Convective transport of biomass burning ... - Wiley Online Library

18 downloads 0 Views 2MB Size Report
Oct 30, 1996 - estimates summed over the first four days of transport com- ..... (ascent 4, descents 1 and 3) have CO mixing ratios typically between 100 and ...
JOURNAL

OF GEOPHYSICAL

RESEARCH,

VOL. 101, NO. D19, PAGES 23,993-24,012, OCTOBER

Convective transport of biomass burning emissions over Brazil during TRACE A Kenneth E. Pickering,• Anne M. Thompson,•,2 Yansen Wang,3 Wei-Kuo Tao,2 Donna P. McNamara, 4 Volker W. J. H. Kirchhoff, 5 Brian G. Heikes, 6

Glen W. Sachse, 7 John D. Bradshaw,8 Gerald L. Gregory,7 and Donald R. Blake9

Abstract. A seriesof large mesoscaleconvectivesystemsthat occurred during the Brazilianphaseof GTE/TRACE A (Transportand AtmosphericChemistrynear the Equator-Atlantic) providedan opportunityto observedeep convectivetransportof trace gasesfrom biomassburning.This paper reports a detailed analysisof flight 6, on September27, 1992,which sampledcloud- and biomass-burning-perturbed regionsnorth of Brasilia.High-frequencysamplingof cloud outflow at 9-12 km from the NASA DC-8 showedenhancementof CO mixingratios typicallya factor of 3 abovebackground(200-

300 partsper billion by volume(ppbv) versus90 ppbv) and significantincreasesin NOx and hydrocarbons.Clear signalsof lightning-generatedNO were detected;we estimate that at least 40% of NOx at the 9.5-km level and 32% at 11.3 km originatedfrom lightning.Four typesof model studieshave been performedto analyzethe dynamicaland photochemicalcharacteristics of the seriesof convectiveevents.(1) Regionalsimulations for the period havebeen performedwith the NCAR/Penn State mesoscalemodel (MM5), includingtracer transportof carbonmonoxide,initializedwith observations.Middle-upper troposphericenhancements of a factor of 3 abovebackgroundare reproduced.(2) A cloud-resolving model (the Goddardcumulusensemble(GCE) model) hasbeen run for one representativeconvectivecell duringthe September26-27 episode.(3) Photochemical calculations(the Goddardtroposphericchemicalmodel), initializedwith trace gas observations (e.g., CO, NOx, hydrocarbons, 03) observedin cloudoutflow,show appreciable0 3 formationpostconvection, initially up to 7-8 ppbv O3/d. (4) Forward trajectoriesfrom cloud outflowlevels(postconvective conditions)put the ozone-producing air massesin easternBrazil and the tropicalAtlantic within 2-4 daysand over the Atlantic, Africa, and the Indian Ocean in 6-8 days.Indeed, 3-4 daysafter the convective episode(September30, 1992), upper troposphericlevelsin the Natal ozonesounding showan averageincreaseof -30 ppbv (3 Dobsonunits (DU) integrated)comparedto the September28 sounding.Our simulatednet 0 3 productionrates in cloud outflow are a factor of 3 or more greater than those in air undisturbedby the storms.Integrated over the 8- to 16-km cloud outflowlayer, the postconvectionnet 0 3 production(-5-6 DU over 8 days)accountsfor -25% of the excess03 (15-25 DU) over the SouthAtlantic. Comparisonof TRACE A Brazilian ozonesondesand the frequencyof deep convection with climatology[Kirchhoffet al., this issue]suggests that the late September1992 conditionsrepresentedan unusuallyactiveperiod for both convectionand upper troposphericozone formation.

•JointCenterfor EarthSystemScience, Universityof Maryland,CollegePark. 2NASAGoddardSpaceFlightCenter,Greenbelt, Maryland. 3Science, Systems, andApplications, Inc,Lanham,Maryland. 4Applied Research Corporation, Landover, Maryland. SlNPE,SaoJosedosCampos, SP,Brazil. 6Graduate Schoolof Oceanography, Universityof RhodeIsland,Narragansett. 7NASA/Langley Research Center,Hampton,Virginia. 8Georgia Instituteof Technology, Atlanta. 9University of California-Irvine, Irvine. Copyright1996 by the American GeophysicalUnion. Paper number 96JD00346. 0148-0227/96/96JD-00346 $09.00

23,993

30, 1996

23,994

1.

PICKERING

ET AL.' CONVECTIVE

Introduction

Deep convectionis a major componentof the atmospheric generalcirculation.Regionsof convectionthat form important upwardbranchesof the Walker circulationare locatedon the eastern and western edgesof the study region for the NASA GTE/TRACE A experiment.This paper focuseson deep convectionthat occurredoverthe westernpart of the studyregion, the interior of Brazil, during the field mission. Chatfieldand Crutzen[1984]hypothesizedthat deepconvection is a major redistributorof trace gasesfrom the boundary layer to the free troposphere,andDickersonet al. [1987]presented observational

evidence

that such redistribution

occurs.

Pickeringet al. [1990] showedthat onceozoneprecursorgases are transportedto the middle and upper troposphereby deep convection,ozoneproductionin the cloudoutflowcanbe substantiallyenhanced.Pickeringet al. [1991] usedmeasurements from a previousNASA/GTE field missionin Brazil (Amazon BoundaryLayer Expedition(ABLE) 2A) in studyingthe effects on ozone production of a dry seasonsquall line over Amazonia, usinga convectivecloud model and a photochemical model in tandem.A pollutionplume from distantbiomass fires was evident in these measurements.When convectively lifted, this pollution was capable of convertinga significant layer from one of 03 destructionto one of 03 production.In the Pickeringet al. [1992a] paper we performed a sensitivity study using the cloud dynamicsof the ABLE 2A event to examinethe effectsof deep convectionon ozoneproductionin two scenarioswith relativelyfresh biomassburningpollution. With an assumptionof a pristineupper tropospherewe found that column ozone production in cloud outflow may be enhancedby at least a factor of 50. Chatfieldand Delany [1990] useda "travelingone-dimensional"model to alsoestimatethe downstreameffectsof convectivetransportof biomassburning pollution.KirchhoffandMarinho [1994]havenotedevidenceof convectivetransport of ozone in ozonesondeprofiles taken over Brazil during the burning season.Chatfieldand Delany [1990] and Pickeringet al. [1992a] hypothesizedthat ozone producedin this manner couldbe transportedover the South Atlantic

