Low-frequency variability and CO2 transient climate change

Climate Dynamics (1993) 8 : 117-133

Uynumla

© Springer-Verlag1993

Low-frequency variability and CO2 transient climate change Part 1. Time-averaged differences Gerald A Meehl 1, Warren M Washington 1, and Thomas R Karl 2

1 National Center for Atmospheric Research*, Boulder, CO 80307-3000, USA 2 National Climatic Data Center, Asheville, NC 28801-2696, USA Received December 13, 1991/Accepted May 15, 1992

Abstract. Results from a global coupled ocean-atmosphere general circulation model (GCM) are used to perform the first in a series of studies of the various time and space scales of climate anomalies in an environment of gradually increasing carbon dioxide (CO2) (a linear transient increase of 1°70 per year in the coupled model). Since observed climate anomaly patterns often are computed as time-averaged differences between two periods, climate-change signals in the coupled model are defined using differences of various averaging intervals between the transient and control integrations. Annual mean surface air temperature differences for several regions show that the Northern Hemisphere warms faster than the Southern Hemisphere and that land areas warm faster than ocean. The high northern latitudes outside the North Atlantic contribute most to global warming but also exhibit great variability, while the high southern latitudes contribute the least. The equatorial tropics warm more slowly than the subtropics due to strong upwelling and mixing in the ocean. The globally averaged surface air temperature trend computed from annual mean differences for years 23-60 is 0.03°C per year. Projecting this trend to the time of CO2 doubling in year 100 produces a warming of 2.3 ° C. By chance, one particular northern winter five-year average geographical difference pattern in the Northern Hemisphere from the coupled model resembles the recent observed pattern of surface temperature and sea-level pressure anomalies. This pattern is not consistent from one five-year period to the next in any season in the model. However, multidecadal averages in the coupled model show that the North Atlantic warms less than the rest of the high northern latitudes, and recent observations may be a manifestation of this phenomenon. Consistent geographic patterns of climate anomalies forced by increased CO2 in the model are more evident with a longer averaging interval. There is also the possibility that the CO2 climate-change signal may itself be a function of time and space. The general

* The National Center for Atmospheric Research is sponsored by the National Science Foundation. Correspondence to: GA Meehl

pattern of zonal mean temperature anomalies for all periods in the model shows warming in the troposphere and cooling in the stratosphere. This pattern (or one similar to it taking into account the rest of the trace gases) could be looked for in observations to verify the enhanced greenhouse effect. A zonal mean pattern, however, could prove scientifically satisfactory but of little value to policymakers seeking regional climatechange forecasts. These results from the coupled model underscore the difficulty in identifying a time- and space-dependent "fingerprint" of greenhouse warming that has some practical use from short climatic records and point to the need to understand the mechanisms of decadal-scale variability.

Introduction

Washington and Meehl (1989, 1991) show a northern winter (December-January-February, or DJF) anomaly pattern of surface air temperature for a 5-year period (years 26-30 in the model) from their coupled model simulation with carbon dioxide (CO2) increasing at 1% per year. They note that the pattern resembles the recent observed temperature-trend pattern with cooling in the North Atlantic and North Pacific and warming elsewhere in the Northern Hemisphere. The model has subsequently been integrated forward in time to year 60. One purpose of this paper is to determine if that pattern in the model is uniquely forced by increased CO2 and, by analogy, if the recent observed surface temperature anomaly pattern is the unique product of increased CO2 and trace gases. A second purpose is to address the question of time scales of natural variability of the coupled climate system in the emergence of a definitive climate-change signal that could have some practical significance. Since many estimates of observed patterns of climate change use simple time mean differences from various averaging intervals (IPCC 1990), that technique will be used in this paper. Subsequent papers in this series will employ other types of analyses.

118

Meehl et al.: Low-frequency variability and CO~ transient climate change

a )

GLOBALLY AVERAGED OCEAN SURFACE TEMPERATURE DIFFERENCE 3.5

,

,

,

,

,

,

,

,

,

,

b)

SURFACE AIR TEMPERATURE DIFFERENCES,TRANSIENT MINUS CONTROL

,

E

3.0

3.0

2.5

2.5

w 2,0

2.0

~

1.5

.~

1.0

INSTANTANEOUS ~ ~

~

i

i

t

i

i

ALL NH ALL SH . . . .

1.5

i~

~

,

1.0

• / J \ v / ~ / - /-J' v/

05 /

o.~

_/ "~

_ ."

