JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 85, NO.

C10, PAGES 5529-5554,OCTOBER 20, 1980

Sensitivityof a Global Climate Model to an Increaseof

CO2 Concentrationin the Atmosphere
SYUKURO MANABE AND RONALD J. STOUFFER

Geophysical
Fluid DynamicsLaboratory/NOAA, PrincetonUniversity,Princeton,New Jersey08540

This study investigatesthe responseof a global model of the climate to the quadrupling of the CO2
concentrationin the atmosphere.The model consistsof (1) a general circulation model of the atmo-
sphere,(2) a heat and water balancemodel of the continents,and (3) a simplemixed layer model of the
oceans.It has a global computationaldomain and realisticgeography.For the computationof radiative
transfer,the seasonalvariation of insolationis imposedat the top of the model atmosphere,and the fixed
distributionof cloud coveris prescribedas a function of latitude and of height. It is found that with some
exceptions,the model succeedsin reproducingthe large-scalecharacteristics of seasonaland geographi-
cal variation of the observedatmospherictemperature. The climatic effect of a CO2 increaseis deter-
mined by comparingstatisticalequilibrium statesof the model atmospherewith a normal concentration
and with a 4 times the normal concentrationof CO2 in the air. It is found that the warming of the model
atmosphereresultingfrom the CO2 increasehas significantseasonaland latitudinal variation. Becauseof
the absenceof an albedo feedbackmechanism,the warming over the Antarctic continent is somewhat
lessthan the warming in high latitudes of the northern hemisphere.Over the Arctic Ocean and its sur-
roundings,the warming is much larger in winter than summer, thereby reducing the amplitude of sea-
sonaltemperaturevariation. It is concludedthat this seasonalasymmetryin the warming resultsfrom the
reductionin the coverageand thicknessof the seaice. The warming of the model atmosphereresultsin
an enrichmentof the moisturecontentin the air and an increasein the polewardmoisturetransport.The
additional moistureis picked up from the tropical oceanand is brought to high latitudeswhere both pre-
cipitation and runoff increasethroughoutthe year. Further, the time of rapid snowmeltand maximum
runoff becomes earlier.

1. INTRODUCTION insolation.They conductedextensivestudiesof the thermal,

Sincethe pioneeringwork of Callender[1938], many studies dynamical,and hydrologicalresponseof the model.
have been made on the climatic impact of an anthropogenic The presentstudyis a natural extensionof the studiesby
increasein the CO2 concentrationin the atmosphere.Earlier Manabe and Wetheraid. It investigatesthe CO2 climate sensi-
studiesof this topic [Plass,1956;Kondratievand Niilisk, 1960;tivity problem by use of a global circulationmodel of the at-
Kaplan, 1960;M611er,1963]containan evaluationof the tem- mospherewith a simplemixed layer ocean,realisticgeogra-
perature change at the earth's surfacein responseto an in- phy, and seasonal variationof insolation.In the interpretation
creaseof the CO2 concentrationbasedupon the consideration of the resultsfrom this model, one shouldrecognizethat the
of the surface radiation balance. One of the basic short- additionalcomplexityof the modeldoesnot necessarily guar-
comingsof this approachis that it cannot properly incorpo- anteethe bettersimulationof the sensitivityof the actualcli-
rate the influence of the atmosphericheat balance upon the mate. However, it is hoped that the presentstudy identifies
temperaturechange of the earth's surface.As was demon- somespecificmechanisms controllingthe sensitivityof the cli-
strated by M611er[1963], this approach leads to rather unre- mate. Specialemphasisof the studyis placeduponthe investi-
liable results. gation of the seasonaland interhemisphericasymmetriesin
Manabe and Wetheraid[1967] avoided this difficulty by us- the responseof the model climate to an increaseof the CO2
ing a radiative convectiveequilibrium model of the atmo- concentrationin the air. Someof the resultsfrom this study
spherein which the heat exchangesamongthe earth'ssurface, were summarizedbriefly in an earlier publicationby Manabe
atmosphere,and outer space are taken into consideration. and $touffer [ 1979].
Studiesof the climate sensitivityproblem with radiative con-
2. MODEL STRUCTURE
vective equilibrium models are extensivelyreviewed by Ra-
manathan and Coakley [1978]. Refer to their review for some As the box diagram of Figure I indicates,the mathematical
of the latest resultsfrom this approach. model of globalclimateusedfor this studyconsists
of (1) a
Obviously, a radiative convectivemodel is a highly sim- spectralgeneralcirculationmodel of the atmosphere,(2) a
plified model of the atmosphereand doesnot contain someof heat and water balancemodel of the continents,and (3) a
the important dynamical and physicalprocessessuch as the mixed layer model of the oceans.The descriptionof those
snow-albedo feedback mechanism and the dynamics of the three parts of the model follows.
large-scaleatmosphericcirculation.To evaluatethe response
AtmosphericModel
of the atmosphereto a CO2 increaseconsideringthese proc-
esses,Manabe and Wetherald [1975, 1980] used a general cir- The generalcirculationmodel of the atmospherepredicts
culation model of the atmospherewith a limited computa- the ratesof the changesin the verticalcomponentof vorticity
tional domain, idealized geography, and annual mean and horizontal divergence,temperature,moisture,and surface
pressurebasedupon the vorticityequation,divergenceequa-
This paper is not subjectto U.S. copyright. Published in 1980 by tion, thermodynamicalequation,and continuityequationsof
the American GeophysicalUnion. moistureand mass.The prognosticequationsassumethe by-
Paper number 80C0663. 5529
5530 MANABEAND STOUFFER.'GLOBALCLIMATE MODEL OF CO2 INCREASE

ATMOSPHERE

CONTINUITY THERMODYNAMIC
EQUATIONOF
EQUATION
OF EQUATION ,

MOTION
WATER
VAPOR RADIATION

HYDROLOGY ,HEAT BALANCE SEA ICE ,HEAT BALANCE

I
i

CONTINENT MIXEB LAYEROCEAN

Fig. 1. Box diagram illustratingthe basicstructureof the mathematicalmodel of global climate.

drostaticapproximationand use a variable, o = [(Pressure)/ Condensationof water vapor is predictedwheneversuper-

(surfacepressure)],as the vertical coordinatefor the conve- saturationis indicatedin the prognosticequationof water va-
nienceof incorporatingthe effectof surfacetopography[Phil- por. Snowfall is predictedwhen the air temperatureat an alti-
lips, 1957]. tude of 300 m falls below the freezingtemperature.Otherwise,
The horizontal distributions of the aforementioned vari- rainfall is predicted.Refer to Manabe et al. [1965] for further
ablesare representedby a finite number of sphericalharmon- details of prognosticsystemof water vapor.
ics.The model predictsthe rate of changeof the spectralcom- For the computation of terrestrial radiation, a scheme of
ponents by computing the tendenciesfrom the prognostic Rodgersand Walshaw[1966] as modified by Stone and Ma-
equationsat all grid points and by transformingthem to the nabe [1968] is used. For the computation of solar radiation
spectral domain. This transform method originally proposed flux, the schemedevelopedby Lacisand Hansen[1974]is used
by Orsag [1970] and Eliassenet al. [1970] yields an accuracy after minor modification. The seasonal and latitudinal varia-
comparable with the conventional spectral method [e.g., tion of insolation is prescribedat the top of the model atmo-
Platzman, 1960]. It also consumesmuch lesscomputer time sphere.For the sake of simplicity, the diurnal variation is re-
than the latter when the spectralresolutionof a model is high. moved from the insolation.The depletionof solar radiation
Obviously, the horizontal resolutionof a spectralrepresenta- and the transferof terrestrialradiation is computedby taking
tion of a field dependsupon the degreeof spectraltruncation. into considerationthe effectsof clouds,water vapor, carbon
For the present model, so-called rhomboidal truncation is dioxide, and ozone. The mixing ratio of carbon dioxide is as-
used. Fifteen waves are retained in both longitudinal and sumedto be constanteverywhere.A zonally uniform distribu-
meridional directions. The vertical derivatives appearing in tion of ozoneis specifiedas a function of latitude, height, and
the prognosticequationsare computed by a finite difference seasonby use of the data compiledby Hering and Borden
method. The model has nine unevenly spacedfinite difference [1964] and London [1962]. Cloud cover is assumedto be zo-
levels in the vertical where o = 0.025, 0.095, 0.205, 0.350, nally uniform and invariant with respectto time. The annual
0.515, 0.680, 0.830, 0.940, and 0.990. mean distributionof cloud cover usedfor this study is deter-
The numerical time integrationof the prognosticequations mined basedupon the studiesof London[1957] and $asamori
are conductedby a semi-implicit method in which the linear and London [1972]. The distribution of water vapor is deter-
and nonlinear componentsof the rate of changeof a variable mined from the time integrationof the prognosticequationof
are separatedand are time-integratedimplicitly and explicitly, water vapor.
respectively.To prevent the growth of fictitiouscomputational
solutions, a time-smoothingtechnique developed by Robert Heat and Water BalanceModel of the Continents
[1966] is applied at each time step.The smoothingconstanta Surfacetemperaturesover the continentsare determinedby
is chosen to be 0.02. the boundaryconditionthat no heat is storedin the soil (i.e.,
The dynamical component of the model describedabove is the net fluxes of solar and terrestrial radiation and the turbu-
developed by Gordonand Stern [1974] and is very similar to lent fluxesof sensibleand latent heat locally add to zero). For
the spectral model developed by Bourke [1974] and Hoskins the computationof the net downward flux of solar radiation,
and Simmons[1975]. The reader is referred to thesepapersfor the surface albedos are prescribedas a function of latitude
further details. The performanceof this spectralmodel of the over oceansand geographicallyover continentsbased upon
atmospheregeneral circulation is evaluated in detail by Ma- the study of Poseyand Clapp [1964]. However, thesealbedos
nabe et al. [1979b]. are replacedby higher valueswheneversnow cover or sea ice
The physical processesincorporated into the model are are simulated.
nearly identical with those used in the general circulation The albedosfor snowcover are mainly determinedby lati-
model of Holloway and Manabe [1971].A brief descriptionof tude and the snow depth. Table 1 givesthe albedo valuesfor
these processesfollows. deep snow as a function of latitude. When the snow depth is
 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5531

TABLE 1. Albedo (%) as Function of Latitude for Snow and Sea
Ice
DegreesLatitude
0ø-55 o 55 ø-66.5 o 66.5 ø-90ø
Deep snow 60 60 ---80* 80
Thick sea ice 50 50 --- 70 70
*--- meanslinear interpolationwith respectto latitude betweenthe
two values.
below a critical value equivalentto 1 cm of precipitablewater,
the albedosare reduced.Referring to the study by Kung et al.
[1964],it is assumedthat snowalbedodecreases from the val-
ues in Table I to the lower albedo of underlying soil sufaceas
a square root function of snow depth (representedin water
equivalent). Note that the albedo of the snow-free surface
over the Greenland and Antarctic ice sheetsis almost as large
as the albedofor deep snowso that there is little albedo feed-
back when the snow melts.
The changeof soil moistureis computedfrom the rates of
rainfall, evaporation,snowmelt,and runoff. A changeof snow
depth is predictedas a net contributionfrom snowfall, sub-
limation, and snowmelt, the last being determined from the
heat budgetrequirement.For further detailsof the hydrologic
computationsover the continental surface,refer to Manabe
[1969a].
Mixed Layer OceanModel
It is well known that the oceanshave a far reaching influ-
ence upon climate. Oceansare the sourceof moisturefor the
hydrologiccycle. The oceansalso have a large heat capacity
and thus reduce the amplitude of the seasonaltemperature
variation. In addition, ocean currents influence climate
through horizontal heat transport. The mixed layer model of
ocean used for this study includes the first two of these three
influences but lacks the third one, the horizontal heat trans-
port. The sensitivityof a climate model, which incorporatesa
three-dimensional ocean model with ocean currents, is the
subjectof future investigation.
In this study the mixed layer of the ocean is simplified as a
vertically isothermal layer of static water with uniform thick-
ness.Over the ice-free region, the prognosticequation for the
mixed layer temperature Tm is
ered by seaice, a highervalue of surfacealbedois assigned.
Table I givesthe albedo values for thick sea ice as a function
of latitude. When the thicknessof ice is less than 50 cm, the
albedo decreases from the values in Table I to the lower al-
bedo of underlyingwater surfaceas a squareroot function of
ice thickness. The sea ice albedo of Table 1 is further reduced
to 45% when the top surfaceof the seaice is melting;this in-
corporatesthe influenceof fresh water ponds (or puddles)
forming on top Of the ice pack.
The thicknessof the mixed layer is chosensuch that it is
equalto the effectivedepthof the seasonal
thermoclineDf
which is definedby the following equation
Df' AAT-•-O-
• = AAT-•xdz= A.•T(z)xdz (3)
whereA• T(z)x is the rangeof the seasonalvariationin zonal
mean water temperatureat depth z. (Refer to Figure 2 for the
graphicalillustrationof the effectivedepth.) Figure 3 shows
the latitudinaldistributionof D• determinedfrom this equa-
tion. The observedoceantemperaturedata, which is used for
this determination,is compiledby Levitusand Oort [1977].
For the sake of simplicity it is assumedthat the thicknessof
the mixedlayeroceanis constanteverywhereandis 68 m (i.e.,
the globalmeanvalueof D• indicatedin Figure3).
The thicknessof seaice, L in the mixed layer oceanmodel
is predictedon the basisof the followingequation
0I
--=
Ot
F- M + S•- Sb (4)
aTA(ZJ
i