and could contribute

to the ozone maximum

detected

by satellite[Fishmanet al., 1990; Watsonet al., 1990]. While contributinga relatively smaller amount to total column 0 3 than similar concentrationsin the lower troposphere,the 0 3 producedin the upper troposphereas a result of convective transporthas greater importancefor climateforcing[Laciset al., 1990]. The aircraftmissionsduringthe TRACE A experimentwere aimed at samplingall of the major featuresof the circulation contributing air to the region of the South Atlantic ozone maximum.During the Brazilian componentof the experiment a major seriesof deep convectivesystemsdevelopedover the cerrado.One of the flights(flight 6) from the TRACE A base in Brasilia was devoted to samplingthe outflow from these systems,althoughsomecloud-processed air was alsonoted on flight 5 (alongthe Braziliancoast)and flight 7 (over the cerrado). These flightsdocumentthe large-scaleeffectsof deep convectionon trace gas distributionsin the South Atlantic roc•{nn(-•7nnocnnclo rnonq•r•rnontqqhowthe impact of unner level transportof ozone from the cerrado region of Brazil to Natal and to a number of more distantlocations[See Thompsonet al., this issue].The combinationof aircraft and ozonesonde measurementsprovidesthe opportunityto verify our

TRANSPORT

OVER BRAZIL

previousmodel resultsconcerningtransportof biomassburning emissionsand subsequentdownstreamozone production. We document the occurrenceof a major series of deep convectiveepisodesduringthe latter part of the cerradoburning seasonin 1992.We presenttrace gasmeasurementstaken from the NASA DC-8 and a Brazilian aircraft that are representativeof the low-altitudepollutionrelatedto biomassburning and of the outflowfrom the deep convectionthat occurred on September26-27, 1992.Becausethe aircraftmeasurements in cloud-processed air were of limited spatial and temporal duration,we use a sequenceof mesoscaleand cloud-resolving model

simulations

to extend the measurements

to determine

the full extent of the effects of the convectivesystemson regional trace gas distributions.A detailed descriptionof the mesoscale(MM5) modelingactivityfor the September26-27 eventsis given by Wang et al. [this issue].We also use the three-dimensional Goddardcumulusensemble(GCE) model to simulateone representativeconvectivecell, providingwind fieldsfor usein transportcalculationsfor a varietyof species. Forward trajectoriesare computedusingthe methodsrecommendedby Pickeringet al. [this issue]to determinethe longrangetransportof the cloudoutflowat upper levels.A photochemicalbox model (similar to the GSFC one-dimensional troposphericmodel used in our previousstudies)is used to estimate the net ozone production during the downstream transport.This ozone productionproceedsat the rate of several parts per billion by volume (pppv) per day, as theseair massesare transportedin upperlevelwesterliesoverthe South Atlantic.

Section2 of this paper presentsa descriptionof the convective eventsleadingup to flight 6, aswell astheir relationshipto the biomassfires in the cerradoregion.In section3 we display and discussthe trace gasmeasurementsmade on flight 6 and on a flight of the Brazilian aircraft. Section4 describesthe use of measurementsfrom undisturbedair to provide initial conditionsfor our mesoscalemodel simulationof the major convective systemsimmediatelyprecedingflight 6. Tracer transport results for CO are also presented in this section and comparedwith DC-8 observations in cloud-processed air. Section 5 presentsa simulationwith the GCE model of one convectivecell containedwithin one of the systemssampledon flight 6. We alsopresentcloud-scaletracer transportresultsfor other speciessuchas NOx, 03, and total nonmethanehydrocarbon(NMHC). In section6 we show8-dayforwardtrajectories from the cloud systemsand from the measurementlocations, as well as photochemical calculations of ozone productionalongthesetransportpaths.Our ozoneproduction estimatessummedover the first four daysof transportcompare favorablywith the upper troposphericozone increaseat Natal, Brazil, noted in ozonesondemeasurementsfollowing the convectiveepisode.In section7 we concludewith a summary of the impact of the Brazilian convectionon ozonein the South Atlantic region.

2.

Synoptic Conditions

2.1.

Meteorology

The 1992 dry seasonover much of the cerrado region of Brazil ended earlier than normal with a number

of cold fronts

arriving from the southwestproducingshoweractivityduring September.Rainfall statisticsfor the regionare givenbyKirchhoff et al. [this issue].An abnormallyhigh frequencyof deep convection

is also noted from satellite

measurements

of out-

PICKERINGET AL.' CONVECTIVETRANSPORT OVERBRAZIL

going longwave radiation (seeFigures la-ld). In September,

almostall of Brazilis seento havemonthlymeanvaluesof outgoing longwave radiation lessthanthe 1979-1988average. Valueslowerthannormalindicate a greateramountof deep convection thannormal.The September anomalyexceeded 30 W m-2 overa largeregionto thewestof Brasilia.In October thepatterncontinued withabnormally lowvalues of outgoing •

23,995

ECMWF 925mb Winds, Sept. 22199212Z

[

• ....

,J .......

-.•--..*--.*-..--..-.. x x \ \ x.x x ', x • x 'rS T, , ,%7h2..

-10

....

longwave radiation (abnormally highfrequency of deepconvection)over all of Brazil between10øSand 25øS,but the anomaly wasnot soextremeasin September. Duringthe TRACE A DC-8 flightsin Brazila majorout-

breakof deepconvection developed overtheregion(September 22-28).The seriesof convective eventsbeganin southern -

50 -100

,

-80

• I

,

-60

30m/s , I





-40

,

I

-20

,

,

,

I

,



0

,

I

,

20

Degrees Longitude

(a)

ECMWl• ?25mb W, inds, Sept. 26199,2 12Z ß

ß

20S

.......

40S 5 90W

0

•.f' •'•'-_' ,. xx......

..

.

............ N%'k'----:,)-_,

90E

(b) , •o-•¾. :10 m/s ' EQ

_50• -100

90W

0

•260

!'



