— — —~

ISSN 1047-3239 J. 41, h Waste Manage. Assoc. 4-4: 831-899

Developmentof Ozone Reactivity Scalesfor Volatile Organic Compounds

William P. L. CarterUniversity of California

This paper discusses methods for ranking photochernical ozone formation reactivities of volatile organic compounds WOC~).Photochernical mechanisms forthe atmospheric reactions of 118 VOC5 were used to calculate their effects on ozoneformatiori undervarious NO, conditions in model scenarios representing 39 different urban areas. Their effects on ozone were used to derive 18different ozone reactivity scales, one of which is the Maximum Incremental Reactivity (MIR) scale used in the new California LowEmission Vehicle and Clean Fuel Regulations. These scales are based on three different methods for quantifying ozone impacts andon six different approaches for dealing with the dependencies of reactivity on NOr The predictions of the scales are compared, thereasons for their similarities and differences are discussed, and the sensitivities of the scales to ND, and other scenario conditionsare examined. Scales based on peak ozone levels were highly dependent on NO,, but those based on integrated ozone were lesssensitive to NO, and tended to be sfrnilar to the MIR scale. It is concluded that the MIR scale or one based on integrated ozone isappropriate for applications requiring use of a single reactivity scale.

IntroductionThefot-madonof ground-levelozoneis a seriousairpollution

problem in many areas.Ozone is not emitted directly, but isformedfrom the photochemicalinteractionsof volatile organiccompounds(VOC5)andoxidesof nitrogen(NO,). Manydifferenttypesof VOCs are emitted into the atmosphere,eachreactiflg atdifferent ratesandwith different reactionmechanisms.lBecauseof this, VOCs candiffer significantly in their effectson ozoneformation.Thesedifferencesin effectson ozoneformation arereferredto asthe ozone“reactivities” of the VOCs. Such differ-enceshaveoftenbeenneglectedin the past and all non-exemptVOC5 havebeenregulatedequally.However, in recentyears, ithas becomerecognizedthat control strategieswhich encouragetheuseoflessreactiveVOCs couldprovidea cost-effectivemeansto achieveozonereductions.An exampleof this is to encouragetheuseof alternativefuelsfor motorvehicles.However,practicalimplementationof suchstrategiesrequiressomemeansto quan-tify thereactivitiesof VOCs.

Thereare a numberof waysto quanti&VOC reactivities,butthe most relevant measureof the VOC effectson ozoneis theactualchangein ozoneformationin an airshed.Thisresultsfromchangingthe emissionsof the VOC in that airshed.Which de-pendsnotonly on how rapidly the VOC reactsandthe natureofits atmosphericreactionmechanism,but also the natureof theairshedwhere it is emitted, including the effects of the otherpollutantswhich arepresentAlthoughthe effectofVOCs on ozoneformationcanbemeasuredinenvironmentalchamberexperiments.

the fact that theseeffectsdependon theenvironmentwheretheVOCs reactmeansthat quantitativeozoneimpacts in the atmo-spherewill notnecessarilybe the sameas thosemeasuredin thelaboratory.However,theeffect of a VOC on ozonein the atmo-spherecanbeestimatedusingcomputerairshedmodels.Whiletheresultsof suchcalculationsare no morereliablethanthe modelsof the chemical reactions and the air pollution episode beingconsidered,modelingprovides the most realistic and flexiblemeansto assessthemany factorswhich affectozoneformationfrom VOCs and for thedevelopmentof VOC reactivity scales.

Theeffectof changingtheemissionsof agivenVOC on ozoneformation in a particularepisodewill, ingeneral,dependon themagnitudeof the emissionchangeand on whetherthe VOC isbeingaddedto. subtractedfrom,or replacingaportionof thebasecase(i.e.. presentday)emissions.To removethe dependenceonthis, it was proposedto use “incrementalreactivity” to quantifyozoneimpactsof VOCs.2This is definedasthechangeon ozonecausedby addinganarbitrarilysmallamountof the testVOC totheemissionsin theepisode,divided by the amountof testVOCadded.Thiscanalsobetho~ightofasthepartialderivatveof ozonewith respecttoemissionsof theVOC. Thisdoesnotnecessarilypredictthe effectsof largechangesin emissions,asmight occur,forexample,if all themotorvehiclesinanairshedwereconvertedto anothertype of fuel. However,ChangandRudy3 found thatincrementalreactivitiesgive good approximationsto effectsonozoneof alternativefuel substitutionscena.riosinvolving chang-ing up to 30 percentof the total VOC emissions.In any case,incrementalreactivitieswill predictthe directionof an initial ozonetrendwhich resultswhen acontrol strategyis beingphasedlit

Incrementalreactivitieshavebeeninvestigatedin a numberofcomputermodelingsrudies.3-9andtheVOC’s reactionmechanismwasfoundto beimportantin affectingits incrementalreactivity.Somecompoundscancausethe formation of 10 or more addi-tionalmoleculesof ozonepercarbonatomreacted,eitherdirectly

Oapydght 1994— Air I Waste Managcmcnt Associadoc

ImplicationsControlstrategieswthch encourage use ofvocs thatform Tess ozone per

gramemittedmayprovidealesscostlywaytoachieveozonereductions.Mexampleof this is to encourageuseof alternativefuels for motorvehicles.Practicalimplementationof suchstrategiesrequiressomemeansto quan-tify ozoneformation potentials of vOCs. This paper discussesvariousmethodsto do this. -

AIR & WASTE • VoL 44 • July 1994 • 881

or through its effects on reactionsof other compounds,whileotherscausealmostno ozoneformation whenthey react,or evencauseozoneformationto bereduced.’ThepredictionsthatVOCshavevariableeffectson ozoneformation,evenafterdifferencesinhowrapidlytheyreactaretakeninto account,andthatsomehavenegativeeffectson ozoneformationundersomeconditions,havebeenverified experimentally)-’0

Themodelingstudiesalsopredictthat incrementalreactivitiesdependsignificantly on the environmentalconditions,particu-larly on therelativeavailability of NO,5~’NO, availability hastraditionally been measuredby the ratio of total emissionsofreactiveorganicgases(ROC)to NO,. In general.VOCshavethelargestincrementalreactivitiesunderrelatively high NO,condi-tions (i.e., low ROC/NO,ratios)andhavemuchlower, iii somecasesevennegative,reactivities underconditionswhereNO, islimited (high ROC/NO,ratios).This is becauseunderhigh NO,conditions,the amount of ozoneformed is determinedby thelevelsof radicalsformedfrom thereactionsof theVOCs. Underlower NO,conditionsit is theavailability of NO,, whichmustbepresentso ozonecanbeformed,that limits ozoneformation.OtheraspectsoftheenvironmentwhereVOCis emitted,suchasnatureoftheotherorganicsemittedinto theairshed,L1Ithe amountofdilutionoccurring.9etc..canalsobeimportantin affectingVOCreactivities,thoughinvestigationsof theseaspectsaremore limited.

Thefactthatincrementalreactivitiesdependon environmentalconditions meansthat nojJngle scalecan predict incrementalreactivjti,es,or evenratios of i~rementalreàctivities,underallconditions.Thustheconceptof a“reactivity scale”oversimplifiesthe complexities of the effects of VOC emissionson ozone

0’ formation. Nevertheless,for someregulatoryapplications,theonly practicalchoiceis betweenusing somereactivity scaleorignoringreactivityaltogether.Thelatterwould betheappropriatechoice if reactivities were so variable that all VOCs could beconsideredthesamewithin this variability. If this is not thecase,andif the policy is adoptedto usea reactivity scale, the issue

~ becomeshow onewoulddevelopascalethatwould resultin thegreates~overallair auality improvementfor therangeof conØi-tionswhereit willbe~p~lT2

Anei~ampIeofacasewhereareactivityscaleusedis the“Low-EmissionVehiclesandCleanFuels”regulationsin California.’In this regulation,non-methaneorganicgas (NMOC) exhauststandardsfor alternativelyfueledvehiclesaredeterminedusingreactivityadjustmentfactors(RAF5), whichareintendedto relatethe differencesin ozone formation potential of the exhaustscomparedto thatofconventionallyfueledvehicles.’2Theregula-tion utilizes the maximumincremental reactivity (MIR) scale

~deveIoped by this authorto calculatetheseRAFs.’3

This paperreportstheresultsof aninvestigationof alternativeapproachesfor developingreactivity scales,and describesthedevelopmentof theMIR andotherreactivity scales.

ChemicalBasisof ReactivityThis sectiongivesasummaryof thechemicalfundamentalsof

03 formation, whichmaybe usefulfor an understandingof thiswork. The only significant processforming 03 in the loweratmosphereis thephotolysis of NO2. which is reversedby therapidreactionof 03 with NO.

NO, + hv—*OQP) + Nb~-c0(3~)+ 0, + M->03 + M;03+ NO-~)’NO-,+°2

This resultsin 03 beingin aphotostationarystatedictatedbythe NO, photolysis rate and the [NO,]/[N0} ratio. If reactive‘OCswerenotpresent,thensignificantamountsof 03 would not

be formed. When VOCsarepresent,they reactto form radicalswhicheitherconsumeNO orconvertNO to NO2. Becauseof the

photostationazystate relationship, this causes03 to increase.Although manytypesof reactionsareinvolved) themajorpro-cessescanbe summarizedas follows:

VOC + 0H*RO, ÷productsR02 + NO-~--~.NO,+ radicals

- radicals-*-*OH + products

Therateof ozoneincreasecausedby theseprocessesis depen-denton the amountsof VOCspresent,the rateconstantsfor theVOC’s initial reactions,andthe level of OH radicalsandotherspecieswith which the VOCs might react.Ozoneproductioncontinuesaslong assufficientNO, is presentso thatreactionsofperoxy radicals(R0) with NO competeeffectively with theirreactionswith otherperoxyradicals.

The OH radical levelsareparticularly importantin affectingthe03 formationratein thepresenceof NO,becausereactionwithOH is amajor(andin manycasestheonly) processcausingmostVOCs to react.Thus if aVOC reactsin suchaway thatit initiatesradical levels (or forms a product which does),then it wouldenhancetherateof ozoneformationfrom all VOCspresent.Thiswould give it a high incrementalreactivity comparedto otherVOCs.If theVOChasradical terminationprocesswhen it reactsin thepresenceofNO,, it will causeall VOCsto reactslowerandform less03. In somecasesthis reduced03 formationfrom otherVOCs may bemore thanenoughto countertheozoneformationformedfrom the VOC’s directreactions.In suchcasesthe VOCwould havea negativeincrementalreictivity in thepresenceofNO,.

Ozoneformation stopsonceNO, is consumedto sufficientlylow levels. Since NO, is removedfrom the atmospheremorerapidly thanVOCs(sincemostVOCsform productVOCswhichalso react), it is NO, availability which ultimately limits 03formation.If the NO, levelsareso high that it is not consumedbeforetheendof the day, then it is mainly therateof theVOC’sreactions,andtheireffectson OHradicals,whichaffectincremen-tal reactivity.Indeed,NO, inhibits 03 underhigh NO, conditionsbecausereactionofOHwith NO, is animportantradicalterminat-ing process.If, however,NO, is consumedbeforethe endof theday, then 03 is NO,-limited. andincreasingNO, would causeincreased03 formation. Under such conditions, if a VOC’sreactionscausedNO, to beremovedmorerapidly than if theVOCwereabsent(suchas, for example,by forming nitrogen-contain-ingproductssuchasPANsfrom aldehydesandnitrophenolsfromaromatics),thenthis would haveanegativeeffecton 03 yields.andtendto reduceaVOC’s incrementalreactivity. UnderhighlyN0,-limited scenarios,this becomessufficiently important tocausenegativeincrementalreactivitiesfor VOCswith significantNO, sinksin their mechanisms— evenfor thosewhichmayhavehighly positive effectson 03 underconditions whereNO, isplentiful.

Thus NO, conditionsarea major factor affectingreactivity.However,otherconditionswill alsoaffectreactivity,by affectinghow rapidly NO, is removed,by affectingoverallradical levelsandthus howrapidly NO,andVOCsreact,andby affectingotherfactorsdeterminingtheefficiency of ozoneformation.This re-sultsin variationsof incrementalreactivitiesamongthediffer-entairshedconditions,eventhosewith similar NO, levels.Therelative importanceof thesefactors are investigatedin thiswork.

MethodsThis paperusesanumberof specializedtermsandabbrevia-

tions. To assistthereaderin following this discussion,Table Igives a summaryof theseterms and abbreviations.Thesearediscussedin more detailbelow.

TECHNItALPAPEff.~ :1.

882 • July 1994 • Vol.44 • AIR & WASTE

Table I. Summaqof terms and abbreviations.

Typesof Scenarios and Scenario Characteristics

EKMA Scenario A model for anair pollution episodewhich canbe representedin the EKMA model formulation.This involvesa single-cell boxmodelwith a fraction of RUG and NO,pollutantspresentinitiaUy andthe remainderemittedthroughouttheday, time-varyingchangesin inversionheightsandentrainmentof pollutantsfrom aloft astheheight raises,andtime-varying humiditiesandtemperatures.

Base RUG Mixture Themixture of ReactiveOrganicGases-(ROGs)initially presentor emittedin the EKMA scenariosexcept for biogenicVOCs, VOCs presentaloft, or VOCs addedfor the purposeof calculatingtheir incrementalreactivities.

NO, (or ROG) inputs The sum of the initial NO, (or baseRUG) and the total emittedNO,(or baseROG) in the scenarios,in unitsof moles(or molescarton)perunit area.

NO,Availability The conditionof whetherNO, is limiting 03 formationor whetherNO, is in excess,and the degreeto which this is thecase.

BaseCaseScenario An EKMA scenariowhoseinputsare derivedto representa specificozoneexceedenceepisodein anareaof the UnitedStates.

MIR Scenario Maximum IncrementalReactivity (PAIR) Scenario.A scenarioderivedby adjusflngthe NO, emissionsin a basecasescenarioto yield the highestincrementalreactivity of the baseRUG mixture.

MOR Scenario Maximum OzoneReactivity (MOR) Scenario. A scenarioderivedby adjustingthe NO, emissionsin abasecasescenarioto yield the highestpeakozoneconcentration,

EBIR Scenario EqualBenefit IncrementalReactivity (SIR) Scenario.A scenarioderived by adjustingthe NO, emissionsinsbasecasescenariosoVOC andNO, reductionsareequallyeffective in reducing03.

Averaged Conditions A scenariowhoseinputsrepresentthe averageof thoseof the basecasescenarios.

Measure: of Reactivity

incrementalReactivity Changein ozoneformedcausedby addinga VOC to the initial andemittedbaseRUG in a scenario,divided by theamountof VOC added.

