooding.
with mid-Cretaceous geography using the version In fact, the majority of the globally-averaged
of the CCM described by Barron and Washington surface temperature increase in the annual cycle
(1984) and discussed above, coupled with a experiment is in response to the removal of the
mixed layer ocean. The ocean model makes use Antarctic ice sheet. If this aspect is ignored,
Geophysical Monograph Series Understanding Climate Change Vol. 52

BARRON 1 53

% CHANGE IN LAND FRACTION warmth-are not restricted to the ocean and the
seasonal cycle. The discussions on the lack of
PRESENT DAY TO CRETACEOUS
agreement
betweenmodelsensitivities to a CO
2
I I I I ' 0 doubling described by Washington and Meehl
(1986) and Schlesinger and Mitchell (1987)
indicate a number of additional factors. The

- -10
two most important factors are (1) cloud
formulations and cloud-climate feedbacks and
(2) snow and sea ice parameterizations and
ice-albedo feedbacks of the control õimulation.
- -20
If the comparison of GCM studies for a CO_
doublingare indicative, the GCM
utilized2by
Barron and Washington (1984) can be
- -30 characterized by a comparatively low
sensitivity. In addition, this model has
little negative lapse rate feedback at high
- -40 latitudes in comparison with Manabe and
Wetheraid (1975). The model by Manabe and
Wetheraid (1975) had a stability-dependent
10-
- -50 vertical diffusion as do more recent versions of
• DJF the Community Climate Model. This high latitude
factor may be very significant for the degree of
polar amplification of warming. As pointed out
by Dickinson (1985) these factors, and the
ß ' JJA differences between climate models, suggest that
:• -10- potential model limitations cannot be eliminated
!
as a solution to the discrepanies between model
90 6'0 3'0 0 experiments and Cretaceous observations. Still
N
LATITUDE the results of the annual cycle simulation and
the lack of global temperature sensitivity in
ZONALLY AVERAGED comparison
with the samehierarchyof CO
2
experiments, indicates that other forcing
SURFACE TEMPERATURE DIFFERENCE factors, in addition to paleogeography, must also
CRETACEOUS - PRESENT DAY be considered as explanations of Cretaceous
warmth.
Fig. 4. Zonally-averaged surface temperature
(o Kelvin) differences (Cretaceous minus the
Are Other Climatic Forcing Factors Important
present day) for annual cycle simulations using
the NCAR CCM coupled with a mixed layer ocean.
With the advent of plate tectonic theory,
DJF - December, January, February average. JJA -
other climatic forcing factors which might
June, July, August average. The surface
explain the contrast between warm episodes and
temperature differences are compared with the
glacial time periods received little attention.
percentage change in land fraction for each 10 ø Atmospheric CO_ variations are the one
latitude belt from the Present day to the
significant exception. Variations in atmospheric
Cretaceous.
carbon dioxide levels were suggested by
Chamberlin (1899), Budyko and Ronov (1979) and
Fischer (1982), but received little credible
the change in geography resulted in a large quantitative or observational support. Then, the
change in the amplitude of the seasonal cycle, geochemical models of Berner et al. (1983) and
but little change in mean, global surface Lasaga et al. (1985), which are based on the
temperature. Evidently, the addition of the carbonate-silicate geochemical cycle, provided
annual cycle of insolation does not solve the much stronger support for the prospect of large
discrepancies between model simulations and variations in atmospheric CO_ levels.
Cretaceous observations. Further, this study Specifically thesemodelsli•ked plate tectonics
continues to develop the theme noted by Rind (rapid sea floor spreading), continental area
(1986), that model sensitivity to geography may (flooding of continents is equated with rapid sea
be quite different frommodelsensitivity to CO
2 floor spreading), higher rates of volcanic
variations. degassing (with rapid sea floor spreading),
geochemical weathering rates and fluxes, CO_
Other Model Factors levels andclimate. Thegeochemical
models
2
predicted a several-fold increase in atmospheric
Potential model limitations which have a
CO
2 during the Cretaceousdependingon a set of
bearinq on the problem of simulating Cretaceous assumptions.
Geophysical Monograph Series Understanding Climate Change Vol. 52

STUDIES OF CRETACEOUS CLIMATE

Much effort is now focussed on improving or paleoclimatic estimates. In fact, the

examining limitations in the geochemical models temperatures in the model simulations are
(e.g. Berner and Barron, 1984; Kasting, 1984; approaching some limits of tropical organisms.
Volk, 1987). The support for higher atmospheric
Evenif CO2 answersthe questionof qlobal
CO
2 levels is not limited to modelevidence,but warmth,the results return us to potential model
may also be supported by observations of the limitations (nature of poleward heat transport
sedimentary record (e.g. Sandberg, 1985) and and the role of the oceans, cloud-climate
independent evidence for higher rates of feedbacks, and the nature of the lapse rate
volcanism (Arthur et al., 1985). Consequently, feedback at higher latitudes).
higher CO
2 levels are a •ogical alternative to Finally, other forcing factors may also have
geography, or an additional climatic forcing contributed to the nature of the Cretaceous
factor. warmth.
However, the large number of assumptions and
limitations in the geochemical cycle models Is the Cretaceous Record Correctly Interpreted?
prevent the CO
s "forcing factor" from being
specified in cIimate models in the same fashion The differences between the Cretaceous model
as the geography. Consequently, the nature of experiments and observations related to
the problem changes: what level of atmospheric temperature could also be explained if the
CO_ would be required to explain Cretaceous interpretations of the Cretaceous record are
2
warmth given Cretaceous paleogeography, and is incorrect in two respects: (1) the amplitude
this level consistent with other sources of of annual temperature in continental interiors
information. and (2) the nature of polar climate, particularly
The results of Barron and Washington (1985) in winter.
indicate that a four-fold increase in atmospheric More specifically, the proposed increased role
carbon dioxide in the mean annual version of the of the oceans in poleward heat transport was
CCM produces a globally averaged surface rejected based on the model experiments described
temperature sufficient to reach the lower limit above. A greater role by the oceans in heat
of Cretaceous temperature observations. transport resulted in a larger amplitude in the
Recognizing that this model is at the lower end annual cycle of temperature in continental
of the GCM model spectrum in sensitivity to a interiors at high latitudes. In addition, the
•02-doubling,moderateandreasonableCO
2 level constraint of continental equability and
].ncreases might well solve the discrepancy essentially frostless or very warm winters forces
between model results and Cretaceous observations estimates of Cretaceous warmth to the upper
described earlier. The results (Figure 5) are limits illustrated in Figure 2.
not without problems, however. The degree of The problems would be distinctly different if
polar amplification of the warming seems the Cretaceous record was strongly biased by
insufficient, and tropical temperatures exceed maritime coastal climates, especially in the case
of warmer polar oceans. In fact, the vast
majority of data come from coastal regions, which
COMPARISON OF CRETACEOUS TEMPERATURES: is evident when the data are plotted on
MODEL AND LIMITS paleogeographic maps. There exists one important

i I i I I / exception.
Alligator-rela%ed
reptiles,
lake
310

- Data "• 1 records

and
other
indicators
from
interior
Asia
.. ........

(Mongolia) support the interpretation

warmth of
• 300 and equability in Cretaceous continental
interiors in the mid-latitudes. Such sites are
rr 290 rare, and much of the real weight of evidence for
the amplitude of the seasonal cycle rests on this
rr 280 sparse data set.
The nature of polar climates is even more
critical. Some of the better paleoclimatic
m 2?0 _-- ---Cretoceous
Geogrophy • evidence from fossil floras (e.g. Spicer and
_ ß......
I
Cretoceous
I
Geogrophy_•nd
I i
4xCO
i
2 •t Parrish, 1986) indicates a north polar (~85 N)
260 mean annual temperature near 10øC. Cool or
90 60 30 O 30 60 90 cold winters are plausible. The interpretations
for Antarctica are much less certain. All the
N LATITUDE S
evidence from Antarctica is associated with the
Fig. 5. A comparison of mid-Cretaceous continental margin or based on the lack of a
temperature (o Kelvin) limits given in Figure 2 record of Antarctic glaciation. The evidence for
with mid-Cretaceous model results described in
a lack of permanent ice is indirect. If
Figure 3 and a second simulation with Cretaceous Pleistocene arguments for the effect of polar ice
geography and specified 4x present levels of on the oxygen isotopic record of the oceans (e.g.
atmospheric
CO2. Emiliani and Shackleton, 1974) are applied to the
Geophysical Monograph Series Understanding Climate Change Vol. 52

BARRON

Cretaceous isotopic record, the occurrence of The model studies summarized in this
substantial ice is unlikely or can be rejected. contribution suggest that Cretaceous geography is
If substantial ice enriched in 0-16 were present, an insufficient forcing factor because (a) the
the ice would be detectable in the oceanic oxygen magnitude of the global warming is insufficient
isotopic record. Warmth at the Antarctic margin and (b) the degree of polar amplification is
and the oxygen isotopic record support a case for insufficient. At least two of the major
the lack of permanent ice on Antarctica. limitations in the model studies (the role of the
A number of cases can be presented which favor oceans and of the annual cycle) were addressed
some permanent ice or at least seasonally without abrogating these deficiencies. However,
subfreezing continental temperatures, either of the strength of the negative response to the
which would substantially reduce the nature of first question is still constrained by model
the problems described here. limitations. Three limitations have been
First, waxing and waning of ice sheets identified: (1) surface-albedo feedbacks, (2)
produces fast and large variations in sea level. cloud-climate feedbacks, and (3) polar lapse rate
We might expect that the character of the record feedbacks. In addition, the nature of the
of sea level variations for glacial and non- sensitivity experiments which attempted to
glacial climates would be different. However, examine the role of the oceans is far too crude
differences are not noted in reconstructions of to eliminate completely a different role by the
sea level by Haq et al., 1987 or Vail et al., oceans as a major causitive factor in explaining
1977) throughout the last 100 million years. Cretaceous climate.

Second, the Antarctic warmth is restricted to Cretaceous geography might also be the major
data from the continental margin. Numerous areas explanation of Cretaceous climate if the nature
(e.g. New Zealand) can be cited as examples of of the Cretaceous warmth is mis-interpreted. Two
coastal, maritime warmth adjacent to interior key climatic characteristics have been identified
glaciers. Certainly, the condition of coastal in this regard: (1) the amplitude of the
warmth on a large polar continent is not seasonal cycle in continental interiors and (2)
necessarily indicative of interior conditions. the nature of polar climates especially in
Third, the oxygen isotopic effect could be winter. In both cases, the data are either
arguable if ice caps were small or the product of sparse or inferential. The Cretaceous was
evaporation from adjacent warm oceans, with a substantially warmer than the present climate,
short path length through the atmosphere (little but the condition of cold continental interiors
distillation and fractionation of isotopes) and during winter at higher latitudes and permanent
if Antarctica was characterized by little ice in the interior of Antarctica would

topographic expression (see Covey and Haagenson, substantially reduce the key differences between
1984, for a discussion of variables which model experiments and observations. The model
influence oxygen isotopic composition of snow and studies in this case have directed attention at
glacial ice). Certainly, the oxygen isotope the two weakest points in the Cretaceous climatic
record does not eliminate isolated glaciers, and reconstructions.

might not eliminate small polar ice caps. If the stated model limitations and the
Even the condition of sub-freezing winter weaknesses in the climatic record are not the
temperatures and summer warmth (a mean near zero) answer to the problems associated with simulating
would substantially reduce the need for polar sufficient global warmth and the degree of polar
amplification of Cretaceous warming in model amplification of temperature then the alternative
experiments. However, all the evidence described is additional or alternative climatic forcing
above is circumstantial and speaks only to factors. The series of sensitivity experiments
possibilities. A substantial increase in indicate that alternatives must be considered and
knowledge from continental interiors and may be likely. The most plausible additional
specifically from Antarctica is required to factor is the possibility of higher COplevels.
determine if the nature of Cretaceous warmth has The level of CO_ required to simulate •ufficient
been correctly interpreted. globalwarmth(•-10x presentday) is within the
range suggested by geochemical models as
reasonable or likely. Even in this case,
Geography as a Forcing Factor however, the polar amplification of warmth may be
for Climatic Change problematic and the degree of tropical warming
becomes a critical factor.
The role of geography in climatic change The weight of the arguments support a
should be divided into two questions: (1) Is conclusion, with several caveats, that Cretaceous
Cretaceous geography the explanation for geography is insufficient to explain fully the
Cretaceous warmth and (2) is geography an observations of Cretaceous warmth. However,
important factor in climatic change? The answer geography remains a substantial climatic forcing
to the first question is probably negative and factor. A hierarchy of models yield an increase
the answer to the second question is almost in globally averaged surface temperature of
certainly positive. several degrees. The amplitude of the annual
Geophysical Monograph Series Understanding Climate Change Vol. 52

156 STUDIES OF CRETACEOUS CLIMATE

TEMPERATURE VERSUS PRECIPITATION, Acknowledqments. The author gratefully

acknowledges the efforts of the convenors and
WORLD WIDE editors, A. Berger, R. Rickinson and J. Kidson.
i.u
This research was partially supported by Grant
D
!
ATM-8715499 from the National Science Foundation.