240/

,.-220• z/

/

'•

1280 200 • • •[••

/• o 90W



220•

-60

:

20m/s 30 m/s

-40

-20

0

20

Figure 2. ECMWF wind vectorplotsat 925 mbar for 1200 UT on (a) September22, (b) September26, 1992.Frontal position(solidline) shownfor September22 andlinesof convergence(dashedlines)shownfor September 26.

90E

(c) •,[20._• X,,,,'" X•_•--"••'"'•.•.•

-80



Degrees Longitude

20S

40S

• /

•.220•



.

•200• 0

•"•'•

220

m 20S

40S

Brazilassociated witha coldfrontalsystem pushing northeastwardfromthe midlatitudes of the southern hemisphere. The frontalpositionon September 22 is depictedon the 925-mbar windvectorplotin Figure2a.Thundershower activityonSeptember22 wasnotedasfar northasCuiaba(--•16øS, 57øW). The frontremainednearlystationary on September 23 and24 withconvective activityto the northof the front,althoughat the surface little thermal contrast remained acrossthe front.

.200 90E

However,a strongzone of low-levelconvergence persisted overtheregioncontinuing to inducedeepconvective activity. The front rapidly advancedto the north between 1200 UT

(d)

September 25 and1200UT September 26 [seeBachmeier and Fuelberg,this issue].Stormsoccurredas far north and eastas

m

8øSand50øWon September 25. Largemesoscale convective systems developedon September26 evenfarther to the north

20s andeast(seeMeteosat4 imageryin Plate 1) andcontinued through theearlymorning hoursof September 27.Thesesystems are associated with the zonesof convergence seenin 40S Figure 2b. 90W

0

90E

Figure 1. Outgoinglongwaveradiation (W m-2) from NOAA 11AVHRR IR windowchannel.(a) September 1992 mean;(b) September 1992anomaly;(c) October1992mean; (d) October1992anomaly. Anomaliesare departures from 1979 to 1988 means for each month.

The morenortherlyline of convection (centeredon about 8øS)developed first(before1530UT onSeptember 26).A line of stormsextendedfrom approximately 46ø to 60øW.The stormsin thisline reachedmaximumintensityat about1830 UT (Platela), withsomecloudtoptemperatures below195K, indicatingcloudtop altitudesof at least16 km. A more southerly (10ø-15øS)north-southorientedline locatedbetween45ø

23,996

PICKERING

ET AL.' CONVECTIVE

TRANSPORT

OVER BRAZIL

TRACE-A M06, 27 Sep. 92

and 50ø W developed by 2130 UT and reached maximum intensity between 2130 UT on September 26 and 0330 UT

500

'

(Plate lb) on September27. Digital imageryis not availablein

[]

this time interval, but low-resolutionMeteosat 3 imagery receivedin the field suggests peak developmentabout 0130 UT on September27. Cloud top temperaturesin this systemwere lessthan 205 K at 2130 UT, equivalentto cloudtop heightsof at least 15 km. Thesesystemsoccurredover regionsof biomass burningactivity,and TRACE A flight 6 was designedto capture the outflow from these storms. The approximateflight

2.2.

ensemble

Biomass

4

100

,

8

10

12

14

GMT(in hours)

Fires

onboard

the NOAA

satellites

• •

200

model.

instrument

10

300

16

0

18

csvc• a/95

I []Polluted BL• Cid Processed XClean Upper Trap. 0Other

A weekly record of fire pixel counts estimated from the AVHRR

[]

400

trackfor flight6 (flownfrom -0900 to -1630 UT on September 27) is superimposedon the satellite imagery.It is these systemson September26 and 27 that we simulatewith MMS. One cell from the secondsystemis simulatedwith the Goddard cumulus

12

(a)

has been

provided by A. Setzer of INPE. Figure 3 showsthe weekly averagedfire pixel countsper day from mid-Augustthrough the end of October 1992, for the region in which the DC-8 flights(flights6 and 7) over Brazilian biomassburningwere conducted.Peakfire countsper day(approximately 2000)were detectedduring the August 28 to September3 period. By the time the DC-8 reached Brasilia (week of September25 to October 1), the daily averagehad declinedto approximately 700 per day,or aboutone third of the peakweeklyvaluein the period shown. The AVHRR datawere processedat NASA GoddardSpace Flight Center for the individualdaysduringthe September25 to October 1 period, usingthe algorithmof Justiceet al. [this issue](SAFARI JGR specialissue).This is the samealgorithm usedfor fire countsin Africa duringthe SAFARI and TRACE A experiments.We use the Justicefire countproduct(avail-

100-' ........

g 80

TRACE-A M06, 27 Sep. 92 , ......... , ..... •', ......... , .........

C1

•2

• 60 o

(b)

- 12

C7 C9

o

10ß

o

40



4

0



8

10

C2

• 12

GMT(in hours)

•6, ...... 14

C8 -,I .....



16

0

18

•src/• a/95

I • Pallured BL• Cid Praeessed XClean Upper Trap. OO•er INPE

AVHRR

Fire Count

Data

6-14 deg S 43-49deg W 2000

1500

1000

5OO

i {

Figure 4.

Summa• of DC-8 chemical measurementson

T•CE A flight6 (a) 0 3, (b) CO. Ascents(A), descents (D), andconstant-altitude (C) legsare indicated.Three regimesare denotedwith indicatedsymbolsand defined as follows:pollutedBL, altitude 300 ppbv;cloud-processed, altitude >6 km and CO >150 ppbv, and cleanupper troposphere,altitude >6 km and CO 1 and only a few are 70

-

70

cloud-scaletracer transport results.Section 4.3 presentsthe resultsof the regional-scaleCO transport calculationsusing the MM5 fields. The regional CO distribution in cloudprocessedair is the basisfor initializingforward trajectoriesto determinelonger-rangetransportof cloudoutflow(seesection 5) and areasof the SouthAtlantic basinwhere 03 producedin the outflow

60

5O

40

-20

30

. 20 .

-25

would be found.

10 Maxfirevalue= 147

4.1.