RelativeReactivity The incrementalreactivityof theVOC dMded by the incrementalreactivity of thebaseRUG mixture.

Kinetic Reactivity Fraction of the VOC which reactsin the scenario.

MechanisticReactivity Changein ozoneformedcausedby.addinga VOC to theinitial andemittedbaseRUG in ascenario,divided by theamountof VOC which reacted.

03 Yield Reactivity Incrementalor relative reactivitybasedon theeffect of the VOC on the maximumamountof ozoneformed.

IntO3 Reactivity Incrementalor relativereactivitybasedon theeffect of the VOC on the 03 concentrationintegratedover time.

intU3>90 Reactivity Reactjyitybasedon the effect of the VOC on thesumof the 03 concentrationsfor eachhour when03?:90 ppb.

ReactivityScales

ReactivityScale A numericalrankingsystemwhereeachVOC is assigneda nunibergiving a measureof how its emissionsaffectozoneformation,

AdjustedNO, Scales Reactivity scalesderivedfrom incremerrtalreactivitiesin scenarioswith a specifiedcondition of NO, availability.

MIR, MUIR, or SIR The adjustedNO,scalesconsistingof the averageof ozoneyield reactivitiesin the MIR, MOR, or EBIR scenarios,respectively.

BaseCaseScales Relativereactivity scalesbasedon incrementalreactitiVies in scenarioswhereNO, inputswere notadjusted.

Base(AR) Scales Base casescalesderivedusing the averagedratio method.Averagesof the relativereactivitiesirithe basecase -

scenarios.

Base(Li) Scales Basecasescalesderivedusingthe leastsquareserror methodwhich minimizes the changein ozonecausedbysubstitutingthe baseROG for the VOC usingreactivityweighting factorswhich the scalepredictshaszero effect onozone.

- AIR & WASTE • VoL 44 • July 1994 • 883

ReactivityScalescant.

Base(U) Scales

Other

Basecasescalesderivedusingthe leastsquareserrormethodwhich minimizesthe changein ozonecausedbysubstitutingthe VOC for the BaseRUG using reactivityweighting factorswhich thescalepredictshas zsroeffectonozone. -

VOC

ROG

NMOG

EKMA

Null test

Volatile OrganicCompound.In this paper,CO is also referredto asaVOC, but strictly speakingit is not.

ReactiveOrganic Gas.VOCs whichreactin theatmospheretoasignificantextent,I.e., VOCs otherthanGO, methane,chiorofluorocarbons,or otherunreactivecompounds. -

Non-MethaneOrganicGases.VOCs excludingmethaneand CO.

Empirical Kinetic ModelingApproach.A method toestimateeffectsof ROG or NO, controlson ozonebasedon boxmodelcalculationsof one-dayepisodesusingaparticularcomputerprogram.In the contextof this work, it referstothebox modelscenariosdevelopedfor this type of modelinganalysis,

A modelsimulationwhereoneVOC or mixture of VOCs is replacedby anotherin a proportionwhichareactivityscalepredictswould haveno effecton ozone.The resulting changein ozoneis away of measuringthe error of a reactivityscale.

Scenarios Usedfor Reactivity AssessmentTheassessmentof ozonereactivitiesof VOCs undera variety

ofconditionsrequirescalculatingtheireffectson ozoneformationusinga setof modelscenarioswhich representa realisticdisuibu-tion of environmentalconditions.An extensivesetof pollutionscenarioshasbeendevelopedforconductinganalysesofeffectsofROG and NO, controls on ozoneformation using the EKMAmodeling approach.”-’8The EKMA approachinvolves usingsingle-cellbox modelsto simulatehow ozoneformationin one-lay episodesis affected by changesin ROG and NO, inputs.Althoughsingle-cellmodelscannotrepresentrealisticpollutionepisodesin greatdetail, they canrepresentdynamicinjection ofpollutants,time-varying changesof inversion heightswith en-trainmentof pollutantsfrom aloft asthe inversionheight increasesthroughoutthe day,andtime-varyingphotolysisrates,tempera-tures,and.humidities.14.16.19Thus, they canbe usedto simulateawide rangeof thechemicalconditionswhich affectozoneforma-tion from ROG andNO,. Thesearethe seineas thoseaffectingVOC reactivity.Therefore,atleast tothe extentthey aresuitablefor theirintendedpurpose,an appropriatesetof EICMA scenarios

I shouldalsobesuitableforassessingmethodstodevelopreactivityI scalesencompassinga wide rangeof condinons.

BaseCaseScenarios.Thesetof EKMA scenariosusedin thisstudyweredevelopedby theUnitedStatesEPA for assessinghowvarious ROG and NO, control strategieswould affect ozonenonattainmentin variousareasof thecountry.”Thecharacteris

ticsof thesescenariosandthemethodsusedto derivetheir inputdataaredescribedin moredetailelsewhere.”’°Briefly. 39 urbanareasin the United Stateswere selectedbasedon geographicalrepresentativenessof ozonenonattainmentareasanddataavail-ability, anda representativehigh ozoneepisodewasselectedfor

- each.Thesewerebasedon 1986-88data-!!Theinitial NMOGandNO,concent.rationLthe aloft 03 concentrations.and the mixingheight inputs were basedon measurementdatafor the variousareas.The hourly emissionsin the scenarioswere obtainedfromthe National Acid PrecipitationAssessmentProgram(NAPAP)emissionsinventory,’$ and biogenic emissionswere also in-cluded,TableII gives asummaryof the urbanareasrepresentedalong with otherselectedcharacteristicsof the scenarios.

The initial NMOG andNO, concentrationsarebasedon airI .1uality data,sotheyarenotaffectedby uncertaintiesandpossible

errorsin the emissionsinventory.Inventoryerrorswould affectamountsofhourlyemissionsafterthebeginningofthesimulation,

which usuallyhavelessofaneffecton the ozonethanthe amountsof NMOGandNO,presentinitially. Thus if theNIvIOG inventoryweretoo low, then thebasecaseROG/NO,ratio would also below, but to a lesserextent.However,this would notsignificantlyaffect the ROCfl~0,ratio in the adjustedNO, scenarios(dis-cussedbelow.)

Severalchangesto the scenarioinputs were madebasedondiscussionswith the California Air ResourcesBoard (CARB)staffandothers.’32’Twopercentofthe initial NO, and0.1%of theemittedNO, in all the scenarioswas assumedto be in the formofMONO. The photolysis rateswere calculatedusingsolar lightintensitiesandspectracalculatedby Jeffries2 for 640meters,theapproximatemid-pointof the mixedlayerduring daylighthours.Thecompositionof theNMOGsentrainedfrom aloft wasbasedontheanalysisofJeffriesetal. ofaircraftdatafrom anumberofurbanareas.23The compositionof the initial and emitted ROGs wasderivedasdiscussedbelow.Completelistingsofthe inputdataforthescenariosare given elsewhere)0

This setof 39 EKMA scenariosarereferredto as“basecase”to distinguishthem from the scenariosderivedfrom them byadjustingNO, inputs to yield standardconditionsof NO, avail-ability asdiscussedbelow.No claimis madeasto the accuracyof’thesescenarios-inrepresentingany real episode,but they are aresultof aneffort to represent,asaccuratelyaspossiblegiven theavailabledataandthe limitationsof the formulationofthe EJUvIAmodel,the rangeof conditionsoccurringin urbanareasouttheUnitedStates.Whendevelopinggeneralreactivityscales,it is more important that the scenariosemployedrepresentarealisticdistribution of chemicalconditionsthanany one accu-rately representingthedetailsof any particularepisode.

BaseROG Mixture. The BaseROG mixture is the mixtureofreactiveorganicgasesusedto representthechemicalcompositionof the initial and emitted anthropogenicreactiveorganic gasesfrom all sourcesin the scenarios.It is referredto as the “base”mixturebecauseit is usedin the simulationswithoutthe addedtestVOC in the reactivitycalculations.(V/henthereactivityofa VOCis assessed,that VOC is addedto this baseROG mixture in thesirnulatiorn)Consistentwith the approachused in the originalEPA EKMA scenarios,the samemixture wasused for all sce-narios_r~xceptfor thecalculationswheretheeffectsof changingthis mixture are assessed.The speciationfor this mixture wasderivedby Croes24basedon ananalysisof the EPAdatabase2zforthe hydrocarbonsand the 1987SouthernCalifornia Air Quality

884 • July 1994 • Vol.44 • AIR & WASTE

I . T~bICII. Surnmaqofconditionsof the EPAiasccasescenarios. NO, conditions,whichis applicableto avarietyofscenarios,NO,

Caic. ROG NO, Final lnlL+Esnlt AloftMax 0~ma. mo,~°~Height Base ROG O~(ppb} (kin) (mmol m-z) (pph)

Atlanta. GA 163 7.3Austin, TX 162 9.3Baltimore. MD 275 £2Baton Rouge,LA 211 6.8Birmingham,Al. 223 6.9Boston.MA 182 55Charlotte. MC 137 7.8Chicago. IL 251 1t6Cincinnati. OH 183 &4Cleveland, OH 220 6.6Dallas, TX 167 4.7Denver. CC 172 62Detroit, Ml 217 6.8El Paso.TX 162 6.6Hartford, CT 160 8.4Houston.TX 256 6.1Indianapolis, iN 187 6.8Jacksonville, FL 141 7.6Kansas City, MO 146 7.1Lake Chajies, LA 257 7.4Los Angeles, CA 483 7.6Louisville, KY 191 5.5Memphis. TN 205 6.8Miami. FL 125 9.6Nashville, iN 155 8.1NewYorIcNY 317 8.1Philadelphia, PA 212 62Phoen’c*, AZ 242 7.6Portland,OR 152 65Richmond. VA 212 6.2Sacramento, CA 184 6.6St Louis. MO 269 6,1Salt Lake City. UT 173 8.5San Antonio,TX 119 3.9San Diego, CA 169 7.1San Francisco, CA 167 4.8Tampa, FL 192 4.4Tulsa, OK 201 5.3Washington, DC 250 5.3

0.8 2.1 12 6303 2.1 11 851.2 1.2 17 841.0 1.0 11 6203 1.8 13 810.7 2.6 14 1050.4 ao 7 920.6 1.4 25 400.8 2,8 17 701.1 1.7 16 891.4 23 18 751.3 3,4 29 570.9 1.8 17 681.1 2.0 12 65nc 2.3 11 781.1 1.7 25 651.0 1.7 12 520.7 1.5 8 400.7 2,2 9 650.7 03 7 401.1 0.5 23 1000.9 2.5 14 - 750.8 1.8 15 580.5 2.7 9 570.5 1.6 7 500.8 1.5 39 1031.0 1.8 19 531.1 3.3 40 600.8 1.5 6 660.9 1.9 16 640.9 1.1 7 601.2 1.6 26 820.7 22 11 8512 2.3 6 601.1 0.9 - 8 902.0 0,7 25 701.3 1.0 8 681.0 1.8 15 700.9 1.4 13 99

6.6 0,9 1.8 15 70Avg. Conditions 206

Study (SCAQS)databasefor the oxygenates.2~’This mixtureconsistsof 52 percent(by carbon)alkanes,15 percentalkenes,27percent aromatics. 1 percentformaldehyde,2 perce’nt higheraldehydes.1% ketones,and2% acetylene.The detailedcomposi-tion of this mixture is givenelsewhere.20

AdjustedNO, Scenarios.Since incrementalreactivitiesarehighly dependentonNO,,5~9”andsincetheNO,conditionsof thebasecasescenariosare variable(as indicatedby theROG/NO,ratioson TableIi), onewouldexpectthe incrementalreactivitiestoalsobehighlyvariable.But if theNO,inputstothesescenarioswere adjustedto yield consistentNO, conditions, one mightexpectthe incrementalreactivities,orat leastthe ratios of incre-mentalreactivities.to bemuch Tessvariable.In this case,thesetof incrementalreactivitiesso obtained may provide a generalreactivity scalewhich is at least applicable to that particularcondition of NO, availability. Comparing different reactivityscalesfor different NO, conditionswould providea systematicmeansto assesshow reactivityscales,andcontrolstrategiesbasedon them,would vary with NO, levels.

To developa set of scenariosfor this purpose,oneneedsameansto assessNO, availability,or to establishequivalencyof

availability is determinedbothby theamountofNO,inputandtherateat which it is removed.The latteris affectednotonly by thereactivity and amountof ROCs which are present,but alsobyfactorssuchaslight intensity,temperature,anddilution, which,ingeneral,will vary fromscenariotoscenario.ThereforetheamountofNO,presentor theROC/NO,ratio arenot necessarilyreliableindicatorsof NO,availability,

However,ifoneexamineshow changesinROCandNO, affectozoneformationasa functionof NO,inputs,essentiallythe samepatternis observedfor all scenarios.This is shown in Figure 1.which plots,againsttotalNO, input, thechangesin ozonecausedby I percentincreasesin ROCorNO, inputsforatypicalscenario.Thefigurealsoincludesa plot ofthepeakdaily ozoneconcentra-tion againstNO, input. In all casesthereis a NO, input level(designated“MIR” on the plot) where the ROC input has thehighestandmostpositiveeffectonozonewhichis nearorthesameas the point wheretheeffectof NO, is the mostnegative;thereisa lowerNO, level (“MOR”) which yields the maximumozoneconcentrationandwheretheeffectof NO,on ozonechangessign;andthereis a yet lowerNO, level (“EBIR”) wheretheeffectsoffractionalchangesof ROG andNO,onozoneformationareequal.Although thesethreepoints, in general,occur at differentNO,inputs or ROG/NO,ratios for conditionsof differentscenarios,they representconsistentNO, conditionsin termsof how ozoneformationisaffectedby ROC and NO, changes.Thus,thesecanbeusedto definethreeconditionsof consistentNO,availability,which yield threesetsof adjustedNO, scenarios.Theseare asfollows:

MaximumIncrementalReactivity (MTR) Scenarios.In thesescenariosthe NO, inputs are adjustedso that the baseROGmixture hadthe highestincrementalreactivity.TheNOx adjust-ment was done by varyingboth the initial NO,and theemittedNO, by the samefactor.20TheseMIR scenariosrepresentNO,conditionswhereemissionsof ROGshavethe greatesteffectonozoneformation,andwhereNO,has thestrongestozoneinhibit-ingeffect.Thusthey representconditionswhereROGcontrol hasthe greatesteffect on ozone.They can also be thought of asrepresentingapproximatelythe highestNO, levels which arerelevant in consideringcontrol strategiesfor ozone, becauseozone is suppressedto low levels if NO, inputs are increasedsignificantly abovethis level,

MaximumOzoneReactivity (MOR) Scenarios.In thesesce-nariosthe NO,inputsareadjustedto yield thehighestpeakozoneconcentration.This representsthe dividing line betweencondi-tionswhereNO, is in excessand whereozoneis NO,limited,orthe “ridgeline” on ozoneisoplethplots.2’ MOR scenariosrepre-sentNO, conditionswhich areoptimumfor ozoneformation.

