Content uploaded by Stefano Caserini
Author content
All content in this area was uploaded by Stefano Caserini on Apr 06, 2015
Content may be subject to copyright.
AtmosphericPollutionResearch4(2013)282‐289
©Author(s)2013.ThisworkisdistributedundertheCreativeCommonsAttribution3.0License.
A
A
tm
spheric
P
P
ollution
R
R
esearch
www.atmospolres.com
Impact of the dropping activity with vehicle age on air pollutant
emissions
StefanoCaserini1,CinziaPastorello1,2,PietroGaifami1,3,LeonidasNtziachristos4
1DIIAR
–
SezioneAmbientale,PolitecnicodiMilano
–
PiazzaL.daVinci,32
–
20133Milano,Ital
y
2Presentaddress:EEA–EuropeanEnvironmentAgency,KongensNytorv6,1050Copenhagen,Denmark
3Presentaddress:RedecamGroup,PiazzaMontanelli,2020099SestoS.Giovanni,Italy
4LaboratoryofAppliedThermodynamics,AristotleUniversity,POBOX458,GR54124Thessaloniki,Greece
ABSTRACT
Roadtransportisamajorsourceofairpollutionespeciallyincities.Detailedcalculationsareneededtosupport
roadtransportemissioninventoriesduetothevarianceoftechnologiesandoperatingconditionsencounteredon
theroads.Theannualdistancedrivenbycarsinrelationtotheircharacteristicsisanimportantvariableinsuch
calculations.Inthiswork,alargeamountofmileagedatawerecollectedfromsecond–handcarsellersinItalyand
werethenanalyzedinordertounderstandtheinfluenceofvehicleageonannualmileagedriven.Theavailable
dataenabledthedevelopmentofdroppingfunctionsofannualmileagewithvehicleage.Itwasfoundthatthe
averagemileageof10yearoldcarsisonlyapproximately40%ofthemileagedrivenonyearone.Thisdropsto
approximatelyonly10%for20–yearoldcars.Thefindingsareofparamountimportanceinenvironmental
calculationsasroadtransportNOXandPMemissionsdropbymorethan20%whenthecorrectedfunctionsare
usedcomparedtousingaconstantmileage.Notintroducingsuchacorrectionmayresulttoanapproximately8%
highernation–wideNOXemissionswithnegativeimplicationstowardsmeetingthenationalemissionceilings.In
termsofpolicyimplications,thedroppingactivitywithageresultstoadecreaseintheimportanceofaccelerated
scrappageschemesandofenvironmentalzonesinairquality.
Keywords:Airquality,emissions,NOX,roadtransport,vehicleactivity
CorrespondingAuthor:
Cinzia Pastorello
:+45‐3336‐7298
:+45‐3336‐7128
:cinzia.pastorello@eea.europa.eu
ArticleHistory:
Received:07February2013
Revised:24April2013
Accepted:07May2013
doi:10.5094/APR.2013.031
1.Introduction
Roadtransportisoneofthemainsourcesofpollutionin
urbanareasinEurope,accountingfor39%,and15%oftotalNOX
andPM2.5emissions,respectively(EEA,2010).Duringthelast
years,benefitsfromtheprogressivereplacementofuncontrolled
gasolinecarswithnewonesequippedwiththreewaycatalystsare
counterbalancedbytheincreasingpenetrationofdieselcarsand
theirhigheremissionlevels,inparticularNOX,comparedtotheir
gasolinecounterparts.Urbansprawlingandthegeneralassociation
ofpersonalmobilitywithqualityoflifeandeconomicdevelopment
(Uhereketal.,2010)haveincreasedthemeanannualdistance
travelledbycars(EC,2012)until2008;thenaslightdecreasehas
beenvisibleduetotheeconomiccrisis.
Emissionmodelsareusedtocalculateemissionsfromroad
transport.ExamplesofsuchmodelsinEuropeincludeHBEFAused
inGerman–speakingcountries(Hausbergeretal.,2009),VERSIT+
usedintheNetherlands(Smitetal.,2007),LIPASTOinFinland
(lipasto.vtt.fi),andCOPERT4whichisusedin22outofthe27
EuropeanUnionmemberstates(Ntziachristosetal.,2009).Insuch
modelsvehicleactivityismultipliedwithemissionfactorsto
calculatetotalemissions.Vehicleactivityisestimatedasthe
numberoffleetvehiclespertypedistinguishedineachmodel
timestheannualmileagedrivenbyeachvehicletype.Vehiclefleet
dataarereadilyavailableinseveralcountries.Forexample,official
statisticsinItaly(ACI,2011)providethenumberofvehicles
distinguishedperfueltypeused,enginecapacity,andemission
controlregulation(EUROcategories).