Initial

Conditions

for Tracer

Studies

-30

Lacking detailed CO emissiondata, we rely on the aircraft CO measurementsto develop initial conditions for model tracersimulations.The DC-8 CO profilemeasurementsin very polluted air and relativelyclean air in the lower troposphere (e.g., Figure 5) and constant-altitude flight leg averages(Figure 6) are usedto constructinitial conditionCO profiles(Figure 10)for "polluted"air, "semipolluted"air, and "clean"air. The "semipolluted"profile is basedon the INPE aircraft profile (Figure 7) taken in moderatelypollutedair. The dailygriddedfire countsare usedto distributethe three CO profiles(Figure 11) The "polluted"profile is assignedto regionswith major concentrationsof fires noted on the September26 fire map. Major areasof firesnotedon the September 25 fire map and not also appearingin the September26 data were assignedthe "semipolluted"profile. Remaining areasin the MM5 fine meshdomainwere designatedas "clean." Uncertaintiesin this assignment includeprobablebias toward lower than actual mixing ratios and toward a polluted region that is smallerthan what may haveexisted.For example,some

20

CO Tracer Initial Conditions 26 Sept. 1992 I

I

-

No.of1Xlkm

-6O

-50

-40

-30

Degrees Longitude

Figure 11. Combined AVHRR fire counts for September 25-26, 1992, from Justiceet al. algorithm.Geographicdistribution of regionsdesignatedas containing"polluted"(dashed line boxes),"semipolluted"(solidline boxes),and "clean"(remainingarea) air.

of the areasfor whichwe usethe "clean"profilewere probably polluted by low-level transport from fire-rich areas. In addition, eventhe "polluted"profileis basedon data from polluted layers already 12-24 hours old. Initial

conditions

for the cloud-scale

tracer

simulation

as-

sume the convectivecell formed over a polluted region becausethe more southerlyconvectivesystemdevelopedover a fire-rich region on September26. The "polluted" CO profile (Figure 10) is usedfor initially undisturbedconditionsalong with NOx, 03, and total NMHC profilesderivedfrom the same flight segmentsusedin developingthe CO profile (see Table 2). 4.2. 26-27

Cloud-Resolving Model Simulation of September Convective

Cell

I

"Polluted"

.......

Clean"

Air

Air

The MM5 simulationcompletesthe CO picture for which cloud outflow

measurements

were restricted

to 300- to 400-km

flight segmentsat 9.5 and 11.3 km. For the vertical structureof 15 _ "Semi-polluted" Air_ other trace gasesrequired in photochemicalcalculationswe turn to the cloud-resolvingmodel. The three-dimensional GCE model is usedto simulateone representativeconvective cell from the more southerlyof the two major convectivesystems of the September26-27 episode.Details of the model can be found in the work of Tao and Simpson[1993]. The initial soundingused in the GCE model simulation is selectedfrom the MM5 mesoscalemodel output [Wanget al., thisissue]at 1930UT September26 at 14øS,49øW),whichis in the region where the second and more southerly convective systemdeveloped.The GCE model showsan intenseconvective cell growingto a height of about 16 km, in good agreement I II 100 200 300 400 with the satellite imagery.At 180 min into the simulationthe [CO] (ppbv) cell beganto dissipate.We ran the tracertransportcalculations from the initiation of the cell throughthe 180-min mark. When Figure 10. CO tracer initial conditionsbasedon DC-8 flight we comparethe averageDC-8 observationsat 9.5 and 11.3 km 6 and INPE aircraftmeasurements. The "polluted"air profile comes from DC-8 ascents 1 and 3, descent 2, and the constant- with calculatedvalues averagedover the distance(---40 km) altitude segments3 and 6. The "clean" air profile is derived coveredby the DC-8 during 3 min, the simulatedCO and NOx from ascent 4, descents1 and 3, and constant-altitudeseg- mixing ratios are too low. Simulated 03 was slightly greater air at 9.5 km; the simulated ments2 and 8. "Semipolluted"profile comesfrom INPE air- than observedin cloud-processed craft measurements. total NMHC was comparableto the measurements.As de-

PICKERING Table

2.

Initial

Altitudes,km 0.00-1.05 1.05-1.73 1.73-2.53 2.53-3.92 3.92-4.99 4.99-6.16 6.16-7.44 7.44-8.12 8.12-8.83 8.83-12.78 12.78-16.42 >16.42

ET AL.: CONVECTIVE

Conditions

for GCE

Tracer

24,001

CO, ppbv

NOx, pptv

03, ppbv

Total NMHC, ppbv C

260 310 223 359 153 182 147 89 95 95 90 30

490 295 156 323 102 112 100 126 177 208 310 500

49 49 49 72 60 77 74 70 74 76 100 150

12.8 13.9 10.4 13.1 4.1 4.7 2.5 2.6 3.1 3.6 2.5 0.5

CO and NOx mixingratiosare too low. Therefore we adjusted the initial conditionvalues of CO and NOx in the polluted layers of the lower tropospheresuch that the model more closelymatchedthe observations (measuredCO and our estimated "No lightningNOx" (Table 1)) at the cloud outflow samplingaltitudes.This processrequireda factor of 1.3 for CO and a factor of 1.8 for NOx. We also adjusted low-level 03 valuesby a factor of 0.85, so that the upper level 03 matched the measurements.Apparently, the NMHC measurements (basedon grab samples)in the pollutedlayerswere represenbefore

OVER BRAZIL

Simulation

scribed in section 4.1 we have reason to believe that the initial

tative of conditions

TRANSPORT

issue]).Vertical transportis achievedusinggrid-scale(30 km horizontalresolution)windsplusthe subgridconvectivetransport informationprovidedfrom the cumulusparameterization in MM5. The regional tracer transportsimulationwas initialized at 1200 UT September26 with the CO mixing ratio distribution described in section 4.1. However, the CO initial

valueson the "polluted" and "semipolluted"profileswere adjusted in the lower tropospherein the samemanner as in the cloud-scaletracer simulation(factor of 1.3) to estimatemixing ratios for the lower tropospherethat may have existedat the onset of convection.