Equal Benefit IncrementalReactivity (EBIR) Scenarios.Inthesescenariosthe NO, inputs are adjustedsothattheeffectonozoneof a givenpercentageincrementalchangein ROGinput is

Flg~re1. Qualitativedependencieson NO, inputsof maximum ozoneandofrelative changesin ozonecausedby 1% increasesin total RUG or total NO,emissions for the AveragedConditionC Scenarios.NO, inputs are shownrelativeto NO,inputswhich give maximumozoneyields. -

CIty, State

S

—

I..

1—

AIR & WASTE •VoI. 44 • Ju~’1994 • 885

-, ...

the sameas theeffect of an equalpercentagechangein NO,. Inotherwords,this is the point wherethe incrementalreactivityofthe baseROG mixture, multiplied by the totalamount of ROCinput (excludingaloft orbiogenicROCs),equalsthe incrementalreactivity of NO,, multiplied by the amountof NO, input. TheEBIR scenariosrepresentthelowestNO, conditionswhereROGcontrol is of equalor greatereffectivenessfor reducingozoneasNO,control.ThustheyrepresentthelowestNO,conditionswhicharerelevanttoROG control,sinceat towerNO,levelsNO,controlbecomesmuchmoreeffectivein reducingozone.

AvengedConditionsScenarios.It is useful for sensitivitystudiesandexamplecalculationstohavea singlescenarioor setofscenarioconditionswhichcanbetakenasbeingrepresentativeof the larger set. For this purpose,we derived an “avengedconditions”scenariofrom the averagesof the relevantinputsofthe 39 basecasescenarios.Thiswas then usedas the basisfordevelopingscenarioswith someinput modified,suchasthebaseROGcompositionor theinitial HONO. TheMIR. MOR, orEBIRversionsof this scenarioaredeterminedasdiscussedabovefor thebasecasescenarios.Notethatwhenconductingsensitivitycalcu-lations on varied scenarioconditions, the NO, adjustmentstodetermineMIR. MOR or EBIR conditionswere doneafter thescenarioconditionwasvaried,sotheeffectof thevariationcanbeassessedon anequalNO, availability basis.Otherwise,theeffectof thevariation on NO,availability may dominatethe result.

Calculation of Reactivities in a ScenarioIncrementalReactivifies.Incrementalreactivitiesin a given

scenarioarecalculatedby conductingmodelsimulationsof ozoneformationin thescenario,andthenrepeatingthecalculationswitha small amountof the testVOC added.Theamountof testVOCaddeddependson how rapidly it reactsin the scenario.Theamountaddedis sufficiently largeso that numericalerrorsin thecomputersimulationdo notsignificantly affectthe results,yetissufficientlysmallsothat theeffectofaddingtheVOCis within thelinear rangewherethe changein ozone is proportional to theamountadded.2°Theincrementalreactivitiesarethenthechange

inozoneformedin the two calculations,dividedby theamountoftestVOC added.The detailedmethodologyusedfor calculatingincrementalreactivitiesin a scenariois given elsewhere.2o

-• Incrementalreactivitiesdependon how theamountsof VOCaddedandhowtheamountsof ozoneformedarequantified.In thiswork, incrementalreactivitiesarebasedon VOCsquantifiedonamassbasis,i.e., by theamountof ozoneformedpergram of VOCadded.This is most relevantfor control strategy applicationsbecauseVOC emissionsare quantified by mass. Alternativequantifications,whichhavea closercorrespOndenceto thechemi-cal processes,are mole basisorcarbonbasis,wherethe latterismorefrequently used.The VOC quantification affects relativereactivitiesof VOCswith differentmolecularweightspercarbon,suchasoxygenatescomparedto hydrocarbons.

The way theamount of ozoneformed is quantifiedwill alsoaffectincrementalreactivities.The following ozonequantifica-

.,x tionsareused:

Ozone Yield Reactivities.Thesearebasedon the maximumnumberof molesorgramsofozoneformedin thescenario.i.e.,themolesor gramsperunit areain the mixed layerat the timeof themaximum ozoneconcentration.This gives the sameratios ofincrementalreactivitiesasreactivitiescalculatedfrom peakozoneconcentrations,but is preferredbecauseit permitsmagnitudesofreactivitiesinscenarioswith differing dilutions tobecomparedonthe samebasis. Most previous recent studies of incrementalrcactivitya.33.9.Shave beenbasedon ozoneyield or peak ozoneconcentrationreactivities,

IntegratedOzone(IntO3) Reactivieies.Theseare basedon theozoneconcentrationsintegratedovertime throughoutthe simu-

latedday. If two VOC5give thesamemaximumozoneconcentra-tion whenaddedin equalamountsin thescenarios,but onecausesozoneto be formedearlier, their IntO3 incrementalreactivitieswouldbedifferent,eventhoughtheozoneyield reactivitieswouldbethesame.

IntegratedOzoneOver90 pp5 (JntO3>90)Reactivities.Thesearebasedon the extentto which theozoneexceedstheCaliforniaambientairquality standardof 90ppb, andthe lengthof time of

I theexceedence.In thiswork, this is quantifiedby thesumof thehourlyozoneconcentrationsfor thehourswhentheozoneexceedsthe standardin the calculationswithout the addedVOC. (Thehourswhen the standardis exceededin thecalculationswith theaddedVOC would bethe sameif the amountof VOC addedweresufficiently small.)

RelativeReactivities.Forcontrolstrategypurposes,theratiosof incrementalreactivitiesfor one VOC relativeto othersareofvtaterrelevancethantheincrementalreactivitiesthemselves.Todefinearelativereactivityscale,oneneedstoselectaVOC to useas the standard.Forexample.Chameideset al.29usedpropene.Russell andco-workers~°used carbon monoxide,andDerwentand Jenkins3lusedethylenefor this purpose.In this work, thestandardusedis thebasecaseROCmixture, i.e., themixtureusedin the model simulations to representthe initially presentandemitted anthropogenicreactiveorganic gasesin the scenarios.Thus,the relativereactivityof a VOC is theratio of theincremen-tal reactivityof theVOCto theincrementalreactivityof thebaseROG mixture. Whendefined in this way, the VOC’s relativereactivitymeasurestheeffecton ozoneof changingtheemissionsofthis VOC comparedto the effectchangingtheemissionsof allVOCsequally.

Kinetic and Mechanistic Reactivities. To provide a moredetailedexaminationof how differing aspectsof VOC reactionmechanismsandscenarioconditionsaffectreactivity, incremen-tal reactivityshouldbeconsideredtheproductof the“kinetic” andthe “mechanistic”reactivities.9The kinetic reactivityis thefrac-tion oftheemittedVOC whichreactsin thescenario,anddependson theVOC’s relevantrateconstantsand thelevelsof theradicalsandspeciesin the scenarioswhich react?viththe VOCs.Mecha-nisticreactivity is thechangein ozoneformedcausedby addingtheVOC, dividedby the amountwhich reacts— or the incremnen-tat reactivity divided by thekinetic reactivity. The mechanisticreactivities are independent(to a first approximation)of howrapidly theVOC reacts,but areaffectedby factors suchas thenumberofconversionsofNO to NO, whichoccurwhentheVOCreacts,whethertheVOC’sreactionsenhanceor inhibit radical orNO, levels,the reactivitiesof the productsthey form,andcondi-tions of the scenariosuchas NO, availability andother factorswhichaffect theoverallefficiencyofozoneformation.~Thesetwocomponentsof incrementalreactivity are affectedby differentaspectsof the VOC reaction mechanismand of the scenarioconditions.Thustheyhelpexplainfactorswhich affectreactivity.andwhy reactivitycanvary from scenarioto scenario.

Derivation of Multi-ScenarioReactivityScalesA total of 1.8 different generalor multi-scenarioreactivity

scaleswere derivedin this work, dependingon which type ofozonequantification,scenarios,or aggregationmethodwasused.

Adjusted NO, Scales. Adjusted NO, reactivity scalesarederivedfrom incrementalor relative reactivitiesin theadjustedNO,scenarios,A total of ninesuchscaleswerederivedbasedonthethreeconditionsof NO, availability (MIR, MOR, andEBIR)andthethreemethodsfor quantifyingozone(03 yield, IntO3,andlntO3>90). Incrementalreactivitiesin thesescaleswerederivedby averagingthe kinetic andmechanisticreactivitiesin theMIR,MOR.orEBIR scenarios,andcombiningthemto yield aggregateincrementalreactivitiesfor thesethreeNO,conditions.Thisgives

a

886 • July 1994 • Vol.44 • AIR & WASTE

essentiallythe sameresultas simply averagingthe incrementalreactivities.but this approachwas used becauseit also giveskinetic andmechanisticreactivitieicharacteristicof the adjustedNO, conditions.This helpsin analyzingreactivitytrends.Incre-mentalreactivitieswerecomputedonly for theozoneyield scales:for theintO3 andIntO3>90scales,only relativereactivitieswerecomputed.Relativereactivitiesfor the nine adjustedNO, scaleswerederivedby averagingtherelativereactivitiesin theadjustedNO, scenarios.

In theremainderof this paper,theterms“MIR scale.”“MOIRscale.”or “EBIR scale”will beusedto refer to the ozoneyieldadjustedNO, scales.(MOIR standsfor “maximum ozoneincre-mentalreactivityjTheIntO3orlntO3>90adjustedNO,scaleswillbe referencedexplicitly as suchwhen they arediscussed..Thisconformsto the terminology usedelsewherefor theMR andMOIR scales.’’3

Base Case Scales.Base caserelative reactivity scalesarederived from incrementalreactivitiesin the basecasescenarios.Only relative reactivities arederivedbecausethe varying NO,conditionsin thebasescenarioscausedincrementalreactivitiestovary widely. For many VOCs, relative reactivities also variedwidely in the basecasescenarios,and different scalescan beobtaineddependingon themethodsusedto deriveasinglescalefromthedistribution ofvaluesamongthescenarios.Threediffer-entmethods,discussedin thenextsection.wereemployed.Com-binedwith thethreemethodsfor quantifyingozone,theseyieldednine differentbasecaserelativereactivity scales.

The “Average Ratio” (AR) Method. This consistsof simplyaveragingtherelativereactivitiesin thebasecasescenarios,with

• eachscenariobeingweighedequally.This is usedto derive the• relativereactivitiesin the adjustedNO, scales.However, unlike

theadjustedNO, scales,thequantitiesbeingaveragedarequitevariable.Thefactthat thismethodweighsthe relativereactivitiesinall scenariosequally, despitethe fact that ozone is much moresensitiveto VOC changesin somescenariosthanin others,suggeststhat this maynot give an optimumscalefor controlapplications.Amoreoptimumscaleshouldgive greaterweight to scenarioswhicharemoresensitiveto the quantitiesbeingregulate&

The “Least SquaresError” Methods. These are basedonminimizing the calculatedsum-of-squareschangein ozonethatwould result if asubstitutionwhichthescalepredictswould havezeroeffectonozonewereappliedthroughoutthe setof scenarios.Model calculationsof substitutionswhich areactivity scalepre-dicts hasno effecton ozonearereferredto as“null tests”of thescale.For example,if the relativereactivity of acompoundin ascalewere0.5.thenthescalepredictsthatsubstitutionoftwo unitsof thecompoundfor oneunit of thebaseROGwould resultin nonetchangein ozone.A null testcalculationwould beasimulationof the effect of this substitution. Since, in general, relativereactivitiesvaryfrom scenarioto scenario,anull testsubstitutionwouldcauseachangein ozonein at leastsomeof thescenariosnomatterwhatrelativereactivitywereused.Thischangecanthenbethoughtofasameasureof the“error” of thereactivityscaleforthescenario.The leastsquareserrorrelative reactivity is the valuewhich minimizesthe sum of squaresof this error, or changeinozone,resulting from this null test.Note that this methodgivesgreaterweight to scenarioswhere ozone is more sensitivetoVOCs.

Sincerelativereactivityis definedasreactivitiesrelativeto thebaseROG.the relevantsubstitutionstrategiesfor deriving thesescaleswould involve either (I) reducingemissionsof theVOCandoffsettingit by anincreasein theemissionsof all ROCs,or(2)reducingall ROGsandoffsetting it by anincreasein theVOC.Leastsquareserrormethod“Ll”is basedon minimizingtheerrorsin null testsof ROG for VOC substitutions,while leastsquareserror“L2” is basedon minimizingerrorsin null testsof VOC for

ROG substitutions,It can be shown2°thatthe Base(LI) relativereactivitiesarethe sameas the weightedaverageof the relativereactivitiesin theindividual basecasescenario,wheretheweight-ing factoris thesquareof the incrementalreactivityof the baseROGmixture.Ontheotherhand,theBase(L2)relativereactivitiesarethereciprocalsof theweighedaveragesof the reciprocalsofthe relative reactivities in the scenarios,where the weightingfactoris thesquareof the incrementalreactivityof the VOC.

Method L2 may seem preferable from a control strategyperspectivebecausemostsubstitutionsinvolve replacingcurrentemissionswith somelessreactiveVOC.whichis thebasisof thenull testtheL2 methodis designedto optimize.However,methodLI is moretractablemathematicallybecausemethodL2 doesnotgivewell-definedresultsfor VOCswhoseincrementalreactivitiesvaryaroundzero.’andbecausetherelativereactivitiesofmixturesin the Base(LI) scalescan be derivedby linearsummationsofrelativereactivitiesof their components.This is not thecasefortheBase(1.2)scales.