OTto Q
• -- (1)
at Co' H

where Cois the heat capacity of water, and H is the thickness

of the mixed layer ocean. Q, the rate of net heat gain by the
ocean,is defined by the following equation
Q --/red- fSH- fLH (2)
where fredis the flux of net downward radiation (including
both solar and terrestrialradiation), fsH and fLH are upward
fluxes of sensibleand latent heat, respectively.The heat ex-
changebetweenthe mixed layer and the deeper layers of the
oceans is not taken into consideration.
Over the ice covered region, the temperature of the mixed
layer ocean remains at the freezing point (i.e., -2øC). Thus
the rate of its change is zero.
The albedo of the mixed layer ocean, necessaryfor the
computationof the net radiation flux at the ocean surface,is Fig. 2. Schematic
diagramillustratingthe methodfor determining
prescribedas a function of latitude. When the oceanis cov- the effectivedepth of the mixed layer ocean.
5532 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE

50

100

150 I , I • I I • I I • , I , [ I [ •
90N 60 30 0 30 60 90S

LATITUDE

Fig. 3. Latitudinal distributionof the effectivedepth of the mixed layer oceanas determinedby the methodillustratedin
Figure 2.

where Sf and So denote the rates of snowfall and sub- ments, respectively.)It is found that the period of numerical
limination (in water equivalent).M and F are ratesof melting time integration required for the model to settle down to a
and freezingof seaice, respectively.In this equationthe effectstable climatic condition is approximately 10-15 years. Since
of ice advection by wind stressand ocean currentsis not taken the straightforwardintegrationof the model consumesa large
into consideration.In the presenceof seaice, the temperature amount of computertime and is thereforevery costly,an eco-
of the mixed layer oceanis at the freezingpoint, and the heat nomical method of time integrationis developed.
conductionthrough the sea ice is balancedby the latent heat For the developmentof an economicalmethod,it is impor-
of freezing or melting at the bottom of the seaice. The freez-tant to recognizethe following characteristicsof a joint ocean-
ing and melting at the ice bottom,togetherwith the meltingat atmospheremodel. (1) The thermal inertia of the atmospheric
the ice top, sublimation,and snowfall determinesthe change part of the model is much shorter than that of the oceanic
of ice thickness.For further details,seeBryan [1969] and Ma- part. (2) The numericaltime integrationof the atmospheric
nabe [1969b]. model over a given time period usuallyconsumesmuch more
computertime than that of the oceanicmodel. With thesefac-
3. NUMERICAL EXPERIMENTS tors in mind, Manabe and Bryan [1969] developedan econom-
ical method in which a very long-term integrationof the oce-
The climatic impacts of an increasein the CO2 content of
anic part of the joint model is synchronizedwith a relatively
the atmosphereare investigatedbasedupon a comparisonbe-
tween the climate of the model with the normal concentration
short-term integration of the atmosphericpart of the model.
Thus they reducedthe disparity betweenthe two parts of the
of CO2 (i.e., 300 ppm) and another model climate with 4 times
model in approachingtoward a statisticallystationaryclimate.
the normal concentration(i.e., 1200 ppm). One could have in-
The consequence is a shorteningof the periodof atmospheric
vestigatedthe consequences of a smallerdifferencein the CO2
integration by a substantialfactor.
concentration.Instead, the CO2 content is altered by a sub-
For the presentstudy their method is modified suchthat it
stantial factor for ease of discriminatingthe CO2 induced
is applicableto the time integration of the joint model with
change from the natural fluctuation of the model climate seasonal variation of insolation. In this modified version the
[Leith, 1973].Following the normal practice,the climateswith
the normal and with 4 times the normal CO2 concentrations
time integrationof the atmosphericpart of the modelover the
period of an acceleratedseasonalcycleis synchronizedwith a
are obtainedfrom the long-termintegrationsof the model de-
full 1 year integrationof the oceanicpart of the model. At the
scribed in the preceding section. (Hereafter, these two in-
beginningof the integration,the periodof the acceleratedsea-
tegrationsare identified as the 1 x CO2 and 4 x CO2 experi-
sonalcyclefor the atmosphericmodel is chosento be 1/16 of
a year (i.e., 365/16 = 23 days).The periodincreasesin several
TABLE 2. Length and Number of SeasonalCyclesin Each Stageof stepsand becomesa full year (i.e., 365 days)toward the end of
the Numerical Time Integrations the integration.Table 2 tabulatesthe length and number of
Length of SeasonalCycle, Year the atmosphericseasonalcyclesfor each experiment.
Number of During a time integration,the quantitieswhich are trans-
Atmosphere Ocean SeasonalCycles reitted from the atmosphericto the oceanicpart of the joint
I X GO2 Experiment
model are the exchangerates of momentum, heat, water, and
•/• 1 4 ice (including snow). On the other hand, the distributionsof
•/8 1 1 sea surfacetemperatureand sea ice thicknessare transmitted
1/4 1 1 from the oceanicto the atmosphericpart of the model.For
'/2 ! 1 further details of the information exchange,refer' to Manabe
! ! 5
4 X C02 Experiment [ 1969b]and Bryan [ 1969].
1/16 1 4 The presenteconomicalmethod usedin this study'differs
1/8 1 1 from an earlier economical method, which is developed by
1/4 1 1 Manabe et al. [1979a] for the seasonaltime integration of a
•/• 1 2
1 1 6 joint model with a deep ocean. It is found that the earlier
method is not effective in acceleratingthe approach of the
 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5533

300

2g0

280

270 _ • LWTOP
2
260 _ \

250

• SW
TOP
24O

230 I I I I I I I I I I I I I
1 2 3 4 5 6 7 8 9 10 11 12 13
YEARS

Fig. 4. Time variationsof the global mean valuesof radiative fluxes at the top of the model atmosphere.Solid line
denotesnet downward solar radiation. Dashed line denotesnet upward terrestrial radiation. To remove the seasonalvaria-
tion, a 1 year running mean operator is applied to both curves.

presentmodel toward a statisticallystationaryclimate. This is to each other. This result implies that the joint model as a
becausethe heat capacity of the mixed layer ocean, which is whole is not far from radiativeequilibrium.Accordingto Fig-
incorporated in the present model, is much smaller than that ure 5, which shows the time variation of the area mean rate of
of the full ocean model. heat gain by the mixed layer ocean-seaice system,this system
The initial condition for the time integration is an isother- also is very closeto the thermal equilibrium toward the end of
mal and dry atmosphereat restoverlyingan isothermalmixed the 1 x CO2 integration. This is underscoredby Figure 6,
layer ocean.The temperatureof this isothermalatmosphere- which showsthat the area mean water temperature of the
oceansystemis chosento be 280 K. To illustratehow the joint mixed layer oceanfrom both the 1 x CO: and 4 x CO2 exper-
model used for this study attains a statisticallystationary cli- iments.This figure showsthat the mixed layer oceantemper-
mate, time variation of some of the key parametersare de- ature hardly changestoward the end of the time integrations.
picted in Figures4 and 5. Figure 4 showsthat toward the end The seasonalvariations of the model climates, which are dis-
of the 1 x CO2 integration, the annual mean values of the net cussedin the following sections,representthe mean annual
incoming solar radiation and that of the outgoing terrestrial cyclesover the last 3 year periodsof both integrations.
radiation at the top of the model atmosphereare almost equal It is importantto recognizethat the approachof the present

140

130

120

110

100

9O

8O

7O

W/M60
5O

t I \i
4O

3O -- i X w • .'•

2O
, , , , ,. ,, /, ,-, ,, ,,, ,., /
lO • ' I I ' t I I t ' ' e' J I ' I
- t , , , , , , , , , , , , ,
0--

'] ' I I I I I I t ' I t I • I I t I •

-10 • tt I I I I • I I I t I I I I I I I t I i I

-20

-30
I I I I I I I I I I I I I
1 2 3 4 5 6 7 8 9 10 ll 12 13
YEARS

Fig. 5. Time va•ation of the global monthly mean rate of net heat gain by the m•ed layer ocean-seaice system(dashed
line). Solid line is obtained by applying 1 year •ng mean operator on the dashedline.
5534 ,MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE
I I I I
296
294
292
290
oK
288
286
284
282
28o
1 2 3 4
Fig. 6. Time variationof the globalmeanwatertemperatureof the mixedlayer oceanfrom I x CO2 and4 x CO2 experi-
ments. A l-year running mean operator is applied to both curves.
atmospheremixed layer oceanmodeltoward a stableequilib-
rium climateis fasterthan the corresponding approachof the
actual joint ocean-atmosphere system,in which the heat ex-
changebetweenthe mixed layer and the deeperlayersof the
oceantakesplace.As wasdiscussed by Thompson and $chnei-
der [1979]and in the recentreportof the NationalAcademyof
Sciences[1979], this heat exchangesubstantiallydelaysthe re-
sponseof the climate to an increasein the CO• concentration
in the atmosphere.
4. SIMULATED CLIMATE
This sectionbriefly describesthe seasonalvariation of eli-
mate, which is obtained from the standard 1 x CO• experi-,
ment. Beforediscussing the comparisonbetweenthe 1 x CO•
and 4 x CO• experimentas planned in the precedingsection,
it is essentialto confirmthat the seasonalclimate variation,
which emerges from the 1 x CO2 experiment, is realistic
enoughto justify a CO2 sensitivitystudy.
Figure 7 compares the geographical distributions of
monthly mean surfaceair temperaturein February and Au-
gustwith the observeddistributionscompiledby Crutcherand
Meserve [1970] and Taljaad et al. [1969]. On the basisof this
comparison,one can identify someof the unrealisticfeatures
in the simulatedclimate.For example,the computeddistribu-
tion of the air temperature indicates an equatorial belt of
maximum temperature in the eastern Pacific. This is in dis-
agreementwith the observeddistribution which containsthe
equatorialminimum of seasurfacetemperaturecausedby the
upwelling of cold water. Also, in the regionseast of Green-
I I I I I' I I I I
4,• CO,
1 ,,CO,
5 6 7 8 9 1o 11 12 13 14
YEARS
land and off the eastcoastof the Asian continent,the simu-
lated temperatureis much too low in both Februaryand Au-
gust,probably causedby a lack of oceanicheat transportin
the mixed layer oceanmodel.This is consistentwith an over-
abundanceof seaice in theseregionsduring winter. On the
other hand, the surfaceair is too warm by severaldegrees
throughoutthe year near the Antarctic continentowing partly
to unrealistically low planetary albedo in this region of the
model. (The prescribedcloud amount turned out to be sub-
stantially smaller than the observedin this region of the
model and accountsfor this low planetary albedo.) Over the
model continents, the surface temperature is too high in
northernEuropein August,whereasit is too low therein Feb-
ruary.
Despitethe failuresof the simulationidentifiedabove,it is
very encouragingthat the modelreproduces the geographical
distribution of surface air temperature reasonablywell. By
comparingthe positionof the 275 K isothermon the com-
puted Februarymapsagainstthe observedisothermposition,
one can seethat the longitudinal variation of the isothermis
simulatedvery well. The latitudinalplacementof the isotherm
alsocomparesfavorablywith the observedposition.Sincethe
mapsof computedsurfaceair temperaturefor both February
and Augustare very similar to the observedmaps,one would
expectthat the geographicaldistributionof the seasonalvaria-
tion of the surfaceair temperaturealsowould be reproduced.
This is evident in Figure 8 which illustratesthe geographi-
cal distributionsof surfaceair temperaturedifferencebetween
Augustand Februaryfor both the modeland the actualatmo-
 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO2 INCREASE 5535

90N

60

30 290

30

60

90S
0 60E 120 180 120 60W 0
90N

60

30

300

30

280
60

90S
0 60E 120 180 120 60W 0
Fig. 7a. Geographicaldistributionsof monthlymeansurfaceair temperature(degreesKelvin) in February.Top: com-
puteddistributionfrom the I x CO2 experiment.Bottom:observeddistribution[Crutcherand Meserve,1970;Taljaad et
al., 1969].The computedsurfaceair temperaturerepresentsthe temperatureof the model atmosphereat the lowest finite
difference level located at about 70 m above the earth's surface.