MM5

simulations

were

conducted

with

the storm. Plate 2 shows the final

the Grell and Kain-Fritsch cumulus parameterizations,and tracer transportcalculationsusingthe output of both schemes the coreupdraft regionof the stormexceeds250 ppbvand the are reported by Wanget al. [this issue].Tracer simulations area perturbedby the cell is ---lee km in horizontaldimension. usingthe Grell schemeproduceda slightlybetter representaLower 03 from the boundary layer appears in the updraft tion of the mean and frequencydistributionof the CO mearegionat 9.5 km, whereasdowndraftsat the outer edgesof the surements,particularlyat 11.3 km. Therefore in this paperwe cell bring higher 03 downward.At 11 km a region coveringa useonlythe tracertransportcalculationsresultingfrom the use distance of ---lee km in the cloud-scale simulation has ---65 cloud-scale

simulation

tracer results for the 9.5-km level. CO in

ppbv 0 3. Measurementsmade in cloud-processed air at 11.3 km 4-5 hoursafter sunriseare >75 ppbv,showingthat appreciable O3 production had already taken place in the cloud outflow.Table 3 presentsstatisticsfrom the simulationfor the 9.5-, 11-, and 14-km levels (the latter used in upper troposphericphotochemicalmodel calculationsof 03 production (seesection5.2)).

of the Grell

scheme.

Plate 3 presentsthe simulateddistributionsof CO at altitudes 9.5 and 11 km at 1200 UT September27. A region of > 1500km in horizontaldimensionhasbeenperturbedat these altitudesby the two convectivesystems;this is about 6 hours after dissipationof the second,more southerlysystem,and about midway between the DC-8 samplingtimes at 9.5 and 11.3 km. With very light upper level winds, CO peaks in the 4.3. Regional CO Transport vicinitiesof individual stormsstill persist.CO perturbations The three-dimensional wind fields from the MM5 simulation extendto the top of the cloud at approximately16 km, having were used in a tracer transport code containing a positive- tapered off in horizontal dimensionabove 12 km to only 400 definite advectionscheme(see details by Wang et al. [this km at 16 km altitude. The cloud-perturbedregion is also Table

3.

Statistics From

Cloud-Scale

CO, ppbv

Tracer

Simulation

NOx, pptv

03, ppbv

Total NMHC, ppbv C

9.5 km

Maximum Minimum

294 95

407 208

85 59

9.5 3.6

120-km average

182

280

72

6.0

11 km

Maximum Minimum

340 95

450 216

85 59

120-km average

243

361

64

Maximum Minimum

254 90

14 km 495 315

120-km average

165

368

108 70

87

10.6 3.5

8.0 8.5 2.5

5.3

24,002

PICKERING

ET AL.: CONVECTIVE

ß

TRANSPORT

- 0•

OVER BRAZIL

qO

TEHP, 23:5-

"20

220-

210

210-

205

205

-

195


150ppbv(Plate 3). Parcelpositionsin 12- to 16-kmlayerat (c) 4 daysand (d) 8 days.Squaresrepresentparcelshavinginitially >250 ppbv CO, while pointsindicateparcelsinitially having150-250 ppbv CO.

the convective episode)hasthe highesttropospheric ozoneand the highest12 km-to-tropopause ozoneof the 14 Natal soundingsin the September-October1992period (Figure 17). Back trajectoriesarriving at four levels between 11 and 16 km at tion amounts. Natal on September30 (Figure 18) all showflow to Natal from TRACE A flight 7 on September28 (---24hoursafter flight the vicinityof the September26-27 eventsover periodsrang6) sampleda regionat 11.3km about300 km eastof the center ing from 2 to 4 days.The box model calculationsshow net of the secondand more southerlyconvectivesystemof Sep- ozoneproductionover the first 4 daysof transportto be ---24 tember 26-27. Ozone mixing ratios on flight 7 at this altitude ppbv at 11.3 km and >8 ppbv at 14 km. Thus we argue that averaged88 ppbv,whereasthe averageon flight6 was78 ppbv. most of the September28-30 increasein ozone at Natal was High CO mixingratios(averageof 210 ppbv) suggestthat the causedby photochemicalproduction in convectiveoutflow flight 7 air masscontainedconvectiveoutflow, althoughthe from the September26 to 27 storms. H202/CH3OOH ratios reach a minimum of slightly>1. Trajectoriesin Figure 18 suggestthat 300 km is approximatelythe 6. Discussion distancetraveledby the convectiveoutflowfrom this systemin 24 hours.Therefore the DC-8 03 measurements showa net 03 Throughthis casestudywe have shownthat mesoscaleconproductionrate of ---10 ppbv for the first day of downstream vectivesystemsover biomassburningregionsof the Brazilian transportand that some H202 has been produced.Our box cerradocancausemajor 03 increasesin the uppertroposphere model calculationof 7-8 ppbv 03 for the first day nearly at Natal. Thompsonet al. [this issue]have noted that Natal matches this rate. experiencedupper level westerliesfor 100% of the soundings Ozonesonde measurements at Natal, Brazil [Kirchhoffet al., duringTRACE A, while at Ascensionthe frequencywas 87%. this issue]showan average31 ppbv increasein upper tropo- Theseresultsand our forwardtrajectories(Figures13 and 14) spheric(11-16 km) 03 betweenSeptember28 and30 (Table 7) suggestthat thesesitesare frequentlydownstreamof the Braand an averageincreaseof 26 ppbvin the 11- to 12-kminterval zilian convectiveoutflowand that 03 producedin the outflow (Figure 16). In fact,the September30 sounding(3-4 daysafter may dominatethe uppertropospheric03 distributionat these incorporationof trajectoryparametersis not expectedto give significantlydifferentresults.Also shownin this figureare the locationsof downstreampostconvectionmeasured03 incrementswhich may be comparedwith the simulated03 produc-

PICKERING

ET AL.: CONVECTIVE

TRANSPORT OVER BRAZIL

locations. For example, to12-km and the 12to16-km a) forwardtrajectories andthe box8model 03 production calculations showthat AscensionIslandcouldhavesampledthis convective outflow4 to 6 daysafter the event(October 1-3) with asmuch as 30 ppbv of excess03. Upper troposphericozone on the October3, 1992,soundingshowsa peak of 20-30 ppbv excess 03 at 13-15 km and other lower-altitudepeaks,but no tracers were measured to allow definitive source assignment.From trajectories it appears that this feature is a combination of outflow from the September26 to 27 cerrado stormsand earlier convectiveevents that occurred farther upstream. The effectsof the September26-27 eventsmay reach the Brazzaville soundinglocationin about8 days(Figure14) in the 12- to

