abundancezcloud

OMIAlgorithmTheoreticalBasisDocumentVolumeIVOMITraceGasAlgorithmsEditedbyK.
ChanceSmithsonianAstrophysicalObservatoryCambridge,MA,USAATBD-OMI-02,Version1.
0,September20012ATBD-OMI-04TableofContentsPREFACE5Background.
5PurposeoftheATBD5ContentsoftheATBDs.
5Summary61.
OVERVIEW71.
1.
NO2,HCHO,BRO,ANDOCLO.
81.
2.
SO2.
111.
3.
REFERENCES.
132.
NO2.
152.
1.
INTRODUCTION.
152.
2.
ALGORITHMDESCRIPTION.
162.
2.
1.
Slantcolumnmeasurements.
17RadiancefittingbyDOAS.
17Referencespectra.
18Fittingwindow202.
2.
2.
Airmassfactorsandverticalcolumnabundances21Airmassfactorcalculations21Binningandsmoothing22Correctionoftheverticalcolumnforpollutedconditions.
232.
2.
3.
Outputs.
242.
2.
4.
Validation.
252.
3.
ERRORANALYSIS.
252.
4.
ACKNOWLEDGMENTS262.
5.
REFERENCES.
272.
A.
APPENDIXA:AIRMASSFACTORSOVERCLOUDEDSCENES.
292.
A.
1.
Effectsofcloudsandaerosols.
292.
A.
2.
Definitions292.
A.
3.
Unpollutedandpollutedprofiles.
302.
A.
4.
Application.
302.
A.
5.
Cloudlesscase.
302.
A.
6.
Clouds.
312.
A.
7.
Temperaturecorrection.
322.
B.
APPENDIXB:ERRORANALYSISDETAILS.
332.
B.
1.
Slantcolumndensity.
332.
B.
2.
AirMassFactor.
342.
CAPPENDIXC:DATAPRODUCTSTABLEANDDATAPRODUCTDEPENDENCIES.
363.
HCHO.
373.
1.
SLANTCOLUMNMEASUREMENTS383.
1.
1.
Nonlinearleast-squaresfitting393.
1.
2.
Re-calibrationofwavelengthscales.
403.
1.
3.
Referencespectra.
413.
1.
4.
Common-modecorrection433.
1.
5.
Radiancefitting:BOAS(baselineoption)433.
1.
6.
Radiancefitting:DOAS(non-baselineoption)443.
2.
AIRMASSFACTORSANDVERTICALCOLUMNABUNDANCES.
443.
3.
ERRORESTIMATES.
453.
4.
OUTPUTS46ATBD-OMI-0433.
5.
VALIDATION.
463.
6.
REFERENCES.
474.
SO2494.
1.
INTRODUCTION.
494.
1.
1.
VolcanicSO2.
504.
1.
2.
AnthropogenicSO2504.
1.
3.
Volcanicashmeasurements.
504.
2.
SELECTIONOFOPTIMUMSPECTRALREGION514.
3.
DETAILEDDESCRIPTIONSOFTHESO2ALGORITHM524.
3.
1.
Inversionstrategy.
524.
3.
2.
Forwardmodel.
534.
3.
3.
Inversiontechnique.
544.
4.
ERRORANALYSIS.
554.
4.
1.
Checkingconsistencyoftheforwardmodelwiththemeasurements554.
4.
2.
Estimatingretrievalerror554.
4.
3.
CorrectionforRingeffect.
574.
4.
4.
Correctionforvolcanicash.
574.
5.
OUTPUTS574.
6.
VALIDATION.
584.
7.
REFERENCES.
595.
BRO615.
1.
SLANTCOLUMNMEASUREMENTS625.
1.
1.
Nonlinearleast-squaresfitting625.
1.
2.
Re-calibrationofwavelengthscales.
635.
1.
3.
Referencespectra.
645.
1.
4.
Common-modecorrection645.
1.
5.
Radiancefitting:BOAS(baselineoption)655.
1.
6.
Radiancefitting:DOAS(non-baselineoption)665.
2.
AIRMASSFACTORSANDVERTICALCOLUMNABUNDANCES.
665.
3.
ERRORESTIMATES.
675.
4.
OUTPUTS685.
5.
VALIDATION.
685.
6.
REFERENCES.
686.
OCLO716.
1.
SLANTCOLUMNMEASUREMENTS716.
1.
1.
Nonlinearleast-squaresfitting716.
1.
2.
Re-calibrationofwavelengthscales.
726.
1.
3.
Referencespectra.
736.
1.
4.
Common-modecorrection736.
1.
5.
Radiancefitting:BOAS(baselineoption)746.
1.
6.
Radiancefitting:DOAS(non-baselineoption)756.
2.
AIRMASSFACTORSANDVERTICALCOLUMNABUNDANCES.
756.
3.
ERRORESTIMATES.
756.
4.
OUTPUTS766.
5.
VALIDATION.
766.
6.
REFERENCES.
774ATBD-OMI-04ATBD-OMI-045PrefacePrefacetothefourOMI-EOSAlgorithmTheoreticalBasisDocuments(ATBDs)fortheOzoneMonitoringInstrument(OMI),tobelaunchedmid-2003onNASA'sEOS-AurasatelliteBackgroundOMI-EOSisaDutch-FinnishozonemonitoringinstrumentthatwillflyonNASA'sAuraMission,partoftheEarthObservationSystem(EOS).
SinceAura'spurposeistocarryoutextensivestudiesoftheEarth'satmosphere,OMI'smeasurementsofozonecolumnsandprofiles,aerosols,clouds,surfaceUVirradiance,andthetracegasesNO2,SO2,HCHO,BrO,andOClOfitwellintotheAuramissiongoals.
Awide-swath,nadir-viewing,nearUV-visiblespectrograph,OMIdrawsheavilyonEuropeanexperienceintheatmosphericresearchinstrumentsGOME(onERS-2),SCIAMACHY(tobelaunchedlaterin2001onEnvisat),andGOMOS(alsoonboardEnvisat).
PurposeoftheATBDThissetoffourOMI-EOSAlgorithmTheoreticalBasisDocuments(ATBDs),isintendedtopresentadetailedpictureoftheinstrumentandtheretrievalalgorithmsusedtoderiveatmosphericinformationfromtheinstrument'smeasurements,sothatthereisaclearunderstandingoftheresults,withinthecommunityofOMIscientists,withintheAuraScienceTeam,andamongtheatmosphericcommunityatlarge.
EachofthechaptersofthefourATBDsiswrittenbyscientistsresponsibleforthedevelopmentofthealgorithmspresented.
TheseATBDswillbepresentedtoagroupofexpertreviewersrecruitedmainlyfromtheatmosphericresearchcommunityoutsideAura.
Theresults—critiquesandrecommendations—ofthereviewersstudyoftheATBDwillbepresentedatanAurameeting,currentlyscheduledfor5November2001.
SincechangestotheinstrumentandtheLevel0to1Bprocessinghaveapotentialforsignificantcostandscheduleimpact,thefirstATBDispresentedforpurposesofcommentandinformation.
However,progressiontothewritingofacceptable,operationalsoftwareforthealgorithmsforwhichthedataproductswillbearchivedintheDAAC(DistributedActiveArchiveCentre)atNASA'sEOSDISintheotherthreeATBDs,dependsontheapprovaloftheATBDbytheAuraMissionScientist.
ContentsoftheATBDsATBD1containsageneraldescriptionoftheinstrumentanditsmeasurementmodes.
Inaddition,thereisapresentationoftheLevel0to1Balgorithmswhichconvertinstrumentcountstocalibratedradiances,groundandin-flightcalibration,andtheflightoperationsneededtocollectsciencedata.
Itiscriticalthatthisiswellunderstoodbythedevelopersofthehigherlevelprocessing,astheymustknowexactlywhathasbeenaccountedfor(andhow),andwhathasnotbeenconsideredintheLevel0to1Bprocessing.
ATBD2coverstheozoneproducts.
AlthoughOMIisnadir-viewing,thisdoesnotlimititsoutputtoverticalcolumns.
Rather,becauseitsoutputisacontinuous,moderate-resolutionspectrumthatincludestherangeofozoneabsorption,itispossibletodevelopanalgorithmforverticalprofileaswell.
ThecontinuousspectrumalsomakesitpossibletouseaDOAS(DifferentialOpticalAbsorptionSpectroscopy)techniquedevelopedinconnectionwithGOME,flyingonERS-2.
Atthesametime,animprovedversionoftheTOMStotalozonecolumnalgorithm—developedandusedsuccessfullyover3decades—willbeusedonOMIdata.
RoundingoutthegroupoffouralgorithmsinthisATBDisaseparate,independentestimateof6ATBD-OMI-04troposphericcolumnozone,usinganimprovedversionoftheTroposphericOzoneResidual(TOR)andcloudslicingmethodsdevelopedforTOMS.
ATBD3presentsretrievalalgorithmsforproducingtheaerosols,clouds,andsurfaceUVradiationproducts.
Retrievalofaerosolopticalthicknessandaerosoltypeispresented.
Aerosolsareofinterestbecausetheyplayanimportantroleintroposphericpollutionandclimatechange.
Thecloudproductsincludecloudtopheightandeffectivecloudfraction,bothofwhichareessential,forexample,inretrievingthetracegasverticalcolumnsaccurately.
EffectivecloudfractionisobtainedbycomparingmeasuredreflectancewiththeexpectedreflectancefromacloudlesspixelandreflectancefromafullycloudypixelwithaLambertianalbedoof0.
8.
Ontheotherhand,twocomplementaryalgorithmsarepresentedforcloud-topheight(orpressure).
OneusesaDOASmethod,appliedtotheO2–O2absorptionbandaround477nm,whiletheotherusesthefilling-inofselectedFraunhoferlinesintherange352-398nmduetorotationalRamanscattering.
SurfaceUVirradianceisimportantbecauseofitsdamagingeffectsonhumanhealth,andonterrestrialandaquaticecosystems.
OMIwillextendthelong,continuousrecordproducedbyTOMS.
ATBD4presentstheretrievalalgorithmsforthe"additional"tracegasesthatOMIwillbeabletomonitor:NO2,SO2,HCHO,BrO,andOClO.
Thesegasesareofinterestbothbecauseoftheirrespectiverolesinatmosphericchemistry,aswellastheirpotentialforpollution.
ExtensiveexperiencewithGOMEhasproducedspectralfittingtechniquesusedinthesenewlydevelopedretrievalalgorithms,eachadaptedtothespecificcharacteristicsofOMIandtheparticularmoleculeinquestion.
SummaryIndividually,thefourOMI-EOSATBDspresentindetailhoweachofOMI'sdataproductsareproduced.
TogethertheydemonstratetheimportantcontributionsOMImakestoaddressingAura'sscientificquestions.
P.
F.
Levelt(KNMI,TheNetherlands)PrincipalInvestigatorG.
H.
J.
vandenOord(KNMI,TheNetherlands)DeputyPIE.
Hilsenrath(NASA/GSFC,USA)Co-PIG.
WLeppelmeier(FMI,Finland)Co-PIP.
K.
Bhartia(NASA/GSFC,USA)USSTLeaderATBD-OMI-0471.
OverviewK.
Chance,T.
P.
Kurosu,andL.
S.
RothmanSmithsonianAstrophysicalObservatoryCambridge,MA,USAOMI-theOzoneMonitoringInstrumentonEOSAura-providesozonemeasurementstocomplementtheotherspeciesmeasurementsfromAura,followingontheheritageoftheTOMSandSBUVinstruments.
Inaddition,itprovidesenhancedinformationontheverticaldistributionofatmosphericozone,includingthetroposphericburden,clouds,radiation,andsurfaceUVinformation,andmeasurementsofanumberofadditionaltracegases.
ThedataproductsforthetracegasesNO2,HCHO,SO2,BrO,andOClOarethesubjectofthisvolumeoftheOMIAlgorithmTheoreticalBasisDocument(ATBD).
TheOMItracegaseshaveallbeenmeasuredfromthegroundusingUV/visiblespectroscopy.
NO2andSO2weremeasuredfromspace,bytheSAGEandtheTOMS/SBUVinstruments,respectively,usingdiscretewavelengthbands.
MeasurementsoftheentiresuiteofmoleculeswereproposedfortheSCIAMACHYandGOMEinstruments,measuringthefullUV/visiblespectrumatmoderateresolution[Chanceetal.
,1991;Burrowsetal.
,1993].
AllspecieshavenowbeensuccessfullymeasuredinthenadirgeometrybytheGOMEinstrument:NO2verticalcolumnabundancesareretrievedoperationally,whiletheothergases(andadditionaldeterminationsofNO2,includingtotalcolumnabundancesandtroposphericabundances)areretrievedinresearchbyanumberofEuropean(e.
g.
,Burrowsetal.
,[1999])andU.
S.
groups(e.
g.
,theSAO).
AnticipatedtracegasdataproductsfromOMIaresummarizedinTable1.
Table1.
1OMITraceGasDataProductSummaryProductTemporalResolutionHorizontalResolution::Coverage1NO2verticalcolumn(cm-2)Once/day26*48km::GDHCHOverticalcolumn(cm-2)Once/day13*24km::GDSO2verticalcolumn(cm-2)Once/day13*48km::GDBrOverticalcolumn(cm-2)Once/day13*24km::GDOClOslantcolumn(cm-2)Once/day26*48km::V1Grepresentsglobalcoverage,Ddaylight,andVvortexInthischapterwesummarizetheproceduresusedtoobtainslantcolumndensities(Ns)andverticalcolumndensities(Nv)forNO2,HCHO,SO2,BrO,andOClOfrommeasuredOMIspectralradiancesandirradiances.
Thebackscatteredradiancesandsolarirradiances,alongwithancillarydataareusedasinputstothealgorithms.
ColumndensityvalueswillbearchivedforeachOMIpixellocationandwillconstitutethebasiclevel2outputsfromthealgorithms.
Tropospherecolumnswillalsobeincludedasoutputs,wherevertheyaredeterminedtocontributesignificantlytothetotal.
OMImakesnadirmeasurementsoftheEarth'sbackscatteredultravioletradiationatspectralresolutionof~0.
42nmintheUV-1channel(whichmaybeusedforpartoftheSO28ATBD-OMI-04retrieval),~0.
45nmintheUV-2channel,thespectralregionwhereHCHO,BrO,andOClOaremeasuredandthebulkoftheSO2informationisobtained,and~0.
63nminthevisiblechannel,whereNO2ismeasured.
OMIspatialresolutionisselectableamongaglobalmodeandspectralandspatialzoom-inmodes.
Inglobalmode,fortheUV-2andvisiblechannels,thespatialresolutionis13kmalong-track·24kmacross-trackatnadir.
Thefullswathwidthis2594km.
Inbothzoom-inmodes,thespatialresolutionis13·13km2atnadir.
Inthespectralzoom-inmode,thewavelengthrangeisreducedto306-364nmplus350-432nm.
Forthespatialzoom-inmodeallwavelengthsarepresent,buttheswathwidthisreducedto725km.
Thesmallergroundpixelsinthezoom-inmodeswillgreatlyimproveourabilitytomeasuresmalleratmosphericfeatures,suchasenhancedtroposphericNO2andHCHOanddetailsofSO2sources.
Mostimportantlyforgeophysics,theOMIinstrumentwillhavecompletespatialcoverage,fortheglobalandthespectralzoom-inmodes,withreasonably-sizedgroundpixels.
Incontrast,thestandardGOMEswathprovides40kmalong-track·320kmacross-trackspatialresolution,with3-dayglobalcoverage[EuropeanSpaceAgency,1995].
TracegasslantcolumndensitiesaredeterminedbymeasuringthevibrationalstructureinmolecularelectronicbandsthatoccurwithintheOMIwavelengthrange:NO22B1←%X2A1;HCHO1A″←%X1A1;SO21B1←%X1A1;BrOA2Π3/2←X2Π3/2;andOClO2A2←%X2B1.
ExtensiverotationalstructureispresentinthespectraofHCHO,SO2,andBrO,butitisnotresolvableattheOMIresolution.
TheOClObandexhibitsagradualonsetofbroadeningbypredissociationintheOMIwavelengthregion,althoughitisnotresolvedbyOMIinthewavelengthwindowwheretheabsorptionissufficientlystrongtobemeasured.
Tracegasretrievalalgorithmsarebasicallyoftwotypes,distinguishedbythosewhichcannormallybeanalyzedassumingthatthescatteringandbroadband(i.
e.
,O3Hartleyband)interferingabsorptioncontributionstothemeasuredradiancemaybeapproximatedasconstantoverthespectralfittingwindow,sothattheslantcolumnfittingandtheerectiontodetermineverticalcolumnabundancesmaybeseparated(NO2,HCHO,BrO,andOClO)andthatwhichmustoftentakeintoaccountthevariationofthesecontributionsoverthespectralfittingwindowduringtheretrievalprocess(SO2).
Foreachofthegasesmeasured,thedataproductisdeterminedfromawavelengthwindowthatisoptimizedfortheparticulargas.
ExperiencewithfittingGOMEspectra,aswellaswithotheratmosphericfieldmeasurementsofspectra,isthatattemptstofitwiderspectralwindowscomprehensivelyformultiplespeciesdeterminationsoftenleadstoinferiorresults.
1.
1.
NO2,HCHO,BrO,andOClOTheintrinsicatmosphericspectrummeasuredbyOMIcanbeapproximatedasIAEeeNNssnnλλσλσλ=+11LHigherOrderTerms,[1-1]whereIisthebackscatteredradiance,Aisthealbedo(includingscatteringcontributions),EistheFraunhofersourcespectrum(irradiance),theNsiarecolumnabundancesoverthemeasurementpathline-of-sight(``slantcolumnabundances''),theσiareabsorptioncrosssections,andthehigher-orderterms(hereafter"HOT")maybemodeledasapolynomialtoaccountforwavelengthdependenceofthealbedo.
Thefirststepineachofthetracegasalgorithmsforthesefourmoleculesistodeterminetheslantcolumnabundances,forthedesiredproductaswellasfortheinterferingspecies.
Thismaybeaccomplishedbyseveralmethods,includingdirectfittingofIbysynthesizingitbeginningwithE,fittingtothelogarithmofI/E,andfittingtoahigh-passfilteredversionofthelogarithmofI/E:ATBD-OMI-049H[ln(I/E)]=-H(-c1σ1)H(-cnσn)+HOT[1-2]whereHdenotesthe(optional)high-passfiltering.
Rayleighscatteringisthemajorcontributiontotheradiativetransferproblemthatmustbeaddressedtodetermineverticalcolumnabundancesfromslantcolumnabundances,asdescribedbelow.
Theanalysistodetermineslantcolumnabundancesmustalsotakeintoaccountthefactthat,forthewavelengthrangeofOMItracegasmeasurements,4%oftheRayleighscatteringisinelastic(the"Ringeffect")andisthusRamanscatteredbythepredominantlyN2andO2molecularscatterers.
Thisresultsinanadditionalspectralcomponent,which,tothelowestapproximation,istheconvolutionoftheFraunhoferandrotationalRamanspectra.
(Theeffectsofhigher-ordercorrectionsmustbeevaluatedindetailduringalgorithmdevelopmentfortheindividualtracegases.
TheseincludemultiplescatteringandthemodificationoftheFraunhofersourcespectrumbyabsorptionofatmosphericgasesbeforeRamanscatteringhasoccurred.
)Ringeffectcorrectionsareincludedasadditionalfittingterms:IAEeeccNNRRRRssnnλλσσσλσλ=+++111122LHOT[1-3]H[ln(I/E)]=-Ns1H(σ1)NsnH(σn)+cR1H(σR1/F)+cR2H(σR2/F)+HOT,[1-4]wheretwoRingeffectcorrectiontermsσR1,σR2areshown(eitheroneortwotermswillgenerallybeincluded,thesecondtermbeingacorrectionforabsorptionbyatmosphericgases-usuallyO3-beforetheRayleighscatteringhasoccurred).
