JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO. D9, PAGES 11,813-11,832, MAY 16, 2000
Effect of biomassburning, convective venting, and transport Reunion
on tropospheric ozone over the Indian Ocean: Island
field
observations
TantelyRandriambelo,Jean-LucBaray,and SergeBaldy Laboratoirede Physiquede l'Atmosph•re,ReunionUniversity,France
Abstract. Relationshipsbetweenverticaldistributionof troposphericozoneat
ReunionIsland(21øS-55øE), satellite(NOAA advanced very highresolutionradiometer)observations of fires,smokeplumes,and convectiveeventsin southeastern
AfricaandMadagascar, andanalyses of meteorological situations(EuropeanCentre forMedium-Range WeatherForecasts) arepresented.Thisstudyis basedon 7 years (1992to 1999)of 2-monthlyPTU-O• radiosoundings at Reunion.Resultsshow,for the first time, that during1995tropospheric ozonecontentroseaboveaverageand that this year shouldbe set apart as atypical. Stratosphericcontributionsare also
ruledout usingan identification methodbasedon considerations of ozone,humidity, vertical stability, and meteorologicalconditions. The seasonalvariation of ozone
profilesduringtypicalyearsand without the stratospheric contributionsuggests that ozonecontamination from biomassburningis a maximumduringOctoberin the wholefree troposphere.During August,beforethe deepconvectionperiod, but alreadywithin the fire period,onlythe middletroposphere is contaminated. by ozoneinputs. By contrast,throughNovemberto December,well within the deep convectionperiod, all the higher troposphereis contaminated.The comprehensive study of the observations in 1993, taken as a typical year, highlightsthe roles of convection and transportin contamination of remoteoceanicregions.August contaminationof the middle troposphereby about 70 ppbv of ozoneis contrasted
to Octoberenhancement of the wholefreetroposphere by about 100 ppbv of ozone after the spreading of deep convectiveevents. Fire satellite data further indicate
that columnintegratedcontamination levelmainlydependson biomassburning intensity. Through August to October the fourfold increaseof ozone concentration is comparablewith the fivefoldaugmentationof fires. The redistributionof ozone
with altitudedependson the convection intensitynearsourceregionsin accordance with convectiondetectionand backtrajectoryanalysis. 1.
transportcouldplay a major role in ozonegeneration
Introduction
Vegetationburningis considered to be responsible for extensivetropospheric ozoneenhancementin the trop-
ics duringthe dry season.Levine[1996a,1996b]and the "Southern Tropical Atlantic Region Experiment
and redistribution, in situ measurementsof ozone ver-
tical distribution are also'important.Thenthe observations that could play a major role in the pollutants redistribution on regional scale in remote oceanic re-
gion in the tropicsare of specialinterest[Crutzenet (STARE):TRACE-A andSAFARI"specialissue(Jour- al., 1979; Chatfield and Delany, 1990; Krishnamufti nal of Geophysical Research,101 (D19), 23,519-24,330, et al., 1996; Pickering et al., 1996b; Thompson et al., 1996)suggest multiplesources of ozoneprecursors (in- 1996; Jenkinset al., 1997]. Indeed, the deep convec-
dustrial, biogenic,and biomassburning). These con- tion zones transport air massesenriched ozone rapidly taminations are monitored by satellite measurements from the planetary boundarylayer (PBL) to the midobtainedfromTotal OzoneMappingSpectrometer [Fish- dle and upper troposphere, where lifetimes of ozone man et al., 1990] and Global OzoneMonitoringEx- and its precursors are extended. However, the source periment[$iddanset al., 1998]. Sinceconvection and of enhancedozone can be of anthropogenicand stratosphericorigin. At midlatitude, stratosphericintrusions into the troposphere induced by tropopause fold are believed to be an important term in the ozone budCopyright2000 by the AmericanGeophysical Union. Paper number 1999JD901097.
0148-0227 / 00/ 1999JD901097509.00
get. This stratosphere-troposphere exchange(STE) is a well-knownphenomenon in temperateregions[Reed, 1955; Danielsen, 1968; Reiter, 1975; Ancellet et al.,
11,813
11,814
RANDRIAMBELO
ET AL.: SEASONAL VARIABILITY
OF TROPICAL
TROPOSPHERIC
O3
1991;Holtonet al., 1995].Until recently, possible impli- quently,mostof the Africanand Madagascaremissions cationof the subtropicaljet streamon STE wasrarely approached because of the limitednumberof available data, and it wassupposed that the role of the subtropical jet streamwas limited. In the tropicsthis con-
are transportedtowardthe westat low altitudes.Middle and upper troposphericozonecontaminationsabove
Whyteand Tyson,1988],and many authorshavesug-
the trade wind inversion,from African and Madagascar fires, are more likely toward the east. Conversely,during the rainy seasonthe stable layers could be broken by deep convection. Ozone and its precursorscan be vented from sourceregionsin southeasternAfrica and Madagascarto the free troposphereby convectionand then could be transported through upper level westerly winds to the Indian Ocean. Since ozonegenera-
gestedthe occurrenceof stratosphericintrusionsfrom
tion and destruction are more dependent on chemical
tribution is not usually estimated in particular situa-
tions[Thompson et al., 1996]or in the largerworldview [Roelofset al., 1997]. However,meteorological disturbanceslike westerlywaves,cut-off low and frontal zones in the southernAfrican continentcan occur [Preston-
ozonemeasurements [Diab et al., 1992;Bachmeierand and radiative conditions,specificstudiesfocusedon the Fuelberg,1996;Loring et al., 1996]. Moreover,during Indian Ocean shouldfurther contributeto clarify inthe last 5 years, tropopausefolds have been observed terconnectionsbetweenbiomassburning, vertical transin subtropics[Gougeret al., 1996;Folkinsand Appen- port, and ozonecontamination. Comparative studieson. zeller,1996;Baray et al., 1998].In 1998,usinga tropo- ozone vertical profile indicate that troposphericozone sphericozonelidarinstalledat ReunionIsland[Barayet concentrationsrecordedat Reunion Island are higher ak, 1999a],daily ozonemeasurements duringJuly show than thoserecordedcloseto sourceregionsin the tropstratosphericozoneintrusionsinto the free troposphere ics (Irene,Brazzaville,Okaukuejo, Natal) [Diab et al., inducedby a quasi-permanentlarge-scalefold structure 1996]. Theseobservations call attentionto transport could be fed by the convergentdescendingbranch of processes, which affect regionsfar from sources. the Hadley cell, and by the upper level monsoonreBaldy et al. [1.996]from 1 year (September1992to turn flow [Barayet al., 2000]. Thesenew observations August1993) data on ozone,temperature,and humidsuggestan important role for STE near the edgeof the ity profilesat Reunion Island, show that high values tropical zones,which should be consideredin the tropo- of ozone concentrationare observedin the free troposphericozonebalance. In addition, when anthropogenic sphere,above the inversionlayer that capsthe marine sourceand stratosphericintrusion work together, large boundary layer,duringSeptember to December. Using enhancement of troposphericozonecouldoccur[Ran- backwardtrajectoriesand the regionaldistributionof
driambeloetak, 1999].
firesin southern Africa(1992)[Justice et al.,1996],this
To better assessthe relationships between fires, convection, transport, and tropospheric ozone enhance-
burning in southeasternAfrica.
ment, SouthernTropicalAtlantic RegionalExperiment, Transportand AtmosphericChemistrynearthe EquatorAtlantic and Southern African Fire-AtmosphereResearchInitiative campaigns havebeenundertaken[Andreaeet al., 1996].Althoughfire contaminations above the Atlantic Ocean have been quite well documented, still few studiesconsiderthe Indian Ocean. Yet, circulationsfrom nearbycontinentsto southernAtlantic and
enhancement of ozone values was attributed to biomass
The presentstudy is basedon a largerdata set from
7 years(1992-1999)of 2-monthlyPTU-Os (pressure, temperature,humidity,and ozone)radiosondes and a more detailedclimatologicalinvestigation. Through complementarymeasurementof fires, smokeplumes,
and of convection zonesfrom satellitedata (NOAA)
in southeasternAfrica and Madagascar,the consideration of meteorologicalsituations(disturbancesfrom Indianoceans arequitedifferent[Chatfieldet al., 1996]. tropical and middle latitude regions)and trajectories Many workershave demonstratedthe predominanceof computedfrom EuropeanCentre for Medium-Range the stable layers during the dry seasonin the African WeatherForecasts(ECMWF) data, this investigation continent[Garstanget al., 1996; Swapet al., 1996a; is focusedon the analysesof the relationshipsbetween Tysonet al., 1996]. Thesestablelayersare responsible fires, deepconvection, and the seasonalvariabilityof for the pollutant distributionsfrom the sourceregion. ozone. Troposphericozoneclimatologyat ReunionIsThen during the dry seasonthe stable layers inhib- land, consideredas a typical site in the Indian Ocean, ited the transport of pollutants to the free troposphere is also further documented. where the westerly winds are predominant. There is no Data and methodologyare presentedin section 2. study about stablelayersin Madagascar,but the mete- The generalozonecontextandthe methodology of idenorologicalanalysesshowthat the trade wind inversion tificationof stratospheric influenceare displayedin seclayers vary as function of location. For example, in tion 3. From considerations of the seasonalvariability the southern African continent and the Atlantic Ocean of tropospheric ozone,a typicalyear (1993)is selected analyzedthroughcasestudy the altitude rangeis 500-2500m [Preston-Whyte and andmorecomprehensively Tyson,1988;Garstanget al., 1996]and2000-4000m for investigationsin section4. Further analysesand conMadagascar [Donque,1975;Raholijao,1987]. Conse- cludingcommentsare givenin section5.
lr•ANDRIAMBELO
ET AL.' SEASONAL VARIABILITY
2. Data and Methodology 2.1.
OF TROPICAL
TROPOSPHERIC
03
11,815
the 5 øN to 40 øS and 20 øE to 90 øE zone, and have been archivedsince 1989. Malawi, Zambia, Zimbabwe, SouthAfrica, Tanzania, Mozambique,and Madagascar
Radiosondes
VerticalozoneprofileshavebeenobtainedsinceSeptember 1992 at Reunion Island from radiosondes us-
are included
in this zone.
