THE "GREENHOUSE" EFFECT AND CLIMATE CHANGE

John F. B. Mitchell
Meteorological
Office,Brackne!l,
England

Abstract.Thepresence
of radiatively
activegasesin the prediction
of changes
in temperature
overthenext50 years
Earth's atmosphere(water vapor, carbon dioxide, and dependsonassumptionsconcerning futurechanges in trace
ozone)raisesits globalmeansurfacetemperature by 30 K, gas concentrations,
the sensitivityof climate, and the
makingour planethabitableby life as we know it. There effective thermal inertia of the oceans. On the basis of our
has been an increase in carbon dioxide and other trace currentunderstanding a furtherwarmingof at least 1 K
gasessincetheIndustrialRevolution,largelyasa resultof seemslikely. Numerical modelsof climate indicate that
man's activities,increasingthe radiativeheatingof the thechanges will notbe uniform,nor will theybe confined
troposphere andsurface byabout 2 W m-2.Thisheating is to temperature.The simulatedwarmingis largestin high
likely to be enhanced by resultingchanges in watervapor, latitudesin winter and smallestover sea ice in summer,
snowand seaice, and cloud. The associated equilibrium with little seasonalvariationin the tropics. Annualmean
temperaturerise is estimatedto be between 1 and 2 K, precipitationand runoff increasein high latitudes,and
there being uncertaintiesin the strengthof climate most simulations indicate a drier land surface in northern
feedbacks, particularlythosedue to cloud. The large mid-latitudes in summer.The agreement betweendifferent
thermalinertiaof theoceans will slowtherateof warming, modelsis muchbetterfor temperature thanfor changesin
so that the expectedtemperature rise will be smallerthan the hydrologicalcycle. Priorities for future research
theequilibrium rise. Thisincreases theuncertainty in the includedevelopingan improvedrepresentation of cloudin
expectedwarmingto date, with estimatesrangingfrom numericalmodels,obtaininga betterunderstanding of
less than 0.5 K to over 1 K. The observed increase of 0.5 verticalmixing in the deepocean,and determiningthe
K since1900is consistent with the lowerrangeof these inherentvariability of the ocean-atmosphere system.
estimates, but the variabilityin theobservedrecordis such Progress in these areas should enable detection of a
that one cannot necessarilyconcludethat the observed man-made"greenhouse" warming within the next two
temperature changeis dueto increases in tracegases.The decades.

1. INTRODUCTION Numericalclimatemodelsindicatethatotherchanges
in
climate would accompanythe increasein globally
Our planetis madehabitableby thepresenceof certain averagedtemperature, with potentiallyseriouseffectson
gaseswhich trap long-waveradiationemitted from the many societaland economicactivities.
Earth's surface,giving a globalmeantemperatureof 15øC, Thisreviewattempts toanswer severalquestions.
as opposedto an estimated-18øC in the absenceof an Whatis thegreenhouse effect(section2)?
atmosphere.This phenomenon is popularlyknown as the Whichgasesareimportant andwhy(section3)?
"greenhouse" effect. By far themostimportantgreenhouse What are theexpected changes in theconcentrationof
gas is water vapor. However, there is a substantial greenhouse gases(section4)?
contribution from carbon dioxide and smaller contributions What are the potentialclimatic effects,and how are
from ozone, methane, and nitrous oxide. theydetermined(section5)?
The concentrationsof carbon dioxide, methane, and Howwill thechanges in climateevolve(section6), and
nitrousoxideare all knownto be increasing,and in recent whenwill webe ableto detectthem(section7)?
years,othergreenhouse gases,principallychlorofluorocar- The aim here is to outlinethe physicalbasisof the
bons (CFCs), have been addedin significantquantifiesto projectedchangesin climate due to enhancingthe
the atmosphere.Thereare manyuncertainties in deducing greenhouse
effect and to identify the main areas of
the consequentialclimatic effects. Typically, it is esti- uncertainty.This review will, therefore,be selectiverather
matedthat increasedconcentrations of thesegasessince thancomprehensive.
For a moredetailedandcomplete
1860mayhaveraisedglobalmeansurfacetemperatures by discussion
of thegreenhouse
effectthereaderis referredto
0.5øC or so, and the projectedconcentrationscould the major reviewsedited by MacCrackenand Luther
producea warmingof about1.5øCover the next 40 years. [1985] andBolin et al. [1986].
,..

This paperis not subjectto U.S. copyright. Reviewsof Geophysics,27, 1 / February1989

pages 115-139
Publishedin 1989 by the AmericanGeophysicalUnion. Paper number 89RG00094

ß 115 ß
116 . REVIEWS
OF GEOPHYSICS
/ 27,1 Mitchell: "GREENHOUSE" EFFECT

2. THE GREENHOUSE EFFECT Watervapor,carbondioxide,andothergreenhouse

gasesarerelatively
inefficient
absorbers
of solarradiation
2.1. Radiative Effects (seeFigure2b). However,theyabsorbthe long-wave
The Earth-atmosphere systemis heatedby solar radiationemitted from the surfaceand reradiateboth back
(short-wave at a meanrateof SO(1 - (z)/4,where to the surface,producing
radiation an additionalwarming,and to
SOis thesolar "constant,"(zis thefraction of radiationspace, maintaining the balance with incomingsolar
reflectedby the Earthand atmosphere, and the factor4 radiation(Figurelb). As a result,thecurrentglobalmean
allowsfor thespherical
geometry of theEarth. Thismust surfacetemperatureis substantiallywarmer than the
be balancedby the emissionof long-wave(thermalor effective
,,
radiating
temperature T•. Thisincreaseinsurface
infrared)radiationto space(Figure la). The rate of temperature aboveT•isknown asthe"greenhouse effect."
coolingisgiven byc•T•4,where c•isStefan'sconstant and The term greenhouseis applied to atmospheric gases
T•istheeffective radiating
temperature ofthesystem. At which are relativelytransparentto solar radiationbut
equilibrium whichabsorblong-waveradiation,in a mannersimilarto
So(1- (:z)/4
= 6• (1) glass in a greenhouse.However,otherprocesses contrib-
ute to the warmingof a greenhouse, especiallythe
whichassuming thecurrentalbedoof 0.30givesa valueof containment of air which is warmer than the environment
Tecorresponding
to255K (-18øC).In theabsence
of an and whichwouldotherwiserise and be replacedby air at
atmosphere,
T•willbetheEarth's
surface
temperature. environmental
temperatures. Thusthe term "greenhouse"
is misleading,as the atmospheric greenhouse effect is
solelydueto radiativeprocesses.Nevertheless,
thetermis
Solar Longwave
universally
accepted andwill be usedhereafter.
Incoming Reflected
2.2. Planetary
Atmospheric
HeatBalance
In practice,of course,the atmosphere is not in pure
radiativeequilibrium. The surfaceis heatedby solar
radiationand downwardlong-waveradiation,much of
which is transferredto the atmosphereby conductionas
sensibleheator throughevaporation as latentheat(Figure
(a) No atmosphere
3). The surfaceheatingis then distributedthroughthe
Figure 1. Schematicillustrationof the greenhouse effect troposphere by convective(vertical)mixing, and thermal
showing(a) no atmosphere, wherelong-wave radiationescapes equilibriumis maintainedby long-wavecooling to the
directly to space,and (b) an absorbingatmosphere, where surfaceand to space. The global mean lapse rate
long-wave radiationfromthesurfaceis absorbed
andreemitted
bothdownward,warmingthe surfaceandloweratmosphere,
and (temperaturechangewith height)shownin Figure4 is less
upward,maintainingradiativebalanceat the top of the thanthevalue
of9.6K km'•, thatfordryair,because
ofthe
atmosphere. releaseof latentheatby condensation.The stratosphereis
heateddirecfiyby absorption of solarradiationby ozone

• • /• •Gaseousabs.
or.ption
andis approximately in radiativeequilibrium.
In the presenceof an atmosphere, muchof the long-
mission wave emissionto spaceemanatesfrom the atmospheric
gasesratherthanthe surface.Thus the level at which the
effectiveradiatingtemperature
T• occurs(the mean
(b) With atmosphere radiatingheight)is raisedabovethe surface. The mean
radiating height varies principally with the infrared
opacity. ThusMars, with a relativelythinatmosphere, has
A perfectemitter,or blackbody,emitsradiationovera a small mean radiating height and a weak greenhouse
rangeof wavelengths.The distribution of energyemitted effect, whereasVenus, with its dense,highly absorbing
with wavelengthis a functionof the temperature of the carbondioxide atmosphere,has a large mean radiating
emitter: the hotterthe emitter,the shorterthe wavelength heightanda very stronggreenhouse effect(Table 1). The
of peak emission. Thus the Sun, which has a surface Earth has a mean radiatingheight of 5.5 km (Figure 4,
temperature of 6000 K, emitsmostradiationin therange point a)andaglobal mean lapserateof6 K km4, giving a
of 0.2-4 !.tm, (includingultraviolet,visible, and near- current
greenhouse warmingof 33 K.
infraredwavelengths), whereasthe Earthat 255 K emits Increasing the concentrationof greenhousegases
mainlyin the range4-100 !.tm(long-waveradiation;see increasesthe meanradiatingheightto a level wherethe
Figure 2a). temperatureis lessthantheeffectiveradiatingtemperature
Mitchell: "GREENHOUSE" EFFECT 27,1 / REVIEWS OF GEOPHYSICS ß 117

BlackBody /,,--%

uJ

0.1
Curve •
0.15
50000

0.2 0.3 0.5

10000

1.0 1.5
Wavenumber(cm-•)

5000

2 3
2000

5
1428 1000

10
663

15 20
333

30
200

50
A

100

100

Wavelength pm

100
Ground
80-
Level

H20 (Rotation)
40-
B
20-
o
o
I
02 02
I I
02
1
03 IJUll LJ II .oJL
I IIoc;,iI.o Iii • •I o,
illill I
II L'! .o
// H20 ,.• I I N20 I Ii co,
OH4
co 4 CF2CI2
Figure 2. (a) Spectraldistributionof long-waveemissionfrom the top of the atmosphere
to the surface. Noticethe compara-
blackbodiesat 6000 K and 255 K, correspondingto the mean tively weakabsorptionof the solarspectrum andthe regionof
emittingtemperaturesof the Sun andEarth, respectively,and (b) weak absorptionfrom 8 to 12 !xm in the long-wavespectrum
percentageof atmosphericabsorptionfor radiationpassingfrom [fromMacCrackenandLuther,1985].

SPACE INCOMING
SOLAR OUTGOINGRADIATION
RADI ATI ON Shortwave Longwave

100 8 17 6 9 40 20

ATMO
S.HR [• / / // " t
\X•Backscattered/ t
Absorbed
by •
/ %••y?ir //
\ X •' /•
Net
Emission
Water
•
Vapor, ' '
by
Emission

Water
Vapor
' 19 \• xReflected/ /, CO
2'O3 byClouds

Water
Vapor, •
• 106
CO2'O3
• La•ent
Absorbedby •
• Reflected r • HeatFlux
Clouds
• bySurface Sensible

• LONGWAVE
Absorbed
46
Heat
•Flux
RADIATION 115 100 7 24
OCEAN, LAND

Figure 3. Schematicrepresentation of the atmosphericheat (thermalIR) fluxesareon the right-handside[fromMacCracken

balance. The units are percentof incomingsolarradiation. The and Luther, 1985].
solar fluxes are shownon the left-handside, and the longwave
1111ß REVIEWS OF GEOPHYSICS / 27,1 Mitchell' "GREENHOUSE" EFFECT

TABLE 1. Greenhouse Effect on Terrestrial Planets

T,. In orderto maintainthe radiationbalanceof the
Earth-atmosphere
system(equation(1)), the troposphere
and surfacemustwarm until the temperatureat the new Solar Mean Effective Observed
Irradiance radiatingradiatingsurface
meanradiating
heightis equalto T, (Figure4, pointc). Planet
SO, Wtn'2Albedo
ct level,
temper-
km ature,
temper-
K ature, K
The surfacewarmingmay be modifiedby accompanying
changesin albedoor lapserate,asdiscussed
in section5.
Mars 589 0.15 1 217 -•220
Earth 1367 0.30 5.5 255 288
Venus 2613 0.75 70 232 -•700
STRATOSPHERE
12
( From National ResearchCouncil [ 1982].
10

Enhanced 3.1.WhatDetermines
theWavelengths
of Absorption?
Gasesabsorb andemitradiationat wavelengths
which
•--•'k•
Present TROPOSPHERE
correspond to transitions
betweendiscreteenergylevels.
The relationbetweenthe changein energylevelAE and
I'- Meanradiating
thewavelength 3•of associated
radiation
isgivenby
heights
AE = h/2x3•

/ I I '• I SURFACE where h is the Planck constant. Hence each transition is

3oo associated
witha discrete
wavelength,
andlargerenergy
255 288
jumps correspond to shorterwavelengths. Diatomic
Temperature (K)
symmetric gasessuchasoxygen havenopermanent dipole
Figure4. Effectof increasing CO2 on theverticalprofileof moment, sothepossibletransitions
arelimitedto changes
temperature (schematic). EnhancingCO2 raisesthe mean
radiatingheightas shown. The temperature
profile warmsuntil in electronic
statewhichinvolvelargeenergiesandhence
the temperatureat the new meanradiatingheightreaches255 K, short wavelengths(Figure 2b). On the other hand,
the currenteffectiveradiatingtemperature
of theEarth. triatomic molecules suchas water and carbon dioxide have
in additionsmallervibrationaland rotationalenergy
transitionscorrespondingto infraredwavelengths.
Water
An important point to note is that the surface and vaporabsorbs strongly
near6.3 gm and2.7 gm becauseof
troposphereare strongly coupled: if the troposphere changes in molecular vibrational energy and at
warms, the surface must warm also and vice versa. Much
wavelengths greaterthanabout18gm because of changes
of the surfacewarmingdue to greenhouse gasesarises in rotationalenergy(Figure2b). Carbondioxideabsorbs
from direct radiativeheatingof the troposphere[seeLal stronglyaround15 gm nearthe peakof the long-wave
andRamanathan,1984]. In orderto includethe changein spectrum(Figure2b).
radiativeheatingat the surfaceand in the troposphere
the
term "greenhouse heating"usedin this paperrefersto the 3.2. What Determinesthe Size of the Contribution to
change in the net downward radiative flux at the the Greenhouse Effect?
tropospause. Althougheach transitionis associatedwith a discrete
wavelength,the intervalover whichabsorption occursis
3. WHICH GASES ARE CLIMATICALLY
"broadened" by additionor removalof energydue to
IMPORTANT, AND WHY?
molecularcollisions (pressure broadening)or theDoppler
The contributionof a gas to the greenhouse effect frequencyshift due to the randomvelocitiesof molecules
dependson the wavelengthat which the gas absorbs (Dopplerbroadening). If absorptionis strong,theremay
radiation,the concentrationof the gas,the strengthof be completeabsorption(saturation)aroundthe central
absorption andwhetheror not wavelength
permolecule(linestrength), of thespectral line(Figure5).
othergases absorbstronglyat thesamewavelengths. As a Bothwatervaporandcarbon dioxideabsorblong-wave
consequence,one molecule of dichlorodifluoromethane radiation overa rangeof frequencies (Figure2b), andthe
(CFC12)
isabout
104times
more
effective
in "trapping"
relativelyhighconcentration
of thesetwogases(Table2)
long-waveradiationthan one moleculeof carbondioxide ensures thatmanyof thespectral linesaresaturated.
Any
in thepresent atmosphere. Thefactors governing absorp- increase in theabsorption dueto enhancing theconcentra-
tion are discussedin more detail below so that the reader tionof these gasesis limitedtothewingsof theabsorption
may understand why somegaesare more importantor lines(Figure5). Consequently, theradiativeheatingdue
potentiallymoreimportantclimaticallythanothers. to thesegases increaseslogarithmically,
notlinearly,with
Mitchell' "GREENHOUSE"
EFFECT 27,1 / REVIEWS
OF GEOPHYSICS
ß 119
TABLE 2. CurrentConcentrations
andGreenhouse
HeatingDueto TraceGases.