Thehigher-ordertermsinthiscaseincludefurthertermsintheexpansionofln[(I+cRσR)/E].
ExperiencewiththefittingofGOMEbackscatteredUV/visiblespectrahasshownthatitmaybedesirabletoincludeimprovedwavelengthcalibrationdirectlyinfittingalgorithms.
ThespectralcorrelationbetweenI(λ)andE(λ)intheindividualfittingwindowsissubstantiallyimprovedoverthewavelengthcalibrationobtainedinthelevel0-1processing[CasparandChance,1997],whichleadstosignificantimprovementinthetracegasfitting.
Additionally,thealgorithmsincludemodelingoftheinstrumenttransferfunctionandfittingtolow-frequencyclosuretermstoaccountforthewavelengthdependenceofthealbedoaswellasinstrumentimperfections[LangleyandAbbot,1900].
Theselectionoftheoptimalfitwindowmustmaximizethesensitivityoftheretrievaltothetargetabsorptionsignatures,whileminimizingerrorsfromgeophysicalandinstrument-relatedspectralfeatures.
Asummaryoftheimportantconsiderationsinselectingawavelengthrangeforthewindowinclude:Locatingregionsofmaximumamplitudeinthestructuresofthecrosssectionsforthetargetgas;Avoidingoverlapwithstrongatmosphericspectralfeaturesfrominterferingspecies,includingpartsoftheRamanscattering(Ring)spectrum;Avoidingregionscontainingspectralstructuresofinstrumentalorigin;Choosingaswideawindowaspossibletomaximizethenumberofsamplingpoints;Selectingaregionofthetargetspectrumthatminimizessensitivityofthemeasurementtothetemperature(especiallytrueforNO2).
Thelevel2tracegasdataproductsareallverticalcolumnabundances,exceptincaseofOClO,whichisexpectedtooccuronlyathighsolarzenithanglesandiscurrentlyonlyenvisagedasaslantcolumndataproduct.
Theamplitudesofthespectralstructuresmeasured10ATBD-OMI-04alongtheOMIline-of-sightdeterminethefittingtoobtainslantcolumnabundances.
Thesemustbecorrectedtotakeintoaccountmeasurementgeometry,thedilutioninthespectracausedbyRayleighscatteringoftheFraunhofersourcespectrum(formostmeasurementgeometries,amajoreffectofRayleighscatteringistoreducetheeffectpathofthebackscatteredlightmeasuredatthesatelliteincomparisontothegeometricpath,althoughinafewcircumstancesRayleighscatteringmayincreasetheeffectivepathandthusamplifythesignal)andobscurationbycloudsandaerosols(insomeinstancesaerosolscatteringmayalsoamplifythespectralsignals),inordertoderiveverticalcolumnabundances.
Thiswillbeaccomplishedbytheuseofpre-calculatedairmassfactors(Ms).
MforgasiisdefinedasM=Nsi/Nvi[1-5]whereNsiistheslantcolumnabundanceandNviistheverticalcolumnabundance.
VerticalcolumnabundancesarethensimplydeterminedasNvi=Nsi/Mi.
[1-6]ThealgorithmsforNO2,HCHO,BrO,andOClOassumeopticallythinabsorptionandRayleighscattering,whichdoesnotvarysignificantlyoverthefittingwindow.
Airmassfactorsarecalculatedusingaradiativetransfermodel[Dave,1965;DeHaanetal.
,1987;Stammesetal.
,1989;Stammes,2001;Spurretal.
,2001].
Foragivenfitwindow,theyarefunctionsofviewingzenithangle,solarzenithangle,surfacealbedo,cloudparameters,andtheconstituentprofile.
Someoftheseparametersincludelatitudinalandseasonaleffects,whichmustbeconsideredduringtheOMIvalidation.
CloudheightandcloudfractionwillbeOMIdataproductsfromotherinvestigations.
Msforstratosphericcomponentsofthegasesconsideredherearewelldeterminedbymeasurementgeometryandclimatologicalverticaldistributionprofiles.
Msforthetroposphericcomponentsaremoreproblematicfortworeasons:(1)Thedistributionsaresubstantiallymorevariable;(2)TheRayleighscatteringcontributiontotheMisrelativelymoreimportantbecauseofthegreateratmosphericdensity.
Fortroposphericmeasurements,higher-levelproductswillrequirefurtherprocessingtotakeadditionalgeophysicalknowledgeaboutmeasureddistributionsintoaccount.
AnexampleofthisisprocessingoftheNO2level2dataproductstoremovethestratosphericoverburden,leavingthetroposphericresidual,whichmaythenbefurtherprocessedtoaccountforthemodeledshapeofthetroposphericprofileanditseffectuponthetroposphericcomponentoftheM.
Suchprocessingwillbeimportantformanyofthetracegasmeasurements,includingthetroposphericcomponentsofNO2andBrO,andallSO2andHCHOmeasurements,sincethesearelikelytobeprimarilytroposphericattheOMImeasurementsensitivity.
Airmassfactorsdependonanumberofparametersthatareinputtotheradiativetransfer.
Theycanbedividedintoparametersthatareassumedapriori:Viewinggeometry,Surfacealbedo,Cloudalbedo,Terrainheightorterrainpressure,andparametersthatmustbeestimatedfromchemicalorphysicalknowledge,e.
g.
,climatologies,predictionsbyassimilationmodels,andotherOMImeasurements:ATBD-OMI-0411AltitudedistributionsinthetropospherehavelargeinfluencesontheairmassfactorsforNO2andHCHO;Cloudfraction;Cloudheight;Aerosolopticalthickness;Ozoneprofile.
Sensitivitystudieshaveshownthatincludingtheozoneslantcolumndensityasafitparameterisrequired,butthattheozonedistributioninducesanegligibleeffectontheretrievedNO2columnsdensities.
Cloudshavevaryingeffectsonairmassfactors.
Firstly,cloudsobscuregaslocatedbelowthecloud,andthusdecreasemeasurementsensitivity.
Secondly,cloudsgenerallyincreasethesensitivitytogasaboveclouds,duetotherelativelyhighcloudalbedo.
Inouralgorithms,weassumethatcloudscanbeapproximatedasopaque,Lambertiansurfaces.
Thincloudswillsimilarlybetreatedasopaquesurfacescoveringonlyasmallpartofthepixel,i.
e.
,theeffectivecloudfractionwillbesmall.
Studiesintothevalidityofthisassumptionfordifferentcloudtypes,cloudopticalthicknessesetc.
arecurrentlybeingperformed.
Allcloudsinagivenpixelwillbecharacterizedbythreeparameters:acloud-topheight,ageometricalcloudfraction,andacloudalbedo.
Multiplecloudlevelsarenotconsidered.
Tocalculatetheairmassfactorforapartlycloudyscene,calculationsforfullycloudedandfullyclearpixelsaremerged.
Undertheindependentpixelassumption,theairmassfactorforpartlycloudyconditionsisgivenby:M=wMcloud+(1–w)Mclear[1-7]whereMcloudandMclearrepresenttheairmassfactorsforcompletelycloudyandclearscenes,respectively,andwistheflux-weightedcloudfraction.
WedefinewasclearcloudcloudIcIcIcw)1(+=[1-8]wherecisthegeometricalcloudfraction,andIcloudandIcleararetherespectiveradiancesofcloudyandclearscenes.
Underbackgroundconditions,aerosolconcentrationsareexpectedtobesmall,andweassumethattheinfluenceofaerosolscanbeneglected.
Wheretroposphericgasconcentrationsareenhanced,aerosolconcentrationsareoftenenhancedaswell,andtheaerosol'seffectontheairmassfactormaybesignificant.
However,highaerosolconcentrationsaretypicallyassociatedwithother,largererrorsources.
Therefore,ourpreliminaryapproachwillbetoneglecttheinfluenceofaerosolsinthederivationofverticalcolumndensities.
1.
2.
SO2TheSO2algorithmissufficientlydifferentfromotheropticallythintracegasalgorithms(NO2,HCHO,BrO,OClO)thatitneedstobeaddressedseparately.
AretrievalalgorithmforSO2mustcontendwiththetransientnatureanddramaticallydifferentscalesofatmosphericSO2emissions.
Theimportantcharacteristicsare:1)adynamicrangefrom0.
5matm-cmto>1000matm-cm(1.
4·1016to2.
7·1019cm-2),makingitthedominantabsorberinvolcanicclouds(whichcanbeopticallythick),2)thelocationintheboundarylayeroftheair-pollutionSO2,and3)volcanicashoraerosolinterference.
12ATBD-OMI-04Theprocessingalgorithmwillproducequantitativedataonlowlevelemissions,whicharewidespreadacrossthenorthernhemisphereandgenerallyfreeofash,andsemi-quantitativeinformationonvolcanicclouds,whicharelocalinsizebutcandriftoverglobalscales.
Asnoconstraintscanbeplacedonthegeographiclocationofeithertypeofsource,alltheOMIdatawillbeprocessed.
TheTOMSSO2algorithmwasdevelopedforsulfurdioxideandozonediscriminationinthenearUVspectralregion[Kruegeretal.
,1995;Kruegeretal.
,2000].
Thealgorithmusesanatmosphericopticalmodelwhichcharacterizestheabsorptionandscatteringprocessesasfourindependentpiecesofinformation;twoabsorptiontermsforozoneandsulfurdioxide,andtwoscatteringtermsforalbedoandwavelengthdependence.
ThefourTOMSradiancemeasurementsrequiredforclosureareselectedfromthesixTOMSbands.
Radiativetransfertablesareusedtodeterminetheopticalpathforeachofthewavelengths.
ThismodelhasprovedveryeffectivewithTOMSdataeventhoughthechannelwavelengthsarenotwellchosenforthispurpose.
However,themeasurementnoiselevelwastoohightodeterminepassivevolcanicemissionsandairpollution.
ThisproblemwasnotpresentintheGOMEdataandtheDOASmethodwasabletoretrieveanthropogenicSO2emissionsunderwintertimeconditions[EisingerandBurrows,1998].
TheOMISO2algorithmisbasedontheTOMSsulfurdioxideandozonemethodsincombinationwiththeDOASmethod.
Wemakeuseof:1)additionalspectralinformationofOMI,2)aprioriinformation,and3)otherOMIproducts(ozone,aerosol,clouds,seeotherATBDvolumes).
SpecialemphasisisplacedonretrievalofsmallamountsoflowertroposphericSO2becauseofthechallengeoftheobservingconditionsandtheimportanceofadatabaseonanthropogenicsulfurdioxideandpassivevolcanicemissions.
Maximumlikelihoodtechniquesareemployedtomakeuseofallinformation.
Itisimportanttonotethatozoneandsulfurdioxidemustbedeterminedsimultaneouslyforvolcaniceruptioncloudsbecauseofthealmostcompleteoverlapoftheirabsorptionbands,comparableabsorberamounts,andthelikelihoodthatthetotalozonecolumnwillbemodifiedbythevolcaniccloud.
Similarly,theopticaldepthofashinthecloudneedstobemeasuredandspecifiedforproductionofradiativetransfertables.
Theopticalpropertiesofashvarybetweenvolcanoesandbetweeneruptionsandrequirelaboratoryanalysisofashsamplesfordetermination.
Thus,off-lineprocessingofsubsetsofdataisrequiredtohandlethegreatercomplexityofvolcanicclouds.
ParallelwavelengthsamplingwithOMIproducesagreaterS/NthantheserialsamplingTOMS.
Also,OMImeasuresthefullUVspectrumallowingselectionofoptimumwavelengthforaTOMS-likeretrieval.
ThesefactorscanproduceanOMISO2retrievalnoiselevelthatisafactorof10lowerthanTOMS(~4DU)andcomparabletoGOMESO2[EisingerandBurrows,1998].
Inaddition,thesmallerOMIFOVwillgreatlydecreasetheminimaldetectableSO2fluxcomparedwiththatfromGOME.
ATBD-OMI-04131.
3.
ReferencesBurrows,J.
P.
,K.
V.
Chance,A.
P.
H.
Goede,R.
Guzzi,B.
J.
Kerridge,C.
Muller,D.
Perner,U.
Platt,J.
-P.
Pommereau,W.
Schneider,R.
J.
Spurr,andH.
vanderWoerd,GlobalOzoneMonitoringExperimentInterimScienceReport,ed.
T.
D.
GuyenneandC.
Readings,ReportESASP-1151,ESAPublicationsDivision,ESTEC,Noordwijk,TheNetherlands,ISBN92-9092-041-6,1993.
Burrows,J.
P.
,M.
Weber,M.
Buchwitz,V.
V.
Rosanov,A.
Ladstatter,A.
Weissenmayer,A.
Richter,R.
DeBeek,R.
Hoogen,K.
Bramstedt,andK.
U.
Eichmann,TheGlobalOzoneMonitoringExperiment(GOME):Missionconceptandfirstscientificresults,J.
Atmos.
Sci.
,56,151-175,1999.
Caspar,C.
,andK.
Chance,GOMEwavelengthcalibrationusingsolarandatmosphericspectra,Proc.
ThirdERSSymposiumonSpaceattheServiceofourEnvironment,Ed.
T.
-D.
GuyenneandD.
Danesy,EuropeanSpaceAgencypublicationSP-414,ISBN92-9092-656-2,1997.
Chance,K.
V.
,J.
P.
Burrows,andW.
Schneider,RetrievalandmoleculesensitivitystudiesfortheGlobalOzoneMonitoringExperimentandtheSCanningImagingAbsorptionspectroMeterforAtmosphericCHartographY,Proc.
S.
P.
I.
E.
,RemoteSensingofAtmosphericChemistry,1491,151-165,1991.
Dave,J.
V.
Multiplescatteringinanon-homogeneous,Rayleighatmosphere,J.
Atmos.
Sci.
22,273-279,1965.
DeHaan,J.
F.
,P.
B.
Bosma,andJ.
W.
Hovenier,Theaddingmethodformultiplescatteringcalculationsofpolarizedlight,Astron.
Astrophys.
183,371-391,1987.
Eisinger,M.
,andJ.
P.
Burrows,TroposphericsulfurdioxideobservedbytheERS-2GOMEinstrument,Geophys.
Res.
Lett.
25,4177-4180,1998.
EuropeanSpaceAgency,TheGOMEUsersManual,ed.
F.
Bednarz,EuropeanSpaceAgencyPublicationSP-1182,ESAPublicationsDivision,ESTEC,Noordwijk,TheNetherlands,ISBN-92-9092-327-x,1995.
Krueger,A.
J.
,L.
S.
Walter,P.
K.
Bhartia,C.
C.
Schnetzler,N.
A.
Krotkov,I.
Sprod,andG.
J.
S.
Bluth,VolcanicsulfurdioxidemeasurementsfromtheTotalOzoneMappingSpectrometerinstruments,J.
Geophys.
Res.
100,14,057-14,076,1995.
Krueger,A.
J.
,S.
J.
Schaefer,N.
Krotkov,G.
Bluth,andS.
Barker,UltravioletRemoteSensingofVolcanicEmissions,inRemoteSensingofActiveVolcanism,ed.
P.
MouginisMark,J.
A.
Crisp,andJ.
H.
Fink,GeophysicalMonograph116,AmericanGeophysicalUnion,Washington,DC,2000.
Langley,S.
P.
,andC.
G.
Abbot,AnnalsoftheAstrophysicalObservatoryoftheSmithsonianInstitution,Vol.
1,pp.
69-75,1900.
Spurr,R.
J.
D.
,T.
P.
Kurosu,andK.
Chance,Alinearizeddiscreteordinateradiativetransfermodelforatmosphericremotesensingretrieval,J.
Quant.
Spectrosc.
Radiat.
Transfer68,689-735,2001.
Stammes,P.
,J.
F.
deHaan,andJ.
W.
Hovenier,Thepolarizedinternalradiationfieldofaplanetaryatmosphere,Astron.
Astrophys.
225,239-259,1989.
Stammes,P.
,SpectralradiancemodellingintheUV-VisibleRange,toappearinIRS2000:CurrentproblemsinAtmosphericRadiation,Eds.
W.
L.
SmithandY.
M.
Timofeyev,A.
DeepakPubl.
,Hampton(VA),2001.
14ATBD-OMI-04ATBD-OMI-04152.
NO2FolkertBoersma1,EricBucsela2,EllenBrinksma1,JamesF.
Gleason21KNMI,2NASA-GSFC2.
1.
IntroductionNitrogendioxide,NO2,isacriticaltracegasintheatmospherebecauseofitsroleinthephotochemistryofozoneinthestratosphereandtroposphere.
NO2isimportantfordirectdestructionofodd-oxygeninthemiddlestratosphere(Reactions2.
1-2.
3).
NO2connectsthehydrogenandchlorinechemicalfamiliesthroughtheproductionofreservoirspecies(Reactions2.
4-2.
7),whicharecriticaltoourunderstandingoflowerstratosphericozonephotochemistry.
NO+O3→NO2+O2(2-1)NO2+O→NO+O2(2-2)Net:O3+O→2O2(2-3)NO2+OH+M→HNO3+M(2-4)NO2+HO2+M→HNO4+M(2-5)NO3+NO2+M→N2O5+M(2-6)ClO+NO2+M→ClONO2+M(2-7)TroposphericozoneisformedafterthephotolysisofNO2inthetroposphericoxidationofhydrocarbons.
Reactions2.
8-2.
12illustrate,usingCO,howozonecanbeproducedinthetroposphere.
CO+OH+(O2)→HO2+CO2(2-8)HO2+NO→NO2+OH(2-9)NO2+hν→NO+O(hν.
WewillsubstituteforunpvN.
2.
A.
5.
CloudlesscaseTheairmassfactorforacloudlessscenecanbeexpressedas:pvvppvunpvpvvpsunpsvsNNMNMNNNNNNNM+>=zcloudbethealtitudeoftheLambertiancloud.
Afractionkpofthepollutedprofilewillbesituatedabovetheclouds,i.
e.
∫∞=zcloudpnpdzznk)((10)And,definedanalogously,afractionkunpoftheunpollutedprofilewillbelocatedabovethecloud.
Supposeaparticularpixelisnotpolluted.
ThenwehaveunpvunpunpvunpcloudunpmeasNMwNkzwXN)1()(+=(11)wherewisthereflectance-weightedcloudfraction(seeChapter5fordefinition),)(cloudunpzXisanairmassfactorforthecloudypixel,definedasunpzcloudunpncloudcloudunpkdzznzmzX∫∞=)()()(,(12)whereweaddedthesuperscriptcloudtomtoindicatethatitdepends(slightly)onthesurfacealbedooftheLambertiancloud.
)(cloudunpzXcanbeinterpretedastheslantcolumndensityobservedoveracompletelycloudedpixel,dividedbythenormalizedverticalcolumndensitylocatedabovethecloud.
However,tosimplifytheequation,wewilldefinetheairmassfactordifferently,namelyas:∫∞==zcloudunpncloudcloudunpunpcloudunpdzznzmzXkzM)()()()((13)whichcanbeinterpretedasthemeasuredslantcolumndensitydividedbythetotalverticalcolumndensity.
Equation(13)willbesubstitutedintoEq.
(11)FromEq.
(11)itfollowsthattheverticalcolumncanbecalculatedusingunpcloudunpmeassunpvMwzwMNN)1()(+=(14)Ifwedealwithapollutedpixel,wehavetoextendEq.
