1

PROGRESSIN THE RETRIE VAL OF SULPHUR SPECIESFROM MIPAS

A.B. Burgess,R.G. Grainger, and A. Dudhia

Atmospheric,OceanicandPlanetaryPhysics,Departmentof Physics,Universityof Oxford,Oxford, OX1 3PU,UKaburgess@atm.ox.ac.uk

ABSTRACT

Operationally only pressure,temperature and six sig-nificant tracegasesare retrieved by ESA from MIPASdata.However, informationonmany other speciesis alsopresentin thespectra.We applya varietyof techniquesandourown retrieval modelto retrieve theconcentrationof threeotherspecies:SO� , OCSandSF� . SulphurDiox-ide (SO� ) is anacidic gas with bothnatural andanthro-pogenic sourcesthatis rapidlyconvertedtosulphuric acidandhencesulphateaerosolsin theatmosphere. CarbonylSulphide(OCS) is producednaturally at the oceansur-faceandby biomassburning and,through stratosphericoxidation, it is thought to bethemaincontributor to non-volcanic stratosphericsulphate aerosols. SulphurHex-afluoride (SF� ) is almost entirely anthropogenic in itsoriginsandshows steadyyear-on-yearincreasesmakingit useful for age of air and tracerstudies. We antici-patethegood globalcoverageandcontinuity of datawillmakeMIPAS usefulfor thedeterminationof changesandtrendsin thequantity anddistribution of thesespecies–both natural andanthropogenic. In this paper we sum-marisethecurrent progressthathasbeenmadein there-trieval of theseimportantsulphur-containing species.Weshow somepreliminaryzonalmeanfieldsandbrieflyout-line themethodsapplied.

Key words:Envisat,MIPAS,OCS,SO2,SF6,Sulphur.

1. INTR ODUCTION

The MichelsonInterferometerfor Passive AtmosphericSounding (MIPAS) is oneof the instruments aboard theEuropeanSpaceAgency’s ENVISAT satellite. Thesun-synchronouspolar orbiting satellitehasa period of onehundred minutes, giving approximately 14 orbits perday. MIPAS measuresinfrared atmospheric limb ther-malemissionspectrafrom 685–2410cm �

�, whichcorre-

spondsto awavelength rangeof 14.5–4.1 � m. Thisrangeenablestheremotemonitoring of a wide varietyof tracespecies.Thenominal altituderangeobserved by MIPASis is 6–68km, coveringtheuppertroposphere,thestrato-sphereandthelowermesosphere.

1.1. Carbonyl Sulphide

Carbonyl Sulphide, OCS, is relatively unreactive in thetroposphere,so is available to enterthe stratosphere insignificantquantities. OCSis directly emittedfrom theoceansasabyproduct of bothbiologicalandphotochem-ical reactionsandhasseasonally-varyingsourceandsinktermsfrom soilsandforests[28, 30]. Anthropogenic ac-tivities,suchasbiomassburning andricecultivation, addto the atmospheric OCSburdenandaluminium produc-tion hasalsobeenhighlightedasa previously unnoticedsource[9, 25]. Recentanalyseshaveshown deepconvec-tive eventscantransport largequantitiesof OCSthroughthe tropical tropopauseinto the stratosphere [16]. Hereit is susceptibleto oxidation by hydroxyl radicalattackandto ultraviolet photolysis [2]. OCShadbeenproposedas the major sourceof Stratospheric Sulphate Aerosols(SSAs)by Crutzen[6]. More recently, however, this al-mostexclusive assignment of the origin of SSAsto hascomeinto question[4]. Large uncertaintiesin theatmo-sphericsulphur budget compound the problemsassoci-atedwith in situ measurement of background SSAsandmakesan exact assessmentof the OCScontribution ex-tremely difficult [10, 25]. In fact, modelssuggestthatunder50%of SSAparticlesoriginatefrom OCS,there-mainingsourcesof sulphatebeingupwardtransport fromthetroposphereandSO� oxidation [17, 24]. A commonobservation by theseauthors,of specialrelevanceto thiswork, is thescarcityof globally distributedprofile infor-mationfor bothOCSandSO� , making thevalidation ofmodel predictions difficult. The solution to this prob-lem is ideallysuitedto satellitemeasurements,with theirglobalcoverage.

