GLOBAL BIOGEOCHEMICAL

CYCLES, VOL. 13. NO 4, PAGES 821-826, DECEMBER 1999

Developmentof the CO2 latitude gradient in recent decades ThomasJ. Conway and Pieter P. Tans ClimateMonitoringand DiagnosticsLaboratory,NationalOceanicandAtmosphericAdministration,Boulder,Colorado

Abstract. Because90% of the CO2 from fossilfuel combustionis emittedin the Northern Hemisphere,annualmeanatmospheric CO2 mixingratiosare higherat middleandhighnorthern latitudesthanin the SouthernHemisphere.The observedCO2 latitudegradientvariesinterannuallyandhasgenerallyincreasedasfossilfuel CO2 emissionshaveincreased.Back extrapolationof the measuredCO2 latitudegradientto zero fossilfuel emissionsgivesa latitude gradientwith the NorthernHemispherelower thanthe Southern.A linearregressionof Mauna Loa minusSouthPole annualmeandifferencesversusfossilfuel emissionsfor 1958 through

1996givesa slopeof 0.5 gmolmo1-1(abbreviated asppmCO2) (Gt C)-1 ((• = 0.03) andan intercept(at zero fossilfuel emissions)of-0.8 ppm (• = 0.2). Shorterdatarecordsyield similar resultswith largeruncertainties.We arguethatthisextrapolatedgradientdoesnotrepresent preindustrialconditionsbut is morecorrectlyviewedasa decadalaveragegradientdueto natural sourcesandsinksthatunderliethe anthropogenic perturbation.We interpretthe extrapolated gradientasevidencefor a contemporary NorthernHemispheresinkthathasbeenproposedon the basisof othermeasurement andmodelapproaches.The slopes(ppmCO2 per gigatonof C from fossilfuel burning)calculatedfrom sufficientlylongrecordstendto agreewith model calculations basedon fossilfuel emissions,suggesting thatanytrendin the NorthernHemispheresink,duringthe periodof the measurements, hasbeensmallrelativeto the trendin fossil fuel emissions.

1. Introduction

Althoughthelong-termrateof increaseof atmospheric CO: is essentially the samein both hemispheres (-1.4 ppm yr-• during 1981-1992[Conwayet al., 1994]), annualmeanmixingratiosare higher in the Northern Hemisphere. The annualmean Barrow,

Alaska-SouthPole difference is currently -4 ppm. This differenceis maintainedbecause-90% of fossil fuel CO: emissionsoccur in the Northern Hemisphere[Marland et al., 1994], and interhemisphericmixing occurson a timescaleof 11.5 years[Law et al., 1996]. Keeling et al. [1989] have shown thatthenorth-south differenceis increasing becauseof increasing anthropogenic emissionsin the NorthernHemisphere. If the measured difference between Mauna Loa and the South Pole is

extrapolated to zero fossilfuel emissions, CO: concentrations in the Northern Hemisphere are lower than in the Southern Hemisphere.Keeling et al. [1989] proposed that such an

extrapolationreveals the preindustrialsurface CO: latitude gradient,and they hypothesized that it was causedby vigorous uptakeof CO: in the North AtlanticOceanbalancedby lossfrom the southern oceans.

These ideas were given a differentinterpretationby Taylor and Orr [1997]. AdoptingKeelinget al.'s [1989] hypothesis that the extrapolationto zero fossil fuel emissionsrepresentsthe preindustrialCO: gradientat the Earth'ssurface,they attributed the lower annualmeanNorthernHemisphereCO: concentrations

Thispaperisnotsubject toU.S.copyright.Published in 1999bythe AmericanGeophysicalUnion. Papernumber1999GB900045. 821

to covariancebetweenseasonalvariationsin atmosphericmixing and a seasonalterrestrialbiospherewhich actsas a sourceduring the winter and a sink duringthe summer. Sincethe preindustrial biosphericsourceandsink are presumedto be annuallybalanced, no net annualmean sink is implied at middle to high latitudesin the Northern Hemisphere. Translated to today's conditions, Taylor and Orr [ 1997] assertedthat no significantterrestrialsink at midlatitudesis necessaryto explain the relatively small observednorth-southgradient[Tans et al., 1990] because,in their view, a balancedbiospherewould already counteractthe fossil fuel imposedgradient. There is generalagreementamongatmosphericmodelersthat the covariancebetweentransportand sources/sinks constitutesan important aspect of the CO2 latitudinal gradient. In a recent intercomparisonof atmospherictransportmodels [Law et al., 1996], resultswere presentedfor simulationsin which all models used the same seasonalsource/sinkpattern for an annually balancedterrestrialbiosphere. The model resultsfell into three groups. One group, including all models with an explicit formulation for the planetaryboundarylayer, obtaineda surface interhemisphericgradient about half as strong and of the same sign as that producedby fossil fuel burning. Denning et al. [ 1995] demonstrated that this wasmostlybecauseof the seasonal difference of vertical mixing processes:RespiratoryCO2 is trappednear the surfaceduring winter, while the summerCO2 drawdown is mixed into a much greater volume of the atmosphericcolumn during summer. This phenomenon is commonlyreferredto as the seasonalatmosphericrectifier effect. A secondgroupof modelsproduceda very slightpositiveor no significant gradient. One model [Taylor, 1989] resulted in a significantlynegative gradient, with the SouthernHemisphere

