JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 95, NO. D10, PAGES 16,851-16,864, SEPTEMBER 20, 1990

Atmosphere-Biosphere Exchange CO2 and03 in the of
Central Amazon Forest
$ONG-MIAOFAN, STEVEN WOFSY,PETER BAKWIN,ANDDANIELJ. JACOB C. $.
Division Applied of Science Department EarthandPlanetary and of Science,
Harvard University, Cambridge, Massachusetts
DAVID R. F1TZJA•ALD

Atmospheric Sciences Research Center, State University NewYorkat Albany of
Measurements vertical of fluxesfor CO2andO3 weremadeat a level 10m above canopy theAmazon the of forest during wetseason, the using eddycorrelation techniques. Vertical profiles CO2andO3 wererecorded of continuously above canopy the soil surface, forest from the to and floor respiration measured was usingsoil enclosures. Nocturnal respiration CO2 by theforestecosystem of averaged 2.57 kgC/ha/h, with about85 % fromtheforest floor. Duringthedaytime, 2 wastakenup at a meanrateof 4.4 kgC/ha/h.Net ecosystem CO

uptake carbon of dioxide increased solar by0.015 with flux (kgC/ha/h)/(W corresponding m-2), tofixation of
0.0076molesCO2 permolephotons (about 0.017molesCO2permoleof absorbed photons photosynthetiat cally activewavelengths). relationship The between ecosystem net exchange solarflux wasvirtuallythe and same the Amazon in forestasin forests Canada in (Desjardins al., 1982, 1985) andTennessee et (Baldocchi et al., 1987a,b). The relatively highefficiency utilization light (about for of 30% of thetheoretical maximum)and the strong dependence net CO2 uptakeon solarflux suggest light may significantly of that regulate ecosysnet tem exchange carbon and storage thetropical in forest.Changes thedistribution cloudcover,associated in of for example with climatic shifts, mightinduce globally significant changes carbon in storage. Rates uptake for

of03 averaged ll molecules 4 inthedaytime hours, 2.3x10 cm-2s (10 700-1700 hours), dropping roughly by a factor 10during 14hours dusk dawn. mean deposition of the from to The 03 velocity40mwas cms at 0.26 4
inthenight 1.8cms in theday.Diurnal and -1 variation O• deposition regulated bystratification of was both of
theatmospheric boundary layerandby stomatal response lightandwaterdeficit.The totalflux of 03 to the to forestwaslim. largelyby supply ited from the free troposphere above. Deposition 03 to the forestcanopy of appears bea regionally, perhaps to and globally, important for tropospherica. sink O

1. INTRODUCTION
tl-,,•

Deposition ratesfor O3 to theAmazonforestwereestimated for
,4..,, " ....

biosphere exchange associated high temperatures chemical are with and production [Jacob andWofsy, 1988]. Deposition ratesat

Tropical represent sources for forests significant or sinkschemi[Gregory 1988; et 1988; et al., Browell al., Kirchhoffet ai., 1988]. tally climatically trace including The and important gases, 03, CH4, analysis indicated a downwardin excess1012 flux of CO2, reactive and hydrocarbons. foratmosphere- cm-2s the Rapid rates molecules but estimates account -1, could not for photo-

humidities, ofbiological and high rates activity,intense sunlight. were night obtained by Kaplan [1988] measured etal. using vertiRecent have that studiesshowndeposition to vegetation ½al isasigni- gradientswithin abovecanopyvertical ofO3 and the and exticant for inthe sink O3 planetary layer the boundary over tropical coefficients from and change derived fluxes gradients The of NO. forest [Gregory 1988; et 1988], may be lack direct etal., Kaplan al., and also of information gas ontrace exchangetropical over an important contributor global oftropospheric makes tothe budget 03 forests itdifficult todetermine the factors regulate that gas [Galbally 1980;Liu, Exchange intropical and assessrole tropical inthe and Roy, 1988]. of CO2 fluxes to the of forests global forests likewisesignificant global cycle may be inthe carbon atmosphere-biosphere system. [Fung, Matthews,Mooney 1987]. 1986; 1983; et al., We present paper first correlation inthis the eddy measureDirectmeasurementsgasexchange of between tropical forestsmentsof 03 deposition CO2 exchange and over a tropicalforest.

and atmosphere reflecting the are sparse, the difficulty ofmount- eddy The correlation has virtue being and method the of direct ing complex experiments locations. rams non-intrusive, may obtained inremote Uptake for and results be continuously over exCO2 photosynthesis estimated using tended by have been indirectly variaperiods. experiment The was conducted forest atDucke tions CO2 of observed and the within above tropical canopy forest reserve Manaus, as ofthe near Brazil part NASA/INPE ABLE2b [e.g., and Odum Jordan, Lemonal.,1970; etal., mission the season 1970; et Wofsy during wet of 1987. Fluxes budgets and of 1988]. Emissions2from which ofCO soils, contributeofthe heal vapor, momentum studied most water and have been previously at nocturnal flux, been CO2 have measured enclosure this byShuttleworth [1984a] Fitzjarrald using tech- site etal., and etal. niques IGoreau DeMello, Keller al.,1986; and 1985; et Schles[1988]. inger, 1977]. FluxesCO2 Oa been of and have measured correlation byeddy
over a varietyof othervegetation typesusingeithertoweror aircraft platforms, e.g. Verma and Rosenberg,[1976]; Ohtaki,

Copyright 1990 by the.american Geophysical Union.
Paper number 90JD00595. 0148-O227/90/90JD-00595505.00

chi et al. [1987a,b]for CO2 overpaddyfield andforest; Weselyet

[1980]; Desjardins Alvoal., et [1982]; et [1984];Baldocal., and

al. [1978, 1981]; Lenschow al. [1981, 1982]; and Droppo, et [1985] for 03 overforest,water,andothersurfaces. Hicksandhis
16,851

16,852 •

[:AN t:] •1.: (;ARB()N [)I()XII)E &N[) (_)Z()Nt I.X½'IIANGE

ABLE-2B

coworkers [e.g.,HicksandMcMillen,1988;McMillen,1988]and calibrations werecardedout at thispressure. modified The CO2 Denmead and Bradley[1985, 1987] havecontributed importantinstrument exhibited artificialmodulation about2 Hz, but an at recentadvances the application eddy correlation in of measure- thisnarrow band,highfrequency component notcorrelate did with ments forests.Thesestudies to providethe background required vertical velocity contribute CO2fluxes nextsection). or to (cf. to design measurement a strategy tropical for rainforests, includ- Air passed through thermoelectrically a cooledglassvolume ingprecautions needed makemeasurementsnonideal to in terrain filledwithglass beads before entering CO2analyzer, the establishusing sensors imperfect with response functions noise. and ingconstant temperature dewpoint(5øC). Liquidwaterwas and We alsomadecontinuous measurements mixingratiosfor drained of through opening thebottom thevolume drawan at of by CO2andO3 at eightaltitudes above withinthecanopy. and The ing off an air flow of 0.3 L/min. 'Thisarrangement assured that

profiles provide measurementsstorage the canopy of in layer. fluxes not haveto be corrected density do for variations to due Canopy storage contributes a significant to thenetecosystem part fluctuations temperature watervaporin the sample of or cell exchangecertain at times day, of particularly CO for 2. [Webb el., 1980]. et Theresults clef'me diurnal the variation rams gas of for exchange TheO3 analyzer (Monitor Labs, Model8410)detects chemibetween forest theatmosphere, provide the and and information luminescence reaction O3withC2H,•. It was of the of modified onthefactors regulate fluxes. examine elements rapid that gas We the for response Gregory al. [1983, by et 1988], attaining a 90%

that control atmosphere-biosphere using canopy exchange a resis- response of 0.8 seconds. time Calibration theinstrument of was
rance modelsimilarto that usedby Hicksand coworkers checked thefieldby comparing O3 concentrations to in with measdescribe deposition 03, SO2,andNO2to vegetated of surfaces ured theDasibi by Model1003AH analyzer 41 m height O3 at
such maize oak-hickory as or forest [Baldocchi el., 1987; et Hicks [Bakwin el., thisissue (b)]. Instrument et (a), response exwas etel., 1987; Meyers Baldocchi, and 1988].Themodel analysis tremely instable during course theexperiment. the of windspeed), CO2,andO3 wereacdicates atmospheric biospheric that and processes both are impor- Rawdataforw (thevertical
usinga PDP 11/73computer system tantin limiting deposition reactive of species thetropical to forest. quired10 timesper second andwererecorded floppy on diskettes. Spectra, cospectra, coand
2. EXPERIMENT