8-12km Layer EndofDay4 EndofDay8 40

o

N 20 O

16-km layer. At theendof8 days theboxmodel shows net03

10

soundings at Brazzaville, indicating thatthepotential impact

0

productionin the 12- to 16-km layer totaling 15 ppbv. Upper level westerlieswere noted for only 23% of the TRACE A

we have seen at this site from the September26 to 27 storms may be a relatively infrequent occurrence.However, none of the upper level back trajectoriesfrom individual Brazzaville

All Cloud Outflow*

Peak 9.5km#

(Equivalent Dobson Units)

Peak 11.3krr•

# 9-min.peakcloudoutflow

12-16km Layer

20

particularlyin the 8- to 12-kin layer, the soundingsite at Reoutflow

Undisturbed

* Avg. of all cloud outflow

soundings show flow from Brazil [Thompson etIndian am., this issue]. •) Becauseour trajectories also show flow to the Ocean, union Island could have received convective

24,009

Endof Day4 I

from the

]

EndofDay8

September26-27 Brazilian events. Pickeringet al. [1992a]estimatedthe amountof 03 production during the first 24 hoursfollowinga prototypedeep convective event over the Brazilian savannaduring the burning season.Chatfieldand Delany[1990]alsosimulated0 3 production in cloudoutflowfor stormsoverbiomassburningregions

••• 10

of Brazil.In thePickering et al. [1992a]analysis we assumed

•'

initial lower-tropospheric mixing ratios (e.g., 500pptv NO,450

(•

I

o

ppbv CO) basedon earlier field expeditionsto Brazil during

biomass burning. In thesimulations withclouddynamics based

s

on a squall line sampled during the NASA GTE/ABLE 2A fieldmissionoverBrazil,cloud-processed air had 600-900 pptv

NO•inthe9.5-to 11-km layer,similar tothepeak9-minNOx

0

measuredin the TRACE A flight 6 cloudoutflow.An average 03 productionvalue in cloud outflow from 5 to 13 km was 7.4 ppbvO3/d,with valuesin the 9.5- to 1l-km layerof 4-5 ppbv/d.

Undisturbed

(EquivalentDobson Units)

14km Cloud Outflow * ß120kin avg. from GCE cloud outflow @ 14kin

Figure 15. Summaryof photochemicalbox model simulation Table

6.

Initial

Conditions

for Box Model

Simulations

in

12- to 16-km Layer 14 km

Peak Outflow

03, ppbv NO, pptv NO 2 pptv CO, ppbv C2H6, ppbv C3H6, pptv C3H8, pptv C2H4, pptv TOL, pptv

87.5 293 77 164 1.80 5.7 339 8.1 0.9

(8 days)of total net 0 3 productionfor days1-4 and 5-8: (a) for 8-to 12-km layer and (b) for 12-to 16-km layer. Model initializedwith measurementsfrom flight 6 in 8-to 12-km layer and with output from cloud-scalesimulationfor 12-to 16-km layer.

Undisturbed

102 228 68 86 0.87 2.7 166 3.8 0.4

Taken from steadystateversionof model in which fluxesare parameterized to match GCE model output for 14 km. Undisturbed NOx, CO, and hydrocarbons at 14 km estimatedby extrapolating9.5 to 11.3 km trend in undisturbedmixingratios. 03 at 14 km taken from UV-DIAL [Browellet al., this issue]measurementsfor this altitude. Invariantsinclude HNO3, PAN, H202, and H2CO which are set to valuesappropriatefor 14 km taken from one-dimensionalmodel simulationsdiscussed by Pickeringet al. [1992a].

Chatfieldand Delany [1990]estimated6 ppbv/dfor the first 24 hours

downstream

of an idealized

"mix-then-cook"

event.

These 03 productionrates are similar to those obtained from the box model using the trace gas measurementsfrom the September26 to 27 cloudoutflowsampledon TRACE A flight 6. We have confirmedthe postconvective upper tropospheric ozone enhancementpredicted in our previouswork and by Chatfield and Delany.

7.

Summary and Conclusions

Observationsin cloud-processed air following the September 26-27, 1992,convectiveepisodeoverthe Braziliancerrado provide evidenceof vigorousconvectivetransportof biomass burningemissionsto the upper troposphere.Enhancementof

24,010

PICKERING

ET AL.' CONVECTIVE

TRANSPORT

Natal Layer Mean Ozone Concentrationsfrom Ozonesondes

Cumulative Ozone Production from 26-27 Sept 92 Event Box Model Calculations

vs. Observations

Sep - Oct 1992

(q.5-12km)

.... Surface -Tropopause

AFRICA

140

• 0

Day 5

7•--• •

Natal 11-12km Ozonesonde Day4



•26 ppbv

D•y•D•v:D•y•/ -10 -



X.0.•/+•.• Day 7

/





Day8 •

lIt•\

120 -

N

[

-

---

X



s km - 12 km

12km-Tropopause

_

•'•'N.+35.31+35.4 X.x



-

_-

+29.81+31.6 Day 6

J

OVER BRAZIL



/ x•x

100



/ z/ : , •

ß

xx

tt ...." •,

i

•t

_



_

-20

-30 ?, -50

-•0

-30

-20

-10

0

,•

DegreesLongitude

•0

•0

'.

....

I .........

250

Figure 16. Cumulativenet ozoneproductioncomputedwith Goddard troposphericchemistrybox model downstreamof September26-27, 1992, Brazilian convectivesystemsis noted at 24-hourintervals(denotedby crosses)alonga characteristic forward trajectory.The photochemicaland trajectorymodels are not actuallycoupledin this simulation,but incorporationof trajectoryparametersis not expectedto give significantlydifferent results.First valuefor eachday is cumulativeproduction usinghighest9-min averagecloud-outflowmixingratiosat 11.3 km (see Table 5) as initial conditions;secondvalue is cumulative productionusing highest 9-min averagecloud-outflow data at 9.5 km (see Table 5). Asterisksshow locationsof measurements;incrementsin measured ozone mixing ratios obtainedby comparingDC-8 flights6 and 7 data and by comparing Natal ozonesondeprofilesfrom September28 and 30,

I,,

'".ff

,,,,,I

260

.........

I ....

270

•,,,,I,,,,,

280

....

290

I,,,,

300

Julian Day

Figure 17. Layer-meanO3 m•ing ratios from Natal ozonesondesoverperiodSeptember-October1992.Note highvalues in the 12 km to tropopauserangeon September30 (day 274) and October2 (day 276) followingthe major convectiveepisodeover the cerradoregion.

