Reprint Series1 January 1982, Volume 215, pp. 56-58

Transient Climate Response to Increasing

Atmospheric Carbon Dioxide

K. Bryan, F. G. Komro, S. Manabe. and M. J. Spelman

Copyright @ 1982 by the American Association for the Advancement of Science

Transient Climate Response to Increasing

Atmospheric Carbon Dioxide

Abstract. The ocean's role in the delayed response of climate to increasingatmospheric carbon dioxide has been studied by means of a detailed three-dimensional climate model. A near-equilibrium state is perturbed by afourfold, step-function increase in atmospheric carbon dioxide. The rise in the sea surfacetemperature was initially much more rapid in the tropics than at high latitudes.However, the fractional response, as normalized on the basis of the total differencebetween the high carbon dioxide and normal carbon dioxide climates, becomesalmost uniform at all latitudes after 25 years. Because of the influence of a morerapid response over continents, the normalized response of the zonally averagedsurface air temperature isfaster and becomes nearly uniform with respect to latitudeafter only 10 years.

Long-term measurements (1) at sever-al stations provide firm evidence that theatmospheric concentration of CO2 is in-creasing on a global scale. The increaseis usually ascribed to the burning offossilfuels, but changes in the carbon contentof the biosphere may also contribute.Studies of the impact of increasing atmo-spheric CO2 on climate have tended tofocus on the global equilibrium climatefor different atmospheric CO2 concentra-tions. The results from mathematicalmodels of climate (2) indicate that adoubling of the present CO2 concentra-tion may cause the globally averaged seasurface temperature to increase by 1.50to 4.5°C. Precise results differ from onemodel to another depending on the de-tails of the albedo feedback effects ofsnow, ice. and cloudiness. Recently (2-

has been developed over a period ofyears. The oceanic and atmosphericcomponents have been tested separatelyin independent studies. This was the firsttime that the model has been used tostudy the time-dependent behavior of theclimate system, and it required a com-pletely synchronous integration with re-spect to time of the ocean and atmo-sphere. Our numerical experiment con-sisted of a "switch on" case in which anequilibrium climate solution is perturbedby a step function increase of atmospher-ic CO2. A "switch on" experiment pro-vides information on the response for awide range of time scales, which can beapplied to more realistic scenarios ofatmospheric CO2 buildup. Unfortunate-Iy, our climate model calculation re-quired many hours of calculations on alarge-capacity computer, limiting consid-eration to only a few cases. It is impor-tant that the search for simpler modelsthat are consistent with both tracer data(5) and more detailed climate modelscontinue.

The coupled three-dimensional cli-mate model used in this study is slightlyless complex than that of some earlierstudies (6, 7). The seasonal variation insolar insolation was not included, andthe geometry of land and sea was highlyidealized. To minimize the amount ofcalculation, the model atmosphere andocean was constrained to be triply peri-odic in a zonal direction with mirrorsymmetry across the equator. Land andocean occupy adjacent 6()° sectors oflongitude, extending from the equator toone of the poles. The atmospheric modelused finite differences in the verticalplane, and variables were represented byspherical harmonics in the horizontalplane. The radiation balance, the hydro-logic cycle, and the transport effects ofatmospheric cyclones and anticycloneswere included explicitly. The oceanmodel did not have sufficient resolutionto resolve mesoscale eddies, which arethe dynamic counterparts of cyclonesand anticyclones in the atmosli>here. Theresolution was sufficient, however, topermit the inclusion of the main featuresof the observed wind-driven and thermo-halinecirculation. Twelve levels in thevertical specified as in (7) provided adetailed representation of the ocean ther-mocline. A simple model of sea ice andland snow cover allowed an importantalbedo feedback mechanism. Because ofthe great difficulty of accurately deter-mining parameters with which to repre-sent the cloud distribution, zonally uni-form cloudiness is fixed to the observedclimatology at each latitude.