variances w, CO2, and03 werecalculated of later. Data weretak-

en in parallelat 1 second intervals a Campbell by Scientific DaThe experiment conducted theReserva was at Florestal Ducke, mlogger. datalogger The stored readings 20-minperiods, for then 25 km northeast Manaus,Amazonas, of Brazil, between April 22 removed mean for eachsensor the and computed variances and andMay 8, 1987. Vegetation climatology the sitehave covariances, and for accounting time delaysbetweeninstruments for

been described in detail Shuttleworth [1984a,b] rounded nearest by et al. and tothe 1 second [Fitzjarrald this etal., issue]. Roberts [1990]. main canopy about m Time etal. The tree extends 30 delays between chemical the the and vertical signals wind above ground some with emergent to 35 m. Foliage were trees is estimated field blowing into robing inthe by breath the inlet presentalllevels, at though vegetation ishighestthe (for density in CO2) byturningand an and on off ozone generatorvapor (Hg upper portionthe of canopy near ground. and the Diurnal varia-lamp) placed theinlet 03). Theestimated near (for delays,
tions temperaturespecific of and humidity deficit strong 8.1-1-0.1 CO2 10.2-1-0.1 03,were influx were at sfor and sfor used calcutopof thecanopy, ranging 22 to30øC -0 to 8 gtKg lations from and throughout the experiment. values conf'mned These were (night-m-day) respectively. variations Diurnal oftemperature bycomputing covariances the data discussed and lagged from raw as

humidity werevery smallneartheground thesoiltemperaturebelow. and

was almost constant at25øC. Mean cloudiness 0.6inthe Concentration for CO andO3 were isabout profiles 2 obtained by dry season 0.7inthe season and wet [Ratisbona, 1976]. sequential sampling ateight altitudes 3, 6, 12,19, 36,41 (0.02, 27, Analuminum scaffolding of 45 m height tower served the m)through 1/4-inch Teflon as fixed OD tubes. inlets the The of top
mainplatform theexperiment. flux measurementsCO2 five levelswere attached rodsextended for The of to aboutI m from the

andO3 wereacquired 39 m on thetower, at about above 9m the tower, whileinlets thelower of three levels were attached a tree to canopy top. A fastresponse single-axis sonic anemometer (Camp- about15 m fromthetowerin orderto avoidinfluence human of

bellScientific was Inc.) mounted theend a beam at of extending activity around base thetower.A manifold solenoid the of of

aboutmfrom tower 3 the [Fitzjarraldal.,this et issue]. Operation valves used switch was to between tubes airsampling, inlet for
of thisinstrument associated and measurements of heatand from lowest highest, a dwell of4 minateach to with time level. momentum atseveral flux altitudes discussed are byFitzjarrald Concentration forO3were et profiles measured a Dasibi using al.[1988, this and issue]. 1003-AH ozone analyzer described Balewin al. [this by et issue A Teflon tube 3/8"outside of diameter mmi.d.) wasat- (a), (b)]. Themeasurements profiles made a (8 of CO2 were with tached thebeam to about cm fromtheanemometer, 30 openingBeckman nondispersed 865 infrared analyzer. Because the of downwards protected a small and by plastic funnel avoid to as- variable lengths the sampling of tubes, cell pressures the in pirating rain. Ambient foranalysis CO2and was air of 03 drawn analyzer varied from to600torrs 580 accordingthealtitude to bethrough Teflon ata total the tube flowrate 25standard per ingsampled. of liters Calibrations carried at several were out pressures
minute, small enough avoid to interference themeasurement with approximately 2 hours.A cooler, every identical thatused to in of vertical windvelocity. Mostof theair flowwasbypassed thefast-response just instrument, conditioned sample and air standards upstream the analytical of instruments, 2.5 L/min drawn toconstant with temperature dew and point.Carbon dioxide concentrathxough CO2analyzer 0.84L/minthrough O3analyzer tionswerecalculated interpolating the and the by between consecutive gain (Monitor Labs, model8410). factors zeros, and andwerecorrected pressure for effects.ProillA BINOS nondispersed analyser used theCO2flux ing procedures IR was for were controlled, datarecorded, an HPand by measurements. instrument The electronics modified give 3421Adataacquisition/control andanHP-85computer. were to system
rapidresponse (90% response tijne4).3 s), so thattheexperimenSoil emissions CO2 were measured sevenlocations of at near

hal response waslimited timerequired flushthe25 mL the baseof the micrometeorological usingTeflon coated time by to tower sample (0.6s)andcold (-0.3 s,see cell trap below).A pressure aluminum of enclosures. enclosures madein two parts: The were about torts 650 wasmaintained sample referew.cells, in and ce and round collars 16.4 cm deepand24.8 cm diameter placed few a

ABLE-2B

FAN ET AL.: (dARB()N DI()XIDI::. AND OZ()NE
!.0

EX('ttANGE

16,853

centimeters into soil, and tops 19 cm high. Each top had two
holes, 1.5 cm in diameter,one connected 1/4-inch Teflon tube to

leading gasanalyzers theother open theair. Theair to and one to
inside the enclosure was mixed gently by an aluminumpaddle drivenby a batterypowered motorat 120 rpm. The collarswere inserted the soil at least2 daysprior to the measurements in and most were left undisturbed throughout experiment. Samples the were takencontinuously a flow rate of 300 mL/min usingthe at Beckman CO2 analyzerimmediately after a top was put on the
collar. Carbon dioxide emissions were calculated from the rate of

increase CO2 concentration of insidethe chamber. Total incidentsolarradiationwasroutinelymonitored a pyby

roheliometer a site2 km awayfrom themicromet at tower(O. Cabrai, unpublished data, 1989). We retrievedsolarflux data from chart records 50 daysin April andMay 1987;thedataweredifor gitized, rectified, and converted to incident flux using the
manufacturer's calibration.

•' -2.o

I

I

I

I

I

3. VERIFICATIONOF FLUX DATA

.4o

.2o

o
SECONDS

2o

40

6O

The verticalflux of a scalar, suchasCO2 or 03, canbe viewed as an imbalm•ce between quantity scalar the of transported across a horizontal planeby air parcels movingup anddown.The time averaged verticalHuxF maybe represented by

I

F=(1/r)lw'(Oc'(t+at)at
o

(•)

whereT is the averaging interval,andAt is thedelaytimebetween themeasured signals verticalvelocity, andconcentration, for w, c. The primes (1) denote in deviations w andc from theirrespecof tive nmning means computed overintervals length The senof T. sorsfor w and c must be sensitiveenoughand respond rapidly enough resolve to smalldeviations thefrequencies at important for turbulent transport. Averaging intervals mustbe sufficiently (T) short resolve to atmospheric variations flux butsufficiently in long to obtainstatistical significance.A periodof 20 min to ! hour is oftenchosen obtainan effective to average overa largenumber of eddies, while allowing observations thechange theflux with of in time[Lumley Panofsky, and 1964]. The observedsignalsare affectedby instrumental artifacts

_111

t_

-20

-60

-40

-20

0
SECONOS

20

40

60

(baseline and drift noise) may that produce inthe errors computed 1.Lagged Fig. covariances wand and between (a) between CO2 (b) wand flux. Exposureorientation anemometer affect calculated period and ofthe may also O3 for the 14-00-1620 2.(Negativewhen on May lags
the quality flux of determinations. Shuttleworth [1984a] measurementand were etal. and ofCO2 O3 delayed.)
Fitzjarrald et al. [1988, and this issue] have demonstrated previouslythevalidityof eddyflux measurements momentum, for heat andwatervaporat thissite. We examine herein detailtheaccura-

cyof thetracer fluxes determined theDatalogger, con- Cross by which correlation functions sLrnilar t.hat to shown Figure in 1 stitute main the components data (only fewhours were of our set a of calculated other for periods. delay The times obtained from rawO3data were acquired thePDPsystem to logistical correlation by due these functions ordy fewtenth vary a seconds one from problems.) sourceserror beexamined: errors period another. indicated Figure anerror thetime Three of will (1) in to As in 1, in delay times bythe used Datalogger,instrumental drift (2) zero and delay 0.1s would of introduceerror fluxof about percent an in 1 selection averaging ofdata interval, (3)statistical and imprecision both and small for CO2 03, compared toother sourceserror. of associated fluctuationtransport and aliasing Instrumental driftrepresents frequency It can with of rates with of zero low noise. high-frequency from sensors. noise the bea serious source error zero occurs both and of if drift for w c Time delays 8 s forCO2 10s forO3 were bythe signals averaging not of and used and does effectively remove contribution its

Datalogger throughout experiment. the Lagged covariancesthefluxestimate. fluxes to The were calculated theDafrom betweenand w CO2 betweenand were and w O3 calculated ta!ogger using measurements bysubtracting means w and 20-min from 10Hz PDPdata investigate sensitivity computed to the of fluxesc, then performing integral (1). In order assess magthe in to the to deviations theassumed from values. magnitude theco- nitude errors The of of introduced theDatalogger by algorithm, comwe variance w' andc' should maximum thecorrect of be at delay puted fluxes CO2and using Hz rawdata, a series of 03, 10 for of time.Examples thecovariancea function delay of as of timeare running mean intervals 1.7to 20 min. Results from (Figures 2a shown Figures andlb, in which in la optimal correlations and2b) indicate the computed is independent the were that flux of
found 8.5 s between andCO2and10.2s between andO3, averaging at w w interval intervals for between and20 min. The com6 in good agreement estimates in thefield. with made puted flux declines intervals than5 minutes. for less When20-

16,854

FAN ET AL.' C'ARB()N DIOXIDE AND O7_.()NEEX('i-tANGE

ABLE-2B

C02 FLUX

lO

5
-4

I'

I

I

I

,I

0

5

10

15

20

Averaging Interval(min)
1

L,

I
0.01

I 0.05

I 0.1

I 0.5

I 1

03 FLUX

0.005

f(hz)
5o
-lO

10 - 2o

5

-3o

1

0.5

-4o

ro

I5

lO

I-

15

i

20

i

AveragingInterval(min)
Fig. 2. Fluxesof (a) CO2 and(b) 03 calculated with running means difof ferent lengths subtract•fromsignals. Dataarefrom 1400-1640, May 2.