< 298 CRET 4xCO2

ReFerences

Arthur, M. A., W. E. Dean, and S. O. Schlanger,

u• • 294 ß
I- o PRESENT CRET Variations in the global carbon cycle during
4x•1•2 the Cretaceous related to climate, volcanism,
CRET PRESENT and changes in atmospheric CO_, in The Carbon
n- 290 JJAß% ß JJA
Cycle and Atmospheric CO_: Natural Variations
PRESENT Archeanto Present, Geophysical
Monoqraph
CRET
Series' 32, edited by E. T. Sundquist and W. S.
< 286 PRESENTßß DJF Broecker, pp. 504-530,AGU, Washington, DC, 1985.
DJF
O i & I i Barron, E. J., A warm, equable Cretaceous- The
2 3 4 5
nature of the problem, Earth Sci. Rev., 19,
AVERAGE CONTINENTAL PRECIPITATION 309-33•, 1983.
Barron, E. J., S. L. Thompson and S. H.
x 1(•5 mm.s- • Schneider, An Ice-Free Cretaceous? Results
from Climate Model Simulations, Science, 212,
Fig. 6. A comparison of model sim lated 501-508, 1981a.
continental
precipitation
( x 10
-• mm.s
-1)with Barron, E. J., C. G. A. Harrison, J. L. Sloan,
model simulated globally-averaged surface and W. W. Hay, Paleogeograph¾, 180 million
temperature (o Kelvin) for a series of NCAR years ago to the present, Eclo•ae Geol. Helv,
CoI•nunit¾ Climate Model experiments. Cret - 74, 443-470, 1981b.
Cretaceous mean annual, Cret 4x - Cretaceous mean Barron, E. J., S. L. Thompson, and W. W. Hay,
annual with 4x present day CO_,
2
Present - Present Continental distribution as a forcing factor
day mean annual, Pres 4x - Present day mean for global-scale temperature, Nature, 310,
annual with 4x present day COp,P - hypothetical 574-575, 1984.
polar land mass, and E - hypothetical equatorial Barron, E. J. and W. M. Washington, The role of
land mass. Present DJF, Present JJA, Cret DJF, geographic variables in explaining
and Cret JJA are annual cycle experiments for the paleoclimates' Results from Cretaceous climate
Present day (Pres) and Cretaceous (Cret) for model sensitivity studies. J. Geoph•s. Res.,
December, January and February (DJF) and June, 89, 1267-1279, 1984.
July and August (JJA) averages. Barron, E. J. and W. M. Washington, Warm
Cretaceous climates: high atmospheric CO_ as a
plausiblemechanism,
in TheCarbon
Cycle•nd
cycle of temperature decreased by as much as 10øC Atmospheric
CO2: Natural Variations Archeanto
in the annual cycle simulations in some latitude Present, Geophysical Monograph Series: 32,
zones. Barron et al. (1984) estimate that the edited by E. T. Sundquist and W. S. Broecker,
range in model simulated globally-averaged pp. 546-553, AGU, Washington, DC, 1985.
surface temperature due to variations in Beaty, C., The causes of glaciation, Am. J. Sci.,
geography in the mean-annual version of the CCM 66, 452-459, 1978.
is greater than 7øC. Each of these results is a Berggren, W. A. and C. D. Hollister,
substantial climatic sensitivity to geography. Paleogeograph¾, paleobiogeography and the
In addition, the focus on temperature has history of circulation in the Atlantic Ocean,
hardly touched the degree to which geography in Studies in Paleo-Oceanography, Soc. Econ.
influences climate. As one example, the range Paleotol. Min. Spec. Publ. 20, edited by W. W.
of CCM experiments described above alter Hay, pp. 126-186, 1974.
continental precipitation (Figure 6) Berner, R. A., A. C. Lasaga, and R. M. Garrels,
substantially, yielding almost an order of The carbonate-silicate geochemical cycle and
magnitude variation in continental run-off. its effect on atmospheric carbon dioxide over
Such differences have considerable implications the past 100 million years, Am. J. Sci., 283,
for plants, plant productivity, continental 641-683, 1983.
weathering and geochemical cycles among other Berner, R. A. and E. J. Barron, Comments on the
factors. Further, the role of geography in BLAG
model' factors affecting atmospheric
CO2
modifying the atmospheric circulation and the and temperature over the past 100 million
nature of the ocean circulation has received years, Am. J. Sci., 284, 1183-1192, 1984.
little attention. Consequently, the ways in Budyko, M. and A. Ronov, Chemical evolution of
which geography alter climate are only beginning the atmosphere in the Phanerozoic, Geokhimiya,
to be identified and understood. •, 643-653, 1979.
Geophysical Monograph Series Understanding Climate Change Vol. 52

BARRON 157

Chamberlin, T. C., An attempt to frame a working Geophysical Monoqraph Series' 32, edited by E.
hypothesis of the cause of glacial periods on T. Sundquist and W. S. Broecker, pp. 397-411,
an atmospheric basis, J. Geol., •, 545-585, AGU, Washington, DC, 1985.
1899. Luyendyk, B., D. Forsyth, and J. Phillips, An
Covey, C. and E. Barron, The role of ocean heat experimental approach to the paleocirculation
transport in climatic change, Earth Sci. Rev., of oceanic surface waters, Geol. Soc. Am.
24, 429-445, 1988. Bull., 83, 2649-2664, 1972.
Covey, C. and P. L. Haagenson, A model of oxygen Manabe, S. and R. T. Wetheraid, The effects of
isotope composition of precipitation' doubling the COA2
concentration on the climate
implications for paleoclimate, J. Geophys. of a general circulation model, J. Atmos. Sci.,
Res., 89, 4647-4656, 1984. 32, 3-15, 1975.
Crowell, J. and L. A. Frakes, Phanerozoic Rind, D., The dynamics of warm and cold climates,
glaciation and the causes of the Ice Ages, Am. J. Atmos. Sci., 43, 3--24, 1986.
J. Sci., 268, 193-224, 1970. Sandberg,P. A., Nonskeletalaragonite andpCO
2
Crowley, T. J., D. A. Short, J. G. Mengel, and G. in the Phanerozoic and Proterozoic, in The
R. North, Role of seasonality in the evolution CarbonCycle and AtmosphericCO?• Na•ura•
of climate during the last 100 million years, Variations Archean to Present• Geophysical
Science, 231, 579-584, 1986. Monograph Series' 32, edited by E. T. Sundquist
Dickinson, R. E., Climate sensitivity, Advances and W. S. Broecker, pp. 585-594, AGU,
Geoph¾s., 28A, 99-129, 1985. Washington, DC, 1985.
Donn, W. and D. Shaw, Model of climate evolution Schlesinger, M. E. and J. F. B. Mitchell, Climate
based on continental drift and polar wandering, model simulations of the equilibrium climate
Geol. Soc. Am. Bull., 88, 390-396, 1977. response to increased carbon dioxide, Rev.
Emiliani, C. and N.J. Shackleton, The Brunhes Geophys., 24, 760-798, 1987.
Epoch' Isotopic paleotemperatures and Schneider, S. H., S. L. Thompson, and E. J.
geochronology, Science, 183, 511-514, 1974. Barron, Mid-Cretaceous continental surface
Fischer, A. G., Long-term climatic oscillations temperatures: Are high COA concentrations 2
recorded in stratigraphy, in Climate in Earth needed to simulate above freezing winter
History, edited by W. H. Berger and J. C. conditions?, in The Carbon Cycle and
Crowell, pp. 97-104, National Academy Press, Atmospheric
CO2' Natural Variations Archeanto
Washington, DC, 1982. Present• Geophys. Mono•r. Ser. 32, edited by E.
Frakes, L. A. and E. Kemp, Paleogene continental T. Sundquist and W. S. Broecker, pp. 554-560,
positions and evolution of climate, in AGU, Washington, DC, 1985.
Implications of Continental Drift to the Earth Spicer, R. A. and J. T. Parrish, Pañeobotanical
Sciences, edited by D. Tarling and S. Runcorn, evidence for cool north polar climates in
pp. 535-550, Academic Press, New York, NY, middle Cretaceous (Albian-Cenomanian) time,
1973. Geoloqy, 14, 703-706, 1986.
Frakes, L. A. and E. Kemp, Influence of Tarling, D. H., The geological-geophysical
continental positions on Early Tertiary framework of ice ages, in Climatic Chanqe,
climates, Nature, 240, 97-100, 1972. edited by J. Gribben, pp. 3-24, Cambridge Univ.
Hag, B. U., J. Hardenbol, and P. R. Vail, Press, Cambridge, 1978.
Chronology of fluctuating sea Levels since the Vail, R. R., R. M. Mitchum, Jr., and S.
Triassic, Science, 235, 1156-1167, 1987. Thompson, III, Seismic stratigraphy and global
Hays, J. and W. Pitman, Lithospheric plate changes of sea level, in Seismic Stratigraphy
motion, sea level changes, and climatic and and Global Chanqes of Sea Level, AAPG Memoir
ecological consequences, Nature, 246, 18-22, 26, edited by C. E. Payton, pp. 49-212, AAPG,
1973. Tulsa, Oklahoma, 1977.
Kasting, J., Comments on the BLAG model- the Volk, T., Feedbacks between weathering and
carbonate-silicate geochemical cycle and its atmosphericCO
2 over the last 100 million
effect on atmospheric carbon dioxide over the years, Am. J. Sci., 287, 763-779, 1987.
past 100 million years, Am. J. Sci., 284, Washington, W. M. and G. A. Meehl, General
1175-1182, 1984. circulation model CO• sensitivity experiments'
Kennett, J.P., Cenozoic evolution of Antarctic snow-sea ice albedo •arameterizations and
glaciation, the circum-Antarctic Ocean, and globally averaged surface air temperature,
their impact on global paleoceanography, J. Climatic Change, •, 231-241, 1986.
Geoph¾s. Res., 82, 3843-3860, 1977. Washington, W. M. and G. A. Meehl, Seasonal cycle
Lasaga, A. C., R. A. Berner, and R. M. Garrels, experiment on the climatic sensitivity due to a
An improved geochemical model of atmospheric doubling of CO
2 with an atmosphericgeneral
CO2 fluctuations over the past 100 million circulation moael coupled to a simple
years, in TheCarbonCycleandAtmospheric CO2i mixed-layer ocean model, J. Geophys. Res., 89,
Natural Variations Archean to Present, 9475-9503, 1984.
Geophysical Monograph Series Understanding Climate Change Vol. 52