sphere.It is expectedthat the differenceyieldsinformationon southernhemisphere,the amplitude of seasonalvariation over

the approximate amplitude of the seasonalvariation except in the northernhemispherecontinentsis computedto be signifi-
the equatorial region where semiannual variation pre- cantly larger in agreementwith the observedcharacteristics.
dominates.Over the oceanicregionsof the model, the ampli- Figure 9 compares the seasonal variation of zonal mean
tude of the seasonalvariation is relatively large in the neigh- surfaceair temperaturein the model atmospherewith the cor-
borhood of the continents, particularly off the east coast of responding variation in the actual atmosphere. This figure
Eurasia and North America, in excellent agreement with the clearly indicates that the model successfullysimulates not
features of the observed temperature difference. Over the only the amplitude but also the phaseof seasonaltemperature
southern hemisphere the model fails to reproduce the zonal variation at most latitudes.
belt of relatively large August-February difference around The global distributionsof the precipitation rate obtained
30øS. However, the model reproducesthe belt of large sea- from the 1 x CO2 experimentare comparedwith the observed
sonal temperature variation along the periphery of the Ant- distributionsin Figure 10. According to this figure, the tropi-
arctic Continent where the area coverage of sea ice changes, cal rainbelt over the eastern Pacific and Atlantic oceans is lo-
and accordingly,the surfaceair temperature varies markedly cated 5ø-10 ø latitude south of the observed position. This
between summer and winter. probably is due to the absenceof the equatorial minimum in
The amplitude of seasonalvariation over the model conti- the temperature distribution of the mixed layer ocean model.
nents is significantly larger than the oceanic amplitude be- In high latitudes the computed precipitation rate is much
cause of the difference in the effective heat capacitiesof the larger than the observedrate in both December-January-Feb-
oceansand continents.It is encouragingthat the model repro- ruary (DJF) and June-July-August(JJA) seasons.This dis-
ducesthe region of maximum differencein the northeastern crepancyis particularlynotablein the winter hemisphere(i.e.,
part of the Eurasiancontinentand northernCanada, although over Siberia and northern Canada in the DJF distribution and
the simulated Eurasian maximum is somewhat smaller than over Antarctica in the JJA distribution). According to recent
the observed. Since the area occupied by continents in the performance tests of a spectral climate model, the over-
northern hemisphereis larger than the continentalarea in the estimation of the high latitude precipitation and the under-
5536 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO2 INCREASE
90N
60
30
300
30
60
90S
60E 120
90N
60
30
30
60
90S
0 60E 120
Fig. 7b. Sameas Figure 7a but for August.
estimationof the tropical rainfall are two of the most serious
biasesof a low-resolutionspectralmodel. (See Manabe et al.
[1979b] for the intercomparisonamong spectralclimate mod-
els with various resolutions.)
Despite the difficultiesin the simulation identified above,
some of the grosscharacteristicsof the global distribution of
precipitationare reproducedby the model. For example,the
model indicatesthe areas of relatively low precipitation rate
near the Saharan, Australian, and central Asian deserts. In
DJF seasonthe model simulatesthe belt of large precipitation
rate extending from south of Japan to the west coastof North
America. In JJA seasonthe distribution of monsoonprecipi-
tation over southeastAsia is roughly reproduced,althoughthe
computed precipitation over northern India is too small.
In general, the global distribution of precipitation as simu-
lated by the model is far from satisfactory.Therefore, it is de-
cided to postponea detailed discussionof the CO2-induced
changeof the geographicaldistributionsof hydrologicalvari-
ables until the resultsfrom a climate model with higher spec-
tral resolution become available.
Snow coverand seaice play an important role in enhancing
the sensitivityof the climate becauseof their high reflectivity
to solar radiation. Figures 1la and I lb showthe simulated sea
ice distributionin the northern hemispherefor February and
August, which are obtained from the I x CO2 experiment.
From the February distribution, one notesthat the thickness
of the sea ice rangesbetween 1.5-2.5 m in the Arctic Ocean.
180 120 60W
280
180 120 60W o
This is comparablewith the observedannual mean thickness
from Koerner[1973].By comparingthe computedand the ob-
served sea ice margins in February on Figure 1la, one sees
that the computedmargin is locatedtoo far southboth eastof
Greenland and in the Sea of Okhotsk. These discrepancies
probablyresultfrom the absenceof horizontalheat transport
in the mixed layer oceanmodel. The simulateddistributionof
sea ice in August on Figure lib indicates that the sea ice
thicknessin the Arctic Ocean ranges between 1.0 and 1.5 m.
This seems to be much less than the observed annual mean
value of Koerner[ 1973].However, the observedseaice margin
in summeris approximatelyreproducedby the model.
In the southernhemispherethe model underestimatesthe
area coverageof the seaice in both Augustand February. Ac-
cordingto Figure 1ld the observedice margin in August is
simulated very well from 80øE-180ø-80øW longitude along
the peripheryof Antarctica. However, the model grosslyun-
derestimatesthe seaice coveragealong the rest of the Antarc-
tic coast.Figure 1l c indicatesthat in February the simulated
sea ice margin is confinedto a few coastalregions,while the
observed ice is extensive in the Ross and Weddell seas. The
generalunderpredictionof the seaice extentaroundthe Ant-
arctic continent is consistent with the overestimation of the
surfaceair temperatureover this region which is described
earlier.
The distribution of snow depth (in centimetersof water
equivalent)in the northernhemisphereof the modelis shown
 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5537

90N

60

30

30

60

90S
0 60E 120 180 120 60W 0
90N

60

30

30

60

90S
0 60E 120 180 120 60W 0
Fig. 8. Geographicaldistributionof surfaceair temperaturedifference(degreesKelvin) betweenAugustand February.
Top: computeddistributionfrom the I x CO2 experiment.Bottom:observeddistribution(from the data compiledby Crut-
cherand Meserve[1970]and Taljaad et al. [1969]).Note that the contourinterval is 2øK when the absolutevalue of the
difference is less than 10øK and is 10øK when it is more than 10øK.

in Figure 12 for February and August.For the sake of com- continent system,Figure 13 is constructed.This figure illus-
parisonthe observedmonthly mean marginsof snowcover as trates the seasonalvariation of hemisphericand global mean
determinedby Wiesnetand Matson [1979]are indicatedin the values of the solar, terrestrial, and net radiative fluxes at the
figure by dashed-dottedlines. Accordingto this comparison top of the model atmosphere.For comparison,the values of
the largest discrepancybetween the observedand the com- the radiative fluxesdeducedfrom satelliteobservationby Ellis
putedsnowmarginsin Februaryoccursover the southeastern and VonderHaar [1976] are indicated in the samefigure. Ac-
part of the Eurasiancontinent.Althoughthe snowmarginin cording to this comparison, the agreement between the ob-
this region has a large interannual variation [Wiesnet and servedand the computed fluxesis good, though both the solar
Matson, 1979], it appearssignificantthat the snow cover ex- and the terrestrialfluxesof the modelare slightlylargerthan
tendsmuch too far equatorwardin this area. It is encouraging the observedfluxes.The agreementis particularlynoteworthy
that over the North American continent, the observed snow if one recalls that the cloud cover distribution in the model at-
margin is simulatedsatisfactorilyby the model. mospherehas no seasonalvariation. The successof the pres-
By August,almostall the snowcoverover the continentsin ent model in simulating the seasonalvariation of the hemi-
the northern hemisphereof the model disappearwith the ex- spheric mean radiation budget suggeststhat the temporal
ception of Greenland. However, snow remains in the Hi- variation of cloud cover has only a secondaryrole in deter-
malyas and the coastalAlaskan mountainsin the observed mining the hemisphericmean heat balance. (Similar results
are obtained by Manabe et al. [1979a] from their joint ocean
Augustdistribution.It is speculatedthat the spectraltrunca-
tion of high wave numbercomponentsfrom the field of orog- atmosphere model with annual mean cloud cover.) One
raphyresultsin the loweringof the heightsof the mountains shouldremember,however,that there alwaysis a possibility
and, accordingly,the underestimationof the snow covered that model defectsmay compensatewith each other to yield a
area. realistic result.
To evaluatethe performanceof the model in reproducing In short, it is shown that despite the failures identified
the overall radiation budget of the joint atmosphere-ocean- above, the model roughly reproducesseasonalvariations of
5538 MANABE AND STOUFFER.'GLOBAL CLIMATE MODEL OF CO 2 INCREASE

90øN

60 ø

30 ø

30 ø

,90øS
Jan Apr Jul Oct Jan Jan Apr Jul Oct Jan
Fig. 9. Latitude-timedistributionsof deviationsin the zonal mean surfaceair temperaturefrom the am•ua!mean value
(degreesKelvin). Left: computeddistributionfrom the 1 x CO2 experiment.Right: observeddistribution(obtainedfrom
the data compiledby Crutcherand Meserve[1970]; Taljaad et al. [1969]).

surfaceair temperature,the marginsof snowand sea ice, and spheretoward the moistadiabaticlapserate. Sincethe moist
the hemisphericmean radiation budget. This encouragedthe adiabaticlapserate reduceswith increasingair temperature,it
authorsto use this model for the presentsensitivitystudy. is expectedthat the CO:-inducedwarming in the upper tropo-
sphere is larger than the warming near the earth's surface.
5. THERMAL RESPONSE
Thus moist convectionis partly responsiblefor the smallness
This sectioncontainsthe discussionof temperaturechanges of the warming near the earth's surfacein low latitudes.
in the model atmospherewhich occur in responseto the quad- Figure 14 alsorevealsthat the high latitudewarmingof the
rupling of CO2 contentin the air. As was explainedin section northern hemisphereis significantlylarger than that of the
3, the model responseis computed as a differencebetween the southern hemisphere.One of the important factors respon-
4 x CO: and I x CO: atmospheres. sible for this interhemisphereicdifference in the warming is
the difference in the reflected solar radiation between the 4 x
Annual Mean Response CO: and 1 x CO: experiment.Accordingto Figure 15, which
The latitude-height distribution of the difference of zonal shows the latitudinal distribution of difference in the net in-
mean temperature between the 4 x CO: and 1 x CO: atmo- coming solar radiation at the top of the model atmosphere,
sphereis shown in Figure 14. This figure indicatesa general the difference in the northern hemisphereincreaseswith in-
warming of the model troposphereand a coolingof the strato- creasinglatitude polewardsof 30øN, whereasin the southern
sphere resulting from the quadrupling of the CO2 content. hemispherethe differencehas local maximum along the pe-
The warming is particularly pronouncedin the lowest layers riphery of the Antarctic continent,but it is smaller over the
in high altitudesand is relatively small in the tropics,resulting continent.As a whole, the northern hemispherechangeof the
in a general reductionof the meridional temperaturegradient net incoming solar radiation (or planetary albedo) which is
in the lower model troposphere.As was discussedby Manabe causedby the poleward retreat of highly reflectivesnow cover
and Wetheraid [1975], this reduction of meridional temper- and sea ice is significantly larger than the corresponding
ature gradient resultsfrom (1) the poleward retreat of highly changein the southernhemisphere.Since the surfacealbedo
reflective snow cover and sea ice and (2) the increasein the of continental ice sheet is assumedto be almost as large as
poleward latent heat transport. that of snow cover, the disappearanceof snow cover over the
In lower latitudes the magnitude of the warming near the Antarctic continentdoesnot resultin the significantreduction
earth's surface is further reduced by the effect of moist con- of surfacealbedo and the further warming of overlying atmo-
vection which tends to adjust the static stability of the atmo- sphere.(Note alsothat partly becauseof the high-surfaceele-
 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5539

COMPUTED (DEC,JAN, FEB)

90N 90N

30 30

30 30

60

90S 905
o 30 60 90 120 150 180 150 120 90 60 30 0

OBSERVED (DEC,JAN, FEB)

90N 90N
....... :::: ....................
..

::::::::::::::::
60 60
::::::::::::::::::::::::::::::::::::::: .......
:::::::::::::::::::::::::::::::::

30 30

::iii:::::::: •

30 :::::::::::::::::
30

60 60

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

..................................................... . .......... .......,,o,,,,,,oooo,,,,,,,o,,,.,..,,,o.,,...,...oo..o.o ....................................................
........................................................... .......................,.............,...,..,,...,,..,............. ..... ...... ...................................
...................................................... ......................... .... ................,.....,.,,o,,o,..,,,,.,.o ..... ...... .....................................

90S 90S
0 30 60 90 120 150 180 150 120 90 60 30 0

Fig. 10a. Geographicaldistributionof December-January-February

mean precipitationrate (centimetersper day).
Top: computed distribution from the 1 x CO2 experiment. Bottom: observeddistribution [M611er,1951]. Slanted shade
indicatesthe area where precipitationrate exceeds0.5 cm/day, whereasdotted shadeidentifiesthe area where it is below
0.1 cm/day.

vation, the surfacetemperature over most of the Antarctic ice There is a possibilitythat the interhemispheric
differencein
sheetis usually too low for the overlyingsnowto melt because temperaturechangediscussedabove is exaggeratedby the
of the warming). Thus the changeof net incomingsolar radia- model. Around the Antarcticcontinentin the 1 x CO: model,
tion is relatively small over the Antarctic continentdespitethe the temperatureof the mixedlayer oceanis too high and the
general warming caused by the CO2 increase.In short, the area of sea ice coverageis too narrow as describedin section
contribution of the snow albedo feedback mechanism is rela- 4. Therefore the contribution of the sea ice albedo feedback
tively small over the Antarctic continent. This accounts for mechanismover the Circum-AntarcticOcean may be less
the interhemisphericdifferencein warming describedabove. than the actual contribution.
In addition, the surface temperature of a continental ice Owingto the interhemispheric
differencein the temperature
sheetcan not exceedthe freezingtemperature(i.e., 0øC) ow- change in high latitudes discussedabove, the area mean
ing to the ice melting. This effectalso actsto reducethe mag- warmingin the northernhemisphereis 4.5øC and is signifi-
nitude of the warming over the outer edgeof the Antarctic ice cantly larger than the area mean warming of 3.6øC in the
sheet and contributes to the interhemisphericdifference in southernhemisphere.The global mean differenceof surface
warming. air temperature between the two experimentsis 4.1øC, im-
5540 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE

COMPUTED (JUN,JUL,AUG)
90N .................... I i i•iiiii!iiji!!!iiiiiiii!il!!iiiiiii!:::ij...]q:•'/•• i 90N
••::Vj•::.]•
........... ':-':••...:j:7•.1 ......:::!!::i!iiii:::
.......