Withregardtomileageestimates,Andreetal.(1999)pointed
outtheimportanceandthedifficultiesofestimatingannual
mileageanditstrendasafunctionofvariousfactors(suchas
vehicleage)fordifferentvehiclecategories.Onthebasisof
inspectionandmaintenancemonitoringprograms,Beydounand
Guldmann(2006),Washburnetal.(2001),andBin(2003)demon‐
stratedthatthetotalmileageofavehicleisstronglyassociatedto
emissionandtestfailurerates.Finally,Sawyeretal.(2000)
underlinedtheimportanceofhavingaccurateactivitydatafor
obtaininganimprovedtransportemissioninventory.
Despiteitssignificancethough,annualmileageisusually
availableasanaveragevaluefortheentiregasolineordieselcar
fleet,withoutdistinguishingintomoreemission–relevantcriteria,
suchastypeofroadorlengthofservicethataffecttheemission
assessment(Ntziachristosetal.,2008).Often,mileagevaluesused
innationalemissioninventoriesarenotbasedonmeasureddata
butarejustcalibratedvalues,estimatedsoastoachieveabalance
betweenthefuelconsumptioncalculatedbythemodelandofficial
statisticsonfuelsold.Althoughthisproducesaninventorywhichis
consistentwithtotalenergystatistics,itdoesnotguaranteea
correctallocationofconsumptiontothedifferentvehicletypes,
neitherthattotalactivityiscorrectlydisseminatedtothevarious
vehicletypesandages.
Caserini et al. – Atmospheric Pollution Research (APR) 283
Anestimateofthemileageofdifferentvehicletypesasa
functionofageisthereforeanimportantinputtoroadtransport
environmentalmodels,inordertoaccuratelyestimatethetotal
emissionsproduced.Itisalsonecessaryinordertoaccurately
predictthereal–worldimpactofpolicymeasurestargetingspecific
vehicletechnologies.Forexample,implementingpoliciestargeting
theremoval(scrappage)orbanoftravelling(environmentalzones)
ofoldvehicletypeswillbelesseffectivethanplannedincasethat
adecreasingfunctionofmileagewithvehicleageisestablished.
Inthisstudyweareproposingamethodologyforprecise
mileageestimationasafunctionofvehicleagethatcanbe
introducedtoemissionmodelsforimprovingthequalityofthe
calculations.Asanexamplecase,weapplythemethodinthecase
ofItalytodemonstratetheextentbywhichmileagemisallocation
betweendifferentvehicleclassesmayaffectemissionsestimates.
Thiscanalsoserveasameasureoftheuncertaintyofaninventory,
whenmileagevaluesarenotbasedonmeasureddata.
2.Material
Real–worlddataonpassengercarmileagewascollectedby
visiting32950individualsecond–handpassengercarsalesonthe
internet(seetheSupportingMaterial,SM,TableS1).Thecollection
andanalysisofthedatatookplaceintheperiodbetweenJuneand
September2010.Thefinaldatasetconsistedof18652gasoline
carsregisteredintheperiodfrom1994to2010and14298diesel
cars,registeredbetween1996and2010.Oldervehicleswere
practicallynotforsaleastheyarenotallowedtocirculatein
severalItalianregionsduringthewinter,duetoair–quality
limitations.Thedatasetconstructedincludedfuelused,yearof
firstregistrationandodometerreading.TableS2(seetheSM)
showssomekeystatisticsofthedataset.
Datacollectedrepresentabout1.1%ofthe2.8–3.0millionof
usedcarssoldinItalyeveryyear.Thiswasasmuchascouldbe
collectedfromthesecondhandmarketwithcompleteinformation
onageandmileage.Noparticularcriteriaweresettoselectthe
vehiclesinthesample.Hence,weexpectthemtomakea
representativesampleoftheactualvehiclestock.Thisisalso
verifiedbythefactthatthedistributionofvehiclesinthedifferent
ageclassesisquitesimilartothedistributionofthecompletestock
ofvehiclesregisteredinItaly.ThisisdemonstratedinTable1
whichcomparestheagedistributionofthesampleusedinthe
currentstudywiththeItalianstockdistribution,basedondata
takenfromtheItaliancarassociationdatabase(ACI,2011).The
frequencydistributionsarequitesimilarbetweenthetwodatasets.
Onemightexpectthattheratioofcarsbeingsoldvs.carsin
operationwouldincreaseastheageincreases.Thiswaspartly
visibleinthecaseofdieselbutnotforgasolinecars.Several
reasonscouldcontributetothesedifferenttrends,suchas
generallylongerlifetimefordieselcars,scrappageandexportingof
vehiclesvs.secondhandsalewithinthecountry,etc.Anyway,the
differencesintheratiobetweensoldandregisteredcarinevery
yeararelow(from0.08%to0.2%fordieselandfrom0.09%to
0.13%forgasoline)andthesamplesizeperageclassis
satisfactory,withtheminimumnumberperclassstillabove400
vehicles(diesel,14yearsold).