1.2. Sulphur Dioxide

SulphurDioxide,SO� , is anacidicgaswith bothnaturalandanthropogenic origins andis of interestto scientistsdueto its closelinks with aerosolformationandacidifi-cationof precipitation in the troposphere.SO� forms amajorpartof thesulphur cyclein theEarth’satmosphere.It is solublein cloudsandaerosoldroplets, whereis itableto reactto form sulphuric acid,H � SO� . In general,themaximum concentrationof SO� is closeto its source

2

andthe amount of SO� decreasesrapidly with distance,indicative of a short troposphericlifetime. In the drystratosphere,especiallyat low altitudeswherethe con-centrationof OH radicalsis small, the lifetime for SO�risesto theorderof a few weeks[22, 26]. Direct injec-tion of SO� into thestratosphere by volcanic activity re-sultsin theincreasedformationof stratosphericaerosols.Photolysisof H � SO� aerosolsleadsto anincreasein SO�in theupper stratosphere, [27].

1.3. Sulphur Hexafluoride

SF� is oneof the mostefficient greenhousegases:it isthreetimesstronger asa ‘greenhousegas’thananequiv-alentvolumemixing ratio (VMR) of CFC–11[11] andathousandtimesthatof CO� . Thecontribution of SF� toradiative forcing is smallbecauseits currentatmosphericconcentrationis lessthanfivepartspertrillion by volume(pptv) [14, 15, 20, 29]. The sourcesof SF� arealmostentirely anthropogenic [11] although tracesmay origi-natenaturally from fluoritic rocks [8]. Thereis evidencethat the atmospheric concentrationof SF� is increasingby about7–8%peryear[15, 19]. However, it hasbeensuggestedthattherateof increasein emission,if notpro-duction,hasslowed in the pastfew years[5, 13]. Re-centwork [3] in applying MIPAS measurements to SF�led to a stratospheric growth rate for October2002 to2003of 6.5� 1.3 % yr �

�anda VMR for October2003

of 4.60� 0.05pptv. This is within the envelopeof fore-castsfrom previous studies,but supports the suggestionthat the rateof emissionis decreasing. Thewell knownemissionrecord of SF� combinedwith its chemical inert-nessmakeit anidealcandidate for useasa tracerandforageof air calculation. Hence,it is usefulin determiningtheflux of otherspeciesinto thestratosphere.

Overall, themajoranthropogenicsulphur-containing gasemissionsarein theform of SO� , whereasthedominantbiogenic emissionsconsistof (mainlyoceanic)DiMethylSulphide[4, 22] andCS� [25]. However, dueto its life-time OCS is thought to account for more than 80% ofgasphasesulphur in theuppertroposphere[16]. Sulphurhexafluoride maybeconsidered aswholly inert over thealtitudeof interest,andassuchformsanexcellent tracerspecies.

2. RETRIEVAL METHODOLOGY

As with all speciestargetedfor retrieval from MIPAS, aninitial feasibilitystudywasperformed.Thisgivesagoodindicationof thebenefitthat will beobtainedby furtherstudyof thespecies.Firstly, a referenceclimatology pro-file [18] thatformsapartof theMIPASprocessingdatasetwasusedto simulatethe radiance contribution from thetargetspeciesexpectedat thesatellite.This is shown forOCSin Figure1. Thereis no latitudinal or seasonalvari-ation in any of the threespeciesfor thereferenceclima-tology, soall retrievedlatitudinalstructureis basedontheeffectof measurements.Figure2 showsthesimulatedra-dianceandcontributionof thesulphurspeciesfor a12kmtangentheight, with theinstrumental noiselevels(NESR)

Figure 1. Carbonyl sulphideclimatology, initially usedin thestudyof retrieval feasibility.