SIMULATIONS OF THE LAST GLACIAL MAXIMUM WITH AN ATMOSPHERIC

GENERAL CIRCULATION MODEL INCLUDING PALEOCLIMATIC TRACER CYCLES

SylvieJoussaume
•, JeanJouzel2, and RobertSadourny•

Abstract. Simulations of the Last Glacial well understood by simple isotope models. Howe-
Maximum have been performed with an atmospheric ver, the isotope atmospheric cycle is quite
general circulation model. We focus on a new ap- complex and is sensitive to the entire atmosphe-
proach of the problem with the modeling of ric circulation. To investigate the sensitivity
important climatic tracers: water isotopes and to a glacial/interglacial oscillation, AGCMs are
desert dust particles. The mean dependency of the only appropriate tools.
the water isotope content of precipitation with As regards desert dust particles, a large in-
temperature is similar to the present-day one, crease in the amount of dust in ice cores has
but can depart locally. The global amount of been observed for the Last Glacial Maximum
dust is little changed for the ice age, but im- [Petit et al., 1984]. Thus AGCMscan help to
portant changes are simulated over some regions. investigate the link between climate (aridity,
atmospheric circulation) and dust deposits.
Introduction

Atmospheric general circulation models (AGCMs) Modeling

can be useful tools to reconstruct past climates.
This approach has been extensively applied to the Evolution in time of a tracer is governed by
Last Glacial Maximum(LGM18000 years Before Pre- source and removal processes, which maybe highly
sent), prescribing boundary conditions (sea sur- dependent on the nature and type of tracers,
face temperature, sea ice, topography, surface large scale transport and diffusion processes -
albedo) given by CLIMAP[1981]. To validate these turbulent vertical diffusion within the planetary
simulations it is then important to comparemodel boundary layer and convective diffusion within
results with palcodata generally obtained through convective clouds. The transport and diffusion
climatic tracers (e.g. water isotopes, dust, pol- processes of tracers have been introduced using
lens). We propose, as a complementary approach, modeling schemes, which are similar to those
to introduce a modeling of important climatic found in our AGCMfor the transport and diffusion
tracer cycles within the AGCM.On one hand, this of water vapor. The AGCMof the Laboratoire de
approach allows us to obtain a more direct M•t•orologie Dynamique is described in Sadourny
comparison with observations. On the other hand, and Laval [1984].
the knowledge of both the climatic response of
tracers and of atmospheric parameters should help Water $sotopes
us in translating palcodata into atmospheric pa-
rameters. This a•roach has been applied to water The modeling of the water isotope cycles in
isotopes (HDO,H2'øO)and desert dust particles. AGCMsfollows the basic water cycle' surface eva-
Present-day observations of water isotopes poration and condensation, water vapor transport,
display a good correlation between isotope ratios atmospheric condensation processes, subsequent
of precipitation and surface air temperature, evaporation of precipitation and ground hydro-
logy. At each phase change, differentiation oc-
16 18
curs between H2 0 and the isotopes H• 0 and HDO
because of the slight differences in their vapor
1Laboratoirede M•t•orologie Dynamique pressuresand their moleculardiffusivities. The
24 rue Lhomond, 75231 Paris Cedex 05, France liquid phase is enriched in heavier isotopes
•Laboratoire de G•ochimie Isotopique/LODYC, relatively to the vapor phase. Advectionis a
CEN/Saclay, 91191 Gif sur Yvette, France critical process in isotope modeling because it
is necessary to avoid the production of negative
Copyright 1989 by values and to transport isotopic ratios. A des-
International Union of Geodesy and Geophysics cription of the modeling and preliminary results
and American Geophysical Union. can be found in Joussaume et al. [1984].

159
Geophysical Monograph Series Understanding Climate Change Vol. 52

160 AGCM SIMUlATIONS OF PALEOCLIMATIC TRACER CYCLES

.. Annual mean oxygen 18 ratio of precipitations Present

..
. >- 2
Do

.,..
... - 4

III
-1& D- ID. - 8

Fig. 1. Simulated oxygen 18 ratio of precipitations for the estimated annual mean
(February + August) present day climate. Isolines every 2% 0 down to -12°/ , every
4% 0 for lower values. Dotted areas above -2 % 0 , hatched areas between -48 / 00 and
-8% ,
0

Desert dust tions, as 20-30 days only are necessary to reach

In this first approach,only one size range of
particles has been modeled, in the order of 1 pm,
to deal with long range transport. Morever, par-
ticles are considered here as passive scalars,
which means that their eventual impact on clima-
te, either as condensation nuclei and through
their radiative properties, is not included.
Dust is raised by winds over desert zone
areas. The extent of source regions is generated
by the model itself as regions that are dry both
in February an August. The surface flux of dust
over source regions depends on the surface wind
speed and on the vertical gradient of dust near
the surface. A constant dust concentration is as-
sumed at the soil-atmosphere interface arbitra-
rily taken equal to I, which defines all dust
quantities in arbitary units.
Removal by the rainout process has been inclu-
ded, assuming the same efficiency for particles
and water vapor. Gravitational settling, although
weak, has been considered. Dry deposition near
the ground, for example by interception or impac-
tion by obstacles, has been introduced depending
on surface wind speed and dust concentration. For
a more detailed description see Joussaume [1985).
Results
Simulations of February and August present-day
and UGM climates (200 days each with tracers in-
troduced over the last 100 days) have been per-
formed using the LMD AGCM with the standard spa-
tial resolution of 64 points in longitude, 50
points in latitude, and 11 levels, that corres-
ponds to a horizontal mesh size of 400 x 400 km2
at 50 0
of latitude. The results presented here
are averaged over the last 60 days of the simula-
an equilibrium regime for tracers.
Climate
Simulated results are in general agreement
with other model results [e.g. Kutzbach and Guet-
ter, 1986; Rind, 1987), for example a decrease in
surface air temperature of 3.1 0 (February) and
4.1 0 (August), a southward shift of storm tracks
in Northern Atlantic, a splitting of the Jet
Stream around the Laurentide ice sheet (February)
and a weaker summer Indian monsoon. This last
feature is corroborated by observations. Global-
ly, the hydrological cycle exhibits nearly no
change in February and a small decrease in August
(3%). However, the latitudinal distribution of
evaporation and precipitation is modified. Note
that simulations are short and that the statis-
tical significance of results cannot thus be
obtained.
Water isotopes
The simulated charts of the oxygen 18 ratio
(H2 18 0/H 16 0 ) of precipitation for present-day
(Figure 1)and UGM climates exhibit a good depen-
dency with temperature in middle to high lati-
tudes. In tropical regions, the oxygen 18 ratio
is modulated by the amount of precipitation. In
high latitudes, the simulated difference (UGM-
Present-day) in the oxygen 18 ratio of precipi-
tation is in good agreement with ice core data
(of the order of -5 % 0 in Antarctica). Consi-
dering the statistical dependency between 18 0 in
precipitation and surface air temperature (Figure
2), no clear difference is simulated between pre-
sent-day and the UGM climates. However, the
variability displayed in this relationship imp-
Geophysical Monograph Series Understanding Climate Change Vol. 52

JOUSSAUME, JOUZEL AND SADOURNY 161

Present Ice age

20.

.
10.
_ 6•8o(O/oo) a 6•8o(O/oo)

-10.

-20.
_

-30. _

-40.

-50.
-60. -40. -20. O. 20. •0. -qO. -20. O. 20.

surface air temperature

Fig. 2. Simulated relationship between the oxygen 18 ratio of precipitations and

surface air temperature for the estimated annual mean (February + August) : a) present
day climate, b) Last Glacial Maximum, and their associated regression lines (below
15øC, slopes a).57 and b).65).

lies some caution when interpreting regional associated with changes in the atmospheric
results. circulation. The simulated change in dust depo-
sits over East Antarctica is weak compared to
Desert dust observations [Petit et al., 1981]; this discre-
pancy is more likely to be due to an under-
The simulated extent of arid regions, con- estimation in the change of source regions for
sidered as source areas of dust, is reasonable, the LGM.
except for an underestimation of the Australian
desert. Simulated modern desert dust results show Acknowledgments. This work was supported by
a clear seasonal cycle, with a double amount of the Programme National d'Etude de la Dynamique du
the atmospheric content of dust in August compa- Climat. The computer time was contributed by the
red to February [Joussaume et Sadourny, in Centre de Calcul Vectoriel pour la Recherche.
press]. During the Last Glacial Maximum, a small
increase is simulated by the model both for the References
source regions (+18%) and the atmospheric dust
content (+8%), with a stronger increase in Feb- Joussaume S., Mod•lisation des cycles des esp•ces
ruary, leading to a weaker seasonal contrast. isotopiques de l'eau et des a•rosols d'origine
However, important changes are simulated in some d•sertique dans un module de circulation
regions, for example over the tropical Atlantic g•n•rale de l'atmosph•re, Th•se de 3•me cycle,
Ocean, and over Europe (Figure 3) and are mainly Universit• de Paris VI, 1983.

Annual mean dust removal Ice age/ Present

::::::::::::::::::::::::::::

•iii'
ß'•- <•',•
' ..'
....:b'ii::
::
.....ii!• -.• lO

=============================================
.....:::......
:::::;:
....................
g:....................
;2;.......
Fig. 3. Simulated total removal of dust (dry and wet) change ß ice age/present day
ratio for the estimated annual means. Isolines 1/2, 1, 2, 5 (shading for values above
1.)
Geophysical Monograph Series Understanding Climate Change Vol. 52

162 AGCM SIMULATIONS OF PALEOCLIMATIC TRACER CYCLES

Joussaume S., J. Jouzel and R. Sadourny, A past 18,000 years, J. Atmos. Sciences, 43,
general circulation model of water isotope 1726-1759, 1986.
cycles in the atmosphere, Nature, 311,24-29,1984. Petit J.R., M. Briat and A. Royer, Ice age aero-
Joussaume S., Simulation of airborne impurity sol content from East Antarctic ice core
cycles using atmospheric circulation models, samples and past wind strength, Nature, 293,
Annals of Glaciology, l, 131-137, 1985. 391-394, 1981.
Joussaume S. and R. Sadourny, Simulation of the Rind D., Components of the ice age circulation,
desert dust cycle using an atmospheric general J. Geophys. Res., 92, 4241-4281, 1987.
circulation model,IAMAP Conference on Sadourny R. and K. Laval, January and July
"Aerosols and Climate", Vancouver 1987, publi- performance of the LMD general circulation
shed by A. Deepak, in press. model. In A. Berger and C. Nicolis (eds) New
Kutzbach J.E. and P.J. Guetter, The influence of perspectives in Climate Modelling, Develop-
changing orbital parameters and surface bound- ments in Atmospheric Sciences, 16, Elsevier,
ary conditions on climate simulations for the 173-198 1984.
Geophysical Monograph Series Understanding Climate Change Vol. 52

PROGRESS AND FUTURE DEVELOPMENTS IN MODELLING THE

CLIMATE SYSTEM WITH GENERAL CIRCULATION MODELS

P. R. Rowntree

DynamicalClimatologyBranch,MeteorologicalOffice
Bracknell,England, RG12 2SZ

I ntr oduc ti on example is the irregular 3-5 year time scale

tropical Pacific ('El Nino') ocean surface
The objective of this paper is to review the temperature variation. Another major use is for
present status of climate models as a tool in the study of the effects of naturally forced
studies of climate change, and to consider the perturbations of climate. These fall into 3 groups
needs for future developments. Progress in (Table 1). Firstly, there are astronomic
development of general circulation models (GCMs) variations:- in solar output these range from
for use in climate studies is a worldwide effort, long trends over billions of years to shorter term
with many of the major contributions coming from variations, 22 or 80 year cycles perhaps; then
the USA modelling groups. For purposes of there are the perturbations of the earth's orbit
illustration here I have mainly used results from which have been discussed in this symposium by
the UK Meteorological Office GCM, and must record Peltier, and in a union lecture by Berger.
my debt to those who did most of the actual work Secondly, there are variations in atmospheric
of running the models and extracting the results. composition - in volcanic aerosols as discussed by
Sigurdsson in this symposium, trace gases such as