0.i
0.05

0.01

•
0.005

•
0.01

I
0.05

I
0.1

I

I

0.5

1

min grandmeans wereremoved from signals averaged the 16for min period(1400-1640)on May 2, 1987,fluxeswereat most15% differentfrom thosecomputed subtracting 6 min running by the mean(seeFigure2). Theseresults weretypicalof othertimeintervals,asindicated Figure4 (discussed by below). The smallerrors associated with zero drift reflect stableinstrument performancein enviromental conditions that variedlitfie duringthe period of experiment.We conclude that subtraction running of meanoverintervals between6 and20 min, or a grandmeanover a

20-min intervals. flux period, provide estimates tobetter consistent than 15% for most
Statistical errors in estimates of the mean flux are associated
0.5

with temporalvariationof the verticalflux and with instrumental noiseproduced fastresponse by chemical sensors. Statistical iraprecision usuallythe largestsource errorfor a flux estimate is of

in a single 20-minute interval, typically least10%under at ideal
conditions [Wesely and Hart, 1985].The errorsin a givenperiod depend thevariances w andc, andon the signal-to-noise on of ratio of eachmeasurement [Lenschow and Kristensen,1985].
0.1
0.005

I
0.01

!

I

I

I

Figures 3a-3c showpowerspectra w, CO2, andO3 sensors, for computed usinga Fouriertransform 10-Hz datafor a 140-min of

0.05

0.1

0.5

1

5

f(hz)

interval theafternoon. spectra CO2and have in The for O3 signi- Fig.3. Examples of power spectra cospectra and atarbitrary (a) units. ticant power highfrequencies is notobserved thespec- Spectrum (b)spectrum (c)spectrum (d)eospeetmm at that in ofw, ofCO2, ofO3, of
tramfor w. HoweverFigures and3e showthatthecospectxa w withCO2, (e)cospectmrn with03. Data from1400-1620, 3d of and of w are w with CO2 and O3 showno significant contributions from fre, May2.

ABLE-2B

FAN El' At..: CARB()N [)!½)Xil_)l•AND OZ()NE EX(•HANGE

16,855

0.0

data (+ connected lines) for the intervalswhen both systems by were operating.Fluxesderivedfrom 10 Hz datavary smoothly over periodsof severalhours. The fluxes obtainedfrom the Damloggerare scattered aboutthe smooth lines,but the magnitude
and diurnal variation of fluxes derived from the 10 Hz data are

-0.5

well reproduced.Errorsfor hourly averaged fluxes are in most cases than-20% of peakvalues.Comparison hourlyfluxes less of from the Datalogger thosecomputed to from 10 Hz data (Figure

4c) indicatesregression a (r2=0.66) witha slope veryclose 1 to
and with zero intercept. The residual standarderror is 1.9 kgC/ha/h,with no systematic trendin theresiduals.
-2.0

-2.5

-3.0

I

I

The influenceof aliasingby the Datalogger algorithmwas furtherexamined computing by CO2 and03 fluxesfromPDP data sampled once per secondto simulatedata collectionby the Datalogger.Mean fluxesandstandard deviations shown Table are in 1 alongwith the corresponding fluxesfrom the Datalogger.The errors introduced the low sampling by rate areevident.The coef5

0.005

0.01

0.05

0.1

0.5

1

f(hz)

ficient of variation for O3 flux is smaller than that for CO2, reflectingthehighersignal-to-noise ratio of the ozonemeasurement.

Nevertheless, fluxes averaged over eight intervalsfrom the Dataloggerand the PDP agreecloselyfor both 03 and CO2, consistent with our analysis errors the techniques. of for Data were obtainedfor varioustime intervalson 12 days, and for each hour we have from the Datalogger19-29 (mostly27), 20-rain intervalsfor CO2, and 24 20-rain intervalsfor 03. Accordingto the analysis errorsoutlinedabove,randomerrorsin of hourly mean values for the whole experiment, due to statistical variance,removalof means,etc., shouldbe lessthan5% of daily maximafor mosthoursof the day. As we shallseebelow, biasin
the data due to variations of natural conditions(cloudiness,wind

speed)andby floodingof the anemometer duringrainstomsmay introduce largererrorsin definingmeanvaluesfor the experimental period.
-5

4. CA."_nON DIOXIDE FL?X .s.N• N•

F.,cosYs,•,¾, PRODUCrI,,', ,,rrl'Y

I

I

I

0.O05

0.01

o .o5

o.1

0.5

Fig. 3.

(continued)

A characteristic diurnalpatternwas observed CO2 flux (12 for daysof data)anda verysimilarpattern wasobserved 03 (parfor tial datafor 6 days),as shownin Figure5. Concentrations above and within the canopy exhibited markeddiurnalpatterns (Figure 6). The meandaily minimumfor CO2 abovethe canopy was340 ppm at midday,7-8 ppm lowerthanthemeanvalueobserved 3 at km altitude from theElectraaircraft[R. Harriss S. Wofsy,unand published data, 1989]. The meandaily maximumat canopy top, more than 370 ppm, was observedas CO2 accumulated the in

quencies above Hz; evidently 1 spurious high-frequency signals forest boundary layerbefore sunrise. Variation totalCO2conof donotcorrelate w sufficiently introduce wi•h to significant errors tentbetween and39 m height related thefluxes thetop 0 is to at intothefluxcalculation, least at whenthesignals oversampled are and bottomboundaries, to production losswithinuhe and and by acquiring readings 10 Hz, high-frequency at noisedoesnot coltomb by
contribute to the mean flux since it does not correlate with w.
39m

periods between and100s, 10 longer would typical than be over
low-stature vegetation [Finnigan,1985].The cospectra w with for

'• Mostpower in 3correspondsdI c(z)dzF•oa+F•,f of in cospectra to the the Figure + = F39m (2)

CO2 Oa, for with and and w temperature vapor and water [Fitzjarwhere isthe ofemission F•oa rate bysoils, isthe F•oa integrated raid al.,this et issue] 'indicate contributions flux that tothe fromrate emissionleaves, F39mthe of by and is upward at39m. flux frequencies 1Hzwere above negligible. thereforeneces- influence It is not The ofhorizontal advection isassumed negligible tobe sary correct to computed forlimited fluxes bandpass trace average. left-hand in(2),the ecosystem ofthe on The side net exchange gas sensor systemsHicks McMillen, [cf. and 1988]. (NEE), becomputed measurements concencan using ofvertical Spurious high-frequency affect Datalogger signals the algorithm profiles flux the ofthe tration and at top tower.
signalsare not oversampled. set of data acquired A onceper c(z)dz)withthemeasured meanfluxes(F39m). Diurnalchanges in second can alias high frequency variations into the flux signal. storage termswere significant the budget CO2, but not for for of Figures and4b showhourlyaveraged 4a CO2 fluxesderived by 03. Storage termswere particularly important duringthe early the Datalogger (squares) fluxesderivedby analysis 10 Hz morning and of hours, whenCO2 respired nightis usedfor photosynat

more the flux than PDP determination, this the Figure because case in 5compares variations (d/dr the diurnal ofstorage1309•

16,856
10

FAN ET AL' CARB()N DI()XIDE AND OZ()NE EX(;}iANGE

ABLE-2B

aaa t

•

-- aaaaa
D

a

•

aaaa ß o a •
D

D

D

D

12

./

-10

1"

!

I

!

..!

0

20

40

60

80

HOURS
10 _

APR 27-29

o1: a I

a

rl

'5
D

ß'
.

a

-10

-15

1"

,

!

.!