Principalabsorbtion
bands
Concentration, Position, Strength,
cm'! Greenhouse
Gas ppm cm
'• atm
'•cm'•STP heating,
W m'2

Water vapor -3000 ~100

Carbondioxide 345 667 (manybands) ~50
Methane 1.7 1306 185 1.7
Nitrous oxide 0.30 1285 235 1.3
Ozone 10-100 x 10'3 1041 376 1.3
CFCll 0.22 x 10'3 846 1965 0.06
1085 736
CFC12 0.38 x 10'3 915 1568 0.12
1095 1239
1152 836

Themainabsorption
bands
andbandstrengths
(ameasure of theprobability
of a molecule
absorbing
aphoton
atthebandwavelength)
areshownforthelessabundantgases.

concentration(Figure 6; see also Augustssonand The mostimportant

of thesegasesarethechlorofluorocar-
Ramanathan[1977]). bonsCFC1
a (CFCll) and CF2C12
(CFC12)(Table2).
The absorption by watervaporandcarbondioxideis so Their low concentrations are suchthatabsorption in the
strong.thatothergasesabsorbing similarwavelengths will middleof thespectrallinesis not saturated, sotheireffect
contributelittle to the greenhouse
effectunlesstheyhave increases linearlywithconcentration (Figure6).
comparable concentrations. However,thereis a regionof
thelong-wave spectrum fromabout8 !xmto 12 gm known
as the "atmospheric window"whereabsorption by water Concentration C02 (ppm)
andCO2 isweak(Figure2b). Othertracegases,including 300 350 400 450 500
4
ozone,whichhasa peakabsorption at 9.6 !xm,methane,
and nitrousoxide, have absorption

tion,despite
theirrelatively
bandsin or near this
rangeand contributeto the trappingof long-waveradia-
lowconcentrations
(Table
2).
I I I I I I I II
Absorption
by thesegases
is relatively
strong
(Table2)
and
close
tosaturation,
and
their
absorption
lines
partially
overlap
those
of other;gases.
Theireffectincreases
asthe
squarerootof theirconcentrations
(Figure6).
Finally,
therearesome
gases
present
in verysmall
quantifies
whichhavestrongabsorption
linesin the
atmospheric
windowand little overlapwith othergases.

I I I I
0 1 2 3 4

Concentration
[ CFC 11,
Methane CFC 12(ppb)
(ppm)
I [ I I I I I [ I
0 0.1 0.2 0.3 0.4

Concentration, Nitrous oxide (ppm)

Figure6. Greenhouse
heating
dueto traces
gases(in wattsper
square
meter),
showing
(topscale)
concentration
of CO2(inparts
permillion;notethatthebaselinevalueis 275 ppmandthecurve
showsthe extraheatingdue to increases abovethat level. The
1
current
heating
duetoCO2isabout
50W m'2.),(middle
scale)
concentration of CFCll and CFC12 (partsper billion) and of
Wave number
methane(partsper million),and(bottomscale)concentration of
Figure 5. Schematicillustrationof weak, intermediate,
and nitrousoxide(parts;per million). The trianglesdenote1985
stronglineabsorption
[fromMacCracken andLuther,1985]. concentrations.
120 ø REVIEWSOF GEOPHYSICS / 27,1 Mitchell: "GREENHOUSE" EFFECT

4. CHANGES IN THE CONCENTRATIONS OF warming,givinga positivefeedbackbetweenatmospheric

GREENHOUSE GASES temperatureand watervapor. This is discussedfurtherin
the next section. The currentgreenhouseheatingdue to
This review is principallyconcerned with the changes water vapor isabout 100Wm'2.
in climate inducedby changesin greenhousegas con- The concentration of water vaporreachesa minimum
centrations causedby humanactivities.However,someof of about 3 ppm in the lower stratosphere.It is thoughtthat
the major greenhouse gasesoccur naturally,and their this minimum is due to the low temperature at the tropical
concentrations vary in time evenin the absenceof human tropopauselimiting the concentrationof water vapor in
intervention.For example,it is knownthat the concentra- tropospheric air mixedinto the stratosphere, in whichcase
tion of carbon dioxide fell to two-thirds its current value the concentrationof stratosphericwater vapor should
duringthe lastice age [Neftelet al., 1985]. As a conse- increase with an increase in tropospherictemperature.
quence,predictingfuturelevelsof tracegasesrequiresan Methaneoriginatingin the troposphere providesa further
understanding of thenaturalchemicalcyclesof the gases, sourceof stratospheric water vapor becauseoxidationof
as well as the clairvoyancenecessary to forecasttrendsin thehydrogenatomsin methaneproduces two molecules of
agriculturaland economicactivity. In this sectionthe water vapor for each methane molecule. The current
increasesin the concentrations of radiativelyactive gases concentration of methane(1.7 ppm) appearsto accountfor
sincethe IndustrialRevolution,usuallytakenasthe middle the occurrence of water vaporin excessof 6 ppm in parts
of the 19th century,are discussed, and the likely changes of the stratosphere [R. L. Joneset al., 1986]. If thisis so,
overthenext50 yearsareestimated.Giventhechangesin increasesin the concentrationof methanemay produce
the concentrationsof the greenhousegases, it is in increases in stratospheric watervapor.
principlea straightforward
matterto calculatethe associ-
atedchangesin long-waveheatingrates,as the equations
.of radiative transfer are well known (see, for example, 4.2. Carbon
Dioxide
(CO:)
Chandrasekhar[1950] and Paltridge and Platt [1976]). As indicatedearlier, the concentration
of atmospheric
Uncertaintiesarisebecauseof the approximations usedto carbondioxide has undergonesubstantialnatural varia-
solvethe equationsandthe measurement of the absorption tions.Carbonis cycledbetweenthreemainreservoirs:the
spectra. Detailed assessmentof these uncertaintiesis atmosphere
containing
720Gt(1Gt= 109metric
tons)
of
beyond the scopeof this review. Instead,we will use carbon in the form of carbon dioxide, the terrestrial
previouslypublishedresults [Ramanathanet al., 1985; biosphere
(1500 Gt), and the oceans(38,000GO. Esti-
Dickinsonand Cicerone,1986] to estimatethe changesin matesof the preindustrial
atmosphericconcentration
of
long-wave heating of the troposphereand surface that carbon dioxide based on measurements of the concentra-
would result from the changesin greenhousegases tionsin bubblestrappedin landice sheetsvary from 265 to
outlined below. 290 ppm [e.g.,Neftel et al., 1985]. Sincethen,concentra-
tions have increasedby about 25% (Figure 7a), initially
becauseof deforestationand more recently becauseof
4.1. WaterVapor(H20) increasedburningof carboniferousfossil fuels (coal, oil,
Most of the-gases consideredhere are well mixed and gas), depictedin Figure 7b. The increasein atmos-
throughoutthe troposphereand stratosphere. Conse- phericcarbondioxidesince 1958, when the first regular
quently,their global concentrations can be determinedby measurements commenced,accountsfor only 58% of that
measurementsat a few selected sites. On the other hand, releasedby burning fossil fuels. The oceansare the
the concentrationof water vapor(specifichumidity)varies primary sink for the excessatmospheric carbondioxide.
from asmuchas 15,000partsper million by volume(ppm) Currentattemptsto modelthe uptakeof carbondioxideby
near the surfacein the tropicsto 3 ppm in the lower theoceans havefailedto account for all the"excess"
CO2
stratosphere and variesconsiderablyon diurnal,synoptic, in the atmosphere.This is a major shortcomingin our
and seasonal time scales as well as with location. understanding of the naturalcarboncycle and henceour
However, evidence from both the observed seasonal ability to predict future carbon dioxide concentrations.
variation [Manabe and Wetheraid, 1967] and numerical The rate of increaseof burningfossil fuels hasdecreased
studies[e.g.,Mitchell et al., 1987] indicatesthat,to a first from about4% per year in the early 1970sto 2% per year
approximation,the relativehumidityof the troposphere is in the 1980s [Marland and Rotty, 1983]. The current
constant.Using thisassumptionin the ClausiusClapeyron greenhouse heatingdue to carbondioxideis about50 W
equation,a 1 K increasein atmospheric temperaturewill m'•. Thishasincreased byabout1.3W m'• since1860
be associatedwith a 6% increasein water vapor. An (Table 3).
increasein watervaporenhances thenet long-waveheating Estimatesof future concentrationsvary considerably
of the troposphereand surface and leads to further [Trabalka,1985;Bolin et al., 1986]. At present,CO:
 Mitchell'"GREENHOUSE"
EFFECT 27,1/ REVIEWS
OF GEOPHYSICS
ß 1:21
TABLE 3. Past and ProjectedGreenhouseGas Concentrationsand Associated
Changesin GreenhouseHeating AQ

Assumed1860 AQ Estimated
2035 Estimated
AQ
Concentration 1860-1985, Concentration 1985-2035,
Gas ppm Wm'2 ppm Wm'2

Carbon dioxide 275. 1.3 475 1.8

Methane 1.1 0.4 2.8 0.5
Nitrous oxide 0.28 0.05 0.38 0.15
CFCll 0 0.06 1.6 x 10's 0.35
CFC12 0 0.12 2.8 x 10's 0.69
Total 1.9 3.5

Ramanathan et al. [1985] and the median scenarioof

350 Nordhaus and Yohe [1983]. This would enhance the
tropospheric/surface
heating
bya further
1.8W m'2(Table
•- 340
3).
._(2

(• 330 4.3. Methane

(CH4)
o
Theconcentration
of methane
wasabout0.7ppm400
d 320
yearsago[CraigandChou,1982]andis now 1.7ppm,
contributing
about
1.7W m'2tothegreenhouse
effect,The
increasesince 1860 alone has addedabout0.4 W m'2
310 --
,,I,,,,1•, ,, I• •1• •1• ,• •1•,• (Table 3). Over the last decade,amountshave increased
1960 1965 1970 1975 1980 1985 by about1% per year [BlakeandRowland,1988]. The
(a) Year
mainnaturalsources foratmosphericmethanearebelieved
Figure 7. (a) Concentration
of atmospheric carbondioxidein to be paddy fields, ruminants,and wetlands. A recent
partspermillion(ppm)of dry air versustimein yearsobserved analysis
of 14½/12½
ratios
and13o12
½ratiossuggests
that
with a continuously recordingnondispersive infrared gas about32%of atmospheric
methane
originates
fromfossil
analyzedatMaunaLoaObservatory, Hawaii[fromKeelinget al.,
1988]and(b) fossilfuelCO2 emissions: [Marland fuelburning
1860-1982 [Loweet al., 1988].Themajorknownsinkis
andRotty,1983](reproduced
by Trabalka[1985]). photochemical
oxidationin the troposphere
by the
hydroxyl
radical
OH. Asthecurrent
methane
budget
isnot
.-. 6 accuratelyknown,the predictionof futureconcentrations
is difficult. If thecurrentrateof increase
is maintained,
ß

concentrations will reach2.8 ppmin thenext50 years,

contributing
anadditional
0.5Wm'2inradiative
heating.