Cnemlcal MechanismThechemicalmechanismusedin thisstudyis thatof Carter,~

with updatesfor severalVOC5.11.20 A completelisting of the• updatedmechanismis given elsewhere.°This mechanismcon-.

t.ainsrateconstantandproductyield assignmentsfor almost 120separateVOCs.It wasevaluatedby simulatingresultsof avariety

• ofenvironmentalchamberexperiments,andwasfoundto beableto simulate maximum ozoneconcentrationsand ratesof NO

• oxidation andozoneformation to within ±30%for 63% of theexperiments.73However, it had a slight bias (—15%) towardsoverpredictingmaximum ozone concentrationsin the experi-

I mentsdesignedto representambientmixtures.33This is compa-rableto or slightly betterthan the performanceof theRADM-ll~4~~andCarbonBondW36mechanismsin simulatingthesame35

or a similar76 database.This is as good as can be reasonablyexpectedgivenourcurrentknowledgeof atmosphericchemistryandcharacterizationof chamberartifacts.3”

This mechanismis consideredappropriateforreactivitycalcu-lations becauseit is at leastas up to dateas the other available

comprehensivemnechanisms3’.)&)’andit is the only onedesignedto representlargenumbersof VOCs which hasbeenextensivelytestedagainstchamberdata.However,its limitationsanduncer-taintiesmustberecognized.Mostofthemechanismrepresentsthestateof knowledgeasof 1989-1990andis out of date in somerespects.At thetime it wasdeveloped,theavailablechamberdataweresufficient to test the representationof —20 representativeVOCs, andthereactionsfor mostof the otherswerederivedbyextrapolationorestimations,32It hasnotbeenupdatedto takeintoaccountresults of recentexperimentalstudies of incremental

reactivities of a variety of VOCs.’°The uncertaintiesin thereactionmechanismobviously must betakeninto accountwhenthe resultsof modelcalculationsof reactivitiesareusedto assessozonecontrolstrategies.To aid in suchassessments,the masterlistingof reactivityresultsgivenin thefollowing sectionincludesfootnotesindicating levelsof uncertaintyin the mechanismsforthe variousVOCs, basedon the amountof experimentaldataavailableto testthemechanismsatthetime theyweredeveloped.

ResultsandDiscussionTableIll givestheincrementalre~ctivitiesin theMIR scaleand

therelativereactivitiesin all nineadjustedNO, scalesfor alIthetypes of VOCs in the mechanism. Comparisonsof relativereactivitiesin thebasecasescalesareshownforselectedVOCsinfiguresgivenlaterin thispaper.Thesedataarediscussedbelow,first in termsof the variability and differencesof incrementalreactivitiesandits components,andthen tn termsof thediffer-encesandvariabilitiesof therelativereacttvities.Finally, reactiv-

AIR & WASTE • Vol.44 • July 1994 • 337

TECHNICAL PAPER. •,.

TableilL Tabulationof tocrerneotalreactivitiesin theMIR scaleandrelativereactivitiesin the adjustedNO~scales,with notesconcerningtheuncertaintyoftheVOCs ~

MIR Relative ReactivIty (hiCompound . lnc.Rct. Ozone Yield Integrated Ozone

fal MIR MUIR EBIR MIll MOIR (SIR

Carbon Monoxldr 0.054 0.018 0.032 0.04.4 0.016 0.023 0.029

AlkanesMethane 0.015 0.005 0.008 0.010 0.004 0.005 0.006Etnane 0.25 0.079 0.140 018 0.061 0.078 0.088Propane 0.48 0.16 0.27 0.33 0.128 0.17 0.19n-Butane 1.02 0.33 0.57 0.70 0.26 0.33 036n-Perrtane 1.04 0.33 0.58 0.71 0.29 0.37 0.42~n-Héxane 0.98 0.31 0.55 0.65 • 0.28 0.35 038n-4leptane 0.81 0.26 0.45 0.49 • 0.22 0.26 0.25n-Octane . 0.60 0.19 0.34 0.33 0.15 0.17 0.137n-Nonane • 0.54 0.17 0.30 0.27 0.132 0.125 0.063n-Decane 0.46 0.146 0.26 0.22 0.109 0.091 0.021n-Undecane 0.42 0.132 0.23 0.19 0.095 0.072 -0.002n-Dodecane 038 0.118 0.21 0.16 0.082 0.058 -0.016n-Tridecarje 035 0.110 0.19 0.15 0.074 0.049 -0.025n-Ietradecane 0.32 0100 0.17 0.136 0.067 0.041 -0.031

lsobutane 1.21 0.39 0.63 0.80 0.33 0.43 0.51Neopentane 0.37 0.117 0.18 0.21 0.092 0.105 0.104Iso-Pentajie • 1.28 0.44 0.74 0.93 0.38 0.48 0552,2-Dimethylbutane 0.82 0.26 0.43 0.51 0.21 0.26 0.272,3-Dimethylbutane 1.07 0.34 0.57 0.72 0.32 0.43 0.512-Methylpentane 1.5 0.48 0.76 0.90 0.42 0.49 0.523-Methylpent.ane 15 0.48 0.80 0.99 0.42 0.52 0.582.2.3-Tzimetylbutane 1.32 0.42 0.68 0.84 0.38 0.49 0.572,3-Dhrneniylpentane 1.31 0.41 0.67 0.79 0.37 0.44 0.482,4-Oimettiylpentane 1.5 0.48 0.73 0.86 0.42 0.48 0.503,3-Olmethyipentajie 0.71 0.22 0.39 0.44 0.18 0.22 0.222-Methylbexane 1.08 0.34 0.57 0.69 0.30 0.37 0.4,0

• 3-Mèthylbexarle 1.40 0.44 0.71 0.82 0.38 0.44 0.462.2,4-Thmemylpentane 0.93 0.29 0.46 • 0.49 0.24 0.27 0.242,3,4-Trimethylpeniane 1.6 0.51 0.78 0.94 0.47 0.55 0.612,3-Oimethythexane 1.31 0.41 0.67 0.79 0.37 0.44 0.482,4-DImethythexaj~e 1.5 0.48 0.73 0.86 0.42 0.48 0.502,5-Oimethyfhexane 1.6 0.52 0.79 0.96 0.49 0.59 0.672-Methylbeptane 0.96 0.30 0.51 0.57 0.26 0.30 0.293-Methylheptane 0.99 0.31 0.53 0.60 0.27 0.32 0.324-Methylheptane 120 0.38 0.59 0.65 0.32 0.35 0.332.4-Oirnethytheptane 133 0.43 0.63 0.72 0.37 0.41 0.402,2,5-TrimethYffiflane 0.97 0.30 0.49 0.54 0.26 0.29 0.274-Ethylheptafle 1.13 0.36 0.54 0.59 0.30 0.31 0.283,4.PropylheptaM 1.01 0.31 0.47 0.50 0.26 0.26 0.223,5-Diethylheptafle 123 0.43 0.63 0.72 0.37 0.41 0.402,6-DictQ’icctane 1.23 039 0.58 0.68 0.35 0.37 0.38Cyclopentanc 2.4 0.76 1.19 1.46 0.67 0.80 0.89Methylcyclopentafle • 2.8 0.90 1.32 1.5 0.82 0.96 1.05Cyclohexane • 1.28 0.41 0.63 0.69 0.35 0.38 0.371,3-Dirnethyicyclohexane 2.5 0.81 1.18 1.43 0.76 0.89 1.00Methytcyclohexane 1.8 0.59 0.84 0.96 0.51 0.56 0.55Ethylcyciopentane 2.3 0.73 1.10 1.31 0.66 0.74 0.78Ethylcyclohexane 1.9 0.62 0.86 0.97 0.55 0.57 0.561-Ethyl.4.MethylcyclOhenne 2.3 0.73 1.00 1.15 0.66 0.70 0.711,3-Diethylcyclohexafle ~.8 0.57 0.79 0.92 • 0.51 0.55 0.561,3~lJiethy1.5_Methy1cyclOt1eXafle 1.9 0.61 0,84 1.00 0.57 0.61 0.651.3,5-Triethylcyclohexane 1.7 0.54 0.74 0.86 0.49 0.53 0.55

Alkanes

Ethene ~ 2.4 2.8 3.2 2.2 24 2.7 2.3 2.6 2.9• Propene , ~ 3.0 3.2 3.7 3.0 3.0 3.2 3.0 3.1 3.4 4

1-Butene • 8~ 2.9 3.0 3.4 2.8 2.7 2.8 2.8 2.8

888 • July 1994 • vol44• AIR & WASTE

-~

Uric.lnt’rJ Ozone>90 ppb Notes

MIll M0~ (BIll (ci

0.017 0.025 0.034 1

0.004 0.006 0.008 20.089 0.091 0109 20.139 0.19 0.23 20.29 0.38 0.44 10.31 0.41 0.48 50.29 0.39 0.44 50.24 0.30 0.30 5017 0.21 018 50.150 0.16 0113 50.128 0.130 0.070 70.114 0.110 0.046 80.101 0.094 0.028 80.093 0.082 0.018 80.085 0.073 0.011 8

0.36 0.48 0.60 70.104 0.123 0.126 70.41 0.54 0.63 70.23 0.30 0.33 70.33 0.45 0.56 50.45 0.55 0.60 70.45 • 0.58 0.67 70.40 0.53 0.65 70.39 0.49 O.55 70.44 0.53 0.58 70.20 0.26 0.27 70.32 0.41 0.46 7 •

0.41 0.49 0.53 70.27 0.31 0.29 70.49 0.60 0.69 70.39 0.49 0.55 70.44 0.53 0.58 70.51 0.63 0.74 70.28 0.34 0.35 70.29 0.36 0.38 70.35 0.40 0.40 70.40 0.45 • 0.47 70.28 0.33 0.33 70.33 0.36 0.35 70.29 0.31 0.28 80.40 0.45 0.47 a037 0.42 0.45 8

0.71 0.89 1.01 70.86 1.03 1.17 70.38 0.43 0.43 70.78 0.95 1.10 80.55 0.62 0.64 50.69 0.81 0.90 80.59 0.64 0.65 80.70 0.76 0.81 80.55 0.60 0.64 80.59 0.65 0.73 8

• 0.51 0.57 0.62 8

2.9 4

MIR Relative Reactivity LW Unc.Compound lnc.Rct. Ozone Yield Integrated Ozone lnt’d Ozone >90 ppb Notes

Cal MIR MUIR EBIR Mill MUIR EBIR MIll MUIR SIR (ci

1-Pentene 6.2 2.0 2.1 2.3 20 1.9 1.7 2S 1.9 1.9 73-Methyl-1-Butene 6.2 2.0 2.1 23 2.0 1.9 1.7 2.0 1.9 iS 71-Hexene 4.4 1.40 1.46 1.5 1.32 1.17 1.00 1.38 1.27 114 41-Reptene 3.5 1.10 115 1.18 1.02 0.88 0.69 1.08 0.97 0.82 81-Octene 2.7 0.85 0.90 0.89 0.76 0.62 0.41 0.82 0.71 0.55 81-Nonene 2.2 0.71 0.74 0.72 0.62 0.48 0.28 0.68 115$ 0.41 8

lsobutene 5.3 L7 1.6 1.9 2.0 21 2.5 1.9 2.0 2.3 52-Methyl-i-Butane 4.9 1.5 t6 1.9 1.7 1.9 2.2 1.6 1.7 ZO 7trans-2-Butene 10.0 3.2 3.2 3.6 3.5 3.6 4.0 3.3 3.3 3.6 5cis-2-Butene 10.0 12 32 3.6 3.5 3.6 4.0 3.3 3.3 3.6 52-Pentenes 8.8 2.8 2.8 3.1 3.0 31 3.2 2.9 23 2S 72-Methyl-2-Butene 6.4 2.1 2.0 2.2 2.4 2.7 • 3.1 2.2 22 2.5 72-F4exenes 6.7 2.2 2.2 2.3 2.3 2.3 2.3 2.2 21 2.1 82-Heptenes 5.5 1.8 12 1.9 1.9 1.9 1.9 1.8 12 12 83-Octenes 5.3 LB 1.6 1.7 L7 1.6 1.5 1.6 1.5 1.5 83-Nonenes 4.6 L46 1.45 15 L48 1.40 1.32 i.46 1.37 1.29 8

13-Butadiene 10.9 3.5 35 4.1 3.5 3.5 3.6 3.5 14 3.6 8lsoprene 9.1 2.9 2.9 3.3 3.2 31 3.2 3.0 2.9 3.1 6Cyclopentene 7.7 2.4 2.3 2.5 2.9 3.1 3.5 2.5 2.6 2.8 8Cyclohexene 5.7 1.8 1.9 21 20 2.0 2.1 12 1.8 1.9 8a-Pinene 3.3 1.04 1.08 1.23 ijI 111 1.17 1.06 1.05 1.09 5p-Pinene 4.4 1.40 1.42 12 1.5 L47 1.5 1.44 1.41 L46 8

Acetylene:Acetylene 0.50 0.16 0.28 0.38 0140 0.19 0.23 0.15 0.21 0.27 5Methylacetylene 4.1 1.31 1.8 2.2 1.14 1.24 1.28 1.21 1.36 1.45 9~Aromatlcs

Benzene • 0.42 0135 0.114 0.051 0.112 0.113 0.097 0.123 0.122 0.101 4

Toltmne 2.7 0.88 0.53 -0.023 0.77 0.71 0.54 0.82 0.72 0.48 4Ethylbenzene 2.7 026 0.52 0.007 0.77 0.72 0.54 0.82 0.72 0.49 7n-Propylbenzene 2.1 0.68 0.41 -0.016 0.59 0.55 0.41 0.64 0.56 0.37 7Isopropylbenzene 2.2 0.71 043 -0.007 0.64 0.59 0.44 0.68 (159 0.40 7s-Butylbenzene 1.9 0.60 0.37 -0.014 (154 0.49 0.38 0.57 0.51 0.33 7

o-Xytene 6.5 21 1.6 1.26 2.0 2.0 1.8 2.1 2.0 1.7 4p-Xylene 6.6 21 1.7 1.29 2.1 2.0 1.8 2.1 2.0 LB 7ni-Xyiene 8.2 2.6 2.1 1.7 2.7 2.7 2.6 27 2.6 2.5 4

1,3,5—Trirnethyibenzene 10.1 • 3.2 2.6 2.4 33 3.8 4.0 3.5 3.5 3.5 41,2,3-Tiimethylbenzene 8.9 2.8 23 1.9 3.0 35 3.0 2.9 2.9 2.8 71,2,4-Tdmethylbenzene 8.8 2.8 2.3 1.9 3.0 3.0 3.0 2.9 2.9 2.8 7

Tetralin 0.94 0.31 0103 -0.23 0.33 0.26 0.147 0.31 . (123 0576 5Naphthalene 1.17 0.37 0.066 -0.43 0.36 0.26 0.064 0.36 $124 -0.009 5Methytnaplithalenes 3.3 1.05 0.65 0.21 111 0.96 0.75 1.07 0.90 1165 82,3-Dimethylnaphthalene . 5.1 1.6 1.16 0.79 1.8 1.7 1.48 1.8 1.5 121 5Styrene 2.2 071 -0.26 -1.8 022 0.54 0.043 0.73 0.32 -0.40 8