(11)tothefollowing:))(1())()((pvpunpvunppvcloudpunpvcloudunpmeassNMNMwNzMNzMwN+++=(15)IfwesubstituteunpvNby,andsolveforpvN,thisyields:32ATBD-OMI-04pLpunpLunpmeassMwzwMvNMwzwMNpvN)1()())1()((+>zcloudpnzcloudpnzcloudpnpzMdzznzmzTdzznzmdzznzmzTT∫∫∫∞∞∞==(17)forthepollutedpartoftheprofile.
SimilardefinitionstoEqs.
(9)and(17)applytotheunpollutedpartoftheprofile(airmassfactorsinthenumerator),butwewillassumethatthetemperatureusedintheDOASfitisdoesnotdiffermuchfromtheunpollutedprofileeffectivetemperature,andthusthatatemperaturecorrectionfortheunpollutedprofileisnotnecessary.
ATBD-OMI-04332.
B.
AppendixB:ErroranalysisdetailsAuthors:FolkertBoersma,EllenBrinksma,KNMIEricBucsela,JamesF.
Gleason,NASA-GSFC2.
B.
1.
SlantcolumndensityErrorsintheinversemodelcomefromapriorierrorsandinstrumenterrors.
TheaccuracyoftheNO2absorptioncross-sectionspectrumandtheassumptionsonthetemperatureoftheNO2intheatmospherearetypicalapriorierrors.
Measurementnoiseandspectralcalibrationareinstrumenterrors.
(1)NO2crosssectionspectrumTheaccuracyoftheNO2crosssectionisestimatedtobe2%[Vandaeleetal.
,1998].
Sincethis2%ismostlyduetooffseterrors,theerrorintheNO2slantcolumndensityisestimatedtobelessthan2%.
(2)TemperaturedependenceInfittingNO2toareflectivityspectrum,a220KeffectivetemperatureisassumedforNO2.
Thefitwindowisoptimizedsuchthatthetemperaturedependenceisminimal.
Forunpollutedconditions,a10Kperturbationoftheeffectivetemperatureresultsina3.
7%fiterror(ascanbeseenfromFigure4).
Inpollutedsituations,thetemperaturecorrectionasdescribedinChapter5willreducetheerrorduetoassuminganeffectivetemperaturetoapproximately6%intheNO2slantcolumndensity.
(3)MeasurementnoiseTheradiometricsignal-to-noiseratioofOMIinthe400–480nmrangeisestimatedtobeoftheorder2800forfourco-addedpixels[DeVriesetal.
,2000].
Testswiththeproposed405–465nmfitwindowshowedthattheerrorintheslantcolumndensityduetonoiseislessthan4%fornominalconditionsandlessthan2%forsituationswithenhancedNO2levels.
(4)SpectralcalibrationandstabilityDOASissensitivetowavelengthcalibrationerrors.
OMIreflectivityspectrawillbecalibratedfromcross-correlatingtheFraunhoferlinesinthesolarirradiancewiththeEarthradiancespectra.
Thespectralcalibrationisestimatedtobebetterthan0.
0021nm[Leveltetal.
,2000].
TheerrorintheNO2slantcolumndensityduetoaspectralcalibrationerrorof0.
0021nmis0.
5%fornominalconditionsand0.
3%forenhancedNO2conditions.
BecauseofinstrumenttemperaturedifferencesoveranOMIorbit,thewavelengthgridoftheEarth'sradiancespectrawillshiftwithrespecttotheirradiancespectra.
Toaccountforthiseffect,irradiancespectrawillbesplineinterpolatedtotheradiancewavelengthgrid.
Thespectralstabilityoveranorbitisestimatedtobebetterthan0.
011nm[Leveltetal.
,2000].
Interpolatingthesolarirradiancespectrumtoa0.
011nm(1/20thofaspectralpixel)shiftedwavelengthgridcausesirradiancelevelstochangebylessthan2·10-5.
Thiserrorhasanegligibleeffectonslantcolumndensityretrieval.
34ATBD-OMI-042.
B.
2.
AirMassFactorThemostimportanterrorsintheairmassfactorcomefromaprioriassumptionsregardingthestateoftheatmosphere.
TheairmassfactorisespeciallysensitivetotheshapeoftheNO2profileandthevalueofthesurfacealbedoinsituationswithenhancedNO2levels.
OthererrorsareduetoassumingthatcloudscanberepresentedasLambertsurfaceswithanalbedoof0.
8andassumptionsregardingtheaerosolmodelthatisusedforsituationswithenhancedNO2levels.
(1)NO2profileshapeThesensitivityoftheairmassfactorontheNO2profileshapeisparticularlyimportantinthecaseofenhancedNO2levels,whensurfaceemissionscangiveriselargevariabilityinthetroposphericNO2burden[Houwelingetal.
,1998].
TheNO2profileismuchbetterknownforunpollutedsituations,whenseasonalvariationsdominatethestratosphericNO2columnandtheuncertaintyintheunpollutedairmassfactorduetoassuminganapproximatelygeometricalairmassfactorisestimatedtobe1%.
ThelargeuncertaintyinthetroposphericNO2profileshapehasbeeninvestigatedbyanalyzingayearlyaveragetroposphericNO2profilethatwasgeneratedfrom12monthlyprofiles(1997)from20regionswithemissionsfromindustrialsources.
ThelargevariabilityinboundarylayerNO2concentrationsaswellastheuncertaintywhetherTM3isrepresentingtheboundarylayerinaproperwayisresultinginanestimateoftheuncertaintyinthetroposphericairmassfactorofapproximately20%forsituationswithenhancedNO2levels.
(2)SurfacealbedoTheairmassfactorissensitivetotheassumedvalueofthesurfacealbedo.
UsingDAKsimulations,thissensitivitywasassessed.
InbothnominalandenhancedNO2cases,thesensitivityislargerfordarksurfacesthanforbrightsurfaces,althoughforthenominalcase,thisdifferenceissmall.
Weassumethatthevariabilitywithinagridboxofthesurfacealbedodatabaseis±0.
03[Koelemeijeretal.
,2001].
Fornominalcases,asurfacealbedochangeof±0.
03toarealisticcontinentalsurfacealbedoof0.
05yieldsanairmassfactorchangeof±0.
5%.
ForsituationswithenhancedNO2levels,thesensitivitytothesurfacealbedoismuchlargerandmuchmoredependentonthesurfacealbedo.
Inthiscase,analbedochangeof±0.
03toatypicalsurfacealbedoof0.
05yieldsanairmassfactorchangeof22%,forabrightsurface(albedo0.
2)thischangeis6%.
SinceenhancedNO2concentrationsareoftenobservedoverland,theaccuracyduetosurfacealbedouncertaintiesisestimatedtobe20%.
(3)CloudalbedoAssumingthatcloudscanberepresentedasLambertiansurfaces,thesensitivitytochangesincloudalbedowastestedbyusingthesamesimulationprocedureasdescribedabove.
Fornominalconditions,thetypicaluncertaintyoftheairmassfactorinducedbyanalbedochangeof±0.
1,is1.
5%.
ForsituationswithenhancedNO2levelsandtypicalcloudalbedoes(0.
6to0.
8)thisis4%.
(4)CloudfractionErrorsinthecloudfractioninduceaninaccuracyintheairmassfactorviatheweightingfactordescribedinequations(8)and(9).
Theaccuracyofthecloudfractionisexpectedtobe0.
1[Leveltetal.
,2000].
Fornominalsituations,anerrorinthecloudfractionof0.
1resultsinanerrorintheairmassfactorof0.
3%.
ForsituationswithenhancedNO2levels,theerrordependsverymuchonthecloudfractionitself.
Forcloudfractionsbelowabout0.
3,whenitisstillpos-sibletodetectNv,p,errorsinthecloudfractionof0.
1resultinerrorsintheairmassfactor<8%.
ATBD-OMI-0435(5)CloudpressureTherequiredaccuracyofthecloudpressureis100hPa[Leveltetal.
,2000].
Fornominalcases,thisinducesaninaccuracyontheairmassfactorof0.
3%attheground,through1%at200hPa.
ForenhancedNO2levels,thelargestsensitivityisfoundwhenthecloudislocatedwithintheregionofhighNO2concentrations.
A100hPainaccuracyonthecloudpressureinducesa50%errorontheairmassfactorforcloudsat900hPa,a0.
6%errorat500hPa,anda1.
2%errorat200hPa.
(6)AerosoleffectRadiativetransfercalculationstoconstructairmassfactorlook-uptablesmakeanassumptionaboutscatteringandabsorptionbyaerosolparticles.
Forbackgroundsituations,whenmostoftheNO2isabovetroposphericaerosollayers,airmassfactorsareestimatedtobeaccurateto0.
5%.
InsituationsofenhancedNO2levels,aerosolconcentrationsareusuallyalsoenhanced(smog,biomassburning)andassumptionsonanaerosolmodelcangiverisetoerrorsintheairmassfactorofabout15%.
(7)EstimateoftheunpollutedcolumnTheuncertaintyindeterminingNv,unpfromthesmoothingandgriddingprocedureisestimatedtobe5%.
36ATBD-OMI-042.
CAppendixC:DataProductsTableandDataProductDependenciesAuthors:EricBucsela,JamesF.
Gleason,NASA-GSFCFolkertBoersma,EllenBrinksma,KNMIPrimaryProduct:TotalNO2verticalcolumndensity+errorSecondaryProduct:TroposphericNO2verticalcolumndensity+errorDiagnosticDataSlantColumnNO2+errorTroposphericSlantColumnNO2+errorCloudfraction&CloudheightClearAMF&CloudyAMFAerosolCorrectionCoefficientsforallbasisfunctionsGeo-locationInformationSolarandViewingGeometryDataDependenciesDataSourceRequiredCloudfraction&CloudheightOMICloudGroupYesSolarandViewingGeometryOMIInstrumentTeamYesSurfaceAlbedoKoelemeijerdatabaseYesAerosolDataOMIAerosolGroupNoTemperatureProfileNCEPNoHIRDLSNO2ProfilesAuraHIRDLSGroup/DAACNoATBD-OMI-04373.
HCHOK.
Chance,T.
P.
Kurosu,andL.
S.
RothmanSmithsonianAstrophysicalObservatoryCambridge,MA,USAHCHOisaprincipalintermediateintheoxidationofhydrocarbonsinthetroposphere,providinganimportantindicatorofbiogenicactivity.
IntheremotemarinetroposphereitmayserveasausefulproxyfortroposphericOH.
HCHOwasproposedformeasurementfromspacebytheGOMEinstrument[Chanceetal.
,1991]andwasfirstmeasuredfromspacebyGOME,inthe337-359nmrange[Thomasetal.
,1998].
HCHOisprominentintheSoutheasternU.
S.
insummertime(fromisopreneoxidation)[Chanceetal.
,2000]andisalsoaprominentproductofbiomassburning[Thomasetal.
,1998].
Figure3.
1ashowsverticalcolumnabundancesofHCHOderivedfromGOMEmeasurementsoverNorthAmericaforJuly1996;Figure3.
1bshowsthemodeledresultfromtheGEOS-CHEM3-dimensionaltroposphericchemistryandtransportmodel.
TheHCHOinthiscaseisprimarilyduetoisopreneoxidation.
Figure3.
2showsinmoredetailtheresultsfromasingleGOMEorbitexhibitinghighisopreneproductionoverthesoutheasternU.
S.
Figures3.
3aand3.
3bshowenhancedHCHOasmeasuredbyGOMEfrombiomassburningepisodesoverthenorthernAmazonbasin,andoverIndonesiaduringtheintensefiresof1997,respectively.
OMIwillmeasureHCHOathigherspatialresolutionthanGOMEandwithbettertemporalcoverage,allowingforimprovedcharacterizationofsourcesandtransport.
GOMEhasdifficultyinmeasuringHCHOatconcentrationstypicaloftheremotemarinefreetroposphereduetosystematiceffects,whichmaybeassociatedwiththeinstrument.
OMImayovercometheseandthussupplybetterdeterminationoftheglobalclimatologyoffreetroposphericHCHO.
AbsorptionsforHCHOarequitesmall(substantiallylessthan1%inmostcases),sothattheyareopticallythintoahighdegreeofaccuracyandsothatinterferencesfromothercauses(e.
g.
,O3absorptionandtheRingeffect)mustbeaccountedforveryprecisely.
ThefittingforHCHOincludestwomajorsteps:(1)thefittingofaselectedwavelengthwindowofspectrumtodeterminetheslantcolumndensity,Nsi,foraparticularspeciesi,and(2)thedeterminationofanappropriateairmassfactor,Mi,toconvertNsitoaverticalcolumndensity,Nvi:M=Nsi/Nvi,[3-1]whereMisafunctionofviewinggeometry,geophysicalcondition(albedo,cloudcoverage),andtheverticaldistributionofthegas.
Generally,gaslocatedatloweraltitudecontributeslesstothesatellite-measuredabsorptionspectrumsinceRayleighscatteringdiscriminatesagainstitsviewing[Palmeretal.
,2001].
Algorithmsaredesignedassumingsequentialprocessingoflevel1dataproducts,includingspectra,sothathigher-levelprocessingwillberequiredinsomecasestofullyexploitthemeasurements.
Thefirststepisaccomplishedbyaspectralfittingprocedure,whichwillbeoptimizedduringthecommissioningphaseoftheOMIinstrument,andwherevariousoptionsdescribedbelowareevaluated.
Thesecondstepismorecomplex:HCHOispredominantlyatroposphericspecies,sothatdeterminationoftheverticalcolumnabundancesrequiresmodelingoftheverticalprofileshapeinthetroposphereinordertodetermineMsappropriatetospecificmeasurementconditions,asdemonstratedforGOMEmeasurements[Chanceetal.
,2000;Palmeretal.
,2001].
38ATBD-OMI-043.
1.
SlantcolumnmeasurementsThedeterminationofslantcolumnabundances,Ns,isaccomplishedbyfittingthemeasuredradianceI,beginningwiththemeasuredirradianceE,molecularabsorptioncrosssections,correctionfortheRingeffect,effectivealbedo(whichincludesthecontributionfromRayleighscatteringforthesemolecules,asdiscussedinAirmassfactorsandverticalcolumnabundances,below),andalow-orderpolynomialforclosure.
ThislattertermaccountsforsmallFigure3.
1.
(a)VerticalcolumnabundancesofHCHOderivedfromGOMEmeasurementsoverNorthAmericaforJuly1996;(b)ModeledresultfromtheGEOS-CHEM3-dimensionaltroposphericchemistryandtransportmodel.
TheHCHOinthiscaseisprimarilyduetoisopreneoxidation[Chanceetal.
,2000;Palmeretal.
,2001].
ATBD-OMI-0439remainingdifferencesinRayleighscatteringversuswavelengthoverthefittingwindow,variationofgroundalbedo,andimperfectintensitycalibrationoftheOMIradianceandirradiancemeasurements.
Theoverallfittingstrategyincludesanumberofoptions,whichwillbefullytestedduringOMIcommissioning,withtheonesthathaveprovedmostsuccessfulintheanalysisofprevioussatellitemeasurementsprovidingthebaseline.
Figure3.
2ElevatedslantcolumnsofHCHOovertheSoutheasternUnitedStatesinJuly1996.
MeasurementsarefromdirectfittingofGOMElevel1spectra.
3.
1.
1.
Nonlinearleast-squaresfittingTheLevenberg-Marquardtnonlinearleast-squaresfittingprocedure(nlls)[Marquardt,1963;Pressetal.
,1986]isusedinseveralofthesubsequentstepsintheanalysis.
Inthisprocedure,theχ2meritfunctionχσ2==)yFxaiiiiN/12[3-2]isminimizedwithrespecttotheparametersa.
Thestrategyforfindingtheminimumistobeginwithadiagonally-dominantcurvaturematrix,correspondingtoasteepestdescentsearchprocedure,andgraduallychangingcontinuouslyovertotheinverse-Hessian(curvature)methodsearchprocedureastheminimumisneared.
Convergenceisreachedwhenχ2islessthanapre-setamount,whenχ2decreasesbylessthanapre-setamountoverseveralsuccessiveiterations,orwhenallparameterschangebylessthanapre-setfractionforseveralsuccessiveiterations.
Iterationisalsohaltedwhenthenumberofiterationsreachesapre-setmaximumwithoutsuccessfulconvergence.
Estimatedfittinguncertaintiesaregivenasσi=iiC,whereCisthecovariancematrixofthestandarderrors.
Thisdefinitionisstrictlytrueonlywhentheerrorsarenormallydistributed.
Inthecasewherethelevel-1dataproductuncertaintiesarenotreliableestimatesoftheactualuncertainties,spectraldataaregivenunityweightoverthefittingwindow,andthe1σfittingerrorinparameteriisdeterminedas40ATBD-OMI-04variedpointspointsiirmsinnnC=εσ[3-3]whereεrmsistheroot-mean-squaredfittingresidual,npointsisthenumberofpointsinthefittingwindow,andnvariedisthenumberofparametersvariedduringthefitting.
Figure3.
3EnhancedHCHOasmeasuredbyGOMEfrombiomassburningepisodesover(a)thenorthernAmazonbasin;and(b)Indonesiaduringtheintensefiresof1997.
3.
1.
2.
Re-calibrationofwavelengthscalesThisisabaselineoption,requiredbythefactthatfittingtosmallroot-mean-squaredifferencesbetweenthemeasurementsandthemodeling(rms),comparabletothemeasurementSNRsrequiresbetterwavelengthcalibrationthanthatprovidedinthelevel1dataproductsATBD-OMI-0441(wavelengthcalibrationforthespecificfittingwindowismoreaccuratethanthatderivedforthespectrumasawhole).
ThemodelforthisprocedurecomesfromtheanalysisofGOMEdata[CasparandChance,1997;Chance,1998].
Theirradiancetobeusedinthesubsequentspectrumfittingisre-calibratedinwavelengthoverthefittingwindowbynllscomparisontoahigh-resolutionsolarreferencespectrumwhichisaccurateinabsolutevacuumwavelengthtobetterthan0.
001nm[ChanceandSpurr,1997].
Aslitwidth(instrumenttransferfunction)parameterisfittedsimultaneously.
Radiancespectraareequivalentlyfitted,withtheslitwidthparameterfrozentothevalues(versusCCDspectralfield)determinedinfittingtheirradiance:SincethewindowregionsforthesefourmoleculesareopticallythininallTelluricabsorptionsandcontainonlyafewpercentofinelastically-scatteredFraunhoferspectrum(Ringeffect),theprocedureworksalmostaswellonradiancesasonirradiances.
ExperiencewithGOMEspectra(whichareathigherspectralresolution)isthatbotharefittedtowithinabout1/50spectralresolutionelement.
Re-calibrationnormallyinvolvesonlythedeterminationofasinglewavelengthshiftparameterforthefittingwindow(baselineoption);inclusionofawavelength"squeeze"parameterisanon-baselineoption.
ThecaseforOMIiscomplicatedbythefactthatsuchcalibrationsaremadefortheseparatespectralfieldsontheCCDdetectorarray(GOMEhaslinearReticondetectors,whichmeasuresinglespectra).
WavelengthcalibrationsaremadeforeachOMIorbitasfollows:1.
Thesetofirradiances,versusCCDpositionarecalibratedinwavelengthoverthefittingwindow;2.
OnesetofradiancesversusCCDposition,selectedbyaninputparameter,usuallyinthemiddleoftheorbit,arecalibratedinwavelengthoverthefittingwindow.
Thiscalibrationisappliedastheinitialwavelengthscaleforallradiancesintheorbit;3.
Furtherfine-tuningoftherelativecalibrationofallradiancesthroughtheorbittotheirradianceisperformedduringthedetailedspectrumfitting,asdescribedbelow(baselineoption).
3.
1.
3.
ReferencespectraReferencespectraaredegradedtotheOMIresolutioneitherinpre-tabulatedform(baselineoption)orusingtheparameterizedslitfunctiondeterminedduringtheirradiancecalibration(non-baselineoption).