4.4. Nitrous
Oxide(N20)
The preindustrialconcentration
has been estimatedto
be0.28ppm[Bolinetal., 1986],though
lowervalues
have
beenquoted[WormMeteorological Organization,
1985].
1860 1880 1900 1920 1940 1960 1980 The increaseto the presentamountaccounts
for an extra
(b) Year heating
of 0.05W m'2. Thecurrent
rateof increase
is
0.2% per year [Weiss,1981],which,if maintained,
would
producea concentration of 0.332 ppm in 2035. The
concentrations
are increasingat a rate of 1.5 ppm sourcesof tropospheric
N20 are believedto include
(equivalent
to 0.48%) per year, which,if maintained,microbial
processes
in soilandwater,nitrogen
fertilizers,
wouldgivea concentration
of 420 ppmby 2035. Most andfossilfuelandagricultural
burning.Themainsinkis
scenarios
assumea gradualincreasein emissionrates. A photolysis
in thestratosphere
withthe O(•D) stateof
growthin emission ratesof 1.4%peryearassuming 50% atomicoxygen. Giventhe sketchynatureof our under-
of fossilfuelCO2 remains in theatmosphere(scenario
B standing
of the N20 budget,projection of the future
[Trabalkaet al., 1986]) wouldgive a concentration
of concentrationsmust be regarded as speculative.
about475ppmby 2035,consistent withthatassumedby Ramanathanet al. [1985] assumea 2030 value of 0.375
122 ß REVIEWSOF GEOPHYSICS/ 27,1 Mitchell: "GREENHOUSE"EFFECT

ppm on the basis of the simple model used by Weiss Since many of the chemicalprocessesare nonlinear,the
[1981]; here we adopt a concentrationof 0.38 ppm by simpleextrapolationsusedin Table 3 may not be valid.
2035,contributing
a further
0.15W m'2tothegreenhouse
This and otheruncertainties,
particularlythoserelatedto
effect(Table 3). industrial
production,
shouldbe bornein mindin assessing
the valuesquotedin Table 3.
4.5. Chlorofluoromethanes (CFC11 and CFC12) The estimatedincreasein long-waveheatingof the
The atmosphericconcentrationsof CFC11 and CFC12 troposphereand surface since the beginning of the
canbe accountedfor by industrialproduction,allowingfor Industrial
Revolution
isabout
1.9W m'2(ignoring
effects
their destructionin the stratosphere and uptake by the of changes in ozone).A further increase of 3.5W m'2
oceans.Their currentcontributionto the greenhouse effect seemslikely by 2035, thoughestimatesfor the next 60
is small (Table 3). If productionis maintainedat present years rangefrom2.2W m'2to7.2W m'2[Dickinson and
levels, the concentrationof CFC11 and CFC12 will rise to Cicerone,1986]. Whereasit is relativelyeasyto estimate
0.5 and 1.0 partsper billion (ppb) within40 yearsand to theincrease in heatinggiventhechanges in greenhouse gas
0.7 and 2.1 ppb at equilibrium[Dickinsonand Cicerone, amounts,it lesseasyto estimatethe resultingchangesin
1986]. The concentration of thesetwo gasesis at present temperature.This is becausethe atmosphere adjuststo the
increasingat 5% per year. If this rate of increaseis perturbedheatingrates in ways which may enhanceor
maintained,concentrations will reach 2.7 and 4.6 ppb, reduce the magnitudeof the response(positive and
consistentwith the upperlimit usedby Ramanathanet al. negativefeedbacks),and the thermalinertia of the oceans
[ 1985]. In Table 3 a 4% per yearincreasein concentration slowsdown the rate at which the atmosphere and land
isassumed, adding 1.04W m'2tothenetglobal heatingsurfacewarm up. Attemptsto quantifythesetwo effects
rateby 2035 (Table3). aredescribedin thefollowingtwo sections.
Note that the potentialcontributionof thesetwo gases
is very sensitiveto the assumedgrowthrate. Without an
increase inemission ratestheycould add0.5W m'2tothe 5. DETERMINATION OF THE EQUILIBRIUM
greenhouseheating rate within the next 50 years. The CLIMATIC EFFECTS
reductionsproposedin the 1987 MontrealProtocolwould
reduce thistoabout 0.3W m'2[Wigley, 1988]. Accordingto Arrhenius[1896] the possibilitythat the
absorptionof long-waveradiationby atmosphericgases
4.6. Ozone(0 3) would influencegroundtemperaturewas recognizedby
Troposphericozonetendsto warm the Earth's surface Fourier as early as 1827. Tyndall [1861] attemptedto
throughthegreenhouse effect,but stratospheric ozonehas measurethe long-waveabsorptionby atmosphericgases,
the oppositeeffect becauseof absorptionof ultraviolet and Arrhenius [1896] tried to estimatethe changesin
(short-wave) radiation. There is evidenceof a slight surface temperaturesresulting from the increasedcon-
decrease in total ozone amounts over the period centrationsof "carbonicacid"(carbondioxide). Arrhenius
1978-1985duemainlyto reductions in the stratosphereused data on the atmosphericabsorptionof moonlight
[Watsonet al., 1988]. About10% of atmospheric ozone collectedby Langley some years earlier. Later studies
occursin the troposphere. Thereare indications of large provideda rangeof estimatesof the surfacetemperature
localincreases in ozoneat the surface[Bolinet al., 1986], changedue to doublingatmosphericcarbondioxide,with
mainly from sitesnear industrialregionsin the northern the value arrivedat dependingvery much on the assump-
hemisphere [VolzandKley, 1988]. The lifetimeof ozone tions made. For example,Plass [1956] consideredthe
moleculesin the troposphere is short,and the spatial surfaceenergybalanceassumingradiativeequilibriumand
variations arelarge,soit is difficultto estimate globally a constantatmospheric specifichumiditygivinga warming
averaged trends. Hence ozone has been omitted from of 2.5 K. Moller [1963], on the other hand, assumedthat
Table 3. Increasesin surface ozone alone would have little relative humidity remained constant,adding a strong
greenhouse effect,buta 25% increase throughthedepthof positivefeedbackto give a value of 9.6 K. Manabe and
thetroposphere wouldadd0.2 W m'2 to greenhouse Wetheraid[1967] pointedout that the groundand atmos-
heating. phereare not in radiativeequilibrium(seealsoFigure 3)
Many other trace gases were considered by and allowed for the convective transfer of heat from the
Ramanathanet al. [1985], but their likely contributionis surfaceto the atmosphere(using a radiative-convective
negligible. Methane,nitrousoxide, and ozone undergo model), givinga warming of2.9K ondoubling CO2. Thus
chemicalcycleswhich involve other gases(for example, the differences in the earlier results could be attributed to
carbonmonoxideandperoxyacetylnitrate)whichare also the different assumptionsmade in formulating the
known to be increasing in concentration [Worm one-dimensional models used in these studies. There have
Meteorological Organization, 1985; Derwent, 1987]. beenmany studiesof climatesensitivityusingradiative-
Mitchell: "GREENHOUSE"EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 123

convectivemodelsproducingincreases in surfacetempera- with radiative-convective

or generalcirculationmodels. It
ture ranging from 0.48 to 4.2 K on doubling carbon is assumed
thatthecontributions
3,i to 3,arelinearin AT
dioxide[Schlesinger, 1985]. Thisrangein sensitivities
can andindependent
ofeachother.Then3,iAT
= fiJ/,where
be explainedby the differentfeedbackprocessesthat were is the changein net downwardradiativeheatingat the
taken into accountin the various studies. The major tropopausedue to the equilibriumchangein componenti
feedbacks aredue to changesin watervapor,snowandsea alone(see,for example,Hansenet al. [1984]. Note that
ice, and cloud, and we now considerthesein more detail. forA//> 0,which
tends
toenhance
thewarming
(apositive
feedback),
3,i is < 0, andviceversa.We nowconsider
the
5.1. Analysis
of ClimateFeedbacks majorclimatefeedbackprocesses.
Giventhechangein concentration of greenhousegases, 5.1.1. Water Vapor As discussed in the previous
it is a relatively straightforwardtask to calculate the section,a 1 K increasein surfacetemperatureincreases
changein radiativeheatingof the troposphere and surface. saturationvaporpressureof water by about6%, which,
The majoruncertainties occurin modelingthe responseof assuming constant
relativehumidity(seethediscussion of
the atmosphere and surfaceto theradiativeperturbation,as watervaporin section4), will give a similarincreasein
is evidentfromthe previoussection. water vapor concentration,leading to a strongpositive
In order to compare results from different climate feedbackbetweentemperature and water vapor. Analysis
models it has been found useful to assess the relative of resultsfrom CO2-doubling
experiments
usingboth
strengthof the various feedbackprocesseswhich may radiative-convective and three-dimensional climate models
enhanceor reducethe radiativewarmingdue to increased indicates
a contribution
to3,ranging
from-1.4W m'2K4
trace gas concentrations.This is usually done using a to-1.8W m'2 [Mitchell,
1987].All theresults
considered
simple globally averaged energy balance model assumeaveragecloudiness,and the general circulation
[Schlesinger,1985; Dickinson, 1986]. The change in models take into account geographicaland seasonal
temperatureAT from equilibriumdue to a heating per- variations in lapse rate which influence the surface
turbationAQ is givenby temperatureresponse.Note that the contribution
to 3, is
negative,reducing3, andincreasing
AT, consistentwith a
C (3AT/3t) + )•AT = AQ (2) positive
feedback.
Fora typical
value
of-1.5W m'2K4
the equilibriumwarming is increasedfrom 1.1 to 1.7 K
whereC is theheatcapacityof thesystem,t is time,and)• (Table 4).
is the feedbackstrength.The significance of 3, may be 5.1.2. Snow and Sea Ice A generalincreasein
illustratedby the simplestcasewhen the only feedbackis temperaturewill reduce the extent of highly reflective
due to long-wavecooling to space. The increasein snowand ice cover,allowingincreasedabsorptionof solar
radiative
heating
atthetropopause
(AQ)is4 W m'2fora radiationand hencefurtherincreasesin temperature.This
doublingof carbondioxide[Lal andRarnanathan,1984]. is thewell-known"temperature albedofeedback."Clearly,
In the absenceof feedbacks(3, = 0), AT would increase the strengthof thefeedbackdependson thearealcoverof
indefinitely. However, the long-wavecooling to space, snow and sea ice and hence on the temperatureof the
c• (see(1)),would
increase
by3.7W m'2foreach1 K unperturbe•climate. Thus Spelmanand Manabe [1984]
riseintemperature
(T•andAT).Then
• = -3.7W-2K-•, foundthatthe colderthecontrolsimulation,the greaterthe
andat equilibrium,
AT = -AQ/3,= 1.1K (Table4). sensitivityof their model. It will also dependon the
Thereareadditional
feedbacks
3,i dueto watervapor, assumptions madeconcerningsurfacealbedowhich may
cloud,etc.,whichcanbe derivedby analyzingexperiments varyfrom 0.8 or morefor freshlyfallensnowto 0.4 or less

TABLE 4. Typical Internal Feedbacks

Contribution to Cumulative Cumulativechangein

feedback
factor•, feedback equilibriumsurface
Feedback Wrn-2K-1 factor
• temperatureAT,* K

Blackbodycooling 3.7 3.7 1.1

Water vapor -1.5 2.2 1.7

Sea ice/snow -0.4 1.8 2.2
Cloud cover -0.9 0.9 4.4
Cloud radiative
properties 0-1.97 0.9-2.8? 4.4-1.47

*AT= AQ/•,where
AQ= 4 W m'2.
124 ß REVIEWSOF GEOPHYSICS
/ 27,1 Mitchell: "GREENHOUSE"
EFFECT

2 32 o 3--2•

•9•9 ' •• ....'•:•::'

'•!!'"i•'•:..:,,,••'.......-.................
:•':•-J::::'•:
'"'"'• ...........................................
'-=:•- - 2 ßi.-ø.
..... 2-
90 ø N 80 ø 70 ø 60 ø 50 ø 40 ø 30 ø 20 ø 10 ø 0ø 10 ø 20 ø 30 ø 40 ø 50 ø 60 ø 70 ø 80 ø 90 øS

Figure 8. Latitude-height
distributionof the differencein zonal cycles. The unitsare in percentof total cover,andregionsof
meanclouddue to doublingCO2, averagedover 10 annual decrease
arestippled[fromWetheraldandManabe,1986].

over old, meltingseaice. Note that the modelingof seaice temperatureand hence provide a negative feedback.
in climatemodelsis still extremelysimple,ignoringeffects Somervilleand Remer [1984] attemptedto relate changes
of salinityor changesin oceancirculation. in cloud water contentto changesin temperatureusing
Estimatesof temperaturealbedofeedbackrange from observationaldata. They thenuseda simpleparameteriza-
-0.2W m'2 K'• to -0.7W m'2K'• [Ingram
et al., 1989], tion of cloudradiativepropertiesin termsof cloudwater
thoughestimatesat thelowerendof therangeappearto be [Stephens, 1978] in a one-dimensional
radiative-convective
more
realistic.
Assuming
avalue
of-0.4W m'2K'l forthis model and found a large negativefeedback. Using the
additional feedback will increasethe warming due to relationshipbetweentemperatareand cloudwaterderived
doublingCO2 to 2.2 K (Table4). from observations (as modifiedby Bohren [1985]), their
5.1.3. Cloud Cover Clouds reflect solar radiation result
suggests
a contribution
to )• of 1.9W m'2K'l, the
and absorb and emit infrared radiation. An increase in
positivevalue denotinga negativefeedback. An increase
cloud enhances both the reflection of solar radiation, in the watercontentof thin highcloudwouldalsoincrease
producinga surfacecooling, and the emissionof long- the long-wave emissionto spacefrom upper levels (and
wave radiation,producinga warming. Current evidence hencereduceemittingtemperatures), tendingto warm the
suggests thatchangesin solarheatingare dominantat most troposphere and giving a positivefeedback. Roeckneret
latitudeswherethereis significantinsolation.An increase al. [1987] parameterizedthe changesin both solar and
in the heightto which cloudsextend,implyingthat cloud long-wavecloud radiativepropertiesin a low-resolution
tops are cooler, will reduce the rate of emission of generalcirculationmodel which they usedto investigate
long-waveradiationto space,leadingto a furtherwarming. the effect of an increasingsolarconstant. In that model,
Simulationswith increasedCO2 using three- cloud radiative propertiesprovide a positive feedback,
dimensional models show a consistent increase in cloud becausethe long-wave effect dominatesthe reflective
height over most latitudesand a decreasein middle and effect.
upper troposphericcloud cover in low and mid latitudes The formationof cloudsand their radiativeproperties
(Figure 8; see also Wetheraidand Manabe [1988] and dependon many microphysicalparametersthat cannotbe
Wilson and Mitchell [1987a]). In general,there is a determinedin large-scalemodels. The parameterization of
reduction in cloud cover, leading to increased solar cloud, cloudwater, and cloud radiativeproperties
in terms
absorption,andan increasein meancloudheight,reducing of large-scalemodelvariablesis at presentextremelycrude
thelong-wavecooling.Thesechanges togethergivea very and providesone of the largestsourcesof uncertaintyin
strong positivefeedback,estimated atbetween -0.8W m'2 thedeterminationof climatesensitivity.
K'l and-1.1W m'2 K'l [seeMitchell,1987].A valueof
-0.9W m'2K'l willenhance
thesurface
warming
dueto 5.2. Three-Dimensional Climate Models
doubling
CO2by 2.2 K (Table4). Notethatthechange
in One-dimensionalmodelshaveprovedextremelyuseful
equilibriumtemperaturedependson the orderin which the in the determination
of the radiativeeffectsof increasing
feedbacksare introduced,whereasthe relativestrengthof the concentrationof trace gasesor in analyzingthe
thefeedbacks isgivenby)•i. strengthsof feedbackprocesses in morecomplexmodels.
5.1.4. Cloud RadiativeProperties Given the Theyareeconomicalto run andrelativelyeasyto analyze.
increasein atmospheric
watervaporwith temperature, one On the other hand, the one-dimensional model cannot
mightalsoexpectan increasein the watercontentof cloud provideinformationon the regionalchangesin climate
and hencecloudreflectivity. This would reducethe solar whichare of economicand societalimportance.Further-
heatingof the atmosphere and surface,tendingto reduce more, given the many nonlinearprocessesinfluencing
Mitchell: "GREENHOUSE"EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 125