Questionsmayariseonthereliabilityoftheodometerreading
asamethodtoinferthemileagedrivenbyeachvehicle.Thisis
onlyreportedbytheownerandsuspicionsmayariseregardingthe
effectofclockingofthecars,i.e.thefraudulentwindingbackof
theodometerreadingtomakethecarappearyoungerthanit
reallyisandnegotiateabetterpricewiththepotentialbuyer.A
recentstudyontheimpactofmileagefraudwithusedcars(EREG,
2010)identifiedmileagefraudasaseriousproblem,butnodata
weremadeavailabletodefinehowmuchthisissuecouldaffectthe
variationofaveragemileagewithtime.Inthiswork,wehave
assumedthatthedistortioneffectisproportionaltothelengthof
service.Thatis,wehaveassumedthatcarclockingtakesplaceat
thesamefrequencyfornewandoldcarsandthatthemileage
correctionisproportionaltotheodometerreading.Inotherwords,
theassumptionisthattheintentiontoimprovethesellingpriceof
thevehicleisthesame,regardlessoftheactualageofthecar,
whichtendstobeareasonableapproach.Hence,thiswillhavean
impactonthetotalmileagereportedbutwillnotaffecttherelative
relationbetweenmileageandage,whichisofimportancetothis
study.Alldatacollectedwerepooledtogetherandastatistical
analysiswasconducted.Theresultsofthisstatisticalanalysisand
themethodsusedarepresentedinthefollowingsection.
Table1.FrequencydistributionofvehiclesintheavailabledatasetandofvehiclesregisteredinItaly,accordingtotheirage
GasolineDiesel
VehicleageDataavailableVehiclesregisteredinItaly
VehicleageDataavailableVehiclesregisteredinItaly
Number(%)Number(%)Number(%)Number(%)
111426.1%10692126.1%111708.2%9378197.3%
212606.8%12681727.2%211848.3%9526487.4%
311236.0%10803426.2%311438.0%11182398.7%
411486.2%11158226.4%411848.3%142517611.1%
511836.3%9825595.6%511037.7%137571910.7%
611035.9%9280705.3%611077.7%130943310.2%
711756.3%9358305.3%710897.6%130626010.1%
811886.4%11284976.4%810697.5%10584848.2%
911706.3%12469257.1%911077.7%8996567.0%
1011916.4%14038958.0%1010847.6%7566015.9%
1111776.3%13799627.9%1110627.4%6606565.1%
1211836.3%12111306.9%129916.9%4943713.8%
1313077.0%12220497.0%135954.2%3461322.7%
1412916.9%11988616.8%144122.9%2428331.9%
1511506.2%7059164.0%
168614.6%6662463.8%
Total18652100%17543488100%Total14300100%12884027100%
16<age<305298935 14<age<301016079
Caserini et al. – Atmospheric Pollution Research (APR) 284
3.CalculationandResults
3.1.Cumulativeandannualmileages
Theaveragecumulativemileage(ACMk)referstothetotal
distancecoveredonaveragebyeachcarofthesameage,orina
moretechnicalway,withthesamelengthofservice(k–inyears).
TherelationshipbetweenACMkandvehicleageisshownin
Figure1asanaverageforallgasolineanddieselpassengercarsin
oursample.ThefigureshowsthatACMincreaseswithvehicleage
almostinalinearfashionupto4–5yearsofage.Beyondthispoint,
therateofincreaseinACMdrops,denotingadecreaseofthe
annualmileageconducted.After14yearsandapproximately
126thousandkilometersforgasolinevehiclesand13yearsand
approximately180thousandkilometersfordieselcars,thereis
onlylimited(ifany)increaseinthecumulativemileage.
Whenlookingatsinglevehiclesonly,totalmileagecanonly
monotonicallyincreasewithvehicleage.Hence,thestabilizationof
mileageafteracertainagecannotbeexplainedonasinglevehicle
basis.ThereasonformileagestabilizationinFigure1isthat
vehicleswithexcessivemileageareremovedearlierfromthe
stock,eveniftheirage–measuredinyears–doesnotjustifythis.
Hence,asthefrequencyofvehiclesbeingscrappedincreaseswith
age,therateofincreaseofmileagewithagegraduallydropsata
fleet–widelevelandasaturationpointisreached.Afterthispoint,
increasingthemeanvehicleagedoesnotcauseanysignificant
increaseinthemileage.Actually,adropinthemileagewouldalso
betheoreticallypossible.