I

TRANSIENT - CONTROL

-0.5 - 1.0

-1.0

-1.5 - 2.0

I

0

I'0

i

2~0

i

3

0=

=

¢

i

40 YEARS

~

i

50

i

60

I

I

IO

I

i

20

I

I

I

i

40

:30

L

L

I

50

60

LFRANSIENT IS A LINEAR I% INCREASE OF CO2 PER YEAR C )

SURFACE AIR TEMPERATURE DIFFERENCES,TRANSIENT MINUS CONTROL 3.5

,

,

r

,

i

i

i

NH 5 5 " 9 ( ~ N SH 5 5 ° ' 9 0 " S

3.0

~Q.

u

i

i

SURFACE AIR TEMPERATURE DIFFERENCES,TRANSIENT MINUS CONTROL

5.5

i

----

2.5

2.5

2.0

2.0

1.5

i~

1.5

1.0

I~

1.0

0.5

~ 03

1.0

-

1.5

~

I

~

,

i

T

i

i

50°N- 50=S LAND - ,50*N- 50 °S OCEAN - - -

° 1,0 -1.5

"2"00

I'D

2'0

'

' 50

~ ' 40 YEARS

'

5"

0

L

~ 60

-20

3.5

'

I

0"

J

2

0.

~

, 30

L

~ 40

,

, 50

A

, 60

DECADAL SURFACE AIR TEMPERATURE DIFFERENCES TRANSIENT MINUS CONTROL 90N

NORTH ATLANTIC REST OF HIGH A T S - - -

3.0

0

YEARS

SURFACEAIR TEMPERATUREDIFFERENCES,TRANSIENTMINUSCONTROL f )

60 2.5

iii!!!!iiiiiiiiiii!!i!iii;iiiii!!! iii!i)iii)i i!! ! ii!i!)ii)iiiii!!!!i)iiiilii!!!iii :,io" ' i

30

2.0 1.5

0

iiiiiiiiiiiiiiiiiiiiiiiiliiiiiiii

50

o.5

go

iiiiiiiiiiiiiiiiiiiiiiiilliiiiiiiiiiliiiiiiiii~ :~' 90S

- 1.0

1.5 0

o.~ ....

60

~-0.5

-2.0

....

I

1.0

-

i

~-0.5

I

~

I

I°

o

-

,

3.0

~-0.5

e)

d)

I0

20

30

40 yFJIRS

50

60

iiiiiiiiiiiiiiiiiiiiiiiiJiii!iii?;,~~:!~°iiiiiii!i 5

15

25 55 45 55 YEAR ( DECADAL AVERAGES)

65

Meehl et al.: Low-frequency variability and COE transient climate change Wigley and Raper (1990) examine the subject of lowfrequency natural variability with a heuristic coupled model that produces inherent natural variability similar to that observed in the globally averaged temperature record. The low-frequency variability in their simple model is attributed to the modulating effects of the ocean on the higher-frequency forcing from the atmosphere. The variability in the model has time scales ranging from decades to centuries, as presumably does the observed variability. For globally averaged temperature, trends of up to 0.3°C occur in the model on the century-long time scale simply from inherent natural variability. Madden and Ramanathan (1980) address the problem of CO2 climate-change detection in terms of climate signal and noise and note that detection of such a signal may have a seasonal dependence. Barnett and Schlesinger (1987) suggest using a climate-change pattern generated by a model to trace the same "fingerprint" in the observations. However, the technique must rely on a model-generated signal that itself may be a function of time and space, as are patterns generated in the observed system. Trenberth (1990) notes that the frequency of interannual events, such as warm and cold events in the Southern Oscillation, can affect climate anomalies on the decadal time scale. Parkinson (1989) generalizes this problem of averaging and notes that time series with inherent variability can yield different anomalies depending on what time samples are included and what period is used for the average. Ghil and Vautard (1991) analyze the observed globally averaged temperature record and document variability on time scales ranging from interannual to interdecadal. These factors of low-frequency variability point to the difficulty of analyzing short data records to search for climate anomalies that are related uniquely to increased CO2 and trace gases. Recently, several global coupled coarse-grid general circulation models (GCMs) have been run for periods longer than 50 years. These include coupled models at the Geophysical Fluid Dynamics Laboratory (GFDL)

119

(Stouffer et al. 1989; Manabe et al. 1991; Manabe et al. 1992), Max Planck Institute for Meteorology in Hamburg, Germany (Cubasch et al. 1991), the Hadley Centre in England (Mitchell personal communication), and the National Center for Atmospheric Research (NCAR) (Washington and Meehl 1989). All of these models exhibit variability on a wide range of time and space scales and begin to mimic some aspects of the observed system. The NCAR and GFDL models include interannual variability that resembles, in some ways, observed E1 Nifio-Southern Oscillation (ENSO) phenomena (Lau et al. 1992 for GFDL; Meehl 1990 for NCAR). Longerterm variability also occurs, suggesting that the present generation of coarse-grid coupled ocean-atmosphere GCMs can begin to address the low-frequency variability problem. This paper is the first in a series to examine various aspects of low-frequency variability (i.e., 5- to 10-year time scales) in the NCAR coarse-grid global coupled GCM. Here, we illustrate very simple aspects of lowfrequency variability in the model and observations by looking at geographic and zonal-mean patterns of surface air temperature anomalies from different seasonal and annual average periods. We also discuss the problem of relevancy or usefulness of CO2 climate-change signals. That is, a certain class of time- and space-scale anomalies could, conceivably, be unambiguously attributed to increased CO2 and trace-gas forcing. But these signals may only be scientifically relevant to prove that the enhanced greenhouse effect actually exists in the observed system and may not be of practical use to policymakers. On the other hand, the predictions of interest to policymakers may be beyond the capabilities of present models because of the low-frequency variability problem and highlight the need to improve various aspects of the models. This would point to an area that requires intensified model and observational studies. Topics that will be addressed in subsequent papers include EOF analyses, variability in the detrended transient integration, and comparison to satellite data.