climate,thereis no guarantee thatchanges evaluatedusing repeated,it will be foundthatthetime-averaged circulation

globallyaveragedparameters will evenbe representative is different. Thus changes in climate can be evaluated by
of the global averageof the changesat each location. comparingthe time-averaged statisticsfrom long simula-
Hence an enormous amount of effort has been directed tions. The simulatedcirculationwill vary slightlyfrom
toward developingthree-dimensional general circulation year to year,as is observed,so it is necessary to carryout
models (GCMs) and using them to determinedetailed statistical tests to demonstrate whether or not the differ-
climatechanges.Much of thiswork hasbeenconcentrated encesbetweena "control"anda "perturbed" simulationare
on determining the climaticeffectsof doublingatmos- likely to havearisenby chance.
pheric carbon dioxide. Hence there follows a brief The fidelityof themodelcanbe evaluated by compar-
descriptionof GCMs and their use and an assessment of ing the seasonalvariationsof the simulatedclimate with
recentnumericalstudiesof the effect of doublingatmos- thatobserved.It is possiblethata modelcouldbe tunedto
phericcarbondioxide. reproducethe presentclimatebut still produceunreliable
Atmospheric general circulation models can be estimatesof climatechange. Someattemptshave been
regardedas numericalweatherpredictionmodelswhich madeto verify climatemodelsagainstpastclimates(see,
have been adaptedfor runningover long periodsof forexample, Hecht[1985]),butthefactorscontrolling past
simulated time(decades) andin whichspecialattention has climates are generallynot well known, and the interpreta-
beengivento therepresentation of thoseprocesses which tion of paleoclimaticindicatorsis not alwaysprecise.
are importantfor climate. Thereare usuallyfive prog- Hence trust in numerical models of climate must rest on
nosticvariables:temperature, humidity,surfacepressure, the extentboth to which they are basedon physical
andthenorth-south andeast-west components of thewind. principles andto whichtheyreproduce thepresentclimate
The valuesof thesevariablesarekeptat regularlocations and its time and spacevariations.
on the globe(the modelgrid) and at variouslevelsin the
modelatmosphere.Currentclimatemodelsuse a grid 5.3. EquilibriumStudiesof the Effectof Doubled
whichtypicallyis approximately 5ø of latitudeby 7ø of Carbon Dioxide Amounts
longitude,with between two and 11 or so levels in the The radiativeperturbation due to increasingcarbon
vertical. The prognostic variablesare determined by the dioxideis qualitativelysimilar to that obtainedby
Navier Stokes equations(wind), the thermodynamicincreasing othertracegases,exceptnear the tropopause
equation(temperature), andequations for the conservation where the other gasesproducea slight warming,as
of water substanceand mass (humidity and surface opposedto a cooling[Ramanathanet al., 1985]. Henceto
pressure)assumingthat the atmosphereis vertically a first approximation the climaticeffectsof carbondioxide
hydrostatic and a perfectgas. The equationsmay be andothertracegasesare likely to be the same. Thusfor
solvedusingfinitedifference or spectraltechniques. Many simplicity, most numerical studiesof the effect of in-
atmospheric phenomena includingclouds,precipitation, creasesin trace gasescarriedout to date have considered
radiativeheating,and surfacefrictionvary on a scale increases in carbon dioxide alone.
smallerthanthemodelgridandsomustbe represented in 5.3.1. EarlyStudiesThefirststudies of theeffectof
an approximate mannerin termsof the grid box variables doubling atmospheric CO2 usingGCMswerecarriedout
(parameterized). by Manabeand Wetheraidat the Geophysical Fluid
In numericalweatherpredictionthe equationsare Dynamics Laboratory (GFDL) in Princeton.Theycarried
steppedforward in time from a set of initial conditions, out a seriesof experiments usingincreasingly complex
and the evolutionof individualdisturbances is usually (and realistic) versionsof the climate model in order to
closeto thatobserved.If theforecast is runfor a periodof isolate the influence of various features on climate
about 10 days or more, it is found that the observedand sensitivity.In their initial study,Manabeand Wetheraid
simulatedevolutions divergebecauseof the exponential [1975]useda modelwithnoorography andwitha domain
growthof errorspresentin theinitialconditions or arising limited to one third of a hemisphere. The seasonal
from shortcomings in the modelformulation.However,if variationof insolationwas ignored,cloudamountswere
the simulation is extended overa periodof severalyears, prescribed, andthe oceanhadno heatcapacity,nor did it
updating theseasurface temperatures fromclimatological transport heat("swamp"ocean).Doublingcarbondioxide
data and the solar zenith angle with season,the time- concentrations increasedthesurfacetemperature by 2.9 K
averagedcirculationaveragedover, for example,all the (Table5), compared to 2.4 K usinga radiative-convective
Januaries is foundtoresemble theobserved climatological model [Manabe and Wetheraid, 1967]. The increased
averagefor January,regardlessof the initial conditions sensitivitywas largely due to the temperature albedo
used.If someaspect of themodelischanged (forexample, feedbackwhich was ignored in the one-dimensional
theconcentration of carbondioxide),andthe simulationis model. The experiment wasrepeated usinga modelwith
126 ß REVIEWSOF GEOPHYSICS/ 27,1 Mitchell' "GREENHOUSE"EFFECT

TABLE 5. GlobalMean SurfaceTemperatureChangesDue to Doubling

Atmospheric
CO2Amountsin GFDL Experiments (SeeText)
Temperature
change
ondoubling
COz,K
Model Originalvalue Revisedvalue*

Swamp ocean
One-dimensional r adiative-conv ective 2.4 2.0
Annual mean, sector 2.9
Annual mean,sector(variablecloud) 3.0
Mixed layer ocean
Seasonallyvarying, sector (2.4)t
Seasonallyvarying,realisticgeography (2.0)t
Seasonally varying,realisticgeography
(variable cloud) 4.0

*Includesresultsusinga revisedradiativetransferscheme.
'•Valueobtained
byhalving
theresults
forquadrupling
CO2.

variable (model generated)cloud cover and a revised on doubling CO2. In therealisticgeography experiment,
radiationscheme,givinga warmingof 3.0 K, as opposed there is a marked increasein high cloud cover at all
to 2.0 K with the one-dimensionalmodel [Manabe and latitudes whichleadsto a reduction
in outgoing long-wave
Wetheraid, 1980]. radiation(due to changesin cloud) and hencea positive
On introducingseasonalvariationbut maintaining feedback.This differenceappearsto be partly a resultof
prescribedcloudamounts, andreplacingtheswampocean the choice of finite difference levels around the
by a 68-m mixed-layeroceanin whichoceantransport
was tropopause:in the global model, there is a level at about
ignored,the warmingwasreducedto 2.4 K with this 95 mbar which coincides with the maximum increase in
versionof the sectormodel[WetheraidandManabe, 1981] highcloudcoverin the tropics,whereasthenearestlevel in
(assuming thatthe magnitudeof theresponse
variesas the the sectormodel is at 70 mbar where changesin cloud
logarithm of the fractionalchangein carbon dioxide cover are small. Wetheraid and Manabe [1988] also note
concentration). This reduction was attributed to the fact thatthe CO2-induced
increase
in low cloudis enhanced
that in the model with seasonal variation the extent of ice when the seasonal variation of insolation is removed from
and snow and hencethe size of the temperaturealbedo their new global model, furtherweakeningthe positive
feedbackare severelycurtailedin the summerhalf of the feedbackdueto cloudprocesses.
year,just whenincominginsolationis largest. Using a The experiments abovedemonstrate that the response
globalmodelwithrealisticgeography, themeanwarming of a model dependson the model used. Furthermore,
wasfurtherreducedto 2 K [Manabeand Stouffer,1980]. conclusions basedon simplifiedmodelsmaybe misleading
With realisticgeography,thereis no temperaturealbedo qualitativelyas well as quantitatively,
as illustratedby the
feedback over Antarctica, as the elevation is sufficient to studies with fixed and variable cloud described above.
ensurethat the surfacetemperatures
remain well below 5.3.2. RecentStudiesThe GFDL experiments show
freezingwithbothpresentandenhanced of the influenceof realisticorography,seasonalcycle, and
concentrations
atmospheric with thepolar changesin cloudamounton climatesensitivity.In recent
carbondioxide. Thiscontrasts
regionsof thesectormodels(andthenorthernhemisphere) years, five studieshave been carried out using models
whicharewarmer,
beingat sealevel,andWhere
tempera- incorporatingthesefeatures.The studieshavebeencarried
turealbedofeedbacks canoperatefor at leastpartof the out at the GoddardInstitutefor SpaceStudies(henceforth
year,leadingtogreater
climatesensitivity. referred to as GISS) [Hansen et al., 1984], the National
WetheraidandManabe[1986]repeated the "realistic Center for AtmosphericResearch(NCAR) [Washington
geography"experimentwith variable cloud cover and and Meehl, 1984], GFDL [Wetheraid and Manabe, 1986;
foundthat the sensitivityof the modelwasdoubled. In Manabe and Wetheraid,1987], the MeteorologicalOffice
contrast, the introductionof variable cloud into the (MO) [Wilsonand Mitchell, 1987a], and Oregon State
annually averaged sector models had little effect on University (OSU) [Schlesingerand Zhao, 1987]. The
climate sensitivity(Table 5). Wetheraidand Manabe models and the results from the first three studies have

withtheannually beenreviewedin detailby Schlesinger

[1988]notethatin their1980experiment andMitchell [ 1985,
averagedsectormodeltherewas little increasein high 1987]. All themodelsusea fixed-depth, well-mixedlayer
cloudcover,and outgoinglong-waveradiationincreased to representthe ocean,apartfrom the GISS model which
Mitchell: "GREENHOUSE"EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 127

TABLE 6. GlobalMeanChanges
in FiveRecentCO2-Doubling
Studies

Surface
Temperature Precipitation
Study Source Change,K Change,%

GISS Hansen et al. [ 1984] 4.2 11.0

NCAR Washingtonand Meehl [1984] 4.0* 7.1
GFDL Wetheraid and Manabe [ 1986] 4.0 8.7
MO Wilson and Mitchell [ 1987a] 5.2 15.0
OSU SchlesingerandZhao [ 1987] 2.8 7.8

*W. M. Washington(personalcommunication,
1988).

prescribesa seasonalvariationof mixed-layerdepth. The (a) NCAR, 2 X CO2- 1 X CO2

GISS and MO modelsprescribeheatingto allow for the
transportof heatby the oceans(assumedto be the samefor
boththe1 x CO2'and2 x CO2 simulations).
Theother
models ignore oceanic heat transport,and the sea ice 0 55
margins in the NCAR control simulation extend much
furtherequatorwardthanobserved. 0•15 121 ©
Here we will presenttypical results,indicatingthose ,
featureswhichare representative
of all five studiesand are 10 •z 265 r'•
well understood, and we will outlineareasof disagree- 5 ' :'•.'";'.-:-
' : ' : 540
ment. In all these studiesthe equilibriumresponseto
0 1013
doublingatmospheric carbondioxidewascomputed, and 90øN 70øN 50øN 30øN 10øN 10øS 30øS 50øS 70øS 90øS

seasonalaveragesfor Decemberto February(DJF) and Latitude

Juneto August(JJA)will be presented.

Figure 9. Latitude-height
distributionof the changesin zonally
The globaland annualaveragewarmingvariesfrom
averagedtemperature due to doublingCO2 (perturbed minus
1.8 K (OSU) to 5.2 K (MO) (Table6). The warmingis control). Contoursare very 1 K. Areasof decreaseare stippled
accompanied by increasesin evaporation
andprecipitation heavily, and increasesgreater than 4 K are stippledlightly.
(Table6). The enhanced radiativeheatingof the surface Distributionsare shownfor (a) DJF and (b) JJA [Washington
due to increases in carbondioxideand water vapor is and Meehl, 1984] (reproducedfrom Schlesingerand Mitchell
[19871).
balancedby increasedcooling due to evaporation,
producinga moreintensegloballyaveraged hydrological (b) NCAR,2 X CO2- 1 X CO2
cycle.
The changein temperature
is not uniformlydistributed
in timeor space.Withenhanced
CO2, thereis increased
emissionof long-waveradiationfrom the top of the
atmosphere to space,producinga coolingof the strato-
sphere(Figure9). The zonallyaveragedwarmingis
generallymost pronouncedin high latitudesnear the
10 265•_.
surfacein winterbecause of severalamplifyingfactors.
3 3 4
First,snowandseaicearelessextensive (thetemperature
albedofeedback). Second,in high latitudesin winter,
5 3•3• • .".":;:";'-".'.
;""' :":.'.. 540
thereis a shallowlayerof colddenseair nearthesurface(a 90øN 71•øN50•ON30•ON10ON10øS 30øS 50øS 70øS 90øS
temperature inversion).The warmingat thesurfacedueto Latitude

increasedradiativeheatingis confinedto this shallow

stablelayer,whereas
at lowerlatitudes
it is mixedthrough
the depthof the atmosphere, givinga correspondinglycondensed
out in regionsof ascending
motion,releasing
smallerwarming.Finally,thewarmer,moisteratmospherelatentheatin themiddleanduppertroposphere. As noted
is moreefficientin transporting latentenergyfrom the earlier,thewarmingof theatmosphere is accompanied by
tropicsto high latitudeswhereit is releasedas latentheat an increase in water content and enhanced latent heat
of condensation [Manabeand Wetheraid,1975]. The release,sothatthewarmingat thesurfaceis amplifiedwith
warmtropicalatmosphere is rich in watervaporwhichis height(Figure9).
128 ß REVIEWSOF GEOPHYSICS/ 27,1 Mitchell: "GREENHOUSE"EFFECT

SURFACE AIR TEMPERATURE DIFFERENCES

GISS, 2 X CO2- 1 X CO2 DdF

90ON I I _. I I I I . I
'•' "'.•'•-I0
'... ß ß '.':
:...•:..::•:".:iit
:?',:!:'..:::
.....
"":':"::::'"
:":":'•:'""•'
.......i Io!%''
..............
.......
30 ø

"6 ' '....4 , .' ..... :: '.".."".:"

9oos
-I i i I i ..... i i ! I
0ø 60 ø 120 ø E 180 ø 120 øW 60 ø 0ø

Figure10. Changes
in surface
temperature
dueto doubling
CO2
for (a) DJF and (b) JJA. Contoursare ever),2 K, and increases
greaterthan 4 K are stippled[Hansenet al., 1984] (reproduced
from Schlesinger andMitchell [1987]).