6O 6O

:' '::: .... 1 .1.:: ..........

30 ::::" . ':iiii•i!iii!iiiii•i: ........... ::::::::::::::::: .:::::::::::::' :::-:-:30
ß' "11iiiii::'::::!iii: iiiiiiiii!iiiiii::
ß .1 :::'" .2 "::::::::::::::: ..:"

0 0

I
30 30

60 60

90S•.....•,...•,....__••
....:::....-"•'
0
......:::::::::::::::::::::::::::::
30
I • "•
60 90 120 150
I
180
I
150
I ..• I 120 90 60
•'•-'•-•-90S
30 0

OBSERVED (JUN,JUL,AUG)
90N 90N

..... :::::::::::::::::: ............... ß........ , ........

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

6O 60

3O 3O

3O 3O

6O 6O

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

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: .... ::::::::::::::::::::: ......................................................................................
90S 90S
0 30 60 90 120 150 180 150 120 90 60 30 0

Fig. 10b. Same as Figure 9a, but for June-July-August.

plying that the warmingcausedby the doublingof CO2 con- SeasonalResponse

tent would be approximatelyhalf as large (i.e., 2øC). (Refer,
for example,to Augustsson and Ramanathan[1977]for the jus-
tiffcation of this inference.)These valuesare significantlyless
than the warming obtained from the general circulation
model of Manabe and Wetheraid[1980] with idealized geogra-
phy and without seasonalvariation of insolation.(They ob-
tained the mean warming of 3øC and 6øC in responseto the
doubling and the quadrupling of CO2 concentration,respec-
tively.) The smallersensitivityof the presentmodel partly re-
suits from the absence of the albedo feedback over the Ant-
arctic continent. In addition, the absenceof the snow albedo
feedback mechanismsover the northern hemisphere conti-
nents in summer can contribute
tivity of a model climate. This topic is discussedagain toward
the end of this section.
Figure 16 showsthe seasonalvariationof the differenceof
zonal mean surfaceair temperaturebetweenthe 4 x CO2 and
1 x CO2 atmosphere.In this figure,three latitude-timedistri-
butions of zonal mean differencesover (Figure 16a) oceans
and continents,(16b) oceans,and (16c) continentsare illus-
trated separately.Figure 16a showsthat in low latitudesthe
warmingowing to the quadruplingof CO2 concentrationis
relativelysmalland dependslittle upon season.However,in
high latitudesit is generallylargerand variesmarkedlywith
season,particularlyin the northernhemisphere.The warming
is at a maximumin early winter and is smallin summer.This
to the reduction of the sensi- impliesthat the rangeof seasonal variationof surfaceair tem-
peraturein theseregionsreducessignificantlyin responseto
the quadruplingof the CO2 contentsof the air. From the sea-
 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5541

eeee

,18C
0

NORTH AMERICA

AUGUST

(c) 9(•øE (cm) 9•øE

AUSTRALIA
AFRICA

eeeeeeeee
ß ß

NEW NEW
ANTARCTICA
ZEALAND ZEALAND

.0 o •)o.
60 ø 60 ø

eeeeeeeeee

SOUTH AMERICA
SOUTH
AMERICA FEBRUARY )øW
AUGUST
/ • 90,•W
Fig. 11. Geographicaldistributionsof seaice thickness(meters)from the I x CO2 experiment.(a) Northern hemi-
sphere-February. (b) Northern hemisphere--August.(c) Southernhemisphere--February.(d) Southernhemisphere--
August.Dashed-dottedlinesindicatethe observedmean seaice boundariesfrom the compilationby U.S. Naval Oceano-
graphic Office [ 1957, 1958].

sonal variations of the zonal mean sea ice thickness shown in
Figure 17, it is observedthat the sea ice from the 4 x CO,• ex-
perimentis everywherelessthan the seaice from the 1 x CO,•
experiment.Therefore it is reasonablethat the 1 x CO,• atmo-
sphereis insulatedby thicker seaice from the influenceof the
underlying sea water so that it has a more continental climate
with a larger seasonalvariation of temperature than the 4 x
CO,• atmosphere.Although the poleward retreat of highly re-
flective snow cover and sea ice is mainly responsiblefor the
large annual mean warming in high latitudes, the change of
the thermal insulation effect of sea ice stronglyinfluencesthe
seasonalvariation of the warming over the polar regionsas is
discussed below.
Figure 18 showsthe seasonalvariation of the zonal mean
difference in surface water temperature of the mixed layer
ocean model between the 4 x CO,• and I x CO,• experiment.
According to the comparison between this figure and Figure
16b,the seasonalvariation of the CO,• warming of the mixed
layer ocean in high latitudes differs from and is much less
than the variation of the warming of overlying air. This result
suggeststhat the magnitude of the thermal insulation effect of
sea ice is highly variable with respectto season,and it de-
serves further examination.
To comprehend the seasonalvariation of air temperature
over the Arctic Ocean, it is necessaryto analyze the heat bal-
ance at the underlying ocean surface. Figure 19a showshow
the zonal mean differences of various surface heat balance
components(at 82øN) between the two experimentschanges
with respectto seasons.The surfaceheat balance components
contained in this analysis are net solar radiation, net terres-
trial radiation, sensible heat flux, latent heat flux, and oceanic
heat flux. In Figure 19 a positive value indicatesa heat gain
for the ocean surface which results from the CO,• increase.
Therefore the increasesin the upward fluxes of sensibleheat,
latent heat, and net terrestrial radiation indicate negative con-
tributions, whereasthe increasesin the upward oceanic heat
flux and net downward solar radiation are shownto be posi-
tive contributionsto the surfaceheat balance. (Here the oce-
anic heat flux includesthe heat exchangeswith not only the
mixed layer ocean but also the sea ice. Over the ice-free re-
gion, the oceanicheat flux is the heat supplyfrom the interior
to the surfaceof the ocean. Over the ice-free region, the oce-
5542 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO2 INCREASE

0.25 o

o
o

eøøw FEBRUARY eøøw AUGUST

Fig. 12. Geographicaldistributionsof snowdepth in water equivalentunits (centimeters)from the I x CO2 experi-
ment. Left: February distribution.Right: Augustdistribution.Dashed-dottedline indicatesobservedmonthly mean snow
margin from Wiesnetand Matson [1979].

anic heat flux is the heat supply from the interior to the sur- oceanic heat flux around November. (As defined earlier, oce-
face of the ocean. Over the ice coveredregion it is the upward anic heat flux includes heat conduction through the sea ice.)
conductiveheat flux through ice minus the rate of latent heat This changein the oceanicheat flux is accompaniedby a large
consumptionowing to the melting of the upper ice surface.) early winter increasein the sensibleand latent heat fluxesinto
Figure 19a indicatesthat in summer,the absorptionof solar the atmosphere.(Note, in Figure 19a, a large negativediffer-
radiation by the Arctic Ocean surfacein the 4 x CO2 experi- ence in the sensibleheat flux around November.) Thus the
ment is much larger than the correspondingabsorptionin the CO2-inducedwarming in the surfacelayer of the atmosphere
1 x CO2 experiment. This differenceresultsfrom the reduc- is most pronouncedin early winter.
tion of surface albedo causedby the disappearanceof sea ice As the winter seasonproceeds,sea ice in both the 1 x CO2
or the formation of water puddles over the sea ice. The figure and 4 x CO2 experimentsbecomessufficientlythick that the
also indicates the net gain of terrestrial radiation energy by differencein upward conductiveheat flux becomesrelatively
the Arctic Ocean surface during summer in the 4 x CO2 ex- small. Thus the magnitude of the CO2-induced warming of
periment in comparisonwith the 1 x CO2 experiment. This surfaceair (through sensibleheat flux) in late winter is less
additional energy of both solar and terrestrial radiation is than that of early winter. Meanwhile, the warming of the
transferred directly into downward oceanicheat flux as is in- mixed layer temperatureis very closeto zero as the temper-
dicated by Figure 19a, and it enhancesthe melting of sea ice ature of water beneath sea ice is assumed to remain at the
and increasesthe warming of mixed layer ocean as revealed freezing point (i.e., -2øC).
by Figure 18. Accordingly, the additional radiation energy Accordingto Figure 16b the time of maximum warming of
does not contribute to the warming of the model atmosphere. the surface layer of the atmospherebecomeslater with de-
(Instead, downward fluxesof sensibleand latent heat further creasinglatitude. (For example,at 70øN, the warming is at a
removes the heat from the atmosphere as implied by small maximum in January rather than early winter when maxi-
gain of sensibleand latent heat by the oceansurfacein sum- mum warming occursnear the North Pole.) This resultis con-
mer.) Thus the CO2 warming of the surfacelayer of the atmo- sistent with the fact that the time of sea ice formation in both
sphereover the Arctic Ocean is relatively small as Figure 16b experimentsbecomeslater with decreasinglatitudes as in-
indicates. However, it is expected,that the additional heat re- dicated in Figure 17.
ceived by the mixed layer sea ice systemdelaysthe formation Figure 20 illustratesthe latitude-timedistributionof zonal
of the sea ice or reduces its thickness. This results in the re- mean differencein the upward flux of sensibleheat between
duction of the thermal insulation effect of the sea ice in early the 4 x CO2 and 1 x CO2 experiments.(Note that in this fig-
winter, when the air-sea temperature differencebecomeslarge ure, the sign of sensibleheat flux differenceis reversedfrom
and enhancesthe warming of the surfaceatmosphericlayer. Figure 19a).Again, this figureindicatesthe reductionof up-
This reduction in the thermal insulation effect of the sea ice ward sensible heat flux in summer and the increase in the late
is indicated in Figure 19a as a large positivedifferencein the fall and winter in responseto the increaseof CO2 concentra-
 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE 5543

1oo I I
75•,. (a)NET NORTHERN HEMISPHERE :

50

25

-25

-50

-75

-lOO
350
I I [ [ [ [ ] •.••1........
325' • (b)SOLAR NORTHERN
HEMISPHERE?" + ._:.

3OO _
xxr .-"'
+
\ / -
275
ß

% ..-'
\\ • •\ :.f+ -
I \ .
250 ß I ..'
2 ß '\ ø ø
225 --

/ • GLOBAL ...' ]q
200 -- / -p:ß / XX --

150 - • •...
....
' ....... .•-'
+
• •sP• • •

•25, i I , I i I I I I , I I I
2751 I I I I
I •c) TERRESTRIAL
I I
__
I
_ •o••
'1 I •sP•g•l
I I I
250F................... x •• ....,
•2 225• • '"•-.•,• x +x •
+ .• .........
• •'--•
2oo• • I I I I I I I I I I /
3 F M A M 3 3 A S O N D •
MONTH
Fig. 13. Seasonalvariation of hemisphericand global mean valuesof radiative fluxesat the top of the atmosphere.
Top: net radiativeflux (netdownwardsolarflux minusupwardterrestrialflux);middle:net downwardsolarflux (incoming
solarflux minusreflectedradiation);bottom:upwardterrestrialflux. Units areW m-2. Observedfluxesare deducedfrom
satelliteobservationsby Ellis and VonderHaar [1976].They are indicatedby crosses
(northernhemisphere),pluses(south-
ern hemisphere),and dots (global domain).

• ••."•
..... :::::-.iiiii::::iiii:i7.-3
"•
,.....
:0
...... ::::
• • ii!ii!i!ii:::::::::::..::
:•-'"••-• •____, :-"-'•
...... --:.:•.::i::
• 205 n-

if)
"' 3so :• ..........
•....... ::........:...........:..........:.....••
.-.....
:.:+:..'•
::::::::::::::::::::::::::::::::::::::::::::::::::::::
iii x
m 515 5
940.1¾
.'•::•.'
'"
'•'"•-.N:"
'o •.... ' "'-''"
' '"' .."
'.".."..'
:l •'• • ==========================
•:.:.','
990 o o o o o o o
90 N 60 30 0 30 60 90 S

Fig. 14. Latitudeheightdistributionof the zonal mean differencein annualmean temperature(degreesKelvin) of the
model atmospherebetweenthe 4 x CO2 and I x CO2 experiment.
5544 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE

•2

0 I I I I I
90øN 60ø 30ø 0 30ø 60ø 90øS
LATITUDE
Fig. 15. Latitudinaldistribution
of thezonalmeandifference
in theannualmeannetincomingsolarradiationat thetop
of the modelatmosphere
betweenthe 4 x CO2 and I x CO2 experiment.