'

I

•

0

20

40

60

80

HOURS

MAY 6-8

Fig. 4. Comparison between hourly CO2fluxes(kgC/ha/h) fromtheDatalogger PDP systems. Time series and (a) fromApril 27 to April 29 (squares, Damlogger, crosses connected linesfor PDP). (b) Time series by from May 6 to May 8 (same labelsasin symbols Figure4alfR). (c) Datalogger versus flux PDP flux (squares), regression:= 0.97 (i0.09, standard ¾ deviation) = 0.0 X

(i0.25,standard deviation), r2=0.66.

thesis whilesimultaneously exhausted theforest the being from to Themorning/evening (am/pm) ratiofor CO2uptake (1.31)was developing atmospheric mixedlayer. virtuallythe sameas the am/pmratio for incident solarflux (see The maximum uptakeof CO2 by the forestsystem occurred Figure reflecting 8), greater cloudiness afternoon for sampling in-

around noontime, average 9.3kgC/ha/h. on about Mean daytimetervals: mean the solar was W m in themorning flux 460 -2 and (6-18hours) NEEforCO2wasabout-4.4 kgC/ha• withgreater360 W m-2 in theafternoon (am/pm=1.28). Neither asymthe
uptake observed before noon (-5.0kgC/ha/h) thanin theafternoonmerry themagnitude thesolar nor of irradiance weretypical the of (-3.8 kgC/ha/h). The flux of CO2 from the forestfloor averaged 50 daysin April andMay 1987,for whichwe obtained solarflux 2,22 kgC/ha/h, with little variation according timeof day(Fig- data: average to values the solarflux in themorning afterfor and

ure7). Thenighttime mean value NEE was2.57kgC/haPn, for noon were 320 W m-2 and 325 W rn respectively -2,
alsowithlittlevariation through night(Figure 'the 5a). (am/pm=0.99).Intervalswith flux data are evidently biased to-

ABLE-2B

FAN ET At..: CARB½)N Dit)X!I)E ANt) ()Z()NE EX½'tt,X, NGE

16,857

002 FLUX- kgC/ha/nr
lO

L O G
G

E R
-5

•

D øa . D

-lO

-5

0

5

10

PDP

y = 0.97 (-,-0.09)x + 0.0 (4-0.,25); = 0.8 R
Fig. 4. (continued)

TABLE 1. Fluxesof CO2 and03 Recorded the by

C/ha/h,and ca--411(:t:l) (wmZ),(r2=0.90). The average halfsaturation value for the forestsystem(ca) is l-dgher than the daynet •,,•.y A•;1,,

Datalogger Calculated thePDPdata and from

COz, kg C/ha/h
Data-

O.½I() molecules/cm2/s mean 389W/m implying thesystem lightlimited: tl time of 2, that is
Data-

Time
1420 1440 1500 1520 1540 1600 1620 1640
Mean Standard

!o!•6er Mean* o'f
-6.29 -5.54 -6.02 -4.73 -6.89 -3.15 2.69 -5.68
-5.12 1.50

lo•6er Mean*
-2.53 -3.18 -3.73 -4.82 -3.13 -4.38 -3.68 -1.53
-3.37 1.04

of
0.08 0.11 0.09 0.14 0.17 0.07 0.11 0.08
0.10

-3.75 -3.29 -5.39 -7.24 -5.55 -2.41 -4.74 -4.64
-4.63 1.50

0.47 0.39 0.33 0.89 1.33 0.34 0.31 0.34
0.55

-1.64 -2.10 -3.06 -5.72 -4.05 -2.66 -2.93 -3.73
-3.23 1.28

tlAaI. tUI Wihh carbong.'•.,:^_ t •..1•1 v a&"y Analysis airborne of flux datafor the Laroseforestin Canada [D½sjardins al., 1985] gave a relationship et remarkably closeto that foundhere for the Amazonforest. They fit a line to uptake data as a functionof solarirradiance, obtaining slopeof -0.014 a

(kgC/ha/h)/(W -z) andan'intercept about kg C/ha/h.If m of 1.2
we treatthe Amazondam the sameway (see the dottedcurvein

Figure weobtain slope -0.015 9), a of (kgC/ha/h)/(W-z) and m intercept 2.2 kgC/ha• (rZ=0.86).Thecoefficient carbon of for uptake observed a Tennessee in forestby Baldocchi aL [1987(a)] et

wasabout-0.028 kgC/ha/• m of PAR),and (W 4 since PAl! accounts abouthalf of the solarflux [Canwbell,1977; Od•m for aL, 1970], thevariationof NEE with solarflux alsocorresponds to

* Meanfluxcalculated theinterval sampling for by 10Hzdataat 1 Hz and
computing valuesof the flux. 10 * Standard deviation.

about-0.014 kgC/ha/h (W Themaximum quantum yieldforphotosynthesis in leaves of species about is 0.06 moleCO•.fixedper molephotons 25øC at

wards sunny periods,consequence a of repeated failures the and of intercellular concentrations and of 230ppm of CO2 O2 and sonic anemometer rain during storms. 21%, respectively [Farquharal.,1980]. et Themean coefficient

Figureshows hourly NEE CO2 9 the mean for observedfor 2uptake the at CO by Amazon 0.015 forest, (kgC/ha/h)/(W mZ),
Reserva Ducke, plotted against 'incident radiationthe isequivalenta quantum of 0.0076 the solar at to yield mole COz/mole phoEmbrapa 2 km away.Carbon site dioxide taken withre- tons, is up using conversion given Campbell the factors by [1977]. The
duced quantum efficiency thehigher at irradiance levels. least albedo the,aanazon was A over forest estimated Shuttleworth by et squares ofthethedata thefunctional fit to form al. [1984a] beabout to 12%, if PARis assumedaccount and to for
NEE = ct +

cz'S
ca S +

50% solar of irradiance, the quantum for yield absorbedat PAR
between and200W/m2, thecarbon zero fixation increases rate by

This 30%of thetheoretical is maximum. irradiance For levels

themeanirradiance levelis about 0.017moleCOz/mole photons.

where NEEis thenetecosystem exchange CO2and isthein- about for S 0.051moles COzfor each of additional of absorbed mole cident solar flux, gavect= 4.1 (:t:l)kgCPna/•c2=-18.4(:!:3.8)kg photons the PAR spectrum; marginal in the quantum yieldfor

16,858

FAN E'I AL.: CARBON DI()XIDE

AND OZ()NE

EX('IIANGE

ABLE-2B

FLUX STO'RAGE (0) (+)

2

x

:3

-4

0 o

-6
'-8

-10

o

5

lO
ß

15

20

LOCAL TIME

FLUX(u) STORAGE (+)
.

•

0

.

• .olo

• ..21:)
o,o_.

'•' -30

I•-40

0

'

'5'

10'

•

'

ß•

15
ß

20

•

...

LOOAL TIME
Fig. 5. Meanhourly fluxes measured 39mheight at (squares) compared change storage with in (crosses) calculated mean from concentrations measured eightaltitudes 2 cmto 41 m above at from ground. NEE (plain The curves) thesum flux andchange is of

in storageß CO2(kgC/ha/h) (b)03 (1010 (a) and molecules-2 s-l). cm

CO2 uptake low light levelsis therefore at nearlyequalto the ecosystem uptake carbon, least of at overshort timeperiods. Light

maximum expected C3plants. for sat,marion arenotdominant effects because forest optically the is The remarkable coincidence results boreal, of for temperate, deep, most theleaves and of function lightlevels at below saturaandtropical forests strongly suggests in a forestwith well- fionandat temperanares limiting that, within values.Thehighefficiendeveloped canopy andadequate water,available PAR controls net cy for utilizationof light by well-watered forestecosystems is

ABLE-2B

FAN E•! a,i•.: CARB()N

Dif)XIDk

AND OZ½)NE EX('tlAN(jE

16,859

CO2 CONCENTRATIONS (ppm)AVERAGED OVER12 DAYS
õ0 .

SOIL CO2 FLUX
,

A 40
3O

• -

T
u

E 20
(M)

o

c•

'
r
0

1.0

1
25

0

Fig.

5 wetseason of

7.

Soil e

•n•

tduring the

5O

40

3O

t
20

{m)
," I I i I I I

4

6

8

10

12

14

16

18

20

LOCAL TIME i
0 5

i
10

i
15

i0
20 25

Local Time

Fig. 8. Incidentsolarradiationaveraged over 50 days (plain curve)and overperiodswhenthe NEE of CO2 was measured (curvewith squares) in April andMay.