Theyarethenre-sampledtotheradiancewavelengthgrid,usingcubicsplineinterpolation[Pressetal.
,1986].
ThecurrentbaselinechoicesforreferencespectratofitHCHOareshowninFigure3.
4:HCHOcrosssectionsCantrelletal.
,1990.
NO2crosssectionsVandaeleetal.
,1997.
BrOcrosssectionsWilmouthetal.
,1999.
OClOcrosssectionsWahneretal.
,1987(correctedtovacuumwavelength).
O3crosssectionsBurrowsetal.
,1999.
O2-O2collisioncomplexGreenblattetal.
,1990(correctedtovacuumwavelength).
Ringeffect:DeterminedspecificallyforOMIapplicationsbyJ.
Joineretal.
Undersamplingcorrection:Calculateddynamically,ifrequired42ATBD-OMI-04Figure3.
4ReferencespectraforfittingtodetermineslantcolumnabundancesofHCHO:(a)HCHO;(b)O3;(c)BrO;(d)NO2;(e)OClO;(f)O2-O2collisioncomplex;(g)Ringeffect.
FittingofGOMEspectragavesystematicresidualsthatweremuchlargerthantheabsorptionofthetracegasesuntilitwasdiscoveredthattheseareduetospectralundersamplingbytheinstrument.
Highfrequencyspectralinformationisaliasedintothespectrumwhenirradiancesarere-sampledtotheradiancewavelengthscale(irradiancesforGOME,andforOMI,aremeasuredatslightlydifferentDopplershiftswithrespecttothesun)[Chance,1998;Slijkhuisetal.
,1999].
ThealgorithmforOMIincludesanon-baselineoption(sinceOMIisnotanticipatedtosignificantlyundersample)forcalculatingundersamplingcorrectionsatthetimeofATBD-OMI-0443referencespectrumsampling.
Thisisaccomplishedbytakingthehigh-resolutionFraunhoferreferencespectrum,convolvedwiththeinstrumenttransferfunction(asdeterminedduringtheirradiancespectralcalibration),whichconstitutesanoversampledirradianceandformingproperly-sampledandundersampledrepresentationsofit,usingcubicsplineinterpolation.
Thedifferencebetweentherepresentationistheundersamplingcorrection;forGOMEthisconstitutesca.
90%ofthesystematicresidual.
3.
1.
4.
Common-modecorrectionRemainingsystematicresidualswhichare,bydefinition,uncorrelatedtothetracegasspectramaybeaveragedandincludedinthespectrumfittingasa``common-mode''spectrum,toreducethefittingrmsand,proportionally,thefittinguncertainties.
Thisisincludedasanon-baselineoptionforOMI.
ForGOME,thisreducesthermsbyaboutafactoroftwoforBrOandafactorof3forNO2.
Thepresumptionisthattheundersamplingcorrectiononlyimperfectlymodelstheinstrumenttransferfunctionandthusleavessomeremainingsystematiceffects.
ComparisonswithGOMEmeasurementsofNO2forcleanmaritime(mostlystratosphericNO2)conditionsshowthatthisreduceduncertaintyistheappropriatechoice.
3.
1.
5.
Radiancefitting:BOAS(baselineoption)Theradianceisnllsfittedtoamodeledspectrumwhichincludestheirradiance,andeffectivealbedo,tracegasconcentrations,Ringeffectcorrection,andalow-orderpolynomial(uptocubic)closureterm:IAEeecccccNNRRssnn()()λλσλλλλλλσλσλ11012233L[3-4]Theσiaretheabsorptioncrosssections,σRistheRingeffectcorrection,andthec0–c3arethecoefficientsoftheclosurepolynomial.
Eachparametermaybeindividuallyselectedtobevariedduringthefittingorfrozenataconstantvalue.
ThedeterminationofslantcolumnabundancesforHCHOisanalogoustothatforBrOandOClOexceptforthefollowingdifference:BecauseoftheweaknessoftheHCHOabsorptionandthecorrelationwithotherspectralfeaturesintheHCHOfittingwindow,theHCHOfittingwillbemadewithfixedvaluesofBrOandO3thathavebeenpreviouslyfittedforthecorrespondingOMIspectra(baselineoption).
BrOvalueswillcomefromtheBrOtracegasfittingalgorithm.
O3slantcolumnvalueswilleithercomefromtheOMIO3totalcolumnalgorithm,orwillbefittedseparatelyforthepresentpurpose(TBD).
ThebaselineisthentovaryA,Ns(HCHO),Ns(NO2),Ns(O2-O2),CR,andc0–c3.
Thenon-baselineoptiontovaryBrOandO3willbepossibleaswell.
Inaddition,therewillbeabaselineoptiontocorrectthefixedO3slantcolumnvaluesforthedifferenceinviewingduetoRayleighscatteringforO3intheO3windowandtheHCHOwindow.
Thebaselinefittingwindowis337-356nm.
Inaddition,thelinearwavelengthscaleoftheirradianceisallowedtovary(baselineoption)tocorrectforsmallwavelengthchangesintheradianceovertheorbit.
Whileitisactuallytheradiancescalethatrequiresadjustment,therelativechangeinwavelengthscaleismuchsmallerthantheinstrumentresolutionorthevariationofmeasurablefeaturesinthereferenceorRingspectra:adjustmentoftheirradiancescaleavoidschangingthemeasuredquantity(theradiance)duringthefittingprocess.
Smoothing(low-passfiltering)oftheirradiances,radiances,andreferencespectraisincludedasanon-baselineoption.
Thisprocedureprovidesanalternatemeanstocorrectforspectralundersamplingeffects.
Itisimplementedbyapplyingarunning5-pointfilter(1/16,1/4,3/8,1/4,1/16)toeachofthespectra.
44ATBD-OMI-04Updatingofthefittingparametersisincludedasabaselineoption:ForeachspectrumafterthefirstinaparticularspectrumfieldoftheCCD,theinitialguessesforfittingparametersareupdatedtothosefittedtothepreviousspectrum.
3.
1.
6.
Radiancefitting:DOAS(non-baselineoption)Thisoptionconvertsthedirect(BOAS)fittingofradiancestotheDOASmethod.
Thelogarithmoftheradiancedividedbytheirradianceistakenandtheresulthigh-passfilteredbythesubtractionofalow-orderpolynomial(uptocubic,withtheactualnumberoftermsdeterminedbytheinputparameterfile):H[ln(I/E)]=-Ns1H(σ1)NsnH(σn)+cR1H(σR1/F)+cR2H(σR2/F)+HOT,[3-5]whereHdenoteshigh-passfiltering.
Thecoefficientsofthepolynomialaredeterminedbylinearfittingtoln(I/E)[Pressetal.
,1986].
ReferencecrosssectionssampledattheOMIwavelengthscalearealsohighpassfilteredbysubtractionofalow-orderpolynomial.
TheRingeffectcorrectionisdividedbytheirradianceforeachspectralfieldandtheresulthigh-passfilteredaswell(notethatthisdeterminesthefirsttermintheexpansionforthelogarithmicquantitywhichincludestheRingeffect;higher-ordertermsarenotincluded-seeEquation[1-4]).
ThefittingisthenentirelyanalogoustotheBOASfitting,exceptthatthefittedquantityisnowHln(I/E).
Iftheradiance/irradiancewavelengthadjustmentisnotselected,thentheDOASfittingwouldbelinearinthefittingparameters.
However,sincewavelengthadjustmentandBOASfittingarebothbaselineoptions,itisnotcurrentlyplannedtoimplementseparatelinearfittingforthiscase;theadditionalcomputertimeinfittingthelinearcasewiththenonlinearmethodisinconsequential,andtheresultsarevirtuallyidentical.
3.
2.
AirmassfactorsandverticalcolumnabundancesHCHOisassumedtobeprimarilytropospheric,withMsthatdependheavilyontheverticaldistributionandthegeophysicalconditions[Palmeretal.
,2001].
Ideally,airmassfactorsappropriatetoverticaldistributionsofHCHOforthemeasurementtimeandgeophysicalconditionswouldbeusedforoperationalprocessing.
ThesearedeterminedusingresultsfromtheGEOS-CHEMglobal3-Dmodeloftroposphericchemistryandtransport[Beyetal.
,2001]andarecalculatedusingtheSAOLIDORTradiativetransfermodel[Spurretal.
,2001;Chanceetal.
,2000;Palmeretal.
,2001].
LIDORTisdesignedtodeliverbothintensitiesandweightingfunctions.
LIDORTsolvestheradiativetransferequationinamulti-layeredatmospherewithmultiple-scatteringusingthediscreteordinatemethod.
Themodelcontainsafullinternalperturbationanalysisoftheintensityfield,allowingallweightingfunctionstobederivedanalyticallytothesamelevelofaccuracyspecifiedfortheintensity.
AlthoughLIDORThasbeendesignedprimarilyasageneralforwardmodeltoolfornon-linearatmosphericretrievalproblems,thecalculationofMsisastraightforwardapplicationofthemodel.
AsinglecalltoLIDORTwilldeliverboththetop-of-the-atmosphereupwellingintensityIBandthecorrespondingsetofweightingfunctionsK(z)requiredfortheMdetermination.
ThebaselineoptionforHCHOprocessingassumesthatthenecessaryassimilateddataforGEOS-CHEMmodeling,andthemodeledresultsforHCHOverticalcolumns,arenotavailableoperationally.
Airmassfactorsaredeterminedfromalookupprocedure,whereMsarepre-calculatedasfunctionsoflocationandseason(and,hence,albedo),andviewinggeometry.
Inordertosimplifythelookupprocess,theviewinggeometryisparameterizedbytheeffectivesolarzenithangle(ESZA),wheresec(ESZA)=sec(SZA)+sec(LOSZA)–1,[3-6]ATBD-OMI-0445whereSZAisthesolarzenithangleofthemeasurementandLOSZAistheline-of-sightzenithangle,andazimuthaldependenceofMisignored.
ExperiencewithfittingGOMEdatashowsthat,exceptforthehighestsolarzenithangles,above80o,thisprocedureaddsnegligibleadditionalerror.
Thelookupprocedurerequiresadatabaseofalbedovalues(TBD-itmaybetakenfromthosedevelopedusingGOMEdataattheKNMIandtheSAO,oreventuallydeterminedfromOMIdataitself,aftercloudanalysis).
ThevariationsofMwithalbedoandwithwavelengthoverthefittingwindowareincludedaserrorterms.
AnadditionalESZA-dependenterrorcontributionofupto20%isincludedtoaccountforthevariationoftheprofileofstratospheric/uppertroposphericHCHOfromtheselectedshape.
3.
3.
ErrorestimatesEstimatederrorsaregivenforagroundfootprintof40·40km2,whichhasbeenadoptedasthestandardforreportingOMIerrorestimateswithinEOS.
S/NvaluesforestimatingfittinguncertaintiescomefromtheOMI-EOSInstrumentSpecificationDocumentRS-OMIE-0000-FS-021,Tables5.
5and5.
6,andareappliedtohebestfittingknowledgefromGOMEanalysis,takingintoaccountthedifferenceinspectralresolution.
Alluncertaintiesaregivenhereas1σ.
The1σfittingerrorinparameteriisdeterminedasσi=iiC+δBrO+δO3[3-7]orvariedpointspointsiirmsinnnC=εσ+δBrO+δO3,[3-8]AsdiscussedinSection3.
1.
1.
,Cisthecovariancematrixofthestandarderrors.
Equations[3-7]and[3-8]includethecontributionfromcorrelationoffittedparameters,andadditionalcontributionsδBrOandδO3fortheuncertaintiesinthepreviously-fittedNs(BrO)andNs(O3)andthecorrelationsoftheBrOandO3concentrationstotheHCHO.
Thiscontributionisdeterminedfromalookupprocedurebaseduponfinitedifferencestudies.
Thepublisheduncertaintiesoftheabsorptioncrosssectionsareaddedtothisinquadraturetoobtainthefinalslantcolumnfittinguncertainties.
Thefittingprecisionforbiogenically-enhancedconditionsis9%.
δBrOandδO3areTBD,butareestimatedhereas5%each,foratotalslantcolumnfittinguncertaintyof11%.
Theuncertaintyinthecrosssections,includingtemperaturedependence,is10%,foratotalslantcolumnuncertainty,inbiogenically-enhancedconditions,of15%.
TheuncertaintyinMisaddedinquadraturetoobtainthefinalverticalcolumnfittinguncertainties.
TheuncertaintyinMduetoassumptionsontheverticalprofileshapeis20%,worstcase[Palmeretal.
,2001].
EffectsonMduetocloudparameteruncertaintiesandalbedouncertaintiesareTBD,butareassumedforthepresenttobenegligiblecomparedtoassumptionsontheverticalprofileshape(theseeffectswillbeaddressedindetailwhentheclouduncertaintiesandalbedodatabaseuncertaintieshavebeenquantified).
Thetotalerrorfortheverticalcolumnuncertaintyunderbiogenically-enhancedconditionsisthen15%+20%=25%.
Theerrorexcludingthesystematiccontributionfromcrosssectionsuncertaintiesis22%,correspondingto3.
5·1015cm-2verticalcolumndensity.
ThiscanbecomparedtotherequirementintheScienceRequirementsDocumentforOMI-EOS(RS-OMIE-KNMI-001,Version2)of1015cm-2.
Thedetectionlimit,onlyconsideringfittinguncertainty,is46ATBD-OMI-041.
8·1015cm-2verticalcolumndensity.
Thus,meetingtheScienceRequirementDocumentspecificationwillrequireimprovementinS/NoverthatgivenintheInstrumentSpecificationDocumentaswellasMswithgreateraccuracy.
3.
4.
OutputsStandardoutputsinclude:Slantcolumnabundanceand1σfittinguncertaintiesforHCHOandtheotherspeciesvariedinthefittingwindow;CorrelationofotherfittedspeciestoHCHO(fromoff-diagonalelementsofthecovariancematrixofthestandarderrors);Fittingrms;Convergenceflagsandnumberofiterations(successfulwhichconvergencecriterion)Geolocationinformation;Versionnumbersofalgorithmandparameterinputfile;Verticalcolumnabundancesand1σuncertainties.
3.
5.
ValidationHCHOslantandverticalcolumndensitiesarecurrentlymeasuredbyGOME[Thomasetal.
,1998;Chanceetal.
,2000;Palmeretal.
,2001]andwillbemeasuredbySCIAMACHY.
Ground-andaircraft-basedmeasurementcampaignswillbenecessaryforOMIvalidation,especiallywhenconcentrationareexpectedtobehigh,i.
e.
,forperiodswithstrongtropospherichydrocarbonemissions.
ApastexampleistheU.
S.
SouthernOxidantsStudy(SOS),measuringcontinentalproductionofHCHOfromisoprene[Leeetal.
,1998].
Measurementsarealsonecessarytoconfirmratesofproductioninthemaritimefreetroposphere,suchasthosefromthe1997SubsonicAssessment(SASS)OzoneandNitrogenOxideExperiment(SONEX)[Singhetal.
,2000].
MeasurementsoverthesoutheasternU.
S.
insummertime,andoverthemidlatitudeoceans(preferablyinsummertimeformaximumproductionfromoxidationofCH4)wouldprovideoptimumdatasets.
Midlatitudemaritimemeasurementscouldbecombinedwithcampaignstostudyintercontinentalpollutiontransport.
ATBD-OMI-04473.
6.
ReferencesBeyI.
,D.
J.
Jacob,R.
M.
Yantosca,J.
A.
Logan,B.
Field,A.
M.
Fiore,Q.
Li,H.
Liu,L.
J.
Mickley,andM.
Schultz,Globalmodelingoftroposphericchemistrywithassimilatedmeteorology:Modeldescriptionandevaluation,J.
Geophys.
Res.
,submitted,2001.
Burrows,J.
P.
,A.
Richter,A.
Dehn,B.
Deters,S.
Himmelmann,S.
Voigt,andJ.
Orphal,AtmosphericremotesensingreferencedatafromGOME-2.
Temperature-dependentabsorptioncross-sectionsofO3inthe231-794nmrange,J.
Quant.
Spectrosc.
Radiat.
Transfer61,509-517,1999.
Cantrell,C.
A.
,J.
A.
Davidson,A.
H.
McDaniel,R.
E.
Shetter,andJ.
G.
Calvert,Temperature-dependentformaldehydecrosssectionsinthenear-ultravioletspectralregion,J.
Phys.
Chem.
94,3902-3908,1990.
Caspar,C.
,andK.
Chance,GOMEwavelengthcalibrationusingsolarandatmosphericspectra,Proc.
ThirdERSSymposiumonSpaceattheServiceofourEnvironment,Ed.
T.
-D.
GuyenneandD.
Danesy,EuropeanSpaceAgencypublicationSP-414,ISBN92-9092-656-2,1997.
Chance,K.
V.
,J.
P.
Burrows,andW.
Schneider,RetrievalandmoleculesensitivitystudiesfortheGlobalOzoneMonitoringExperimentandtheSCanningImagingAbsorptionspectroMeterforAtmosphericCHartographY,Proc.
S.
P.
I.
E.
,RemoteSensingofAtmosphericChemistry,1491,151-165,1991.
Chance,K.
,andR.
J.
D.
Spurr,RingEffectStudies:RayleighScattering,IncludingMolecularParametersforRotationalRamanScattering,andtheFraunhoferSpectrum,Appl.
Opt.
36,5224-5230,1997.
Chance,K.
,AnalysisofBrOMeasurementsfromtheGlobalOzoneMonitoringExperiment,Geophys.
Res.
Lett.
25,3335-3338,1998.
Chance,K.
,P.
I.
Palmer,R.
J.
D.
Spurr,R.
V.
Martin,T.
P.
Kurosu,andD.
J.
Jacob,SatelliteObservationsofFormaldehydeoverNorthAmericafromGOME,Geophys.
Res.
Lett.
27,3461-3464,2000.
Greenblatt,G.
D.
,J.
J.
Orlando,J.
B.
Burkholder,andA.
R.
Ravishankara,Absorptionmeasurementsofoxygenbetween330and1140nm,J.
Geophys.
Res.
95,18,577-18,582,1990.
LeeY.
-N.
,etal.
,Atmosphericchemistryanddistributionofformaldehydeandseveralmultioxygenatedcarbonylcompoundsduringthe1995Nashville/MiddleTennesseeOzoneStudy,J.
Geophys.
Res.
103,22,449-22,462,1998.
Marquardt,D.
L.
,Analgorithmforleast-squaresestimationofnon-linearparameters,J.
Soc.
Indust.
Appl.
Math.
2,431-441,1963.
Palmer,P.
I.
,D.
J.
Jacob,K.
Chance,R.
V.
Martin,R.
J.
D.
Spurr,T.
P.
Kurosu,I.
Bey,R.
Yantosca,A.
Fiore,andQ.
Li,AirMassFactorFormulationforSpectroscopicMeasurementsfromSatellites:ApplicationtoFormaldehydeRetrievalsfromGOME,J.
Geophys.
Res.
106,14,539-14,550,2001.
Press,W.
H.
,B.
P.
Flannery,S.
A.
Teukolsky,andW.
T.
Vetterling,NumericalRecipes,ISBN0-521-30811-9,CambridgeUniversityPress,1986.
Singh,H.
,etal.
,DistributionandfateofselectedoxygenatedorganicspeciesinthetroposphereandlowerstratosphereovertheAtlantic,J.
Geophys.
Res.
105,3795-3805,2000.
48ATBD-OMI-04Slijkhuis,S.
,A.
vonBargen,W.
Thomas,andK.
Chance,CalculationofundersamplingcorrectionspectraforDOASspectralfitting,Proc.