822

CONWAY ANDTANS:DEVELOPMENT OFTHECO2LATITUDE GRADIENT

surface havinga higherconcentration thanthenorthern, opposite

3. Results

to the resultof Denninget al. [ 1995]. AnnualmeanCO2 mixingratiosrelativeto the SouthPole are In this paperwe reexamine the trendin the CO2 latitude for MaunaLoa gradient in recentdecades andoffera newinterpretation of the plottedasa functionof fossilfuel CO2 emissions extrapolation to zero fossilfuels.Usingseverallong CO2 in situdata(1958-1996) and Barrow (1976-1997) andAscension measurementrecordsand fossil fuel emissiondata, we have extrapolated the latitudegradientto zerofossilfuel emissions.In

agreement withKeelinget al. [1989],we findthatoverthepast severaldecades,in the absenceof fossil fuel emissions,the

Island (1980-1997) flask data in Figure 1. The solid lines are linear regressions to the data. Similar regressions have been calculatedfor each record, and the slopes,intercepts,and uncertainties(1 (• from Bevington[1969]) are summarizedin

oqe.eans andbiosphere wouldhaveproduced a gradient withless

Table

CO2 in the NorthernHemisphererelative to the South Pole. Whatdoesthistell usaboutCO2sources andsinks?

Althoughthere is significantcorrelationbetweenthe latitude gradient and fossil fuel emissionsfor all sites, it is clear from

1.

Figure1 andTable 1 that the slopesandintercepts determined from the shorter NOAA

2. Data

CMDL

flask records are much less

certainthanthosedeterminedfrom the 39 year SIO MaunaLoa

In thisanalysis weuseannualmeanatmospheric CO2mixing and SouthPole records. The resultsfrom the relativelyshort ratios obtainedfrom flask samplescollectedat the National Oceanic and AtmosphericAdministration(NOAA) Climate

flask records must be viewed with caution because the rate of

fossil fuel combustiononly increasedfrom 4.9 to -6.7 Gt C

however, to notethe Monitoring andDiagnostics Laboratory (CMDL) sampling sites year-1from1976to 1997. It is encouraging, betweenthe independent (exceptfor at Barrow,Alaska(BRW); Mould Bay, Canada(MBC); Station relativelygoodagreement M (STM); Niwot Ridge, Colorado (NWR); Terceira Island, Azores(AZR); Key Biscayne,Florida (KEY); Mauna Loa and Cape Kumukahi,Hawaii (MLO and KUM); Guam (GMI); Ascension Island(ASC); AmericanSamoa(SMO); andthe South Pole(SPO);thesemixingratioscoverthetimeperiodsindicated in Table 1. These sites were chosen because their measurement

calibration)39 and 22 year SIO and NOAA CMDL recordsat MLO: The slopes and interceptsagree to within 1 c• uncertainties.We havealsoperformeda regression on the 1976-

1996subset of theSIO MLO record.Theslopeandintercept still agreewith the flask and in situ recordsto within 1 c•, while the uncertaintiesare larger and are similar to thoseobtainedfrom the

resultis the good Shorter and shorterflask records. Anotherencouraging discontinuous records areinsufficient for performing meaningful agreementof the slopeand interceptbetweenKUM (sealevel) records are long and mostly continuous.

regressions. The NOAA CMDL air samplingnetwork and measurementsare described in Conway et al. [1994 and

and MLO (3400 m), which are both on the island of Hawaii.

references therein], and the data are available for electronic

tudein Figure2. It is interesting to notethatthelargestslopes (-1.0 ppm CO2 (Gt C)-1) are observed at middleand high

transfervia ftp from the CMDL server(ftp.cmdl.noaa. gov). Annualmeanatmospheric CO2 mixingratiosmeasured at Mauna

The slopesand their 1 o uncertainties are plottedversuslati-

northern latitudes. This is consistent with the increase in fossil

Loa(in situ)andtheSouthPole(flasks)from1958to 1996by the ScrippsInstitutionof Oceanography (SIO) [Keelinget al., 1994]are alsoused. The globalannualCO2 emission datafor fossilfuel combustion andcementproduction through1996are fromMarlandet al. [1994] andsubsequent updates.The 1997 value is a preliminary estimate (T. Boden, personal

fuel emissions havingoccurredprimarilyat midlatitudes of the NorthernHemisphere. Notealsothata significant negative slope is observedat ASC (Figure 1), indicatingthat as fossil fuel

communication, 1999).

in Figure 3. The averagestationminusSouthPole differences

emissions increase, the difference between ASC and SPO is

decreasing.