Fig. 6. Diumal variationsof the mean vertical distributions (a) CO2 of (ppm)and(b) O3 (ppb).

bon,more than1.2x10 kgC/yr •2 overthe5x106 2 of theAmakm
zon Basin. This flux may be balancedin part by emissions of

consistent thisview. It might possible usethecoeffi- volatile with be to hydrocarbons. Zimmerman al. [1988]estimated et that cient NEEversus for solar asa diagnostic i.e.,if a sys- emissions isoprene terpenes flux tool, of and should account 0.010and for tern were significantly efficient, less another limiting factor (e.g., 0.002 kgC/ha/h during dryseason, the respectively, onexbased water) might suspected. be tensive measurements planetary in the boundary layer,Jacob and Net dailyuptake CO2wasobserved be 0.93kgCflna/h Wofsy [1988] obtainedsimilar resultsfor isoprene(0.016 of to for the periodof experiment, representing significant a imbalancekgC/ha/h) usinga photochemical modelandaircraft observations between carbon fixation mineralization. imbalance a of isoprene and This is concentrations. Emissions reactive of terpenoid hyfunction the sampling of bias towardsunnyintervals.The 12- drocarbons alone maythusaccount 5-6% of theestimated for net hour-mean solarflux for intervals withflux dataexceeded 50- carbon the uptake. Carbonmonoxide [Kirchhoff al., this issue], et

daymean 90 W rn equivalentstimulationCO2uptakemethane, particulate dissolved by -2, to of and and organic carbon also are exby 0.67kgC/ha/h (averaged 24 hours) over according theresults ported from the forest, and may accountfor much of the to in Figure9. We wouldexpect therefore that,overthe50 daysin remainder.

April andMay for which havesolar we data, uptake CO2 Theresponse NEE to variations solar net of of in irradiance suggests wouldbe only0.25kgC/ha/th close thedetection for this thatimportant to limit shifts global in carbon storage could induced be by experiment. changes thedistribution cloudiness, for example cliin of due to
The small residualuptakeof CO2 duringthe wettestmonths, mate fluctuations such as the E1 Nino-Southem Oscillation. ff

0.25 kgC/ha/h, correspondsa globally to significant of car- cloud flux cover increased 10%,(insolation by decreased about by 35

16,860

FAN EI AL.: CARBON DIOXIDEAND OZONEEX('HANtiE
CO;, NEE'
3.0

ABLE-2B

2.$

2.0

I •,,-10 I
-15

0

200

400

I

600

.... I=:.... .... I
800

0.0

I

0

5

10

15

Wlm2

LOCAL TIME

Fig. The 9. NEE CO2 of versus insolation. measurements Fig. Diumal Our (squares) 10. variation ofozone deposition (cm-1)measured velocity s
andcurvilinear regression (solidline) are shown, alongwith a linearfit in theAmazon forest. (dotted line) compared the text with the treatment borealforestdata in of
by Desjardins al. [1985]. et

takearethe high density vegetation, hightemperature of the and humidity(favoringstomatal opening), and rapid turbulent tran-

W -2) might net of kgC/ha• basis) at m we expect release 0.25 (24-h sport top.of O5 measured are reflectscontribucanopy various 39 examined These factors below. The downward flux at m or than the annual release from fossil fuel factor fiGns deposition andthe and about a2 1.2x10 kgC/yr. Although onlyburning, it this a is of4 from to vegetation ground, to chemismaller
cal reactions in the air column between zero and 39 m. Chemical

represents about ofthe lossonly 0.5% standing [Mediof biomass reactions are negligible contributors inthe tothe flux daytime na period. A1983] year mightsustained and Klinge, per cloudcovermight stimulatecarbon when ozone islarge and this but and be for signifi- the flux a [JacobWofsy,issue],at cant decrease of
storage, enhanced although evaporation might stress significant et this (a), The F39mbe induce that [Bakwin al., issue (b). flux can
counteracted the influence ofhigher flux. solar written as:
5. FACTORS CONTROLLING DEPOSITION 03

night the contribution from reaction with NO emitted soil is by

The xx daytime LT)s arather forest fluxeslarge canopy =I Rt.(z) +I k[NO](z)[O3](z)dz dL(z) + Rs (4) averaged (0700-1700 O3the consid- Fag=' 2.3x10 molecules -1,of to value cm-:
ering the low O5 concentrations canopytop (=6 ppb). Mean at

rr [O3](z)

[O3](1)

nighttime (1700-0700 were fluxes LT) lower more afactorwhere = 30m isthe of the by than zr top canopy, isthe resisRt.(z) leaf

of 10, 2.2x10molecules s while concentrations todeposition xø cm -l, -• O5 at rance per square centimeter area, isthe ofleaf L(z)
canopy werelower onlya factor 2. Uptake O3by leafarea top by of of index between zr,k isthe constant the z and rate for teac-

night. daytime5 fluxes about factor 6 lower The O are a of than based theO5concentration on ([O3](1)) I m altitude. leaf at The themid-morning average estimated Gregory al. [1988] by et for resistance canbe decomposed contributions leafRL into from thedryseason. difference beexplainedpart larger atmosphere This can in by boundary stomatal mesophyllic and (R•), (R,), (R,•), O3concentrations planetary in the boundary (PBL) layer duringcuticular resistances,follows (Re) as [Meyers Baldocchi, and
thedryseason, 20 ppbin thedaytime about [Gregory al., 1988]. 1988]: et
The O5 flux F, at altitude maybe assumed z proportional the to concentration [Oa](z) at that altitude[Galballyand Roy, 1980], wheretheproponionality constant defines deposition the velocity

vegetation isevidenfiy more much efficient the than fion + 05,and s isthe during day at NO R resistance todeposition ground atthe

RL=Rb+ D• •c'c +

[

]-'

(5)

F: = [O3](z)V,•

(3) where and are molecular D D• the diffusivities03 and for water
vapor,respectively.The term representing reactionwith NO is

We show Figure10 thediurnal in variation V,•at 39-maltitude, notstrictly of correct thedaytime, in since some theproduct of NO2 computed thehourly from meaneddycorrelation fluxes O5 mayphotolyze, theterm negligible and but is except night. at

concentrations. Deposition velocities averaged cms in the Weapproximate integrals (4) by subdividing atmo1.8 -x the in the daytime (0700-1700 and cms atnight LT) 0.26 -x (1700-0700 LT). sphere between and39 m intodiscrete zero layers uniform of
The daytime deposition velocities at the highendof values vegetation are density uniform concentration: and O3 previously reported vegetation for surfaces [Galbally andRoy,

ured deposition a velocity 1.0cms over GulfCoast of -x a pine
forest, using eddy the correlation technique. appears O5upIt that

• [O:•]•,AL, " 1980; 1980]. Lenschow Sehmel, In particular, [1982] F3•mR• I+•x et meas- =c=• al. •k[NO]i[Oa]•AZi +[0:•] (4') Rs

take theAmazon isparticularly by forest efficient comparedwhere to [Oa]i, [NO]t, Ratrepresent and quantities averaged over
otherenvironments. Possible factors determhn. thisrapidup- layeri, AL•is theleaf areaindexof thelayer,andAZ•is thelayer ing

ABLE-2B

FAN EI AL.: CARBON D!t•Xiœ)E AND OZONE EX(TIANGE

16,861

thickness. Inspection canopyarchitecture of verticalO3 of and concentration profiles suggestssubdivision thecanopy 3 Time a of into layers: 0-2 m (layer 1), 2-20 m (layer 2), and20-30 m (layer3).

TABLE 2. LeafResistances (s/m) Used theModel in
30-20 m 20-2 m 2-0 m

Layerconsists 1 ofundergrowth vegetation and 6-7 with high density 0-6
low lightlevels, layer2 consists mainlyof treenunks hasthe 7-8 and lowestvegetation density, layer3 includes crowns trees, 8-9 and the of which intercept most of the incoming light. The atmosphere 9-10

above canopy, the between and m,isdescribed fourth10-11 30 39 asa 11-12 layer 4 = 0), where may consumed (AL O3 be atnight reaction by 12-13
with NO. The concentrations NO and O3.usedin (4•) are the 13-14 of mean values measuredat the site as a function of time of day 14-15

proximated aerodynamic between 1mob-16-17 as the resistance 0 and 17-18
rained CO and from 2 radon (Rs=l cm atnight, 2.5s 18-24 data s 4 and
cm in thedaytime 4 [Trumbore, 1988].Thesurface resistance of

[Bakwinal.,this (a),(b)].The et issue ground resistance ap- 15-16 R is s

2.59 1.00 0.75 0.69 0.60 0.69 0.71 0.69 0.78 0.84 0.87 1.07 1.34 2.59

5.27 3.20 2.34 1.95 1.45 1.57 1.76 1.82 2.07 2.13 2.60 2.97 5.10 5.26

11.70 11.36 11.36 7.56 5.85

5.59 5.95 5.59 5.95 5.99 5.84 7.16 11.36 11.70

the ground assumed is negligible. Robertset al. [1988] as functionof altitudeand time of day, and distribuThe leaf areaindexof theforestcanopy wasnotmeasured dur, tion leafarea of index (LAI)discussed text. in the