Definition of Climate Models

the CO2 variations associated with glacial cycles,
and tropospheric aerosols in which variations may
arise due to changes in vegetation or surface
Firstly, I must say what is generally meant by winds -these may also be associated with glacial
a climate model in the GCM context. It contains
cycles. Thirdly, there are changes in surface
models of the atmosphere, ocean, sea-ice and land; boundary conditions - those due to continental
since the pioneering GFDL models constructed by drift as Barron has discussed in this symposium
Smagorinsky, Manabe and Bryan in the 1960s the and those in vegetation, ice caps etc. associated
variables have been generally at least surface with glacial cycles.
pressure, wind, temperature and moisture in the Turning now to man-made perturbations (Table 2)
atmosphere, current, temperature and salinity in there are again three groups. Firstly,
the ocean and snow, surface temperature and soil atmospheric composition- study of the effects of
moisture on land. There have been increases in
increases in trace gases is probably the most
resolution and complexity. For example, the important use of climate models and is discussed
present UKMeteorological Office climate model has elsewhere in these symposium
proceedings.
11atmosphere
1/2ø layer•
latitudeby 3 /4and 17oceanlayerswitha 2
ø longitude
horizontal
Secondly, man's tendency to change the land
surface, like his effect on CO
2, imposeson
resolution, and a much more complete climatologists the responsibility of assessing the
representation of the land surface is being effects of such changes (see Nicholson, this
incorporated. symposium). Thirdly, there are changes in water
management - diversion of rivers, construction of
reservoirs, drainage of wet lands. I have
Uses of Climate Models
indicated under each item in Table 2, the
processes in a climate model affected by the
Let us now consider the uses of climate models.
changes. This is relevant to the next topic -what
An important area is the investigation of natural,
requirements these uses of climate models place on
internal variability of the climate system; one the models. Table 3 shows these for some
applications. There are two types. Firstly, to
be able to study the effects of these changes it
Published 1989 by the American Geophysical is necessary to be able to represent the
Union. perturbation. Thus, we cannot study the effects

163
Geophysical Monograph Series Understanding Climate Change Vol. 52

164 PROGRESS IN MODELLING CLIMATE

TABLE 1. Natural Perturbations of Climate TABLE 3. Requirements on Climate Models

(a) Astronomic variati ons :- (a) Represent perturbations

Solar output e.g. deforestation, trace gas radiation
(long term trends, oscillations) effects.
Ear th' s orb i t (b) Describe effects
(b) Atmospheric composition variations e.g. Agriculture -
Volcanic stratospheric aerosols radiation
Trace gases, tropospheric aerosols temperature
(glacial cycle variations in vegetation etc) soil moisture (precipitation,
(c) Surface boundary conditi ons evaporation)
Continental drift snow cover, frost
Glacial cycles sea level
(icecaps, sea level, lakes, vegetation) similarly for:
Fisheries, Water, Power, Transport,
Construction, etc.

of aerosols on precipitation formation without a

sufficiently realistic representation of cloud
microphysics, in which the effect of aerosol size strong winds for ship icing. However, today I
distribution and other factors is taken into shall restrict my discussion to time averaged
account. Deforestation cannot be properly results.
simulated if the model's hydrology parametrization
cannot represent the effects of a change from Some Results
forest to, say, pasture.
The other requirement is that the effects be Let me now take you fairly quickly through some
properly described. Here the need is not only for recent results, most of which provide some
adequate model diagnostics but also sufficiently illustration of progress in modelling. One quite
realistic simulations of the required model instructive model field is that of sea level
parameters. Just taking one example, agriculture pressure. The two top frames in Figure 1 show a 4
(Table 3), you see that the list is quite year observed northern winter normal - very like
extensive. Note also how what is needed is nearly longer periods - and an 8 year mean from our model
always surface data - not upper winds and as it was around 1984. Note the excessive westerly
temperatures; indeed, of the non-surface flow, a problem common to relatively high
variables, only clouds will occur frequently in resolution models at that time. This was not a
the list of requirements and then mainly because problem in the Southern Hemisphere - not shown
of the impact on surface radiation. Also, it is here (see Slingo and Pearson [1987]) but the
often frequency distributions and interannual simulation there is very good with mean
variability that are required, not means; and the sub-Antarctic pressures of about 980 mb and
frequency distributions may be complex - e.g., realistic sub-tropical ridges. This suggests a
occurrence of low temperatures together with problem with orography, as this is relatively
unimportant in the Southern Hemisphere. The first
solution proposed was to enhance the orography
(e.g. Hills [1979]) in order to represent the
TABLE 2. Manmade Perturbations of Climate
barrier effects of mountains better. Later, the
envelope orography technique [Wallace et al, 1983]
was tested [Slingo and Pearson, 1987]. This
(a) Atmospheric composition worked quite well in winter but the northern
Trace gases
summer circulation was degraded, probably because
(fossil fuel, CFCs) the elevated heat sources were too elevated.
- Radiation
The second solution was to incorporate a
Aerosols
representation of the vertical transfer of
(land use, ocean biota) momentum by gravity waves, •n other words gravity
- Radiation, cloud physics wave drag [Palmer et al., 1986, S•ingo and
(b) Land surface Pearson, 1987]. This also had a dramatic effect
(deforestation, afforestation,
(Figure lc) on the simulation as easterly stress
desertification, irrigation) or momentum flux convergence, applied at the
- Radiation, turbulence, hydrology earth's surface over rough (i.e., mountainous)
(c) Water management areas, was transferred to higher levels, where it
(river diversions, reservoirs)
decelerated the westerly flow. With this change,
- Runoff, salinity, evaporation how realistic are model winds? The results in
Figure 2 are for a 4 winter mean from our latest
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 165

a
1024

1032•
1024•

b
•,992 984

103 16
1024

1008

C _

!ooo

1024

1024 1016
1Ol

oo8
Fig. 1. December-February sea level pressure: (a) observed, based on Meteorological
Office (MO) operational analyses for 1983-86; (b) modelled, average for 8 winters
without gravity wave drag; (c) modelled, average for 4 winters with gravity wave drag.

long run of the model. The pattern of zonal mean
winds is quite close to observations, though the
jets are too strong. Differences greater than 5
m/s are shownin Figure 3 -mostly they are
confined to the stratosphere. The other contours
here are for temperature errors - mostly these are
less than 4 K except in the polar stratospheres. A
similar result is obtained in June to August
(Figure 4). This stratospheric coolness is not a
problem peculiar to our model - all three of the
models used for CO2 experiments r•viewed by
Schlesinger and Mitchell •1985] suffered from such
errors, especially in the Southern Hemisphere.
possible cause is the rather poor resolution of
the stratosphere - we have only three layers above
200 mb - in view of the important role of vertical
motions in the stratospheric heat balance.
However,Mahlmanand Umscheid[1984] have error•
approaching 10 K in the January of their
simulation with a much higher resolution (40
level) model. One might also suspect radiation as
a cause. The ICRCCM(Intercomparison of Radiation
Codesfor Climate Models) organised a valuable
comparison of line-by-line models and
parametrizations for climate models; Figure 5
comparestwo Line-by-line comparisonsof long wave
cooling - GFDLand LMD. Though this shows
generally good agreement, there are differences.
Apart from the coarse resolution, the MO model's
differences from the line-by-line calculations are
of comparable magnitude; however there is too much
A cooling from 60 to 370 mb. Note that these
schemesomit the water vapour continuum. Though
this is a poorly understood area of radiation
theory, a better representation would not affect
the stratospheric cooling.
Further intercomparisonprogrammes are underway
- for boundaryand surface fluxes, as discussed by
Geophysical Monograph Series Understanding Climate Change Vol. 52

166 PROGRESS IN MODELLING CLIMATE

200

600

90N 60 30 0 30 60 90S
Latitude

200

600

90N 60 30 0 30 60 90S
Latitude

Fig. 2. December-February zonally averaged zonal winds: (a) modelled 4 year average;
(b) observed 3 year mean (1983/4-1985/6). Contour interval 5 m/s; easterlies shaded.

MeBean (this symposium). In this context, it is stress on the ocean surface, averaged over the
r•levant to •xamlne the surface winds which are of year, compared with Hellerman and Rosenstein's
importance for driving an ocean model. Figure estimate from observed data. There are
shows the zonal mean of the eastward surface differences, but mostly these probably owe as much
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 167

200 k----1I---+---+-+:i=--'""""+-'-+--+

600 F4rl-----1r--+---l\--#-+---+---+--+--+-----lf----+--+---+---It--H-----1r--I-'~

90N 60 30 o 30 90S
Latitude

Fig. 3. December-February zonally averaged model errors for temperatures (solid) and
zonal winds (dashed). Contour interval 2K for temperature, 5 m/s for winds.

to uncertainties in the data and the differences
in the stress formulation as to errors in the
model winds. An exception is probably near 50-60°
S, where the model's peak westerlies are a few
degrees too far from the pole, and north of 60° N,
where some easterly bias is present.
Rainfall (Figure 7) is important for
agriculture over land and over the ocean because
of salinity's role in driving ocean currents.
This simulation, though not as good in several
respects as some of our earlier ones, nevertheless
serves to demonstrate that broadly the patterns
are realistic. The 'desert' areas with less than
1 rom/day and the rainy areas with more than 5
rom/day are broadly similar in the simulated and
observed maps. Features which were worse here
than in most of our experiments included several
parts of the tropics, for example the excessive
rain over South America and northern Australia and
the deficient rain over the tropical Atlantic and
southern tropical Africa. For many purposes, it
is also important to simulate the runoff well - we
have recently started to calculate runoff for
selected basins, simply summing the runoff of grid
bOKes within the catchments. One basin we
considered was the Mackenzie River which runs
through Arctic Canada. The result (Figure 8) was
disappointing with too strong a summer snow melt
peak and too 11 ttle the rest of the year.
However, the model is not necessarily so wrong in
high latitudes. Another comparison (Figure 9) we
made was for the region of Siberia from 60-105° E
north of 50° N, which closely apprOKimates the
basins of the Ob'Irtysh and Yenisei. Here, the
model had nearly 50% too much run off, but the
apportionment through the year is close to
observed.
An important need in climate change studies,
especially those involving CO 2 , may be sea level
information. To obtain this, it is essential to
have an ice 'sheet model and a prerequisite for
that is realistic snow accumulation. This is
obviously very dependent on getting precipitation
right but other factors are also important, as
shown by the ice budget analysis for the North
American ice sheet at the last glacial maximum in
Manabe and Broccoli [1985]. Their results show
the sublimation to be nearly half as large as
snowfall. We have calculated snow budget data for
Antarctica and compared it to the estimates by
Schwerdtfeger [1970] (Figure 10). The broad
pattern of a minimum of less than 5 g cm- 2 yr- 1
Geophysical Monograph Series Understanding Climate Change Vol. 52

168 PROGRESSIN MODELLING CLIMATE

2OO

600

90N 60 30 0 30 60 90S
Latitude

Fig. 4. As Figure 3 except for June-August.

over most of eastern Antarctica6 with peak values northwest Atlantic and cool northeast Atlantic.
near the coast of over 40 g cm
-= yr-1 in places, Most of these errors are, I think, not well
is simulated quite well. The South Pole understood.
apparently has too little - there is a maximum on
the south edge of the Ross Ice Shelf, but like Future
much of western Antarctica, magnitudes are too
small. Now I want to touch on a few of the many
The results discussed so far are from challenges and problems in modelling climate.
atmosphere models. It is obvious that the Table 4 indicates some likely future developments
simulation of the surface climate from coupled
models cannot, except through chance cancellation
of errors, be expected to be as good as that for
atmospheric GCMs, which use climatologically TABLE 4. Future Developments, Problems
correct sea surface temperatures (SSTs) and sea
ice. So the crucial question is how good are the
SST and ice simulations in coupled models - I will (a) Ocean: - eddies
discuss only SST, referring to Han et al. [1985]'s Tropical (El qino)
useful comparison of annually averaged SST errors. Extratropical (glacial cycles)
Their maps show rather large areas with errors (b) Cloud: - water content -> albedo
exceeding 4 K -i.e., greater than the largest (c) Mountains: - elevation
observed anomalies - and there is much in common (barrier or heating)
between the simulations which is only partly gravity waves
disguised by the range of up to 2 K in the mean (d) Land surface: - soil, etc datasets
errors. These common features include a warm (e) Other possible developments:
Antarctic Ocean and eastern southern hemisphere vegetation models
ice sheet models.
oceans, cool water in the northwestern tropical
and sub-tropical oceans, warm north Pacific and
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 16 9

I I I and problems. One of the most important may be

ocean eddies - how important for modelling climate
is it to get the eddy transports right? And if
they are crucial, how can they be simulated
without the enormous increase in computing power
needed to represent the eddies explicitly - about
3 orders of magnitude with a 10 km mesh instead of
CO: + O) + H:O (no continuum)
a 300 km one? Possibly they can be parametrized
Mid-latitude summer atmosphere _
40 -though the methods have not yet been developed.
Table 4 also lists some other ocean modelling
problems. Firstly initialization: can WOCE provide
60
a suitable data set for, say, initializing the
80 ocean for 100 year CO
2 runs starting in 19957
Secondly, the tropical oceans -do we need to
I00
simulate the delicate balance illustrated in real
data by the existence of two quasi-steady states
in the tropical Pacific? Thirdly, the high
MO latitude oceans - the large and quite rapid

GFDL
200
variations of North Atlantic circulation, for
example around the Younger Dryas period, suggest
another delicate balance.