The intercepts and 1 o uncertainties fromTable 1 areplotted

Table 1. Summaryof LinearRegression Results Slope,

Site

Type

MBC BRW STM NWR AZR

Flask Flask Flask Flask Flask

KEY MLO MLO MLO KUM GMI ASC SMO

Flask Flask In situ In situ Flask Flask Flask Flask

Time Period 1981-1996 1976-1997 1981-1997 1976-1997 1980-1988

Intercept ppm

ppmCO2 (Gt C)'1

Oslop e

1.05 0.76 0.76 0.91 0.58

0.24 0.20 0.18 0.19 0.39

-2.4 -0.8 -1.6 -2.7 -0.9

1.4 1.2 1.1 1.1 2.2

0.48 0.56 0.51 0.44 0.58 0.39 -0.30 0.19

0.16 0.10 0.03 0.12 0.12 0.19 0.12 0.11

0.6 -1.2 -0.8 -0.4 -1.0 0.3 3.1 -0.1

0.9 0.6 0.2 0.7 0.7 1.1 0.7 0.6

{Jint

1995-1997

1976-1997 1976-1997 1958-1996 1976-1996 1977-1997 1980-1997 1980-1997 1976-1997

CONWAY AND TANS' DEVELOPMENT OF THE CO2LATITUDE GRADIENT ß

,e

'

I

....

I

....

I

....

I

....

and their

823

standard errors for the 1981-1984

and 1994-1997

NOAA CMDL data are plottedfor comparison.The measured gradientsclearly showthe increaserelative to the SouthPole over the past two decades. The gradientextrapolatedto zero fossil fuel emissions(intercepts) is dramatically different from the observedgradients.The minimumin the extrapolatedgradientis -2.7 ppm at NWR. At higherlatitudes(STM, BRW, and MBC) the surfaceCO2 is 1-2 ppm lower than at the SouthPole, and at MLO and KUM, CO2 is -1 ppm lower than at SPO. For KEY, GMI, and SMO the gradient is slightly positive or zero.

Barrow ".•. • "ø .

ß

Mauna Loa

•,•,,.'i'-•,

ß .....

i

...

," ,

Extrapolation of the negativeslopefor ASC resultsin a CO2 value3.1 ppmhigherthan SPO, with a relativelysmallstatistical uncertainty. As was noted with the slopes, there is good agreementamong the interceptsfor the MLO (flask), MLO (in

Ascension

....

•

i

i

•

•

.

.

i

i

I

i

i

i

i

•

situ), and KUM (flask) records.

7

Fossil Fuel Emissions (Gt C) Figure l.

4. Discussion

AnnualmeanCO2 differencesfrom the SouthPole

It is clear from the scatterof the data in Figure 1 and the

plotted against global fossil fuel emissionsfor Mauna Loa

(squares),Barrow (circles),and AscensionIsland (triangles). uncertaintiesof the slopesin Table 1 that there is significant interannualvariabilityin the CO2 latitudegradient. Most of this variability is not randomor due to measurementerror but reflects regionalvariationsin marine and terrestrialCO2 sourcesand

The solidlines are linear regressions fitted to the data. The time periods for each site and the slopes,intercepts,and their uncertaintiesare given in Table 1.

Latitude 90"S

1.5

30'S

I

I

I

EQ

30øN

90ON

i

I

I

1.5

(station- SPO)vs ff slope

I--I (MLO - SPO) vs ff slope

•

1958-1996 (MLO- SPO)vsff slope

1.0

1.0

0.5

0.5

0.0

0.0

'7,

o -0.5

,

I

- 1.0

,

I

-0.8

•

I

-0.•

•

I

-0.4

•

I

,

-0.2

I

•

0.0 Sine

I

0.2

• ,

,I

0.4

=•[ •

I

0.•

/

I

0.8

,

-0.5

1.0

Latitude

Figure2. Theslopes fromlinearregressions of annualmeans minusSouthPoledifferences versus fossilfuel(ff) emissions plottedas a functionof latitude. The errorbar represents +1 (5. The MLO valuefrom the Climate Monitoring andDiagnostics Laboratory (CMDL)flaskdatais represented by anopensquare to distinguish it from KUM. Thestarrepresents theMLO slopefromtheScripps Institution of Oceanography (SIO)(1958-1996)record.

824

CONWAY AND TANS: DEVELOPMENT OF THE CO2LATITUDE GRADIENT Latitude 9oøs

30'S

EQ

30øN

90øN

1994- 1997averagestation- SPOdifference

1981- 1984averagestation-SPO difference station- SPOextrapolated to zerofossilfuel 1958- 1996MLO- SPOextrapolated to zerofossilfuel

A

1

E

O

0

-1

-1

-2

-2

-3

-3

-4

-4

-5

-5

•

-6 - 1.0

-0.8

-0.6

-0.4