tion. It is significantthat the time of maximum increasebe- (a) OCEANS

ANDCONTINENTS
90N • , • • •. • , ' .....
comes later with decreasinglatitude and accounts for the
,,/'.J; ", il
phasevariationin the latitude-timedistributionof air temper- '1/,:/"'',-,..',\ 2J,'\
ature difference as discussed above.
• / I ', • .-' ) '"'-•,,]o.L-'
The seasonal variation of the difference in the surface air 60 :......... •- 6 '"-' " •6
temperature between the two experimentsover the model
continentsis significantlydifferentfrom the variation over the
model oceans.Accordingto Figure 16c,which showsthe lati-
tude-time distribution of the difference in zonal mean surface -...... • ......... .. 3__- .-•... s•-
air temperatureover continents,the CO2 warmingin high lati-
tudes is at a maximum in early winter is being influenced by
the large warming over the Arctic Ocean discussedabove.
However, the figure also indicatesa secondarycenter of rela- 30
tively large warming around 65øN in April. To determinethe
mechanismfor this springmaximum, the seasonalvariation of
zonal mean differences of the continental heat balance com-
60
ponents(at 65øN) betweenthe two experimentsare illustrated 4 '"- 6 '• 4
in Figure 19b in the samemanner as Figure 19a. This figure
indicatesthe largespringmaximumin the differenceof net 90S
downward solar radiation betweenthe two experiments.Since J F M A M J J A S 0 N D J
the incomingsolar radiation of late springis large in high lati-
tudes,the reduction of highly reflectivesnow coveredarea in (b) OCEANS
springcausesa large increasein the amountof solarradiation
absorbedby the continentalsurface.This additional solar en-
ergy immediatelyraisesthe temperatureof continentalsurface
and is responsiblefor the increasein the upward fluxes of
both sensibleand latent heat into the model atmosphereas in-
6o
dicatedin Figure 19bby the large negativedifferencesof these
fluxesaround May. (Note that the heat capacityof continen- 30-

tal surface is assumedto be zero.) Thus the aforementioned

springmaximufia
in warmingoccurs.
.

o
It is important to recognizehere that the influenceof the al-
bedo feedback mechanismupon the change in net incoming
solar radiation is pronouncedin the spring or early summer
over the oceansas well as over the continents.This is evident 30
in Figure 21, which showsthe latitude-time distribution of the
zonal mean differencein net incoming solar radiation at the
topoftheatmosphere
between
the4 x CO•and1 x CO•ex- 60
periments.The zonal mean differenceis shownseparatelyfor
continental and oceanic regions. According to this figure the 9OS
increasein the net incoming solar radiation over continentsis J F M A M J J A S O N D J
mostpronouncedin March at around 50øN, in May at around
70øN, and in June near the North Pole. Over oceans the in- Fig. 16. Latitude-time distribution of zonal mean differencein
surfaceair temperature(degreesKelvin) betweenthe 4 x CO2 and 1
creaseis most pronouncedin the beginning of May at about x CO2 experiment.(a) Oceansand continents,(b) oceans,and (c)
70øN and in August near the North Pole. Since the high lati- continents.The differenceis computedat the lowestfinite difference
tude insolation has pronounced seasonal variation and level located at about 70 m above the earth's surface.
 MANABE AND STOUFFER.'GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5545

90N
(c) CONTINENTS
90N

60
6o

30

30 30

6O 60

90S 90S •-
J F M A M J J A S 0 N D J J F M A M J J A S 0 N D J
Fig. 16. (continued) Fig. 18. The latitude-time distribution of the differencein zonal
mean surfacewater temperature (degree Celsius) of the mixed layer
reachesits maximum between the spring and the summer, a model ocean between the 4 x CO2 and I x CO2 experiments.
changein the surfacealbedo becauseof the poleward retreat
of snow cover and sea ice results in the particularly large the growthof seaice accountsfor thislargewinterwarmingof
change in the net incoming solar radiation in these seasons. the atmosphere.
The large increaseof the absorbedsolar energy in late spring It is of interest that the continental difference in the net in-
or early summer results in the CO:- fall maximum in the coming solar radiation betweenthe two experimentsis very
warming of the mixed layer oceanwhich, in turn, is followed small around July. This is indicative of the absenceof snow
by the maximum warmingof the surfaceatmosphericlayer in cover over the continentsduring the summer in both 1 x CO•
early winter (i.e., November at the North Pole and January at and 4 x CO: experiments.The smallnessor absenceof a con-
70øN). As was discussedalready, the CO2-induced delay in tribution from the snow albedo feedback mechanism over the
northern hemispherecontinentsin summeris partly respon-
(a) lxCO 2 sible for the relatively small sensitivityof the presentseasonal
model which is discussedearlier. This assertionis supported
by a forthcomingstudyof R. T. Wetheraidand S. Manabe (in
preparation,1980) which comparesthe sensitivityof a sea-
u r.......... "•••'":•::'•'.'••.•:".":•ii:.::.•ii!:•::::::
!i:•
......
sonal model with that of a model with annual mean in-
solation.As waspointed out before,the small differencein the
net solar radiation over the Antarctic continent is another fac-
tor reducingthe sensitivityof the presentmodel.
45ø
f Over the oceanic region surroundingthe Antarctic conti-
60ø[- nent, the differencein solar radiation betweenthe two experi-
ments varies with seasonsin a manner which is qualitatively
90øS
similar to the correspondingvariation in the northern hemi-
J F M A M J J A S O N D J spheredescribedabove.However,the differencehasa signifi-
(b) 4xCO2 cant magnitudein a narrow zonal belt around the continent
90øN -,r-.••-.•-..•.••.•..••••••••••:.....•::•::..:.•.•.•,=.....•..
........•••:•:.:..:.:.:.:;.s..:•.::
•......• , , ' """"•"
'•.
•"•'
'.'•':'
:•••.'.'.1
':'
ß' ßß.--ßß' 0 where the seasonalvariation of sea ice takes place. These re-
"-'"'- '" ............. sults are consistent with the latitudinal distribution of the an-
nual mean difference in the solar radiation between the two
experimentsshownin Figure 15 and discussed
in the preced-
60ø
• ......................
.":'
......
':'•.'
"':.'-";::
•............ ing subsection.
It is reasonable that the seasonal variation of the CO,•
warming of the surfacelayer of the model atmosphereover
45ø
f the Circum-AntarcticOceanis qualitativelysimilar to the cor-
60ø
I respondingvariationover the Arctic Oceanas Figure 16bin-
dicates.Again, the regionof large winter warmingis limited
90os
to a narrow latitude belt around the Antarctic continent.
J F M A M J J A S O N D J Up to thispoint the discussion
hasbeendevotedto the dif-
Fig. 17. Latitude-timedistributionsof seaice thickness(centime-
ference in surface air temperaturebetween the two experi-
ters). Top: I x CO2 experiment.Bottom:4 x CO2 experiment.Shad- ments. To examine the temperature differencein the higher
ing indicatesthe regionswhere seaice thicknessexceeds0.1 m. levels of the model atmosphere,the latitude-heightdistribu-
5546 MANABEAND STOUFFER.'
GLOBALCLIMATEMODEL OF CO2 INCREASE

90øN
80
j I 1(82N
70
)I I
(a) Ocean I I I I I I
........... SENSIBLE HEAT FLUX
..... LATENT HEAT FLUX
60 ...... NETTERRESTRIAL
RADIATION 60 ø
.... NET SOLAR RADIATION
50 OCEANHEATFLUX
I
I 30 ø
40 I
I
I
30 I
I
I o
20 I
lO

w 30 ø

60 ø-
-20 ß
ß

-30
90øS

-40 \'""...../ JAN APR JUL OCT JAN

-50
Fig. 20. Latitude-timedistributionof zonalmeandifferencein the
upwardflux of sensibleheat(W/m 2) overmodeloceansbetweenthe4
-60 x CO2 and I x CO2 experiment.

-70

-80 I I I i i I
this sectionmay be comparedwith the resultsfrom somere-
50 I • i ! I i I i i I cent studieswhich use simpleclimatemodels.For example,
(b) Continents (65N) •" SNOW MELT
Rarnanathanet al. [1979] investigatedthe responseof a sea-
sonalenergybalanceclimatemodel to an increaseof the CO•
3o- I
/ ',% - concentrationin the atmosphere.Accordingto their results,
I the CO• warmingof the surfacetemperaturein high latitudes
20- I
/ \ - of their modelis at a maximumin earlysummer.Held [1977]
W In - / examined the responseof his energy balance model to a
I• .,./ _ %..,"'"
..... ',,, •,• changein the solar constantand found that the surfacewarm-
I "--• / • •--•'• --
ing is at a maximumin the late summerand fall. (Note that
%... both energy balancemodelscarry surfacetemperatureas a
basicvariable and do not discriminatethe temperatureof air

-301 I I I
,'-.,,
'•.•/./'
I' I • I I I I /
from that of underlyingsurface.)Sincetheseenergybalance
modelsincorporatethe thermal inertia of the mixed layer
J F M A M J J A S O N D
ocean, there should be a time lag between the maximum
TIME (month) springsolarheating(causedby the albedofeedback)and the
Fig. 19. Annual marchesof the zonal mean differencesin various maximumsurfacetemperaturedifference.In the experiment
heat balancecomponentsat the earth'ssurfacebetweenthe 4 x CO2 of Ramanathan et al. which has an early summer maximum
and I x CO2 experiments.
(a) Oceansurfaceat 82øN,(b) continental temperaturedifference,the time lag is short becausethe ther-
surface at 65øN. Positive difference indicate an additional flux to-
wards the surface. Units are in W m -•. mal inertia in high latitudes of their model is assumedto be
small.On the otherhand, Held'smodelexhibitsa time lag of
approximately3 monthsbecauseof its larger thermal inertia.
According to a comparisonbetween the results of Held and
tion of the zonal mean temperaturedifferencebetweenthe 4 thoseof presentstudy,the time of maximumsurfacetemper-
x CO• and 1 x CO• atmospheres is shownin Figure22. This ature differencein high latitudesof Held's model approxi-
figure indicatesthat the CO•-inducedwarmingin winter is matelycoincideswith the time of maximumCO• warmingof
limited to the lowestatmospherelayer in high latitudesbe- the mixed layer ocean,but precedessignificantlythe time of
cause,asManabeand Wetheraid[1975]suggested,
stablestratifica- maximum warming of the atmosphericsurfacelayer of the
.tiontendsto preventthe penetrationof heat into the upper presentmodel.This comparisonsuggests that it is necessary
to
modeltroposphere. On the otherhand,the warmingtendsto treat the temperaturesof the atmosphereand mixed layer
spread over the entire model tropospherein lower latitudes oceanseparatelyto simulatethe winter maximumin warming
where moist convection predominates.In summer, similar as influencedby the thermal insulatingeffect of seaice.
spreadingof warmingtakesplace at mostlatitudesexcept It is of interestthat the olderresultsfrom an energybalance
over the very narrow region around the North Pole where model of Sellers[1974] indicatesthe winter maximum of sen-
slightcoolingappearsin the surfacelayer.The influenceof sitivitywhichisin qualitative
agreement
withtheresults
from
stablestratification
mentionedaboveis an additionalfactor thepresent study.However, hiswintermaximumappears to
which enhancesthe high latitudewarmingin winter. resultfrom a somewhatdifferentmechanismwhich has not
The seasonal
variationof the CO2warmingdescribed
in beenfully identifiedby thepresentauthors.
 MANABE AND STOUFFER.'GLOBAL CLIMATE MODEL OF CO 2 INCREASE 5547

(a)CONTINENT (b) OCEAN

90øN

• • •s'"'-' s• so

50

90ø$ .
JAN APR JUL OCT JAN JAN APR JUL OCT JAN

Fig. 21. Latitude-timedistributionof zonalmeandifferencein net incomingsolarradiation(W/m 2) at the top of the

model atmospherebetweenthe 4 x CO2 and 1 x CO2 experiment.(a) Continents,(b) oceans.

GeographicalResponse The difference in the surface air temperature for the De-
cember-January-Februaryperiod is shownin Figure 23b, and
The geographicaldistribution of the difference in annual it revealssignificantlongitudinal variations.Accordingto this
mean surface air temperature between the 4 x CO2 and I x figure the CO:-induced warming is particularly large along
CO2 atmospheres is shownin Figure 23a. This figureindicates the east coasts of both the Eurasian and North American con-
that the distribution of the CO2-inducedchangein the annual tinents. The intenseArctic warming discussedin the preced-
mean surfaceair temperatureis highly zonal and reveals the ing subsectionresultsin the warming of air flowing southward
characteristicswhich are identified in the previous subsection along the periphery of both the Aleutian and Icelandic lows,
on the zonal mean response(i.e., large polar warming, rela- causingthe relatively large increasein the surfaceair temper-
tively small warming in the tropics,and the interhemispheric ature along the east coast of the Eurasian and North Ameri-
asymmetryin warming betweentwo polar regions). can continents.This Eurasian coastal warming is further en-

25 25

..-. '•,-6 ............................... 20

E 9,5 ............................ :::'•;:::::•................ E

u.J 1 15 "-'
n- 2 I--
• 205 -r-
u') lO c9
cu 350

515 ...:.:.:.:
.......... 5
660
830
T

90øN 60 ø 30 ø 0ø 30 ø 60 ø 90øS
25 25

•.• . 20
• 1,5
"-'
g

O0 U :""":"2:':':':':':':':':':':':':':':':':':':':':':':':':':':" '"':':':':':':':':':':':':':':':':':
',,••":':g':•'•
:.:.:.' ..:.:.:.'.'.....:.:-'

940 ':':':':':':':':':':':':':"
99•90øN 60'":':':'"":':':':':'
ø ":':':':':"
30ø 0•' 30o 60o 90oS
Fig. 22. Latitudeheightdistributionsof differencein zonal mean temperature(degreesKelvin) betweenthe 4 x CO2 and
1 x CO2 model atmospheres.Top: (December-January-February)difference.Bottom: (June-July-August)difference.
5548 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF C02 INCREASE

90N
I•(a)
ANN::•..%-r.•'"'"'••:':••.•:
.......
:,!,•i..•i:.•.'•,:•:.a:,::'•:As.7::::::::::::::::::::::::
:i:5
.......
:.::.:...::..
•:!:•.....'"-..:.'..':.:::..'/;::
60
.'...-•
"'•;....' '":':':'"
'""••ii•iii:.':.::'/.'.
• •-•'-' •-'"'•i! i''"""
-'.......'
."• "..•'"-'-'-'••

30 -'•. ' : • ' -i - - •- -•' .... '" -'"'""

""•
' .........