Alcoholsand Ethers

Methanol 0.56 0.18 0.23 0.28 0.149 0.18 0.19 0.17 0.20 0.23 • 1Ethanol 1.34 0.43 0.61 0.72 0.33 0.36 0.37 0.38 0.43 0.45 1n-Propyi Alcohol 2.3 0.72 0.95 1.07 0.60 0.63 0.61 0.66 0.71 0.70 7Isopropyl Alcohol 0.54 017 0.27 0.35 . 0.17 0.22 0.27 0.17 0.23 0.30 7n-ButyI Alcohol 2.7 0.86 1.09 1.25 0.74 0.77 0.76 0.79 0.84 0.86 7lsobutyl Alcohol 1.9 0.62 0.80 0.92 0-54 0.58 020 0.58 0.64 0.67 7t-Butyi Alcohol 0.42 0.132 0.21 0.27 0114 0.148 018 0125 0.17 0.21 7Dirnethyl Ether • 0.77 0.24 0.48 0.68 0.24 0.35 0.47 0.24 0.38 0.53 7Methyl. t’Butyl Ether 0.62 0.20 0.34 . 0.47 0.18 0.25 0.31 0.19 0.27 0.36 1Ethyl t-Butyl Ether 2.0 0.64 0.88 111 0.62 0.77 0.92 0.63 (180 0.98 • 7

AIR & WASTE • Vol.44 • July 1994 • 889

Aromatic Oxygenatai

ECHNICALFAPm

Methyl Nitrite

Base ROG Mixture

9.5 3.1 3.5 5.2 5.2 6.9 10.2 3.5 4.7 6.9 3

3.1 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

[a} Incremental reactivities in units of grams ozone formed per gram VOC emitted for the ozone yield reactivity scale for the MIR scale.Note that there are small differences in the last digit in the tabulated values compared to those used in the California ARB Regulationsl2for some VOC5. This is within the numerical uncertainties of these calculations, and is because of minor changes in the softwareused.2O

[b] Incremental reactivities of the VOCs (in units of ozone per gram of VOC) divided by incremental reactivities of the base ROG mixture.[ci Notes concerning the uncertainty of the mechanism are as follows:

1 Least uncertain mechanism, and tested against chamberdata2 Mechanism probably not uncertain, but was not tested.3 Laboratory data are available for the major reactions in the mechanism, but the mechanism was not tested.4 Uncertain portions of the mechanistn are adjusted or parameterized to fit chamber data.5 The mechariisni is uncertain, and only limited or uncertain data were available to test it.6 The mechanism was not optimized to fit existing chamber data.7 The mechanism was estimated and was not tested.8 The mechanism was estimated and was not tested, and must be considered to be highly uncertain.9 The mechanism was estimated and was not tested, and is likely to be incorrect. Suitable only icr estimating reactivities of mixtures

where this is a component

ity adjustmentfactors forselectedvehicle exhaust mixturescalcu-fated using these various scales are compared.

Distributions of NO1 and Reactivity In the ScenariosFigure 2 shows the distribution plots of the maximum ozone

concentrations,two measures of the NO1 levels, and the baseROGincremental reactivities for the various scenarios. Note that thewde distributions of ROG/N01 ratios in the adjusted NO, sce-narios indicate that the ROG/N01 ratios are, by themselves, poor

• predictorsof NO,availability. A muchbetterpredictor is the ratioof the NO, input to the NO, yielding maximum ozone concentra-

• tions,or the NO,INO,MOR ratio. This rario. which is I by definitionfor the MOR scenarios, was found to be narrowly distributedaround 1.5 for theMIR scenarios, and 0.7 for the EBIR scenarios.On the other hand, as expected, it varies widely among the basescenarios. By this measure, most of the EPA base scenarios usedin this work are between MOR and EBIR conditions.

The distribution of maximum ozone levels is very similar in allthe scenarios, being only slightly lower in the MIR scenariosrelative to the others. Thus, while MIR conditions are nor opti’

urn for ozone formation, levels of ozone which exceed air.~alitystandards are still formed. On the other hand. the distribu-tion ofbase ROG incremental reactivities are significantly higher

• in the MIR scenariosthan in the lowerNO, MOL0rEBIRscenarios.

The wide distribution of base ROG reactivities in the base scenariosis expected, given their distribution of NO1 conditions.

In a previous derivation ofgeneral VOC reactivity scales,” adifferent set of scenarios obtained from the studies ofGery et al.’7

and ~ were used. To show how the distribution of NO~conditions from these scenarios compare with the EPA-derivedscenarios used in this work, Figure 2 also shows plots of NO,!NO,M0R for the Gery et al.” and WhitteniS scenarios. These aredesignated as “base (199l)”on the figure. This shows that these.scenarios have a much wider distribution of NO1 conditions thanthose used in this work. The implications of this on reactivityscales are discussed later.

Factors Affecting ReactivityDependence of Reactivity on EnvironmentalConditions. Table

Ill shows that, as expected, the incremental reactivities of a givenVOC vary significantly with NO, conditions. However, it does notshow theextent to which the reactivities or their components varyamong the individual adjusted NO, and base case scenarios. Anillustration of this is shown in Figure 3, which shows distributionplots of kinetic, mechanistic, incremental, and relative ozoneyield reactivities for carbon monoxide and toluene. Althoughthese are only two of the many types of VOCs, together theyillustrate the trends which are characteristic of most other VOCs

Mill Relative ReactivIty tbllnc.Rct. Ozone YIeld Integrated OzoneCompound

Aldebydes

FormaldehydeAcetaldehydeC3AldehydesGlyoxalMethyl Glyoxal

Keto no:

AcetoneC4 Ketones

lnt’d Ozone >90 ppbLa! MIll MUIR SIR MIR MUIR SIR MIR MUIR SIR

7.2 23 12 1.7 2.8 2.9 3.2 2.5 2.5 2.65.5 1.8 1.8 2.2 1.7 1.5 1.31 1.8 LB LB6.5 2.1 2.1 2.4 2.1 1.8 L49 2.1 1.9 1.72.2 0.71 0.61 0.63 OSO 0.90 0.95 0.75 0.80 0.8514.8 4.7 4.0 3.9 6.5 7,1 6.2 5.3 5.5 6.1

0.56 0.18 0.17 0.18 0.16 0.142 0.134 0.16 0.16 0.149 51.18 0.38 0.46 0.53 0.32 0.30 0.29 0.34 0.34 0.34 5

tinc.Notes

(ci

I4533

BenzaldehydePhenolAlkyl Phenols

Othen

-0.57 -0.181.12 0.362.3 0.7•4

-158 -2.7 -0.31 -0.74 •1.5 -0.25 -0.74 -1.7-0.45 -1.7 0.32 0.091 -0.42 033 0.0147 0.65-0.53 -2.6 0.69 0.33 -0.53 0.70 0.20 -0.92

575

a~o• July 1994 • Vol.44 • AIR & WASTE

FIgure 2. Distribution plots of maximum ozone, the base ROG incremental reactivity, the BOG/NO, ratio, andthe ratio of NO, inputs to MOB NO, inputs for the MIR. MOR. EBIR and base case scenarios. Base scenariosused previouslyt1 are shown on the No,/MOR NO1 plot.

Carbon Monoxide~- —

.4 1~~~

— —

Toluene—

—.4. I....—,

~4 1—4 ..q — 1__,..

— —

~— .4 1~~~

— —

~44 1—

— —

I1~:E.E~L1~~ •

— —

It. — — 1

to varying degrees.The sensitivity of kinetic reacti)?ities

to scenario conditions depends on howrapidly the VOC reacts, with slowly re.acting compounds being most sensitive,and having kinetic reactivities which areessentially proportional to the integratedlevels of species with which the VOCreacts. Rapidly reacting compounds havekinetic reactivities approaching unity.Most slowly reacting compounds reactonly with OH radicals, so variations intheir kinetic reactivities reflect variationsin integrated OH radical levels. CO reactsso slowly that its kinetic reactivity is es-sentially proportional to the integratedOH. and thus the distribution plot of itskinetic reactivities is the same as the dis-tribution plot of integrated OH. The dis-tilbution for kinetic reactivities of toRi-ene, which reacts more rapidly, is qualita-tively similar but varies over a narrowerrange.

The distribution plots show that ki-netic reactivities, and thus integrated OHradical levels, are lower in the MIR sce-narios compared to MOR and EBIR con-ditions. This is attributed to the fact thatNO, is involved in a number of radicaltermination reactions. However, the ki-netic reactivities do not significantly in-crease as NO, is reduced from MOR toEBIR levels. Reducedterzninationcausedby lower NO, when going from MOR toEBtR is offset by the increased termina-tion due to HO2 + HO2 and otherperoxy +peroxy reactions which become moreimportant once NO, is consumed. On theother hand, the wide distributions of ki-netic reactivities in the adjusted NO, sce-narios indicate that kinetic reactivities aresignificantly affected by other factorsbesides NO,. Factors such as light inten-sity, temperature, and dilution mightbeofequal or greater significance as NO, inaffecting radical levels and thus kineticreactivities.

Since the dependence of kinetic reac-tivity on NO, is in the opposite directionas that for incremental reactivities, themechanistic reactivity must be thedomi-nant factor affecting how incrementalreactivities vary with NO,. In contrast tothe case with kinetic reactivities, for mostVOCs there is also almost no overlap inthe distribUtions ofmechanisticreactivitiesin the MW and MOR scenarios; the datain Figure 3 are typical in this regard. Thus,at least when NO, is above MOR levels,NO, availability dominates scenario con-ditions in affecting mechanistic reactiv-ity. While NO, is still important in affect-ing mechanistic reactis~tieswhen NO, isbelow MOR levels, the other factors be-come relatively more important, as indi-catedby theoverlap in the distributions of

—

t__L c~r rr— ~4 —

._ ICC *

1~ i~i_~— — 1—’——

I1

aI

FIgure 3. Distribution plots of kinetic, mechanistic, incremental, and relative reactivitiesforcarbon monoxideand toluene in the MIR, MOB, EBIR and base case scenarios.

•AIR & WASTE • Vol. 44 • July 1994 • 691

• • •the MOR and EBIR mechanistic reactivities for both CO andtoluene.

VOCs form ozone by producing HO2 and other peroxy radicalswhich react with NO to shiftthe NO-NO2-O3 photostationary statetowards ozone formation. The mechanistic reactivity of CO pro-vides a direct illustration how much ozone is formed by thisprocess, since the reaction of CO involves only the formation ofa single HO2 radical. Thus, no more than one molecule of ozonecan be formed from each molecule of CO which reacts. Thistheoretical maximum is almost achieved under MIR conditions,but the yield of 03 fromHO2 decreases rapidly as NO, is reducedfrom MWto MOR levels. This is because the H02+HO2 reaction,forming H2O2, begins to compete with the reaction of HO2 withNO in the scenarios where all the NO, is consumed before the endofthe simulation. Since essentially all VOCs react to form peroxyradicals, this factor contributes to the NO, dependencies of mecha-nistic reactivities of almost all cases,

For most other VOCs, this is not the only factor affecting howtheir mechanistic reactivities depend on NO,. If a VOC is asignificantradical source, itwillhavehigh mechanisticreactivitiesunder high NO, conditions where final ozone concentrations aredetermined by how rapidly ozone is formed. However, this en-hancement would be less important under lower NO, conditionswhere ozone yields are determined less by radical levels and moreby NO,availability. On the other hand, if a VOC has NO, sinks in

0037T.Id e b*~so3 a~’4Q3>$O,I. Orn,...ts* 2 ~,,,, ,3 3 ~ a

4 A** d S *. CC S Mall SOC 4

II

CAIC0~ICaCa a ~ 1* L. U a ‘~7a

• s—cl,

a a — 45 u.~ u ,.~ a

its mechanism, it will cause low, and even negative, mechanisticreactivities under low NO, conditions, but will not significantlyaffect high NO, (MW) reactivities. Toluene is an example of acompound where both factors are operative, and consequently itsmechanistic reactivity, and thus its incremental reactivity, de-creases much more rapidly as NO, is reduced than is the case forCO and other VOCs. Note that the NO, sink effect becomesdominant when NO, is sufficiently low, causing reactivities tobecome negative. This is despite the fact that toluene is stillcalculated to form radical-initiating products and radicals thatreact with NO. Forthe mechanism used in this work, the crossoverfor toluene reactivity occurs at some NO, level around the equalbenefit point, though the exact level appears to be highly variabledepending on other scenario conditions. (The crossover occurs athigher NO, levels when the Carbon Bond IV mechanism isused.39)

One would expect relative reactivities to be much less depen-dent on scenario conditions than incremental reactivities, at leastfor scenario conditions which affect reactivities of all VOCs insimilar qualitative ways. Thus, while decreasing NO, levelscauses decreased incremental reactivities in all compounds, thiseffect, at least to some extent, cancels out when consideringrelative reactivitjes. However, if a VOC differs significantly inhow its incremental reactivities vary withNO, than is the case forthe base ROG mixture, then its relative reactivity will also vary

________________ with NO,. For example, the mechanisticreactivities of CO are less dependent onNO, than most VOCs so its relativereactivides increase with decreasing NO,,while the opposite is true for toluene. Forsimilar reasons, one would expect thedistribution of relative reactivities in theadjusted NO, scenarios to be much nar-rower than the corresponding distribu-

• lion of incremental reactivities. This isindeed the case for the MW scenarios,but the relative reactivities appear to bemuch more variable in the MOR and(especially) the EBIR scenarios. Thusnon-NO, scenario conditions appear toaffect incremental reactivities of differ-ent VOCs similarly underhighNO,, MIT(conditions, but this is apparently not thecase undermore NO,-liinited conditions.The variations in relative reactivity arediscussed further in the following see-lion.

RelativeReactivitiesThe relativereactivities ofa variety of

VOCs in the various scales are comparedgraphically in Figures 4-6. (To aid incomparisons of the different scales, thedashed and dotted lines show the relativereactivities in theMIR and MOW scales,respectively.) Figures 7-9 show how wellor poorly the relative reactivities of anumber of different VOCs compare indifferent scales by plotting the relativereactivities of the VOCs in one scaleagainst those in theother. The position ofthe points for the VOCs should be com-pared to the I:! line where they wouldfall if they had equal relative reactivitiesin the two scales. The error bars in Fig-tires 4-9 indicate the standard deviations

- C

Figure4. Comparison ofrelative reactiviti~sof carbon monoxide. ethane, n-butane, n-octane, n-pentadecane,and ethene calculated using various methods. Points on rigbt are Ozone yield relative reactivities forthe variedaveraged conditions scenarios.