Simulationof the September26-27 convectivesystems with MM5 reproducesthe observedfeaturesof cloud outflow and showsthat at 6 hoursafter the decayof the two major convective systems, the perturbedCO extendedover a regionof 1500 km. Cloud-scaleandregional-scale simulationsexpandthe lim1992. ited samplingrange and showenhancedtrace gasconcentrations in all upper troposphericlayersextendingto the cloud tops at approximately16 km. CO mixing ratios by a factor of 3 over backgroundvalues The regionalextent of the convectivesystemsand trajectooccurred in widespreadconvectiveoutflow. The NOx mea- ries from sampledcloud outflow point to a significantcontrisuredin outflowwas a mixture of that transportedfrom the bution toward the SouthAtlantic ozone maximum.A photolower troposphereby the convectiveupdrafts and that pro- chemical box model, initialized with cloud outflow ducedby lightning.We estimatethat the lightningcontribution observationsand tracer output from the GCE model, shows amountedto at least 40% of the measuredNOx mixingratios severalparts per billion by volume 03 per day producedas at 9.5 km and 32% at 11.3 km in regionswith a detectable outflow is transporteddownstreamfor severaldays.Forward lightning signaturebasedon the NOx/CO ratio. trajectoriesfrom the cloud-perturbedregion show eastward

Table 7. Upper TroposphericOzone From Natal, Brazil, Ozonesondes Altitude Interval, km 11.25-11.5 11.5-11.75 11.75-12.0 12.0-12.25 12.25-12.5 12.5-12.75 12.75-13.0 13.0-13.25 13.25-13.5 13.5-13.75 13.75-14.0 14.0-14.25 14.25-14.5

September28 03, ppbv

September30 0 3, ppbv

September30-28 Delta 03, ppbv

91.37 82.49 87.64 91.77 101.90 99.07 100.96 92.28 86.50 88.03 96.50 93.27 85.92

96.70 114.91 128.69 132.04 119.83 109.32 115.10 123.17 126.48 131.87 130.73 134.61 135.52

5.33 32.42 41.05 40.27 17.93 10.25 14.14 30.89 36.98 43.84 34.23 41.34 49.60

14.5-14.75 14.75-15.0

Msg. Msg.

130.69 131.0

Msg. Msg.

15.0-15.25 15.25-15.5 Mean

96.97 97.24 92.79

127.77 134.19 124.86

30.80 36.95 31.07

PICKERING ET AL.: CONVECTIVE

10

TRANSPORT OVER BRAZIL

24,011

air: Cloud transport of reactive precursors,J. Geophys.Res., 89,

TrajectoryOutput, Julian Day 274-266 Theta (start P/end P)

(137/138) • 359K 349K (176/172) (117/119) • 370K 363K (129/130)

7111-7132, 1984.

Chatfield,R. B., and A. C. Delany, Convectionlinksbiomassburning to increasedtropical ozone:However,modelswill tend to overpredict 03, J. Geophys.Res., 95, 18,473-18,488, 1990. Dickerson,R. R., et al., Thunderstorms:An importantmechanismin the transportof air pollutants,Science,235, 460-465, 1987. Fishman,J., C. E. Watson,J. C. Larsen,and J. A. Logan, Distribution of troposphericozone determinedfrom satellite data, J. Geophys. Res., 95, 3599-3618, 1990.

Fishman,J., V. G. Brackett,E. V. Browell,and W. B. Grant, Tropospheric ozone derived from TOMS/SBUV measurementsduring TRACE A, J. Geophys.Res., this issue. Grell, G. A., J. Dudhia, and D. R. Stauffer, A descriptionof the fifth-generationPennState/NCARmesoscale model(MM5), NCAR

-2o

Tech.Note, NCAR/TN-398 + STR, Natl. Cent. for Atmos. Res., Boulder, Colo., 1994.

-30

-90

-80

-70

-60

-50

-40

-30

-20

Heikes, B. G., et al., Hydrogenperoxideand methylhydroperoxide distributionsrelated to ozone and odd hydrogenover the North Pacificin the fall of 1991,J. Geophys.Res., 101, 1891-1905, 1996. Figure 18. Backwardtrajectoriesarriving at Natal on Sep- Jacob,D. J., et al., Origin of ozone and NOx in the tropical tropotember30, 1992,on four potentialtemperaturesurfaces.Boxes sphere:A photochemicalanalysisof aircraft observationsover the SouthAtlantic basin,J. Geophys.Res., this issue. showlocationsof September26-27 convectiveregions. Justice,C. O., J. D. Kendall, P. R. Dowty, and R. J. Scholes,Satellite remote sensingof fires duringthe SAFARI campaignusingNOAA AVHRR data, J. Geophys.Res., this issue. flow from the 9.5 and 11.3 measurement altitudes across the Kirchhoff,V. W. J. H., and E. V. A. Marinho, Layer enhancements of SouthAtlantic and reachingAfrica in 8 days.Ozone measuretroposphericozone in regionsof biomassburning,Atmos. Environ., DegreesLongitude Backtraj., ECMWF data, starttime12Z

mentson flight 7 at 11.3 km (downstreamof the more southerly convectivesystem)were --•10ppbvhigherthan on flight 6, suggestingan 03 production rate similar to our box model calculationfor the first 24 hoursof transport.Three daysafter flight 6 the September30, 1992,ozonesondeat Natal showsan averageincreaseof ---30ppbv(---3DobsonUnits) in the upper tropospherecomparedto two days earlier. Back trajectories confirmthe sounding-convection link. Air arrivingat Natal on September30 in a 5-km-deep upper troposphericlayer was transportedfrom the convectiveregionover the cerradoin 2-4 days.Trajectoryanalysesby Thompsonet al. [this issue]show that upper troposphericozone in soundingsat Natal, Ascension Island, and African stations(south of 15øS)throughout TRACE A was transportedfrom South America. This paper complementsthat studywith direct evidenceof ozoneforming in the upper troposphereafter deep convectiondetrainsbiomassburning emissions.Kirchhoffet al. [this issue]showthat the 1992wet seasonhad an early onsetand middle and upper troposphericozonewas higher than usual.This suggests that the role of deep convectionin the South Atlantic regional ozonebudgetfor the TRACE A period may have been more prominentthan in a typicalyear.