The methods
of detection
of fires and smoke plumes in southeasternAfrica and
ing electrochemical cell (ECC) ozonesonde with a 2- Madagascar are detailedby Randriambelo et al. [1998]. monthlylaunching frequency near0800UT. Sevenyears The analysisof firesdetectedfrom 1992to 1998displays (September 1992- February1999)oftropospheric mea- a seasonalvariability for eachregion. The m•imum of surements,starting on September1, 1992,are presented fire intensitycanbe split into two peaks:onein Septemhere. The ozonesondeis coupledto a standard mete- ber correspondingto the southeasternAfrica contribuorologicalradiosondeto simultaneously transmit, with tion and the other in October-November for Madagasozonedata, atmospherictemperature,pressure,and hu- carfires[Randriambelo et al., 1998].Theseperiodsfall midity from the groundto the burst altitude of the bal- in the secondhalf of the dry seasonfor each contributloon. The characteristics of the radiosonde are detailed
ing region. These results are in agreementwith those
elsewhere [Baldyet al., 1996]. Ozoneannualdistribu- obtained by other authors for the periods 1992-1994 tion obtainedfrom 149 ozonesoundingseriesshowa [Arino and Melinotte,1995]and 1992-1993[Brivio and strongseasonalenhancementduring the dry seasonin Grdgoire,1997]and supplementthe analysesof Justice southeasternAfrica and Madagascar. This enhance- et al. [1996]for May to October(1989and 1992),which ment essentiallyobservedin the free troposphereabove
found that fires are very weak in October for southeastern Africa. The contribution of fires from Madagascar al., 1999]is in agreement with Thompson et al. [1996] during Octoberto Decembershouldnot be ignored. results.Sincewindandtemperatureinversions capping During clear-skyconditionsallowingsmokedetection, the marine boundarylayer limit ozoneinputs, we will fire peaks are associatedwith well developedsmoke focuson ozonevariationsin the free troposphere. plumes during September for the southeasternAfrica and during October-November for Madagascar(Figure
the marineboundarylayer[Baldyet al., 1996;Taupinet
2.2.
Satellite
Measurements
1) [Randriambelo et al., 1998].This furtherdeterminaInformationaboutfires,smokeplumes,and deepcon- tion of smokeplumes gives additional information on
vective clouds is obtained from National
Oceanic and
AtmosphericAdministration-advanced very high resolution radiometer-local areacoverage(NOAA-AVHRRLAC) 1-km resolutiondata usinga specificantennaof the Surveyof Environment Assisted by Satellite(SEASORSTOM) centerat ReunionIsland. Thesedata cover
the dispersionof biomassburning by-products. It is to be noted that savannafires are observedat the beginning of fire period, that is, in June for the studied region. Conversely,forest fires are observedmore toward the end of the dry season,that is, Septemberfor southeasternAfrica and October-Novemberfor Madagascar.
Figure 1. Grey levelimagesshowingfires(blackspots)and smokeplumes(in white) on the east coastof Mozambique,(left) September18, 1993,and (right) October12, 1993.
11,816
RAND•AMBELO
ET AL.' SEASONAL VARIABILITY
Forest fires are more likely to generate intense smoke plumesthan savannafires. Hence during October, that is, after the September fire peak for the southeastern Africa region, developmentof intensesmokeplumesis
still observed(Figure 1). N OAA-AVHRR imagesare alsousedto detect deep
OF TROPICAL TROPOSPHERIC Oa
120 100
80 60
convectionzones. These zones are characterizedby a
high reflectancevalueof channelI (> 70%) and a low brightnesstemperaturevalue of channel4 (< 190 K).
40-
Deep convectivecloudsare intenseduring the rainy season, which beginsat the end of the year in southeastern Africa and Madagascar.Yet deepconvectivecloudsare appearingat the end of the dry season,that is, October
20-
1992
1993 1994 1995 1996 1997 1998 1999
in the studiedregion[Randriambelo et al., 1998].Con- Figure 2. Eight years(1992-1999)time evolutionof versely,during the dry season,we do not detect these clouds. During this period, only middle and lower level
integrated total troposphericozone.
clouds are observed.
The annual variability of integratedtroposphericozone (Dobsonunit, DU) indicatesthat the ozonerecorded The meteorologicalconditions and backtrajectories in the year 1995 representsthe highest valuesfor the presentedin this paperare obtainedfrom ECMWF (1 ø period 1992-1999(Figure 2). For the rest of the year grid, 15 levels of pressurebetween 1000 and 10 hPa, the valuesrecordedof each individual profile are in ac6 hours time resolution)data sets. Trajectoriesused cordancewith Thompsonet al. [1997],whichfound59 in this study are provided from two programs: one DU for polluted regionsduring biomassburning con2.3.
Global
Model
from MeteoFrance and another developed at Reunion.
The MeteoFrancetrajectorieswerealreadyvalidatedby
Baldy et al. [1996]with other computationlike isentropic trajectoriesfrom Goddard SpaceFlight Center and trajectories from the Laboratoire de Meteorologie Physique of Clermont-Ferrand, France. Backtrajectories calculated by the program developedat Reunion are donedirectly from three-dimensional ECMWF wind fields. At each point of the trajectory, value of wind components (horizontaland vertical)is interpolatedin time and space. Trajectory is then determinedby iterative advection of the air mass with a 5-min time step. Many workershave been made comparisonbetweendif-
tamination in tropics. One must keep in mind that this value has been obtained from each individual profile, and not from annual means like in Table 1. The evolu-
tion of interannual means of column-integratedtropo-
sphericozone(Table 1) further corroborates the atypical characterof 1995. The year 1995 is characterizedby botha 49-DU mean(abovevaluesobtainedduringother years,around30-35 DU), and a 22-DU standarddeviation well abovethe 6 to 7-DU typical value. Since1992
data (September to December)and 1999data (January to February)do not includeprofilesduringthe whole year, the mean valuescorresponding to theseyearsare biasedand should not be considered.High mean ozone
ferenttrajectorycomputations[Kuo et al., 1985;Fuel- valueobtainedfor 1992 is likely to be explainedby high berget al., 1996].Kuo et al. [1985]showed that thecom- biomassburning contaminationsoccurringduring the parisonbetweenisentropicbacktrajectoriesand trajec- measurementperiod of Septemberto December. Contories simulated by a mesoscalemodel displayeda good versely,the low mean of ozonevalue corresponding to agreementfor 2-days backtrajectories. Fuelberget al.
1999 can be explainedby a data record limited to the
[1996]and D'Abreton[1996]havefoundthat the kine- first 2 monthsof the year that correspondto the rainy
matic trajectories undergoconsiderablygreater vertical displacementsthan their isentropic counterparts. The comparisonbetween our trajectory computation with
season.
Then it is of special interest to study ozone concentration variation as a function of altitude. Figure 3 the otherof Baldyet al. [1996]showsa goodagreement presentscolumn-integratedozonewithin differentlay-
(data not shown).