The coupled model Fig. 1. a Time series of globally averaged annual mean ocean surface layer temperature differences from the instantaneous CO2 doubling (upper line) and the transient (CO2 increasing at 1% per year) experiments, each differenced from the control case with CO2 held constant. The cold-start phenomenon (very little warming in the first 15 or so years of the transient experiment) is evident; b same as a except for surface air temperature (from the model level closest to the surface at a=0.991) from years 23-60 for the Northern Hemisphere (solid line) and the Southern Hemisphere (dashed line); c same as b except for average of the Northern Hemisphere north of 55°N (solid line), and Southern Hemisphere south of 55°S; d same as b except for all land points between 50°N and 50°S (solid line), and all ocean points in that same latitude zone (dashed line); e same as b except for the North Atlantic sector (55°N-90°N, 65°W-45°E, solid line), and all other areas of the northern latitudes north of 55°N outside the North Atlantic sector (dashed line); f time evolution of decadal zonal mean differences of surface air temperature, transient minus control; dashed line indicates start of time series data; light stippling denotes areas less than 0.5°C warming; dark stippling areas greater than 1.0° C

The coupled model includes a global spectral atmospheric GCM with rhomboidal 15 (R15) resolution (about 4.5 ° latitude by 7.5 ° longitude), realistic geometry, nine layers in the vertical, and parameterized landsurface processes. Details of this model are given by Washington and Meehl (1984, 1989). The global ocean GCM has a coarse grid and a latitude-longitude resolution of 5 ° , realistic geography, bottom topography, four layers, and a simple thermodynamic sea-ice formulation. Basic sensitivities to some of the parameters in the ocean model are given by Meehl et al. (1982) and in the appendix of Washington and Meehl (1989). The coupled model reproduces ENSO-like phenomena with transitions from warm to cold events occurring in the eastern Pacific with associated teleconnections to the midlatitudes (Meehl 1990; Meehl et al. 1992). The atmospheric model, when coupled to a simple nondy-

120

Meehl et al.: Low-frequencyvariabilityand CO2 transient climate change

namic, slab, mixed-layer ocean, simulates seasonal lowfrequency variability associated with persistent height anomalies (blocking) reasonably well in the North Pacific, North Atlantic, and Southern Hemisphere extratropics and less well over north central Asia (Bates and Meehl 1986). The coupled model does not use any correction terms at the ocean and atmosphere interface and, therefore, is left with certain systematic errors in the simulation. These are documented in part by Washington and Meehl (1989). Meehl (1989) shows net heat flux systematic errors at the air-sea interface for the tropical Indian and Pacific sectors for various model configurations and concludes that, in the coupled model, most of the systematic errors in those regions are the result of deficiencies in the ocean GCM. Meehl et al. (1992) show how ENSO-like phenomena simulated in the coupled model change with doubled CO2. I n the model, warm and cold events continue to occur in the tropics but with mean sea-surface temperatures (SSTs) higher by about 1°. Thus, in the tropics, absolute climate effects associated with model warm events are more intense with increased CO2 (i.e., anomously dry areas get drier, wet areas wetter). However, one of the systematic errors in the coupled model is the lack of development of the western Pacific warm pool. Thus, changes in that region with increased CO2 must be viewed with caution in the model. In the extratropics, the climate anomalies (in relation to the new 2 x CO2 mean climate) are altered compared with 1 x CO2 with more zonalization of atmospheric teleconnection anomaly patterns. Such changes of ENSO-like phonomena could play a role in changes in low-frequency variability but are beyond the scope of the present paper and will be addressed in a subsequent study. The results from the coupled model are from two integrations. In one, CO2 is held at a constant 330 ppm. In the other, CO2 is increased linearly at 1% per year (as discussed by Washington and Meehl 1989) starting with a concentration of 330 ppm. Earlier results by Washington and Meehl (1989) were from the model integrated to year 30. Since then, the integrations have been extended and most results are shown here to year 60.