....,...
'"' , ..
"-,-".,.,,..."-'._
30
..:: _
, ..- . ;' :: -- •.,• ,, •_• ---•
[.....-- .•g.:.• ---. •.•..&.:.:...• .... •
[---'.:_...'::• '"'"'-• ß':;:;:::;:
,• •',.-,; ,•"'""*:•.":'":•:••••'.•....:::••
7 5::::::::::';
..•...a•. /--'3.75
! ß• -'
90S
••......
'"'"'"'"'"'"'"•'
'"''''' '
5 ...=======================
'"'""•'"'..•:::::::::::'"'"':':.
ßßß ßß.
:"::-
..:.:.:.:.:.:.:..
3.75••:""-
;•:...........'•-•:•..."'"'"'•"'''"'
:..I '.'•,.'....'.?;:;::•_•..
---:;:::::::::;:;:"::::•'"'•:".'::-
-.-":-':-'::..'.;.-'.• ....
"-••••" I
':":"'• • :_•i:::':-::.--/...'.:•
.............. :'.".":
....
0 60E 120 180 120 60W 0
90N

60

30

30

0 60E 120 180 120 60W 0

Fig. 23. Geographical

distributionof the differencein surfaceair temperature
(degreesKelvin) betweenthe 4 X CO2
and 1 x CO2atmospheres.(a) Annualmeandifference; (b) (December-January-February)difference;
(c) (June-July-Au-
gust) difference.Shadeidentifiesthe regionswhere the differenceexceeds5øC.

hanced by the melting of highly reflective sea ice in the July-Augustseasonis shownin Figure 23c. Althoughone can
OkhotskSea.As pointedout in section4, the seaice coverage identify some longitudinal variation in the CO:-induced
over the Okhotsk Sea is exaggeratedin the 1 x CO2 simula- warming, it is not certain that these small-scalevariations are
tion. Therefore it is probable that the northwestPacific coastal
significantin view of the interannual variability of the 3-
warming is overestimatedowing to the exaggeratedinfluence month mean surfaceair temperature.Confirmingthe descrip-
of the seaice albedofeedbackmechanism.In Figure 23b one tion in the precedingsubsectionon the zonal mean response,
can identify small areasof CO2-inducedcoolingover sub- Figure 23c indicatesthe small Arctic warming and the belt of
tropical portionsof the southernhemispherecontinent.These relatively large warming at around 60øN.
areas approximately coincide with the area where soil be-
6. HYDROLOGIC RESPONSE
comeswetter, so that the evaporativeventilation is enhanced
in responseto the CO2 increase.In view of the predominance This sectiondescribes the changesin the hydrologiccharac-
of small-scalecomputationalnoisein the spacialvariation of teristicsof the model in responseto the quadruplingof the
the CO2-inducedchangeof soil moisture,it may not be rea- CO: content in the atmosphere.
sonable,however,to take thesegeographicalcharacteristics of
the temperaturedifferencein the southernhemisphereof the Annual Mean Response
model too literally. Figure 24 illustratesthe latitudinal distributionsof the an-
The differencein the surfaceair temperaturefor the June- nual mean ratesof both precipitationand evaporationfrom
 MANABE AND STOUFFER: GLOBAL CLIMATE MODEL OF CO2 INCREASE 5549

.eoj
(a)
PREC•i[•TION
50 ......... 2 "

.40

.:30
.20

.10

0
90øN 60 ø 30 ø 0ø 30 ø 60 ø 90øS
LATITUDE
.6O
(b) EVAPORATION

.50 - '........4xC02 -
lxCO2

.40 -

• .:30

.20

.10

0 I I I I I
90øN 60 ø 30 ø 0ø 30 ø 60 ø 9005
LATITUDE

Fig. 24. Latitudinaldistributionsof (a) zonal mean precipitationrate and (b) zonal mean evaporationrate in units of
centimeters/day.Solidline and dottedline indicatethe resultsfrom the 4 x CO2 and I x CO2 experiment,respectively.

both 4 x CO2 and 1 x CO2 experiments.(As was pointed out Wetherald and Manabe [1975] which describes the climate
in section4, the precipitation rate from the present low-reso- changeof a model causedby an increaseof the solarconstant.
lution spectral model is too small in the tropics and is too According to Figure 24 the distribution of the CO•_-induced
large in high latitudes.) This figure reveals that the rates of increasein the zonalmeanevaporation
rate/• doesnot have
both precipitation and evaporation increaseat most latitudes systematiclatitudinal variation. On the other hand, the in-
in responseto the quadrupling of the CO•_concentrationin creasein the zonalmeanprecipitation
rate/5 in highlatitude
the air. It is found that the global mean increasein the precip- is larger than the correspondingincrease in lower latitudes
itation (or evaporation)rate is 0.018 cm/day which implies a and has significantlatitudinal variation. These resultsimply
6.7% increasein the overall intensityof hydrologiccycle.This that the 4 x CO•_atmospherereceivesmore moisturefrom the
is smallerthan the intensificationof 12%obtainedby Manabe earth's surface in low latitudes and returns more moisture to
and Wetheraid [1980] from their simple general circulation the surfacein high latitudesthan the 1 x CO•_atmosphere.As
model without seasonal variation of insolation. This differ- discussedby Manabe and Wetherald [1980], the poleward
ence between the two results is reasonable if one recalls that transportof moistureincreasesmarkedly in responseto an in-
the global mean surfaceair temperatureof the Manabe-Weth- creaseof the CO•_concentrationin the model atmosphere,and
erald model risesby as much as 6øC in responseto the quad- thus it compensatesfor the difference in the moisture ex-
rupling of CO•_concentration,whereasa warming of about change with the earth's surface in higher latitudes. The in-
4øC occursin the presentmodel as discussedin the preceding creaseof the moisturecontent in the air, which is causedby
section. For a discussionof the mechanismsresponsiblefor the CO•_-inducedwarming of the model atmosphere,accounts
the intensificationof the hydrologic cycle, see the paper by for this increasein the poleward moisturetransport.
5550 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE

.O6

.O5

.O4

.O3

.02

.Ol

:Ol

-.02

-.03

.04
(b) %

.03 L •,"'.),
ß i
ø02

>, .01
o

• 0
E
u -.01

-.02 - ".
'•'.• .•
-.O3- --------CONTINENTS
ONLY
........... OCEANS ONLY
-.04
- CONTINENTS
ANDOCEANS
-.05 I I I I I
90N 60 30 0 30 60 90S
LATITUDE
Fig. 25. Latitudinaldistributionof zonalmeandifference
in theannualmeanvalueof (P - E) betweenthe4 x CO2
and I x CO2experiment in unitsof centimeters/day.
(a) Bothcontinents
andoceans(solidline),(b) oceans
(dottedline),
and continents(dashedline).

The results described above indicate that the latitudinal dis- Accordingto Figure 25, (P - E) fails to decreasesignifi-
tributionof the waterbalanceat the earth'ssurfaceis signifi- cantly over the continentsin low latitudesdespitethe prob-
cantly alteredfrom the 1 x CO2 to the 4 x CO2 experiment. able lossof moisture resultingfrom the increasein the mois-
This is evident in Figure 25, which illustratesthe latitudinal ture transporttoward high latitudes.This lossof moistureis
distributionsof zonalmeandifferences in (P - E) betweenthe compensatedby the enhancementof the monsoonaltransport
two experimentsover continents,oceansand both regions. of moisturefrom oceansto continentsin low latitudesmainly
(Note that (P - E) is equalto the rateof the net watergain by owing to the increaseof the moisturecontentin the air (see
the earth'ssurface.The balancerequirementof soil moisture Manabe and Wetheraid [1980] for further discussionof the
overthe modelcontinents requiresthat the annualmean(P - changein the monsoonaltransport).
E) is equal to the annual mean rate of runoff in the absenceof The distributionsof the differencein (P - E) betweenthe 4
interannualvariationof soilmoisturestorage.)As onecan in- x CO2 and 1 x CO2 experiments,which is describedabove,
fer from Figure24 and the discussion in the precedingpara- resemblesqualitatively the correspondingdifferenceobtained
graph, the earth's surface (including both oceansand conti- by Manabe and Wetheraid [1980] from their simple general
nents)losesmoisturein low latitudesand gainsit in high circulation model with a limited computationaldomain and
latitudesin response to the CO2increase.Furtherexamination idealized geography.However, there are somedifferencesbe-
of Figure25 revealsthat the increasein (P - E) in high lati- tween the two results.For example, the change of (P - E)
tudesoccursover both oceansand continents, whereas(P - over continentsin the subtropics,which is obtained from the
E) in low latitudesreducessignificantlyonly over the oceans. study of Manabe and Wetheraid [1980], has a small positive
Theseresultsmay be summarizedas follows.In the 4 x CO2 value, whereas it is difficult to determine the sign of the
experiment, moremoisture is takenfromthetropicaloceans changefrom the presentmodel. In Figure 14cin the paper by
than the 1 x CO2 experiment. This additional moisture is Manabe and Wetheraid, one can identify a zonal belt of large
brought not only to the oceansbut also to the continentsin negative (P- E) differencearound 40ø latitude, which results
high latitudesand thus accountsfor an increasein the rate of from the poleward shift of the rain belt in middle latitudes. In
runoff over the high latitude portionsof the continents. the presentstudya similar belt of enhancedaridity (i.e., nega-
 MANABEAND STOUFFER:GLOBALCLIMATE MODEL OF CO2 INCREASE 5551

(a) CONTINENTS
& OCEANS (b) CONTINENTS
90øN

60 ø
" ß . ....
ß • .. .• • ....

•'.-.'.'.'-'•e'"•N'"""
ßß-•-'"-'• •-'• •-'-'"n ß-,:.-•.::::
.:.:.:.:.:.:.•:
:. '.•.'.',
'.:•:•
•0 ø •:..•:::. k•----
•-•
w+•.:.:.:.:.'
ß. -.-.:.:.•.• :.:.:.'•.. '"'¾•
:.:. •-:.:-:-:-:-:-:-:-:-•5

.....

30 ø
• •,..-.-........................ ......
.:.:.:.•..z....•.:....:.:c..•::.:.•:.:.•:•:•':':':'¾c••' •:::.
:.:.:•:•.•.:.:..:,•.••:• --•'•'•:•::•: ::•::'$?•:•:.•.:.:.::::•
. .•...v ..-•• ß.........

60 ø
.• , _•
.... •.. • . • , .••"• .... • ---
.... .025
90øS
JAN APR JUL OCT JAN JAN APR JUL OCT JAN

Fig. 26. Latitude-timedistributionof zonal mean differencein (P - E) betweenthe 4 x CO2 and I x CO2 experiments.
(a) Continentsand oceansand (b) continents.Units are in centimeters/day.The areasof negative(P - E) are shaded.

tive (P - E)) is evident at approximatelythe same latitude
(45ø) in both hemispheres.However, the magnitude of the
negative (P - E) is significantlysmaller than that from the
study of Manabe and Wetheraid. This difference appears to
result from the incorporationof the effect of the seasonalvari-
ation of insolation into the presentmodel (R. T. Wetheraid
and S. Manabe, manuscriptin preparation, 1980).
SeasonalResponse
The latitude-time distribution
in (P - E) between the 4 x CO2 and 1 x CO2 experimentsis
shownin Figure 26. As pointed out already, (P - E) increases
in high latitudesand decreasesin low latitudesin responseto
the increaseof CO2 concentration.Though there are someex-
ceptions,the general characteristicsof the latitudinal distribu-
tion of the (P- E) differencedescribedabovepersistthrough-
out the year.
Over the continents,the beltsof both positiveand negative
(P- E) differencesundergo large latitudinal shifts from one
seasonto another (see Figure 26b). For example, the belt of
negative difference is located at about 25øN in January,
whereasa wide belt of negative(P - E) differenceextends
from 30øN to 70øN in May. Concurrently,the belt of positive
difference in high latitudes retreats poleward as the season
proceedsfrom winter to summer.
Over the southern hemispherecontinents, the distribution
of the (P- E) differencein Figure 26b contains small-scale
characteristics.Since the fractional coverageof continentsis
very small poleward of 30øS, it is probable that most of these
small-scalefeatures are not significant.
The seasonalvariation of the (P- E) change described
above causesa correspondingvariation of the changein soil
wetness.Figure 27 showsthe latitude-time distribution of the
zonal mean difference of soil moisture over continents
tween the 4 x CO2 and 1 x CO2 experiments.Accordingto
this figurethe regionof enhancedsoilwetnessin the high lati-
tudesof the northern hemisphereexpandstoward the mid-lat-
itudes in winter but retreats poleward in summer. On the
other hand, the zonal mean soil wetnessreducesin two sepa-
rate regionscenteredaround 45ø and 70øN during the warm
season.The zone of enhanced drynessaround 45øN in sum-
mer results from the poleward shift of the middle latitude
rainbelt as suggestedin the preceding section.The enhanced
drynessaround 70øN resultspartly from the elongationof the
warm drying seasonowing to the earlier timing of the late
spring snowmelt describedbelow.
The increasein (P - E) in high latitudes of the northern
of the zonal mean difference hemispheremanifestsitself as a generalincreasein the runoff
rate. This is seenas a large area of positivevaluespolewardof
60øN on Figure 28, which illustratesthe zonal mean differ-
ence in the runoff rate betweenthe 4 x CO2 and 1 x CO2 ex-
periments. The belt of negative runoff difference, which ap-
pearsparallel to the belt of positiverunoff differencein high
latitudesin the springand early summer,is apparently caused
by a shift in the time of maximum runoff rate. One can see
this shift by comparing Figures 29a and 29b, which show the
zonal mean runoff ratesfor the 1 x CO2 and 4 x CO2 experi-
ments, respectively.On the 1 x CO2 distribution, Figure 29a,
the belt of maximum runoff, which is causedby snowmelt, ex-
tends from 60øN in May to 80øN in July. While for the 4 x
CO: distribution, Figure 29b, the belt of maximum runoff
stretchesfrom 60øN in March to 80øN in June. This implies
that the maximum runoff rate in the 4 x CO: case is about 1
month earlier than for the 1 x CO: case. This shift is caused
by the time of snowmelt occurring earlier in responseto the
general increaseof surfaceair temperature.
GeographicalDistribution
As was discussedin section4, the geographicaldistributions
of precipitation rate from the 1 x CO: experiment contain
many unrealisticfeaturescausedpartly by the poor spectral
be- resolution of the climate model used for this study. Further-
more, the geographicaldistributionof the differencein precip-
itation rate between the 4 x CO2 and 1 x CO2 experimentsis
dominated by small-scalecharacteristics.It is suspectedthat
these small-scale features result from the bias of the low reso-
5552 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO 2 INCREASE