LMD Another major problem area is clouds. As

400 - -
discussed by Schlesinger in this symposium,
-

current GCMs mostly omit any dependence of cloud

radiative properties on water or ice content, and

600
_-I
oo_- ,
I), - so omit a possibly important
as suggested by Somerville
albedo tends to increase
negative feedback

with temperature. The

if,
and Remer [1984],
1000 ! 1
0 ! 2 3 4 5 6 problem is not with calculating the water or ice
Cooling rate (K/day) content. There are ways of doing this already.
Fig. 5. Longwave radiative cooling rates from Figure 11 shows a calculation of zonally averaged
line-by-line calculations with GFDL (sotid cloud water content for September (R N B Smith,
curve) and LMD (dashed) schemes (Luther, 1984) personal communication) with a scheme developed
and the MO GCM broadband scheme of Slingo and from that of Sundqvist's. It verifies quite well
Wilderspin (1986) (A. Slingo, persona.l against SMMRdata for ocean only, except that the
communication) (thin solid lines). falloff towards high latitudes is too early and

1.5

.. == o.s /

I
90 ø N
.........
........
.,.,./.` o
90 øS

-I.0

Fig. 6. Zonally averaged annual mean zonal wind stress (sea only), observed (Hellerman
and Rosenstein, 1983) (dotted), and modelled (• year mean for 80 mode•) (solid).
Geophysical Monograph Series Understanding Climate Change Vol. 52

PROGRESS IN MODELLING CLIMATE

b ::::::::::::::::::::::::::::::::::
::•::..-...••:;.:-•
'•--•'.-'-•'
' -.-•.•
'....''
:--.-.. :i•::i:::"
'• •::::::::::::::::....•:::::::::::::::::
:::::::::::::::::::::::::::::::
...........-...-.......-...-..•........••.•
ß'.'.'.'.'.'.'.'.'.'.'.'.'.'..• .'.'.' ß ß•.-.-.-...---•......................•-
ß ß '.'.'. ß ß ß -•.... .... • - j •
..2 ••...-.-.-...-.-.•.
ß .'.'.'.' ' ' ----•.....-.-..•...........-•
ß .'..'.'.:.:-:.:.......- ... -.-. '-', ..•=•,•.-.......-.-.-.-...-.
..... .'.'-'.'-:-:..'.:-:-:.

ß...=====================
:::::::::::::::::::::::::::::::
.........•
;'.... '- o -:-,.....
:::::::::::::::::::::::
======================
:.'.-
"'-'•1• '•) liiii!!ii!::.:!ii;,•.fc---'• '• '• $ • '__ '.-:-'•:::'•i:i•i:i:i:i::':'•f.:
:::::::::::::::::::::::::::::::::::
::.,•.
)_.."i-,

• • 0":'::
i::
i:::::
:'-"
?.0-.•.••
••t•"'"•i•;;•
...•...]
i '"::i:i:i:i:i:i:i:i:i:i:!
•-• '.i:i:i:i:i:i:i:i:!:i:::i:::'
:::::i:i:i:i
'.i''
''"'
:"•
'f': :"".'•
..............
, =========================
.• .:.:.:.:.:.:.:.:.:.:.:.i:i:i:i:i::
<.:" 2 •--_ - ? . . ;.
.::'.!:!:!:
::::::::::::::::::::::::::::::::::
.ii:::i:
-
:- o
.':•" _,,'--
•i!i!!.!.!.!..!.!.........
""::.. ":':':':'"
":::i:i:!:i:!
i:i:':
..i:i:i :,?'-',

•: _
-•
"'.'.'-:-:-:.:.:-:-:-:-:.:.'..i'
:•"
'"'::{!!iiiiii
•
iiii::i!•'
-,
============================i ========================:••
:::::::::::::::::::::::::::::::::::::::::
ß ' ' '--",,/ :::::::::::::::::::::
i:i:i: --" fi/'/' .....:.:.:.:::::::
!:!:!:i:i:i:i:i:i::'

Fig. 7. December-February precipitation (ram/day): (a) 4-year mean from MO model;

(b) observed from Jaeger (1976).
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 171

,oF
40

Model

"20

•Observed
I0

o
Aug Oct Dec Feb Apr Jun Aug

Fig. 8. Modelled (solid) and observed (Mackay and Loken, 1974) (dashed) runoff
for the Mackenziebasin (103m3s
-1).

30

= 20

•' I0

0 I I
Aug Oct Dec Feb Apr Jun Aug

Fig. 9. As Figure 8 but as percentage of annual total and for Ob'-Irtysh and
Yenisei basins (observed) and 60-105 E north of 50 N (modelled).
Geophysical Monograph Series Understanding Climate Change Vol. 52

PROGRESS IN MODELLING CLIMATE

40
•0
20

40
30

2O 3O

Snowaccumulation
(gcm-2yr
-')

0ø
b

90øW• 90øE

60

70 ø S-
30

180 ø

Fig. 10. Annual

snow
accumulation
(g/cm2/yr)for Antarctica:(a) modelled
fromMO
model; (b) observed according to Giovinetto from Schwerdtfeger (1970)
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 173

240 -

220

2OO

•'• 180 F \
I
/
140

• 120
.--.

'-- I00

o
/
-• 80
' /
o 60 - /
/
40
7

20

I I i I I I I I I I I I I I IN• I I I
90 80 70 60 50 40 30 20 I0 0 -I0 -20 -30 --40 --50 -60 -70 --80 --90

Latitude (degreesnorth)

Fig. 11. Zonallyaveraged

cloudliquid water (g m-2) as observed(oceanonly) from
Nj•ku and Swanson (1983) for July 11-August 10 1978 (solid) and modelled for
June-August from MO model (all longitudes) (dashed).

too fast compared with the satellite data. These results are consistent with paleoclimatic
However, the water path must still be interpreted data indicating high lake levels over the Saharan
in terms of cloud radiative properties. This is region at about 9000-6500 years BP. Mitchell et
not straightforward as Figure 12 (W. Ingram, al. E1988] used the UK Meteorological Office
personal communication), showing several proposed model with a 50 m slab mixed layer ocean to
relationships between reflectivity and water simulate the effects, but although rainfall
content, illustrates. Clearly, estimates of drop increased over southern Asia and eastern Africa,
size effective radii are needed if uncertainties they obtained little change, even some decreases
of up to 10 per cent in albedo are to be avoided. over the central and western Sahara. The
Land surface parametrizations are becoming much explanation of this difference between the models
more comprehensive. However, the land surface was traced to the formulations in evaporation: the
poses some other interesting problems. Let me NCAR model used one quarter of the potential
illustrate these with the simulation of the evaporation, however little soil moisture there
climate 9000 years ago when the solar radiation at was, so that evaporation was able to respond to
the top of the atmosphere differed from that of increases of radiation and allow the model to
the present day by about as much as it has in the develop a moist climate. In the Meteorological
last few tens of thousands of years. There were Office scheme, any increase in evaporation had to
increasesof 30-40Wm
-2 overparts of the Northern be metfromsoil moistureand, as there usually is
Hemisphere in summer. This was due mainly to the little over' the model Sahara, no evaporation
different time of perihelion - in July compared to increase occurred. However, if the albedo was
January at present. Kutzbach and Guetter [1986] decreased to values typical of grassland,
simulated the effects of this in the NCAR CCM and precipitation did increase fairly widely as a
obtained precipitation increases in summer across result of the Charney albedo feedback mechanism.
the Sahara and Arabia, as well as southern Asia. Now, such a vegetation change could occur
Geophysical Monograph Series Understanding Climate Change Vol. 52

174 PROGRESS IN MODELLING CLIMATE

1.0

As
Stephens .... Cu
.8 eee
eeeee ••x' St
7#m

ß_ .......... 30 #m
0.6

Ci

0.4

0.2

0.0
0 I O0 200 300 400 500 600 700 8O0
Watercontent
(g/m:)
Fig. 12. Estimates of the relation between reflectivity and cloud water content for
different clouds (stratus, cumulus, altostratus and cirrus from Lieu and Wittman
(1979), Charlock and Ramanathan (1985) 7 and 30um effective drop radius and Stephens'
(1984) spectral average.

gradually as a result of a slow increase in climatic effects of natural and manmade

moisture availability but, without interactive perturbations of the forcing. Such
vegetation, the climate model cannot represent it. applications mostly require diagnostic
This brings me to my final point, that we must output relevant to climate at the surface,
expect to expand climate models. Such an rather than higher in the atmosphere.
expansion to include modelling of vegetation (2) Model characteristics are obviously
appears necessary. Eagleson's talk in this determined in part by the requirement for a
symposium suggests possible directions for such a realistic simulation. However, the need to
development. However, note that the role of man represent perturbations of the forcing and
in determining vegetation implies a need for the nature of the guidance required are also
models of society. The same applies to other important constraints.
expansions - not perhaps to ice sheet models, but (3) Atmospheric GCMs are generally adequate for
certainly carbon cycle models (see Bolin's and climate simulation purposes though
Toggweiler's papers), and ozone chemistry models developments are needed in the
needed for experiments exploring the effects of representation of cloud radiative properties
chl or of 1 uor ocarb ons. and mountains and in the simulation of
regional variations. Important problems lie
ahead in ocean modeiling, at least if eddies
Sum•nary and Conclusions need to be resolved, and in the
incorporation of vegetation and ice sheet
(1) There are many possible uses for climate models and the representation of
models. These include the study of the interactions between climate and society.
Geophysical Monograph Series Understanding Climate Change Vol. 52