Time-dependent climate change Washington and Meehl (1989, their Fig. 4b) show the time evolution of global annually averaged ocean surface temperature anomalies for the 2 x CO2-minus-control and transient-minus-control experiments up to year 30. Figure la updates those results through year 60 in the coupled model. The ocean surface temperature differences in the 2 x CO2 case rise rapidly in the first 5 to 10 years and then experience a slower secular increase afterward. In the transient experiment, the temperature differences show almost no increase in the first 15 or so years and then rise more rapidly later in the experiment. This effect in the transient experiment, sometimes referred to as "cold start", could be an artifact of the experimental design in that the system exhibits an initial lag

for some period after CO2 begins to increase. A more plausible scenario of what the observed climate system is experiencing and will experience in the future may involve starting the integrations with pre-industrial levels of CO2. The experiment could then continue into the future so that the time scale and lag of the response would be more comparable to what we are now observing. That is, the observed system at present has already experienced nearly a century of CO2 increase. Because of the time delays in the response of the coupled climate system, a model that started "cold" with present-day amounts of CO2 could be expected to experience some time lag in its response to the start of the slow but steady increase of CO2. Thus, the slow increase followed by the more rapid increase of temperature anomalies could simply be an artifact of this effect. This possibility has been noted in other models (e.g., Cubasch et al. 1991). In the present paper, results are shown for periods after year 20. As indicated by Fig. la, this avoids the possible cold-start phenomenon in the first 20 years. In the instantaneously doubled CO2 case, the sudden increase of temperature in the first 10 years is a result of the relatively rapid response of most of the upper ocean to warming in the atmosphere. Only certain areas of the ocean (the North Atlantic and the high latitudes around Antarctica) are well-mixed on time scales of many decades. As these regions adjust more slowly, the system continues to experience a secular warming trend after the initial 10-year rapid warming in the 2× CO2 case. The globally averaged SST differences in Fig. la show variability on the order of several tenths of a degree on the decadal time scale, not unlike observed surface temperatures (IPCC 1990). Annual mean differences of surface air temperature (temperature from the lowest atmospheric model layer at sigma=0.991), transient minus control, are shown for various regions in Fig. lb-e for years 23-60 of the coupled model simulation. Unfortunately, time series data prior to year 23 are not available due to data-archival problems while the model was running. Both the transient and control runs contain interannual and other time scales of variability. This is a nontrivial point since changes in relative variance are contained in differences of the annual means. Therefore, absolute measures of variability will be performed in a subsequent paper using the detrended time series from the model integrations. Here the time series of the differences illustrate the sensitivities of various regions in the coupled model. Figure lb shows the time evolution of Northern Hemisphere surface air temperature compared to Southern Hemisphere temperatures. The Northern Hemisphere warming is somewhat greater than that in the Southern Hemisphere. A 30-year average (years 31-60) of the differences is 0.89°C for the Northern Hemisphere and 0.72° C for the Southern Hemisphere. Figure lc shows the contribution from the northern high latitudes (solid line) compared to the southern high latitudes (dashed line). The northern high latitudes show greater warming as well as higher variability. A 30-year average (years 31-60) for the high northern latitudes shows warming of 1.42 ° C compared to only 0.70 ° C for

121

Meehl et al.: Low-frequency variability and CO~ transient climate change the high southern latitudes. The difficulty of this model in maintaining sea ice around Antarctica may be contributing to this difference. But, as noted earlier by Washington and Meehl (1989), Stouffer et al. (1989), Manabe et al. (1991), and Manabe et al. (1992), the well-mixed ocean around Antarctica inhibits warming there compared to the high latitudes of the Northern Hemisphere. Because of the greater heat capacity of the upper ocean compared to the land surface, earlier studies have noted that the land warms faster than the ocean (e.g., I P C C 1990). Figure ld shows this feature for land points versus ocean points between 50°N and 50 ° S. The year 31-60 average is 0.83 ° C for land points and 0.71 ° C for ocean points• As could be expected f r o m the properties of the respective surfaces, the standard deviation of the land temperature differences is somewhat greater than the ocean: 0.16°C versus 0.10°C, Clearly for this model, attributing statistical significance to the differ-

a)

TREND OF ANNUAL T , 7 0 0 MB (KAROLY, 1989) 90 E

o)

TRANSIENT MINUS CONTROL (YR 26-50) •

,,-

.,

-',~.i/i

\ ~

.

:i,i':: < " . . . . .

'°;S ._, .:;,77i. i

'

,~oL

/\

.

.

.:

• ~'~,o,

/ ao

"

b '}

OBSERVED, RECENT MINUS 20's - 50's •

.

,..,,

..: .......

i ~\

...!> :::."

.

.....j <

,. ,,.. :

' , / ~ / / ~ I E . , ' : • ~ ", ." _ ~ W - - / J -

\-,,

;.....

........

¢~.::;\

/

V , ......

oY/

s

/

.:.,

,, / .

•

/

"

.