Figure3. Assignmentof variouserror sourcesto thefinal errorbudget as calculatedduring microwindowselection.Thetotalerror is shown(solid line) together with therandomerror com-ponent(dottedline) basedon theinstrumentNESR.Thetotal ofthe systematicerror contributionsis alsoshown(dashedline).Significantsystematicerror sources include: ‘TEM’ – retrievedtemperature uncertainty, ‘GRA’ – error froma horizontal tem-perature gradient, ‘H2O’ – water vapour uncertainty, ‘O3’ –ozone profile uncertainty, ‘SHIFT’ – error from spectral cal-ibration, ‘SPREAD’– uncertaintyin the ILS and ‘SPECDB’–uncertainty in spectroscopic line parameters. Thea priori erroris assumedto be100%.

shown per band. OCS is unusualamongst speciesthatemit within the MIPAS spectralrange as it exhibits itsmostsignificantemissionin band‘D’, between1850 and2410cm�

�, sois theonlyspecieswhoseprimaryretrieval

informationoriginatesin thisband.

3

Figure 2. A radiativetransfersimulationof therelativeintensityof thesulphur speciesfeaturescomparedto thetotal radiancefor a12kmtangent altitude, basedonclimatological abundances.Theinstrumentnoiselevel (NESR)is shownasthedashedhorizontalline.

Figure 4. Plot indicating the position and altitude range(shaded blocks, left axis) of the selectedmicrowindows,over-plottedwith the total radiance(grey line, right axis)within thesignificant spectral region for a 12 kmtangent height.Thelow-ermostred line (right axis) showsthe radiancecontributionfromOCSto this total radiance. TheMIPASnominalmeasure-mentaltitudesare shownaspalehorizontal lines.

2.1. Micr owindows

Theselectionof microwindows enableda moredetailedconsideration of the feasibility of retrieval. Microwin-dows are small regions of the spectrumno more than3 cm�

�wide. Theseregions areoptimisedto maximise

the information obtainedon the target species,whilstminimisingbothrandomandsystematicerrors.An infor-mative discussionof the development of the microwin-dow selectionalgorithm may be found in Dudhia et al.

[7] andanupdatedversionis availablein theseproceed-ings. Together with themicrowindow regions,a detailederroranalysisis alsoobtained,showing the relative sig-nificanceof all the known error sources. Again, OCSis usedasanexample, although thesameprocedureanderroranalysesexist for theotherspecies.Theerroranal-ysis is shown in Figure3. As a result of the analysis,we canconclude that the rangeover which an OCSre-trieval shouldbeworthwhile is 6–25 km, with 30 km asan upper limit. The microwindow spectralandaltituderangesareshown in Figure4. Similar resultswereob-tainedfor theotherspecies.Themicrowindow selectionwasvalidatedusingcontrolledretrievalsfrom ‘blind test’spectra,originally createdaspartof theE.U.FrameworkV ‘AMIL2DA’ study[1].

In orderto perform the retrievals,we focusedon an op-timal estimationapproach, formalised by Rodgers [21].In addition asimplecloud-screeningprocedure,basedonthe ratio of two small spectralregions [23] wasalsoap-plied. This removed thosescanscontaining cloud. Oncompletion, testsareperformedon thediagnosticsof theretrieval to remove any profilesarephysically unreason-able.Useof theaveragingkernelsallowsustocheckboththeamount andtheverticaldistributionof informationinthefinal resultthathasbeencontributedby themeasure-ment.

3. RESULTS

Theresultsaregroupedinto threesections,by species.

4

Figure 6. Zonal meanOCSdistribution, August2003. Approximately10,000retrievedvalueswere binnedto a 5� latitudeby 3 kmaltitude grid covering the globe. For each bin large statisticaloutliers, if present, were removed. Note the higher valuesover theNorthern latitudesat low altitude (6 km), possiblydueto anthropogenic emissions.Thelow valuesrecorded in a bandaround 70Smaybedueto a mixtureof dynamicaland(photo-)chemicalprocesses– there areobservedcorrelationswith boththevortex range andterminator– but the low polar winter temperaturescoupledwith the locationof themicrowindowsin the ‘D’ band(on the tail of thePlack curve)maybe significant. Correlativedata for validation is lacking so this is an area for further investigation.Althoughnotshownin this figure, high variability wasobservedin the southern polar region, probably originating from the fact that the meaniscomputedfrommeasurementsthat are bothinsideandoutsidethevortex.