Since we are interested in the response

SCIENCE. YOLo 215, I JANUARY 1982

4), attention has been drawn to the non-equilibrium climate problem. In otherwords, if an increase in the globallyaveraged temperature will eventuallytake place for a specified increase in theatmospheric CO2 concentration, howrapidly will the change actually occur?The moderating effect of the ocean playsa key role. In a recent report of theNational Academy of Sciences (2), it isestimated that a climatic response toincreasing atmospheric CO2 will have adelay of the order of decades because ofthe thermal inertia of the ocean. In thisstudy, we attempt to investigate thistransient response problem by examin-ing the results from numerical experi-ments with a coupled ocean-atmospheremodel.

The climate model used in this study

).55

0.5'...' .., , ~..."",-v".., " , ...,..", : j, ,..- ~ "-'~

" ' ,","

.-'......"~...: ~- .

". ."'\ ...'..M .

..., ...."" ~,.

.

0.4

0.3

0.2

0.1

1

0.0~""'-L'-.~.- ~~.."

-0.1

0 5 10 15Years

Fig. I. The normalized response of the globally averaged sea surface temperature plotted at 30-day intervals (upper curve); the response for a calibration experiment (lower curve)~

56 0036-8075/82/0101-0056$01.00/0 Copyright @ 1981 AAAS

Table I. Zonally averaged surface air tem-perature of the ocean-atmosphere climatemodel.

Tem-

per-aturediffer-ence

(high -normal

CO2climate)

rC)

Tem-per

aturefor thenormal

CO2climate

(OC)

Ob-served

tem-

per-ature(DC)

Lati-tude

(deg)

800600400200

00

turbed by a sudden quadrupling of theatmospheric CO2. The atmosphere andocean models are numerically integratedin a synchronous mode as opposed to thenonsynchronous, economical methodused to obtain climate equilibrium. Thenormalized response of the globally av-eraged sea surface temperature is shownas a function of time in Fig. I (uppercurve). The normalized response, R,may be defined as

of a quick rise in temperature over thefirst 3 years to nearly 30 percent of thetotal difference between equilibriumstates. This rapid rise is associated withthe heating of the mixed layer and isfollowed by a much slower rise as theeffect of vertical transfer of heat to lowerlevels is felt. The bottom curve of Fig. Ishows the response of a calibration ex-periment in which the initial conditionsare the same as those of the climateresponse experiment but without the in-crease in atmospheric CO2. The calibra-tion experiment shows a downwardtrend in R that is equivalent to 0.5°C over15 years. An analysis of the ocean modelsolution, however, shows that the tem-perature trend .is confined to the nearsurface. Although the climate of themodel is not in perfect equilibrium, thedeparture is relatively small. Evidencefor this is found in the net heating rate ofthe ocean, which is equivalent to a heatflux of -0.06 W/m2 at the ocean surfaceaveraged over 15 years as compared to+6.5 W/m2 for the "switch on" experi-ment.

The results of an earlier study (3), inwhich a Budyko-Sellers climate modeland a passive ocean with a heat capacityvarying with latitude were used, suggest-ed the possibility of a more rapid re-sponse at low latitudes. The zonally av-eraged normalized response as a func-tion of time and latitude (Fig. 2a) sup-ports this idea in the early stage of theexperiment. After 20 years, however, asystematic difference in the normalized

T- ToR=

T~ -To

where T is the globally averaged seasurface temperature, To is its initial equi-librium value, and T ~ is its final equilib-rium value for a fourfold increase inatmospheric CO2, The response consists

of the coupled model as it makes thetransition from one equilibrium state toanother, the first requirement was tocalculate equilibrium climates corre-sponding to normal and high atmospher-ic CO2. The increase in CO2 (four timesnormal) was deliberately chosen to belarge relative to any foreseeable actualchange in the composition of the earth'satmosphere in order to discriminate be-tween COr induced change and back-ground temporal fluctuations of the mod-el climate. The economical method usedto obtain climate equilibrium has beendescribed in (6). Briefly, a nonsynchro-nous method is used to couple the atmo-sphere, upper ocean, and deep ocean toallow for the extremely disparate timescales of the climate system. The deepocean has a time scale of millennia. Theatmosphere has a time scale of about Imonth, whereas time scales of the upperocean lie in an intermediate range. Non-synchronous coupling may be thoughtof as a relaxation procedure to hastenthe convergence to equilibrium. In thisstudy, I year in the atmospheric modelwas taken to correspond to 110 years inthe upper ocean. One year in the upperocean, in turn, was taken to be equiva-lent to 25 years in the deepest levels ofthe ocean.