Themileagestabilizationisofimportancetoroadtransport
emissionmodels.Forexample,asaninputinfunctionswhich
correctemissionfactorsaccordingtothemileagecovered.Figure1
showsthat,despitetogeneralbelief,theactualaveragemileageof
afleetofpassengercarsdoesnotincreasebeyondacertainpoint
astheygrowolder,henceemissionfactorsshouldnotdegrade
aboveacertainlevel.
BydividingtheACMvaluewiththelengthofservice,one
obtainstheaverageannualmileage(AAMk)whichistheaverage
annualdistancecoveredbyeachcarofthesamelengthofservice.
Thatis,
(1)
AAMonlydependsonvehicleageanditisthesameforeach
yearduringthelifetimeofthevehicle.AAMkvaluesasafunctionof
vehicleageareshowninFigure2.Thisshowsthattheaverage
mileagedrivenperyeargenerallydropswhilethevehiclesbecome
older.Thefunctiondoesnotdropmonotonicallyfordieselcarsbut
carsofoneyearofageappeartobedrivenlessthancarsoftwo
andthreeyearsofage.Wehavenoevidencethatthisisasample
artifact,howeverasatisfactorynumberofmorethan1000diesel
vehicleswasavailableforeachageclassinourdataset.
Ifvehicleageisnottakenintoaccount,thentheaverage
mileageofourgasolinecarsampleis10636kmand18685kmfor
dieselcars.However,asshowninFigure2,thetrueaveragewill
dependontheaverageageofthevehiclesconsidered.Thisisnot
alwaystakenintoaccountinrelevantstudies.
3.2.Mileagetobeusedforemissionmodeling
Justbecausecarsareuseddifferentlythroughouttheir
lifetime,themaximumlengthofservice,j,i.e.theageinyearsat
whichthevehicleisremovedfromthestock,mayvaryfrom
vehicletovehicle.Themaximumlengthofservicewillhavean
impactontheincreaserateofACMwithage.Thatis,vehicles
whicharescrappedfromthestockearly(shortlengthofservice)
shouldbeexpectedtoaccumulatemileagefasterthanvehicles
withalongermaximumlengthofservice.Therefore,themaximum
lengthofserviceisaparameterthathastobetakenintoaccount
whenexpressingthefunctionofmileagewithvehicleage.
Figure1.
A
veragecumulativemileage(ACM)fordieselandgasolinepassengercars,asafunctionoftheirage.Fittedparabolas
passingthoughtheaxesoriginhavebeendrawnforbothdatasets.
Caserini et al. – Atmospheric Pollution Research (APR) 285
Figure2.
A
verageannualmileage(AAM)drivenbyvehiclesofdifferentage.
Inordertointroducethemaximumlengthofservice(j)inthe
calculation,onemaystartbyobservingthatthemileagevaluesin
Figure1seemtofollowaparabolicfunctionwithage.Hence,ACM
canbeapproachedbyabinomialfunctionpassingthroughtheaxes
origin,i.e.afunctionofthetypeshowninEquation(2):
(2)
Thebinomialcurvesthatbestfitthedataresulttotheaandb
parametersareshowninFigure1forbothdieselandgasoline
passengercars.Thisequationcanbeconsideredrepresentativefor
astockofvehicleswithanaverageend–of–lifeageatthepoint
wherethebinomialequationbecomeslevel.Thisismathematically
expressedbythefunction:
0
kj
dACM
dk
(3)
Forsuchacurve,parametersaandbcanbeeasilycalculated
byEquations(2)and(3)as:
(4)
2(5)
UsingEquations(4)and(5)andthebestfitparametersfor
dieselandgasolinecars,thismethodresultstojvaluesof16(years
ofservice)forgasolinecarsand14yearsofservicefordieselcars,
withcorrespondingACMvaluesof126300kmand172600km,
respectively.Thesevaluesareveryclosetheonesestimatedinthe
previoussectionwhichmeansthatthebinomialfunctionverywell
describestheevolutionoftheACMasafunctionofvehicleage.
Havingestablishedaparabolicdevelopmentofmileagewith
age,withparametersdefinedinEquation(4)and(5),thiscanbe
usedtoestimatetheevolutionofthemileageofvehiclesthathave
amaximumusefullifemorethanwhatisshowninFigure1(see
alsotheSM,FiguresS1andS2,forgasolineanddieselcars,
respectively).Themaximumaveragemileageremainsconstant,as
thiswastheevidencefromtheexperimentaldatainFigure1;
hence,theevolutionofmileagealongthelifeofthevehiclecanbe
predictedhavingtheend–of–lifeageofthevehicleastheonly
independentparameter.Allgasolinecarswhichareyoungerthan
16yearsoldanddieselcarswhichareyoungerthan14yearsold
areexpectedonaveragetofollowtheoriginalcurve.