28, 69-74, 1994.

Kirchhoff, V. W. J. H., J. R. Alves, F. R. de Silva, J. Fishman, Observationsof ozone concentrationsin the Brazilian cerradoduring the TRACE A field expedition,J. Geophys.Res.,this issue. Lacis,A. A., D. J. Wuebbles,and J. A. Logan, Radiative forcing of climateby changesin the verticaldistributionof ozone,J. Geophys. Res., 95, 9971-9981, 1990. Luke, W. T., R. R. Dickerson,W. F. Ryan, K. E. Pickering,and L. J. Nunnermacker,Troposphericchemistryover the lower Great Plains of the United States,2, Trace gasprofilesand distributions,J. Geophys.Res., 97, 20,649-20,670, 1992. O'Sullivan,D., M. Lee, K. B. Noone, and B. G. Heikes, Henry's Law solubilitiesfor hydrogenperoxide, methylhydroperoxide,hydroxymethylhydroperoxide, ethylhydroperoxide, and peroxyaceticacid in water, in press,Environ. Sci. Technol.,1995. Pickering,K. E., A.M. Thompson,R. R. Dickerson,W. T. Luke, D. P. McNamara, J. Greenberg, and P. R. Zimmerman, Model calculationsof troposphericozoneproductionpotentialfollowingobserved convectiveevents,J. Geophys.Res., 95, 14,049-14,062, 1990. Pickering,K. E., A.M. Thompson,J. R. Scala,W.-K. Tao, J. Simpson, and M. Garstang,Photochemicalozoneproductionin tropicalsquall line convectionduring NASA/GTE/ABLE 2A, J. Geophys.Res.,96, 3099-3114, 1991.

Pickering,K. E., A.M. Thompson,J. R. Scala, W.-K. Tao, and J. Simpson,Ozone productionpotential followingconvectiveredistribution of biomassburning emissions,J. Atmos. Chem., 14, 297-313, 1992a.

Acknowledgments. This work was supported under the NASA TroposphericChemistryProgram(to AMT), the NASA PhysicalClimate Program(to WKT), the NASA AtmosphericEffectsof Aviation Program(to KEP), and an EOS Interdisciplinary Study.We thankA. Motta of INPE and L. Cavalcante, F. Diniz, and O. Chiesa of the

Pickering,K. E., J. R. Scala, A.M. Thompson,W.-K. Tao, and J. Simpson,A regional estimateof convectivetransportof CO from biomassburning, Geophys.Res.Lett., 19, 289-292, 1992b. Pickering,K. E., A.M. Thompson,D. P. McNamara, M. R. Schoeberl, H. E. Fuelberg,R. O. Loring Jr., M. V. Watson,K. Fakhruzzaman, and A. S. Bachmeier, TRACE A trajectory intercomparison,1, Effectsof different input analyses,J. Geophys.Res., this issue.

Brazilian Departamento Nacional de Meteorologia in Brasilia for assistanceduring the field mission.Thanks to A. Setzer for providing Prins, E. M., and W. P. Menzel, Trends in South American biomass weeklyfire count data and to J. Kendall and C. Justicefor processing burningdetectedwith the GOES visibleinfrared spinscanradiomdaily fire counts.Thanks also to T. Kucserafor chemicalmodeling eter atmosphericsounderfrom 1983 to 1991,J. Geophys.Res., 99, assistance.

References Bachmeier,A. S., and H. E. Fuelberg,A meteorologicaloverviewof the TRACE A period, J. Geophys.Res., this issue. Browell, E. V., et al., Ozone and aerosol distributions and air mass

characteristicsover the South Atlantic basin during the burning season,J. Geophys.Res., this issue. Chatfield, R. B., and P. J. Crutzen, Sulfur dioxide in remote oceanic

16,719-16,735 1994.

Tao, W.-K., and J. Simpson,The Goddard cumulusensemblemodel, 1, Model description,Terr.Atmos. OceanicSci., 4, 35-72, 1993. Smyth,S., et al., Factorsinfluencingthe upper free troposphericdistribution of reactive nitrogen over the South Atlantic during the TRACE A experiment,J. Geophys.Res., this issue. Thompson,A.M., K. E. Pickering,D. P. McNamara, M. R. Schoeberl, R. D. Hudson, J. H. Kim, E. V. Browell, V. W. J. H. Kirchhoff, and

D. Nganga,Where did troposphericozoneoversouthernAfrica and the tropical South Atlantic come from in October 19927 Insights

24,012

PICKERING

ET AL.: CONVECTIVE

TRANSPORT

OVER BRAZIL

from TOMS, GTE/TRACE A and SAFARI 1992,J. Geophys.Res.,

B. G. Heikes, Graduate School of Oceanography,University of Rhode Island, Narragansett,RI 02882-1197. Wang, Y., W.-K. Tao, K. E. Pickering,A.M. Thompson,J. S. Kain, V. W. J. H. Kirchhoff,INPE, Sao JosedosCampos,SP, Brazil. D. P. McNamara, Applied Research Corporation, Landover, MD R. F. Adler, J. Simpson,P. R. Keehn, and G. S. Lai, Mesoscale model (MM5) simulationsof TRACE A and PRESTORM convec- 20785. tive systemsand associatedtracer transport,J. Geophys.Res., this K. E. Pickering(corresponding author), Joint Center for Earth Sysissue. tem Science,Department of Meteorology, University of Maryland, Watson, C. E., J. Fishman,and H. G. Reichle Jr., The significanceof Code 916, College Park, MD 20742. biomassburning as a sourceof carbonmonoxideand ozone in the W. K. Tao and A.M. Thompson,NASA Goddard Space Flight southernhemispheretropics:A satellite analysis,J. Geophys.Res., Center, Greenbelt, MD 20771. 95, 16,443-16,450, 1990. Y. Wang, Science,Systems,and Applications, Inc, Lanham, MD this issue.

20706.

D. R. Blake, University of California-Irvine, Irvine, CA 92717. J. D. Bradshaw, Georgia Institute of Technology, Atlanta, GA 30332.

G. L. Gregory and G. W. Sachse,NASA LangleyResearchCenter, Hampton, VA 23681.

(ReceivedApril 12, 1995;revisedDecember29, 1995; acceptedDecember29, 1995.)