3. General Context and Methodology of Analysis 3.1. the
Ozone Content Study for
in the whole free troposphere. Two examples of extreme contamination of troposphericozoneduring 1995 have been studied; the first case was producedfrom stratosphere-tropospherelinked to a tropical cyclone
the sources of ozone enhancement
[Barayet al., 1999b],andthe secondcaseinvolvedsev-
Tropospheric Period
To better
1992-1999 evaluate
ers. Annual variations in all layers (2.5-5, 5-9, 912.5, and 12.5-16.5 km) corroboratethat 1995 is an exceptionalyear characterizedby high ozone contents
in the tropics, integrated tropospheric ozone concen- eral mechanismssuch as anthropogenicand dynamiworkingtogether[Randriambelo et al., trations observed over Reunion Island are analyzed. caldisturbances
tL4NDRIAMBELO
ET AL.' SEASONAL VARIABILITY
OF TROPICAL
TROPOSPHERIC
O3
11,817
Table 1. Annual Integrated TroposphericMean Ozone spheric ozone, it is necessaryto identify and to sepaand Standard Deviation for 1992-1999 rate profiles contaminated by stratosphericintrusions. Stratosphericcontributionson this study are justifiable Year Ozone, DU Standard Deviation by meteorologicalcondition agreedwith many workers 1992 1993 1994 1995 1996 1997 1998 1999
42.7 31.6 32.1 49.8 33.1 34.4 35.5 28.3
6.3 7.6 7.8 22.6 6.1 7.6 7.5 4.3
Interpolation has beenapplied in order to correctthe temporal inhomogeneityof measurements
in the southernAfrican continent[Preston-Whyte and Tyson,1988; Diab et al., 1992; Bachmeier and Fuelberg,
1996;Loringet al., 1996]. Indeed,in the studiedregion, different stratosphere-troposphereexchangescould take place during a large part of the year and could greatly complicatethe study of biomassburning influenceon troposphericozone. Due to the positionof the Reunion Island near the tropic edge, it is likely to observeseveral types of stratosphere-troposphere exchanges:tropopausefoldsinducedby subtropicaljet stream, cut-off
low, and tropical cyclonedetrainments[Baray et al., 1998,1999b,2000].
as normal yearsexhibiting near-averageozoneconcentrations in accordancewith typical values observedin the tropics(Figure 2). In the tropicsthesevaluesindicate that column-integrated ozone is about 50 DU for the polluted region and about 30 D U for pristine air
Although anthropogeniccontributionsare seasonal and are likely to overridethe ozonedistribution during the fire period, occasionalstratospheric inputs occurring in the higherpart of the troposphereshouldnot be neglected. These occurrencescould confusethe comprehensionof the mechanismof fire contamination to
[Thompson et al., 1996;Diab et al., 1996].
remote oceanic regions.
1999]. Conversely, the other yearscan be considered
The study of typical years is based on the consideration of mean profilesof ozoneobtainedby averaging 1992 to 1998 profiles and focusingon the fire period of potential sourceregions. To further scrutinize the analysis of anthropogeniccontributions to tropo-
In order to undertake climatologicalstudies of tro-
popausefoldsat midlatitude, Van Hayer et al. [1996] (hereinafterlabeledVH) established automaticcriteria to detect midlatitude tropopause folds, based on ra-
diosoundings (ozone,humidity,and temperature),and
40
35 3O
2.5-5
km
5-9km
35 30
25÷
•
2O
25
g 20
15÷
•
15
¸ 10
10
5
1992'1993'1994'1995'1996'1997'1998 '1999
0
1992'1993' 1994'1995'1996'1997'1998 • 1999
25.
9-12.5
km
20
20] 12.5-16.5 km 15
15
•0
!0.
1992'1993 ' 1994"1995'1996'1997'1998'1999
1992
1993
1994
1995
1996
1997
1998
1999
Figure 3. Time evolutionof integratedtroposphericozonefor differentlayers' 2.5-5 km, 5-9 km in the middle troposphere,9-12.5 km in the high troposphere,and 12.5-16.5 km near the tropopausefrom 1992 to 1999.
11,818
tLhNDRIAMBELO ET AL.: SEASONAL VARIABILITY OF TROPICAL TROPOSPHERIC Oa
onmodeldata (potentialvorticity(PV) andwindfields). beganto be dissipated. Thus the detectionlevel is lowSince stratospheric contamination mechanismsin the eredto 20%both for detectionagainsta climatological tropics could differ from tropopause folds observedin midlatitude, midlatitude criteria of automatic detection of stratosphericinputs usedcurrentlyshouldbe adapted to the presentstudy. Hence,in the followingtwo subsections, VH criteria are presented and then are modified to fit the tropical region.
profileand againstvaluestaken I km aboveor belowthe peak height. The climatologicalprofilesare thoseused
by Baray et al. [1999b].Hencewe accountbetter for contaminationaffectinga large part of the free troposphereand are not restrictedto a thin peak. Moreover, the ozonecriterionis appliedto the wholetropicalfree
troposphereand under the thermal tropopause. $.3.2. Relative humidity criterion. The huVH Criteria midity VH criterion is adjusted to midlatitude studVH identifiesa midlatitude tropopausefold as a well- ies. In the tropics the water vapor vertical distribution defined tongue of ozone-richstratosphericair, situated differsconsiderablyfrom the midlatitude distribution. around400 hPa, linked to a very dry air layer and char- High-humidity valuesobservedduring summerin the acterized by a distinct temperature inversion,below a wholefree tropospherecontrastwith the low-humidity region of strong vertical wind shear. In order to sys- contentabovethe trade wind inversionduring winter. tematically identify tropopausefolds, VH have incor- Hence without any stratosphericinput a large number porated four main criteria into an automatic detection of relative humidity profilesare under 25% in the free troposphere. Since subtropicaltropopausefolds occur algorithm: 1. An ozone peak must be detected in the free tropo- essentiallyin winter, a 20 or 25% humidity threshold, sphere, betweenthe 3 km height above sea level and 1 givenby VH for midlatitude, is not sufficientto detect km below the thermal tropopause. The maximum ozone tropopausefolds near the tropic edge. Therefore the by a mixing ratio must be at least 25% largerthan a climato- present25% humidity thresholdis complemented logicalprofile,or 30% largerthan the mixingratio of the correspondencecheck between ozone peak and humidsame sounding taken I km above and below the peak ity minimum, that is, layers located above and under height, in order to also detect ozonepeakswhich could the ozonepeak layer have to be more humid. 3.3.3. Vertical stability criterion. The VH staoccur on a very low troposphericozonebackground. 2. The relative humidity at the height of the ozone bility thresholdis againsuitablefor the midlatitudetroposphereand has to be modified for the tropical tropeak has to be lower than 25%. 3. The verticalstability(-dO/dP) calculatedovera posphere. Convection often occurs under conditional vertical range of 500 m in the vicinity of the ozonepeak instability situationsin the tropicsand quite high gradientsof potential temperature could be observedin the hasto exceed11.5 K/100 hPa. 4. The wind speed between the ozone peak and the tropical troposphere,especiallyin its upper part. Hence tropopausemust reachat least 20 m/s. the 20 K/100 hPa thresholdis then better adaptedto However, these criteria are only designedto detect the tropical free troposphere conditions below 12 km. tropopause folds. The aim of the present investiga- Between12 and 17 km (the tropopauselevel), values tion is actually slightly different: we need to extract up to 30 K/100 hPa are currentlyobservedwithout all stratosphericinfluencein a region which is dynami- any other stratosphericsignaturessuch as ozone, hucally dissimilar,that is, influencedby cut-off low, sub- midity, P V, etc. Therefore we will confinethe use of tropical tropopausefolds, and stratosphere-troposphere the vertical stability threshold below 12 km. For the exchangeslinked to the Intertropical ConvergenceZone uppertroposphere(over 12 km), we will only consider (ITCZ) or to the tropical cyclones.Hencethesecriteria the vertical temperature gradient. The normal tropo3.2.
Identification
of Stratospheric
Intrusions:
sphericgradientin the tropicsis around-6 K/km. A
have to be modified.
well-pronouncedinversionof the temperature gradient 3.3. Identification of Stratospheric Intrusions: Modification of VH Criteria for Tropical Region 3.3.1.
Ozone
criterion.
The aim of VH
was to
(lessthan 4 K/km in the vicinityof the ozonepeak)is taken as a sign of stratosphericinput. 3.3.4. Wind criterion. Since the building of the present identification schemeis designedto discard stratosphericinputs from any mechanism,the VH wind criterion is specificto midlatitude tropopausefolds,and is over-selective.Hence, the wind criterion is not con-
detect tropopausefold; in the present study our aim is slightly different. We want to study the impact of biomassburning, and consequently,we need to extract all stratospheric influences(cut-offlow, subtropicaltro- sidered. popausefold, stratosphere-troposphere exchangelinked to the ITCZ or to the tropical cyclones).Consequently, 3.4. Identification of Stratospheric Intrusions: Results our criterion must be less selective than that of VH, in order to detect other ozone enhancement mechaFigure4 presentsa typical casestudyof stratospheric nisms and ozone inputs from subtropical tropopause input from a tropopausefold observedabove Reunion fold which occurredjust beforethe radiosoundingand Island characterizedby an ozoneenhancement,a clear
RANDRIAMBELO ET AL.: SEASONAL VARIABILITY
OF TROPICAL TROPOSPHERIC Oa
lOO
200
300 400
5OO
600 700
800 900
1000 -100
-50
0
50
0
20
Temperature (øC)
40
60
80
100
0
Humidity (%)
50
100
150
200
Ozone(ppbv)
Figure 4. Vertical profilesof temperature,humidity,and ozonecorresponding to October 1, 1996, radiosoundingat ReunionIsland and illustrating stratosphericair input between10 and 12 km.