Theparabolicfunctionsdefinedcaninturnbeusedtomodel
howtheannualmileageofcarsdropswiththeirage.Thiswasnot
possiblebysubtractingtheACMvaluesoftwoconsecutiveyearsas
thisstabilizedafteracertainageandwouldresulttozero,oreven
negativevalues.Withthemodeldeveloped,ifonespecifiesthe
end–oflifeageofvehicles(j),thentheactualaverageannual
mileage(AAAM)thatthesevehiclesconductedwhenatagekwill
be:
,
1
1
(6)
Theparametersaandbinthisfunctiondifferforgasolineand
dieselpassengercars.ThemodeldevelopedinEquation(6)canbe
appliedtopredicttheevolutionofmileageoftheItalianpassenger
carstock.Thedistributionofvehiclesaccordingtoyearof
registrationisavailableatnationallevelintheItaliancar
associationdatabase(ACI,2011).Thisdatabaseshowsthatthere
areveryfewvehiclesregisteredabove30yearsofage,whichis
consideredasthemaximumendoflifeageinouranalysis.The
probabilityofvehiclestoreachacertainendoflifeageisrequired
inordertoapplyEquation(6)ontheItalianstockdata.However,
thisisnotknowna–priori.Wecanassumethatthisprobabilityis
equaltothepercentageofvehiclesintheparticularagebin
registeredinItalyin2010.Thisapproximationisactuallyalso
theoreticallyaccurateifthenumberofvehiclesregisteredinItalyis
constantthroughouttheyears.Inreality,intheperiod2000–2011,
theItaliannewpassengercarregistrationshavebeenfallingwith
anaveragerateof2%.Thisisaverymildchangewhichmeansthat
ourapproximationisveryclosetothetheoreticalaccurateandcan
besafelyusedinourcalculations.
Withthisassumption,andthefactthatallvehiclesbelowa
certainage(J*=16forgasolinecarsandJ*=14fordieselcars)
followthesameACMcurve,theAAAMvalueofallvehicles
registeredinItalyasafunctionoftheiragecanbecalculatedby
meansof:
Caserini et al. – Atmospheric Pollution Research (APR) 286
30
*Jj j
30
*Jj jj,k
kf
fAAAM
AAAM (7)
wherefjistheprobabilityofcarstoreachanendoflifeagej.
ThegraphicalrepresentationofEquation(7)isshowninFigure
3.TheAAAMkvaluesdropwithvehicleageandpracticallyreach
zeroatanageof30years.Regressioncurvessplitindifferent
regionshavebeendrawnthatallowusingthesetrendsindifferent
applications.Aquadraticfithasbeenassumeduntilthe22ndyear
ofageandthenalinearfittotheminimumofmileageinthe30th
year.Thishasbeenselectedinordertomaximizethefittingofthe
curveswiththedata.Severalattemptshavebeendoneandthe
split<22and>22yearofageresultedasthebestone.
3.3.Comparisontootherstudies
Despiteitssignificanceforemissioncalculations,information
onvehiclemileageasafunctionofageisratherscarceinthe
literature.Fewdataareavailable,mostlyestimatedonthe
assumptionthatvehiclesaredrivenforthesameannualdistance
duringtheirwholelives.Asalreadysaid,thisassumptiondoesnot
correctlyrepresenttheknowndependenceofmileageonvehicle
agewhichisdescribedbyAAAM.Ontheotherhand,nottaking
intoaccountthedropofannualmileagewithagelikelyresultsto
anoverestimationoftotalemissionsasthecontributionofolder
vehicletechnologiesisoverstated.
AdatasetofmileagedatahasbeendevelopedatEuropean
levelintheframeworkoftheTREMOVE(EC,2005)andMEET
(Andre,1999;Andreetal.,1999)activities.TREMOVEisapolicy
assessmenttoolthatprovidesthebackgroundcalculationsfor
impactassessmentstudiesintheareaoftransportpolicy
interventions.IntheMEETproject,thedependenceofmileageon
vehicleagewasalsocollectedfromafewnationaldata,without
offeringdistinctiontovehiclecategories(Andreetal.,1999).For
example,forSwedenthedependenceofmileageonvehicleage
(Figure4)wascalculatedusingdatasetsfromtwoconsecutive
years(1987–1988)andmatchingcarwiththeregistrationnumber.
Althoughnoinformationisavailableontheamountofdata
processed,inthiscasetheresultistheactualannualmileage
definedpreviously.