ESAMS'99-EuropeanSymposiumonAtmosphericMeasurementsfromSpace,563-569,1999.
Spurr,R.
J.
D.
,T.
P.
Kurosu,andK.
Chance,Alinearizeddiscreteordinateradiativetransfermodelforatmosphericremotesensingretrieval,J.
Quant.
Spectrosc.
Radiat.
Transfer68,689-735,2001.
Thomas,W.
,E.
Hegels,S.
Slijkhuis,R.
Spurr,andK.
Chance,DetectionofbiomassburningcombustionproductsinSoutheastAsiafrombackscatterdatatakenbytheGOMEspectrometer,Geophys.
Res.
Lett.
25,1317-1320,1998.
VandaeleA.
C.
,C.
Hermans,P.
C.
Simon,M.
Carleer,R.
Colin,S.
Fally,M.
F.
Mérienne,A.
Jenouvrier,andB.
Coquart,MeasurementsoftheNO2absorptioncross-sectionfrom42000cm-1to10000cm-1(238-1000nm)at220Kand294K,J.
Quant.
Spectrosc.
Radiat.
Transfer59,171-184,1998.
Wahner,A.
,G.
S.
Tyndall,andA.
R.
Ravishankara,AbsorptioncrosssectionsforOClOasafunctionoftemperatureinthewavelengthrange240-480nm,J.
Phys.
Chem.
91,2734-2738,1987.
Wilmouth,D.
M.
,T.
F.
Hanisco,N.
M.
Donahue,andJ.
G.
Anderson,FouriertransformultravioletspectroscopyoftheA2Π3/2←X2Π3/2transitionofBrO,J.
Phys.
Chem.
103,8935-8945,1999.
ATBD-OMI-04494.
SO2A.
J.
Krueger,JointCenterforEarthSystemsTechnologyN.
A.
Krotkov,GoddardEarthScienceTechnologyCenterUniversityofMaryland,BaltimoreCounty4.
1.
IntroductionSulfurdioxidereleasesintheEarth'satmospherearealwaystransientbecauseofchemicalconversiontosulfatewithtimescalesofhoursintheboundarylayertoweeksinthestratosphere.
Volcanicsulfurdioxidecloudsareeasytomeasurebecausetheyareathighaltitudes,large,andhavelonglifetimes.
PassivedegassingandSO2fromanthropogenicpollution(e.
g.
,burningofcoal),ismoredifficulttomeasureasthegasiscontainedintheboundarylayerbelowthemeanlevelofpenetrationforUVsunlight.
Thesebackgroundsourcesareimportanttomeasurebecausetheycontributemoretotheglobalannualsulfurbudgetthanexplosiveeruptions.
Modelcalculations[Chinetal.
,2000]ofbackgroundSO2columnamountsareshowninFigure4.
1andcomparedwithfarlargeramountsmeasuredwithTOMS[Kruegeretal.
,I2000]fromfourlargevolcaniceruptions(yellowcolors).
Figure4.
1GlobalmonthlymapofairpollutionSO2fromGOCARTmodel[Chinetal.
,2000]superimposedonglobalcompositemapofseveralvolcanicSO2eruptioncloudsmeasuredbyTOMS[Kruegeretal.
,2000].
50ATBD-OMI-044.
1.
1.
VolcanicSO2Volcaniceruptionshavebeenconsideredanimportantfactorinclimatechangeeversinceabnormalweatherwasobservedtofollowmajoreruptions.
Thefactorforforcingclimatewassuspectedtobestratosphericsulfateaerosolsfromsulfurdioxideratherthanthemoreobviousvolcanicashbecauseofthelong-lastingeffects.
Thishypothesiscouldnotbevalidatedquantitativelybecausenomeanswasavailableformeasuringtheamountofsulfurloftedintothestratosphereuntiltheeraofsatellites.
ForthefirsttimethesulfurdioxideoutputforallsizesofmagmaticeruptionscouldbemeasuredwiththeTOMSinstrument[Krueger,1983;Kruegeretal.
,2000]becauseofitscontiguousmappingability.
Nootherinstrumentortechniquehaseverapproachedthisuniquerecord.
The14-yearNimbus7TOMSmissioncapturedeveryeruptionsince1978greaterthan10ktSO2,including2super-sizederuptions.
The1991eruptionofMt.
Pinatuboisbelievedtobethelargestofthecenturyinsulfuroutput(20Mt),althoughonlyindirectestimationsexistforeruptionspriorto1978,sothisissuemayneverberesolved.
Follow-onTOMSmissions(Meteor-3,ADEOS,EarthProbe)haveextendedtherecordto21years(withan18monthgapin1995-6,whichmaybepartiallyfilledusingGOMEdata[Burrowsetal.
,1993,Burrowsetal.
,1999],).
Theuncertaintyintheaverageannualfluxofsulfurfromeruptions[Bluth,etal.
,1993;Bluthetal.
,1997]hasbeenreducedfrom100%to30%.
Thisaverageisrepresentativeofthelate1900'sbutisgreatlyinfluencedbytheoccurrenceoflargeeruptionsinthedatabase.
ThisdatabasewillcontinuewithQuikTOMSandOMI,asthesuccessortoTOMS.
TheOMImissioncouldextendthevolcanicrecordthroughthenextdecadewhentheNPOESSinstrumentsareexpectedtooperationallymeasurevolcaniceruptions.
4.
1.
2.
AnthropogenicSO2ThesmallcolumnamountsofairpollutionSO2predictedfromthemodelshowninfigure4.
1areamajorchallengeforsatelliteinstruments.
However,columnabundanceslessthan2.
6DobsonUnits(DU,1DU=2.
6868·1016cm-2)duetotheburningofsulfur-richlignitecoalinSoutheasternEurope[EisingerandBurrows,1998]havealreadybeenmeasuredinwinterconditionsbyGOME.
ThisresultwasduetoGOME'shighsignaltonoiseratioinspiteofitslargefootprintwhichdilutesthesignalfrompointsources.
WithOMI'ssmallerfootprintandcomparablesensitivity(ca.
0.
4DUverticalcolumnabundance)itwillbepossibletoobservemanyofthesourcesofanthropogenicsulfurdioxideandbegintobuildasatellitebaseinventory.
4.
1.
3.
VolcanicashmeasurementsAshispresentinsomevolcanicclouds,particularlyfromexplosiveeruptions,andaffectstheradiativetransferpathsignificantly.
Sulfurdioxideretrievalsthatignoretheashproduceseriousoverestimatesdependingontheashopticaldepth(Kruegeretal.
,1995).
Theopticaldepthdependsontheindexofrefractionoftheashwhichvariessignificantlybetweenvolcanoesanderuptions.
Fortunately,thepresenceofashcloudsisidentifiedintheOMIdatabydiscriminationfromRayleighscatteringusingtheUVabsorbingaerosolindex(AAI)at340/380nmwavelengthpair[Seftoretal.
,1997;Krotkovetal.
,1999a]ordistancefromtheRayleighcurveonthespectralslope-reflectivitydiagram[Krotkovetal.
,1997a].
Quantitativealgorithmsweredevelopedtocomputeashopticalthickness,effectiveparticlesize,andcolumnmass[Krotkovetal.
,1999a,b].
ThesewillbeenhancedtakingadvantageofthebroaderspectralspanofOMItodeterminewater/icepresence.
Withtheashspecifiedwewillproducetablesincorporatingtheparticularashpropertiesforreprocessingofthevolcanicclouddata.
ATBD-OMI-04514.
2.
SelectionofoptimumspectralregionTheoptimumspectralregionforSO2retrievalsisdeterminedfromthedifferentialradianceduetoaddedSO2comparedwiththenoiseradianceoftheinstrument.
Figure4.
2isanexampleofthedifferentialradiance(bluecurves)duetotheadditionof1DUofSO2inthestratosphere(20km)orintheboundarylayer(1km)atmidlatitudeconditions.
Anorder-of-magnitudelossofsensitivityatthelesseraltitudeisveryapparent.
However,theoptimumspectralregionbetween310and320nmremainsnearlythesame.
WhenthedifferentialradianceiscomparedwiththeexpectedOMInoise(redcurve)basedonOMIDemonstrationModeldata[DeVries,2000]forminimumradianceconditions(clearsky)theradianceexceeds.
Figure4.
2SelectionofoptimumwavelengthregionforSO2determinationatmidlatitudeswithclearsky.
Radianceofthemidlatitudeatmosphere(blackcurve)anddifferentialradianceduetoSO2layers(bluecurves)comparedwiththenoiseradianceofOMI(redcurve).
thenoiseradianceonlyforwavelengthslessthan320nm.
Thebreakinthenoisecurveat310nmlocatedattheintersectionoftheUV-1andUV-2channelsisnearthemiddleoftheoptimumspectralregion.
However,differinggroundfootprintspreventdirectcombinationofthetwochannels.
TheUV-2channelat310-317nmispreferredforairpollutionsourcedatabecauseofthesmallerfootprint;whereastheUV-1channelat305-310nmhasagreatersignalabovethenoise.
Thus,dualprocessingmayberequiredOzoneabsorptionvariesstronglyacrosstheoptimumspectralregionsofbothchannelsandtheuseofairmassfactorsisproblematic.
Thealgorithmdescribedinthenextsectionmakesuseofwavelengthdependentpathsasanimplicitpartofradiativetransfertables.
52ATBD-OMI-044.
3.
DetailedDescriptionsoftheSO2algorithm4.
3.
1.
InversionstrategyForOMIwehavedevelopedanapproachwhichcombinestheTOMSandDOAStechniques.
Thegeneraldescriptionofthealgorithm(dataflow)isshowninFigure4.
3.
Figure4.
3SO2inversionflowchartTheSO2inversionstrategyisbasedonthestatisticalapproachgivenbyDubovikandKing[2000].
ThestrategyistoconsiderOMImeasurementstogetherwithaclimatologicaldataasasinglesetofmulti-sourcedata.
Theinversiontechniqueisdesignedasasearchforthebestoverallfitofalldataconsideredbyourforwardmodel(inaleast-squaressense)takingintoaccountdifferentaccuraciesofthefitted(measuredandapriori)data.
Thebasicmodelcontainstheterms:+=+Σ=Σ++Σ=++Σ=Σ*****;;;;λλλλλλλRkSkRkSkkkNNN[4-1]()λ*kN-OMImeasuredNvalues:N1*λ()=100log10IF;(k=1forUV1andk=2forUV2)Σ-theconcentrationofSO2,Σ*-aprioriestimateoftheconcentrationofSO2;,-theconcentrationofO3,*-aprioriestimateoftheconcentrationofO3;Asterisksdenotethatthedataaremeasured(orknown)withsomeuncertainty,k:SkandRkarethesystematicandrandomNvalueerrors,andσΣ2andσ2denotethevariancesinaprioriATBD-OMI-0453SO2andO3estimates.
Fordiscretespectralmeasurements,()λ*kNisconsideredasavectorandEq.
[4-1]canbesolvednumerically.
WeassumethatsystematicNvalueerrors(includingaerosolandcloudeffects)inEquation[4-1]canbedescribedbya2nddegreespectralpolynomialplusknownhighfrequencyterms(Ringeffect):Sk=ak+bk(λ-λ0)+ck(λ-λ0)2+Ring(λ)+….
.
WealsoassumethatrandomNvalueerrors(Rk)aredistributednormallywithzeromeansandknowncovariancematricesandvariances,Rk.
Toachievethestatisticallyoptimumsolution,x=[Σ,,ak,bk,ck,],wesolveEquations[4-1]usingtheMaximumLikelihoodapproach.
ThesolutionofEquations[4-1]isperformedusingaleast-squaresfittingmethodwithweighting(S/N)-2(λ,S(λ)).
TheaprioriozoneandSO2datasupplementtheOMIradiancedata,withtheirrelativecontribution(Lagrangemultipliers)weightedaccordingtotheratioofthevariancesofthemeasurementsanda-prioriestimates:γ=Rk/σ2andγΣ=Rk/σΣ2.
4.
3.
2.
ForwardmodelDetailedradiativetransfercalculationsareusedtobuildlookuptablesofN(λ,Σ,)ofbackscatteredradiancesinthe305to340nmspectralregionattheOMIspectralresolutionasafunctionofO3andSO2verticalprofiles(Figure4.
4)andtheconditionsofthemeasurement:geometry,surface/cloudpressure,reflectivityandlatitude.
Figure4.
4Forwardmodel:ozoneandbackgroundsulfurdioxideprofilesTheinversionalgorithmworkswithanarbitrarysetofOMIwavelengthswithin300-310and310-330nmfittingwindows,whichcanrangefromjustfour"TOMS-like''wavelengthstothefullrangeofwavelengthsatthehighestOMIspectralresolution.
LiketheTOMSalgorithm(andOMIstandardozonealgorithm),theclouds,aerosolsandsurfacealbedovariationswithinOMIFOVaretreatedimplicitlyintheforwardmodelthroughtheconceptofLambertianEquivalentReflectivity(LER)(knownfromtheOMIcolumnozonealgorithm).
TheradiancesandsensitivityfunctionstotroposphericSO2arepre-calculatedfortwoconstantmixingratioSO2verticalprofiles:a)fromthegroundto900mbarandb)fromthegroundto700mbar(Fig.
54ATBD-OMI-044.
4).
WhenhighSO2amountsarefound,theyareassumedtobevolcanicandwilluseprecomputedtablesforanSO2layerat15km.
AbsorptioncrosssectionsinthenearUVofozoneandsulfurdioxideareshowninFig.
4.
5togetherwiththeratioofthecrosssections.
BothozoneandSO2cross-sectionsaretemperaturedependent.
Therefore,forbestaccuracywehavetoaddressthegeographicalandseasonalchangesintheboundarylayertemperature.
Inaddition,theairtemperaturedropswithaltitude.
Wepresentlyassumethatexplosivevolcaniccloudsrisetothetropopausewherethetemperatureisnear210K.
Sulfurdioxidecrosssectionsat210and295KaretakenfromMcGeeandBurris[1987],augmentedbyManattandLane[1993]atwavelengthslongerthan320nm.
CurrentworkatintermediatetemperaturesbyJ.
HalpernwillbeusedwhenpublishedFigure4.
5Ozoneandsulfurdioxidecrosssections4.
3.
3.
InversiontechniqueGeneralsolutionofnon-linearsystemofEq.
(4-1)canbere-writtenasfollows(thisisgivenforclarityofthealgorithm):f*=fx4-2]SinceEq.
[4-2]isnon-linear,wesolveititerativelyonthebasisofTaylorexpansioninsmallvicinityateachstep:xp+1=xptpxp,[4-3]wherexpiscorrectionwhichissolutionoflinearsystemateachp-thstep:UpTC1Up()xp=UpTC1fxp()f*().
[4-4]where,UpistheJacobimatrixofthefirstderivatives((sensitivityfunctions)inthenearvicinityofthevectorxp,tp–Levenberg-Marquardtmultiplier[Marquardt,1963](ifnecessarythemoresophisticatedLevenberg-MarquardttypestatisticalcorrectioncanbeincludedinthematrixATBD-OMI-0455jijjiijiSOSnSnγγλλσ+=22222),(UpTC1Up(),thedetailsaregiveninisgiveninthepaperDubovikandKing[2000]).
Thismultiplierisusedtoprovidemonotonicconvergenceanditistypicallysmallerthan1(tp<1).
TheinitialguessoftheeffectivereflectivityandtotalozonevalueistakenfromtheOMIozonealgorithmandtheinitialguessforSO2istakenfromthestandardDOASfittingprocedureoftheOMIradiancesintheUV2spectralregion.
Thentheiterations[4-3]areperformedtoimprovetheSO2solution.
Inordertocontrolconvergencethefollowingresidualshouldbecontrolledateachstep:Ψxp()=fxp()f*()TC1fxp()f*().
[4-5]IftheconditionΨxp+1()≤Ψxp+1()isnotsatisfiedthentpshouldbedecreaseduntiltheconditionissatisfied.
Theiterationsareneededtoimprovethesolutionandreducetheresidual[4-5].
TheconvergenceissatisfiedwhentheSO2differencebetweeniterationsislessthanacertainthresholdvalue.
4.
4.
Erroranalysis4.
4.
1.
CheckingconsistencyoftheforwardmodelwiththemeasurementsThemagnitudeoftheminimumofΨx()isnotpredefinedandusuallyisusedtoestimatethevarianceofthemeasurementserrors,σ2,(forthesimplestcaseofthecovariancematrix:C=σ2Ι,Ιistheunitymatrix).
Incaseofuncorrelatedrandommeasurementerrors,theresidual,Ψx(),isdistributedaccordingtoχ2,i.
e.
Ψxp()=fxp()f*()TC1fxp()f*()≈MN()σ2.
[4-6]Correspondingly,thevarianceofthemeasurementerrorsσ2canbeestimatedfromtheresidualvalue:σ≈Ψxp()M-N.
[4-7]whereMisthenumberoffittingwavelengthsandNisthenumberofretrievedparameters(N=5inourcase).
Thisestimateofthemeasurementerrorisalsoveryusefulforcheckingboththemeasurementerrorandconsistencyofforwardmodel.
Namely,forareasonablefittingoneshouldexpect:σ2~ε2[48]whereε2isanticipatedmeasurementaccuracy(measurementerrorvariance).
Ifthecondition[4-8]isnotmet,theforwardmodelisnotconsistentwiththeobservationalconditions.
Thus,thecontroloftheresidual[4-7]isusefulforaninternalqualitycontroloftheretrieval.
4.
4.
2.
EstimatingretrievalerrorThelookuptablescanbealsousedfortheestimationoftheSO2retrievalnoiseduetorandomnoiseintheOMImeasurements.
Table4.
1showsthecalculatedSO2noise(1σSO2)fordifferentpairsofOMIchannelsfromthefollowingequation:[4-9]56ATBD-OMI-04whereSi=dN(λi)/dissensitivitytoozoneatwavelengthλi,(knownfromJacobimatrixofthefirstderivatives);γi(HS)isthesensitivitytoSO2,dN(HS,λi)/dΣ,normalizedtoSi(HSistheSO2altitude),andniis1/(signal-to-noise)oftheOMImeasurements.
Table4.
1Estimated1σOMISO2retrievalnoiseusingdifferentpairsofOMIspectralchannelsforconditionsinFigure4-2.
λλλλ2λλλλ21σσσσSO2at20km[DU]1σσσσSO2at1km[DU]304.
5305.
70.
412.
76305.
7306.
60.
271.
55306.
6307.
60.
251.
53307.
6308.
70.
201.
04308.
7309.
60.
211.
14309.
6310.
80.
241.
14310.
8311.
850.
201.
00311.
85313.
20.
210.
93313.
2314.
40.
241.
11314.
4315.
30.
512.
18315.
3316.
20.
723.
28316.
2317.
20.
943.
82317.
2319.
20.
652.
81319.
2319.
81.
214.
86308.
7311.
850.
181.
02310.
8314.
40.
231.
22308.
7314.
40.
201.
22307.
6310.
80.
231.
03307.
6313.
20.
230.
96Thetypical1σSO2noiseforaclearsceneinonepixelis0.
2DUifonly2wavelengthsareusedandforanSO2layerat20km.
Thenoiseincreasesto1DUiftheSO2layerat1km.
TheincreaseinretrievalnoiseresultsfromthedecreaseinradiometricsensitivitywithdecreaseinSO2plumeheight.
UsingmultipleOMIwavelengthsreducestheradiometricnoisecomparetothecaseofjust2wavelengths.
Theestimatedcovariancematrixoftheretrievalerrorestimatedinlinearapproximationisgivenbythefollowingequation(seeDuboviketal.
[2000]):Cx≈UpfTC1Upf()1σ2[4-10]wherepfisthenumberofthelastiteration,andσ2canbeobtainedfrom(4-7)orfromtheanticipatedmeasurementaccuracy(4-8).