I.1II

2~’~a

N—WI—

4* U 13 ~

II

,IOCTAMC

— ,, ‘2 ‘., —ia

._.2~J,4

IIII

S

11 ~ a a.~,a

ç4I

III

892 • July 1994 • Vol.44 • AIR & WASTE

of theaverages or the derivations, and thus for the adjusted NO,scales. they indicate the importance of non-NO, scenario condi-tions in affecting relative reactivities. These results are’discussedin more detail below.

Ozone Yield Relative Reactivities — Adjusted NO1 Scales.The results show that the ozone yield relative reactivities candepend significantly on NO, for many VOCs. with the trenddepending on the type ofVOC. Relative reactivitiesof the aromat-ics and other compounds with major NO1 sinks decrease signifi-cantly as NO, is reduced, with the effect being largest for thecresols and benzaldehyde, the compounds with the strongest NO,sinks. The higher alkenes apparently have similar balances offactors in affecting reactivity as the base ROG mixture, since theirrelative reactivities appear to be almost independent of NO1, atleast in the MIR to MOR regimes. The relative reactivities ofcompounds which have weaker than average NO2 sinks, such asCO. ethene, and methanol, tend to increase with decreasing NO,because their incremental reactivities are less sensitive to NO,than that for the base ROG rilixture, which includes a significantcontribution from aromatjcs. In addition, relative reactivities ofslowly reacting compounds such as CO and ethane tend to in-crease with decreasing NO1 because kinetic reactivities. whichincrease as NO1 is reduced, are relatively more important inaffecting reactivities ofslowly reacting compounds. Because COis both slowly reacting and has essentiaily no NO, sinks, it

provides the most extreme case of a compound whose relativereactivity increases with decreasing NO4.

The distribution plots in Figure 3 and the lengths of the errorbars (the standard deviations) in Figures 4-6 provide an indicationof how other scenario conditions affect relative reactivities Formost VOCs, the MIR relative reactivities are quite insensitive toscenario conditions, with the distributions shown in Figure 3being fairly typical. In general. the sensitivities to scenario condi-tions increase as the NO, decreases, with the most extreme cases.being the compounds. such as toluene. cresols. and benzaldehyde,with the large NO, sinks in their mechanisms.

Figure 7 shows the extent to which the relative reactivities ofpositively reactive compounds in the MIR and MOW scalescorrespond to each other. Although these relative reactivities arehighly correlated, the MIR scale tends to underpredict the MOWrelative reactivities for CO. the alkanes, and the alcohols, andoverpredict them for the aromatics, onafairly consistent basis. Onthe other hand, FigureS shows that, except for toluene. the MOIRand EBIR relative reactivities correspond very well. These twoscales are essentially equivalent to within the uncertaintiescausedby variabilities in non-NO, scenario conditions if one considersonly compounds which are positively reactive in both scales.These two scales also agree in indicating that phenols and cresolsare negatively reactive, while they are positively reactive in theMW scale. However, the discrepancy in the MOIR and EBIR_______________________ scales in the relative reactivities of tolu-

ene — and by extension the othermonoalkylbenzenes, which the currentmechanisms3.’36 assume have similarreactivities as toluene — is not insignifi-cant in view of the relatively large emis-sions of these compounds.

Ozone Yield Relative Reactivities —

Base Case Scales. Figures 4-6 show thevarious base case ozone yield relativereactivities for the representative VOCs,where they caE be compared with thosefor the adjusted NO, scales. As before,the “errorbarC show the standard devia-tions ofthe avenges or derivations. Theavenge ratiobase case [Base (AR)]ozoneyield reactivities have high standard de-viations because of the variation in rela-tive reactivities in the scenarios due tothe variation of NO, conditions. In mostcases the least squares error methods(Base (LI) and Base (L2)j give morewell-defined values, having standard de-

•viations which are comparable to orsmaller than those for the adjusted NO,scales. For most VOCs, the Base (L2)relative reactivities are esserhially thesame as the Base (LI) values. Thus, asone might expect, a reactivity optimizedfor assessing substitutions involving re-placing currentemissions with emissionsof a less reactive VOC is essentially thesame as one optimized forassessing sub-stitutions of highly reactive VOC5 forincreased emissions of all ROCs.

There are a few apparently anoma-lous Base (L2) values which can be seenfrom the data for the cresols and xi-pentadecane. These are cases when theincremental reactivities of the VOC aredistributed aroundzero, when the method

FIgure 5. Comparison of relative reactivities of propene, trans-2-butene, toluene, m-xylene, 1.3,5-trimethylbenzene. and formaldehyde calculated using various methods. Points on right are ozone yield relativereactivities for the varied averaged conditions scenarios.

— a ‘A Li IS ~., a Yy a

C 03 ~1Sd 0 wtd 23 - ~td OS > NO —IN. Orr.,s*I,. 2 O.fltw,n .3 3 Ia.’C* ~2

4 1ANS .2 ~. ICC S AlolI *00 8

I—:—anoc. LI U ,,a ,,a

IIII

.1III

— ~ — —

TCUKICa a 4* Ij 22 a 21 a

IIII

C’

I4fltt~

‘-, — l.~ a

III

_I.v.4

SS — — — — —

135-nd IOCb~— .2 .3 IN.y a ~ a

I— — — _J?_— _.,_~_* .1

III

AS 4, 12 ~a y,a

- T •‘

4..,

AIR & WASTE • VoL 44~July 1994 • 893

•TECHNICALPAPEK:

used to compute the Base (L2) reactivities is most sensitive to themost extreme values in the distribution. Because of the poorperformance ofthis method in these cases, and the fact that in mostother cases it yields essentially the same result as the LI method,it is concluded that the LI method is the better method to deriveleast squares error relative reactivities.

In most cases, the relative reactivities in the Base (AR) casescale tend to fall between those in the MOW and EBIR scale. Thisis as one would predict from the distribution ofNO,/NO,M0R ratiosin the base case scenarios. On the other hand, all ofthe Base (LI)and most of the Base (L2) relative reactivities lie somewherebetween the MIR and MOW values. More Mm-like values for theleast squares error scales are expected because the method used toderive them puts more weight on reactivities in scenarios whereozone is more sensitive to VOCs, i.e., which are closer to MIRconditions. However, unlike the least squares error scale given ina previous study,” where the Base (LI) scale corresponded muchbetter to the MIR scale than MOW, in this case the Base (Li)reactivities are somewhat closer to the MOIR scale.

The reason for the differences between this result and thosegiven previously’ arises from the fact that the scenarios employedin the previous study represented a more varied set of NO,conditions. This is apparent from the distribution plots of NO,!NOIMOR ratios in Figure 2. which include the distribution for thebase scenarios from Gery et al.” and Whitten3t used in the

‘a a AR U U ‘—7 — ~ —

previous’1 study (“base (1991)”), where they can be comparedwith distribution for the EPA scenarios used in this work. Al-though both sets have average NO,INO,MOR ratios near the MORrange, the much wider distribution of NO1 conditions in the base(1991) set results in a larger fraction ofscenarios which have near-MW or higher-than-MW NO, conditions. Since reactivities inthese high NO, scenarios are weighed most heavily in computingthe least squares error scales, these scales are highly sensitive tothe number of such scenarios in the distribution. In general, thewider the distribution of NO, conditions in a set, the closer theleast squares errorreactivity scale derived from it will correspondto an MW scale.

The appropriateness of base case reactivity scales from thisworkobviously depends on how well these EPA scenarios repre-sent the distribution of conditions where ozone pollution episodesoccur. Ttshould berecognized that MIR conditions probablyoccurin the atmosphere much more frequently than representedby theseEPA episodes. Each of these scenarios is based on the EPA’sassessment of the conditions of a near-worst ozone episode insome area, and thus represents a meteorological condition whichis near to the most favorable for ozone formation in that area. Thusmost other days would have less favorable meteorological condi-tions for ozone, including many days when unacceptable ozonelevels may still be formed. These would include days with lowerrates of NO1 removal because lower temperatures or light inten-

sities cause lower rates ofphotochemicalreactions. Slower NO, removal meansmore NO, availability, and thus moreMIR-like conditions. Since, as shown inFigure 2. ozone can still exceed air qual-ity standards underMiR conditions, suchmeteorological conditions, while notworst case, are not irrelevant to the prob-lem of urban ozone formation. If theseconditions were represented in a morecomprehensive set of scenarios, the re-sulting least squares error scales wouldcorrespond much more closely to theMW scale than observed in this work.

Integrated Ozone and IntO3>90 Rela-tive Reactivities. Figures 4-6 show rela-tive reactivities derived from the effectsofthe selected VOCs on integrated ozoneconcentrations (the IntO3 scales) andfromthe effects of the VOCs on integratedozone above 90 ppb (JntO3>90). wherethey can be compared with the ozoneyield reactivities discussed above. TheIntO3, fntO3>90, and ozone yield relativereactivities tend to agree well underMiRconditions, but then they tend to divergeas NO, is reduced, with the IntO, andIntO,>90values changing less as NO, isreduced than is the case for ozone yieldvalues. In a number of the most reactivecompounds. such as formaldehyde, m-xylene and trimethylbenzene, the IntO,relative reactivities are essentially inde-pendent of NO,. despite the fact that theNO, dependencies in the ozone yieldrelative reactivities are significant. TheIntO, relative reactivities of most othercompounds show the same trend withNO, as the 03 yield reactivities, exceptthat the change with NO2 is less extreme.This lower sensitivity oflntO, reactivities

o 03 2~N45

4 — .2

a

nAhaoftoca AR LI U YIy — ~ a

&C4 03

— ICC

— b,Vd 03 ~ NO pe3 As*.,s*. l2* A—I *00 .5

I

II

— ~_•~ _~.a~_~a—

*

v-, — , a

—‘4

C’4

.-t

. .

I

II

AR

I

U LI ,—~a ~—, a

II t • . ,S~43I

. .;-t _J&4g4

Id

U 22 y—~a ~., a

II

J;r ~‘C.

,3 ~II

---C KNZALDCIIY2C

I• S.—..R

til~.S~

FIgure 6. Comparison of relative reactivities of acetaldehyde, acetone, methanol, ethanol, cresols, andbenzaldehyde calculated using various methods. Points on ‘ight are Ozone yield relative reactivities for thevaried averaged conditions scenarios

894 • July 1994 • Vol. 44. AIR & WASTE

to NO,meansthat theya~lessvariablein thebasecasescen~os. sameaspectsof the VO~5•mech~isms.therelativereactivitiesBecause of this, the base case relative IntO, reactivities are lesssensitive to the method used to derive them, except for the fewanomalous Base (L2) cases discussed above.

The reason forthis lowersensitivity ofIntO3 relative reactivitiesto NO, — and their tendency to correspond to MW relativereactivities — is that integrated ozone levels are sensitive to thesame mechanistic factors which determine ozone yields underhigh NO,. MW conditions. These are the factors which affect howrapidly 0, is formed, as opposed to those which affect the ultimate0, yield when NO, is limited. In a high NO, scenario, both theozone yield and the integrated ozone would be determined by howrapidly 03 is formed. In a lowerNO, scenario, the integrated ozonewould still be sensitive to the ozone formation rate, but if the 0)is NO, limited the maximum ozone yield is more sensitive to theNO, availability than the ozone formation rate. While NO, avail-ability has some influence on integrated ozone under low NO,conditions, it tends to be less important a factor than the amountof time that the high—i ‘~-“!zof ozone were present. Thus, sinceIntO, reactivities and MW reactivities are both sensitive to the

i:ix o.zo

o CC.o ~,.T,QIICI

X Ai4•n,d.’

IS 0.30 Lx 2.30 LXMl loath, 2.odm (~nm~)

£ McCA~o A5~.~,••

V M.Ioq,e

tend to correspond to each other. NO, sinks in the VOCs’ mecha-nisms, which become the dominant factor a.ffecting MOW andEBIR reactivities, are only of secondary impol’tance in affectingIntO, reactivities.

One would expect the lntO,>90 reactivity scales to havecharacteristics somewhere between those for the IntO, and theozone yield scales, and this is indeed what is observed. However.the lntOp’90 scales are closer to IntO3 scales than the ozone yieldscales, and all the discussion above for the IntO7 scales areapplicable to IntOp9O. There are a few cases, such as formalde-hyde. trimethylbenzenes. and (to a lesser extent) acetone, ethanol.and methanol, where there is a non-negligible difference betweenthe ozone yield and the integrated ozone reactivities under maxi-mum reactivityconditions. In those cases, the 1ntO~>9Oreactivitiestend to be closer to the MIR reactivities.

Because of this. MW reactivities tend to give very goodpredictions of IntO,>90 reactivities in the base case scenarios.This is shown in Figure9. which gives plots of base case IntO,>90relative reactivities computed using the average ratio method___________________ against the values predicted by the MW

scale for selected representative posi-tively reactive VOC5. Agreement towithin the standard deviations are at-tained for all but two VOCs,and forthosethe agreement is within 1.5 standard de-viations. This is much better than thecorrespondence of the base case IntO,>90

reactivities with the MOIR or EBIRscales.

Effects of Variations of OtherScenario Conditions

The comparisons of reactivities in theadjusted NO, scenarios provide directinformation on their dependencies onNO, availabilities, and also, through thestandard deviations of the averages, pro-vide indirect information on the impor-tance of other conditions which werevariable in the scenarios. However, thecomposition of the base ROG mixture,the level and cornpositionsofROGs aloft,and the initial nitrous acid (HONO) as afraction of the NO, inputs were heldfixed in all these scenarios, and thusthese data provide no information on thesensitivities of reactivities to these in-puts. To assess this, modified versions ofthe averaged conditions scenario werederived by varying these inputs as de-scribed below. Next NO, inputs for eachversion were adjusted to derive corre-sponding MW or MOR scenarios. Fi-nally these were used to assess how thesevariations affect the MW and MOIRreactivities. Sensitivities of EBIRreactivities to scenario conditions are notdiscussed here, but they were generallyfound to be sin’iularto. though often greaterthan, those for MOIR reactivities.

Four different modifications of thecomposition of the base ROd were ex-amined, all involving relatively largechanges to this mixture. These involvedonly changes to the ROds associated

Figure 7. Plots of relative reactivities in the MOR scale against relative reactMties in the MIS scale forselected VOCs. The left plot shows the fuil range of relative reactivities, while the right plot has an expandedscale to show the less reactive VOCs more clearly.

Ii.,i.

0 —7:’

2C

I‘N

i.

+

E’g.

150 _tD, 0.12IS 8*th. bod,Iy (%.0 Os.)

o CD. Afran.Io AtC.nCflfl2 AAA.hyd.*

a Alcai,olt0 AJk.nn

V MSOaA,e

2C

0

tIC 0.20 0.30 0.40 t50 i~ 030 15icc o*e, LoaMy (g,m b~)

FIgure 8. Plots of relative reaclivities in the EBIR scale against relative reactivities in the Main scale forselected VOCs. The left plot shows the full range of relative reactivities, while the right plot has an expandedscale to show the less reactive Vacs more clearly.