Vehiclesurvivabilityandmileageforpassengercarswerealso
developedonthebasisof1977to2002registrationsand2001
mileagesurveydata(NHTSA,2006)inUS.Thisanalysisshowsthat
atypicalpassengercartravelsforatotaldistanceofapproximately
150000miles,reachedafter25years,whileinthepresentstudy
thelifetimemileageforgasolineanddieselvehicleswasrespect‐
tively126300km(after14years)and172600km(after16years).
AsitisshowninFigure4,thedecreaseofAAAMmileageinthe
firstyearsofserviceishigherthanfortheothercurves.
Thecomparisonofthedatageneratedinthisworkwith
resultsfromotherstudiesgenerallyshowsthattheAAMvalues
generatedareclosetothemaximumoftherangeofthedata
collected.Actually,therelativeincreaseofdieselcarsupto3years
ofageisuniqueinourdataset.Ontheotherhand,theAAAM
values–whichshouldcloserreflecttheactualdropinmileagewith
age–areatthelowendoftherangecollected,andcomparableto
datafromUS.Ithasnotbeenpossibletoexactlyidentifywhich
methodhasbeenusedbyotherstudiesforestimatingthemileage
functionwithage.However,thevaluesobtainedinthisworkwith
twodifferentmethodsaregenerallywithintherangeofvalues
reportedelsewhere.
Thiscomparisonshowsthatthedefinitionofthemileage
functionwithageisimportantandthatdifferentdroppingmileage
ratescanbeobtained,dependingonthedefinition.Thisleadsto
twoimportantconclusions.First,themethodusedtoestimatethe
functionofmileagewithagehastobereportedand,second,the
environmentallyrelevantcalculationswilldependonwhich
methodhasbeenusedfortheassessment.
Figure3.
A
ctualaverageannualmileage(AAAM)bydieselandgasolinecarsintheItalianvehiclestock.Bes
t
‐fitcurves
havebeendrawn,splitintworegions.
Caserini et al. – Atmospheric Pollution Research (APR) 287
4.Discussion
Therelationshipbetweenmileageandvehicleageallowsthe
estimationoftheaveragemileageofvehiclesthatbelongto
specificlegislative(i.e.Euro)classes.Thisisthestandardvehicle
classificationinallEuropeanroadtransportemissionmodels(e.g.
COPERT,HBEFA,VERSIT+).Multiplicationofthismeanmileagewith
thenumberofvehiclesintheclassandwithanappropriate
emissionfactorcalculatedbythemodelleadstotheestimationof
thetotalemissionsproducedbythevehiclesintheparticularclass.
Themeanmileagepercategoryisderivedastheweighted
averageofvehiclesofdifferentagewhichcomplywiththesame
emissionlimit,astheyareregisteredintheofficialstatistics(ACI,
2011).Thepercentagedistributionofvehiclesduringdifferent
yearsofage(Table1)wasusedinordertoestimateaveragevalues
ofAAAMseparatelyforthedifferentlegislativeclassesofdiesel
andgasolinevehicles,fromEuro0(non–catalyticvehicles)toEuro
5,aswellasanaveragemileageforthewholefleet(Figure5).It
shouldbenotedthattheestimateisrelatedtotheyearinwhich
theestimationofmileagewascarriedout,thatistheyear2010in
ourcase,andsooneyearoldvehiclesaretheonesregistered
during2009.Aswouldhavebeenexpected,themeanmileagefor
vehiclesregisteredbefore1992(Euro0)isonlyasmallfractionof
newEuro5vehicles.
Figure4.Comparisonbetweentheresultsofthepresentstudyanddatafromtheliteraturereviews:USAdatacollectedbetween1997
and2002(NHTSA,2006),TREMOVEdata(EC,2005)andSwedishdatacollectedbetween1987and1998(Andreetal.,1999).
Figure5.
A
AAM(1
0
3km/y)fordifferenteurocategoriesandaverageforthewholefleetintheyear2010.
Caserini et al. – Atmospheric Pollution Research (APR) 288
ThemileagevaluesestimatedinFigure5canthenbeusedas
inputtoemissionmodelstoestimateroadtransportemissions.We
haveusedCOPERT4asanexample,todemonstratetheimpactof
thedroppingmileagewithageontotalemissionestimates.The
maininputdatausedforsuchacalculationareshowninTable2.