Applicationofequation(4-10)forthecaseofboundarylayerSO2anduncorrelatedandspectrallyindependentrandomnoiseinOMImeasurementswithsignal-to-noiseratioof300providesthetheoreticalSO2errorestimateof~0.
1DU(1σSO2noise).
Thiserrorestimatedoesnotincludeanygeophysicalorinstrumentalbiasesthatshouldbeproperlytakenintoaccountbytheforwardmodel[4-1].
ATBD-OMI-04574.
4.
3.
CorrectionforRingeffectThemodelNvaluesinEquations[4-1]arecorrectedfortheRingeffectusingapre-calculatedlookuptable.
TheRingcorrectionalgorithm[Joineretal.
,1995]willbeappliedtotheentirespectrum,butwillonlyincludeO3absorption.
Figure4.
6(left)showsthestandardRingcorrection[J.
JoinerandP.
K.
Bhartia,privatecommunication]intheSO2fittingwindow.
Figure4.
6(right)showstheadditionalcorrectionduetoSO2absorption[calculatedbyD.
Flittner].
TheadditionalRingcorrectionduetoSO2upto10DU(maximumexpectedfornon-volcanicsources)issmallandonecanusethestandardRingcorrectionasafirstapproximation.
Figure4.
6RingcorrectionintheSO2spectralfittingwindow:a)StandardRingspectrumtakingintoaccountonlyozoneabsorptionandRayleighscattering;b)Additionalcorrectiondueto10DUofSO2.
AtriangularslitfunctionwithaFWHM=0.
45nmwasusedforthiscalculation.
4.
4.
4.
CorrectionforvolcanicashAshinvolcaniccloudscanproduceoverestimationinTOMS-likesulfurdioxideretrievals[Kruegeretal.
,1995].
IntheinitialversionoftheSO2algorithmtheash/aerosoleffectsaretreatedimplicitlythroughtheeffectivereflectivity(aby-productoftheOMIozonealgorithm)andcoefficientsofthequadraticcorrectiontermin(1).
ThiswillbechangedinthevolcanicSO2algorithmthroughtheexplicitradiativetransfermodelingofthespectraldependenceoftheasheffect[Krotkovetal.
,1997a;Kruegeretal.
,2000].
Afterinitialashparametersestimationusingoneofthemethodsdescribedin[Krotkovetal.
,1999a,b],anewsetoflookuptableswithexplicitinclusionofashparticleswillbeconstructedforthespecificvolcaniccloud.
4.
5.
OutputsStandardoutputsinclude:VerticalcolumnSO2anduncertaintyPrescribedSO2layerheightFittingrmsConvergenceflagsTotalozone58ATBD-OMI-04CommonOMIdata:GeolocationdataSolarzenithangleSatellitezenithangleSurfacereflectivityCloudtopheightVersionnumbersofalgorithmandparameterinputfile;4.
6.
ValidationTheOMISO2algorithmperformancewillbeconfirmedthroughtestingwithsyntheticradiances,andcomparisonwithDOASretrievalsusingGOMEdata.
TheMKIIBrewerspectrophotometerisdesignedtomeasureSO2andtotalozonefromdirectsunirradiancesat5nearUVwavelengths.
However,backgroundSO2amountsareusuallybelowthedetectionlimit.
BetterstraylightrejectionindoublemonochromatorMKIIIBrewerinstrumentsmaybeadequatetomeasureamountsassmallas1matm-cm.
SevendoubleBrewersatNorthernHemispherestations(oneattheGoddardLaboratoryforAtmospheres)arecollectingdata.
Anewdirectsunspectralscanmethodappearscapableofdetectingsubmatm-cmamounts(J.
Kerr,privatecommunication,2001).
Ifsuccessful,thiswillbeinvaluableforvalidatingbackgroundSO2amounts.
ValidationofvolcanicSO2cloudamountsdependsondriftofthecloudoveranobservingstation.
Normallyvolcaniccloudsdrifttoorapidlytoscheduleinstrumentedaircraftflightsintheirpaths.
Othersatellitedatacanoftenprovidesupportinginformationorcoincidentobservationstolendconfidenceintheretrievals.
ATBD-OMI-04594.
7.
ReferencesBluthG.
J.
S.
,C.
C.
Schnetzler,A.
J.
Krueger,andL.
S.
Walter,Thecontributionofexplosivevolcanismtoglobalatmosphericsulfurdioxideconcentrations,Nature366,327-329,1993.
BluthG.
J.
S.
,W.
I.
Rose,I.
E.
Sprod,andA.
J.
Krueger,Stratosphericloadingofsulfurfromexplosivevolcaniceruptions,J.
Geol.
105,671-683,1997.
Burrows,J.
P.
,K.
V.
Chance,A.
P.
H.
Goede,R.
Guzzi,B.
J.
Kerridge,C.
Muller,D.
Perner,U.
Platt,J.
-P.
Pommereau,W.
Schneider,R.
J.
Spurr,andH.
vanderWoerd,GlobalOzoneMonitoringExperimentInterimScienceReport,ed.
T.
D.
GuyenneandC.
Readings,ReportESASP-1151,ESAPublicationsDivision,ESTEC,Noordwijk,TheNetherlands,ISBN92-9092-041-6,1993.
Burrows,J.
P.
,M.
Weber,M.
Buchwitz,V.
V.
Rosanov,A.
Ladstatter,A.
Weissenmayer,A.
Richter,R.
DeBeek,R.
Hoogen,K.
Bramstedt,andK.
U.
Eichmann,TheGlobalOzoneMonitoringExperiment(GOME):Missionconceptandfirstscientificresults,J.
Atmos.
Sci.
,56,151-175,1999.
Chin,M.
,R.
B.
Rood,S-J.
Lin,J-F.
Muller,andA.
M.
Thompson,AtmosphericsulfurcyclesimulatedintheglobalmodelGOCART:Modeldescriptionandglobalproperties,J.
Geophys.
Res.
,105,24671-24687,2000Dave,J.
V.
Multiplescatteringinanon-homogeneous,Rayleighatmosphere,J.
Atmos.
Sci.
22,273-279,1965.
DeVries,J.
,S/Nstatusprediction,SE-OMIE-0437-FS/00,2000.
Dubovik,O.
andM.
King,Aflexibleinversionalgorithmforretrievalofaerosolopticalpropertiesfromsunandskyradiancemeasurements,J.
Geophys.
Res.
105,20,673-20,696,2000.
Eisinger,M.
,andJ.
P.
Burrows,TroposphericsulfurdioxideobservedbytheERS-2GOMEinstrument,Geophys.
Res.
Lett.
25,4177-4180,1998.
Joiner,J.
,P.
K.
Bhartia,R.
P.
Cebula,E.
Hilsenrath,R.
D.
McPeters,andH.
Park,RotationalRamanscattering(Ringeffect)insatellitebackscatterultravioletmeasurements,Appl.
Opt.
34,4513-4525,1995.
Joiner,J.
andP.
K.
Bhartia,Thedeterminationofcloudtoppressuresfromrotational-Ramanscatteringinsatellitebackscatterultravioletmeasurements,J.
Geophys.
Res.
100,23,019-23,026,1995.
Krotkov,N.
A.
,A.
J.
Krueger,andP.
K.
Bhartia,Ultravioletopticalmodelofvolcaniccloudsforremotesensingofashandsulfurdioxide,J.
Geophys.
Res.
102,21,891-21,904,1997.
Krotkov,N.
A.
,O.
Torres,C.
Seftor,A.
J.
Krueger,A.
Kostinski,W.
I.
Rose,G.
J.
S.
Bluth,D.
Schneider,andS.
J.
Schaefer,ComparisonofTOMSandAVHRRvolcanicashretrievalsfromtheAugust1992eruptionofMt.
Spurr,Geophys.
Res.
Lett.
26,455-458,1999a.
Krotkov,N.
A.
,D.
E.
Flittner,A.
J.
Krueger,A.
Kostinski,C.
Riley,W.
Rose,andO.
Torres,Effectofparticlenon-sphericityonsatellitemonitoringofdriftingvolcanicashclouds,J.
Quant.
Spectrosc.
Radiat.
Transfer63,613-630,1999b.
Krueger,A.
J.
,SightingofElChichonsulfurdioxidecloudswiththeNimbus7TotalOzoneMappingSpectrometer,Science220,1377-1379,1983.
Krueger,A.
J.
,L.
S.
Walter,P.
K.
Bhartia,C.
C.
Schnetzler,N.
A.
Krotkov,I.
Sprod,andG.
J.
S.
Bluth,VolcanicsulfurdioxidemeasurementsfromtheTotalOzoneMappingSpectrometerinstruments,J.
Geophys.
Res.
100,14,057-14,076,1995.
Krueger,A.
J.
,S.
J.
Schaefer,N.
Krotkov,G.
Bluth,andS.
Barker,UltravioletRemoteSensingofVolcanicEmissions,inRemoteSensingofActiveVolcanism,ed.
P.
MouginisMark,J.
A.
60ATBD-OMI-04Crisp,andJ.
H.
Fink,GeophysicalMonograph116,AmericanGeophysicalUnion,Washington,DC,2000.
Manatt,S.
L.
andA.
L.
Lane,Acompilationoftheabsorptioncross-sectionsofSO2from106to403nm,,J.
Quant.
Spectrosc.
Radiat.
Transfer,50,267-276,1993.
McGee,T.
J.
andJ.
BurrisJr,SO2absorptioncrosssectionsinthenearUV,J.
Quant.
Spectrosc.
Radiat.
Transfer,37,165-182,1987.
Marquardt,D.
L.
,Analgorithmforleast-squaresestimationofnon-linearparameters,J.
Soc.
Indust.
Appl.
Math.
2,431-441,1963.
ATBD-OMI-04615.
BrOK.
Chance,T.
P.
Kurosu,andL.
S.
RothmanSmithsonianAstrophysicalObservatoryCambridge,MA,USABrOwasfirstmeasuredfromspacebyGOME,intheregion344-360nm.
WhileitwasanticipatedthatBrOcouldbemeasuredglobally[Chanceetal.
,1991],itwasalsothoughtthatBrOwouldbeofinterestprimarilyasastratosphericgas.
However,lowertroposphericozonedestructionintheArcticpolarsunrisehasbeencoupledwithbrominechemistryassociatedwiththeicepack[Barrieetal.
,1988].
ER-2observationsshowthepresenceofenhancedBrOinthefreetroposphereduringtheArcticpolarsunrise[McElroyetal.
,1999],andGOMEmeasurementshavenowconfirmedandfurtherquantifiedenhancementsinBrOinbothArcticandAntarcticspring[WagnerandPlatt,1998].
Figure5.
1showsanexampleofenhancedtroposphericBrOovertheicepackintheNorthernhemispherelatespringin1997.
OMImeasurementsofBrOwillmakesuchobservationsathigherspatialresolutionthat,coupledwithclouddetermination,willpermitthelocationandpersistenceofenhancedpolartroposphericBrOtobestudiedinsynergywithtroposphericO3inordertoquantifytheeffectsontroposphericO3.
Figure5.
1ThedistributionofBrO(verticalcolumnincm-2)intheNorthernhemisphereforApril30-May2,1997[Chance,1998].
Largeenhancements,almostcertainlyduetotroposphericBrO,areobviousoverseveralareas,includingHudsonBayandtheArcticiceshelf.
62ATBD-OMI-04ThehigherspatialresolutionofOMIwillalsopermitdetailedobservationstobemadeoftherelationofBrOtothepolarvortexstructure.
AbsorptionsforBrOarequitesmall(substantiallylessthan1%inmostcases),sothattheyareopticallythintoahighdegreeofaccuracyandsothatinterferencesfromothercauses(e.
g.
,O3absorptionandtheRingeffect)mustbeaccountedforveryprecisely.
ThefittingforBrOincludestwomajorsteps:(1)thefittingofaselectedwavelengthwindowofspectrumtodeterminetheslantcolumndensity,Nsi,foraparticularspeciesi,and(2)thedeterminationofanappropriateairmassfactor,Mi,toconvertNsitoaverticalcolumndensity,Nvi:M=Nsi/Nvi,[5-1]whereMisafunctionofviewinggeometry,geophysicalcondition(albedo,cloudcoverage),andtheverticaldistributionofthegas.
Generally,gaslocatedatloweraltitudecontributeslesstothesatellite-measuredabsorptionspectrumsinceRayleighscatteringdiscriminatesagainstitsviewing[Palmeretal.
,2001].
Algorithmsaredesignedassumingsequentialprocessingoflevel1dataproducts,includingspectra,sothathigher-levelprocessingwillberequiredinsomecasestofullyexploitthemeasurements.
Thefirststepisaccomplishedbyaspectralfittingprocedure,whichwillbeoptimizedduringthecommissioningphaseoftheOMIinstrument,andwherevariousoptionsdescribedbelowareevaluated.
Thesecondstepismoreproblematicforsomeconditions.
BrOisnormallypredominantlylocatedinthestratosphereanduppertroposphere,andairmassfactorspre-tabulatedversusmeasurementgeometryfornominalgeophysicalconditionsandprofileshapegiveresultscorrecttowithintheuncertaintiesfromstep(1).
However,duringthepolarspringthereareenhancementsintroposphericBrO[WagnerandPlatt,1998;Richteretal.
,1998;Hegelsetal.
,1998;Chance,1998].
Forthese,theMscalculatedassumingstratospheric/uppertroposphereBrOsignificantlyunderestimatethetroposphericconcentration(byaboutafactoroftwo).
Itisanticipatedthatcorrectionforthesecaseswilloccurinhigher-levelprocessingratherthandynamicallyasapartofthelevel1-2operationprocessing,sinceeffectivediscriminationofthesecasesinvolvesprocessingthatisnotsequentialinanalysis.
Additionally,thereissomeevidencethatfreetroposphericconcentrationsofBrOmaybelargerthananticipated[Fitzenbergeretal.
,2000].
Ifthisprovestobethecase,appropriatechangestothealgorithmwillberequiredinordertoappropriatelyweightthecontributiontothemeasuredspectrafromBrOatthesealtitudes.
5.
1.
SlantcolumnmeasurementsThedeterminationofslantcolumnabundances,Ns,isaccomplishedbyfittingthemeasuredradianceI,beginningwiththemeasuredirradianceE,molecularabsorptioncrosssections,correctionfortheRingeffect,effectivealbedo(whichincludesthecontributionfromRayleighscatteringforthesemolecules,asdiscussedinAirmassfactorsandverticalcolumnabundances,below),andalow-orderpolynomialforclosure.
ThislattertermaccountsforsmallremainingdifferencesinRayleighscatteringversuswavelengthoverthefittingwindow,variationofgroundalbedo,andimperfectintensitycalibrationoftheOMIradianceandirradiancemeasurements.
Theoverallfittingstrategyincludesanumberofoptions,whichwillbefullytestedduringOMIcommissioning,withtheonesthathaveprovedmostsuccessfulintheanalysisofprevioussatellitemeasurementsprovidingthebaseline.
5.
1.
1.
Nonlinearleast-squaresfittingTheLevenberg-Marquardtnonlinearleast-squaresfittingprocedure(nlls)[Marquardt,1963;Pressetal.
,1986]isusedinseveralofthesubsequentstepsintheanalysis.
Inthisprocedure,theχ2meritfunctionATBD-OMI-0463χσ2==)yFxaiiiiN/12[5-2]isminimizedwithrespecttotheparametersa.
Thestrategyforfindingtheminimumistobeginwithadiagonally-dominantcurvaturematrix,correspondingtoasteepestdescentsearchprocedure,andgraduallychangingcontinuouslyovertotheinverse-Hessian(curvature)methodsearchprocedureastheminimumisneared.
Convergenceisreachedwhenχ2islessthanapre-setamount,whenχ2decreasesbylessthanapre-setamountoverseveralsuccessiveiterations,orwhenallparameterschangebylessthanapre-setfractionforseveralsuccessiveiterations.
Iterationisalsohaltedwhenthenumberofiterationsreachesapre-setmaximumwithoutsuccessfulconvergence.
Estimatedfittinguncertaintiesaregivenasσi=iiC,whereCisthecovariancematrixofthestandarderrors.
Thisdefinitionisstrictlytrueonlywhentheerrorsarenormallydistributed.
Inthecasewherethelevel-1dataproductuncertaintiesarenotreliableestimatesoftheactualuncertainties,spectraldataaregivenunityweightoverthefittingwindow,andthe1σfittingerrorinparameteriisdeterminedasvariedpointspointsiirmsinnnC=εσ[5-3]whereεrmsistheroot-mean-squaredfittingresidual,npointsisthenumberofpointsinthefittingwindow,andnvariedisthenumberofparametersvariedduringthefitting.
5.
1.
2.
Re-calibrationofwavelengthscalesThisisabaselineoption,requiredbythefactthatfittingtosmallroot-mean-squaredifferencesbetweenthemeasurementsandthemodeling(rms),comparabletothemeasurementsignal-to-noiseratios(SNRs)requiresbetterwavelengthcalibrationthanthatprovidedinthelevel1dataproducts(wavelengthcalibrationforthespecificfittingwindowismoreaccuratethanthatderivedforthespectrumasawhole).
ThemodelforthisprocedurecomesfromtheanalysisofGOMEdata[CasparandChance,1997;Chance,1998].
Theirradiancetobeusedinthesubsequentspectrumfittingisre-calibratedinwavelengthoverthefittingwindowbynllscomparisontoahigh-resolutionsolarreferencespectrumwhichisaccurateinabsolutevacuumwavelengthtobetterthan0.
001nm[ChanceandSpurr,1997].
Aslitwidth(instrumenttransferfunction)parameterisfittedsimultaneously.
Radiancespectraareequivalentlyfitted,withtheslitwidthparameterfrozentothevalues(versusCCDspectralfield)determinedinfittingtheirradiance:SincethewindowregionsforthesefourmoleculesareopticallythininallTelluricabsorptionsandcontainonlyafewpercentofinelastically-scatteredFraunhoferspectrum(Ringeffect),theprocedureworksalmostaswellonradiancesasonirradiances.
ExperiencewithGOMEspectra(whichareathigherspectralresolution)isthatbotharefittedtowithinabout1/50spectralresolutionelement.
Re-calibrationnormallyinvolvesonlythedeterminationofasinglewavelengthshiftparameterforthefittingwindow(baselineoption);inclusionofawavelength``squeeze''parameterisanon-baselineoption.
ThecaseforOMIiscomplicatedbythefactthatsuchcalibrationsaremadefortheseparatespectralfieldsontheCCDdetectorarray(GOMEhaslinearReticondetectors,whichmeasuresinglespectra).
64ATBD-OMI-04WavelengthcalibrationsaremadeforeachOMIorbitasfollows:1.
ThesetofirradiancesversusCCDpositionarecalibratedinwavelengthoverthefittingwindow;2.
OnesetofradiancesversusCCDposition,selectedbyaninputparameter,usuallyinthemiddleoftheorbit,arecalibratedinwavelengthoverthefittingwindow.
Thiscalibrationisappliedastheinitialwavelengthscaleforallradiancesintheorbit;3.
Furtherfine-tuningoftherelativecalibrationofallradiancesthroughtheorbittotheirradianceisperformedduringthedetailedspectrumfitting,asdescribedbelow(baselineoption).
5.
1.
3.
ReferencespectraReferencespectraaredegradedtotheOMIresolutioneitherinpre-tabulatedform(baselineoption)orusingtheparameterizedslitfunctiondeterminedduringtheirradiancecalibration(non-baselineoption).
Theyarethenre-sampledtotheradiancewavelengthgrid,usingcubicsplineinterpolation[Pressetal.
,1986].