90øN 90øN

.20 •j •;•F•.W•.•fr•F.•--• ß

60" 60"

"".....:•::i:
'"' :!::::::'i'.iiiii:iiii::::•... .......... C":':"'"'
ßß
30" - :..:.:..•::...:.:.:.: 30"
.'.:? __ ..•-•.....-:-:.:-:.:-:-•-.-.- •--•..' :•
'""'"'"'"'i".'.'.'.'..'."•-0 ' '.'.'.'.'.'.'.'.'.'.'.'.'.'.'.'.-•'.'.'.'.'.'.'.'.'.',
"•:'--•._-':!•::.:.:.:.:..:.:.:.:...'..:..'.':•
:'-'--' •:i.•.:.:.:.:.:.:.:.:.:.:.:...' ::::: :::' i!!!iiiiii:i:!::::.:.
:::':':':'""' ..........o...../:.v_'
:::::::::::::::::::::::::::::
.....i"••:!:: 'x-:-:-.'-."-•
:.;-:-.-Z.;:;:;:;;;:::::;-::::
::::::::::::::::::::: '"•::':::::::::::-::::::;:.:"
..",./'":-•-""•.':'.
ß:.;.;.:.;........ 0,._•_
ß.......................•..• .:...........:.,.........:...--.•..._.....•.;. :::::::::-:.'.'.'.'.'-'.6'-'
' 4:::
...:.:.:.:.::.:.:.'.'.'.' .....
• J•""'"'"'"'"'"•'"
"'• 0..fi• k...',,...)...x •.-:._'.:.:.'_•
- • - ,•..;.:.•--?;.:.;.;•--•. ,'"- , ),.:-;.;..',,,;• '.---....,4D,;.;.;.;-;.;.;.)'"• :':"'•:'"*:::*:*:*:':'
.':':':'-'T-
......:•:::::.x.
ß-':': ß ß .•D.... . '."'"'.:.:.:..."=%....:.:.:..
•ß.....
'-'-' ' .'-'-'---"'.'.'".'.'-•
' '-1' i":':':""" ' ' ' l& ½:.'-'.'-'.'.'
,•)'.'.'-'.'-':•0-
ß''•-'
----
..,-.:..•••
' ...... "':
:::-::':•';'•
•l•
• c_• -'-'-'-:.'
%:• t..<.''•..: -:-:.:-'
:..:: ' '.:.....
:..?:•
.-./77..;-:.:-:-."•:..:.;-:,'.'-.-2•0''•----•
•--• • /:-'•-:-O'----:Q_:::-:-:-:':".......'---,-m•_._
30"
%%." '•-"•"7.•:!1!::.'•"""• •:::::::::•' .o.s;.',-'::-'.,,,,•
' -4•..:::::::-:.:::
:'.:-i
::::::::::::::::::::::
:::':•:.--'.'.--•":-•' '_ .'.':':':':':'. 0............ .c..=.:..•
•;:;;:;:;:;:;:......
....., ......... ::. ;.: ':': :%
................
:..:.=. :. ,,.:..::
:o..½iiii
60"

90os - .•
1

-- 90øS
'.......:./I
:::• r
II [.2•x:'.'.'.'.'.'.'.'.'
I!I
........?...o.......-•....
JAN APR JUL OCT JAN JAN APR JUL OCT JAN

Fig. 27. Latitude-time distribution of zonal mean difference in Fig. 28. Latitude-time distribution of zonal mean difference in
soil moisture (centimeters)over continentsbetweenthe 4 x CO•_and runoff rate over continentsbetweenthe 4 x CO•_and I x CO•_experi-
I x CO•_experiments.Note that in this study,the field capacityof soil ment. Units are in centimetersper day.
is assumedto be 15 cm everywhere.

lution spectralmodel which tends to accumulatekinetic en-
ergy at relativelyhigh wavenumbers(i.e., near the wave num-
bers of spectral truncation [Manabe et al., 1979b]).
Accordingly,it is decidedto postponethe discussionof the
geographicalaspectsof the hydrologicresponseuntil the re-
suits from a model with higher computationalresolutionbe-
comes available. The numerical time integration of such a
model is under progress.
7. SUMMARY AND
It is shownthat the sensitivityof a global climate model to
an increaseof the CO,_content in the atmospherehas signifi-
cant seasonal and latitudinal variations. For example, the
CO•_-inducedwarming of the surfaceair is particularly large
in high latitudes owing mainly to the poleward retreat of
highly reflectivesnowcoverand seaice. However,the warm-
ing over the Antarctic continent is significantlylessthan the
warming over the Arctic Ocean partly becauseof the small-
ness of a snow albedo feedback
Over the Arctic Ocean and its neighborhood,the warming
of the surfacelayer of the model atmosphereis much larger in
the winter than in the summer. It is found that a reduction of
seaice thicknessis an important factor responsiblefor the sea-
sonal asymmetriesin the atmosphericwarming. Owing to the
reduction (or the disappearance)of highly reflectivesea ice,
the surfacealbedo reducessignificantly,and the net incoming
solar radiation increasesmarkedly in summer in responseto
an increaseof the CO2 concentrationin the atmosphere.How-
ever, the additional solar energyis usedeither for the melting
of sea ice or the warming the mixed layer oceanwhich has a
large heat capacity.Thus the increaseof the surfaceair tem-
peraturein summertums out to be relativelysmall. However,
the additional heat energy, which is absorbedby the oceans
during the warm season,delaysthe appearanceof the sea ice
or reduces its thickness. This reduces the thermal
effect of the sea ice in early winter, when the air-sea temper-
ature differencebecomeslarge, thereby enhancingthe warm-
ing of the surfaceatmosphericlayer. The winter warming is
enhanced further by the stable stratification of the model at-
mospherein the winter which confinesthe warming to the
surface layers.
Over the northern hemispherecontinentsthe CO2-induced
warming in high latitudes also is at a maximum in early win-
ter, being influenced by the large warming over the Arctic
Ocean. However, one can identify a secondarycenterof rela-
tively large warming around 65øN in April. This resultsfrom
CONCLUSIONS
a large reductionin the surfacealbedo in springwhen the in-
solationacquiresa near maximum intensity.The reductionof
snow cover owing to the generalwarming of the surfaceair is
responsiblefor this albedo change.
The increaseof the global mean surfaceair temperaturein
responseto the quadruplingof the CO2 concentrationin the
air is about 4øC. This result suggeststhat the warming caused
by the doublingof CO2 contentwould be approximately2øC.
This is significantlyless than the warming of 3øC obtained
mechanism over Antarctica. from the generalcirculationmodel of Manabe and Wetheraid
[1975] which has idealizedgeographyand is without seasonal
variation. It is suggestedthat the smaller sensitivity of the
presentmodel in comparisonwith the annual mean model is
due partly to the absenceof or the smallnessof the contribu-
tion from the albedo feedback mechanism over the Antarctic
continent throughout the year and over the northern hemi-
spherecontinentsin the summer.
The general warming of the model atmosphereresultsin
the enrichment of the moisture content in the air and an in-
crease in the poleward moisture transport. The additional
moistureis picked up from the tropical oceansand brought to
high latitudeswhere both the precipitationand runoff rates
increasethroughoutthe year. Also, the time of maximum run-
off resultingfrom the rapid snowmeltbecomesearlier.
It is of interest that the sea ice disappearscompletely from
insulation the Arctic Oceanduring a few summermonthsin the 4 x CO2
experiment. Since the presentmodel tends to exaggeratethe
seasonalvariation in the thicknessof the sea ice as pointed out
 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE 5553

(a) 1,,C0
2 (b) 4,,C02

...........,..........................•, ............................ .'.:..'.:.:.:.:-:

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

60 ø
..'-:.:.:.:.:-:-:-.'-:-:..'-:.:.:

•"--'
.0
;-:-'"

<• •
..,.,.•.-. .__.............................
• .'C:'.•.:d.•.'.:.:.:.:.:
:.:.:.:.:.:.:.:.:.:.:.:.:-:.

-•,-;,'<-;.:-;.•'-:,•:.;.;.;..;.;...•..-•.'7...;•;•
'•
':':':':':':':':':
.05:'"•

••.'.'.'.'.'. -;-'.;-;.;.;-;.;.;-;.;..'..•,,
• lt•
:,:,•:.:,:.• -_._.
a..•.._..:.0 "-'-'"-'"'"'

5•• •-•••••• v.v.v.-.v.v.v.......•

v....:, 0$.v ..........-...•,.ua
•'•'; '
0
., .•0•
.7.•.•• •• • _.._ •%:':':'"
,..-.•..•...__.
••..,v.- .... .--, •x, • ,':';':>•.• • ,-:...•..?..-.-..•..?...-:
•..F..7...-..T.x•X•• •,• •,:.'..7.--..?..:,..-..-..-v.,7...-...0'.':'.:
'..':'.':'.'.'.'.'.'.';'.'.'.'.'
';';';'";'.'-".'-•.•-.---
,...... 05;;;;'-;.;;;
;.."'..•..--•..--.':.-?.r.....Z .;...'...;.;.;.;.;.;.;.;.;.;.;
;.;.;.;.;.;.;.;.;.;.x•,•
...,,•?..;.;.:.;.;.;.;....;..-....;.;.;.;,;.;.;.;.;.;.
i:i:i:i:!:i:i:i:i:!:i:!:i:i:i
i:i:!:i:i:i:i:i:i:i:::'•:•'
•'*'•':'•'""::::::::
:::::::::::::::::::::::::::::: =============================
:::::::::::::::?Y ================================

"•*'•:':':'¾'•'•'*a
;-•.•0,"'- -•C "'ff';'•:•""•'-'-'i _
•.30.
•
)• I "%".":-:':':-:-:
.•,,•....... :-:•-.'•..--.-_'"•---•
10 •--' •,,._.•.•-.•'
•1 •.--•"•
•.•0;
30_/- 10e,--•"'"....-•,.•<
ß. •..:.:.:.:.:
..?..-..-. __
•.30 '-.-.
\ V .-.'
• • ,,•,.-..•:-:.:.:.:.:..:.:.:.:.:.:.:.:.:.:?.-.
• _•i
•'.'.'.'.'.'.'.'.'.'.'.' ................................?..•.•............• •v•_•_•j•,•_ • -.o•:, ,............................•:•..•..•
.-.,,'.-.-.-.-.-.-•. .20

_-7'.'•.'."C.'?.'.•'•'.•.•.'•"•.•.':':-•':-:-:-:-:'•-:-:':
:':'X-X-X-X-X-X- '.'.'. ß-- .... '.'.'.'.'-' '.'.'.'-'.'.'-'.'-? ..... •-'-•-?C-•-•.?.•-•-•.7 •'•':-•C-•-•'•'_ _?-_'.'_'
30' -;-:-:-:-:-;-:-;-:¾;¾:-;'
,,-:-;-:¾:-:-:-:-;-:-:-:-:-;
:-;-;-:-;-:-:-:-:-:-:-:-:-:-
-:-:-:-:-:-;':':-:-:-:-:-:-:' ß........................................................
:::::::::::::::::::::::::::::
:::::::::::::::::::::::::::::
-:._-:.;:½.05;_;:-:-::::::;.:
:.:.;-:.:.:.:.:.:.:-:.:-:.:-; :::::::::::::::::::::::::::::
::;::;;;::;:;;::::?•;•'
;;•.'•'-";:.;.05•:;:;:::
::::::::::::::::::::::::::::
:.:::.:.:-:.:.x;;;:":"•.-•""--•-•::.;:
::::::::::::::::::::::::::::::
......................,,
.•,••...- •i......... .:.:.:.:.:.:.;.;.:.;.:.;.::½.,c.,..•_•""
..............
'•.,,.;.;.:.:.;.:.;.:.;.;.;.;.:.;