18
16
14
12
•1o
8
August
September October November i i iii
0
20
40
60
i i i
I
i
80
100
December
120
140
Ozone (ppbv)
Figure 5. Mean monthly troposphericozoneprofilesobtainedby radiosondesat Reunionfrom August to December.
11,819
11,820
RANDRIAMBELO ET AL.: SEASONAL VARIABILITY OF TROPICAL TROPOSPHERIC
temperatureinversion,a relativeminimumof humidity, temberthan in October.In addition,it shouldbe furandassociated with largevaluesof the verticalstability ther noted that deep convectivecloudsin southeastern (morethan 50 K/100 hPa). The layersjust belowand Africa are lessdevelopedin Septemberthan in Octoabovethe tonguedonot displaystratospheric signature. ber. Most of the studieswerefocusedon the transport The test run of the presentstratosphericidentification in the Atlantic Ocean. Generally,at low levels,withand discardingschemeresults in the elimination of the out disturbance,the pollutants are transportedto the 9 to 12-km layer of this casestudy. The application west, and at high levelsto the east. The altitude of the of these criteria to the Reunion Island data results in the identification
of an ozone enhancement that could
possiblybe attributed to a stratosphereinput in 41% of the 50 selectedprofiles. On these 50 profiles, 42 respond to the humidity criterion and 32 to the stability criterion. Before attributing a stratosphericorigin to a layer, we observeall the meteorologicalsituations of the casesrespondingto the ozone criterion, and at
least one other criterion (O3-•humidity,O3-•stability, or O3-•humidity-•stability).When elementsof the meteorologicalsituation confirm the stratosphericorigin (proximity of the subtropicaljet stream, upper level frontal system, intense convectivezone, tropical cyclone,or cut-off low), we discardthe layer whichis affected by the stratospherein the computation of the climatologicalprofilesused for the study of the impact of biomassburning products.
wind inversionlayer dependson the latitude and on the meteorologicalsituation and can reach 500 hPa in some
cases[Jenkinset al., 1997].Indeed,duringthe dry season, anticyclonicmeteorologicalconditionslead to the recirculation
of air masses over the African continent
[Garstanget al., 1996; Thompsonet al., 1996; Tyson et al., 1996],then pollutantsare trapped underneath the subsidence inversion over southern Africa.
Once
the pollutantsleavesouthernAfrica to the east, they tend to ride above the trade wind inversion and to be
cappedby stable layersabove 500 hPa. During recirculation, air massescan reach the westerlies and then
are advectedto the Indian Ocean. A large part of the
flowfromthe Africancontinent,evaluatedontrajectory statistics, could be advected toward the Indian Ocean
[Garstanget al., 1996;Tysonet al., 1996;Pikethet al., 1999].As notedby Stockset al. [1996],in the savanna in southern Africa the convection column does not ex-
3.5.
Evolution
of the Vertical
Distribution
of
Tropospheric Ozone During the Biomass
ceed 3000-4000 m. Then few pollutants can reach the wcsterlies and reach the Indian Ocean. In the west coast
of Madagascar,savannasare dominated,then the pollutantsproducedby biomassburningin the dry season By usingthesecriteria,stratospheric intrusionlay- cannot reachthe freetroposphere.However,Swapet al. ersare removedfrom meanmonthlyprofilesconsidered [1996a]pointoutthat the eastward transports shouldbe in this study. Mean monthlyprofilesresultingfrom furtherelucidated.Yet numerousobservations suggest the processingof the Reunion radiosondedata set durthat under dry conditions,ozonemaximaare likely to Burning Period
ing typicalyears,that is, omitting1995,are givenin extend over the southern part of the African continent Figure5. To furtherstudybiomass burninginfluence, and over the Indian Ocean,by synoptic-to-planetaryozoneprofilesfrom Augustare presented.Indeed,in- scalecirculationfields[Fishmanet al., 1996;Kim et al.,
dividual ozone profiles show that the enhancedozone
1996; Swapet al., 1996a; Thompsonet al., 1996;Hud-
justabovethetradewindinversion appearfromAugust sonand Thompson,1998]. Conversely, in the presence whereozoneconcentrations reachmorethan 50 ppbv. of deepconvective clouds,all theseauthorsagreethat The standarddeviationduringthis periodisabout20%.
rapid transports occurred and penetrated these stable
Beforethisperiod,ozoneprofiles aremostlyinfluenced layers,and pollutants can reachthe westerliesin the free
by stratosphere-troposphere exchange.
troposphere [Preston-Whyte and Tyson,1988;Garstang et al., 1996;Swapet al., 1996a;BachmeierandFuelberg, tweenthe tradewindinversion and10 km) is contami- 1996; Tysonet a/.,1996]. nated by enhancedozone. The ozoneconcentrationsinThe meteorologicalconditionsof the sourceregion
DuringAugust,onlythe middletroposphere (be-
crease in September andthec,•'c:x•amination is reaching are illustrated by the low ozoneconcentrationobserved upto 12km. Ozoneconcentration isstillenhancing and at Reunion during the dry seasonin the upper tropospreading upward,reaching16 km, duringOctober.A sphere. When the rainy seasonbeginsduring October
decreaseof ozone concentration is observedin November, but the upper altitude of contaminationremains
in sourceregions,the upper troposphericozoneconcentration increaseswith the deep convectionoccurrence. very high (16 km). In Decemberthe ozoneconcentra- The strong ozone concentrationsobservedat Reunion
tion is furtherdecreasing; yet, the maximumaltitudeof contamination remainsin the vicinityof 16 km. These monthly profilesindicate that when fires are
peakingin southeastern AfricaduringSeptember, the ozone recorded at the remote Reunion site is not a maximum but is less than in October. It should be noted
that in Madagascar, fewerfiresare detectedin Sep-
during Octobercan be explainedby the presenceof intense fires and of deep convectiveclouds both in the
southeastern Africancontinentand Madagascar.Moreover,most of the firesobservedduringthis periodare likely to be forestfiresthat generatehigh concentrationsof by-productsfrom biomassburning. The presenceof deepconvection makespossible the ventingof
RANDRIAMBELO
ET AL.' SEASONAL VARIABILITY
poilutams into the higher troposphere through which they can be advected by westerly winds to the Indian Ocean, under chemical and radiative conditions favorable to ozonegeneration,namely,isolationfrom humidity and boundary layer compoundsby the trade wind inversion. The development of deep convective clouds and vertical transport observed close to the source regionsis likely to explain the rising of ozonecontamination to higher altitudes at the remote Reunion site. The decreaseof ozone concentrationduring November and December is expected to follow from the decrease of fires in the southeastern African continent that dominates the relative
increase of forest fires in Mada-
gascar. Still, the high altitude reachedby ozonecontamination can be explained by the presenceof forest fires at Madagascar associatedwith abundant smoke plumes, comparable to those observed in southeastern Africa during October (Figure 1) and deep convective clouds with brightness temperature less than
OF TROPICAL
TROPOSPHERIC
Oa
11,821
forestfire activity in Madagascar,throughdeep convectioninjection and possiblymix-then-cookamplifica-
tion [ChatfieldandDelany,1990],is responsible for the ozone contamination
still observed at Reunion
Island
up to high levelsin the troposphere.Indeed, during November and December the trade winds decrease due
to the southwarddisplacementof the ITCZ, and of the jet stream toward the high latitude, and also to the slowingdown of the southernHadley cell. Then the transportsbetweenReunionand Madagascarare likely
to take morethan I day [Baldyet al., 1996; Taupinet al., 1999;Randriambelo et al., 1998]. In addition,the recirculationand the accumulationof pollutantsin the boundarylayer are possible.The forest regionsin the east coast of Madagascarare situated near mountains.
Then in the end of the dry seasonthe meteorological conditionsare favorable to the deep convectiveclouds
formation(brigthnesstemperaturelessthan 190 K). During October to December, strong developmentof
190 K [Goldammeret al., 1996; Randriambelo, 1998]. smokeplumes associatedto deep convectivecloudsare We further note that the ozone concentrationsduring November and December are still high. Interestingly, severalstudieshave shownthat during this period, fires in southeasternAfrica are very rare and that fires are
still observedfrom SpaceShuttle SpaceTransportation
System(STS 61) [Goldammeret al., 1996]and from satellitedata (NOAA) [Randriambelo et al., 1998]at Madagascar. This is alsothe same caseof the east coast
peakingin the northernpart of Africa [Jenkinset al., of Mozambique(Figure 1) wherea strongdevelopment 1997].The presentfire detectiondata suggestthat for of forest smokeplumes associatedwith deep convective this period the primary sourceof biomassburning is cloudsare observedduring Septemberto October. As Madagascar which is located about 1000 km to the suggested by Stockset al. [1996],deepconvective clouds west of Reunion Island. It is probable that the residual could inject pollutants from the PBL to the stratosphere
18
16
14
12
•1o
8
Mean August
4 -
August 24th, 1993 Mean October
October 12th, I993
0
20
40
60
80
1 O0
120
140
160
180
200
Ozone (ppbv)
Figure õ. Comparison of meanmonthlytropospheric ozoneprofilesand radiosondes obtained on August24, 1993,and on October12, 1993.