Figure 7. A ‘slice’ throughFigure 6 showingtheglobal distribution of OCSbetween11 and13 km(150 to 250mb). Approximately10,000 retrievedvalueswere binnedto a 20� by 10� grid covering the globe. For each bin large statisticaloutliers, if present, wereremoved. Thehigh valuesover Indonesiamaybe anomalous,as there is very poor coverage in this region (dynamicaleffectscausestrongascentin this region, leadingto almostall profilesbeinginvalidatedby cloud). However, thevaluesare consistentout into theIndian Ocean,which hasmuch more acceptablecoverage. In addition, the structure over the NorthernAtlantic and Europe is notunreasonable. Thelower values(around300pptv– blue)centeredover thesouthernwinter poleare indicativeof descending air fromaround 15 to 20 kmasthemainOCSlossmechanisms,such photolysis,do not operatein thepolar night.

5

Figure 8. Zonal meanSO� distribution, August2003, displayedin a similar mannerto Figure 6. We can seea clear NorthernHemisphericsourcetermthat hasbeenassignedto anthropogenicemissions,althoughtheextentto such high latitudesis lesscertain.Thetropopauseis distinctasSO� hasa shortlowerstratospheric lifetime. Theincreaseabove10 mb,or 30 km,is thoughtto beduetothephotolysisof gas-phasesulphuricacid that hasboiledfromthetop of thestratosphericaerosol layer. Themid-stratosphere valuesstronglyreflect thea priori valuesasthesignalhasdroppedbelowthedetectionthresholdfor such low VMRs.TheenhancementovertheSouthpoleis yetto beassigned,andis thought to arisefromthelow temperatures. However, somemechanismshavebeensuggestedandinitial investigations into publishedresultsfrommodelshavesuggestedthat there is an enhancementin this region (e.g. Liao etal.[12]).

Figure9. A similar plot to Figure7, but for SO� . TheAntarctic valuesseemunreasonablyhigh,asdo thoseover Indonesia– althougha proportion maybeassignedto deepconvectiveeventsandbiomassburning. TheNorth AmericanandEuropeanenhancementsarenot quitecentredon therespective continents,which givessomecausefor concern. In additionexpectedenhancementsdueto effusivevolcanic activity and other industrialisedregions such as India havenot beenobserved. However, the interhemispheric variationshowingmore (anthropogenic)emissionin theNorthernHemisphere is directly in line with expectations.Asthere wasno latitudinalvariation in theprior informationusedin theretrieval, this is an encouraging result.

6

Figure 10. Zonal meanSF distribution, August2003, displayedin a similar mannerto Figure 6. Expectedfeatures,such as thelow over theAntarctic poledueto thedownward motionof air fromthemesosphere, and higher NorthernHemisphericvaluesat thelowermostaltitudescanbeseen.Thetropical mid troposphere hadbeenmaskedin this plot dueto cloudissues.

3.1. OCS

Oneof thefirst resultsfor OCS,a meanprofile, is shownin Figure 5. This resultcanbecomparedto theshapeof

Figure 5. Sampleof resultsfrom the OCSretrieval, showingthemeanretrievedprofile, with thestandard deviationshownastheshadedregion. Pointsrepresent thedistribution of retrievedvalues.Someof this scatterarisesfrom variability aroundtheorbits used.Theprofile reportedin Chin & Davis[4] is similarin the lower atmosphere but, probably dueto theextrapolationthat wasperformed,reaches1 pptvat around37 kminsteadof33 km.

the climatology shown in Figure1. The lower altitudesaresimilar in valueandcontainalmostentirelymeasure-mentinformation. Towards thetop of theprofile, around35 km, the contribution from the measurement is muchreduced. However, theresultis still animprovementoverthe previousdatawhich werepurely anextrapolation ofthe20to 25km values.

Collatedglobal resultsfor OCSareshown in Figures6and 7, which containapproximately 10,000 successfulprofile retrievals over the month of August2003. Thiscorresponds to around 200 orbits of data,which is justover half thenumber of orbits performedby thesatellitein that period. It is hoped to usethe remainingdataasit becomesavailable.Figure6 shows thezonalmeanfora 5 latitudebin sizeagainst retrievedpressure,which isthe resultof the joint pressureandtemperature retrievalthat is performedbefore the tracegasretrieval. Thekeyfeaturesare described in the figure caption. Figure 7shows a ‘slice’ through this zonalmeanfield to show theglobal distribution. In this casethe third retrieval ele-mentin eachprofile wasselectedbefore beingbinnedtothe 20 longitude by 10 latitudegrid according to thegeolocation recorded in the ESA level 1B dataproduct.This elementcorrespondsto profile points whosealti-tudeslie in the range 11 to 13 km (approximately150to 250mb). Additional zonalmeanplotswerecreatedtoexaminethevariation within themonth, in orderto checkfor internalconsistency. Thecomparisonof thefirst halfof themonthwith thesecondhalf showedsimilar struc-ture.Dueto thelifetime andlossmechanismsof OCS,nodiurnal variationwasexpected. The daytimeandnight-timezonalmeanswereindeedsimilarbut containedsomesignificantdiscrepancies. Someoriginate from the biasin globalcoveragebetweendayandnight, thatoriginates