,,oo, :::::::::::::::::::::::::::::
:.:.:.:.:.:.:.:.:.:.:.:.:.:..:.:.:.:.:.:.:.:.:.:.:.:.:.:,
JAN APR JUL OCT JAN JAN APR JUL OCT JAN
Fig.29. Latitude-time
distribution
ofzonal
mean
runoff
rateovercontinents.
(•) 1,•<,
CO:experiment,
(b)4 x CO:ex-
periment. Units are in centimetersper day.

in section4, this result requires further confirmation by the
use of more realistic models.
In view of the natural fluctuation of the model climate, it is
desirableto evaluate the statisticalsignificanceof the CO:-in-
ducedresponseof the model which is describedin the preced-
ing sections[Leith, 1973]. Such an evaluation has not been
completed.However, it is encouragingthat R. T. Wetheraid
and S. Manabe (manuscript in preparation, 1980) have re-
cently obtained the thermal and hydrologic responseswith
qualitatively similar characteristicsby the use of a seasonal
model of the joint atmospheremixed layer ocean which has a
limited computationaldomain and idealized geography.
The global model of climate used for this study contains
many simplificationsand idealizations.For example,the hori-
zontal heat transportby oceancurrentsis not taken into con-
sideration.Also, the prognosticsystemof sea ice doesnot in-
dude the influence of snow cover and sea ice leads upon the
heat exchange between the atmosphere and underlying sea
water. Further, the distribution of cloud cover in the model
atmosphereis prescribedas a function of latitude and height
and is assumed to have no seasonal variation.
these idealizations may be necessaryfor acquiring a reliable
quantitative estimateof climate changeresultingfrom a CO:
increase.
In addition, it is important to recognize that the present
model has a very coarsecomputational resolution. Therefore,
one should not accept literally the details of the geographical
distribution of climate change which is obtained from the
the climate, it is nebessaryto develop a joint ocean-atmo-
sphere model in which the interaction between the mixed
layer and the deeper layers of the oceans are incorporated.
Such a model has been developed at the Geophysical Fluid
Dynamics Laboratory [Manabe et al., 1979a].
Acknowledgment. Using a simple energy balance model of the
joint atmosphere-mixedlayer oceansystem,I. Held and D. Linder re-
examined some of the results from the present study. R. Wetheraid
helpedwith the comparisonof the responseof the presentmodel with
the correspondingresponseof another model which possesses a simi-
lar structurebut has a limited computational domain and idealized
geography.These studieswere very usefulin assessing the reliability
of the presentstudy. T. Gordon generouslymade available his spec-
tral model of the atmospherefor this study. L. Holloway, Jr. repro-
gramed the spectralmodel for the adaptationto the presentclimate
model. K. Bryan instigatedand contributedto the developmentof the
economical method of the numerical time integration of the present
seasonalmodel describedin section3. M. Spelman made a contribu-
tion by settingup the computerprogramwhich combinesthe atmo-
sphericand oceanicpart of the model. D. Schwarzkopfand S. Fels
constructedan algorithm for computingthe transferof solar radiation
and made it available for this study. C. Rooth, R. Wetheraid, K.
The removal of Bryan, H. Levy II, and S. Schneiderread earlier versionsof the manu-
script and gave many valuable commentsfor improving it. J. Ken-
nedy, P. Tunison, J. Connor, M. Zadworny, and W. Ellis contributed
to the preparationof the manuscriptand were very willing to incorpo-
rate many major revisions.J. Smagorinsky,the Director of the Geo-
physicalFluid DynamicsLaboratory,gavewholeheartedsupportfor
this research.
REFERENCES

present study. As mentioned earlier, it is planned to repeat

Augustsson,T., and V. Ramanathan, A radiative-convectivemodel
this study by use of a similar climate model which has a studyof the COe climate problem,J. Atmos.ScL,34, 448-451, 1977.
higher computationalresolution. Bourke, W., A muir-level spectralmodel, 1, Formulation and hemi-
A topic for a future study is the processof transition from sphericintegrations,Mon. WeatherRev., 102, 687-701, 1974.
the presentclimate to CO:-rich warm climate. As was pointed Bryan, K., Climate and ocean circulation, 3, The ocean model. Mon.
Weather Rev., 97, 806-827, 1969.
out by Thompsonand Schneider[1979],it may not be possible
Callender, G. S., The artificial production of carbon dioxide and its
to infer satisfactorily the characteristicsof the intermediate influence on temperature. Quart. J. Roy. Meteorol. Soc., 64, 223-
climate from the CO:-rich equilibrium climate describedin 240, 1938.
this study. To investigatethe long-termtransientresponseof Crutcher, H. L., and J. M. Meserve, Selectedlevel heights, temper-
5554 MANABE AND STOUFFER:GLOBAL CLIMATE MODEL OF CO2 INCREASE

ature, and dew points for the Northern Hemisphere,NA FAIR 50- Manabe, S., K. Bryan, and M. J. Spelman, A global ocean-atmo-
IC-52, U.S. Nav. Weather Ser., Washington,D.C., 1970. sphereclimate model with seasonalvariation for future studiesof
Eliassen, E., B. Machenhauer, and E. Rasmussen, On a numerical climate sensitivity,Dyn. Atmos. Oceans,3, 393-426, 1979a.
method for integrationof the hydrodynamicalequationswith a Manabe, S., D. G. Hahn, and J. L. Holloway, Climate simulation
spectralrepresentation of the horizontalfields,Reœ.2, Dep. of Me- with GFDL spectralmodels of the atmosphere,GARP Publ. Ser.
teorol., CopenhagenUniv., Denmark, 1970. 22, World Meteorol. Org., Geneva, 1979b.
Ellis, J. S., and T. H. Vonder Haar, Zonal average earth radiation M611er,F., Quarterly charts of rainfall for the whole earth (in Ger-
budgetmeasurementfrom satellitefor climatestudies,Atmos.$ci. man), Petermanns.Geograph.Mitt., 95, 1-7, 1951.
Pap. 240, Colo. StateUniv., Fort Collins,Colo., 1976. M611er,F., On the influenceof changesin air on the radiation balance
Gordon, T., and B. Stem, The GARP programme on numerical ex- of the earth'ssurfaceand on the climate, J. Geophys.Res.,68, 3877-
perimentation,Reœ.7, World Meteorol. Org., Geneva, 1974. 3886, 1963.
Held, I. M., Time DependentStatisticalDynamical Model of Various National Academy of Sciences,Carbon dioxide and climate:A Scien-
Shape and Sizes, World Meteorological Organization, Geneva tific assessment,Washington, D.C., 1979.
1977. Orsag, S. A., Transform method for calculation of vector-coupled
Hering, W. S., and T. R. Borden, Jr., Ozonezondeobservationover sums:Application to the spectralform of the vorticity equation, J.
North America, Environ. Res. Pap. 38(2), Air Force Cambridge Atmos. $ci., 27, 890-895, 1970.
Res. Lab., Hanson Field, Mass., 1964. Phillips, N. A., A coordinatesystemhaving some specialadvantage
Holloway, J. L., Jr., and S. Manabe, Simulationof climateby a global for numerical forecasting,J. Meteorol., 14, 184-185, 1957.
generalcirculationmodel. Mon. WeatherRev., 101, 69-78, 1971. Plass, G. N., The influence of 15-micron carbon dioxide band on the
Hoskins, B. J., and A. J. Simmons,A multi-layer spectralmodel and atmosphericinfrared cooling rate, Quart. J. Roy. Meteorol., Soc.,
the semi-implicitmethod, Quart. J. Roy. Meteorol. Soc., 101, 637- 82, 310-324, 1956.
655, 1975. Platzman, G. W., The spectralform of the vorticity equation,J. Mete-
Kaplan, L. D., The influenceof carbondioxidevariationson the at- orol., 17, 635-644, 1960.
mosphericheat balance,Tellus,12, 204-208, 1960. Posey,J. W., and P. F. Clapp, Global distribution of normal surface
Koemer, R. M., The mass balance of the sea ice of the Arctic Ocean. albedo, Geof Int., 4, 33-48, 1964.
J. Glaciol., 12, 173-185, 1973. Ramanathan, V., and J. A. Coakley, Jr., Climate modeling through
Kondratiev, K., and H. I. Niilisk, On the questionof carbon dioxide radiative-convectivemodels, Rev. Geophys.Space Phys., 16, 465-
variation in the atmosphere,Geofis.Pura. Appl., 46, 216-230, 1960. 489, 1978.
Kung, E. C., R. A. Bryson,and D. H. Lenshow,Studyof a continen- Ramanathan,V., R. D. Lian, and R. D. Cess,Increasedatmospheric
tal surfacealbedo on the basisof flight measurementand structure CO2: Zonal and seasonalestimatesof the effecton the radiation en-
of the earth's surface cover over North America, Mon. Weather ergy balance and surfacetemperature,J. Geophys.Res., 84, 4949-
Rev., 92, 543-564, 1964. 4958, 1979.
Lacis,A. A., and J. E. Hansen,A parameterizationfor the absorption Robert, A., The integrationof a low order spectralform of the primi-
of solarradiation in the earth'satmosphere,J. Atmos.Sci., 31, 118- tive equations,J. Meteorol. $oc. Jap., 44, 237-244, 1966.
133, 1974. Rodgers, C. D., and C. D. Walshaw, The computation of infrared
Leith, C. E., Standard error of time-average estimatesof climatic cooling rate in planetary atmospheres,Quart. J. Roy. Meteorol.
means,J. Appl. Meteorol., 12, 1066-1069, 1973. $oc., 92, 67-92, 1966.
Levitus, S., and A. H. Oort, Global analysisof oceanographicdata, Sasamori,T., J. London, and D. V. Hoyt, Radiation budget of South-
Bull. Amer. Meteorol. Soc., 58, 1270-1284, 1977. ern Hemisphere,in Meteorologyof the SouthernHemisphere,Mete-
London, J., A study of the atmosphericheat balance,final report, orol. Monogr., vol. 13, edited by C. W. Newton, American Meteor-
ContractAF19(122)-165,DDCAD 117227,Coil. of Eng., New York ological Society,Boston,Mass., 1972.
Univ., 1957. Sellers, W. D., A reassessment of the effect of CO2 variations on a
London,J., MesosphericDynamics,3, The distributionof total ozone simpleglobalclimatemodel, J. Appl. Meteorol.,13, 831-833, 1974.
in the Northern Hemisphere, Final Report, Contract AF19(604)- Stone, H. M., and S. Manabe, Comparisonamongvariousnumerical
5492, Dep. of Meteorol. and Oceanogr.,New York Univ., 1962. models designed for computing infrared cooling, Mon. Weather
Manabe, S., Climate and ocean circulation, 1, The atmosphericcircu- Rev., 96, 735-741, 1968.
lation and the hydrology of the earth'ssurface,Mon. WeatherRev., Taljaad, J. J., H. van Loon, H. L. Crutcher, and R. L. Jenne, Climate
97, 739-774, 1969a. of the upper air, 1, Southern Hemisphere,NA FAIR 50-1C-55, U.S.
Manabe, S., Climate and ocean circulation, 2, The atmosphericcircu- Nav. Weather Ser., Washington, D.C., 1969.
lation and the effect of heat transport by ocean currents, Mon. Thompson, S. L., and S. H. Schneider,A seasonalzonal energy bal-
Weather Rev., 97, 775-805, 1969b. ance climate model with an interactive lower layer. J. Geophys.
Manabe, S., and K. Bryan, Climate calculation with a combined Res., 84, 2401-2414, 1979.
ocean-atmospheric model, J. Atmos. Sci., 26, 786-789, 1969. U.S. Naval OceanographicOffice, OceanographicAtlas of the Polar
Manabe, S., and R. J. Stouffer, A CO2-climate sensitivitystudy with a Seas, 1, Antarctic, Hydrogr. Office Publ. 705, Washington, D.C.,
mathematical model of the global climate, Nature, 282, 491-493, 1957.
1979. U.S. Naval OceanographicOffice, OceanographicAtlas of the Polar
Manabe, S., and R. T. Wetheraid, Thermal equilibrium of the atmo- Seas,2, Arctic, Hydrogr. OfficePubl. 705,Washington,D.C., 1958.
spherewith a givendistributionof relativehumidity,J. Atmos.Sci., Wetheraid, R. T., and S. Manabe, The effect of changingsolar con-
24, 241-259, 1967. stant on the climate of a general circulation model, J. Atmos. Sci.,
Manabe, S., and R. T. Wetherald, The effectsof doubling the CO2- 32, 2044-2059, 1975.
concentration on the climate of a general circulation model, J. Wiesnet, D. R., and M. Matson, The satellite-derivedNorthern Hemi-
Atmos. Sci., 32, 3-15, 1975. sphere snow cover record for the winter of 1977-1978, Mon.
Manabe, S., and R. T. Wetheraid, On the distribution of climate Weather Rev., 107, 928-933, 1979.
change resulting from an increase in CO2-content of the atmo-
sphere,J. Atmos.Sci., 37, 99-118, 1980.
Manabe, S., J. Smagorinsky,and R. F. Strickler,Simulatedclimatol- (ReceivedJanuary23, 1980;
ogy of a general circulation model with a hydrologiccycle, Mort. revisedMay 5, 1980;
Weather Rev., 93, 769-798, 1965. acceptedMay 11, 1980.)