Computedfrom stomataland leaf boundaryresistances measuredby

ing the ABLEYB experiment, previous but irradiance measurements at the site [Shuttleworth al., 1984b] indicatethat about et

1.4%of thelight at canopy reaches ground top the between 0900 40 and 1000LT (the time of day whencloudiness leastfrequent). is Assuming uniform a angular distribution leaves, of corresponding

..
,

. '

to anextinction coefficient0.5normalized area of toleaf index
Shuttleworth al. [1984b] corresponds a leaf areaindexof 7 et to

[Verstraete, the penetration 1987],that we f'md light observed by
1, 2,layerof canopybottom sensitivity layer 3) the from to top; 2.10
'ß r• i 1. , ' i i I

0930LT). We assme(1,2,4)asthedistribution leafarea of index
in thecanopy, wherethenotation refersto leaf areaindexin (layer
calculationswill be conductedwith distributions(1,1,5) and

(taking account correction zenith into the for angle, = 38 at • 20. 0 ø

(1,3,3).

theindividual contributing resistances(5). Leaf-boundary in (Rb)
eta!. [1990] Values for R• varied w;,h o•,;,,a•, •om i '• s/cm

The resistances becomputed knowledge leaf Rt.i can from of

0

and stomatal resistanceswere (Rs) measured site Roberts atthe by

LOCAL TIME

near ground 0.3 s/cm canopy values R, varied Fig. Uptake ozone forest the to at top; for 11. of by canopymeasured m(squares) as at39 both altitude time day, with and of reflecting stomatal response and estimated to as bymodel from (lines bottomtop, to cumulative uptake

illumination other and parameters. minimum The stomatal resis-rates ground layerstop canopy). from through to of
rancein mid-morning at canopytop was about 1.5 s/cm. We adopted average values theresistances of measured Roberts by et

resistance10scm was R, = 4 adopted, as determinedob- Our from analysis that daytime ofO3 vegetashows the removal by
served flux night. total resistances layer,fion becomputed observed O3 at The leaf foreach can using stomatal resistances. GenRt•a• are significanfiy in thedaytime atnight, eralizafion approach lower than as of this to other environments is however
shownTable due openingthe in 2, to of stomata the during day. hampered thescarcity data stomatal by of for resistances. A
We compare Figure11 theO3 fluxes in computed from(4•) to number models of havebeenproposed predictthe functional to

al.[1990] each and ofday.Mesophyllicresistances (Table associated partial for layer time resistance 2), with stomatal closure were assumed negligible [Wesely, 1989]. uniform A cuticular caused likely water [Roberts 1990]. most by stress etal.,

theobserved values. Reasonable agreement is found throughout dependences of stomatal resistances On environmental variables, the day.Deposition upper tothe canopy (20-30 accounts forvarious m) for vegetation a particularly types; elaborate has model
about 75%of thetotalflux, dueto thehighvegetation density been in proposed Baldocchi al. [ 1987a, which by et b] includes the

that layer the stomatal and low resistances. The distribution oi?effects photosynthetically of active radiation, temperature, vapor
leafareaindexin thecanopy onlya small has effect thecom- pressure on deficit,•and water leaf potential. Figure shows 12 stomaputedflux: the 24-houraverage flux is 11% higherwith the tal resistances. calculated usingthe modelof Baldocchi al. et
(1,1,5) distribution, and 11% lower with the (1,3,3) distribution. [1987a, b], adopting environmental data for the /Xanazon forest The agreement between observed computed and fluxesat nightre- andmodelparameters stomatal for response derived theseau, by flectsmainly the adjustment the cuticular of resistance 10 s thorsfor an oakforest. Watervapordeficits to werecomputed from

cm a value 4, much lower than previously estimated deciduous for dataon air temperature humidity. and Leafwater potential was
forests [MeyersandBaldocchi, 1988;Wesely, 1989]. Despite this assumed remainhigh throughout day. Computed to the stomatal relativelylow cuticular resistance, daytime the flux is still much resistan• agreewell with observations, exceptnearthe ground higherthanthenighttime flux because thelowerleafresistancewherethe modeldoesnot predictsufficient of stomatal areaat the in theuppercanopy(factorof 4) andthe higherO3 concentrationlow ambient light intensity.The modelsuccessfully simulates the at 39 m (factorof 2) duringtheday. The decrease theflux from increase resistance in in between morning and afternoon, indicating morning afternoon be explained an increase stomatal good to can by in representation theeffectof waterdeficit of

16,862
0-2m
10

I-AN E'I At..:CARBf)N [)I{)XII•I:: AND()Z()NE !-:.X('ttANGI•:

ABLE-2B

I

I

al., thisissue] suggest thesupply O3 to thecanopy that of maybe ultimately ]knited therateof downward by transport thefree from troposphere, although interpretation Lhese of gradients compliis cated potential by photochemical production loss 03. or of Thefactors regulating fluxof O3 to theforest the maybeelucidated studying sensitivity theO3 flux to values canoby the of for py surface resistance rate of verticaltransport, or usingtheonedimensional photochemical modelof Jacoband Wofsy[thisissue]. The model simulates meteorologically undisturbed conditions observed frequently during wetseason, diurnal even the with growthanddecayof a well-defined mixedlayer. Deposition to
leavesandground treated is with the multi-layer modelpresented above,and verticaltransport described a seriesof aerois by dynamic resistances extending from the ground the topof the to planetary boundary layer(2000m). A fixedconcentrationat the is upperboundary adopted from observations 2500 m for each at constituent ppbfor O3). (25
Results of model runs are summarized in Table 3. The 24-hour

i

, i,

i

i

,

8

10

12

14

16

LOCAL TIME
2-20 m
10 .

average3 flux to thecanopy O computed "standard" using model parameters [Jacob Wofsy, issue], and this 1.07x10 molecules •l
cm s agrees withobserved -2 -1, well values. Photochemical productionand loss ratesfor ozoneare significant throughout the

6 _

column [Jacob Wofsy, issue] and this despite verylow levels the
6 .

of NO; howeverthe net effectof photochemistry smalldueto is

cancellationnetproductionlowaltitudes netloss higher of at by at levels. Net O3 loss,duemainlyto deposition the canopy, to is onlypartlybalanced supply by fwm thefreetroposphere at (flux 2500 m), leading decay theozone to of column with a timeconstantof about2.5 weeks. We believethatozoneis replenished ep-

0•
6

I

I

i_

ß

I

I

6

I0

12

14

16

LOCAL TIME
20-30 m
10 _

8 _

$

E

J
0
œ

t
I I t • J

isodically deepmixingof thetroposphere convective by in storms [r:_•,,o,,,,g ,t ,,,;,,;..... • indicat;.ng ;'""po .,, o" ....... '•'" deep •'"" convection supplying in ozone theloweratmosphere. to The factors regulating O3 flux wereassessed imposing the by arbitrary changes thevalues to adopted leaf areadistributions for andfor ratesfor turbulent transport (seeTable 3) in thismodel. Increasing leafarea the index a factor 2 increased fluxto by of the thecanopy 12%,decreasing a factor 2 decreased flux by by of the by 16%. Results were more sensitive rateof vertical the transport: increasing by a factorof 2 throughout boundary R, the layerdecreased ozone the flux by 30%,decreasing by 2 increased R, flux by 40%. Theresponse R, reflects to changes theconcentration in of ozone above canopy i.e. changes thegradient just the top, in of O3 in the "mixed" layer. We conclude theflux of O3 to the that Amazon forest significantly is limitedby therateof supply from the free troposphere, although canopysurface the resistance also has an effect. Note that the average concentration O3 in the of boundary layeris regulated largely theupper by boundary condition imposedin the model, implying an important connection betweenO3 deposition fluxes and global-scale processes that
determine ozoneconcentrations themiddle troposphere. in
6. CONCLUSIONS
16

I

6

8

10

12

14

16

LOCAL TIME

Thenet ecosystem exchange CO2in thetropical of forest un-

Fig. Observed resistances inthe 12. stomatal for leaves Amazon dergoes forest a well-defined variation bytheinput diurnal driven of (range by given vertical from lines, Roberts [1990]) values solar etal. and calcuradiation. curvilinear A relationship found was between

lated the of usingmodel Baldocchi1987a, et [ al. hi.

solar irradianceuptake CO2, net uptakea and of with CO2 at
given solar irradiance torates equal observed forests other over in

mwere order scm (Trambore, S.M. Fan S.C. anced releasenight, soils of 0.1 -1 1988; and by at with accounting formore 90% than Wofsy, unpublished 1989], data, smaller thecanopy than surface nocturnal of respiration. resistance 2),which (Table implies the fluxisnot that O3 limited The carbon balancethe of system appeared very tobe sensitive byturbulent transportand in above canopy. the However, tocloud on time ofthe the cover the scale experiment.global The im-