"--~" • . •

• , ; : ~ , , .....

/

% /

, /o

i

~ ..... / :i' ~ - "

'

'

'

.. .... • "::Tk:;::.\

.

'

.

Fig. 3. a SLP difference for DJF, transient (years 26-30) minus control. Solid contours indicate positive SLP anomalies in the transient case (after Washington and Meehl 1991); b Northern winter seasonal difference in SLP, mean pressure of the 1920s and 1930s minus mean pressure during more recent years (after Rogers 1989)

> •

" z

.~:,~:!

.......

tow

b)

.

.--e.c:s-,:.iF...~FFH-FF '~i:~: :. .,.*~:1.=========================================================================================

: ~ ~ " o .074

~--,.074 ::::::::::::::::::::::::::::::::::::::~iiiiii!i!~ ...........'"'0 [email protected] ~

..:~iiiiiii!i!!!

.!ii:i:iii!i!iii:

::2::::::

:~;:i:i::::.:,~i:::!:i::::~

:......

~

_

,:1:; : ::: : 1:11

I

. i:ii~i::::!i~iiii:ili~:i ~ ,8.o~

O,5J ~~1::::::::::::::::::::::

• 189 I~====================)===

12. I

.336 ~

.,ooph .664 [-- L ,

.8"k

~

,","~.-]

90N

, , .,-,..,

60

C )

30

, , , ,.-,.o.m

0 30 LATITUDE

"

60

8,3

,.,

.500

5.5

3.3

i~

~~

,.r

90S

90N

60

30

0

30

1.7 90S

60

LATITUDE

d)

AT,YR 51-60, ANNUAL MEAN

.oo9 ~ ; ~ i N ~ i ~ i ~ i ~ i ~ i : ~ i_~ . : ~

~.o

&T,YR 31-60, ANNUAL MEAN .009

:::::l~,l;-~i~.~.PY3:-:-:-l.:.:-:.l.7~l,::.:.:l:.:.:.}:i:

:..~,~ ~,~.. [. i i:l...} -~kl. • -I" ":-iI': 1 "~ t3

:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: "1. . . . . . . . . . . . ~. . ~ .............. .... ~ * + 4 ... " . ~ : ~......... i4~ . . . . :::::: .............m..~¢~..-,-... ,~. / •. .*~'*~ .: ::::: ...... ~. .:. .:. :" . . . . . . . . . . . . . :. ~ . ' - . . ; ~ - : , " ~ ~ ..

,,...............~............................~.,.-~,,~..........~,...... ••...~,~......... :::::::::::::::::::::::::::::::::::::: ================================================== =========================================================

~:':':'

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

• • : ::::2:~ ~:

~

:;,,~ ~

2

....-y'~....... . . . . . . .

m,~'-'-'.'.'-.~;,. I

....... •........................... •... •.............~.~.....,,,.,., . . . . . . . . . ..... / :::::::::::::::::::::::::::::::::::: ~ : ~ : ========================= ~i ~ : ~ : ~i.:.:.::::::::i:i:i:::i:i::::::::::::2'::::::2 :::::::::::2::::::::::::.:,:.i:~:i:i::::.i~~ l

~'~======:=========================='t,~===============

:~i~:~i~!~i:~:i~::::::::::~:i:~i~:~:~i:i:i~i~i:i:::::~ = =============================================== ::::::::::::::::::::::: ====== ::::~

:,:.:.:~ :~:.:.-,:~.:'~:.~:~..~- I =============================================================== :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

~ : ~ : ~ =========================================~==, . ~ ,

=============================

E

18.0 "-* tl

.:,:,:,:.:,:.:. :.:.:.

"~'

:iiiiiiiii!iii!:.

:ii:i:i:i:i!i:i:i:i:it:?!i:i!

_

.-

::!ii::iii::i::ii!#:~i

:iil;iii!i!i!!ili

LO

12.1 -1-

.

1 :::::::8::::::::: 9

~

\

~

:i,i!!iii!i!:!) i:i! 12.1 .r -

C

1664

.811 I-,,,-'-~ 1.5~ 90N 60

~

30

0 30 LATITUDE

60

3.3

1.7 90S

Fig. 1 2 a - d . Z o n a l m e a n t e m p e r a t u r e differences for 10-year annual means from the coupled model, transient-minus-control,

dashed lines are negative differences (cooling in transient case); Zonal-mean patterns

Since there is a problem with discerning unambiguous CO2 regional climate-change signals in the model on time scales that m a y be of interest to policymakers, can any unambiguous CO2 climate-change signal be found on shorter time scales with time-averaged differences? Figure 10 shows zonal-mean differences, transient-minus-control, for 5-year D J F and J J A averaging periods. In a qualitative sense, a singular pattern stands out in all the frames of Fig. 10. That is, in every D J F and J J A 5-year averaging period, there is a general pattern o f warming in most o f the troposphere and cooling in the stratosphere. (This pattern is typical of all GCMs with increased C O 2 - - I P C C 1990). The size of the anomalies varies in a similar fashion to the geographical decadal annual means in Fig. 8, but there is more consistency in the patterns in Fig. 10 than in Fig. 4. Similar patterns are present in transition seasons as well (not shown). As