7

from an incompletedatasetfor themonth. Theremain-ing differencesareunder investigation.

3.2. SO�

Theinitial resultsfor SO� shown in Figures8 and9 wereconstructedin the samemanner asfor OCS.The figurecaptionscontain a description of the key featuresalongwith suggestionsas to the origin of someunexpectedanomalousfeatures.

The atmospheric extremes in the southernpolar regiongive rise to greateruncertainties and higher variabili-ties in all the species.Part of the reasonfor this is thestrongreduction in thestrengthof the thermalemission,on which MIPAS relies,whenthetemperaturesdropbe-low 200K. In additionthereis a highly perturbedchem-ical environment which is currently only consideredasaseriesof ‘polar climatology’ profiles. Finally, thereex-ist very high altitudeclouds,suchaspolarstratosphericclouds,whicheitherreducethenumberof samples(if de-tectedascloud andflagged) or subtly perturb the spec-trumin waysthatourretrieval is unableto accommodate.

3.3. SF�

Finally, the zonal meanresult for SF� shown in Figure10. Thereis morevariability in the field thanwould beexpectedfrom suchalong-livedspecies,although theen-tire scalecoversonly 5 pptv. Themostinterestingfeatureis the low above the Antarctic polar vortex attributedtodescending mesospheric air. The low valuesat higheraltitudesarea mixture of the greaterageof air andthefirst significantSF� lossmechanismwhich occursin thelowermesosphere. As with all threespecies,theprior in-formation contains no latitudinal structure,so this ‘low’originatesentirelyfrom features within theMIPAS spec-tra.

4. CONCLUSIONS

We haveshown theprogressthatis currently beingmadein extendingthelist of speciesthatmayberetrieved fromMIPAS data.All threespeciesarecurrently availableforAugust2003 with informationon latitude,longitudeandaltitudeavailable,allowing bothzonalmeansandglobalmapsto becreated.Theinitial validationwork, basedontheretrieval diagnosticsandliteraturesearchesfor modeland measurement data is encouraging. The fields pre-sentedherecontain significantlymoregeolocatedmea-surementdatathanareavailablein theliterature.As thepreviously highlighted issueswith eachof the species–suchasthepersistentenhancementandlow measurementdensityover Indonesia– areresolved, this work shouldform the basisof a valuabledatasetcovering the mea-sured 3-dimensional distributions of thesespeciesovertheglobe.

Someprovisional work hasbeenperformedin applyingMIPAS observations to the partitioning and massbud-getsof OCSandSO� in theuppertroposphereandstrato-sphere,andasmoreretrieval databecomesavailablewehopeto pursue thisarea.Indeed, first resultsindicatethatthepartitioning of thesespeciesgiven in the2002Scien-tific Assessmentof OzoneDepletion is outby 30to 50%.

5. FURTHER WORK

An assessmentof thesimilarity betweenthesamemonthsin 2002, 2003 and 2004 will contribute towards boththevalidation andtheuseful applicationof MIPAS data.Comparisonagainstmodelfieldswould assistvalidationof the dataas will the useof any in situ or alternativecolocatedmeasurements.In addition, calculationof OCSand SO� fields for the entire year of 2003 would use-ful from both a validation and an applicationpoint ofview. Relatedto this is the investigation of theSouthernHemisphericOCSvariability andstructure.Otherconse-quencesof more accurateabundancesandvariability areimprovedestimatesof atmosphericlifetimes.