Daytime aerodynamic resistances between 41 climate Uptake the was average balmeasured 27and zones. during day on closely

strong vertical gradients 3observed aircraft forO from [Gregoryportantecarbon et of storage tropics long recognized inthe has been

ABLE-2B

FAN [:'i AL.: CARB½)N DI½}XiI_)E AN[) OZf)NE

EXf'i-iANGE

16,863

TABLE3. Ozone Budgets theAmazon over Forest

Standard
[O3] g O3 flux at 2500 m 03 flux at canopy top

Leaf Index Area
12.9 6.6(10) 1.10(11) 14.5 4.9(10) 8.4(10)

Rax2* .
14.0 2.9(10) 6.9(10

R•x2õ
13.6 1.10(11) 1.4(11)

[Not

13.6 5.9(10) 9.9(10)

3.4

3.3

3.3

3.4

Photochemical production #

-2.5(9)

-6.5(8)

-2.6(9)

-1.0(10)

2.3(9)

oaz

-4.25(10) 4.46(10) -3.76(10) -5.0(10) -2.77(10)

Results from simulations usingthe modelof Jacoband Wofsy[1990], usingvaluesof canopy leaf areaindex (LAD and aerodynamic resistances from their standard (Rs) modelas well as higherandlower valuesof LAI. All results shownas 24-houraverages. NO and03 concentrations represent meanvaluesbetween and2000 30 m, in pptv and ppbv respectively.Photochemical production given as the columnintegralbetween30 and is

2000 in molecules -1. Fluxes column derivative also m, cm-2s and time are given molecules -1. The in cm-2s expression 5.6(10) denotes 1010. 5.6x * Withleafarea distribution (2,4,8). t Withleafarea distribution (0.5,1,2). • All aerodynamic resistances the from ground 2500 are to m multiplieddivided afactor 2. or by of õAveraged between and 30 2500 m.
# Column lossratebetween and2500 m. 30

[e.g.,Houghtonet al., 1985], andpresent results suggest glo- Baldocchi, S.B.Verma, D.E. Anderson, that D.D, and Canopy photosynthesis hal carbon storage mightbe affected changes insolation by in as- andwater-use efficiency a deciduous J. Appl. in forest, Ecol., 25124, sociated with tropicalclimaticfluctuations suchastheENSO. 260,1987.

Theforest anefficient for O3during day. A large Browell, G.L.Gregory, Hatriss, V.W.J.H. was sink the E.V., R.C. and Kircchoff, Tro-

fluxwasobserved (2.3'10 molecules q) despite aml• cm-2s low

pospheric and ozone aerosol distributions theAmazon J. across basin,

bient concentrations atcanopy The deposition (- 6ppbv) top. high Geophys.93, Res., 1431-1451, 1988. velocity O3during day cm -• atcanopy appeared for the (1.8 s top) Campbell, An G.S., Introduction toEnvironmental Heidelberg Biophysics, toreflect large the stomatal available dense area inthe foliage Science 1977. of Library,
the absence significant photochemical of net production close and Denmead, T., and E. F. Bradley,On scalar O. transport plantcanopies, in couplingto the middle troposphere, suggest that the Amazon Irrig. Sci.8, 131-149,1987.

the trapical f•est. Denmead,and F.Bradley, O. T., E. Flux-gradient inaforest relationships Rapid deposition over Amazon combined canopy,Hicks,pp. 421-442, D. Reidel,Hingham,Mass.),1985. ofO3 the forest, with sonandB. The inB. Forest-Atmosphere by A.HutchinInteraction B. (edited

forests beasignificant global during wet Desjardins, Braek, and Schuepp, monitorcould sink for ozone the R.L., E.J. P. Alvo, P.H. Aircraft period (typically 9 monthsthe of year). sink 03isnon- ing surface dioxide This for of carbon exchange, 216, Science, 733-735, 1982.
existent the season, during dry because photochemical production Desjardins,J.L. R.L., MaePherson, and Schuepp, P.Alvo, P.H. Measureof03 isenhancedhigher by levels NO,derived higher of form soil ment turbulent and 2 exchange forests aircraft, of heat CO over from in:
emissions from biomass and burning[Crutzen al., 1985;Jacob TheForest-Atmosphere et Interactions, by B.A. Hutchinson B.B. ed. and andWofsy,1988;Andreae al., 1988]. Deforestation coloni- Hicks, Reidel, et and D. Hingham, Mass., pp645-658, 1985. zationof theAmazon Basincoulddramatically regional, alter and Droppo, Jr.,Concurrent J.G., measurements ofozone deposition dry using

perhaps global, budgets 0 3, by simultaneously for decreasing the

eddy correlation profile methods,Geophys. 90,2112and flux J. Res.,
of photosynthetic CO2 assimilation leavesof C3 species, in Planta, 149,

potential deposition increasing for photochemical for and rates 2118, 1985. production. Farquhar, S.yon G.D., Caemmerer,J.A. and Berry, biochemical A model
Acknowledgements. Wegratefully acknowledge collaboration ongoing 78-90, 1980.

with ABLE2B team with many the scienceand the individuals who helped Fi,'migan, Turbulent in flexible canopies, J.J., transport plan in The inthe (UNESP, Baum,study OsvaldoM.conditions, executionfield and difficult Cabral (EMBRA- Forest.Atmosphere by M.Hutchinson ofthe S.P.) under particularly Interaction, B. edited and B. B. JoseScolar R.
Hicks,pp. 443-480,D. Reidel,Hingham, Mass.,1985. PA, Manaus,AM). This work was supported NASA grants by NAG1-55 Fitzjarrald, D.R., B.L Stormwind, Fisch,andO.M.R. Cabral,Turbulent G. andNAGS-719,andby NSF grantATM-8413153to HarvardUniversity, observed abovethe Amazon just forest, Geophys. J. Res.,93, and by NASA grant NAG-I-692 to the Atmospheric Sciences Research transport 1551-1563, 1988. Center theStateUniversity New Yorkat Albany. at of FitzjarraldD.R., K.E. Moore, O. Cabral,J. Scolar,A.O. Manzi, and L.C.

de AbreuSa, Daytime turbulent exchange between Amazon the Forest andtheatmosphere, Geophys. J. Res.,thisissue.

Alvo,P.R.,R.C. Desjardins, Schuepp, J.I MacPherson, P.H. and Aircraft Fung, Analysis theseasonal geographical I., of and patterns atmospheric of measuremenu CO2 exchange various of over ecosystems, Boundary CO2distributions a three-dimensional model, TheChangwith tracer in Layer Meteorol., 167-183, 29, 1984. ingCarbon Cycle:A Global Analysis, edited J.R.Trabalka D.E. by and Bakwin, P.S.,S.C.Wofsy, S.-M.Fan, and Measurements of reactive nitro- Reichle, 459-473, pp. Springer-Verlag, York,1986. New gen oxides (NOy)within above tropical and a forest canopy, thewet Galbally, and in I.E., C.R.Roy,Destructionozone theEarth's of at surface,
season, Geophys. J. Res.,thisissue (a). Q.J. R. Meteorol. Soc.,106,599-620,1980. Bakwin,P.S.,S.C. Wofsy,S.-M. Fan,M. Keller,S. Trambore, J.M. Goreau, and TJ., andW.Z. DeMello,Effects deforestation sources of on and daCosta, Emission nitricoxide of (NO) fromtropical forest soils exand sinksof atmospheric carbon dioxide, nitrous oxide,andmethane from change NO between forest of the canopy atmospheric and boundary some Amazonian andsoils, Workshop Biogeochemistry biota in on of

layers, Geophys. this J. Res., issue (b).

Tropical Forest: Rain Problems Research, for edited D. Athaie, by T.E.

Baldocehi, D.D, B.B, Hicks,andP. Camaro, canopy A stomatal resistance Lovejoy,and P. deM. Oyens,Universidad Sao Pauloand World de modelfor gaseous deposition vegetated to surfaces, Atrnos. Environ., 21, Wildlife Fund,Diracicaba, S.P.,Brazil,1985.

91-101,1987.