.500-

}

.664 " 5~1 .811 - a l . ~ ~ I , ~ i J ~ , ~ ~ I 90N 60 30 0 ~0 60 LATITUDE

~

8.3 5.5 3.3

1.7 90S

solid lines (stippled areas) are positive differences (warming in transient case); a years 31-40; b years 41-50; c years 51-60; d same as parts a-c except for 30-year annual mean, years 31-60 noted earlier, Wang et al. (1991) have found that this general pattern is somewhat different for the various minor greenhouse gases in their model simulation. Even if the pattern is altered somewhat by the effects of additional greenhouse gases, the point to be made is still valid in that a zonal mean pattern is likely to be more stable than a geographical one. Standard deviations of the 5-year summer and winter averages f r o m the control case are shown in Fig. 11. The high latitudes in the lower and upper troposphere as well as the stratosphere show the largest values of greater than 0.5°C in winter. The tropical stratosphere also shows relatively large variability on the 5-year time scale year-round. Most areas of warming in the troposphere are greater than 0.5°C in Fig. 10, with stratospheric cooling increasing with height to values generally larger than - 2 . 0 ° C, considerably more than the standard deviations in Fig. 11.

132

Meehl et al.: Low-frequencyvariabilityand C O 2 transient climate change

Decadal zonal annual mean temperature differences, transient-minus-control, are shown in Fig. 12a-c. These averaging intervals correspond to the geographical plots in Fig. 8. Also shown in Fig. 12d is the 30-year zonal annual mean temperature difference pattern corresponding to the 30-year geographical temperature difference pattern in Fig. 9. The same pattern of warming in the troposphere and cooling in the stratosphere is present as in the seasonal means. Unlike the geographical patterns, the zonal means are characterized by a more consistent qualitative general pattern for both the 5- and 10-year time scales, and for seasonal and annual means. Perhaps this is a candidate for an unambiguous climatechange pattern uniquely forced by increased CO2 and trace gases. This possibility is explored in more detail by Karoly et al. (1992). Once again, the Wang et al. (1991) results suggest a somewhat different pattern with the addition of more greenhouse gases, but the point remains that a zonal mean pattern of climate change may be easier to detect than a geographical one. Zonal-mean temperature patterns may be pleasing scientifically (i.e., a unique CO2-forced pattern that can be verified with models and observations from relatively short records), but a zonal mean may not be of much interest to a policymaker interested in time-dependent climate change for a particular region. Therefore, a satisfying scientific result in this case may not be of much practical use to policymakers. More elaborate "fingerprint" detection tests (e.g., Barnett and Schlesinger 1987) could, conceivably, identify a geographical timedependent CO2 climate-change pattern. However, such a technique relies on some CO2 climate-change pattern from a GCM. As we have seen, due to inherent low-frequency variability, such patterns themselves may be a function of time and space and are dependent on the length of the time series used to form them. For example, a CO2 climate-change signal could be different from a 20-year time period versus a 50-year time period, and the time-dependent emergence of those patterns would differ in a fingerprint analysis. Furthermore, because of the large observed decadal variability, rather different results could be expected depending on the decade chosen to begin the analysis. For example, a geographic detection test that began in 1967-76 versus 1950-59 (Fig. 8d, e) would lead to rather strikingly different results if stopped at 1990. This does not rule out the application of such a technique. We are simply noting the difficulties inherent in documenting geographical time-dependent climate-change signals that could be of use to policymakers on relatively short time scales, and pointing to the importance of the variability problem. Clearly, interdecadal variability should be a high research priority in both coupled models and observations, and nonlinear coupled interactions that could produce a time- and space-dependent CO2 climate-change signal should be explored.