AIR & WASTE • Vol.44 • July 1994 • 895

TECHNICALPAPEff

with anthropogenic emissions, a fraction of which (—60%) werepresent initially and the remainder emitted throughout the day.The compositions and amounts ofaloft and biogenic ROds inputwere not varied. The variations, and the code numbers used todesignate them, are as follows: (I) “No Oxygenates”: the aide-hydes and ketones were removed without modifying the levels ofthe other components: (2) “Oxygenates x3”: the aldehydes andketones were increased by a factor of 3 without modifying thelevels of the othercomponents; (3) “Aromaticsx2”: the aromaticswere increased by a factor of 2, and the alkanes and olefinsreduced to keep the total carbon the same; and (4) “Alkenes x2”:the olefins were increased by a factor of 2 and the alkanes andaromatics reduced to keep the total carbon the same.

FIgureS. Plots of relative reactivities in the Base (Li), lntC,>90 scale against relative reactivities in the MIRscale for selected VOCs. The left plot shows the full range of relative reactivities, while the right plot has anexpanded scale to show the less reactive VOCs more clearly.

ODSflMd 0 L$fdO3 — ~S’dD3,20ptI ~i. 0,w..,.tn 2 0.*1 ‘S S ij,.”.~,, .2£ A&~~~x2 S ,*.,00 I .s,n,o,s

MeS Laho,d

SI -Le Cs, £5 LSL VI~y ,a.o a

S

3~~_i_~__•~_

One of the changes made to the EPA scenarios was assumingthat -2% ofthe initial and 0.1% of the emitted NO, was in the formof nitrous acid (HONO). which is a powerfiul photoinitiator whichcould help initiate the photochemical processes early in the thy.The EPA scenarios as received from Baueues,~°and the scenariosused previously,” assumed no initial or emitted HONO. The (5)“No HONO”modification, where the initial nitrous acid (HONO)was assumed to be zero and no HONO was subsequently emitted.was used to assess the effect of this change.

The scenarios as received from Baugues all assumed a standard30 ppb of VOCs aloft, and the same chemical composition wasused for this aloft mixture in all scenarios. The (6) “Aloft ROdsxS” modification, where the concentrations ofall ROGsaloftwere

_____________________ increased by a factor of 5, was used toassess how important aloft ROCs are in

affecting reactivity calculations in thesescenarios.

The MIR and MOIR relativereactivities calculated using these modi-fied scenarios are shown for the repre-sentative VOCs on the right hand side ofthe plots in Figures 4-6. where they canbe compared with the values for the cor-responding averaged conditions scenario.The relative reactivities in the variedscenarios are indicated by the code num-bers on the plots, while the standard, oravenged conditions, values are indicatedby the dashed (for MW) or dotted (forMOW) lines.

The data in Figures 4-6 show thatthese relatively large variations in thebase ROd mixture had, in most cases,only small effects on the relativereactivities. The variation which had thelargest effect was the increase in the

aromatics (“3” on the plots), whose two-fold increase. forexample. caused a—l9%decrease in the relative MIR reactivity offormaldehyde. Removing the oxygen-ates from the base ROd (“I”) increased

the relative MW reactivity of formalde-hyde by —7%. The effects of these varia-tions on the other VOCs were generallysmaller.

The removal of initial and emittedHONO from the scenarios (“5”) had al-most no effect on any of the results ex-cept for formaldehyde, who~eMW and

MOIR relative reactivities increased by—15%, and whose integrated ozone rela-tive reactivities (not shown) increased‘by —20%. This is a large sensitivity inview of the almost complete insensitivi-ties of the other results to initial HONO.Since both HONO and formaldehydeprovide early radical sources in the simu-lations, this shows that removing onesuch radical source increases the sensi-tivity of the scenarios to the other.

The fact that these scenarios haveinitial HONO, while those given previ-ously’ do not might partly explain whythe formaldehyde reactivity in these sce-narios is less sensitive to changes inaldehyde emissions than calculated pot-

OC En—a a LI t~ — — ~-, —

3 ~~i:11t

H -

£25 C—

ITc.R

FIgure 10. Comparisons of reactivity adjustment factors for selected vehicle exhaust mixtures calculatedusing various methods. Points on right are calculated from ozone yield RAFs for the varied averagedconditions scenarios.

o CD. Akan~0 An~I,*S

2 A4dfl�,.

‘o.

S

ItI.

£ AJc0t~01t

0 A&*nISV atO.,

I‘—4

C.

C5

00A

001.1 1.30 LX 220 IX 220 LX ~ ~ a.,: tia 0.20 031 ~ usS bdw Locfm (Ps’ ~) IS toat~.1.ocMy (~b~)

_~_~e~

LPC C.h.u,tA. L~ 4-, — ~ —

896 • July 1994 • Vol.44 • AIR & WASTE

viously. in the absence of the initial HONO. we calculate that therelative reactivity of formaldehyde increases by — 11% when thebase ROC oxygenates are removed. (This is not shown on theplots.) This is greater than the —7% effect observed when the

- HONO is present, and indicates that adding radical initiators suchas HONO to the scenario reduces the sensitivity of formaldehydereactivities on initial aldehydes.

The fivefold increase in the aloft ROCs was found to have aninsignificant effect on the relative reactivity results. In view ofthis, the sensitivity to the composition of the aloft ROC mixturewould also be expected to be small.

Examples of ExhaustReactivityAdjustmentFactorsAn example of a regulatory application of a reactivity scale is

the utilization of reactivity adjustment factors (RAFs) in thealternative fuel vehicle exhaust standards recently adopted inCalifornia.’ The mass emissions of exhausts from alternatively-fueled vehicles are multiplied by these RAFs to place them on thesame ozone impact basis as e:,.:;c.;~sfrom vehicles using con-ventional gasoline. The RAFs are calculated from the ratios ofincremental reactivities (as ozone per gram) for the exhaustmixtures from alternative-fueled vehicles, relative to that for amixture characteristic of exhaust from vehicles using industry-avenge gasoline. The regulations as adopted utilize the MIR scaleto calculate these RAFs,’~but it is of interest to see how thesewould differ if other scales were used. This is shown on Figure 10,which gives RAFs for selected vehicle exhaust mixtures calcu-lated using the various reactivity scales. The example mixtures,which are based on analysis provided by the CARD,’3 includeexhausts from vehicles fueled with 85% methanol and 15%gasoline (M85). compressed natural gas (CNC), liquifred petro—leum gas (LPG), and 85% ethanol, 15% gasoline (E85). The RAF5are calculated relative to the standard exhaust mixture used by theCARD.’3

The format for the data in Figure lOis similar to that inFigures4-6. The MIR. MOW, EBIR, and Base (AR) RAPs are theaverages of theRAFs for the individual adjusted NOx or base casescenarios, calculated from ozone yield, IntO3, and IntO3,’90incremental reactivities. The least squares error (LSE) RAFsshown are those which give the least squares error in ozone yield,integrated 03, or integrated 03>90 ppb in the null tests where thealternative fuel exhaust is substituted for the standard exhaust.This is analogous to the Base (L2) method except that the reactiv-ity ofthe standard exhaust is used in place ofthat for the base ROC.The points under “Vary MIR” and “Vary MOffi” show the effectsof varying the various scenario conditions on the MW or MOIRRAFs, as discussed in the previous section.

For all four exhausts shown, the RAPs tend to increase withdecreasing NO, conditions, except that the integrated 03 RAPs forE85 are almost independent of NO,. The M85 RAF is leastsensitive to the scale used, but is somewhat unique in that theintegrated ozone and ozone yield RAPs have about the samedependence on NO1. The RAPs for the other mixtures are moretypical of relative reactivities in general, in that the integrated 03RAPs are less sensitive to NO, than ozone yield values. The factthat the MIR scale predicts the lowest RAP in all cases (exceptE85)may suggest to some that the MIR has a bias towards givingundue credits to alternative fuels. This increase in RAP withdecreasing NO, is due to the fact that most alternative exhaustshave more slowly reacting compounds whose reactivities areaffected by the lower radical levels in MIR scenarios, and by thefact that the alternative exhausts tend to have less species withstrong NOx sinks (e.g.. aromatics) than the standard exhausts.However, regardless of this variability, the range of RAPs do notoverlap unity except for E85. which overlaps unity only in someextremely NO,-lin’rited scenarios.

The points on the right hand side of the plots show that thevariations in the base ROC mixture, the removal of initial HONO,and the increase in aloft ROCs will not significantly iffect theRAFs in these cases. Thus the main issue in affecting RA$s doesnot relate to these uncertainties but to what type of scale is mostappropriate.

Calculating reactivity adjustment factors forexhausts is not theonly regulatory application where a reactivity scale might beuseful. However, this is the only regulation where such ascale hasbeen applied to date. l’he results of this study should aid inassessing the appropriateness of reactivity scales in other regula’tory contexts.

Discussion Of Issues and Research NeedsA quantitative reactivity scale which compares the effects of

different types of VOCs on Ozone formation could be useful for anumber of ozone control strategy applications. However, thedevelopment of such a scale has a number of difficulties. Thesecan be categorized into three major areas. The first is that the gas-phase chemical mep1ianisms by which VOCs react in the atmo-sphere to form ozone are in many cases highly uncertain. Thisresults in uncertainties in the model predictions of the reactivity ofa VOC in any given scenario, The second is that the effects ofVOCs on ozone formation — their reactivities — depend on theenvironment inwhich they are emitted. This means that even if weare capable of reliably predicting the reactivity of a set of VOCsin a set of scenarios, it is not obvious how these results should beused in developing a single reactivity scale — or even whether ause of a single reactivity scale has any validity. The third is thatthere are uncertainties in conditions of airsheds and episodeswhere unacceptable levels of ozone are formed,The uncertaintiesin conditions of a specific episode affect predictions of VOCreactivities for that episode, and uncertainties in distribution ofconditions affect the development of appropriate methods foraggregating scenario-specific reactivities into a generali2ed reac-tivity scale. -

The focus of this paper has been on the second of theseproblems, that ofderiving a reactivity scale given that reactivitiesdepend on environmental conditions. This has been studied byderiving reactivity scales using several different techniques, givena single chemical mechanism and a single set of representativeairshed scenarios. The chemical mechanism employed is uncer-tain for many VOCs. but it incorporates our current best estimateof their atmospheric reactions, and represents most of the majortypes ofspecies which need to be incorporated in reactivity scales.The representative environmental scenarios employed are evenmore uncertain,but they represent their developers’ best estimateof the conditions of a wide variety of representative pollutionepisodes, given the limitations in available data and the con-straints of the simplified physical formulation of the model used.This is sufficient, at least for evaluating methods, for derivingreactivity scales.

Consistent with results of previous studies, it was found thatthe NO, conditions can significantly affect relative as well as

-absolute reactivities. in addition, it was also found that relativereactivities can depend on how ozone impacts are quantified,especially under low NO, conditions. Because of this, differentreactivity scales give different reactivity rankings for VOCs andin a few cases different orderings of VOCs in these rankings.However, in mostcases thequalitative rankings among the differ-ent scales are very similar, and the quantitative differences be-tween them are small compared to the full range of reactivities ofthose VOCs which are now regulated as ozone precursors. Theresults ofthis study do not support theconclusion that reactividesare so strongly dependent on scenario conditions that all VOCscan be considered equal within this variability. Therefore, use of1~

AIR & WASTE • Vol.44 • July 1994’aal

- TECHNICAL PAPET. - ;- - -

some appropriate type of scale will yield more a more efficientozone control strategy than regulating all VOCs equally. Themore difficult issue is what is the optimum type of scale to use forthis purpose.

Although a total of 18 different scales based on various NO,conditions and methods for quantifying ozone were derived,essentially the choice is between scales which are sensitive toeffects of VOCs on how rapidly ozone is formed and scales whicharesensitive to effects ofVOCs on themaximum amount ofozonewhich is formed when ozone is NO,-limited. Scales in the firstcategory are the MW and the various integrated ozone scales.Scales in the second category are the MOW. EBIR. and theaverage ratio base case ozone yield scales. (Least squares errorbasecase ozone yieldscales are in the first category ifthe base casescenarios represent a varied set of NO, conditions, but are inter-mediate between the two if they represent only near-worst-caseconditions.) Although there are arguments for each type of scale,it is concluded that if only one scale can be used, a scale like MWis more appropriate)0 While the MOW and scales like it are mosteffective at addressing peak ozone levels under conditions whichare the most favorable for ozone formation, the scales like MW itare more optimal when applied to the wide variety of conditionswhere ozone is sensitive to VOCs, or when one is concerned withreducing exposure to integrated ozone or ozone over the airquality standard.

Although these conclusions are based on reactivities calcu-lated for highly simplified single-day scenarios whoseaccuraciesare unknown, the scenarios employed are sufficientlyvaried so itis not unreasonable to expect that similar results would be ob-tained if more detailed and accurate scenarios were employed,This is supported by the results of Russell and co-workers.~°.~’who calculated integrated ozone reactivities using a much morecomplex physical model,~147and obtained results which corre-sponded very closely to the MIR and integrated ozone scalescalculated in this work. The calculations of Russell and co-workers30-4’ also increased the level of confidence in the validityof the reactivity scale derivation because they showed that adetailed physical modej4-33 with condensed chemistry” can giveessentially the same reactivity scale as a simplified physicalmodel with detailed chemistry. However, further work is neededto develop and utilize a more comprehensive and physicallyrealistic set of scenarios for VOC reactivity assessment. All thescenarios used in this work represented the reactions ofthe VOCsonly overasingle day, and scenarios involving multi-day episodesand regional models are needed to assess the total impact of VOC5on ozone over their lifetimes. The work of Russell and co-workers30.4’ is an important start in this regard, but these resultsneed to be further evaluated using physically detailed models ofother areas, and using regional models which can assess theimpacts of VOCs over longer time periods and in long rangetransport scenarios. -

Regardless of which approach or set of airshed conditions isused fordeveloping a reactivity scale. model calculations ofVOCreactivities are no more reliable than the chemical mechanismused to calculate them. Modeling studies may give us an indica-tion of the magnitudes ofthe effects ofthese uncertainties, butwillnot reduce them. To reduce these uncertainties. expenmental dataare needed to test the mechanisms used to derive the reactivityfactors, or at a minimum to test their predictions of maximumreactivity. Such experiments are underway in our laboratories. 0

r,onc!usjons -

Practical implementation of ozone control strategies whichtake into account differences among VOCsin their effects onozone require use of some quantitative reactivity ranking scheme.Use of incremental reactivity, or more particularly ratios of

incremental reactivities or relative reactivities, is an appropriatemeans to do this. However, relative reactivities can vary depend-ing on environmental conditions, how ozone impacts are quanti-fied, and on what approaches are used to derive single scales fromreactivities under a variety of conditions. Although these varia-tions can be significant, in most cases they are smaller than theranges of reactivity among non-exempt VOCs. Thus, despite thevariabilities, use of any appropriate reactivity ranking schemewould yield a more efficient ozone control strategy than ignoringreactivity altogether.