Convergence to a climatic equilibriumfor the normal and high CO2 cases wasobtained after an integration of the nu-merical models over the equivalent of 6years in the atmosphere. 650 years in theupper ocean, and 16,000 years for thedeep ocean. Zonally averaged surface airtemperatures for the equilibrium statesare shown in Table I. The model climatefor normal CO2 compares quite well withobservations (8) except in high latitudeswhere warming by the active ocean inthe model led to a less extensive snowand sea ice cover than observed. Theexcessive warming at high latitudes oc-curred because the ocean in the climatemodel extended directly to the pole. Ta-ble I also shows that the difference be-tween the zonally averaged surface airtemperatures for the normal CO2 andhigh CO2 climates increased with lati-tude. The pattern of sensitivity was simi-lar to that found in earlier studies (9), butthe amplitude was somewhat less in mid-dle and high latitudes. This reduced sen-sitivity occurs because the present mod-el climate for normal CO2 was warmerand had less snow and ice at high lati-tudes than models without an activeocean.

The determination of two climaticequilibrium states provides the basis fora "switch on" experiment in which thenormal CO2 equilibrium climate is per-

I JANUARY 1982

Fig. 2. Time-latitude variation of the zonally averaged normalized response of (a) sea surfacetemperature and (b) surface air temperature averaged over continent and ocean. The normal-ized response, R, is calculated from Eq. I with the use of zonally averaged temperatures.

57

2595

205

350

515

680

-4-

-8

2015 E~

E~"Qi~

..\!~~0.0)0E<

10

,"

~ , ', I

3.5 II ,, ""-" 5

~.,-I

2.5I

I!

~

=.0

.§Q)

~....Q)

0:.3

"

I,I\

.,, '

"~'830 ,~:" ..3.5

~~o,. :!:!~~~:::i~~"" ~3-= jb-V°

0

~I

6

'\ 0:

900N 450

Latitude

Fig. 3. Latitude-height distribution of the zonally averaged temperature at 10 years minus theinitial temperature, showing the response to a step-function increase in atmospheric CO2. Unitsare degrees Celsius.

750 60' 30" 150

200400

600 :§

800 =Q

1000 GI."1000 c

2000 ~()

13000 0I

4000

5000

induced warming of sea surface tempera-ture is limited to the thin surface layerbecause of the stable stratification. Atabout 600 latitude where the ocean has aweak stratification, the heat anomalyfinds pathways for relatively deep pene-tration. The greater stratification of thetropical ocean prevents much verticalheat exchange at low latitudes.

Our calculations represent a first at-tempt to predict the transient response ofclimate to increasing atmospheric CO2with a model that includes an activeocean. The aim of our study has not beento provide reliable calibration for simplermodels, since the precise results ob-tained depend on the particular geometryand other assumptions inherent in themodel. The importance of our study liesin the greater insight that a more detailedmodel can provide, which in turn cansuggest how simpler models should bedesigned and how the available data canbest be used to test them.

An important finding in our calculationis that the initial response to a sudden"switch on" of atmospheric CO2 is arapid rise of sea surface temperature inthe tropics. After 10 years, however, seasurface temperature rises decrease in thetropics but persist at higher latitudes.The normalized sea surface temperatureresponse becomes almost uniform withlatitude after 25 years. However, thenormalized response of zonal mean sur-

response between low and high latitudesis no longer obvious. Since R is thetemperature increase normalized interms of the total difference betweenhigh CO2 and normal CO2 climates ateach latitude, Fig. 2a implies that nor-malized sea surface temperature in-creases will have almost the same latitu-dinal distribution for time scales in ex-cess of 25 years. According to Fig. 2bwhich illustrates the normalized re-sponse of zonal mean surface air tem-perature, R becomes almost uniform af-ter only 10 years, int1uenced by the fast-er response over the continent. If thisfinding is supported by further numericalstudies with even more detailed oceanmodels, it is a significant result. Theconclusion would be that, barring anunforeseen drastic acceleration of therate of CO2 increase in the atmosphere,sensitivity studies of climate equilibriumcan be used as an approximate guide forpredicting the latitudinal pattern of seasurface temperature trends.