Conclusions Since time-averaged differences are often used to identify climate anomalies in the observed system, the issue of low-frequency variability is addressed here via time-averaged differences of varying averaging intervals from a global coupled coarse-grid ocean-atmosphere GCM. Time series of annual mean surface air temperature differences for various regions in the model show the Northern Hemisphere warming more than the Southern Hemisphere, with greatest warming occurring in the northern high latitudes outside the North Atlantic sector. Land areas warm faster than ocean areas, with the lowest-magnitude warming occurring in the high-latitude southern oceans and equatorial tropics. After the initial cold-start period (about 15 years), the linear increase of CO2 in the model (1°70 per year) is reflected by a 30-year global warming trend of 0.03°C per year. At year 100, when CO2 doubles, the projected warming in the model based on that trend is 2.3 ° C. From one 5-year period to the next, the model exhibits variability that corresponds to observed variability on that time scale with largest values at high latitudes and over the continents in the winter hemisphere. For the geographic patterns of seasonal surface temperature difference for transient-minus-control, however, there is little discernible consistency of geographical patterns of surface temperature anomalies from one 5-year period to the next in any season. Therefore, the pattern from DJF years 26-30 in the coupled model that resembles the recent observed temperature trends in the Northern Hemisphere (cooling in the North Atlantic and North Pacific, warming elsewhere) is related to low-frequency variability on the 5-year time scale in the model. This suggests that the recent temperature trend pattern is also probably the product of inherent low-frequency variability in the observed system with a somewhat longer time scale. However, the coupled model shows that, for long averaging periods (years 31-60), there is a tendency for the North Atlantic to warm less than the rest of the northern high latitudes. This could be expected to be reflected in the observations over long time periods. Annual mean temperature anomalies are computed from the model for three decades. The signal of small consistent warming in the tropics and subtropics, less warming around Antarctica, and greater warming at high northern latitudes is apparent in every decadal annual mean average. This pattern generally resembles that of the instantaneously doubled CO2 results after 30 years in the coupled model (Washington and Meehl 1989). This geographical pattern does not increase uniformly with time in the decadal averages as CO2 slowly increases and is indicative of low-frequency variability on the decadal time scale in the coupled model. A longer averaging period (30 years) shows an even more consistent pattern that resembles the CO2 doubling result. This aspect of averaging related to variability is consistent with the earlier studies of Parkinson (1989) and Ghil and Voutard (1991). Zonal-mean temperature differences show a generally consistent pattern in both 5- and 10-year averaging peri-

Meehl et al.: Low-frequency variability and

CO 2

transient climate change

ods f o r s e a s o n a l a n d a n n u a l m e a n s with w a r m i n g in the t r o p o s p h e r e a n d c o o l i n g in the s t r a t o s p h e r e . This suggests t h a t t h a t q u a l i t a t i v e p a t t e r n (or one like it if the effects o f the o t h e r t r a c e gases a n d o z o n e are t a k e n into a c c o u n t ) c o u l d be used to search for a CO2 s i g n a t u r e in the o b s e r v a t i o n s . Such a p r o c e d u r e is i n v e s t i g a t e d b y K a r o l y et al. (1992). Liu a n d S c h u u r m a n s (1990) p o i n t o u t s o m e p r o b l e m s with this a p p r o a c h a n d suggest searching for a c o o l i n g t r e n d in the s t r a t o s p h e r e t h a t increases with height, as i n d i c a t e d b y the m o d e l results. T h e results o f W a n g et al. (1991) suggest t h a t the a d d i tion o f o t h e r g r e e n h o u s e gases m a y p r o d u c e a s o m e w h a t d i f f e r e n t p a t t e r n . T h e p o i n t to be m a d e is t h a t a z o n a l m e a n p a t t e r n o f c l i m a t e c h a n g e m a y be m o r e c o n s i s t e n t a n d easier to detect t h a n a g e o g r a p h i c a l one. In a n y case, this raises the issue o f a s a t i s f a c t o r y scientific verifi c a t i o n o f the e n h a n c e d g r e e n h o u s e effect (zonal m e a n t e m p e r a t u r e differences) n o t being useful to p o l i c y m a k ers w h o are m o r e interested in p r e d i c t i o n s o f t i m e - d e p e n d e n t r e g i o n a l c l i m a t e - c h a n g e p a t t e r n s . T h e results p r e s e n t e d here s h o w the latter t a s k to be difficult because o f the i n h e r e n t n a t u r a l l o w - f r e q u e n c y v a r i a b i l i t y in the c o u p l e d o c e a n a n d a t m o s p h e r e system. I n t e n s i f i e d s t u d y o f l o w - f r e q u e n c y v a r i a b i l i t y o n these t i m e scales in c o u p l e d m o d e l s a n d the o b s e r v e d c l i m a t e system is necessary if we are to a d d r e s s the r e g i o n a l c l i m a t e - c h a n g e p r o b l e m in a n y k i n d o f m e a n i n g f u l way. S u b s e q u e n t p a p e r s in this series will d o c u m e n t time e v o l u t i o n o f E O F s in the c o u p l e d m o d e l , d e t r e n d e d time series analysis, a n d c o m p a r i s o n o f r e g i o n a l v a r i a b i l i t y in the m o d e l to u p p e r - a n d l o w e r - t r o p o s p h e r i c satellite data.

Acknowledgements. A portion of this study is supported by the Office of Health and Environmental Research of the US Department of Energy as part of its Carbon Dioxide Research Program. The authors acknowledge comments that contributed to the improvement of the text by Bryant McAvaney and Richard Wetherald, and helpful discussions with Roland Madden, David Karoly, Kevin Trenberth, Harry van Loon, David Baumhefner, and Michael Glantz. Ann Modahl edited the text, and Suzanne Whitman drafted the figures.