GISS, 2 X CO2- 1 X CO2 JJA

I I I I I I I

(b) 90ø
N-::•

60ø-
••4 '•.:: • 4•• •"'
'•'•••4 •::: 4. ". .

0o

30ø
-• 4

I I I
0ø 60 ø 120 ø E 180 ø 120 ø W 60 ø 0ø

The simulated changes in temperaturealso vary the onsetof freezingandleadingto thinnerseaice. More
considerably with longitudeand season(Figure 10). The heatcandiffusethroughthe thinnerseaice, alsocontribut-
largestwarmingoccursover seaice in winter;the smallest ing to the enhanced surfacewarmingin winter [Manabe
over sea ice in summer. In summer,sea ice in the Arctic is andStouffer,1980].
maintainedat freezinglevel by the meltingof ice. With There is also a considerable
geographicalvariationin
enhanced CO2, eithermoreseaiceis melted, producing no warmingwithin individualcontinents. In regionswhere
change in surface temperature,or the sea ice melts the soil becomesdrier, evaporationmay be restricted,
completely,leaving the underlyingoceanicmixed layer leadingto increasedwarmingand vice versa[Manabeet
which limits the rise in temperaturethroughits large al., 1981]. Local changesin cloud amount may also
thermal inertia. On the other hand, the reductionin surface contributeto the regional variations,especiallywhere
albedodue to removingseaice leadsto increasedabsorp- insolationis significant.Althoughthelarge-scale features
tion of solarradiation. This extraenergyis storedin the of the simulatedtemperaturechangesare consistentfrom
mixed layer and releasedin autumnand winter,delaying modelto model,thedetailsvaryconsiderably.
Mitchell: "GREENHOUSE"
EFFECT 27,1 / REVIEWSOF GEOPHYSICSß 129

180 ø 120 øW 60 ø 0ø 60 ø 120 ø E 1 •30 ø

(a) 90øN ! • i I [ • I I I I • • 90øN

60 ø

30ø o -.....
.ii'"'•:. • -..::!:i"
•
0 ::::::::::::::::::::::::::
.........
...::i:i:i:i:iiii!iiiiiiiiiiiiiiii!iiii
::'"'"'::!'i!i':
......
::::::::::::::::::::::
30 ø

-0 o
o
30 ø 30 ø

60 ø - 60 ø

90øS i • i • i i 90øS
•oo •oow •oo ' oo •oo ' '
120 ø E 180 ø

Figure11. Changes
in precipitation
dueto doubling
CO2 for (a)
DJF and (b) JJA. Contoursare at 0, + 1, and + 2 mm d-l, and
areasof decrease
arestippled[fromWilsonandMitchell, 1987a].

180 ø 120øW 60 ø 0o 60 ø 120øE 180 ø

(b)
90
øNi ===================================:'•1
IL '___l' ' ' F-I' _ 90
øN
o.1' (A-l "0 ' o
1 1 ..:!iiiii!iiii!:ii!iii!::i:!:::::'::.i:::::i::::'i:i:i:::::-'.'
0 •-..•
(• .....'."
.½:.:!:...:::
['!:!:i:::.-.
1 ½, C..••
30
ø- ::)•- 30':'
0o =======================================================
.... •.___...__./_
0o

30
ø -.............
'"•- 30':'
600-L '""• - ,.._,_
• •____.•_•_•
1• • 60ø

90øS I , • ' • ' • ' • , • i 90 S

180 ø 120 øW 60 ø 0ø 60 ø 120 øE 180 ø

With theincreasein atmospheric

moistureaccompany- tion simulatedby the differentmodels. Precipitation
ing thewarmingdueto enhanced CO2 onewouldexpect dependson many small-scalephenomena: orography,
precipitationto increasein the main regionsof low-level
atmosphericconvergence,including the extratropical
depression beltsand the IntertropicalConvergence Zone
[Mitchell et al., 1987]. There is a generalincreasein
precipitation in highlatitudes,especiallyin winter(Figure
11), consistent with increased moisturetransportfrom low
latitudes. The changesin the IntertropicalConvergence
Zone (ITCZ) vary considerablyfrom model to model,
with, for example,the GFDL and MO modelsproducing
markedincreases in monsoonprecipitationover southeast
Asia and the NCAR and GISS models producing
decreases.In the subtropics, precipitation
changeslittle or
decreases slightly. Apart from high latitudes,thereis, as
yet, little agreementin the regionalchangesin precipita-
land-seacontrasts,and transientatmospheric
disturbances,
all of which are degradedin low-resolutionmodels.
However,manyof the featuresfoundin the MO experi-
ment, including the enhancementof the summer mon-
soons,have also been obtainedusing higher-resolution
modelswith prescribed increasesin seasurfacetempera-
tures [Mitchell and Lupton, 1984; Mitchell et al., 1986,
1987].
In determiningthe availability of moisture at the
surface,changesin evaporationas well as precipitation
have to be taken into account. In current models,
precipitation
and snowmeltare accumulatedat eachgrid
pointin whatcanbe regardedas a "bucket,"typically15
cm deep. Any "overflow,"corresponding
to saturationof
130 ß REVIEWSOF GEOPHYSICS
/ 27,1 Mitchell: "GREENHOUSE"
EFFECT

the soil,is lostanddiagnosed asranoff,andthewaterlevel

tintropical
continentsin winter,resultinglargelyfromthe
is depletedby evaporation.(The GISS modelhasa more
increasedprecipitation. However, in summer,there is
sophisticated considerable
treatmentof soil moisture;seeHansenet al. disagreement (Figure 12), with the GFDL,
[1983].) With doubledCO2, all five modelsproduce MO, and OSU models all simulatinga drier surfacein
enhanced soil moisture over most of the northern ex- northern
mid-latitudes,
whereas
theGISSandparticularly

GFDL, 2 X CO2- 1 X CO2

60 ø--

60 o --

90øS I I I I I I I
0o 60 ø 120OE 180 ø 120 øW 60 ø 0ø

G ISS, 2 X CO2- 1 X CO2

(b) 90øN
t30 ø - --

30 ø

•øø
90øS i • • I
••
'" I •
0ø 60ø 120
ø[ 180
ø 1•0øW 60ø 0ø

NCAR, 2 X CO2- 1 X CO2

90 ø N- I I I I I I I ,_

(c) 60 ø -

30 ø -

0 o-

30 ø -

60 o -

90os - I i i I I ! I
0o 60 ø 120 ø E 180 ø 120 ø W 60 ø 0ø

Figure12. Changein soilmoisture CO2 (JJA) The modelsshown are (a) GFDL, (b) GISS, and (c) NCAR.
dueto doubling
from three differentmodels. Contoursare every 1 cm (GFDL, Note the discrepancies
in northernmid-latitudes(reproduced
NCAR) andevery4 cm (GISS). Areasof decreaseare stippled. from SchlesingerandMitchell [1987]).
Mitchell: "GREENHOUSE" EFFECT 27,1 / REVIEWSOF GEOPHYSICSß 131

the NCAR modelsexhibit a widespreadincreasein soil

moisture. As this area containsthe world's major grain-
4F Standard
run
off
3 !- • ---- Modified
runoff
growingregions,whichalreadyare vulnerableto drought,
there is considerable interest as to which of the simulations
are correct.
=O
In thosestudiesproducinga summerdryingin northern •o 0 -
'• o•
latitudesthe following factorshave been found to con-
tribute: increasedevaporationdue to enhancedradiative ßo•• •=
O
-
heatingfromthe warmer,moister,CO2-enriched
atmos-
phere;earlier snowmelt,especiallyin high latitudes;and •o
reducedcloudandprecipitation,at leastin moresoutherly -4
latitudes [Manabe et al., 1981; Manabe and Wetheraid,
1987]. Mitchell and Warrilow [1987], using a revised
versionof theMO model,carriedout two CO2-doubling -6 i • I I I I I I I I I I
experimentswith differenttreatmentsof snow•nelt.In the Month J F M A M J J A S O N D
firstexperiment,
rainandsnowmelt
wereusedto augment Figure 13. Changein monthlymean soil moisturedue to
soilmoisture
in theusualfashion.On doubling
CO• the doublingatmospheric
carbondioxideconcentrations,
averaged
soildriedoutmorerapidlythanthecontrolin late spring over land between 45 ø and 60øN. The solid line shows the
and remaineddrier into the autumn(Figure 13). In the standardtreatmentof runoff (seetext); the dashedline showsall
secondexperiment,rain and snowmeltwere run off if the rain andsnowmeltranoffoverfrozenground[fromMitchelland
Warrilow, 1987].
near-surface
soil layerswere frozen,producinga much
driersoilin thecontrolsimulation.
On doubling
CO• the
soil driedout muchlessrapidlyandremainedwetterthan
thecontrolfor mostof theyear. Thisis consistent
with the
30
resultsfrom the NCAR modelwhich alsoproducessoil
moisturelevels which are generallyfar from saturation
over the northern mid-latitude continents in the control 20

simulation.The GISS modelalsorestrictsthe absorption

of snowmeltover frozenground,which may explainthe
increasein summersoil moisturein regionsof heavy
winter snow accumulation over western North America
andnorthwest Europedepictedin Figure12. Theseresults
indicate the dependenceof model results on the
parameterizations used. Recently,much more detailed 30

representationsof the land surface which allow for the

effects of vegetationhave been developed[Dickinson,
1984;Sellerset al., 1980; Warrilow et al., 1986]. I
Theserecentstudieshaveemphasized the changesin
mean climate. For agriculturalpurposes,changesin
variabilitymaybe at leastasimportant aschanges in mean
climate.WilsonandMitchell[1987b]investigated changes
in climatevariabilityover westernEuropein an experi- Jul 1 Jan 1 Jul 1 Jan 1 Jul 1 Jan 1

mentin whichCO• amounts werequadrupled,

and sea Figure 14. GCM-simulateddaily precipitationtotals (in
were increasedaccordingly[see millimeters)simulatedoversouthernItaly for (top) controland
surfacetemperatures
MitchellandLupton,1984]. For example,theincreases
in (bottom)
4 x CO2[fromWilson
andMitchell,1987b].
CO• andseatemperatures
produce
a reduction
in summer
precipitationover southernItaly, manifested in a reduction 5.3.3. EquilibriumOceanic Effects The models
in daily precipitationtotalsand a reductionin the number used to derive the results reviewed above have either
of dayswith rain(Figure14).However,in orderto provide assumed that the heattransport by the oceansremained
realisticinformationon changesin climatevariability, constant or ignoredoceanicheattransport altogether.
The
there is a need for the horizontal resolutions used to be oceaniccirculation
is drivenby the transferof heat,
finerthanthe500 km typicalof currentclimatemodelsand moisture,andmomentum fromtheatmosphere acrossthe
for long simulationsto be run to fully establishthe oceansurface.Clearly,changes in atmospheric
circulation
statistical
significance
of thechanges. arelikely to includechangesin oceaniccirculation
which
13:2 ß REVIEWSOF GEOPHYSICS/ 27,1 Mitchell: "GREENHOUSE" EFFECT

may in turnfeedbackto the atmospherethroughchanges models. The existing studies have used numerical
in ocean surface temperature. It is known that the techniquesto acceleratethe approachto equilibrium
circulationof the oceanhaschangedin the past,as, for [Bryan,1984]. The zonallyaveraged equilibriumchanges
example, during the last glacial maximum when the in atmospheric temperature on quadrupling CO2 from
Atlantic oceanicpolar front was displacedequatorward, Spelman andManabe[1984](Figure15) arequalitatively
and duringthe periodof maximumnorthernhemisphere similarto thoseobtainedusingatmosphere-only models
insolationaround9000 yearsagowhencoastalupwelling [e.g., Manabe and Wetheraid, 1975], with an enhanced
associated with the Asian monsoon was more intense. surfacewarmingin highlatitudesdue to the temperature
Hence the climatic effects of increased concentrations of albedofeedbackanda weakeningof thesurfaceinversion,
greenhouse gasesmay includesignificantchangesin ocean andincreases in temperaturewith heightin thetropicsdue
heat transport with substantial impacts on climate, to moist convectiveprocesses. The changesin ocean
especiallylocally. surface
temperaturefollowthosein theatmosphere, though
The use of three-dimensionalcoupled ocean- they are smaller,especiallyin high latitudes. The en-
atmospheremodelsto study the effectsof increasedtrace hancedsurfacewarmingin the polar region is mixed
gaseshas largely been limited to low-resolutionsimula- downward in the high-latitudeoceans and advected
tions using annuallyaveragedinsolationand idealized equatorwardto give a pronouncedreductionin vertical
orography [Spelman and Manabe, 1984; Manabe and temperature gradientat mid-latitudes.Despitethereduced
Bryan, 1985]. The deepoceanrequiresseveralcenturiesto meridionaltemperature gradientevidentin Figure 15,
reach thermalequilibriumbecauseof its large thermal Manabe and Bryan [1985] found little changein the
inertia. At present,no institutionhas the computing intensityof the oceanicmodel'smeridionalcirculationfor
resourcesto perform lengthy experimentswith such CO2concentrations above300 ppm. The coefficient of