11,822
RAND•AMBELO ET AL' SEASONALVARIABILITY OF TROPICAL TROPOSPHERICO3 lOO
200
300 1
400
!
5OO
600 710 8OO
9OO 1000 -100
i
-50
0
Temperature (øC)
50
0
!
20
40
60
80
100
0
50
Humidity(%)
100
150
200
Ozone (ppbv)
Figure 7. Tropospheric temperature,humidity,and ozoneprofilesobtainedby radiosondes at Reunion,February23, 1993 (thick line), August24, 1993 (thin line), and October12, 1993 (dottedline).
in temperateforestregions.Hencea largeamountof lowerthan 15 ppbv in lower troposphereand around ozonecouldbe released in the freetroposphere during 50 ppbv in the uppertroposphere.The humidityprothese periods. file displaysquite elevatedvaluesup to the high troTo further analyzethe mechanism leadingto this posphere. The weather conditions show disturbances ozonecontamination,a more comprehensive study of in the vicinity of Madagascarand of Reunion Island. a typical year is necessary.From annual variation of Thetotal rainrecorded onthisdaywashigh(97.5ram) ozonecontent(Figure2) it couldbe observed that the and is consistentwith the presenceof a strongdistur-
year 1993is about average.Furtherjustificationof this banceabovethe Indian Ocean. This ozoneprofile is selectionis givenin section4 by a comparisonof 1993 representativeof the wet seasonin southeasternAfrica profiles againstmeanmonthlyprofiles (Figure6). and Madagascar.Sincecontaminations by vegetation firesare not expectedin this season,this profilecan be consideredas a minimum backgroundprofile. 4. Analysis of a Typical Year' 1993
Representative profilesfor the threeseasons during the yearare selected to analyze1993asa typicalyear. Thesearesummer(February 23), winter(August24), andspring(October12) as shownin Figure7. Before analyzingthesecasestudies,wepresentthe generalcontext of transportsfrom sourceregionsat low altitude. Thewindfieldsat 850hPa (Figure8) demonstrate that easterlywindsare predominant.The smokeplumedi-
4.2.
August 24, 1993
The winterozoneprofilerepresented by August24 radiosonde is different. For low altitudes,ozoneconcentrations remainweak(approximately 28 ppbv)and are associated with highvaluesof humidity(> 75%).
Abovethetemperature inversion at 3.8km (650hPa), the ozoneprofileshowsa strongenhancement to about 75 ppbv. This enhancement is followedby a slowerde-
rectionfieldsare in agreementwith the wind fieldsand crease to lowozoneconcentrations at 10km (500hPa) the tradewindinversion layers[Donque, 1975;Raholi- that prevailthroughthe hightroposphere.The weather jao, 1987;Preston-Whyte andTyson,1988;Garstang et datado not showa disturbance on the launching day al., 1996]. and on the days before. Satellite data from N OAAAVHRR provideevidenceof vegetationfiresin south4.1. February 23, 1993 easternAfricaand Madagascar (Figure9). Fewdeep The tropospheric ozoneprofileduringthis seasonis convective cloudsareobserved duringthisperiod(Figcharacterized byverylowvaluesofozoneconcentration,ure10). Theonlyclouds detected ontheNOAAimages
RANDRIAMBELO
ET AL.: SEASONAL VARIABILITY
OF TROPICAL
TROPOSPHERIC
Oz
11,823
by Machadoet al. [1992]for Meteosatdata (10.5-12.5 •um,5 km resolution)have beenadjustedfor NOAA data (10.5-11.5•umand 11.5-12.5•um,I km resolution). Deep convectivecloudsare situatednear the tropopause(brightnesstemperature< 190 K and reflectance> 70%), high cloudsbetween7 and 13 km,
-10
middle clouds between 3 and 7 km, and low cloudsbelow 3 km. This casestudy suggeststhat the injections
-2o
of the pollutantsfrom biomassburningto the free tropospherereachthe middle tropospherenear the source
regions.This suggestion is in agreementwith the flow to 500-hPaanalyzedby Garstanget al. [1996]. After this convection venting, ozone and its precursorsare advected to the Indian Ocean below the middle tropo-
-3o
sphere. Backtrajectoryanalysesin the vicinity of the 20
30
40
50
60
500 hPa level corroborate this contamination pattern
and suggestthat air massesoriginatefrom the burnt areas in southeasternAfrica and Madagascar(Figure 12). The low valuesof ozoneconcentrations at low altitude can be justified by the sources,which moved awayfrombiomass burningregion(Figure12,700hPa).
-JO
These results thus document the ozone contamination
of the middletroposphereoverthe Indian Ocean. Then it would be valuable to contrast this simple occurrence of ozonecontamination with a more complex casestudy
-2O
of intensebiomassburning and deep convectioncoincidence.
4.3.
-3O
October
12• 1993
The springrepresentativeozoneprofiledisplaysquite highvalues(around100 ppbv) in the wholefreetropospherefrom the trade wind inversionat around 3 km to the tropopause. The correspondinghumidity distribution is characterizedby very low values in the middle and in the high troposphere.No disturbanceis observed
-10
on the days precedingthe radiosondelaunchingin the studied zone. As is apparent in Figure 13, more vegetation fires and smokeplumes, compared with those in
August, are present in Africa and Madagascara few days beforethis launching[Randriambelo, 1998]. In addition, deep convectivecloud zones are detected in the vicinity of burning zones(Figure 14). In order to track back the origin of these air masses,backtrajectories starting from different altitudes are plotted over -3O the Reunion site as shown in Figure 15. These trajectoriesindicatethat from 250 to 660 hPa, air parcel originsare to the west of Reunion Island. Moreover, these air massesare crossingover southeasternAfrica Figure 8. Wind fieldsat 850hPa for (top) February22, and Madagascarwhere biomassburning and deep con1993,(middle)August24, 1993,and (bottom)October vective clouds are intensive. It is to be noted that the global model can not cap12, 1993 showingthat airflows in the low altitude are mostly westward. ture mesoscale eventslike deepconvection[Fuelberget aL, ].996]. Moreover,due to the limited observations availablein the oceans,the globalmodeldata are poorly are located in the low and middle troposphere(Fig- initialized in the southern tropics. Then the transports or subsidence) within the deepconvection ure 11). The cloudclassification is basedon thresholds (ascendance obtained from brightnesstemperature and reflectance zoneare not exactly describedin trajectoriespresented -20
values.
The thresholds
used on the infrared
channels
in this section.
Then the use of the satellite
data com-
11,824
tLANDRIAMBELO
ET AL.: SEASONAL VARIABILITY
OF TROPICAL
TROPOSPHERIC
Oa
10%
20øS
20øE
30øE
40øE
50OE
Figure 9. Map of fires detectedfrom NOAA satelliteduring August 20-24, 1993.
pletesthe analyses. Indeed, the NOAA satellitedata bustionsin southeasternAfrica and Madagascar.This characteristicprofileis obtainedat the beginni,ng of the
can give more information about the convection event
with I km x I km resolution (AVHRR,Figure14). deep convectionperiod, when the depressioncellsreach Consequently, if trajectories crossed overthe deepcon- the southernAfrican continent. During this period, a vectionregion,we can suggestthat the pollutantscan large quantity of fire by-productsreleasedin the boundbe injectedinto the freetroposphere. ary layer can be vented to the high tropospheredue to By comparing firesobtainedat the endof August deepconvection.Then upper level westerlywinds can
(Figure 9) andat thebeginning ofOctober (Figure 13),
advect these ozone-generating contaminantsto the In-
we note that fivefold more fires are observedin October
dian Ocean over the trade wind inversion.
than in August(Figure16). In addition,by contrast ECMWF weather data indicate that there was no with October,duringAugustvery little deepconvec- jet stream near potential sourceregionsand that aptive cloudsare observed.Theseanalyses suggest that proximatelythe samemeteorologicalsituation occurred high values of ozone concentrations observedover Re-
during the casestudy period for August and October.
unionIslandduringOctoberresultfrom deepconvec- Following Pickeringet al. [1994],backtrajectory anal-
tion occurrencesassociated with intense biomass com-
yses from cluster points are consideredto determine
10os
20oS
20OE
30OE
40OE
50OE
Figure 10. Map of deepconvective clouds detected fromNOAAsatelliteduringAugust21-24, 1993.