ROWNTREE 175

Acknowledgments. The author acknowledges the T. Matsuno), (Terra Sci./Reidel) pp. 501-525,
considerable help from staff of the Dynamical 1984.
Climatology Branch who developed and ran the model Manabe, S., and Broccoli, A.J., The influence of
and diagnostic programs used for many of the continental ice sheets on the climate of an ice

results •n this paper, and also the constructive age, J. Geophys. Res., 90, 2167-2190, 1985.
suggestions made by Dr Howard Cattle. Mitchell, J.F.B., Grahame, N.S., and Needham,
K.J., Climate simulations for 9000 years before
present; seasonal variations and the effect of
Re fe re nce s
the Laurentide •ce sheet, J. Geophys. Res., in
press, 1988.
Charlock, T.P., and Ramanathan, V., The albedo Njoku, E.G., and Swanson, L., Global measurements
field and cloud radiative forcing produced by a of sea surface temperature, wind speed and
general circulation model with internally atmospheric water content from satellite
generated cloud optics, J. Atmos. Sci., 42, microwave radiometry, Mon. Weath. Rev., 111,
1408-1429, 1985. 1977-1987, 1983.
Han, Y.-J., Schlesinger, M.E., and Gates, W.L., An Palmer, T.N., Shutts, G.J., and Swinba•k, R.,
analysis of the air-sea-ice interaction Alleviation of a systematic westerly bias in
s•mulated by the OSU coupled atmosphere-ocean general circulation and numerical weather
general circulation model, Coupled prediction models through an orographic gravity
Ocean-Atmosphere Models (ed. J. C. J. Nihoul), wave drag parametrization, Quart. J.R. Met.
Elsevier, pp 167-182, 1985. Soc., 112, 1001-1039, 1986.
Hellerman, S., and Rosenstein, M., Normal monthly Schlesinger, M.E., and Mitchell, J.F.B., Model
wind stress over the world ocean with error projections of the equilibrium climatic response
estimates, J. Phys. Oc., 13, 1093-1104, 1983. to increased carbon dioxide, Projecting the
Hills, T.S., Sensitiv•ty of numerical models to climatic effects of increasing carbon dioxide,
mountain representation, ECMWF Workshop on US Dept. of Energy, pp 81-147, 1985.
mountains and numerical weather prediction, pp Schwerdtfeger, W., The Climate of the Antarctic,
139-161, 1979. Climate of the polar regions, World Survey of
Jaeger, L., Monthly maps of precipitation for the Climatology, Elsevier, Vol. 14, pp 253-355, 1970.
whole world, Berichte des Deutschen Slingo, A., and Pearson, D.W., A comparison of the
Wetterdienstes, 18, No. 139, 1976. impact of an envelope orography and of a
Kutzbach, J.E., and Guetter, P.J., The influence parametrization of orographic gravity-wave drag
of changing orbital parameters and surface on model simulations. Quart. J. R. Met. Soc.,
boundary conditions on climate simulations for 113, 847-870, 1987.
the past 18000 years, J. Atmos. Sci., 43, Slingo, A., and Wilderspin, R.C., Development of a
1726-1759, 1986. revised longwave radiation scheme for an
Liou, K.-N., and Wittmann, G.D., Parametrization atmospheric general circulation model, Quart. J.
of the radiative properties of clouds, J. Atmos. R. Met. Soc., 112, 371-386, 1986.
Sci., 36, 1261-1273, 1979. Somerville, R.C.J., and Remer, L.A., Cloud optical
Luther, F., Intercomparison of radiation codes in thickness feedbacks in the CO• climate problem,
climate models (ICRCCM): Longwave clear sky J. Geophys. Res., 89, 9668-9672, 1984.
calculations, WMO/ICSU WCP-93, 1984. Stephens, G.L., The parametrization of radiation
Mackay, D.K., and Loken, O.H., Arctic hydrology, for numerical weather prediction and climate
Arctic and alpine environments (ed. J.D. Ives models, Mon. Weath. Rev., 112, 826-867, 1984.
and R.G. Barry), Methuen, pp. 111-132, 1974 Wallace, J.M., Tibaldi, S., and Simmons, A.J.,
Mahlman, J.D., and Umscheid, L.J., Dynamics of the Reduction of systematic errors in the ECMWF
middle atmosphere; Successes and problems of the model through the introduction of an envelope
GFDL 'SKYHI' general circulation model, Dynamics orography, Quart. J. R. Met. Soc., 109, 683-718,
of the middle atmosphere (ed. J.R. Holton and 1983.
Geophysical Monograph Series Understanding Climate Change
QUANTITATIVE ANALYSIS
MODEL
Michael E. Schlesinger
Departmentof AtmosphericSciencesand ClimaticResearchInstitute
Oregon StateUniversity,Corvallis,Oregon 97331
Abstract. Why do climate models, even within
the same category of the climate model hierarchy,
simulate different climatic changes for the same
forcing? The first step toward answering this
question is to calculate quantitatively the feedbacks
of the physical processesin the models. This paper
therefore presents the concept and terminology of
classical feedback analysis; three different feedback
analysis methods, one for radiative-convective
models and two others for general circulation
models (GCMs); and the application of these
feedback analysis methods to the simulations of
C O •.-induced climatic change. A rigorous
intercomparison of the quantitative values of the
feedbacksfor the GCMs is made difficult by the fact
that these feedbacks have been determined by two
different methods. This notwithstanding, the
intercomparison indicates that the contribution of
the individual feedback processesto the simulated
climatic changes is not the same for different
models, even when they are within the same
category of the climate model hierarchy.
Consequently, based on these and future feedback
analyses, the second step toward answering the
above question is to intercompare the
parameterizations of the most important feedback
processes among themselves, with more-detailed
models, and with observations.
Copyright 1989 by
International Union of Geodesy and Geophysics
and American Geophysical Union.
Vol. 52
OF FEEDBACKS IN CLIMATE
SIMULATIONS
Introduction
A hierarchy of climate models has been and is
being used to simulate past, present and potential
future climates. For example, the potential
climatic changes induced by projected future levels
of carbon dioxide (CO•.) have been simulated by: 1)
energy balance models (EBMs), which calculate
only the temperature at the Earth's surface,
generally in terms of what is identified as the global
mean; 2) radiative-convective models (RCMs),
which calculate the vertical profile of temperature
at a point, again in terms of what is identified as the
global mean; and 3) general circulation models
(GCMs), which calculate the global geographical
distributions of a wide variety of climatic quantities.
Yet these models generally do not depict the same
climatic changes for a given forcing such as a CO2
doubling, even for the models within each of the
above categories. Why is this?
One way to answer this question is to identify
quantitatively the contribution to the climatic
changes by each of the physical processes in the
models, and then rank these physical processes in
decreasing order in terms of their contributions.
Doing this will identify which of the processes in
each model is important for the simulated climatic
changes and, thereby, permit a systematic
intercomparison of the models. Insofar as any
physical process contributes differently to the
models' simulated climatic changes, further
investigation of the treatment of that process will be
177
Geophysical Monograph Series Understanding Climate Change Vol. 52

178 QUANTITATIVE ANALYSIS OF FEEDBACKS

warranted. In particular, such further AQ AT.

investigation should intercompare the Go
parameterizations of the contributory physical
processes among themselves, with more-detailed
models, and with observations. It is only through Ad

such a systematic approach that the causes for the

models' different simulated climatic changes can
be identified, understood and ameliorated.
purpose of this paper is to present a method
whereby the contributions of the individual physical
processes to a simulated climatic change can be
determined quantitatively, namely, by the
evaluation of their feedbacks.
The concept of feedback had its origins in
Electrical Engineering [e.g., Bode, 1975] and was
concerned with the design of electronic amplifiers
such as those employed in radio. However, a
somewhat different definition of feedback
used in the study of climate [Dickinson, 1981].
this climatic context, a feedback parameter k has
been defined which decreases as the sensitivity of
the climate system increases and increases as the
sensitivity of the climate system decreases. More
recently, the classical definition of feedback has
been introduced into the study of climate [Hansen et
al., 1984], however, with a terminology that is the
reverse of the classical terminology. Therefore, the
second objective of this paper is to introduce both the
concepts and terminology of classical feedback
analysis. Earlier expositions of classical feedback
analysis in the climatic context have been presented
by Schlesinger [1985, 1986 and 1988a]. Here we
present the classical concepts and terminology of
feedback and develop a method whereby feedbacks
can be quantitatively analyzed. This method is then
applied to analyze the feedbacks in RCM and GCM
simulations of climatic change. Finally, the
conclusions of this study are presented.
The
Fig. 1. Block diagram for the climate system. AQ
and AT. = Go J are the forcing and surface-
temperature response of the climate system,
respectively, with GO the gain of the system in the
absence of feedback and J = AQ + AJ, with AJ = F
AT.. The feedbackof the climate systemis f = GOF.
has been
In the "output" from the climate system in the form of
a change in the global-mean surface temperature.
The quantity GOis the gain of the climate system,
defined as the "output"/"input", when there is no
feedback, and F is a measure of the feedback in the
climate system. If there is no feedback, F = 0, and
the input is transferred directly to the output so that
J = AQ. Then
AT. = (AT,)ø = G0 AQ ' (1)
where (AT.) o is the surface temperature change in
the absence of feedback. If there is feedback, F • 0,
and part of the output is transferred through the
feedback loop back to the input as AJ. Then J = AQ +
AJ, where AJ = F AT., hence
AT. = Go J = Go (AQ + FAT,) . (2)

Solving for AT. then gives

Feedback: Concept and Terminology

The concept of feedback can be most easily G

introduced with the aid of the block diagram for the AT.- I .of AQ, (3)
climate system shown in Fig. 1. In this figure AQ
represents the "input" to the climate system in the where f = GoF is the feedbackfactor [Bode, 1975, p.
form of a radiative forcing due to an external 32] or, more simply, the feedback. Equation (3) can
perturbation such as a doublingof CO2, and AT. is also be written as
Geophysical Monograph Series Understanding Climate Change Vol. 52

SCHLESINGER 179

6 I

Negative -0.2 Positive I

Feedback • Feedback
•""! ß

I /
I ß
-0.6
" I /
I '
/
I
-I.0
-
I
Fig. 2. The feedback/no-feedback
responseratio Rf= AT./(AT.)o versusthe feedbackf,
where AT. is the surface-temperatureresponseof the system with feedback and (AT.) o is
surface-temperature response without feedback. There is a change in the scales of both
axes at the origin.

AT. = GfAQ , (4) feedback to the temperature change without

feedback, •
where
G AT. Gf I
(6)
Gf- I -f (5) Rf=(AT*)o G
o
1-f

is the gain of the climate system with feedback.

The effect of the feedback can be characterized on the latter by Eqs. (1), (4) and (5). This ratio is shown
the basis of the ratio of the temperature change with in Fig. 2. For f = 0 there is no feedback and Rf = 1.
Geophysical Monograph Series Understanding Climate Change Vol. 52

QUANTITATIVE ANALYSIS OF FEEDBACKS

For f < 0 the feedback is negative and 0 _<Rf < 1.
Consequently, when there is negative feedback,
regardless of its magnitude, the sign of the
response is the same as the sign of the forcing.
This is in contrast to what has sometimes
erroneously inferred for the outcome of negative
feedback. For 0 < f < 1, the feedback is positive and
Rf > 1. However, for positive feedbackwith f > 1, Rf
< 0 and the sign of the response is opposite to the
sign of the forcing. This outcome, while
mathematically possible (and actually obtained by
one improperly formulated EBM; see Schlesinger
[1985, 1988a]), is not physically consistent and must
therefore be rejected.
The classical expression for feedback given by
Eq. (4) can be contrasted with the definition of the
"feedback parameter k" used by Dickinson [1981],
namely,
AT,= 1 AQ.
From Eqs. (4), (5) and (7) it can be seen that
I 1-f
• - - (8)
Gf G o '
Consequently, as f increases and the climate system
becomes more sensitive, k decreases, while k
increases as f decreases and the climate system
becomes less sensitive. This "upside-down"
behavior of k is unnecessarily confounding and can
be avoided by the use instead of the gain of the
The Feedback f For The Climate System
The net radiation at the top of the Earth's
atmosphere N can be expressed from the viewpoint
been of a planetary energy balance model as N = N(E, T.,
I). Here E is a vector of quantities that can be
regarded as external to the climate system, that is,
quantities whose change can lead to a change in
climate, but which are independent of climate. I is
a vector of quantities that are internal to the climate
system, that is, quantities that can change as the
climate changes and, in so doing, feed back to
modify the climatic change. The external
quantities include, for example, the solar constant,
the optically-active ejecta from volcanic eruptions
and the CO2 concentration(although eventually it
may change as a result of climatic change). The
internal quantities include all the variables of the
climate system other than the surface temperature
(7) T. such as the atmospheric temperature, water
vapor and clouds. BecauseT. is the only dependent
variable in this model, the internal quantities must
be represented therein by I = I(T.).
A small change in the energy flux, AN, can be
expressed as
AN= AQ- (G•- F)AT., (9)
where

AQ
=i• 0NAE. (10)
climate system with feedback, Gf, together with Eq.
(4). In fact, Cess and Potter [1984] employed the ß •E. '
1

relation given by Eq. (4), but instead of Gfthey used

the symbol k, identified as the "surface response is the change in N due to a change in one or more
function," which was therefore easily confused with external quantity, AEi,
the same symbol used by Dickinson [1981].
Moreover, Hansen et al. [1984] used an expression
equivalent to Eq. (6), but with the equivalent of f - G'1AT 3N AT, (11)
(their g) identified as the "gain" of the climate o * - 3T,
system, and Rf (their f) as the "net feedbackfactor."
In this paper we will use the classical
is the change in N due to the change in T. alone,
terminology in which f is the feedback and G is the
and
gain. Also, by virtue of Eq. (6), we will call Rf the
feedback/no-feedback response ratio. In the
followingwe will derive expressionsfor f and Go for FAT.= •3N dI.
• AT.
. •. dT. (12)
the climate system.
Geophysical Monograph Series Understanding Climate Change Vol. 52

SCHLESINGER

is the change in N due to the change in the internal physicalprocess

with • = 0.2 wouldincrease
AT, by
variables I through their dependence on T,. When 1.6oC if added to a system with an existing feedback
the equilibrium AT, is reached in responseto the of 0.5, but would increase AT, by only 0.5øC if added
forcing AQ, AN = 0 and to a system with no existing feedback.