On a moretechnical side,thereremaina varietyof sub-tleties to investigate.Firstly, it is expectedto begin thetransitionto a newer retrieval code,which is ableto per-form joint retrievalswith otherspecies.Thesecondareaof investigationis pushing at the cloudboundary, to tryandreducethesizeof thedata‘gaps’around theequator.Finally, asalready discussed,investigationsinto thevari-ousanomalies,suchashigh southpolarSO� , needto beexplained.

FURTHER INFORMA TION

The Oxford group group performs a varietyof trace gas retrievals and MIPAS data health-monitoring. A large amount of diagnostic in-formation together with preliminary informationfor other months is available from our website,http://www.atm.ox.ac.uk/group/mipas/.

ACKNOWLEDGMENTS

We would like to acknowledgeNERC(UK) for fundingthis researchand the Royal Meteorological Societyforproviding travel assistance.

REFERENCES

[1] AMIL2DA, 2003, AMIL2DA Final report, Tech.rep.,IMK, co–ordinatedby T. vonClarmann

[2] BarnesI., Becker K., PatroescuI., 1994,Geophys.Res.Lett., 2389–2392

[3] BurgessA., GraingerR., DudhiaA., PayneV., JayV., March 2004, Geophys. Res.Lett., 31, L05112,doi:10.1029/2003GL019143

8

[4] Chin M., Davis D., 1995, J. Geophys. Res.,100,8993

[5] ConnellP.O., Heil F., Henriot J., et al., 2000, SF�in the electric industry, status2000, www.cigre-sc23.org/SF6/status2000.pdf

[6] Crutzen,1976, Geophys.Res.Lett, 3, 73[7] DudhiaA., JayV., RodgersC., 2002, Applied Op-

tics,41,3665[8] HarnischJ., Eisenhauer A., July 1998, Geophys.

Res.Lett., 25,2401[9] HarnischJ., BorchersR., FabianP., Kourtidis K.,

1995, Environ. Sci.Pollut.Res.,2, 161[10] Kettle A., Kuhn U., von HobeM., KesselmeierJ.,

AndreaeM., November 2002, J. Geophys. Res.-Atmos.,107, 4642

[11] Ko M., SzeN.D., WangW.C., et al., 1993, Journalof GeophysicalResearch,98,10499

[12] Liao H., Adams P., Chung S., et al.,2003, J. Geophys. Res., 108(D1), 4001,doi:10.1029/2001JD001260

[13] MaissM., BrenninkmeijerC., October1998, Envi-ron.Sci.Technol., 32,3077

[14] MaissM., Levin I., April 1994,Geophys.Res.Lett.,21,569

[15] MaissM., SteeleL., Francey R., et al., May 1996,Atmos.Environ.,30,1621

[16] NotholtJ.,KuangZ., RinslandC.,etal.,April 2003,Science,300,307

[17] Pitari G., Mancini E., Rizi V., ShindellD., 2002,J.Atm. Sci.,59,414

[18] Remedios J., 2001, www.atm.ox.ac.uk/RFM/rfm downloads.html, Downloads of Ref-erenceAtmospheres

[19] Rinsland C., andM.C. Abrams M.G., Lowes L.,ZanderR.,MahieuE.,November1993, J.Geophys.Res.-Atmos.,98,20491

[20] RinslandC., Goldman A., StephenT., et al., April2003, J.Quant. Spectrosc.Radiat.Transf., 78,41

[21] RodgersC., 2000, Inverse Methods for Atmo-sphericSounding: Theory andPractice,World Sci-entific

[22] Seinfeld,Pandis,1997, Atmospheric ChemistryandPhysics: From Air Pollution to Climate Change,Wiley

[23] SpangR., RemediosJ., Barkley M., 2004, AdvSpaceRes,33,1041

[24] Timmreck C., November2001, J. Geophys. Res.-Atmos.,106, 28313

[25] WattsS.,2000,Atmos.Environ.,34,761[26] WayneR., 1999, Chemistryof Atmospheres,3rd

edition,Wiley[27] WeisensteinD., YueG., Ko M., et al., June1997, J.

Geophys.Res.,102,12019[28] WeissP., AndrewsS.,1995, Geophys.Res.Lett,22,

215[29] ZanderR., RinslandC., DemoulinP., August1991,

J.Geophys.Res.-Atmos.,96,15447[30] ZeppR., Andrae M., 1994, Geophys.Res.Lett, 21,

2813