ThecurrentbaselinechoicesforreferencespectratofitBrOareshowninFigure5.
2:NO2crosssectionsVandaeleetal.
,1997.
BrOcrosssectionsWilmouthetal.
,1999.
OClOcrosssectionsWahneretal.
,1987(correctedtovacuumwavelength).
O3crosssections(neededtocorrectforspectralinterference)Burrowsetal.
,1999.
O2-O2collisioncomplexGreenblattetal.
,1990(correctedtovacuumwavelength).
Ringeffect:DeterminedspecificallyforOMIapplicationsbyJ.
Joineretal.
Undersamplingcorrection:Calculateddynamically,ifrequiredFittingofGOMEspectragavesystematicresidualsthatweremuchlargerthantheabsorptionofthetracegasesuntilitwasdiscoveredthattheseareduetospectralundersamplingbytheinstrument.
Highfrequencyspectralinformationisaliasedintothespectrumwhenirradiancesarere-sampledtotheradiancewavelengthscale(irradiancesforGOME,andforOMI,aremeasuredatslightlydifferentDopplershiftswithrespecttothesun)[Chance,1998;Slijkhuisetal.
,1999].
ThealgorithmforOMIincludesanon-baselineoption(sinceOMIisnotanticipatedtosignificantlyundersample)forcalculatingundersamplingcorrectionsatthetimeofreferencespectrumsampling.
Thisisaccomplishedbytakingthehigh-resolutionFraunhoferreferencespectrum,convolvedwiththeinstrumenttransferfunction(asdeterminedduringtheirradiancespectralcalibration),whichconstitutesanoversampledirradianceandformingproperly-sampledandundersampledrepresentationsofit,usingcubicsplineinterpolation.
Thedifferencebetweentherepresentationistheundersamplingcorrection;forGOMEthisconstitutesca.
90%ofthesystematicresidual.
5.
1.
4.
Common-modecorrectionRemainingsystematicresidualswhichare,bydefinition,uncorrelatedtothetracegasspectramaybeaveragedandincludedinthespectrumfittingasa``common-mode''spectrum,toreducethefittingrmsand,proportionally,thefittinguncertainties.
Thisisincludedasanon-baselineoptionforOMI.
ForGOME,thisreducesthermsbyaboutafactoroftwoforBrOandafactorof3forNO2.
ThepresumptionisthattheundersamplingcorrectiononlyimperfectlyATBD-OMI-0465modelstheinstrumenttransferfunctionandthusleavessomeremainingsystematiceffects.
ComparisonswithGOMEmeasurementsofNO2forcleanmaritime(mostlystratosphericNO2)conditionsshowthatthisreduceduncertaintyistheappropriatechoice.
Figure5.
2ReferencespectraforfittingtodetermineslantcolumnabundancesofBrO:(a)BrO;(b)NO2(c)OClO;(d)O3;(e)O2-O2collisioncomplex;(f)Ringeffect.
5.
1.
5.
Radiancefitting:BOAS(baselineoption)Theradianceisnllsfittedtoamodeledspectrumwhichincludestheirradiance,andeffectivealbedo,tracegasconcentrations,Ringeffectcorrection,andalow-orderpolynomial(uptocubic)closureterm:IAEeecccccNNRRssnn()()λλσλλλλλλσλσλ11012233L[5-4]Theσiaretheabsorptioncrosssections,σRistheRingeffectcorrection,andthec0-c3arethecoefficientsoftheclosurepolynomial.
Eachparametermaybeindividuallyselectedtobevariedduringthefittingorfrozenataconstantvalue.
ThebaselineforBrOistovaryA,Ns66ATBD-OMI-04(BrO),Ns(NO2),Ns(O3),Ns(O2-O2),cR,andc0–c3.
Inaddition,thelinearwavelengthscaleoftheirradianceisallowedtovary(baselineoption)tocorrectforsmallwavelengthchangesintheradianceovertheorbit.
Whileitisactuallytheradiancescalethatrequiresadjustment,therelativechangeinwavelengthscaleismuchsmallerthantheinstrumentresolutionorthevariationofmeasurablefeaturesinthereferenceorRingspectra:adjustmentoftheirradiancescaleavoidschangingthemeasuredquantity(theradiance)duringthefittingprocess.
Thebaselinefittingwindowis345-359nm.
Smoothing(low-passfiltering)oftheirradiances,radiances,andreferencespectraisincludedasanon-baselineoption.
Thisprocedureprovidesanalternatemeanstocorrectforspectralundersamplingeffects.
Itisimplementedbyapplyingarunning5-pointfilter(1/16,1/4,3/8,1/4,1/16)toeachofthespectra.
Updatingofthefittingparametersisincludedasabaselineoption:ForeachspectrumafterthefirstinaparticularspectrumfieldoftheCCD,theinitialguessesforfittingparametersareupdatedtothosefittedtothepreviousspectrum.
5.
1.
6.
Radiancefitting:DOAS(non-baselineoption)Thisoptionconvertsthedirect(BOAS)fittingofradiancestotheDOASmethod.
Thelogarithmoftheradiancedividedbytheirradianceistakenandtheresulthigh-passfilteredbythesubtractionofalow-orderpolynomial(uptocubic,withtheactualnumberoftermsdeterminedbytheinputparameterfile):H[ln(I/E)]=-Ns1H(σ1)NsnH(σn)+cR1H(σR1/F)+cR2H(σR2/F)+HOT,[5-5]whereHdenoteshigh-passfiltering.
Thecoefficientsofthepolynomialaredeterminedbylinearfittingtoln(I/E)[Pressetal.
,1986].
ReferencecrosssectionssampledattheOMIwavelengthscalearealsohighpassfilteredbysubtractionofalow-orderpolynomial.
TheRingeffectcorrectionisdividedbytheirradianceforeachspectralfieldandtheresulthigh-passfilteredaswell(notethatthisdeterminesthefirsttermintheexpansionforthelogarithmicquantitywhichincludestheRingeffect;higher-ordertermsarenotincluded-seeEquation[1-4]).
ThefittingisthenentirelyanalogoustotheBOASfitting,exceptthatthefittedquantityisnowH[ln(I/E)].
Iftheradiance/irradiancewavelengthadjustmentisnotselected,thentheDOASfittingwouldbelinearinthefittingparameters.
However,sincewavelengthadjustmentandBOASfittingarebothbaselineoptions,itisnotcurrentlyplannedtoimplementseparatelinearfittingforthiscase;theadditionalcomputertimeinfittingthelinearcasewiththenonlinearmethodisinconsequential,andtheresultsarevirtuallyidentical.
5.
2.
AirmassfactorsandverticalcolumnabundancesValuesofMarecalculatedversusviewinggeometryandgroundalbedousingtheLIDORTmultiplescatteringradiativetransfermodel[Spurretal.
,2001].
LIDORTisdesignedtodeliverbothintensitiesandweightingfunctions.
LIDORTsolvestheradiativetransferequationinamulti-layeredatmospherewithmultiple-scatteringusingthediscreteordinatemethod.
Themodelcontainsafullinternalperturbationanalysisoftheintensityfield,allowingallweightingfunctionstobederivedanalyticallytothesamelevelofaccuracyspecifiedfortheintensity.
AlthoughLIDORThasbeendesignedprimarilyasageneralforwardmodeltoolfornon-linearatmosphericretrievalproblems,thecalculationofMsisastraightforwardapplicationofthemodel.
AsinglecalltoLIDORTwilldeliverboththetop-of-the-atmosphereupwellingintensityIBandthecorrespondingsetofweightingfunctionsK(z)requiredfortheMdetermination.
Inordertosimplifythelookupprocess,theviewinggeometryisparameterizedbytheeffectivesolarzenithangle(ESZA),wheresec(ESZA)=sec(SZA)+sec(LOSZA)–1.
SZAATBD-OMI-0467isthesolarzenithangleofthemeasurementandLOSZAistheline-of-sightzenithangle,andazimuthaldependenceofMisignored.
ExperiencewithfittingGOMEdatashowsthat,exceptforthehighestsolarzenithangles,above80o,thisprocedureaddsnegligibleadditionalerror.
BrOisassumedtobeprimarilyinthestratosphereanduppertroposphere,andisdescribedbyaGaussianverticalconcentrationprofileshapewithacenterat19kmandahalfwidthat1/eintensityof9km(takenfromBrasseurandSolomon,1986).
Calculationshavebeenmadeforarangeofalbedovaluesfrom0.
01to1.
0.
Resultsforthealbedorange0.
01-0.
30varybylessthan2%fromthoseforalbedo0.
1,sothisisselectedasthebaseline.
Foralbedo=1.
0,theerrorcanbeaslargeas6%.
Calculationforthelong-andshort-wavelengthsidesofthefittingwindowandthemiddlevarybylessthan3%.
Anaverageofthreevalues(long,short,andmiddle)isusedforthetabulationofMs.
Apre-calculatedgridofMvaluesversusESZAisusedinalookupschemewheretheindividualMsaredeterminedbycubicsplineinterpolation.
ThevariationsofMwithalbedoandwithwavelengthoverthefittingwindowareincludedaserrorterms.
AnadditionalESZA-dependenterrorcontributionofupto5%isincludedtoaccountforthevariationoftheprofileofstratospheric/uppertroposphericBrOfromtheselectedshape.
ThedepartureofMfromthegeometricvalue(1+sec(ESZA))forstratosphericBrOisquitesmall,≤3%forvaluesofESZAupto70o.
CorrectionfortroposphericBrOisnotincludedinthisalgorithm.
IdentificationofpolarspringenhancementsandcorrectiontotheMsfortheiranalysiswilltakeplaceinhigher-levelprocessing.
AccountingforhigherlevelsoffreetroposphericBrOwillawaitfurtherconfirmationofitsexistence.
5.
3.
ErrorestimatesEstimatederrorsaregivenforagroundfootprintof40·40km2,whichhasbeenadoptedasthestandardforreportingOMIerrorestimateswithinEOS.
S/NvaluesforestimatingfittinguncertaintiescomefromtheOMI-EOSInstrumentSpecificationDocumentRS-OMIE-0000-FS-021,Tables5.
5and5.
6,andareappliedtohebestfittingknowledgefromGOMEanalysis,takingintoaccountthedifferenceinspectralresolution.
Alluncertaintiesaregivenhereas1σ.
The1σfittingerrorinparameteriisdeterminedasσi=iiC[5-6]orvariedpointspointsiirmsinnnC=εσ,[5-7]AsdiscussedinSection5.
1.
1;bothdefinitionsincludethecontributionsfromcorrelationoffittedparameters.
Cisthecovariancematrixofthestandarderrors.
Thepublisheduncertaintyoftheabsorptioncrosssectionsisaddedtothisinquadraturetoobtainthefinalslantcolumnfittinguncertainties.
TheuncertaintyinMisaddedinquadraturetoobtainthefinalverticalcolumnfittinguncertainties.
TheglobalaveragefittinguncertaintyforBrOis11%.
Theuncertaintyincrosssectionsis7%.
ForstratosphericanduppertroposphericBrO,theuncertaintyduetoassumingastandardverticalprofileshapeis<1%forESZAvaluesupto70oand<4%upto80o.
Thisincludeserrorfromalbedoandclouduncertainty,whichisaminorcontributor.
TotalerrorforESZAvaluesupto80o,assumingupperstratosphericandtroposphericBrO,isthequadraturesumof11%+7%+4%=14%.
Theerrorexcludingthesystematic68ATBD-OMI-04contributionfromcrosssectionsuncertaintiesis12%,correspondingto2.
0·1013cm-2verticalcolumndensity.
ThiscanbecomparedtotherequirementintheScienceRequirementsDocumentforOMI-EOS(RS-OMIE-KNMI-001,Version2)of1013cm-2;meetingthisrequirementwillrequireanimprovementofafactorof2intheS/NoverthatgivenintheInstrumentSpecificationDocument.
ThepresenceofsubstantialBrOlowerinthetroposphereandintheplanetaryboundarylayerwillrequireadditionalanalysisinordertocorrectlyquantifytheabundance.
Inparticular,thecontributiontoMfromboundarylayerBrOcaneasilybeoffbyafactoroftwofromthatderivedwiththepresentassumptions.
5.
4.
OutputsStandardoutputsinclude:Slantcolumnabundanceand1σfittinguncertaintiesforBrOandtheotherspeciesvariedinthefittingwindow;CorrelationofotherfittedspeciestoBrO(fromoff-diagonalelementsofthecovariancematrixofthestandarderrors);Fittingrms;Convergenceflagsandnumberofiterations(successfulwhichconvergencecriterion)Geolocationinformation;Versionnumbersofalgorithmandparameterinputfile;Verticalcolumnabundancesand1σuncertainties.
5.
5.
ValidationGround-basedUV/visibleBrOcolumnmeasurementsprovidetheprimaryvalidationsource.
ER-2measurementsoftheuppertroposphericabundanceaswellasvertically-integratedOClOprofilesfromballoon-basedUV/visibleSAOZinstrumentswillalsoproveuseful.
Ground-basedmeasurementsshouldincludehigh-latitudelocationswhereconditionsinsidethepolarvortexmaybesampledandwhereenhancedboundarylayertroposphericBrOeventsoccur.
5.
6.
ReferencesBarrie,L.
A.
,J.
W.
Bottenheim,R.
C.
Schnell,P.
J.
Crutzen,andR.
A.
Rasmussen,OzonedestructionandphotochemicalreactionsatpolarsunriseinthelowerArcticatmosphere,Nature334,138-141,1988.
Brasseur,G,andS.
Solomon,AeronomyoftheMiddleAtmosphere,Secondrevisededition,ISBN90-277-2343-5,D.
Reidel,Dordrecht,1986,p.
284.
Burrows,J.
P.
,A.
Richter,A.
Dehn,B.
Deters,S.
Himmelmann,S.
Voigt,andJ.
Orphal,AtmosphericremotesensingreferencedatafromGOME-2.
Temperature-dependentabsorptioncross-sectionsofO3inthe231-794nmrange,J.
Quant.
Spectrosc.
Radiat.
Transfer61,509-517,1999.
Caspar,C.
,andK.
Chance,GOMEwavelengthcalibrationusingsolarandatmosphericspectra,Proc.
ThirdERSSymposiumonSpaceattheServiceofourEnvironment,Ed.
T.
-D.
GuyenneandD.
Danesy,EuropeanSpaceAgencypublicationSP-414,ISBN92-9092-656-2,1997.
ATBD-OMI-0469Chance,K.
V.
,J.
P.
Burrows,andW.
Schneider,RetrievalandmoleculesensitivitystudiesfortheGlobalOzoneMonitoringExperimentandtheSCanningImagingAbsorptionspectroMeterforAtmosphericCHartographY,Proc.
S.
P.
I.
E.
,RemoteSensingofAtmosphericChemistry,1491,151-165,1991.
Chance,K.
,andR.
J.
D.
Spurr,RingEffectStudies:RayleighScattering,IncludingMolecularParametersforRotationalRamanScattering,andtheFraunhoferSpectrum,Appl.
Opt.
36,5224-5230,1997.
Chance,K.
,AnalysisofBrOMeasurementsfromtheGlobalOzoneMonitoringExperiment,Geophys.
Res.
Lett.
25,3335-3338,1998.
Fitzenberger,R.
,H.
Boesch,C.
Camy-Peyret,M.
P.
Chipperfield,H.
Harder,U.
Platt,B.
-M.
Sinnhuber,T.
Wagner,andK.
Pfeilsticker,FirstprofilemeasurementsoftroposphericBrO,Geophys.
Res.
Lett.
27,2921-2924,2000.
Greenblatt,G.
D.
,J.
J.
Orlando,J.
B.
Burkholder,andA.
R.
Ravishankara,Absorptionmeasurementsofoxygenbetween330and1140nm,J.
Geophys.
Res.
95,18,577-18,582,1990.
Hegels,E.
,P.
J.
Crutzen,T.
Kluepfel,D.
Perner,andJ.
P.
Burrows,GlobaldistributionofatmosphericbrominemonoxidefromGOMEonEarthObservingSatelliteERS-2,Geophys.
Res.
Lett.
25,3127-3130,1998Marquardt,D.
L.
,Analgorithmforleast-squaresestimationofnon-linearparameters,J.
Soc.
Indust.
Appl.
Math.
2,431-441,1963.
McElroy,C.
T.
,C.
A.
McLinden,andJ.
C.
McConnell,EvidenceforbrominemonoxideinthefreetroposphereduringtheArcticpolarsunrise,Nature397,338-341,1999.
Palmer,P.
I.
,D.
J.
Jacob,K.
Chance,R.
V.
Martin,R.
J.
D.
Spurr,T.
P.
Kurosu,I.
Bey,R.
Yantosca,A.
Fiore,andQ.
Li,AirMassFactorFormulationforSpectroscopicMeasurementsfromSatellites:ApplicationtoFormaldehydeRetrievalsfromGOME,J.
Geophys.
Res.
106,14,539-14,550,2001.
Press,W.
H.
,B.
P.
Flannery,S.
A.
Teukolsky,andW.
T.
Vetterling,NumericalRecipes,ISBN0-521-30811-9,CambridgeUniversityPress,1986.
Richter,A.
,F.
Wittrock,M.
Eisinger,andJ.
P.
Burrows,GOMEobservationsoftroposphericBrOinnorthernhemispherespringandsummer1997,Geophys.
Res.
Lett.
25,2683-2686,1998Slijkhuis,S.
,A.
vonBargen,W.
Thomas,andK.
Chance,CalculationofundersamplingcorrectionspectraforDOASspectralfitting,Proc.
ESAMS'99-EuropeanSymposiumonAtmosphericMeasurementsfromSpace,563-569,1999.
Spurr,R.
J.
D.
,T.
P.
Kurosu,andK.
Chance,Alinearizeddiscreteordinateradiativetransfermodelforatmosphericremotesensingretrieval,J.
Quant.
Spectrosc.
Radiat.
Transfer68,689-735,2001.
VandaeleA.
C.
,C.
Hermans,P.
C.
Simon,M.
Carleer,R.
Colin,S.
Fally,M.
F.
Mérienne,A.
Jenouvrier,andB.
Coquart,MeasurementsoftheNO2absorptioncross-sectionfrom42000cm-1to10000cm-1(238-1000nm)at220Kand294K,J.
Quant.
Spectrosc.
Radiat.
Transfer59,171-184,1998.
Wagner,T.
,andU.
Platt,SatellitemappingofenhancedBrOconcentrationsinthetroposphere,Nature395,486-490,1998.
Wahner,A.
,G.
S.
Tyndall,andA.
R.
Ravishankara,AbsorptioncrosssectionsforOClOasafunctionoftemperatureinthewavelengthrange240-480nm,J.
Phys.
Chem.
91,2734-2738,1987.
Wilmouth,D.
M.
,T.
F.
Hanisco,N.
M.
Donahue,andJ.
G.
Anderson,FouriertransformultravioletspectroscopyoftheA2Π3/2←X2Π3/2transitionofBrO,J.
Phys.
Chem.
103,8935-8945,1999.
70ATBD-OMI-04ATBD-OMI-04716.
OClOK.
Chance,T.
P.
Kurosu,andL.
S.
RothmanSmithsonianAstrophysicalObservatoryCambridge,MA,USAOClOisausefulindicatorforchlorineactivationinthestratosphere.
ItwasfirstmeasuredfromspacebyGOME,inthe363-393nmrange.
TowithincurrentGOMEmeasurementandretrievaluncertainties,itisfoundentirelywithinthepolarvortices.
Becauseofthehighsolarzenithanglesassociatedwiththeoccurrence,measurementstodatehavebeenlimitedtoline-of-sightslantcolumnmeasurements[Wagneretal.
,2001].