G Zero-Feedback Gain Of The Climate System, Go

AT,= GfAQ= o AQ, (13)
1-Zf. The planetary radiative energy budget can be
J written as

I - O•
where
N= 4
P S - eo"Id,
'
(16)
aN dI.
f.=G J (14) whereS is the solarconstant,
ap the planetary
j o 3I. dT.'
albedo, e the effective emissivity of the Earth-
atmosphere system, and • the Stefan-Boltzmann
Equation (13) with constant. From this equation and Eq. (11) we can
calculate the zero-feedback gain as

f = Z f. (15)
J
(17)
is identical to Eqs. (4) and (5) which were obtained Go:I•T,J=(1-•p)S
'
solely from our consideration of the feedback block
diagram of Fig. 1. Taking
S= 1370Wm-2,ap= 0.3andT. = Ts= 288K,
It should be noted that this classical feedback
where T s is the observed global-mean surface air
analysis, together with its representation by Fig. 1,
temperature, Eq. (17) gives GO = 0.3oC/(Wm-2).
is linear because it neglects the second- and higher-
order derivatives of N. Because this is equivalent to Thus, for AQ = ART = 4 Wm-2 for a doublingof the
CO2 concentration, where ART is the change in the
theassumption
that32N/aIjaIk
= 0,where
j andk
are any two physical processes, the effects on N of net longwave radiation at the tropopause, Eq. (1)
the individual physical processes are considered to gives(AT,)o = 1.2øC. (The reasonsfor evaluatingAQ
be independent. at the tropopause rather than at the top of the
As a result of the classical linear feedback atmosphere are described subsequently.) This
analysis, Eq. (14) shows that the feedback of a value of (AT,) o for a CO2 doubling is in agreement
physical process j depends on three quantities: 1)
with what has been obtained by RCMs without
the sensitivity of the net flux N to the process as
feedbacks (see below and Schlesinger [1985, 1988a],
measured by 3N/aIj,2) the sensitivity
of the process and Schlesinger and Mitchell [1985]).
to the surfacetemperature as measured by dIj/dT.,
and 3) the zero-feedback gain of the climate system Methods Of Feedback Analysis For
Go. Consequently, the feedback of any physical Climate Model Simulations
process is zero if either the net flux N is
independent of that physical process or the physical Three methods have been used to evaluate the
process is independent of the surface temperature. feedbacks in climate model simulations, one
As shown by Eq. (13) and Fig. 2, the influence of method for simulations by radiative-convective
any particularfeedback• on the response
of the models, and two other methods for simulations by
climate system depends nonlinearly on the sum of general circulation models. In the following we
the other feedbacks. For example, the addition of a describe each of these feedback analysis methods.
Geophysical Monograph Series Understanding Climate Change Vol. 52

l e2 QUANTITATIVE ANALYSIS OF FEEDBACKS

Radiative-Convective Models processes are sequentially added one at a time.

Although this feedback analysis method is
Schlesinger [1985, 1988a] evaluated the feedbacks practicable for radiative-convective models because
in radiative-convective model simulations of C02- of their computational economy, it is not viable for
induced climatic change on the basis of Eq. (13) general circulation models because of the large
which may be written as amount of computer time that would be required.
Therefore, two methods different from that of Eq.
G
(20) have been used to analyze the feedbacks in
)j _ o AQ (18)
(AT
s- j• 1-
i=1
f.-f. ' climatic-change simulations performed by general
circulation models. In one method
radiative-convective model as a surrogate for the
use is made of a

and general circulation model. In the other method,

G only the radiative transfer model of the GCM itself
o AQ (19) is used. These two methods are described below.

(ATs)j-
1=1- j• fi ' i=1 al.
Use of a radiative-convective model. Hansen et
[1984] used a radiative-convective model to
analyze the feedbacks in the GISS (NASA Goddard
Institute for Space Studies, New York, New York)
where (ATs)j and (ATs)j_x are the surfaceair
temperature changes from two 1xCO2-2xCO 2 general circulation model simulation of the global-
simulation pairs, the first with physical process j mean surface air temperature change induced by a
and the second without. (Here and in the following
doubling of the C02 concentration. The basis for
this analysis is Eq. (6) which can be written as
we replace the surface temperature change AT.
with the surface air temperature change ATs. We J

AT -(ATs)
ø= J=E• •(ATs)
do this because the results for RCMs and GCMs are

given for AT s and not AT., and because the f= s , (21)

difference between the two temperature changes is AT
$
AT
$

small.)Solvingfor• fromEq. (18),substituting

for
the sum of the feedbacks from Eq. (19), and use of
where 5j(ATs) is the contribution to the total
Eq. (1) then yields
temperature change AT s by feedback process j
alone. Substituting Eq. (15) for f then gives
j = 1 .....J, (20)
J (ATs)j
-x (ATs
f.=5j(ATs)
J
, j =1.....
AT
J. (22)
s

where J is the total number of feedback processes.

By defining ATs = Ts(perturbation) - Ts(control), this To use the feedbackanalysismethod given by Eq.
feedback-analysis method can be used for a (22), Hansen et al. [1984] employed a radiative-
climatic-change simulation other than that for a convective model which had the same radiation
CO2 doubling. code as the GISS GCM, and first ran a 1xCO2-
2xCO 2 simulation pair with this RCM without any
General Circulation Models feedbacks. The result of this was (ATs)o = 1.2oC,
which is in agreement with the zero-feedback
The feedback analysis method defined by Eq. (20) estimate given previously. The effect of each
requires that J+l pairs of control-perturbation feedback was then determined by adding the GCM-
simulations be performed to generate the values of simulated global-mean2xCO2-1xCO2 changein the
the (ATs)
j, onepair forthe caseofzerofeedback
(j = appropriate quantity to the same quantity of the
0), and J pairs (j = I ..... J) in which the feedback RCM lxCO 2 simulation without feedback, and
Geophysical Monograph Series Understanding Climate Change
rerunning the RCM lxCO2 simulation with this
fixed modified value. (During this rerun of the
RCM lxCO 2 simulation, the temperaturelapse rate
in the troposphere was adjusted to the global-mean
value of the GCM lxCO2 simulation if it exceeded
that value.) The difference between this RCM
lxCO2 simulationwith the prescribedchangein the
appropriate quantity and the initial RCM lxCO2
simulation
withoutfeedback
gives•j(ATs).
Use of a radiative transfer model. The
principal shortcomingof analyzing the feedbacksof
a GCM simulation with an RCM is the need to use
the global-mean values of the GCM as input to the
RCM. This would be of no importance if the climate
system were linear because then [T (X)]ave =
T(Xave),wherex and T representthe input to and
output from the RCM, respectively, and "ave"
denotes the global average. However, because
radiative transfer is a nonlinearprocess,[T(X)]ave
2xCO2~1xCO2 inducedwarming give values of ATs
•eT(Xave).Thus,the RCM analysisof the feedbacks
of a GCM simulation is likely to produce erroneous
results. Another method which does not suffer this
shortcoming was introduced by Manabe and
Wetheraid [1980]. This method, which is based on
Eq. (14), can be written using our nomenclature as
f.=G 3N dI.J -G [•jN]av
e (23)
j o3Ij dTs o [ATs
]ave
where[SjN]av
e is the global-mean
changein the net
radiation at the tropopause due to feedback process
j, and [ATs]ave is the total 2xCO2-1xC02 change in
the global-mean surface air temperature.
To use the feedback analysis method given by Eq.
(23), Wetheraid and Manabe [1988] employed the
radiative transfer code of their GFDL (Geophysical
Fluid Dynamics Laboratory, Princeton University,
Princeton, NJ) GCM together with the monthly-
mean quantities over the period of one annual cycle
for both their lxCO2 and 2xCO2 GCM simulations.
For each calendar month the monthly-mean
geographical distributions of the relevant climatic
quantities for the lxCO2 simulation were input to
the radiation code, the geographical distribution of
N was calculated, and the resultant 12 "monthly-
means" of N were averaged both annually and
Vol. 52
SCHLESINGER
globally. Then this radiative computation was
repeated individually for each feedback process by
replacing the appropriate monthly-mean quantity
from the GCM lxCO2 simulation with the
corresponding monthly-mean quantity from the
GCM 2xCO2 simulation. The difference between
the latter and former calculations then gives
[•jN]ave.
FeedbacksFor A C02 Doubling
In this section we present results, obtained by
the above analysis methods, of the feedbacks in
radiative-convective and general circulation model
simulations of the temperature changes induced by
a doubling of the C02 concentration.
Radiative-Convective Models
Radiative-convective model simulations of a
which range from 0.48 to 4.2oC. Thus by Eq. (6) and
(ATs) o = 1.2oC, f = -1.5 to 0.7. Several feedback
mechanisms are likely to be the cause of this range
in f, including as T s increases: 1) the increase in
the amount of water vapor in the atmosphere as a
consequence of the constancy of the relative
humidity, 2) the change in the temperature lapse
rate, 3) the increase in the cloud altitude as the
clouds maintain their temperature, 4) the change
in cloud amount, 5) the change in the cloud optical
depth, and 6) the decrease in surface albedo.
The values of these feedbacks have been
determined by Schlesinger [1985, 1988a] using the
method given by Eq. (20), the results of which are
presented in Table 1. This table shows that: 1) the
water vapor, cloud altitude and surface albedo
feedbacks are positive, with values that decrease in
that order; 2) the cloud optical depth feedback is
negative; 3) the temperature lapse rate feedback is
either positive or negative, depending on whether
the lapse rate is controlled by baroclinic adjustment
(BADJ) or convective (MALR or PC) processes;and
4) the cloud cover feedback is unknown.
General Circulation Models
Analysis by use of an RCM. For the particular
analysis performed by Hansen et al. [1984], the
Geophysical Monograph Series Understanding Climate Change Vol. 52

184 QUANTITATIVE ANALYSIS OF FEEDBACKS

TABLE 1. Summary of the feedbacksf in RCM and GCM

simulations of CO2-induced surface air temperature change

Feedback
Mechanism RCM a GCM
GISS b GFDL c

Water Vapor 0.3 to 0.4 0.66 0.41

LapseRated
BADJ 0.1
MALR -0.25 to-0.4
PC -0.65
Total -0.26 0.05
Cloud
Altitude 0.15 to 0.30 0.12
Cover Unknown 0.10
Altitude & Cover 0.22 0.09
Optical Depth 0 to -1.32
Surface Albedo 0.14 to 0.19 0.09 0.13

Total -1.5 to 0.71 0.71 0.68

a. Based on the analysis of Schlesinger [1985, 1988a].

b. Based on the results of Hansen et al. [1984]
c. Based on the results of Wetheraid and Manabe [1988] with
modifications as described in the text.
d. BADJ, MALR and PC denote baroclinic adjustment, moist
adiabatic adjustment and penetrative convection,respectively.