References Barnett TP, Schlesinger ME (1987) Detecting changes in global climate induced by greenhouse gases. J Geophys Res 92:1477214780 Bates G, Meehl GA (1986) The effect of CO2 concentration on the frequency of blocking in a general circulation model coupled to a simple mixed-layer ocean. Mon Weather Rev 114:687-701 Cubasch U, Hasselmann K, H6ck H, Maier-Reimer E, Mikolajewicz U, Santer BD, Sausen R (1991) Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model. Max-Planck Institut far Meteorologic Report No. 67, Hamburg, FRG Ghil M, Vautard R (1991) lnterdecadal oscillations and the warming trend in global temperatures. Nature 350:324-327 IPCC (1990) Climate change: the IPCC scientific assessment. In: Houghton JT, Jenkins G J, Ephraums JJ (eds) Cambridge University Press, Cambridge Karoly DJ (1989) Northern Hemisphere temperature trends: a possible greenhouse gas effect? Geophys Res Let 16: 465-468

133

Karoly D J, Cohen JA, Meehl GA, Mitchell JFB, Oort AH, Stouffer R J, Wetherald RT (1992) An example of fingerprint detection of greenhouse climate change. Clim Dyn (in press) Lau N-C, Philander SGH, Nath MJ (1992) Simulation of ENSOlike phenomena with a low-resolution coupled GCM of the global ocean and atmosphere. J Climate 5:284-307 Liu Q, Schuurmans CJE (1990) The correlation of tropospheric and stratospheric temperatures and its effect on the detection of climate change. Geophys Res Lett 17 : 1085-1088 Madden RA, Ramanathan V (1980) Detecting climate change due to increasing carbon dioxide. Science 209:763-768 Madden RA, Shea D J, Branstator GW, Tribbia JJ (1992) The effects of imperfect spatial and temporal sampling on estimates of global mean temperature: Experiments with model and satellite data. J Climate (submitted) Manabe S, Spelman MJ, Stouffer RJ (1992) Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric C Q . Part II: seasonal response. J Climate 5 : 105126 Manabe S, Stouffer R J, Spelman M J, Bryan K (1991) Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2. Part I: annual mean response. J Climate 4 : 785-818 Meehl GA (1989) The coupled ocean-atmosphere modeling problem in the tropical Pacific and Asian monsoon regions. J Climate 2: 1146-1163 Meehl GA (1990) Seasonal cycle forcing of E1 Nifio-Southern Oscillation in a global coupled ocean-atmosphere GCM. J Climate 3 : 72-98 Meehl GA, Branstator GW, Washington WM (1992) Tropical Pacific interannual variability and CO2 climate change. J Climate (in press) Meehl GA, Washington WM (1988) A comparison of soil-moisture sensitivity in two global climate models. J Atmos Sci 45 : 1476-1492 Meehl GA, Washington WM, Semtner AJ Jr (1982) Experiments with a global ocean model driven by observed atmospheric forcing. J Phys Oceanogr 12: 301-312 Parkinson CL (1989) Dangers of multiyear averaging in analyses of long-term climate trends. Clim Dyn 4: 39-44 Rogers JC (1989) Seasonal temperature variability over the North Atlantic. Proceedings of the Thirteenth Annual Climate Diagnostics Workshop, Cambridge, MA, 31 October-4 November 1988, US Dept. Commerce, Washington, DC, pp 170-175 Stouffer R J, Manabe S, Bryan K (1989) Interhemispheric asymmetry in climate response to a gradual increase of atmospheric CO2. Nature 342- 660-662 Trenberth KE (1990) Recent observed interdecadal climate changes in the Northern Hemisphere. Bull Amer Meteor Soc 71 : 988-993 Wang W-C, Dudek MP, Liang X-Z, Kiehl JT (1991) Inadequacy of effective C Q as a proxy in simulating the greenhouse effect of other radiatively active gases. Nature 350:573-577 Washington WM, Meehl GA (1984) Seasonal cycle experiment on the climate sensitivity due to a doubling of CO2 with an atmospheric general circulation model coupled to a simple mixedlayer ocean model. J Geophys Res 89:9475-9503 Washington WM, Meehl GA (1989) Climate sensitivity due to increased CO2: experiments with a coupled atmosphere and ocean general circulation model. Clim Dyn 4:1-38 Washington WM, Meehl GA (1991) Characteristics of coupled atmosphere-ocean CO2 sensitivity experiments with different ocean formulations. In: Schlesinger ME (ed) Greenhouse-gasinduced climate change: a critical appraisal of simulations and observations. Elsevier, Amsterdam Wigley TML, Raper SCB (1990) Natural variability of the climate system and detection of the greenhouse effect. Nature 344: 324-327