25

95

205
.........
' ,œ ,*
...:::::•.••••.. . ......
..... , 2•3•--
4•'-• - 10
350

515

680
,...,
:....::..•••
':'••'"'
:....••
:•
:'''"''
:...:
:•• 6 5T
83O
94O '-•." "::::::-."::::'.
::::::::::::::::::::::::::::::: 4 Mean
990
ß
::::.-'•---,'.].',•.•ii-
:•'•;'•"'""'
, ,4.0 _•- :3.5- 200
,
i!i_7.5
!ii8.5- - ½
•.'.'.'.-.'. ..:.:.:.:-:-:-:-:-:-:.:.:.:.:-:-:-:.:-:.:.:-:-: ' . ..:.:-:.:.............................
600 400

:.•:•.:.•F:•:.:•:•:M:•:•:•:•:•:•:•:•:M:M:•:•:•:M:M:i:•:•:•:•:•:M:•:•:•:•:•:•:•:•:•:•:•:•:•:M:•:•:M:•:M:•:•:M:•:•:•:•:
.......-...-.-..
3000
2000
.... •.•.•n.•..-•.....•.•.•.•.•.•.•.:.:.:.:•:.:•:.:.:.:•:•:.:•:•:.:•:•:•:.:.:•:.:.:.:.:.:.:.:•:.:•:•:.:•:.:.:•:.:•:.:•:.:•:•:•:.:.:.:.:.:•:•:•:•:.:•:•:.:•:•:.:•:•:.:•:•:.:.•.•:..
ß:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.2:.:.:.:.:.:'.'.'..'
.• '."::::::::::.
:.:.:.:.:.:.:.:.'.'.'.'.'.'.:.:.:.:.:.:.:.:4.:.:.:.:.:':-:':':':'::',•-:"',a,•::•;i:'
',444.
',444.
',4•. '•b4.:444.
:.-•.•
:.:.:.:.:.:.:•:•:•:•:.:•:.:•:•:.:.:.:•:.:•:.:•:.:.:•:•:•:•:.:.•...':.•..•.n4.•..•..•.•
'• •' ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
"'.n'•.
"nnn"'-'-nn" 'nnnn"w'"w"
.?..:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.::::::.:.........•....::?..
/. O ......... '"'""'"'"'"""":'"":':':':':':':':':':':':
4000
.ooo
90 ø N 75 ø 60 ø 45 ø 30 ø 15 ø 0ø

Latitude

Figure 15. Height-latitude diagram of zonally averaged 1 K upward. Extra dashedcontoursare at 3.5 K, 6.5 K, and7.5
equilibrium
changes
in temperature
dueto quadrupling
CO2 in an K. Areasof decrease
arelightlystippled,andwarminggreater
idealizedcoupledocean-atmosphere
model [f•om Spelmanand than7 K is heavilystippled.
Manabe, 1984]. Contoursare at-8 K,-4 K, -1 K, and then every
Mitchell: "GREENHOUSE" EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 133

thermalexpansionof seawaterincreasesapproximately oceanicdensity structureand varies with latitude, it is

quadraticallywith temperature. Hence the greater possiblethat the transientresponsecoulddiffer from the
warmingin highlatitudes is compensated by theeffectof equilibrium response,as suggestedby Schneiderand
thelargercoefficient
of expansionin lowlatitudes,andthe Thompson[1981]. If the oceansare assumedto have a
meridionaldensitygradientchangeslittle. (For lower constantheatcapacityC (equation(2)), thenthe evolution
concentrationsof CO2 themodel's seaicemarginextends of theresponseto an instantaneous increasein heatingAQ
substantiallyequatorward,insulatingthe oceanfrom heat is givenby
loss to the atmosphere.Consequently, the meridional
circulation
is weakened,especially
in highlatitudes.) AT = (AQ/X) [1 - e (-)•t/C)] (3)
The aboveresultsshouldbe viewedwith caution,given
the potentiallymisleadingresultsobtainedusing atmos- The time x taken to reach [1 - (l/e)] of the equilibrium
pheric modelswith a simplified domain, the neglect of responseis C/Z. The oceanshave an averagedepthof
seasonalvariations,and the very coarseresolutionocean about 3500 m. If the warming were mixed uniformly
modelsusedto date. Coupledocean-atmosphere models throughout
thisdepth,
C = 1.4x 10•øJ m'2,andfrom(2)
incorporatingrealistic geographyhave been developed the time scalefor approachto equilibrium,x, would be
[Manabe et al., 1979; Washingtonet al., 1980; Gates et about
440years,
assuming
)•= 1W m'2K'•. Theocean
is
al., 1985; Foreman, 1986; Sausenet al., 1988], but at the
heatedat the surfaceand so is stablystratified. Hence the
time of writing, no resultsfrom equilibriumstudiesof the
initial warmingis likely to be confinedto the upperlayers.
effectsof increased CO2 havebeenpublished in theopen The observedseasonalwell-mixed layer at the ocean
literature. S. Manabe(personalcommunication,1987) has, surfacehas a depth of order 100 m, corresponding to a
however, carriedoutanexperiment withenhanced CO2in response time x of order 10 years.
which he finds that the warming in the southernhemi- Thesetwo estimatesprovideupperand lower limits to
sphereextratropicaloceansis substantially
lessthanfound thethermalresponse time. The actualresponse dependson
in studieswhich neglect changesin ocean circulation. the efficiency with which heat is propagateddownward.
Broecker [1987] has suggestedthat changesin ocean Vertical mixing in the oceanis achievedby a varietyof
circulationcouldproduceabruptlocal changesin climate. processes, including large-scale vertical circulations,
Thus improved modeling of the ocean remains a high turbulencedue to inertial wavesor wind mixing, and, in
priorityfor research. high latitudes,convectiveoverturning. The large-scale
One oceanic effect of considerableimportance is motionscanbe modeledexplicitly,but the otherprocesses
possibleincreasesin sea level. Two climatic processes cannotbe resolvedby a generalcirculationmodelgrid and
may contribute: thermalexpansionof the oceanand the so must be parameterized.Attemptshave been made to
meltin
g of landice,including
thaton Greenland
and verify thefidelityof theseparameterizationsby comparing
Antarctica. The contributionfrom thermal expansion the observed and simulated evolution of the distribution of
dependson the magnitudeof the oceanicwarmingand its tritium releasedat the ocean surfacefollowing nuclear
distributionwith latitude and depth. The net effect of weapontestsin theearly 1950s[Sarmiento,1983].
changesin land ice are uncertain;increasedmeltingover It is possible,however,thata parameterization
whichis
Greenland would tend to increase sea level, but the suitablefor a passivetracer suchas tritium may not be
enhanced accumulation of snow over the cold Antarctic appropriatefor heat. Furthermore,mostof the dispersion
plateaumay more than offset increasedmelting at lower of tritium appearsto have occurredalong, rather than
levels aroundthe periphery. Estimatesof the increasein across,isopycnal(constantdensity)surfaces.This results
sealeveldueto doublingCO2varyconsiderably andare in an apparentverticaldiffusion,as the isopycnalsurfaces
based either on simple models or on extrapolationof are slightlytilted from the horizontal. Temperatures are
existingdata. Bolin et al. [1986] quotea rangeof 20-165 relativelyconstantalongisopycnals,andhencethe vertical
cm with a most probablevalue of 80 cm, with the main mixing of heat dependsmore on cross-ispoycnal mixing,
contributioncomingfrom thermalexpansion of theocean. as opposedto along-isopycnal mixing which is dominant
in the case of tritium.

6. HOW WILL THE CHANGES EVOLVE z. From(3) it canbeseen

So far, we have only consideredthe equilibrium
response of climateto increasesin tracegases.In practice,
the trace gasesand the associated radiativeheatingare
increasing steadily,whereasthe climaticresponse will be
slowedby the large thermalinertiaof the oceans. Since
the effectivethermalinertiaof the oceansdependson the
thattheresponse
timedepends
on both)• andtheeffectiveheatcapacityC. Hansenet al.
[1984], using a one-dimensionalbox diffusion ocean
model,obtainedresponse
timesof 27, 55, and 102 years,
corresponding
to 3.equalto2.1,1.4,and1.0W m'2K'•,
respectively.This quadraticdependence of the response
timex on 1/)•, the inverseof the climatesensitivity,
has
also been reproducedby Wigley and Schlesinger[1985]
134 ß REVIEWS OF GEOPHYSICS / 27,1 Mitchell: "GREENHOUSE" EFFECT

usinga similar box diffusionmodel. Smallervaluesof A,
lead to longertime scalesx, allowingdiffusionof heat to
greater depths and hence increasingthe effective heat
capacityC of the oceanequation. This is equivalentto a
furtherincreasein thetimescalex (= C/•; equation(3)).
Box diffusionmodelsproducea uniformtemperature
with depthat equilibriumandhencecannotreproducethe
observedgloballyaveragedverticaltemperature profile.
To remove this shortcoming, one-dimensionalbox
diffusionmodelshavebeenrun, includinga background
upwelling velocity. This increasesthe rate of transient
responseof the upperoceanfor the sameA,and vertical
mixing coefficients[Harvey, 1986] and also reducesthe
sensitivityof the response
time to 3.. Clearly,to takeinto
accountadvectiveand regional effects, one has to use a
moresophisticated
oceanmodel.
Spelmanand Manabe [1984] rerantheir equilibrium4
x CO2 experiment(see section5.3.3), removingthe
devicesusedto acceleratethe convergenceto equilibrium
and"switching
on"theincrease
in CO2. After25 yearsthe
zonallyaveragedtemperature changes werecomparedwith
thoseobtainedat equilibrium. The fractionalresponsein
temperaturein the atmosphereand oceanwas generally
uniform with latitude (Figure 16) and consistentwith a

25
2O
15

205"'-'••'-'"'-'.'.•.:
'- .......'•.'.'.'.0
-'}.-----'--••
2•
• • 10
3•
E 350 •'- ,

515
5

68O

83O
½
Mean
990 sea level

2OO
1- i 4OO
6OO
8OO
I I I 1000

35O

515

940 !!iiii!!iiiiiii!iii!.O
7:.:.:.:.:.:.:.:.
:::::::::::::::::::::::::::::::::::
...............................
.................... .o......o.o......O..oO...o........oOo.,.....o..O..o ...... 1 Mean
99 ========================
' ...................... sea level

200
04
400
01 600
8OO
1000
90 øN 75 o 60 o 45 o 30 o 15 o 0 ø

Latitude

Figure 16. (Top) Height-latitudediagramof zonally averaged is an extradashedcontourat 3.5 K. Also shownis the percent-
changes
in temperature
25 yearsafterquadrupling
CO2 amounts age of the equilibrium response(contoursevery 10% with
in an idealizedcoupledoceanmodel. Contoursare at -8 K, -4 K, increases
greaterthan50% stippledheavily)[fromSpelmanand
-1 K, andthenevery 1 K upward. Areasof decreaseare lightly Manabe, 1984].
stippled,andwarminggreaterthan4 K is heavilystippled.There
Mitchell' "GREENHOUSE"EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 135

responsetime 'cof about25 yearsusing(3). The greatest 3. There is a lack of knowledgeconcerningthe

penetrationof the warminginto the oceanoccursnear 55ø. effective thermal inertia of the oceans. There is consider-
In higherlatitudes anexcess of precipitation
overevapora- ableuncertaintyin the degreeto whichthe thermalinertia
tion produceda stablelayer of fresh,lessdensewater at of the oceanwill slowthe warming. This is compounded
the surfaceandconfinedthe warmingto the surface. The by the uncertaintyin the estimatesof 3,. As discussedin
oceanicresponseis discussed in moredetailby Bryan and section6, the responsetime of the idealized system
Spelman [1985]. Schlesinger[1986] estimatedthat the described in section5 isC/X= CAT•/AQwhere AT•isthe
thermalresponsetime of the oceanwas 75-100 yearson equilibriumtemperature change.As a result,thelargerthe
thebasisof a 16-year"switchon"experimentwith doubled equilibrium response, thesmallerthefractionalresponse in
CO2 carriedoutusinga coupled ocean-atmosphere model a giventime (equation(3)), so that the magnitudeof the
which includedthe diurnalcycleand realisticgeography. response on a shorttimescaleis onlyweaklydependent on
Harvey [1986] reanalyzedSchlesinger's [1986] resultsand the magnitudeof the equilibriumresponse.As a conse-
arrived at a time scalenearerto 50 years. An accurate quence,a small range in the magnitudeof the current
estimateof the thermalresponsetime is essentialfor the observedresponseis compatiblewith a large range of
detection of man-induced climate change, which is equilibriumresponses.Thus the observedtemperature
discussed later. changeswill only provide a weak constrainton the
Spelmanand Manabe [1984] point out that if the magnitudeof climatesensitivity.(This in turn meansthat
characteristic time scaleof the oceanicresponseis less one can only make a roughestimateof the final equilib-
thanthatof theincreasein tracegases,thenthe distribution rium warmingwhichwouldoccurwith no furtherincreases
of the zonallyaveragedtemperature responseto a gradual in tracegases.)
increasein tracegasesmay resemblethe distributionof the The magnitudeof the expectedwarmingto date has
equilibrium response. Hence both the rate at which been estimatedusingone-dimensional models[Hansenet
climate respondsto increasesin trace gases and the al., 1981;Wigleyand Schlesinger, 1985]. If one ignores
geographicaldistributionof the time-dependent response the increase in chlorofluorocarbons,which has been
may dependcritically on the magnitudeof the feedback confinedto the last two decades, the increasein heating
strength 3,. since 1880isabout 1.7W m'2,equivalent totheeffect of
increasing
CO2 from250ppmto itscurrent
value.Onthe
7. WHEN WILL WE BE ABLE TO DETECT THE basis of Wigley and Schlesinger'sassumptionsthe
CLIMATIC EFFECTS? warmingto date rangesfrom just under 0.5 K (large
vertical
mixingcoefficient,
)•= 2.0W m'2K'•) toover1 K
Given that the estimatedincreasein trace gasessince
(small
vertical
mixingcoefficient,
3,= 1.0W m'2K4).
1860is equivalent
to a 40% increase
in CO2,it is pertinent
0.5
to ask if the climatic effects can be detected in the
observationalrecord. The answer to this question is
obscured by the accumulated effects of several
uncertainties. øC 0.0