Theemissionfactorsfornitrogendioxide(NOX)andparticulate
matter(PM)werederivedasaggregatesfromthedetailedCOPERT
4methodology,usingdetailedactivityandenvironmental
informationcorrespondingtotheItalianconditions.Thenumberof
vehiclessplitbyemissionlegislationwasderivedfromnational
statistics,andtherelativemileagepertechnologystepwas
estimatedusingtheapproachesdescribedabove(constant
mileage,AAMandAAAM).Basically,themileagevaluesestimated
pereachclassbasedeitherontheAAMortheAAAMmethods
wereproportionallyadjustedtoleadtothesamefuelconsumption
astheconstantmileagecase.Thisisthetypicalprocedurefollowed
inemissioninventories,i.e.arelativemileagepertechnologyclass
isfirstestimatedandthenitisproportionallyadjustedtomeetthe
fuelconsumptionreportedbytheofficialstatistics.Withthis
method,allemissionresultsshowninTable3correspondtothe
sametotalfuelconsumption.
Despiteallcalculationscorrespondtothesamefinalenergy
utilization,therearesignificantdifferencesinthetotalemissions
calculationsforbothpollutants.BothNOXandexhaustPM
emissionsdropbymorethan20%whenAAAMisusedinsteadof
fixedmileage.ThedifferenceinNOXmostlycomesfromgasoline
carsasthetrueNOXemissionfactors(Table2)ofdieselcarsdonot
consistentlydropwithanimprovingemissionstandard.Hence,the
allocationofmileagetogasolinevehiclesismoreimportantthan
dieselonesinthecaseofNOX.TheoppositeoccursincaseofPM
emissionswheredifferencesingasolinevehiclePMemission
factorsareminimalandthemainreductionsoriginatefromdiesel
vehiclesonly.RelevantdifferencesalsooccurwhentheAAM
estimateofmileageisused.
Suchemissiondifferencesarenottrivial.Forexample,outof
theelevenmemberstatesthatemittedmorethantheirNOX
targetsaccordingtotheemissionceilingsdirective(EEA,2011),
eightofthemonlyexceededtheirlimitsbylessthan20%.Taking
intoaccountthatroadtransportaloneissome40%oftotal
nationalNOXemissions(EEA,2010)thedifferenceofmorethan
20%thatwecalculatedduetomileageestimationonlyinthis
studywouldbeequivalentofmorethan8%oftotalnation–wide
NOXemissions.Suchadifferencewouldbringanumberof
countriescloserorwithintheirallowedlimits.
Table2.AggregatedNOxandPM10emissionfactorsandmileagevaluesusedforenvironmentalmodeling
FuelLegislative
category
NOX
(mg/km)
PM10
(mg/km)
Vehicle
share(%)
Constant
mileage
(km/y)
AAM
(km/y)
AAAM
(km/y)
Gasoline
euro019582.418%697042191829
euro14262.38.4%697051893038
euro22232.226%697063034909
euro3951.117%697076537854
euro4571.128%6970919312088
euro5461.11.9%6970996314548
Diesel
euro06952164.6%1363861772267
euro1691892.4%1363884704023
euro27345413%13638109846961
euro38044331%136381405711966
euro46003645%136381690017922
euro54331.84.0%136381876722525
Table3.NOxandPM10emissionscalculatedwithdifferentaveragemileageestimates
NOX PM
Constant
mileageAAMAAAM Constant
mileage AAMAAAM
Gasoline
euro073.2%63.7%46.6% 24.8%16.4%8.1%
euro17.6%8.1%8.0%11.4%9.3%6.1%
euro212.3%16.0%20.9% 34.4%34.1%30.0%
euro33.4%5.4%9.3%10.6%12.8%14.8%
euro43.4%6.4%14.2%17.6%25.5%37.8%
euro50.2%0.4%1.0%1.2%1.9%3.1%
Total(t/y)678254716027949 243221196
Diesel
euro04.6%2.0%0.8%20.2%9.8%4.2%
euro12.4%1.4%0.7%4.4%2.9%1.6%
euro213.9%10.7%7.3%14.4%12.4%9.2%
euro336.9%36.1%33.3% 27.9%30.9%30.6%
euro439.6%46.5%53.5% 32.9%43.7%54.1%
euro52.5%3.3%4.3%0.1%0.2%0.3%
Total(t/y)129219136159125524 923086067382
Caserini et al. – Atmospheric Pollution Research (APR) 289
Moreover,itwouldbeinterestingtoexplorewhatwouldbe
theimpactofintroducingameasurethateliminatesnon–catalyst
vehiclesfromtheroad.Suchmeasurescouldbeanincentive–
basedscrappageschemeoranenvironmentalzoneenforcedina
partofacity.Ifonemakestheusualassumptionthatallvehicles
aredrivenforthesamedistanceeitherannuallyoronastreet
network,theneliminationofnon–catalystvehicles(Euro0)would
haveledtoassumethat28%ofNOXand20%oftotalPMemissions
fromgasolineanddieselpassengercarsshouldbereducedwith
suchameasure.However,takingintoconsiderationthatolder
vehiclesaredrivenless,theactualimprovementwouldonlybe9%
and4%respectively.Thisentirelychangesthecost–benefitratios
ofsuchmeasures.