RANDKIAMBELO ET AL.: SEASONAL VARIABILITY OF TROPICAL TROPOSPHERIC O3
11,825
140
file, the maximum of ozonecontaminationoccursin the 650 to 450-hPa layer characterizedby high and approximately constantconcentrations.The high concentrations observedin this layer are likely to be explainedby its situation just above the trade wind inversionlayer. Largequantitiesof biomassburningby-productsare expectedto be vented by low or medium intensity convection to this level and then can be advectedby westerly
120
100
80
60
4O
winds to the Indian Ocean under conditions favorable 20
to ozonegeneration,namely, isolationfrom boundary layer compoundsby the trade wind inversion. Then 0 the August profile presentsa decreasein ozoneconcenDeep Deep High High Mddle I•ddle low low tration, as a function of altitude, in the 450 to 300-hPa Figure 11. Comparison of the top of convective cloud layer. This decreaseshouldstem from the intensityof occurrences obtainedin August (shaded)and October convectionnear the source regions which is given by (solid)1993. cloud altitude computedfrom NOAA data during this period. Since convectivecloudsnear sourceregionsare
observed to peakin the mi,'Jdle troposphere (Figure11),
the originof air masses.Thesetrajectories,computed a decreaseof fire by-products abundance is expected in from clusterspositionedaroundReunionIsland, do not evidencea large dispersionduringthe transport,and air massesare observedto come from biomassburning zonesin southeastern Africa and Madagascar(Figures 12 and 15).
the middletroposphere.After the easterlytransport,a similar decreasein ozone concentration is likely in the Indian Ocean troposphere.
Conversely, the Octoberprofiledisplaysstrongand quite constantmixingratiosin the freetroposphere up to very high, that is, 650 to 150 hPa, levels. These By usingthedataforthistypicalyearasthe database but omittingthe stratospheric contributions, it is valu- strongand approximatelyconstantvaluesof the mixof deepconvective ableto investigatethe influenceof fire and deepconvec- ing ratio accordwith the observation cloudsnearburningzones(Figure11). Deepconvection tion on ozone contamination as a function of altitude. Ozonemixingratio profiles,givenin Figure 7, suggesta next to firesis expectedto ventfire by-productsinto the Then upperlevelwesterlies are likely different ozone contaminationmode for each layer. For freetroposphere. to advect this fire contamination to the Indian Ocean. Augustand October,threelayerscanbe distinguished: During the venting and the transport, biomass burning 650-450hPa (3.7-6.7km), 450-300hPa (6.7-9.5km), areexpected to generate largequantities of and 300-150hPa (9.5-17 km). For the Augustpro- by-products
-10
-15
-20
-
-25
-
-30
20
•
25
30
•
35
•
40
•
45
•
50
•
55
60
Figure12. Cluster backtrajectories (latitude range:20.5ø-21 øS;longitude range:54ø-55 øE) ending onAugust 24,1993,overReunion Islandfordifferent levels'(a) 700hPa,(b) 500hPa, and(c)300hPa,with24-hour timeincrements given byasterisks, and(d)altitude corresponding to these transports.
11,826
-5
-10
-15
-20
-
-25
-
t) I I i I
-30 20
30
35
40
45
5O
55
6O
30
35
40
45
50
55
60
-5-
-!0
-
-15
-
-20
-25
-30 20 100
d
20O
300 400
500
60O
700
800
1000
21
22
23
Day
Figure 12. (continued)
24
Reunion
RANDRIAMBELO
ET AL.: SEASONAL VARIABILITY
OF TROPICAL
TROPOSPHERIC
Oa
11,827
10øS
20øS
20øE
30øE
40øE
50øE
l•igure 13. Map of fires detected from NOAA satellite during October 8-12, 1993.
ozonethrough the mix-then-cookamplificationprocess fires)in the studiedregion(Figure16). In addition,this result further validates the biomassburning contamina[Chat. fieldand Delany,1990]. Since stratospheric contributions are discarded and tion patterns in remote regionsin the Indian Ocean. are not likely to obscurethe ozonecontaminationanalysis, it is valuable to further quantify the relation5. Concluding Discussions ship betweenozoneenhancementsand vegetationfires. The analysisof a large set of complementarydata By computing the column ozoneenhancementover the backgroundvalue in the free troposphere,that is, from (PTU-O3 radiosoundings from 1992to 1999, NOAAthe trade wind inversion to the tropopause, we obtain AVHRR satellites record from 1989 to 1998, meteoro7.89 DU for Augustprofileand 28.75DU (about 4 times logicaldata, and ECMWF models)offerssomeinsight larger)for October(Figure17). This amplificationra- into the respectiverole of fire, smoke plumes, convectio is in reasonableagreementwith the increaseof fires tion, and transport which determinethe seasonalvaridetectedin August (1240 fires) and in October (6395 ability of the troposphericozone observedin remote
10ø$
20øS
20øE
30øE
40øE
50øE
Figure 14. Map of deepconvective cloudsdetectedfrom NOAA satelliteduringOctober8-10, 1993.
11,828
KANDRIAMBELO ET AL.' SEASONAL VARIABILITY OF TROPICAL TROPOSPHERIC Oo
10øS
20øS
• 661 mbar
20øE
30øE
40øE
50øE
20øE
30øE
40øE
50OE
20OE
30øE
40OE
50øE
10øS
20øS
10øS
20øS
Figure 15. Backtrajectories endingon October12, 1993overReunionIslandfor differentalti-
tudesfrom200hPato 660hPa,with(a) 24hours timeincrements, and(b) altitudecorresponding
to these transports.
region at Reunion Island. The 7-year measurements of ozone content in different layers led us to set apart atypical mechanisms.In order to further clarify contributionscomingfrom different mechanisms,it has been necessary to first identify and then extract stratospheric intrusions. We modified some criteria of VH, who established statistical studies of midlatitude tropopause
folds, and used the resulting schemeto filter stratosphericinputs beforeanalyzingthe biomassburning influenceon troposphericozone. The analysis of profiles of the typical years documents the relationships between, on one hand, the seasonalvariation of troposphericozone profilesabove the wind inversion level at Reunion Island, and on
ILiNDRIAMBELO
ET AL'
SEASONAL
VARIABILITY
OF TROPICAL
TROPOSPHERIC
03
11,829
lOO
b
200
300 400 500
600
700 800
1 ooo
9
10
11
12
Reunion
Day ,
Figure 15. (continued)
the other hand, anthropogenicsourcesof ozoneprecursors and deep convective clouds in southeastern Africa and Madagascar. The enhancedvaluesof tropospheric ozonedepend on the quantity of vegetationfire and the intensity of the convectiveclouds. The redistribution of ozoneas a function of altitude dependson the maximum height of convectivecloudsnext to sourceregions. During the dry season,the remote sensingof smoke plumesindicatesthat the dominant dispersionof pollutants at low levels is to the northwest
of their sources.
Due to the presenceof the stable layers, few pollutants could be injected into free troposphereand hence
reach the Indian Ocean. The only disturbances,which could perturb these stable layersduring the dry season, are cumulus clouds and frontal perturbations. These clouds are likely to inject biomass burning pollutants from low altitudes to the middle troposphere. From August, biomassburning beginsto influencefree tropospheric ozone concentrations, which reach more than 50 ppbv just above the trade wind inversion. During this
period,middlelevelconvectiveclouds(cumulusclouds) near the sourceregionsplay an important part in middle tropospherecontamination of the Indian Ocean. Then 14
12
lOO
10
8o
o
6o
N
o
4O
2o 0,5- 3,7
3,7-6,7
6,7-9,5
9,5-17
Altitude (km)
Abica
Madagascar
Figure 17. Enhancement of ozone over Reunion for
August(shaded)and October(solid)for differentlay-
Figure 16. Comparisonof firesin southeasternAfrica ers. Abovethe trade wind inversion,the total enhanced and Madagascarobserved on August20-24 (1240fires, ozoneis 28.75 DU in October and 7.89 DU in August
shaded)and October8-12 (6395fires,solid)1993.
which correspondsto 27.5% of the October value.
11,830
RANDRIAMBELO ET AL.: SEASONALVARIABILITY OF TROPICAL TROPOSPHERIC Oa
convectionintensityincreaseswith time to reacha deep convectivestage in the rainy season. During September,althoughfiresin southeasternAfrica are peaking, the ozone concentration observedat Re-
inatingthe forestfiresin Madagascarwhichare always intense.In southeasternAfrica, from November,which corresponds to the rainy season,the moistureincreased, and the meteorologicalcondition is not favorableto the
union does not reach its maximum. However, it is to
fire extent, then the ozoneformation could alsodecrease
be noted that during this period Madagascaris only affectedby savannafires. In addition, it is not a straightforward processof moving biomass burning products into the free troposphere in this region given the pres-
becausethe lackof ozoneprecursors and of incomingsolar radiation.