Gregory, G.L, C.H. Hudgins, R. A. Edahl, Laboratory and Jr., evaluation

16,864

FAN E'I •,L.: CARBON DIt)XIDE

AND OZ()NE

EX('t!ANGE

ABLE-2B

of an airborne ozone instrument which compensates for mation, Atomic U.S. Energy Commission, Washington, D.C., 1970. altitude/sensitivity effects, Environ. Technol., 100-103, Sci. 17, 1983. Odum, H.T., G. Drewry J.R.Kline,Climate E1Verde, and at 1963-1966, in Gregory, G.L., E.V. Browell, L,S. Warren, and Boundary layerozone: a A tropical Rain Forest,edited H.T. OdumandR.F. Pigeon, by pp. airborne survey above Amazon the basin, Geophys. 93, 1452- BB47-B418, J. Res., Divisionof Technical Information, U.S. AtomicEnergy 1468,1988. Commission, Washington, D.C., 1970. Gregory, G.L., E. Browell,L. Warren,and C. Hudgins, Amazon Basin Ohtaki, Turbulent E., transport carbon of dio•dde a paddy over field,Bounozone andaerosol:Wet season observations, Geophys. J. Res.,thisisdaryLayerMeteorol., 315-336,1980. 19,
sue. Ratisbona, L.R., The climateof Brazil, in World Surveyof Climatology, Hicks,B. B., andR. T. Mc3/Iillan,On themeasurement drydeposition vol. 12,edited W. Schwerdfferger, of by Elsevier, NewYork, 1976. using impeffectsensors and in non-idealterrain,Boundary Layer Roberts, O. M. R. Cabral, L. F. deAguiar, J., and Stomatal boundaryand Meteorol. 79-94, 1988. 42, layerconductances measured a TerraFirmerainforest, Appl.Ecol., in J. Hicks, B.B.,D.D. Baldoechi, Meyers, Hosker, D.R.Matt,A T.P. R.P. and in press, 1990. preliminary multiple resistance routine deriving deposition for dry velo- Schlesinger, W.H., Carbon balance terrestrial in detritus, Annu. Rev. Ecol. citiesfrom measured quantities, Water Air Soil Pollut., 36, 311-330, Syst.,8,51-81,1977. 1987. Sehmel, A., Particle G. andgasdry deposition: review, A Atmos.EnHoughton,R. A., R. D. Boone, J. M. Melillo, C. A. Palm, G.M. viron.,14, 983-1011,1980.

Woodwell, Myers, Moore and L. Skole, fluxof carbonShuttleworth, J.H.C. N. B. 1ll, D. Net W.J., Gash, Lloyd, Moore, Roberts, C.R. C.J. J. A.O. dioxide fromtropical forests 1980, in Nature 316,617-620, 1985. Marques, Fisch, F. V.P. deSilva, M.N.G.Ribeiro, L.C.B.Molion, L.D. Jacob, D.J.,andS.C.Wofsy, Photochemistry of biogenic Emissions over Abreu J.C.A. Sa, Nobre, O.M,R.Cabral, Patel, J.C.deMoraes, S.R. and

the Amazon J.Geophys. 93,1477-1486, forest, Res., 1988. Eddycorrelation measurements of energy partition Amazonian for Jacob, and D.J., S.C.Wofsy, Budgets reactive of nitrogen, hydrocarbons, forest, J.R.Meteorol. 110,1143-1162, Q. Soc., 1984a. and ozone theAmazon during wetseason,Geophys. over forest the J. Shuttleworth, J.H.C. W.J., Gash, Lloyd, Moore, Roberts, C.R. C.J. J. A.O. Res., issue. this Marques,Fisch, deSilva, F. V.P. M.N.G. Ribeiro, L.C.B. Motion, L.D. Kaplan, S.C. W.A., Wofsy, Keller, J.M. Costa, M. and da ,Emission ofNO Abreu J.C.A. Sa, Nobre, O.M.R. Cabral, Patel, J.C. Moraes, S.R. and de and deposition in a tropical system,Geophys. 95, Observations of O3 forest J. Res., of radiation exchange and above below Amazonian forest, 1389-1395, 1988. Q.J.R.Meteorol. 110, Soc., 1163-1169, 1984b. Keller, W.A.Kaplan, S.C. M., and Wofsy, Emissions ofN20,CH4, and Trambore,E.,Carbon S. cycling gas and exchange in softs. Ph.D. thesis,
COg from tropical forestsofts, Geophys. J. Res.,91, 11,791-11,802, 194 pp., ColumbiaUniv., New York, 1988. 1986. Verma,S.B., andN.J. Rosenberg, Carbon dioxideconcentration flux and

Kirchhoff, V.WJ.H.,E.V. Browell, G.L. Gregory, and Ozone measure- in a largeagricultural regionof greatplainsof North America, GeoJ.
ments the troposphere an Amazonrainforestenvironment, Geoin of J.
phys.Res.,81,399-405, 1976.

phys.Res., 15,850-15,860, 93, 1988. Verstraete, M.M., Radiation transfer plant canopies: in Transmission of Kirchhoff, V.W.J.H.,i.M.O. da Silva,andE.V. Browell, Ozone measure- directsolarradiation theroleof leaf orientation, Geophys. and J. Res., meatsin Amazonia:Dry season versus season, Geophys. wet J. Res., 92, 10,985-10,995, I987. this issue. Webb,E.K., G.I. Pearman, R. Leuning, and Correction flux measureof Lemon, E.H., L.H. Allen,andL. Miller, Carbon dioxide exchange a of meatsfor densityeffectsdue to heat and watervapourtransfer, J. Q.
tropical rainforest,1I,Bioscience, 1054-1059,1970. 20,
Roy. Meteorol.Sot., 106, 85-100, 1980.

Lenschow, andL. Kristensen, D.H., Uncorrelated in turbulence noise Wesely, Improved M.L., parameterizations resistance forsurface to gasemeasuremeats, Ocean. J.Atmos. Technol., 2, 68-81, 1985. ous deposition dry in regional numerical scale, models, Agreement EPA Len•chow, R. Pearson, and D.H., Jr., B.B.Stankov, Estimating ozone DW89932394-01, the Environ. Agency, Prot. Washington, 1989. D.C., budget theboundary byuse aircraft in layer of measurements of ozoneWesely, and Hart, M.L., R.L. Variabilityshort eddy-correlation of term eseddy, and flux mean concentration, J. Geophys. 86,7291-7292, timaresmass Res., of exchange, Forest-Atmosphere edited inThe Interaction, 1981. byB.H.Hutchinson B.B. and Hicks, 591-612, Reidel, pp. D. Hingham,
Lenschow, D.H., R. Pearson, and B.B. Stankov, Jr., Measurements of
1982.

Mass., 1985.
tionsof ozoneeddyfluxesto maize,Boundary LayerMeteorol., 15,

'ozone vertical toocean forest, Geophys. 87,8833-8837,Wesely, flux and J. Res., M.L.,J.A.Eastman, Cook, B.B.Hicks, D.R. and Daytime variaLiu,S.C.,Model studies background formation, Tropospheric 361-373, of ozone in 1978. Ozone, Regional Global Interactions, and Seal NATO Series, ASI edited Wesely, M.L., D.R.Cook, E.M.Williams, and Field measurement of small byI.S.A.'Isaksen, D.Reidel, Hingham, Mass., 1988. ozonefluxes to snow,wet bare soil, and lake water,Boundary Layer Lumley, J.L.,andH.A. Panofsky, Structure Atmospheric The of Tur- Meteorol., 20, 459-471, 1981.
bulence, John Wiley, New York, 1964.
Wofsy,S.C.,R.C. Harriss, W.A. Kaplan,Carbon and dioxide the atmoin

Matthews, Global E., vegetation landuse: and Newhigh resolution data sphere theAmazon over basin, Geophys: J. Res.,93, 1377-1387, 1988. bases climate for studies,Appl. J. Meteorol., 22,474-487, 1983. Zimmerman, P.R., J.P. Greenberg, C.E. Westberg, and Measurements of MclVIiilen,T., Aneddy R. correlation technique extended with applicabili- atmospheric hydrocarbons biogenic and emission fluxesin the Amazon
ty tonon-simple terrain, Boundary LayerMeteorol., 231-245, 43, 1988.
boundary layer,J. Geophys. Res.,93, 1407-1416, 1988.

Medina, and Klinge, E., H. Productivitytropical of forests tropical and
woodlands, Encycl. PlantPhysiology, 281-304, 12D, 1983.

Meyers,and Baldocchi, T.P., D.D. Acomparison for of models deriving Fan, Wofsy, Bakwin, J. Division s.-M. S.C. P.S. and Jacob of D. Applied
dry deposition ofO3and toa forest fluxes SO2 canopy, 40B, SciencesDepartment and Tellus, and ofEarth Planetary Sciences, UniverHarvard
270-284, 1988. sity,Cambridge, 02138. MA

Mooney, P.M. H.A., Vitousek, P.A. and Matson, Exchange ofmaterials Fitzjarrald, D.R. Atmopsheric Research State Sciences Center, University between terrestrial ecosystems atmosphere, 238, and the Science, 926- ofNew atAlbany, NY 12205. York Albany,
932, 1987.

Odum, and Jordan, H.T., C.F. Metabolism evapotranspiration and ofthe
lowerforest a giantcylinder, A Tropical in in RainForest, edited by

(Received 8,1989; June
revised February 1990; 20,

H.T.Odum R.F.Pigeon, 1165-1189, and pp. Division Technical of Infor-

accepted 5, 1990.) March