Inadditiontocontinuingthemeasurementrecord,providinganindicatorforthetrendinstratosphericchlorineloading,OMIwillmeasureOClOathigherspatialresolutionthanGOME,whichwillhelptoelucidatethedetailsofitsformation,persistence,andcorrelationwithPSCsandtemperature.
AbsorptionsforOClOarequitesmall(substantiallylessthan1%inmostcases),sothattheyareopticallythintoahighdegreeofaccuracyandsothatinterferencesfromothercauses(e.
g.
,O3absorptionandtheRingeffect)mustbeaccountedforveryprecisely.
Thefittingfortracegasesusuallyincludestwomajorsteps:(1)thefittingofaselectedwavelengthwindowofspectrumtodeterminetheslantcolumndensity,Nsi,foraparticularspeciesi,and(2)thedeterminationofanappropriateairmassfactor,Mi,toconvertNsitoaverticalcolumndensity,Nvi:M=Nsi/Nvi,[6-1]whereMisafunctionofviewinggeometry,geophysicalcondition(albedo,cloudcoverage),andtheverticaldistributionofthegas.
ForOClO,thedataproductsareslantcolumndensities,sothatthissecondstepisnotinvoked.
ThisisduetothefactthatOClOisgenerallyfoundatveryhighsolarzenithangles,sothatthedeterminationofappropriateairmassfactorsforagivencaserequiressubstantialoff-lineanalysis.
Theslantcolumndeterminationisaccomplishedbyaspectralfittingprocedure,whichwillbeoptimizedduringthecommissioningphaseoftheOMIinstrument,andwherevariousoptionsdescribedbelowareevaluated.
6.
1.
SlantcolumnmeasurementsThedeterminationofslantcolumnabundances,Ns,isaccomplishedbyfittingthemeasuredradianceI,beginningwiththemeasuredirradianceE,molecularabsorptioncrosssections,correctionfortheRingeffect,effectivealbedo,andalow-orderpolynomialforclosure.
ThislattertermaccountsforsmallremainingdifferencesinRayleighscatteringversuswavelengthoverthefittingwindow,variationofgroundalbedo,andimperfectintensitycalibrationoftheOMIradianceandirradiancemeasurements.
Theoverallfittingstrategyincludesanumberofoptions,whichwillbefullytestedduringOMIcommissioning,withtheonesthathaveprovedmostsuccessfulintheanalysisofprevioussatellitemeasurementsprovidingthebaseline.
6.
1.
1.
Nonlinearleast-squaresfittingTheLevenberg-Marquardtnonlinearleast-squaresfittingprocedure(nlls)[Marquardt,1963;Pressetal.
,1986]isusedinseveralofthesubsequentstepsintheanalysis.
Inthisprocedure,theχ2meritfunction72ATBD-OMI-04χσ2==)yFxaiiiiN/12[6-2]isminimizedwithrespecttotheparametersa.
Thestrategyforfindingtheminimumistobeginwithadiagonally-dominantcurvaturematrix,correspondingtoasteepestdescentsearchprocedure,andgraduallychangingcontinuouslyovertotheinverse-Hessian(curvature)methodsearchprocedureastheminimumisneared.
Convergenceisreachedwhenχ2islessthanapre-setamount,whenχ2decreasesbylessthanapre-setamountoverseveralsuccessiveiterations,orwhenallparameterschangebylessthanapre-setfractionforseveralsuccessiveiterations.
Iterationisalsohaltedwhenthenumberofiterationsreachesapre-setmaximumwithoutsuccessfulconvergence.
Estimatedfittinguncertaintiesaregivenasσi=iiC,whereCisthecovariancematrixofthestandarderrors.
Thisdefinitionisstrictlytrueonlywhentheerrorsarenormallydistributed.
Inthecasewherethelevel-1dataproductuncertaintiesarenotreliableestimatesoftheactualuncertainties,spectraldataaregivenunityweightoverthefittingwindow,andthe1σfittingerrorinparameteriisdeterminedasvariedpointspointsiirmsinnnC=εσ[6-3]whereεrmsistheroot-mean-squaredfittingresidual,npointsisthenumberofpointsinthefittingwindow,andnvariedisthenumberofparametersvariedduringthefitting.
6.
1.
2.
Re-calibrationofwavelengthscalesThisisabaselineoption,requiredbythefactthatfittingtosmallroot-mean-squaredifferencesbetweenthemeasurementsandthemodeling(rms),comparabletothemeasurementSNRsrequiresbetterwavelengthcalibrationthanthatprovidedinthelevel1dataproducts(wavelengthcalibrationforthespecificfittingwindowismoreaccuratethanthatderivedforthespectrumasawhole).
ThemodelforthisprocedurecomesfromtheanalysisofGOMEdata[CasparandChance,1997;Chance,1998].
Theirradiancetobeusedinthesubsequentspectrumfittingisre-calibratedinwavelengthoverthefittingwindowbynllscomparisontoahigh-resolutionsolarreferencespectrumwhichisaccurateinabsolutevacuumwavelengthtobetterthan0.
001nm[ChanceandSpurr,1997].
Aslitwidth(instrumenttransferfunction)parameterisfittedsimultaneously.
Radiancespectraareequivalentlyfitted,withtheslitwidthparameterfrozentothevalues(versusCCDspectralfield)determinedinfittingtheirradiance:SincethewindowregionsforthesefourmoleculesareopticallythininallTelluricabsorptionsandcontainonlyafewpercentofinelastically-scatteredFraunhoferspectrum(Ringeffect),theprocedureworksalmostaswellonradiancesasonirradiances.
ExperiencewithGOMEspectra(whichareathigherspectralresolution)isthatbotharefittedtowithinabout1/50spectralresolutionelement.
Re-calibrationnormallyinvolvesonlythedeterminationofasinglewavelengthshiftparameterforthefittingwindow(baselineoption);inclusionofawavelength"squeeze"parameterisanon-baselineoption.
ThecaseforOMIiscomplicatedbythefactthatsuchcalibrationsaremadefortheseparatespectralfieldsontheCCDdetectorarray(GOMEhaslinearReticondetectors,whichmeasuresinglespectra).
WavelengthcalibrationsaremadeforeachOMIorbitasfollows:ATBD-OMI-04731.
Thesetofirradiances,versusCCDpositionarecalibratedinwavelengthoverthefittingwindow;2.
OnesetofradiancesversusCCDposition,selectedbyaninputparameter,usuallyinthemiddleoftheorbit,arecalibratedinwavelengthoverthefittingwindow.
Thiscalibrationisappliedastheinitialwavelengthscaleforallradiancesintheorbit;3.
Furtherfine-tuningoftherelativecalibrationofallradiancesthroughtheorbittotheirradianceisperformedduringthedetailedspectrumfitting,asdescribedbelow(baselineoption).
6.
1.
3.
ReferencespectraReferencespectraaredegradedtotheOMIresolutioneitherinpre-tabulatedform(baselineoption)orusingtheparameterizedslitfunctiondeterminedduringtheirradiancecalibration(non-baselineoption).
Theyarethenre-sampledtotheradiancewavelengthgrid,usingcubicsplineinterpolation[Pressetal.
,1986].
ThecurrentbaselinechoicesforreferencespectratofitOClOareShowninFigure6.
1:NO2crosssectionsVandaeleetal.
,1997.
BrOcrosssectionsWilmouthetal.
,1999.
OClOcrosssectionsWahneretal.
,1987(correctedtovacuumwavelength).
O3crosssections(neededtocorrectforspectralinterference)Burrowsetal.
,1999.
O2-O2collisioncomplexGreenblattetal.
,1990(correctedtovacuumwavelength).
Ringeffect:DeterminedspecificallyforOMIapplicationsbyJ.
Joineretal.
Undersamplingcorrection:Calculateddynamically,ifrequiredFittingofGOMEspectragavesystematicresidualsthatweremuchlargerthantheabsorptionofthetracegasesuntilitwasdiscoveredthattheseareduetospectralundersamplingbytheinstrument.
Highfrequencyspectralinformationisaliasedintothespectrumwhenirradiancesarere-sampledtotheradiancewavelengthscale(irradiancesforGOME,andforOMI,aremeasuredatslightlydifferentDopplershiftswithrespecttothesun)[Chance,1998;Slijkhuisetal.
,1999].
ThealgorithmforOMIincludesanon-baselineoption(sinceOMIisnotanticipatedtosignificantlyundersample)forcalculatingundersamplingcorrectionsatthetimeofreferencespectrumsampling.
Thisisaccomplishedbytakingthehigh-resolutionFraunhoferreferencespectrum,convolvedwiththeinstrumenttransferfunction(asdeterminedduringtheirradiancespectralcalibration),whichconstitutesanoversampledirradianceandformingproperly-sampledandundersampledrepresentationsofit,usingcubicsplineinterpolation.
Thedifferencebetweentherepresentationistheundersamplingcorrection;forGOMEthisconstitutesca.
90%ofthesystematicresidual.
6.
1.
4.
Common-modecorrectionRemainingsystematicresidualswhichare,bydefinition,uncorrelatedtothetracegasspectramaybeaveragedandincludedinthespectrumfittingasa``common-mode''spectrum,toreducethefittingrmsand,proportionally,thefittinguncertainties.
Thisisincludedasanon-baselineoptionforOMI.
ForGOME,thisreducesthermsbyaboutafactoroftwoforBrOandafactorof3forNO2.
Thepresumptionisthattheundersamplingcorrectiononlyimperfectlymodelstheinstrumenttransferfunctionandthusleavessomeremainingsystematiceffects.
74ATBD-OMI-04Figure6.
1ReferencespectraforfittingtodetermineslantcolumnabundancesofOClO:(a)OClO;(b)BrO;(c)NO2;(d)O3;(e)O2-O2collisioncomplex;(f)Ringeffect.
ComparisonswithGOMEmeasurementsofNO2forcleanmaritime(mostlystratosphericNO2)conditionsshowthatthisreduceduncertaintyistheappropriatechoice.
6.
1.
5.
Radiancefitting:BOAS(baselineoption)Theradianceisnllsfittedtoamodeledspectrumwhichincludestheirradiance,andeffectivealbedo,tracegasconcentrations,Ringeffectcorrection,andalow-orderpolynomial(uptocubic)closureterm:IAEeecccccNNRRssnn()()λλσλλλλλλσλσλ11012233L[6-4]Theσiaretheabsorptioncrosssections,σRistheRingeffectcorrection,andthec0-c3arethecoefficientsoftheclosurepolynomial.
Eachparametermaybeindividuallyselectedtobevariedduringthefittingorfrozenataconstantvalue.
ThebaselineforOClOistovaryA,Ns(OClO),NsATBD-OMI-0475(BrO),Ns(NO2),Ns(O3),Ns(O2-O2),cR,andc0–c3.
Inaddition,thelinearwavelengthscaleoftheirradianceisallowedtovary(baselineoption)tocorrectforsmallwavelengthchangesintheradianceovertheorbit.
Whileitisactuallytheradiancescalethatrequiresadjustment,therelativechangeinwavelengthscaleismuchsmallerthantheinstrumentresolutionorthevariationofmeasurablefeaturesinthereferenceorRingspectra:adjustmentoftheirradiancescaleavoidschangingthemeasuredquantity(theradiance)duringthefittingprocess.
Thebaselinefittingwindowis344.
5-393nm.
Additionally,thereisabaselineoptiontoonlyfitspectrawhereSZAandlatitudearewithinpre-setlimits.
Smoothing(low-passfiltering)oftheirradiances,radiances,andreferencespectraisincludedasanon-baselineoption.
Thisprocedureprovidesanalternatemeanstocorrectforspectralundersamplingeffects.
Itisimplementedbyapplyingarunning5-pointfilter(1/16,1/4,3/8,1/4,1/16)toeachofthespectra.
Updatingofthefittingparametersisincludedasabaselineoption:ForeachspectrumafterthefirstinaparticularspectrumfieldoftheCCD,theinitialguessesforfittingparametersareupdatedtothosefittedtothepreviousspectrum.
6.
1.
6.
Radiancefitting:DOAS(non-baselineoption)Thisoptionconvertsthedirect(BOAS)fittingofradiancestotheDOASmethod.
Thelogarithmoftheradiancedividedbytheirradianceistakenandtheresulthigh-passfilteredbythesubtractionofalow-orderpolynomial(uptocubic,withtheactualnumberoftermsdeterminedbytheinputparameterfile):H[ln(I/E)]=-Ns1H(σ1)NsnH(σn)+cR1H(σR1/F)+cR2H(σR2/F)+HOT,[6-5]whereHdenoteshigh-passfiltering.
Thecoefficientsofthepolynomialaredeterminedbylinearfittingtoln(I/E)[Pressetal.
,1986].
ReferencecrosssectionssampledattheOMIwavelengthscalearealsohighpassfilteredbysubtractionofalow-orderpolynomial.
TheRingeffectcorrectionisdividedbytheirradianceforeachspectralfieldandtheresulthigh-passfilteredaswell(notethatthisdeterminesthefirsttermintheexpansionforthelogarithmicquantitywhichincludestheRingeffect;higher-ordertermsarenotincluded-seeEquation[1-4]).
ThefittingisthenentirelyanalogoustotheBOASfitting,exceptthatthefittedquantityisnowH[ln(I/E)].
Iftheradiance/irradiancewavelengthadjustmentisnotselected,thentheDOASfittingwouldbelinearinthefittingparameters.
However,sincewavelengthadjustmentandBOASfittingarebothbaselineoptions,itisnotcurrentlyplannedtoimplementseparatelinearfittingforthiscase;theadditionalcomputertimeinfittingthelinearcasewiththenonlinearmethodisinconsequential,andtheresultsarevirtuallyidentical.
6.
2.
AirmassfactorsandverticalcolumnabundancesItisnotcurrentlyplannedtoimplementtheseforOClO.
6.
3.
ErrorestimatesEstimatederrorsaregivenforagroundfootprintof40·40km2,whichhasbeenadoptedasthestandardforreportingOMIerrorestimateswithinEOS.
S/NvaluesforestimatingfittinguncertaintiescomefromtheOMI-EOSInstrumentSpecificationDocumentRS-OMIE-0000-FS-021,Tables5.
5and5.
6,andareappliedtohebestfittingknowledgefromGOMEanalysis,takingintoaccountthedifferenceinspectralresolution.
Alluncertaintiesaregivenhereas1σ.
The1σfittingerrorinparameteriisdeterminedas76ATBD-OMI-04σi=iiC[6-6]orvariedpointspointsiirmsinnnC=εσ,[6-7]AsdiscussedinSection6.
1.
1;bothdefinitionsincludethecontributionsfromcorrelationoffittedparameters.
Cisthecovariancematrixofthestandarderrors.
Thepublisheduncertaintyoftheabsorptioncrosssectionsareaddedtothisinquadraturetoobtainthefinalslantcolumnfittinguncertainties.
TheaveragefittinguncertaintyforOClOinenhanced,polarvortexconcentrationsis9%.
Theuncertaintyincrosssectionsis10%.
Thetotalslantcolumnuncertaintyisthenthequadraturesum:9%+10%=13%.
Theerrorexcludingthesystematiccontributionfromcrosssectionsuncertaintiesis9%,correspondingto9·1012cm-2slantcolumndensity.
ThiscanbecomparedtotherequirementintheScienceRequirementsDocumentforOMI-EOS(RS-OMIE-KNMI-001,Version2)of1013cm-2verticalcolumndensity;performanceasspecifiedbytheInstrumentSpecificationDocumentisadequatetoobtainthesciencerequirement.
6.
4.
OutputsStandardoutputsinclude:Slantcolumnabundanceand1σfittinguncertaintiesforOClOandtheotherspeciesvariedinthefittingwindow;CorrelationofotherfittedspeciestoOClO(fromoff-diagonalelementsofthecovariancematrixofthestandarderrors);Fittingrms;Convergenceflagsandnumberofiterations(successfulwhichconvergencecriterion)Geolocationinformation;Versionnumbersofalgorithmandparameterinputfile.
6.
5.
ValidationGround-basedUV/visibleOClOcolumnmeasurementsprovidetheprimaryvalidationsource.
Vertically-integratedOClOprofilesfromballoon-basedUV/visibleSAOZinstrumentswillalsoproveuseful.
Ground-basedmeasurementsmustbefromhigh-latitudelocationswhereconditionsinsidethepolarvortexmaybesampled.
ATBD-OMI-04776.
6.
ReferencesBurrows,J.
P.
,A.
Richter,A.
Dehn,B.
Deters,S.
Himmelmann,S.
Voigt,andJ.
Orphal,AtmosphericremotesensingreferencedatafromGOME-2.
Temperature-dependentabsorptioncross-sectionsofO3inthe231-794nmrange,J.
Quant.
Spectrosc.
Radiat.
Transfer61,509-517,1999.
Caspar,C.
,andK.
Chance,GOMEwavelengthcalibrationusingsolarandatmosphericspectra,Proc.
ThirdERSSymposiumonSpaceattheServiceofourEnvironment,Ed.
T.
-D.
GuyenneandD.
Danesy,EuropeanSpaceAgencypublicationSP-414,ISBN92-9092-656-2,1997.
Chance,K.
,andR.
J.
D.
Spurr,RingEffectStudies:RayleighScattering,IncludingMolecularParametersforRotationalRamanScattering,andtheFraunhoferSpectrum,Appl.
Opt.
36,5224-5230,1997.
Chance,K.
,AnalysisofBrOMeasurementsfromtheGlobalOzoneMonitoringExperiment,Geophys.
Res.
Lett.
25,3335-3338,1998.
Greenblatt,G.
D.
,J.
J.
Orlando,J.
B.
Burkholder,andA.
R.
Ravishankara,Absorptionmeasurementsofoxygenbetween330and1140nm,J.
Geophys.
Res.
95,18,577-18,582,1990.
Marquardt,D.
L.
,Analgorithmforleast-squaresestimationofnon-linearparameters,J.
Soc.
Indust.
Appl.
Math.
2,431-441,1963.
Press,W.
H.
,B.
P.
Flannery,S.
A.
Teukolsky,andW.
T.
Vetterling,NumericalRecipes,ISBN0-521-30811-9,CambridgeUniversityPress,1986.
Slijkhuis,S.
,A.
vonBargen,W.
Thomas,andK.
Chance,CalculationofundersamplingcorrectionspectraforDOASspectralfitting,Proc.
ESAMS'99-EuropeanSymposiumonAtmosphericMeasurementsfromSpace,563-569,1999.
VandaeleA.
C.
,C.
Hermans,P.
C.
Simon,M.
Carleer,R.
Colin,S.
Fally,M.
F.
Mérienne,A.
Jenouvrier,andB.
Coquart,MeasurementsoftheNO2absorptioncross-sectionfrom42000cm-1to10000cm-1(238-1000nm)at220Kand294K,J.
Quant.
Spectrosc.
Radiat.
Transfer59,171-184,1998.
Wagner,T.
,C.
Leue,K.
Pfeilsticker,andU.
Platt,MonitoringofthestratosphericchlorineactivationbyGlobalOzoneMonitoringExperiment(GOME)OClOmeasurementsintheaustralandborealwinters1995through1999,J.
Geophys.
Res.
106,4971-4986,2001.
Wahner,A.
,G.
S.
Tyndall,andA.
R.
Ravishankara,AbsorptioncrosssectionsforOClOasafunctionoftemperatureinthewavelengthrange240-480nm,J.
Phys.
Chem.
91,2734-2738,1987.
Wilmouth,D.
M.
,T.
F.
Hanisco,N.
M.
Donahue,andJ.
G.
Anderson,FouriertransformultravioletspectroscopyoftheA2Π3/2←X2Π3/2transitionofBrO,J.
Phys.
Chem.
103,8935-8945,1999.
78ATBD-OMI-04