effect of the 33% increase in total water
simulated by the GCM was estimated by increasing
the water vapor at each level of the RCM by 33%,
thereby giving the result for õi(ATs). To determine
the effect of the change in the vertical distribution of
water vapor simulated by the GCM, the latter was
inserted into the RCM and the
temperature change was decreasedby õ1(ATs) to
obtain•2(ATs). To determinethe effectof the change
in lapse rate simulated by the GCM, the latter was
inserted into the RCM and gave õ3(ATs). Similarly,
õ4(ATs) for the GCM-simulated change in surface
albedo was obtained. The total cloud
temperature was obtained by changing the cloud
amounts at all levels in the RCM in proportion to
the changes obtained in the GCM. The effect of
changingonly cloud cover,õ6(ATs),was obtainedby
vapor inserting a uniform cloud change in the RCM equal
to the total change in the GCM. Lastly, the effect of
the cloud altitude change,õ5(ATs),was obtained by
subtractingõ6(ATs)from the total cloudeffect.
The results of the feedback analysis using the
method given by Eq. (22) together with the Hansen et
resulting
al. [1984]valuesof•(ATs)arepresented
in Table1.
The feedback due to the changes in water vapor
amount and vertical distribution is fw = 0.66. This
is considerablylarger than the fw = 0.3 to 0.4 given
by the RCMs. The much larger fw estimated for
the GISS GCM indicates that the relative humidity
effect on increased with doubled C02 in that model, unlike
the constant relative humidity assumed by the
RCMs. Indeed, Hansen et al. [1984] state that the
average relative humidity increased by 1.5% with a
maximum of 6% at the 200 mb level. The estimated
Geophysical Monograph Series Understanding Climate Change
lapse rate feedback,fLR = -0.26, lies at the smaller
limit given by the RCMs for the moist adiabatic
lapse rate case, perhaps because the change in
lapse rate of-0.2øC/km in the GCM simulation is
less than the change in the moist adiabatic value of
-0.5oC/km. The cloud altitude feedback,fCA = 0.12,
also lies at the lower limit given by the RCMs. The
cloud cover feedback estimated for the GCM
positive. This indicates that the global-mean low
and middle clouds, whose albedo effect dominates
their longwaveeffectgiving•N/•Ij < 0 in Eq. (14),
decreased and/or the global-mean high clouds,
whose longwave effect dominates their albedo effect
giving•N/•Ij > 0, increased.Finally,the surface and Manabe [1988] presentedvalues of [AQToA] and
albedo feedback, due largely to the reduced sea ice, [•jNTOA]av
is estimated as fSA = 0.09 which is somewhat
smaller than the estimates given by the RCMs. The
total feedback estimated for the GCM is f = 0.71, of
which water vapor feedback,fw = 0.66, is the single
most important positive contributor, followed by
cloud feedback, fc = 0.22, and surface albedo
feedback, fCA = 0.09, with the lapse rate feedback,
fLR = -0.26, making a negative contribution.
Analysis by use of a radiative transfer model.
Wetheraid and Manabe [1988] present results for: 1)
the change in the net radiation at the top of the
atmosphere (TOA) due to the doubling of the C02
concentration,[AQTOA]av
e = 2.28 Wm-2;2) the total
C02-induced temperature change [ATs]ave = 4.0øC,
and 3)the [•ijNTOA]av
e values obtainedfor the
2xCO2-1xC02 change in surface albedo and the
changes in the vertical profiles of temperature,
water vapor and fractional cloud cover. From Eqs.
(3) and (23) the total feedback can be expressedby
G
1- o[AQTOA]ave
(24a)
[ATs]av
e average over the five profiles of these ICRCCM
fTOA--
o
[ATs]av
e j=l
[•jNTOA]av
e (24b)to [AQTROP]av
UsingEq. (24a) togetherwith Go= 0.3oC/(Wm
-2) and
the above values of [AQTOA]av e and [ATs]ave gives
fTOA = 0.829. As can be seenfrom Fig. 2 and Table
Vol. 52
SCHLESINGER
1, this is a very large value of the feedback, much
larger than the maximum value for the RCMs, f =
0.71, and the same value for the GISS GCM. The
reason for this large value is the use above of
[AQTOA]av
e, that is, the changein the net radiation
at the top of the atmosphere. Consequently, for the
feedback analysis of the Wetheraid and Manabe
is [1988] results to be consistent with the two other
feedback analyses, it is necessary to convert the top-
of-the-atmosphere results, [AQTOA] and
[•jNTOA]ave, to the correspondingtropopause
results,[AQTROP] and[•ijNTROP]av
e. (Wetheraid
e insteadof [AQTRO
P] and[•jNTROP]ave
because the profiles of the changed quantities
extended throughout both the troposphere and
stratosphere, and because it is difficult to define the
tropopause in a GCM simulation. However,
because: 1) the change in the net radiation flux is a
maximum at the tropopause, 2) the sign of the C02-
induced change in the stratospheric temperature is
opposite to that of the tropospheric and surface
temperature changes, and 3) the changes in the
stratospheric quantities have little effect on the
troposphere and the surface [Schlesinger 1985,
1988a], it is more correct to use [AQTRo P] and
[•jNTROP]av
e than [AQToA]and[•SjNTOA]ave.)
We
do this as described below.
For the forcing we take [AQwRoP]av e =
[AQTOA]ave + C, where C is a conversionfactor.
The value of C was obtained from the
Intercomparison of Radiation Codes used in
Climate Models (ICRCCM) study in which the
radiation model of the GCM was used to compute
the longwave fluxes for five atmospheric profiles
[Ellis, 1987, personal communication]. From the
results for the difference between the changesin the
longwave fluxes at the tropopause and at the top of
the atmosphere, C = 2.03 Wm -2. (The contribution
e and [AQTOA]av
e by the changein the
absorbedsolar radiation due to the C02 doubling is
negligible.) Thus for [AQTOA]av
e = 2.28 Wm-2, the
above conversiongives [AQTROP]av e = 4.31 Wm-2.
Using this value in
Geophysical Monograph Series Understanding Climate Change Vol. 52

QUANTITATIVE ANALYSIS OF FEEDBACKS

G [AQTROP]av
o e Conclusion
fTROP
=I - (25)
[ATs]av
e In the Introduction we recommended that the
quantitative analysis of the feedbacks in climatic-
gives fTROP = 0.677 which is comparable to the change simulations should be the first step of a
maximum value for the RCMs and the same value systematic approach to answering the question:
for the GISS GCM. Why do models, even within the same category of
Unfortunately, we do not have the analog of the the climate model hierarchy, simulate different
ICRCCM results with which to convert the climatic changes for the same forcing? To support
[•jNTOA]ave values to [•jNTROP]ave values. this recommendation we have presented in this
Consequently,we computethe components
of fTOA paper: 1) the concept and terminology of classical
from Eq. (24b) and convert them into the feedback analysis; 2) three different feedback
componentsof fTROPby analysis methods, one for radiative-convective
models and two others for general circulation
models; and 3) the application of these feedback

fTROP,
j TMfTOA
fTOA,
j =0.817 j (26)analysis
fTOA, methodsto the simulationsof CO2-induced
climatic change.
Although this conversion is by no means unique, at
least the resultantvaluesof fTROP,
j andfTROP
satisfy Eq. (15).
The results of this feedback analysis of the
Wetheraid and Manabe [1988] study are presented
in Table 1. It can be seen that the water vapor
feedback for the GFDL GCM is comparable to that of
the RCMs, and is about two-thirds that of the GISS
GCM. Interestingly, the total lapse rate feedback is
positive in the GFDL GCM in contrast to the
negative lapse rate feedback of the GISS GCM. The
combined cloud altitude and cloud cover feedback
positive for the GFDL GCM but is only about 40% of
the corresponding positive feedback of the GISS
GCM. Finally, the surface albedo feedback of the
GFDL GCM is comparable to that of the GISS GCM
and is on the low side of the RCM results.
A rigorous intercomparison of the quantitative
values of the feedbacks for the GFDL and GISS
GCMs is made difficult by the fact that the
feedbacksof these models have been analyzed by two
different methods, namely, the radiative-convective-
model method for the GISS model and the radiative-
transfer-model method for the GFDL model.
notwithstanding, the intercomparison indicates
that, although these models simulate similar
valuesof [ATs]av
e, they do so with •valuesof their
feedbacks
whichdifferin bothmagnitude
andsign,
this despite the approximate agreement of their
simulatedvalues oft s for the lxCO 2 climate.
The intercomparison of the results of these
feedback analysis shows that the contribution of the
individual feedback processes to the simulated
climatic change is not the same for different
models, even when these models are within the
same category of the climate model hierarchy. This
is a clear demonstration of the utility of the analysis
of the feedbacks in climatic-change simulations.
To improve our understanding of the behavior of
feedbacks,the feedbacksin the simulations of CO2-
induced climatic change obtained by the GISS,
NCAR, OSU and UKMO (NCAR is the National
is
Center for Atmospheric Research, Boulder, CO;
OSU is Oregon State University, Corvallis, OR; and
UKMO is the United Kingdom Meteorological
Office, Bracknell, Berkshire) GCMs should be
determined using the radiative-transfer-model
method and compared with the feedbacks of the
GFDL model which have been determined by this
method. However, the correct application of this
feedback-analysis method requires use of the
changes in the net radiative fluxes at the
tropopause, not at the top of the atmosphere.
After completion of these additional feedback
This
analyses, the feedback processes can be ranked in
terms of their contributions to the simulated
climatic changes. Then it will be time to take the
secondstep in the systematic approach, namely, the
intercomparison of the parameterizations of the
highest-ranked processes among themselves, with
more-detailed models, and with observations.
Geophysical Monograph Series Understanding Climate Change Vol. 52

SCHLESINGER • s7

Acknowledgments. I thank John Mitchell for models, in The Potential Climatic Effects of
his constructive review of this paper. This study Increasing Carbon Dioxide, M.C. MacCracken
was supported by the U.S. National Science and F.M. Luther (Eds.), U.S. Department of
Foundation and the Carbon Dioxide Research Energy, DOE/ER-0237, pp. 280-319, 1985.
Division, Office of Basic Energy Sciences of the U.S. (Available from NTIS, Springfield, Virginia.)
Department of Energy under grant ATM 87-12033. Schlesinger, M. E., Equilibrium and transient
warminginducedby increasedatmo.spheric
CO2,
References Climate Dynamics, 1, 35-51, 1986.
Schlesinger, M. E., Quantitative analysis of
Bode, H. W., Network Analysis and Feedback feedbacks in climate model simulations of CO2-
Amplifier Design, Krieger, New York, 577 pp., induced warming, in Physically-Based
1975. Modelling and Simulation of Climate and
Cess, R. D., and G. L. Potter, A commentary on the Climatic Change, M.E. Schlesinger (Ed.), NATO
recent CO2-climate controversy, Climatic Advanced Study Institute Series, Reidel,
Change, 6, 365-376, 1984. Dordrecht, pp. 653-736, 1988a.
Dickinson, R. E., Convergence rate and stability of Schlesinger, M. E., Model Projections of the
ocean-atmosphere coupling schemes with a zero- Climatic Changes Induced by Increased
dimensional climate model, J. Atmos. Sci., 38, Atmospheric CO2, in NATO Climate/Geo-
2112-2120,1981. Sciences Symposium, A. Berger and S.H.
Hansen, J., A. Lacis, D. Rind, G. Russell, P. Stone, Schneider (Eds.), Kluwer Academic (in press),
I. Fung, R. Ruedy and J. Lerner, Climate 1988b.
sensitivity: Analysis of feedback mechanisms, in Schlesinger, M. E., and J.F.B. Mitchell, Model
Climate Processes and Climate Sensitivity, Projectionsof the EquilibriumClimatic Response
Maurice Ewing Series, 5, J. E. Hansen and T. to Increased Carbon Dioxide, in The Potential
Takahashi (Eds.), American Geophysical Union, Climatic Effects of Increasing Carbon Dioxide,
Washington, D.C., pp. 130-163, 1984. M.C. MacCracken and F.M. Luther (Eds.), U.S.
Manabe, S., and R. T. Wetheraid, On the
Departmentof Energy,DOE/ER-0237,pp. 81-147,
distribution of climate change resulting from an 1985. (Available from NTIS, Springfield,
increase in CO2 content of the atmosphere, J. Virginia.)
Atmos. Sci. , 37, 99-118, 1980. Wetheraid, R. T., and S. Manabe, Cloud feedback
Schlesinger, M. E., Feedback analysis of results processes in a general circulation model. J.
from energy balance and radiative-convective Atmos. Sci., 45, 1397-1415, 1988.