In Fig. 3 the zonally averaged re-sponse at 10 years is shown for all levelsin the atmosphere and the ocean. In thepolar region when sea ice exists at theair-sea interface, the atmosphere exhib-its the largest response at the earth'ssurface whereas the temperature of theunderlying seawater remains at thefreezing point. Near the sea ice marginlocated at about 750 latitude, the CO2-

~R

References and Notes

I. C. D. Keeling and R. B. Bacastow. Enerf(.v andClimate (National Academy of Sciences. Waslt-ington. D.C.. 1977). p. 72.

2. J. G. Cltarney. Ed.. Carbon Dio.ride and Cli-mate: A Scientific As.,essment ~ National Acade-my of Sciences. Wasltington. IJ>.C.. 1979).

3. S. Schneider and S. L. Thompson. J. Geoph.vs.Res. 86. 3135 (1981).

4. M. I. Hoffert. A. J. Callegari. C.-T. Hsieh. ibid.85.6667 (1980).

5. W. Broecker. in E,.oliition ofPh.vsical Ocean')f(-raphv. B. A. Warren and C. Wunsclt. Eds. (MITPress. Cambridge. Mass.. 19811. p. 434.

6. S. Manabe and K. Bryan. J. Atmos. Sci. 26. 786(1969); -.M. J. Spelm~n. D).n. Atm,)s.Oceans 3. 393 (1979).

7. K. Bryan and L. J. Lewis. J. Geoph.vs. Res. 84.2503 (1979).

8. H. Crutcher and J. Meserve. NA VAIR Pllbl. 50-IC-52 (U.S. Naval Weather Service. Washing-ton. D.C.. 1970).

9. S. Manabe and R. T. Wetlterald. J. At)n,)s. Sci.37.99(1980); R. T. Wetherald and S. Manabe. J.Geophys. Res. 86. 1194 (1981).

10. A detailed analysis for this effect will be present-ed elsewhere (K. Bryan. F. G. Komro. S. Man-abe. M. J. Spelman. in preparation).

II. We thank E. Williams. M. Jackson. P. Tunison.W. Ellis. and J. Conner for assistance in prepar-ing the manuscript; C. Rootlt for valuable dis-cussions; and I. Held. P. Lemke. and J. Sar-miento for detailed reviews.

31 August 1981; revised 25 September 1981

SCIENCE. YOLo 215

face air temperature is even faster andbecomes nearly uniform with respect tolatitude after only 10 years, because ofthe influence of a faster response overthe continents. The results of an earlierstudy (3) with a much simpler ocean-atmosphere model raised the possibilitythat the ocean could cause a transientresponse very different from that indicat-ed by sensitivity studies of normal CO2and high CO2 equilibrium climates. Ourconclusion, based on Fig. 2, is that Rwould be essentially uniform with lati.tude except in the unlikely event of veryrapid increases of CO2, much more rapidthan those indicated in present measure-ments. Of course, the existence of oceancurrents does change the sensitivity ofthe present model, compared to earlierstudies in which a full ocean was notincluded (10).

Our test calculation shows that theuptake of heat by the ocean causes asizable delay in the response of climateto a sudden increase in atmosphericCO2. Quantitative results of our calcula-tions must be considered tentative, sincethis is a process study rather than anexhaustive examination of all model pa-rameters. With this reservation, we con-clude that the oceans can delay but by nomeans eliminate a strong climatic reosponse to increasing atmospheric CO2.

K. BRYAN, F. G. KOMROS. MANABE, M. J. SPELMAN

Geophysical Fluid DynamicsLaboratory, National Oceanicand Atmospheric Administration,Princeton University.Post Office Box 308,Princeton, New Jersey 08540