The dispersionof pollutants during transport dependson dynamicaldisturbancesand on chemicalproenceof stability[Stocks et al., 1996].Mostfiresin south- cesses.High concentrationsof ozoneare still observed ern Africa occur in the savannaor dry deciduousforest on Reunion Island during November-Decemberdespite duringthis period [Fuelberg et al., 1996; Thompson et longtransportdistances(1000 km for Madagascarand al., 1996;Swapet al., 1996b;Levine,1996a,b]. How- 2000 km for southeasternAfrica) from the sourcereever, there are yet forest fires detectedin the east coast gions which induce horizontal and vertical dispersions of Mozambique(Figure1) [Randriambelo et al., 1998]. of fire by-products,althoughthe intensityof fire is lower Forestfiresin Africa are not inducingas muchof a quan- at that time. The importance of ozonegenerationdurtity of ozoneto remote areas as one could expectfrom ing transport, which counteractsthe dilution effect, is fire countsand the mix-then-cookprocessproposedby further corroboratedby the quantitative relationship ChatfieldandDelany[1990].This pointis callingatten- obtainedbetweenfire counts(fivefoldincrease)and ention to the importanceof dilution during the transport hancedozone(fourfoldrise) observedat Reunionfrom of fire products into upper levels and to the need for a August to October. changein the currently accepteddynanaicsof transport from
western
African
contaminations
to the Atlantic
Ocean. Fire contaminationscomingfrom the southern part of the African continent to the southernAtlantic Ocean are observedto mostly travel through mediumlow levels,while southernAfrican fire contaminationsto the Indian Ocean are carried by westerlieshigher levels and above the trade
wind inversion.
Conversely, the maximumof ozoneprofileobservedduring October at Reunion is likely to comefrom biomass burning in southeasternAfrica and from intenseforest fires in Madagascar. In addition, convectiveclouds, which occur during this period, enhancethe distribution of pollutants to higher levels. Indeed, at the end of the dry season,conditionsare favorableto cumulonimbus cloud formation. In addition, forest fires are observednear these deep convectivezones. During this period, a large quantity of pollutants can be injected into the free troposphere.It is noticeablethat the studied regionscontainforestryzonescharacterizedby well-
developed smokeplumes[Randriambelo et al., 1998]. Ground observations indicate that around these zones,
Acknowledgments.
The authors would like to thank
ORSTOM-Reunion station for providing NOAA-AVHRR data, LPA radiosoundingteam (G. Bain, D. Faduilhe, J. M. Metzget, T. Portafaix, F. Posny,and S. Roumeau),and ECMWF for providing model analyses. We also wish to thank M. Bessaft for useful discussions. We are also indebted
to the two reviewerswhosesuggestions and correctionshave contributed to the improvementof the paper.
References Ancellet, G., J. Pelon, M. Beekmann, A. Papayannis,and G. Megie, Ground-based lidar studiesof ozoneexchange between the stratosphereand troposphere, J. Geophys. Res., 96, 22,401-22,421, 1991. Andreae, M. O., J. Fishman, and J. Lindesay,The Southern Tropical Atlantic Experiment (STARE): Transport and Atmospheric Chemistry near the Equator-Atlantic (TRACE-A) and SouthernAfrican Fire-AtmosphereResearchInitiative (SAFARI) :An introduction,J. Geophys. Res., i01, 23,519-23,520, 1996. Arino, O., and J. M. Melinotte, Fire index atlas, Earth Ohs. Q., 50, 11-16, 1995. Bachmeier, A. S., and H. E. Fuelberg, A meteorological overviewof the TRACE period, J. Geophys. Res., 10i,
23,881-23,888, 1996. thick smogsare generated in the afternoon. These Baldy, S., G. Ancellet, M. Bessaft, A. Badr, and D. Lan smogsevidencewidespreadbiomassburning emissions Sun Luk, Field observationsof the vertical distribution of in theseregions.In addition, lightningobservedin deep troposphericozoneat the island of la Reunion(southern convectivecloudscan producesignificantNOx leading tropics), J. Geophys.Res., 101, 23,835-23,849,1996. to the photochemicalproductionof ozonein the middle Baray, J. L., G. Ancellet, F. G. Taupin, M. Bessaft,S. Baldy,
and in the uppertroposphere [Pickeringet al., 1990]. In the free troposphere,ozonehas a long life time (90 days)and can thereforeinfluencemeasurements at both local and long-rangescales.Henceconvectiveuplifts and horizontal transports in the free troposphere could inducelarge amountsof troposphericozoneover wind and temperatureinversionsin the Indian Ocean
[Garstanget al., 1996].The decrease of the ozonecontaminationbeginningin Novembercan be explainedby the decrease of fire counts in southeastern Africa dom-
and P. Kechut, Subtropicaltropopausebreak as a possible stratosphericsourceof ozonein the tropical troposphere, J. Atmos. Sol. Terr. Phys., 60, 27-36, 1998. Baray, J. L., J. Leveau, J. Porteneuve,G. Ancellet, P. Keckhut, F. Posny and S. Baldy, Description and evaluation of a troposphericozonelidar implementedon an existing lidar in the southern subtropics, Appl. Opt., 38, 68086817, 1999a.
Baray, J. L., G. Ancellet, T. Randriambelo,and S. Baldy, Tropical cyclone Marlene and stratosphere-troposphere exchange,J. Geophys.Res., 10•, 13,953-13,970, 1999b. Baray, J. L., V. Daniel, G. Ancelletand B. Legra•, Planetary-
RANDRIAMBELO
ET AL.' SEASONAL VARIABILITY
scale tropopause folds in the southern subtropics, Geophys. Res. Left., 27, 353-356, 2000.
OF TROPICAL
TROPOSPHERIC
Oa
11,831
ozonemapping from total ozone spectrometerby a modified residual method, J. Geophys. Res., 103, 22,129-
Brivio, P. A., and J. M. Gr•goire,Spatialanalysisof vegeta22,145, 1998. tion fire patterns: A multiannualrosediagram atlas fire Jenkins,G. S., K. Mohr, V. R. Morris, and O. Arino, The for Africa, in Fire in Global Resource and Environmenrole of convectiveprocessover the Zaire-Congobasin to tal Monitoring, Rep. EUR17699, Joint Res. Cent., Ispra, the southern hemispheric ozone maximum, J. Geophys. Italy, 1997. Res., 102, i8,963-18,980, 1997. Chatfield, R., and A. C. Delany, Convectionlinks biomass Justice, C. O., J. D. Kendall, P. R. Dowry, and R. J. Scburning to increasedtropical ozone' However, modelswill holes,Satelliteremotesensingof firesduringthe SAFARI tend to overpredict Oa, Jo Geophys. Res., 95, 18,478campaignusingNOAA advancedvery high resolutionra18,488, 1990. diometer data, J. Geophys. Res., 101, 23,851-23,863, 1996. Chatfield,R. B., J. A. Vastano,H. B. Singh,and G. Sachse, A generalmodelof how fire emissions and chemistrypro- Kim, J. H., R. D. Hudson, and A.M. Thompson, A new duceAfrican/oceanicplumes(Oa, CO, PAN, smoke)in method of deriving time-averaged troposphericcolumn TRACE-A, J. Geophys.Res., 101, 24,279-24,306, 1996. ozone over the tropics using total ozone mapping spectrometer • ,,., .... ) radiances:.n•e. Crutzen, P. J. , L. E. --•.,,•, u•:n, j.p. .,,o•,,e•, •r W. H. Pollock, i * .... comvo•loon 'o and analyand W. Seiler,Biomassburningas a sourceof atmospheric sisusingTRACE-A data, J. Geophys.Res., 101, 24,317gasesCO, H2, N20, NO, CHaC1,and COS, Nature, 282, 24,330, 1996. 253-256, 1979. Krishnamurti,T. N., M. C. Sinha, M. Kanamitsu,D. OostD'Abreton, P. C., Lagrangiankinematicand isentropictraerhof, H. Fuelberg,R. Chatfield, D. J. Jacob,and J. Lojectory modelsfor aerosoland trace gastransport studies gan, Passive tracer transport relevant to the TRACE-A in southernAfrica, $. Aft. J. $ci., 92, 157-160, 1996. experiment,J. Geophys.Res., 101, 23,889-23,907, 1996. Danielsen,E. F., Stratospheric-tropospheric exchangebased Kuo, Y., H. Skumanich,P. Haagenson,and J. Chang, The on radioactivity, ozone and potential vorticity, J. Atraos. accuracyof trajectory modelsas revealedby the observing $ci., 25, 502-518, 1968. systemsimulationexperiments,Mon. WeatherRev., 113, 1852-1867, 1985. Diab, R. D., J. Barsby, G. Bodeker, M. Scourfield,and L. Salter, Satellite observationsof total ozone above South Levine, J. S. (Ed.), BiomassBurning and Global Change, vol. 1, MIT Press,Cambridge,Mass., 1996a. Africa, S. Aft. J. Sci., 7•, 13-18, 1992. Diab, R. D., et al., Vertical ozonedistribution over southern Levine, J. S. (Ed.), BiomassBurning and Global Change, vol. 2, MIT Press,Cambridge,Mass., 1996b. Africa and adjacent oceansduring SAFARI-92, J. GeoLoring, R. O., Jr., H. E. Fuelberg, J. Fishman, M. V. Watphys. Res., 101, 23,823-23,833, 1996. son, and E. V. Browell, Influence of a middle-latitude Donque, G., Le type de temps d'aliz• actif • Madagascar, cycloneon troposphericozone distributionsduring a peBull. Acad. Malgache,51/2, 55-59, 1975. riod of TRACE-A, J. Geophys.Res., 101, 23,941-23,956, Fishman, J., C. E. Watson, J. C. Larsen, and J. A. Logan,
Distribution of tropospheric ozone determined from satellite data, J. Geophys.Res., 95, 3599-3617, 1990. Fishman, J., J. M. Hoell, R. D. Bendura, R. J. McNeal, and V. W. Kirchoff, NASA GTE TRACE-A Experiment
(September-October1992)' Overview, J. Geophys.Res., 101, 23,865-23,879, 1996. Folkins, I., and C. Appenzeller, Ozone and potential vorticity at the subtropical tropopausebreak, J. Geophys.Res., 101, 18,787-18,792, 1996.