1. There are uncertainties

in the preindustriallevelsof
trace gases,particularlycarbon dioxide, methane,and
troposphericozone(section4). For example,the rangeof
theestimates
of preindustrial
CO2 concentration,
260-290
1900 1920 1940 1960 19 0 2000
ppm,introduces
anerrorof + 0.3W m'2 in thecurrent Year

greenhouse effect(Table3).
2. Thereare uncertainties in the sensitivityof climate. Figure17. Observed globalsurfaceair temperature
anomalies,
Estimatesof theequilibriumsensitivityof climatediffer by 1901-1987 [Joneset al., 1988]. Copyright 1988 Macmillan
a factorof 3 [Bolin et al., 1986]. Much of this uncertainty MagazinesLtd;reprintedwith permission.
is due to the uncertaintyin cloud feedbacks. Current
modelsimulations indicatea feedbackstrength3,of about
1 W m'2 K'• whichcouldbeincreased•to 2.0 W m'2 K'• if 4. There is considerablevariability in the observed
the net cloudfeedbackis neutral(seeTable 4). Assuming temperature record.Globalmeansurface temperaturehas
AQ= 1.9W m'2 (Table3) anda feedback strengthof increasedby about0.4 K since1900 (seeFigure 17 [from
between 1.0and2.0 W m'2 K'•, wecanestimate thatthe Jones et al., 1988]). The warmestyear on record was
equilibriumwarmingdueto increases in tracegasesto date 1987. The increase has been far from uniform, with
lies between 1.0 and 1.9 K. markedyear to year variability and a period of cooling
136 ß REVIEWSOF GEOPHYSICS / 27,1 Mitchell: "GREENHOUSE" EFFECT

after 1940. Clearly,thiscoolingis not due to increasesin The atmosphericconcentrations

of manygreenhouse
trace gases. It may be a manifestationof the natural gases,includingcarbondioxide, tropospheric
ozone,
variabilityof the ocean-atmosphere system,or it may be methane,and nitrousoxide, havebeenobservedto increase
due to external causes such as a decrease in solar overthelastfew decades.Theseincreases aregenerally
luminosityor an increasein volcanicaerosols.Prior to this believedto be dueto increased
industrial
andagricultural
period,land data analyzedby P. D. Joneset al. [1986] activity, thoughall these gaseshave natural chemical
suggestthat global mean temperatures were relatively cycles,and with the exceptionof carbondioxide our
constant from 1860 to 1900, whereas marine air data understanding of the detailedcausesof the increasesis
[Parkeret al., 1988]suggest thattemperatures weretwo or lacking. In the pasttwo decades, othergases,notably
threetenthsof a degreewarmerin 1860thanin 1900. chlorofluorocarbons, havebeenaddedto theatmosphere.
The observedwarmingfrom 1900 is slightlysmaller Although theyarepresent in verysmallquantifies, theyare
than the lower estimatesgiven above. However, the severalordersof magnitude moreeffectiveper molecule
waxmingdue to tracegasesmay be maskedby natural thancarbon dioxidein trapping thermal radiation,asthey
variations and the effects of external factors. Thus one absorbstronglyat wavelengthswhere there is little
cannotconcludethat theoreticalestimates of the warming absorption by theothergases (theatmospheric "window").
are too high, nor can one concludethat the climaticeffects The direct radiativeeffectsof increasingmajor
of tracegaseshavebeendetected. greenhouse gasesare well known. It hasbeen estimated
Estimatesof the warmingdue to increasedgreenhouse thattheadditional heatingdueto increases in tracegases
gas concentrations over the next 50 yearsare obviously since theIndustrial Revolution isabout 2 W m-2,andinthe
dependenton the projectedconcentrations of thesegases. next50 yearsthiscouldeasilyincrease by a further4 W
First,a continuedwarmingis expectedasclimatecomesto m-2,though thisfigureisheavily based onextrapolation of
equilibriumwith the currentconcentrations of tracegases. existing trends.Theclimatic response maybeamplified or
Second, the estimatesof further increasesin trace gas reducedby changesin watervapor,snowand ice, cloud,
concentrations over the next 50 years(Table 3) imply an andsoon. Estimates of thesensitivityof surfacetempera-
additionalwarming,which on reachingequilibriumwould ture to enhanced radiative heating vary from lessthan0.5
rangefrom 1.8 to 3.5 K. A preciseestimateof the actual K W4 m2toover 1 K W-• m2. Mostofthisuncertainty
changein surfacetemperatureover the next 50 years is arises from difficultiesin representing cloud effects
precludedby uncertainties in the sensitivityof climateand faithfullyin numerical models.Estimates of theequilib-
the effect of the thermal inertia of the ocean, but an riumwarmingduetochanges since1860rangefrom 1 to 2
additionalwarmingof 1 K or moreseemslikely. If this is K, but the actualwarmingis estimatedat abouthalf these
the case,then an unambiguousdetectionof the warming values because of the thermal inertia of the oceans. This is
due to trace gasesshouldbe possiblewithin the next two not inconsistentwith the observationalrecord, but one
decades. Even if the trace gas amountsremain constant cannotprovethata greenhouse waxminghasbeendetected
thereafter,the warming will continueuntil the oceanand becauseof the temporalvariabilityof climate. Estimates
atmosphere reachthermalequilibrium. of thechangeoverthenext50 yearsaredependent on the
projectedincreases in tracegasesand the assumptions
madeconcerningclimatesensitivityand thermalinertia,
buta warmingof 1 K or moreseemslikely.
8. SUMMARY AND CONCLUDING REMARKS The possibleclimaticeffectshavebeenstudiedusing
generalcirculation
modelsof the atmospherecoupledto
The surfaceof our planetremainshabitablebecauseof simplified
models of theoceanandseaice. Thefollowing
the presenceof water vapor,carbondioxide,and ozone in physicalresponses(wherereported)are commonto all
the atmosphere. These gasesabsorb and emit thermal recentexperiments.
radiation,raisingtheglobalmeansurfacetemperaturefrom 1. Thereis a warmingof the troposphere andsurface.
225 K (-18øC) to 288 K (15øC). The absorptionoccurs In thetropicsthewarmingincreases withheightandvaries
because of excitation of vibrational and rotational molecu- littlewithseason.Thelargestwarmingoccursoverseaice
lar states. Becausetropospherictemperaturesgenerally and surrounding regionsin winter,becauseof theremoval
decreasewith height,the additionof greenhouse gasesto or thinningof seaiceandtheinherent low-levelstabilityin
the atmosphereraisesthe mean level from which thermal highlatitudes.The warmingis generallya minimumover
radiationescapesto space,hencereducingthe temperature sea ice in summer.
of emissionand the rate of emission. Troposphericand 2. Snowmelt occurs earlier, and sea ice is less
surfacetemperaturesincreaseuntil the outgoingthermal extensive.
radiationagainbalancesthe incomingsolarradiation. This 3. The increasein atmospheric
temperature
is accom-
is knownas the greenhouseeffect. paniedby an increasein atmospheric
watervaporcontent,
Mitchell: "GREENHOUSE" EFFECT 27,1 / REVIEWSOF GEOPHYSICS ß 137

further enhancingthe radiative heatingat the surface. Sustainedresearchover the coming decadewill be
Global mean rates of precipitationand evaporation requiredto reducethe currentuncertaintiesin the predic-
increase. tions of climate change. Numerical modelsprovide the
4. The geographicaldistributionof precipitation mostpromisingmeansof determiningthe climatic effects.
changes is far from uniform.Thereis a generalincreasein Whereas numerical weather prediction models can be
precipitationand runoff in high latitudes. Precipitation calibrated by verification of short-term forecasts,our
increases in low latitudes,thoughlocallythereare regions confidencein the reliabilityof climatemodelsdependson
of increaseanddecreasewhichvary from modelto model. the extentto whichtheyarebasedon scientificprinciples.
In most studiesto date, there is a reductionin soil moisture The potentialwarmingdue to man-madechangesover
over the northern mid-latitude continents in summer. the last 100 yearsis 1-2 K; hencethe questionis not will
5. The magnitude of theresponse is greatlyenhanced the changesoccur,but when and where will the changes
in most simulationswith model-generatedcloud. A occur? Cooperationbetweenmanydifferentdisciplinesis
decrease in total cloud and an increase in mean cloud neededto predictthe futureconcentration of gasesand the
heightbothcontributeto thisenhancement. consequent climaticandeconomiceffects.
Thereare severalprioritiesfor futureresearch. First,
we need to reducethe uncertaintyin the magnitudeof
climate sensitivity. Here the major shortcomingis our ACKNOWLFDGMFNTS. Andrew Gilchrist, William
Ingram, and Geoff Jenkinsmade usefulsuggestions on the first
poor understandingof the processesgoverning the
draft of this paper. Dick Wetheraid,Marcia Neugebauer,and
formationandradiativepropertiesof clouds,compounded two other anonymousrefereescontributedconstructivecom-
by thefact thatcloudsare notresolvedby the modelgrid. mentswhichhavebeenincorporated in the test.
The radiativepropertiesof water cloudsdependon the Ann Henderson-Sellers was the editor responsiblefor this
cloud water content, which has been parameterizedin paper. She thanksR. E. Dickinsonand R. Wetheraidfor their
generalcirculation models,andtheeffectivedropletradius, help in evaluatingits technicalcontentandE. Clifton for serving
whichhasnot. Even lessis knownaboutthe propertiesof as a cross-disciplinary referee.
ice clouds,an areawherethereis a requirementfor further
experiment andobservations.
REFERENCES
The secondmajor sourceof uncertaintywhich needs
attentionis the degreeto which the oceanswill slow the Arrhenius, S., On the influence of carbonic acid in the air on the
expectedwarming. Whereascurrentglobaloceanmodels temperature of theground,Philos.Mag., 41,237-276, 1896.
can reproducethe broad featuresof the oceancirculation, Augustsson,T., and V. Ramanathan,A radiative-convective
the parameterization of convectionand othersubgridscale modelstudyof theCO2 climateproblem, J. Atmos.Sci.,34,
488-451, 1977.
verticalmixingare notwell developed. Blake, D. R., andF. S. Rowland,Continuingworldwideincrease
A third factor which hinders the detection of climate in troposphericmethane; 1978 to 1987, Science, 239,
1129-1131, 1988.
changeis a lackof knowledgeof theinherentvariabilityof
Bohren,C. F., Commenton "Cloudopticalthicknessfeedbacks
the Earth's climate on a time scale of several decades.
in theCO2climateproblem"
byRichard
Somerville
andL. A.
There is muchto be learnedin thisrespectfrom analyzing Remer,J. Geophys.Res.,90, 5867, 1985.
long simulations of coupledocean-atmosphereGCMs and Bolin, B., B. R. Doos, J. Jaeger,and R. A. Warrick (Eds.), The
Greenhouse Effect,ClimaticChangeandEcosystems, SCOPE
from a carefulstudyof observations.Oncethe magnitude œep.,29, 1-541, 1986.
of the inherentvariabilityof climatehasbeendetermined, Broecker, W. A., Unpleasantsurprisesin the greenhouse?,
and the temporalevolutionof the warmingdue to trace Nature, 328 123-126, 1987.
gasescan be estimatedmore accurately,the chanceof Bryan, K., Accelerating the convergenceto equilibrium of
ocean-climatemodels, J. Phys. Oceanogr.,14, 666-673,
detectingthe effect of increasesin trace gaseswill be 1984.
greatlyenhanced. Bryan, K., and M. J. Spelman,The ocean's responseto a
As yet, little can be said concerningthe regional CO2-induced
warming,J. Geophys.
Res.,90, 11, 679-11,
688, 1985.
changesin climate,particularlychangesin precipitation.
Chandrasekhar,S., Radiative Transfer, 393 pp., Clarendon,
The horizontal(andperhapsvertical)resolutionof current Oxford, 1950.
climatemodelsandthe accuracyof physicalparameteriza- Craig, H., andC. C. Chou,Methane: The recordin thepolar ice
tions must be increased before realistic simulations of the cores,Geophys.Res.Lett., 9, 1221-1224, 1982.
Derwent,R. G. (Ed.), Ozonein the United Kingdom,report, 112
regionaldistributionof climateand climatechangecan be pp., Photochem.Oxidants Rev. Group, IJ.K. Dep. of the
made. Substantialimprovementin representations of the Environ.,Her Majesty'sStationaryOffice, London,1987.
boundarylayerandsurface,particularlyvegetation,will be Dickinson, R. E., Modeling evapotranspirationfor three-
dimensional climate models, in Climate Processes and
neededif modelsare to be capableof predictingthe
ClimateSensitivity,Geophys.Monogr.Ser., vol. 29, editedby
behaviorof thoseparametersrequiredin climate impact J. E. Hansen and T. Takahashi, pp. 58-72, AGU,
assessments. Washington,D.C., 1984.
138 ß REVIEWSOF GEOPHYSICS/ 27,1 Mitchell: "GREENHOUSE"EFFECT