The availability of NO, in the environment is the most impor-mat single factor affecting reactivity rankings. This is oftenmeasured by the ROC/NO, ratio, though this is not always a goodpredictor of reactivity characteristics because of variability offactors affecting rates of NO, removal. The ratio of NO, to NO,levels giving maximum ozone concentrations is a better measureof this. Variations in the compositionofthe base ROG mixture, theamount of initial HONO, and level of aloft VOCs are relativelyunimportant in affecting reactivity when compared to the NO,effect. However, the effect of NO, is less when ozone impacts arequantified by integrated ozone concentrations, or by integratedozone above air quality standards, than is the case when ozone isquantified by peak ozone concentrations or ozone yields. Underhigh NO, conditions where VOCs have their greatest effect onozone, which is the basis for deriving the Maximum IncrementalReactivity (MI?,) scale, the relative reactivities are not stronglyaffected by how ozone is quantified, and are also relativelyinsensitive to other scenario conditions. Under lower NO, condi-tions, relative reactivities tend to become more sensitive to otherscenario conditions, and tend~odiffer depending on how ozone isquantified.

Thus the MIR scale is relatively well defined in the sensethat it is fairly insensitive to the choices of scenarios used toderive it. In most cases, it gives a reasonably good approxima-tion to scales based on integrated ozone under lower NO,conditions. The MOIR scale gives better predictions of effectsof VOCs on peak ozone yields in base case scenarios, but gavepoor predictions of effects on integrated ozone or integratedozone above the standard. It is also more sensitive to the set ofscenarios used to derive it. Based on these considerations, andthe fact that the MIR scale is based on environmental condi-tions where VOC control is mostimportant foraffecting ozone,we conclude that the MI?, scale (ora scale similar to it, such asone based on integrated ozone over the standard) is appropriatefor regulatory applications where a reactivity scale is required.Airshed model calculations using a much more detailed physi-cal scenario (but a simpler chemical mechanism) lead to simi-lar conclusions.30.~’

AcknowledgmentsThe author acknowledges Mr. Bait Croes of the Research

Division ofthe California Air Resources Board (CARB) forhiscontributions to this work regarding the establishment of thescenario conditions and for many helpful discussions. Themembers of the CARB’s Reactivity Advisory Panel also con-tributed to the establishment of the scenario conditions andprovided helpful input. The author thanks Mr. Keith Bauguesfor supplying the EKMA scenarios used in this study; Dr.HarveyJeffries forsupplying the actinic flux data used in thesecalculations and other useful data files and information; Dr.Michael Gery for helpful discussions; Dr. Roger Atkinson forhelpful discussions and for reviewing this manuscript; and Mr.Dennis Fitz for reviewing this manuscript. Finally, the authoracknowledges Mr. Alvin Lowi for initiating the project whichultimately led to the adoption of an incremental reactivity scaleas a means to assess ozone impacts of vehicle emissions.

U

898 • July 1994 • Vol. 44 • AIR & WASTE

This work was supported by the California Air ResourcesBoard through Contract No. A934-094. and incorporates results ofwork carried previously out under EPA Cooperative AgreementCR814396-UI-U. The statements and conclusions in this reportare entirely those of the author and not necessarily those of theCalifornia Air Resources Board, the United States EPA, or any ofthe individuals listed above.

ReferencesI. L Atkinson.‘Gas-phase tropospheric chemistryof organic compounds:

review.’ A:nws. Environ. 24A: (990).2- w. P. L. carter. L Atkinson. “An experimental study of incremental

hydrocarbon reacttvity.’ Environ. Set Technol. 21.670t 9871.3. T. Y. Chang. SI. Rudy. ‘Ozone-footing potential of organic emissions

ftotn altcrnasivc.fucled ‘chides.’ A:n,os, Environ 24A.2421 (19901,4. 1.1. Bufalini. M. C. Dodge. ‘Ozone-forming potential of light saturated

hydrocarbooC Environ. Sc:. Technol. 17:308 (1983),5, st c Dodge. tombined effects of organic reactivity and NMHCJNOs

ratio on photochcmical oxidant formation — a modeling study? Asians.

Environ. 13:1657(1984).6. A. SI. Rough. L 0. Derwent. ‘Computer modelling studies ofthe disusbution of

pbotocheenical ozone production between different hydrocarbons.’ Aonos.Environ, 21:2015(1987).

7. W. P. L. Caner. Estimates of photochemical reactivities of solvent speciesused in architectural coatings.” presented at the 82nd Annual Meeting of the Airand waste Management Association, Anaheim. California. June 25-30, 1989.

8. 3. R. weir. A. 5. Rosenbaum, L A- Gardner. C. Z. Whinen. w. caner‘Architectural Coatings in the south Coast Air Basin: Survey. Reactivity, and

• Toxicity Evaluation,’ Final Report to the South Coast Management District.SYSAPP-88/137. Systems Applications. Inc.. San Rafael. CA. December 1988.

9. W. P. L- Caner. R. Atkinson. ‘A computer modeling study of incrementalhydrocarbon reactivity.’ Environ. Set Technol. 23:864 ((989).

IC. W. P. L Carter. 1. A. Pierce.!. L. Malkina. D. Lao. W. 0. Long. ‘EnvironmentalChamber Studies of Maximum Incremental Reacti’.itiea of Volatile OrganicCompounds,’ Report on Coordinating Research council Project ME-9 andCalifornia Air Resources Board contract No. A032.0692. April I. 1993.

II. W. P. L- Caner. “Development of Ozone Reactivity Scales for volatile OrganicCompounds.’ EPA-600/3.911050. August 1991.

It CARS. ‘Proposed Regulations for Low-Emission vehicles and Clean Fuels —

StaffReport aisdTechnicalSupport Documenl.’CaltfomiaAirResourcesBoard. -

Sacramento,CA.August 3,1990.5cc alsoAppendix vIliof’CaliforniaExhaustEmission Standards andTest Procedures for 1988 and Subsequent Model Passen.ger Can. Light Duty Trucks and MeØiuni Duty vehicles: as last amendedseptember 22. 1993. Incorporated by reference io Section 1960. (k) ofTitie 13.California Code of Regulations.

13. CARS. “Proposed Reactivity Adjustment Factors for Transitional Low-Emis-sions Vehicles —Staff Report andTechnical Support Document.’California AirResources Board. Sacramento. CA. September 27. 991.

14. 0. L Cipson. w. P. Freas. R. A. Kelly, S. L. Meyer. ‘Guideline for Usc ofCity- -

Specific EICMA in Preparing Ozone Sips. EPA4SO/4-80-027. March 1981.IS. 6. L. Gipson. W. P. Freas. ‘Use of City-Specific EKMA in the Ozone RIA.’ U.

S. EPA. July 1983.16. 1)3. EPA. ‘Guideline for Using the Carbon Bond Mechanism in City-Specific

EK,MA? EPA-450/4.84-005. February 1984.17. M. W. Gezy. R. 0. Edxnond. C. z. V/bitten. 1’ropospheric Ultraviolet Radiation,

Assessment of Existing Data and Effects on Ozone Formation.’ Final Report -

EPA’600/3-87-047. October 1987.18. K. Baugues, ‘Preliminary Planning Information for Updating the Ozone Reg-uls’

tory Impact Analysis Version ofEKMA.’ Draft Document. Source ReceptorAnalysis Branch. Technical Support Division. U’S, EPA. Research Triangle

Past. NC. January 1990.19, H- Hogo, M. w. Gcry.’Cuidelines for UsingOZiPM-4 withCBM-tVorOptional

Mechanisms. Volume 1. Description ofthe Ozone Isopleth Plotting PackageVersion 4,’ Final Report for EPA Contract No. 68-024136. AttnosphtricSci-ences Research Laboratory. ResearchTriangle Part. MC January 1988. Sec also:C. L. Cipson. ‘Users Manual for OZIPM-2: Ozone Isopleth Plotting PackageWith Optional Mechanism/Version 2,’ EPA-450/4-8-4.024,August 1984.

20. W. P.L. Caner. ‘Development and Application ofaa Up-To.Date PhotochemicalMechanism for Airshed Modeling and Reactivity Assessment.” Final report forCalifornia Air Resources Board Contract No. A93-4-’094, Research Division.Sacramento. CA, April 1993.

21. CARS. ‘Proposed Reactivity Adjustment Factors for Transitional Low-Esnis- -

sions Vehicles — Staff Report and Technical Support Document. IS DaysNotice? California Air Resources Board, Sacramento. CA, April 21. 1992.

22. H. B. Jeffries. Department ofEnvironmental Science and Engineering, School ofPublic Health. University of North Carolina. Chapel Hill. NC. private communi’cation, 1991.

23. H.E. Jeffries, K. C. Sexton. J. R. Arnold, T. L. Kale.’validation Testing ofNewMechanisms with Oistdoor Chamber Data. Volume 2; Analysis of VOC Data forthe CS-i and CAL PhotocheotiesiMechanisms? Final Report EPA-600/3.89’OlOb. 1939.

24. B. B. Croes, Technical Support Division, California Air Resource Board, privatecommunication. 1991.

25. B. E. Croca.. s-rat. ‘Southern California Air Quality Study Data Archive,’Research Divtason. California Air Resources Board. 994.

26. F. w. Lunnann. H. H. Main. ‘Analysts ofthe Ambient voc Data Collected in theSouthern CaliforniaAir QualityStudy.’ Final Reporsto California Air ResourcesBoard Contract No. A832. 30. Febnsary 1992.

27. M. CDodge. ‘Combined Use ofModeling Techniquca and SmogChamber Datato Denve Ozone. Precursor RelationshipC in Proceedings. Intemational Confercnce on Ptsocoehcmical Oxidant and its Control. Research Triangle Park, NorthCarolina, EPA-6C4’3.80-023a. t977.

2t A. L Lowi. W, P. L. Carter. ‘A Method for Evaluating the Atmospheric OzoneImpact of Actual vehicle Emissions,’ presented at the SAS InternationalCongress and Exposition. Detroit. Michigan. February 26- March 2. 1990.

29. w. L. Chameidea. F. Fthaenfeld. M. 0. Rodgers. C. Cardelina 1. Maninez. 0.Pamah. W. Lonneman. D- L Lawson, R A. Rasmussen. P. Zimmerman. I.Creenburg. P. Middleton T. wang. ‘Ozone precursor relationships inthe ambientatmosphere.” J~Groph~s. Rca. 91:6037(1992).

30. L McNair. A. Russell, M. t Odmait. ‘Airstied calculations o(the sensitivity ofpollutant fornsassoo to organic compound classesand oxygenates associated withalternative fueLs?). Air Wo.z:eManage. Assoc..42:174(1992).

3!. R. C. Derwent. M. B. Jenkins. Hydrocarbons and the long range transport ofozone and PAN over Europe? Agmos Environ. 25A:1661 (1991).

32. W• P.1. Carter, ‘A dctailcd mechanism for thegas-phase aunosphaie reactionsof organic compounds.’Armor. Environ. 2’IA:48l (1990).

33. W. P. 1. Carter. F. w. Lurrnann ‘Evaluation of a detailed gas.phase auno’tpheric reaction mechanism using environmentalchamberdata.’Aintos. Environ.

25A:277l (1991).34. w, R. Stockwelt. P. Middleton.1.5. Clung. X. Tang ‘The second generation

regional acid deposition model chemical mechanism for regional air qualitymodeling.”). Geophvs. Re:. 9S:l6343 (1990).

35. W. P. L. Caner. F. W. Lurmann. ‘Evaluation of the RADM Gas-Phase ChemicalMechanism.’ Final Report EPA.600i3’go-OoI, 1990.

36. M. w. Cery. 0.2. Whitten. 1. P. Killus. “Development andTesting ofthe CSM-IV For Urban and Regional Modeling.’ EPA-600/ 3-88-012, January 1988.

37. A. M. Hough. ‘As intcreomparisoo of mechanism for the production ofphotochemical oxidants.’ J. Geopkvs. Re:. 93:3789 (1988)-

38. C. 2. ‘A/bitten. ‘Evaluation of the Impact of Ethanol Gasoline Blends on UrbanOzone Fonnasiots.’ Syssems Applications. Inc., San Rafael. CA, SYSAPP-88/029. 1988.

39. 0. M. Smylie.G.Lwhinenjt S. Morris. L. F. Feiber. K~O’Connor,’TechnicalAnalysis of the ARB Stair Proposal for Hydrocarbon Reactivity AdjustmentFactors for Clean Fuel/Low Emission vehicle Regulations.’ Final Report toWestern States Petroleum Association. Glendale.CA. Systems Applications Inc.Report SYSAPP-90i092. September 1990.

40. K. Baugues. US. EPA. Office of Air Quality Planning and Standards. ResearchTriangle Part, NC;Letterto BartCroes.CaliforniaAirResoureesSoard, May IS.1991.

41. LA. McNair. AG. Russell. SI. T. Odmaa. B. E. Croes. L. Kao, “Airahed modelevaluation of reactivity adjustment factors calculated with the maximum meremental reactivity scale for transitional low-emissions vehicles.’). Air wasteManage. Assoc.. in press.

42. A. Russell. I. Harris, I. Milford. 0. St. Pierre, ‘Quantitative Estimate of the AirQuality Impacts of Methanol Fuel Use.” Final Report. California Air ResourcesBoard Agreement No. A6’048-32, Carnegie-Mellon University. Pittsburgh. PA,April t989.

43. A. C. Russell. D. St. Pierre.). B. Milford,’Ozone control and methanol fueluse,’Science, 237:201 (1990).

44. F. w. Lurmann. w. P. L. Carter, R. A. Coyner. ‘A Surrogate Species ChemicalReaction Mechanism for Urban-Scale Air Quality Simulation Models. Volume I‘Adaptation ofthe Mechanism,’ EPA.6~/3-87-0l4a,1987.

About the AuthorWilliam P. 1.. Carter is a Research Chemist with a joint

appointment at the Statewide Air Pollution Research Centerand the College of Engineering Center for EnvironmentalResearch and Technology at the University of California,Riverside CA 92521.

AIR & WASTE• VoL 44 • Ju& 1994 .899