Fuelberg,H. E., R. O. Loring, M. W. Watson, M. C. Sinha, K. E. Pickering, A.M. Thompson, G. W. Sachse,D. R. Blake, and M. R. Schoeberl,TRACE-A trajectory intercomparison, 2, Isentropic and kinematic methods, J. Geophys. Res., 101, 23,927-23,939, 1996. Garstang, M., P. D. Tyson, R. Swap, M. Edwards, P. Kalle-
berg, and J. A. Lindsay,Horizontal and vertical transport of air over southernAfrica, J. Geophys.Res., 101, 23,72123,736, 1996. r•.,
j. r•
J. L. t•c...,.•
•,
t•
u..,c....,
•,' p. T
and STS-61 Mission Crew, Use of the Earth observation systemin the spaceshuttle programfor researchand documentation of global vegetationfires: A casestudy from Madagascar,in BiomassBurning and GlobalChange,vol. 1, edited by J. S. Levine, pp. 236-240, MIT Press,Cambridge, Mass., 1996. Gouger, H., J.P. Cammas, A. Marenco, R. Rosset, and I. Jonquilres,Ozonepeaksassociatedwith a subtropical tropopause fold and with the trade wind inversion: A casestudy from the airborne campaignTROPOZ II over the Caribbeanin winter, J. Geophys.Res., 101, 25,97925,993, 1996.
1996.
Machado,L. T. A., M. Desbois,and J.P. Duvel, Structural characteristics of deep convective systems over tropical Africa and Atlantic Ocean,Mon. WeatherRev., 120, 392406, 1992. Pickering, K. E., A.M. Thompson, R. R. Dickerson, W. T. Luke, D. P. McNamara, J.P. Greenberg,and P. R. Zimmerman, Model calculations of tropospheric ozone production potential following observedconvectiveevents, J. Geophys.Res., 95, 14,049-14,062, 1990. Pickering, K. E., A.M. Thompson, D. P. McNamara, and M. R. Schoeberl,An intercomparisonof isentropictrajectories over the South Atlantic, Mon. Weather Rev., 122, 864-879, 1994. Pickering, K. E., A.M. Thompson, D. P. McNamara, M. R. Schoeberl,H. E. Fuelberg, R. O. Loring, M. V. Watson, K. Fakhruzzaman, and A. S. Bachmeier, TRACE-A trajectory intercomparison, 1., Effects of different input analyses•J. Geophys.Res., 101, 23,909-23,925, 1996a. Pickering, K. E., et ai., Convective transport of biomass burning emissionsover Brazil during TRACE-A, J. Geophys. Res., 101, 23,993-24,012, 1996b. Piketh, S. J., H. J. Annergan, and P. D. Tyson, Lower troposphericaerosol loadings over South Africa: The relative contribution of aeolian dust, industrial emissions,and biomassburning, J. Geophys.Res., 10•, 1597-1607, 1999. Preston-Whyte, R. A., and P. D. Tyson, The Atmosphere and Weather of Southern Africa, Oxford Univ. Press, New York, 1988.
Raholijao, N., Etudedel'•coulement surl']ledeMadagascar en situations d'aliz• hivernal non perturbS: Approche ex-
perimentaleet num•rique (Simulationbidimensionnelle), Ph.D. thesis,Univ. of Clermont II, Aubi•re, France, 1987. Randriambelo, T., S. Baldy, M. Bessaft,M. Petit, and M. change,Rev. Geophys.,33, 403-439, 1995. Hudson,R. D., and A.M. Thompson,Tropicaltropospheric Despinoy,An improved detection and characterizationof
Holton,J. R., P. H. Haynes,M. E. Mcintyre,A. R. Douglass, R. B. Rood, and L. Pfister, Stratosphere-troposphere ex-
11,832
RANDRIAMBELO
ET AL-
SEASONAL VARIABILITY
active fires and smokeplumes in southeasternAfrica and Madagascar,Int. J. Remote $ens., 19, 2623-2638, 1998a. Randriambelo, T., D•tection satellitaire desfeux de v•g•tation et des zonesde convectionen zone tropicale: Application k l'•tude climatologiquede l'ozonetroposph•rique, Ph.D. thesis, Reunion Univ., Reunion Island, France, 1998b.
Randriambelo, T., J. L. Baray, S. Baldy, P. Br•maud, and S. Cautenet, A casestudy of extreme troposphericozone contamination in the tropics using in situ, satellite, and meteorologicaldata, Geophys.Res. Left., 26, 1287-1290, 1999.
OF TROPICAL
TROPOSPHERIC
Oa
and biogenicemissions to the tropical SouthAtlantic, in BiomassBurning and GlobalChange,vol. 1, editedby J. S. Levine,pp. 396-401, MIT Press,Cambridge,Mass., 1996b.
Taupin, F. G., M. Bessaft,S. Baldy, and P. J. Br•maud, Troposphericozoneabove the southwesternIndian Ocean
is stronglylinked to dynamicalconditionsprevailingin the tropics,J. Geophys.Res., 10•, 8057-8066, 1999. Thompson,A.M., K. E. Pickering,D. P. McNamara,M. R. Schoelberl, R. D. Hudson,J.-H. Kim, E. V. Browell,J. H. Kirchoff,and D. Nganga,Where did troposphericozone over southernAfrica and the tropical Atlantic comefrom
Reed, R. J., A study of characteristictype of upper level in October19927:InsightsfromTOMS, GTE/TRACEfrontogenesis,J. Meteorok, 12, 226-237, 1955. A andSAFAKI-92,J. Geophys. Res.,101,24,251-24,278, 1996. Reiter, E. R., Stratospheric-troposphere exchangeprocess, Rev. Geophys.,13, 459-474, 1975. Thompson,A.M., W. K. Tao, K. E. Pickering,J. R. Scala, Roelofs, G. J., J. Lelievield, H. G. J. Stair, and D. Kley, and J. Simpson,Tropical deep convectionand ozoneforOzone production and transports in the tropical Atlantic mation,Bull. Am. Meteorol.$oc., 78, 1043-1054,1997. region during the biomass burning season,J. Geophys. Tyson,P. D., M. Carstang,and R. Swap,LargescalerecirRes., 102, 10,637-10,651, 1997.
Siddans,R., B. J. Kerridge, W. J. Reburn, and R. Munro, Height-resolved ozone retrievals in the troposphere and lower stratospherefrom GOME, Earth Ohs. Q., 58, 1113, 1998.
culationof air oversouthernAfrica, J. Appl. Meteorol., 35, 2218-2236, 1996.
Van Haver,P., D. De Muer, M. Beekmann,and C. Mancier, Climatology oftropopause foldsat midlatitudes, Geophys. Res. Left., 23, 1033-1036, 1996.
Stocks,B. J., B. W. V. Wilgen, W. S. W. Trollope, D. J. McRae, J. A. Mason, F. We[rich, and A. L. F. Potgieter, Fuels and fire behavior dynamicson large-scalesavanna fires in Kruger national park, South Africa, J. Geophys. Res., 101, 23,541-23,550, 1996. Swap, R. J., M. Garstang, S. A. Macko, P. D. Tyson, W. Maenhaut, P. Artaxo, P. Kallberg, and R. Talbot, The long-rangetransport of southern African aerosolsto the tropical South Atlantic, J. Geophys. Res., 101, 23,777-
S. Baldy, J. L. Baray and T. Randriambelo, Laborato[re de Physique de l'Atmosph•re, Reunion University, B.P. 7151, 15 avenue Ren• Cassin, 97715 Saint-Denis-dela-R•union, France. (
[email protected])
23,791, 1996a.
Swap,R. J., M. Garstang,S. A. Macko,P. D. Tyson, and (ReceivedMarch 24, 1999; revisedOctober22, 1999; P. Kallberg, Comparisonsof biomassburning emissions acceptedOctober27, 1999.)