Dickinson,R. E., How will climatechange?, The Greenhouse Manabe,S., and R. J. Stouffer,Sensitivityof a globalclimate
Effect,ClimateChangeandEcosystemss, editedby B. Bolin, modelto an increasein the CO2 concentration in the
B. R. Doos,J. Jaeger,andR. A. Warrick,SCOPERep.,29, atmosphere, J. Geophys.Res.85, 5529-5554, 1980.
206-270, 1986. Manabe,S., and R. T. Wetherald,Thermalequilibriumof the
Dickinson,R. E., and R. J. Cicerone,Futureglobalwarming atmosphere with a givendistribution
of relativehumidity,J.
from atmospheric tracegases,Nature,319, 109-115, 1986. Atmos.Sci., 24, 241-259, 1967.
Foreman, S. J., Oceanandatmosphere Meteorol.Mag., Manabe,S., and R. T. Wetheraid,The effectsof doublingthe
interact!,
115, 358-361, 1986. CO2 concentration on a general circulation
model,J. Atmos.
Gates,W. L., Y. J. Han, and M. E. Schlesinger, The global Sci., 32, 3-15, 1975.
climate simulatedby a coupledatmosphere-ocean general Manabe, S., and R. T. Wetheraid, On the distributionof climate
circulationmodel, preliminaryresults,in CoupledOcean change resulting fromanincrease in theCO2 content of the
Atmosphere Models,ElsevierOceanogr.Ser.,vol. 40, edited atmosphere, J. Atmos.Sci.,37, 99-118, 1980.
by J. C. J. Nihoul,pp. 131-151, Elsevier,New York, 1985. Manabe,S., and R. T. Wetherald,Largescalechanges of soil
Hansen,J. E., D. Johnson,A. Lacis, S. Lebedeff,P. Lee, D. wetnessinducedby an increasein atmosphericcarbon
Rind, andG. Russell,Climateimpactof increasing atmos- dioxide,J. Atmos.Sci., 44, 1211-1235, 1987.
phericcarbondioxide,Science,213, 957-966, 1981. Manabe,S., K. Bryan, and M. J. Spelman,A globalocean-
Hansen,J., G. Russell,D. Rind, P. Stone,A. Lacis,S. Lebedeff, atmosphereclimate model with seasonalvariation for future
R. Ruedy,and L. Travis,Efficientthree-dimensional
global studiesof climate sensitivity,Dyn. Atmos. Oceans, 3,
models for climate studies: Models I and 1I, Mon. Weather 393-426, 1979.
Rev., 111, 609-662, 1983. Manabe,S., R. T. Wetherald,andR. J. Stouffer,Summerdryness
Hansen,
J., A. Lacis,D. Rind,G. Russell,
P. Stone,I. Fung,R. dueto an increase
of atmospheric
CO2 concentration,
Clim.
Ruedy, and J. Lerner,Climate sensitivity: Analysisof Change,3, 347-385, 1981.
feedback mechanisms, in Climate Processes and Climate Marland,G., andR. Rotty,Carbondioxideemissions
from fossil
Sensitivity,
Geophys.
Monogr.Ser.,vol. 29, editedby J. E. fuelCO2emissions,
report,U.S.Dep.of Energy,
Washington,
HansenandT. Takahashi,
pp. 130-163,AGU, Washington, D.C., 1983. (Available as NTIS DOE/NBBO036 from Nail.
D.C., 1984. Tech.Inf. Serv.,Springfield,Va.)
Harvey,L. D., Effectof oceanmixingon the transientclimate Mitchell,J. F. B., Climatesensitivity
andpastclimates;
Evidence
response to a CO2 increase:Analysisof recentresults,J. fromnumerical studies,
in AbruptClimaticChange,editedby
Geophys.Res.,91, 2709-2718, 1986. W. H. Bergerand L. D. Labeyrie,pp. 383-398, D. Reidel,
Hecht,A.D. (Ed.),PaleoclimaticAnalysisandModelling, 445 Hingham,Mass.,1987.
pp. JohnWiley, New York, 1985. Mitchell,
J. F. B., andG. Lupron,
A 4 x CO2 experiment
with
Ingram,W. J., C. A. Wilson,andJ. F. B. Mitchell,Modeling prescribed
changes
in seatemperatures,
Prog. Biometeorol.,
climatechange:An assessment
of seaice andsurfacealbedo 3, 353-374, 1984.
feedbacks,J. Geophys.
Res.,in press,1989. Mitchell,J. F. B., and D. A. Warrilow,Summerdrynessin
Jones,P. D., T. M. L. Wigley, and P. B. Wright, Global northern mid-latitudesdueto increasedCO2,Nature,330,
temperaturevariations,1861-1984, Nature, 322, 430-434, 238-240, 1987.
1986. Mitchell,J. F. B., C. A. Wilson,W. J. Ingram,and W. M.
Jones,P. D., T. M. L. Wigley, C. K. Folland,D. E. Parker,J. K. Cunnington,Numerical studiesof the effect of increased
Angell, S. Lebedef,and J. E. Hansen,Evidencefor global atmosphericcarbondioxideon climate,final report,contract
warmingin thepastdecade,Nature,338, 790, 1988. CL-114-UK(H), Comm. of the Eur. Communities,Meteorol.
Jones,R. L., J. A. Pyle, J. E. Hardes,A.M. Zavody,J. M. Office,Bracknell,England,1986.
Russell1II, andJ. C. Gille, The watervapourbudgetof the Mitchell,
J.F.B.,C.A. Wilson,
andW. M. Cunnington,
OnCO2
stratosphere
studiesusingLIMS andSAMS satellitedata,J. climatesensitivityandmodeldependence
of results,Q. J. R.
œ.Meteorol. Soc., 112, 1127-1144, 1986. Meteorol. Soc., 113, 293-322, 1987.
Keeling, C. D., R. B. Bacastown,and T. P. Whoif, Measure- Moller,F.,Ontheinfluence
ofchanges
inCO2concentrations
in
ments of Concentration of carbon dioxide at Mauna Loa air on the radiative balanceof the Earth's surfaceand on the
Observatory,Hawaii, in Carbon Dioxide Review, 1982; climate,J. Geophys.Res.,68, 3877-3886, 1963.
updatedby C. D. Keeling, D. J. Moss, and T. P. Whoif, National ResearchCouncil, Carbon Dioxide and Climate: A
Measurements of the concentrations
of atmosphericcarbon SecondAssessment, 72 pp., NationalAcademyPress,
dioxideat MaunaLoa Observatory,Hawaii 1958-1986,final Washington, D.C., 1982.
report for the CarbonDioxide Informationand Analysis Neftel,A., E. Moor, H. Oeschger, andB. Stauffer,Evidence
Center, subcontract 19X-89696L,Mar•tin-MariettaEnergy frompolaricecores fortheincrease in atmosphericCO2 in
Systems,Inc., Oak Ridge,Tenn.,April 1987. Latestupdate, thepasttwo centuries,Nature,315, 45-47, 1985.
NationalOceanicand Atmospheric Administration/Scripps Nordhaus, W. D., andG. W. Yohe,Future pathsof energyand
Institution
of Oceanography, Boulder,Colo.,May 1988. carbon dioxideemissions,
in Changing Climate,pp.87-153,
Lal, M., and V. Ramanathan,Effects of moist convectionand NationalAcademyof Sciences, NationalResearch Council,
water vapourradiativeprocesses on climatesensitivity,J. Washington,D.C., 1983.
Atmos.Sci., 41, 2238-2249, 1984. Paltridge,G. W., and C. M. R. Platt, RadiativeProcessesin
Lowe, D.C., C. A.M. Brenninkmeijer, M. R. Manning,S. Meteorology andClimatology, Dev.Atmos.Sci.Ser.,vol. 5,
Sparks and G. Wallace, Radiocarbon determination of 318 pp., Elsevier,New York, 1976.
atmospheric
methaneat BaringHead,New Zealand,Nature, Parker,D. E., C. K. Folland,and M. N. Ward, Worldwide
332, 522-525, 1988. surfacetemperaturevariations1984:1987in relationto E1
MacCracken,M. C., and F. M. Luther(Eds.),Detectingthe Nino,tropical
rainfallandlonger
termtrends,
in Proceedings
climaticeffectsof increasing
carbondioxide,Rep.DOE/ER- of Second
WMOWorkshop ontheDiagnosisandPrediction
0235,U.S.Dep.of Energy,Washington, D.C., 1985. of Monthlyand Seasonal
AtmosphericVariationsOver the
Manabe,S., andK. Bryan,CO2-induced
change
in a coupled Globe,WorldMeterological
Organization,
Geneva,1988.
ocean-atmospheremodelanditspaleoclimatic
implications,
J. Plass,G. N., The carbon
dioxidetheoryof climaticchange,
Geophys.Res.,90 11,689-11,707, 1985. Tellus, 8, 140-154, 1956.
Mitchell: "GREENHOUSE"
EFFECT 27,1 / REVIEWS
OF GEOPHYSICS
ß 139
Ramanathan,V., R. J. Cicerone,H. B. Singh, and J. T. Kiehl, Tyndall,J., On theabsorption
andradiation
of heatby gasesand
Tracesgastrendsandtheirpotentialrole in climatechange,J. yapours, and on the physical connexion of radiation,
Geophys.Res.,90, 5547-5566, 1985. absorptionand conduction,
Philos.Mag., 22, 169-194 and
Roeckner,E., U. Schlese,J. Biercamp,and P. Loewe, Cloud 273-285, 1861.
opticaldepthfeedbacksand climatemodelling,Nature, 329, Volz, A., and D. Kley, Evaluationof the Montsourisseriesof
138-139, 1987. ozonemeasurements madein thenineteenthcentury,Nature,
Sarmiento,J. L., A simulationof bomb tritium entry into the 332, 240-242, 1988.
AtlanticOcean,J. Phys.Oceanogr.,13, 1924-1939, 1983. Wardlow,D. A., A. B. Sangster,
andA. Slingo,Modellingof
Sausen,R., K. Barthel, and K. Hasselmann,Coupled ocean- land surfaceprocessesand their influenceson European
atmospheremodels with flux corrections,Clim. Dyn., 2, climate,Met 0 20 Dyn. Climatol.Tech.Note 38, Meteorol.
145-163, 1988. Office,Braeknell,England,1986.
Schlesinger,M. E., Feedbackanalysisof resultsfrom energy Washington,
W. M., and G. A. Meehl, A seasonalcycle
balance and radiative-convectivemodels, Projecting the experiment
on the climatesensitivitydue to a doublingof
Climatic Effectsof IncreasingAtmosphericCarbonDioxide, CO2withanatmospheric general
circulation modelcoupled to
edited by M. M. MacCracken and F. M. Luther, NTIS a simplemixed layer oceanmodel,J. Geophys.Res., 89,
DOE/ER-2037, pp. 280-319, Nail. Tech. Inf. Serv., 9475-9503, 1984.
Springfield,Va., 1985. Washington,
W. M., A. J. Semtner,D. J. Knight,and T. A.
Schlesinger,M. E., Equilibriumand transientclimatic warming Mayer, A generalcirculationexperimentwith a coupled
induced
by increased
atmospheric
CO2,Clim.Dyn.,1, 35-51, atmosphere,
oceanandseaicemodel,J. Phys.Oceanogr.,10,
1986. 1887-1908, 1980.
Schlesinger,
M. E., and J. F. B. Mitchell, Model projectionsof Watson,R. T., F. S. Rowland,and J. Gille, Ozone Trends Panel
equilibrium
climaticresponse
to increased
CO2 concentration, pressconference
report,NASA Headquarters,
Washington,
D.
Projectingthe ClimaticEffectsof IncreasingCarbonDioxide, C., March 15, 1988.
edited by M. C. MacCrackenand F. M. Luther, Rep. Weiss, R. F., The temporal and spatial distributionof
DOE/ER-0237, pp. 81-147, U.S. Dep. of Energy, troposphericnitrousoxide,J. Geophys. Res.,86, 7185-7195,
Washington,D.C., 1985. 1981.
Schlesinger,M. E., and J. F. B. Mitchell, Climate model Wetheraid,R. T., and S. Manabe, Influenceof seasonalvariation
simulationsof the equilibriumclimaticresponseto increased uponthesensitivity of a modelclimate,J. Geophys.
Res.,86,
carbondioxide,Rev.Geophys.,25, 760-798, 1987. 1194-1204, 1981.
Schlesinger,M. E, and Z. Zhao, Seasonalclimate changes Wetheraid,R. T., and S. Manabe,An investigation of cloud
inducedby doubledCO2 as simulated
by the OSU atmos- coverchangein response
to thermalforcing,Clim.Change,8,
phericOCM/mixedlayermodel,Rep. 70, 73 pp., oreg. State 5-24, 1986.
Univ. Clim. Inst., Corvallis, 1987. Wetheraid,
R. T., andS. Manabe,Cloudfeedback
processes
in a
Schneider,
S. H., and S. L. Thompson,
Atmosphere
CO2 and general circulation model, J. Atmos. Sci., 45, 1397-1415,
climate: Importanceof the transientresponse,J. Geophys. 1988.
Res., 86, 3135-3147, 1981. Wigley, T. M. L., FutureCFC concentrations
undertheMontreal
Sellers,P. J., Y. Mintz, and A. Dalcher,A simplebiosphere Protocoland their greenhouse-effect implications,Nature,
model (SiB) for use within generalcirculationmodels,J. 335, 333-335, 1988.
Atmos.Sci., 43, 505-531, 1986. Wigley,T. M. L., andM. E. Schlesinger,
An analyticalsolution
Somerville,R. C. J., and L. A. Remer,Cloud opticalthickness fortheeffectof increasing
CO2 onglobalmeantemperature,
feedbacks
in theCO2 climateproblem,
J. Geophys.
Res.,89, Nature, 315, 649-652, 1985.
9668-9672, 1984. Wilson,C. A., andJ. F. B. Mitchell,
A doubled
CO2 climate
Spelman, M. J., and S. Manabe, Influence of oceanic heat sensitivity
experiment
with a GCM includinga simpleocean,
transportuponthesensitivityof a modelclimate,J. Geophys. J. Geophys.Res.,92, 13,315-13,343, 1987a.
Res., 89, 571-586, 1984. Wilson, C. A., and J. F. B. Mitchell, Simulatedclimate and
Stephens,G. L., Radiationprofilesin extendedwater clouds,lI, CO2-induced
climatechangeover westernEurope,Clim.
Parameterization schemes, J. Atmos. Sci., 35, 2123-2132, Change,10, 11-42, 1987b.
1978. World Meteorological
Organization,Atmospheric ozone1985,
Trabalka, J. R., (Ed.), Atmosphericcarbon dioxide, Rep. GlobalOzoneRes.Monit. Proj. Rep. 16, 1095pp., Geneva,
DOE/ER-0239, U.S. Dep. of Energy, Washington,D.C., 1985.
1985.
Trabalka,J. R., J. A. Edmonds,R. H. Gardner,J. Reilly, and D.
E. Reichle,Atmospheric
CO2 projections
with globally
averagedmodels,in The ChangingCarbonCycle: A Global
Analysis,editedby J. R. Trabalkaand D. E. Reichle,pp. J. F. Mitchell,Met 01 20, Meteorological
Office,London
534-560, Springer-Verlag,
New York, 1986. Road,Bracknell, BerkshireRG122SZ,England.