Despitesuchsignificantimpacts,widespreadandrobust
estimatesofmileageasafunctionofspeedarestilllackingandour
strongrecommendationisthatsuchinformationhastobemore
reliablyandthoroughlyassessedandcollected.
5.Conclusions
Inordertoimprovetrafficemissionestimatesand,
consequently,fordefiningthestrategiesaimedtocontrolair
pollutionevents,thepresentworkhighlightstheimportanceof
increasingourknowledgeonvehiclemileagebehavior.Basingon
anextensivedataset(morethan33000data),arelationship
betweenvehiclemileageandagehasbeendefined,bothfordiesel
andgasolinepassengercars,withtheexampleoftheItalianstock.
Amethodologyhasalsobeenpresentedwhichcanbeappliedto
thenationalconditionsinothercountries.Theresultsofthis
methodologyshowthatannualmileagedropssignificantlywith
mileageage.Bothdieselandgasolinecarsdrivehalftheannual
distancewhentheyhavereachedanaverageageofapproximately
8years.Vehiclesof20yearsofageonlydriveapproximately10%
oftheannualdistancetheyusedtodrivewhentheyarenew.
Theimpactofthedroppingmileagewithageissignificantin
assessingtheenvironmentalimpactsoftransportandthepotential
impactofenvironmentalpolicies.NOXandPMemissionsof
passengercarsdropbymorethan20%whenadecreasingfunction
ofmileagewithageisused,insteadofafixedmileageforeach
environmentalclass.Also,theemissioncontributionfromold
vehiclesdecreaseswhichworsensthecost–effectivenessofair
qualityrelatedpolicymeasurestargetingsucholdvehicles.These
findingsdemonstratetheimportanceofperformingprecise
estimatesofmileagepervehicleclassifrobustroadtransport
emissioninventoriesneedtobeproduced.
SupportingMaterialAvailable
Listofwebsitesusedtocollectdata(TableS1);Meanmileage
(km)andstandarddeviation(km)ofthevehiclesampleperagebin
(TableS2);Averagecumulativemileageofgasolinecarsasa
functionofend–of–lifeage(FigureS1);Averagecumulative
mileageofdieselcarsasafunctionofend–of–lifeage(FigureS2).
ThisinformationisavailablefreeofchargeviaInternetat
http://www.atmospolres.com.
References
ACI(AutomobileClubd’Italia),2011.http://www.aci.it/laci/studi–e–
ricerche/dati–e–statistiche/autoritratto.html,accessedinJanuary
2013.
Andre,M.,1999.AirPollutionduetoTransport–StatisticsofVehicleFleet
andTransportActivityinFrance,ReportINRETS–LTEn°9930,24pages.
Andre,M.,Hammarstrom,U.,Reynaud,I.,1999.DrivingStatisticsforthe
AssessmentofPollutantEmissionsfromRoadTransport,INRETSReport
LTE9906,194pages.
Beydoun,M.,Guldmann,J.M.,2006.Vehiclecharacteristicsandemissions:
logitandregressionanalysesofI/MdatafromMassachusetts,
Maryland,andIllinois.TransportationResearchPartD:Transportand
Environment11,59–76.
Bin,O.,2003.Alogitanalysisofvehicleemissionsusinginspectionand
maintenancetestingdata.TransportationResearchPartD:Transport
andEnvironment8,215–227.
EC(EuropeanCommission),2012.StatisticalPocketBook2011,
http://ec.europa.eu/transport/publications/statistics/pocketbook–
2011_en.htm,accessedinJanuary2013.
EC(EuropeanCommission),2005.DevelopmentoftheTREMOVE
environmentandtransportmodel,http://ec.europa.eu/environment/
archives/tremove/tremove_model_dev.htm,accessedinJanuary2013.
EEA(EuropeanEnvironmentAgency),2011.Distance–to–targetforEEA
MemberCountries.,Copenhagen,2pages.
EEA(EuropeanEnvironmentAgency),2010.StateandOutlook2010.Air
Pollution,Copenhagen,228pages.
EREG(EuropeanVehicleandDriverRegistration),2010.ImpactStudyof
MileageFraudwithUsedCars&AdaptabilityoftheCar–Passmodelin
otherEUCountries,ReportoftheAssociationofEuropeanVehicleand
DriverRegistrationAuthorities,Brussels,140pages.
HausbergerS.,RexeisM.,ZallingerM.,LuzR.,2009.EmissionFactorsfrom
theModelPHEMfortheHBEFAVersion3,ReportoftheInstitutefor
InternalCombustionEnginesandThermodynamics,GrazUniversityof
Technology,76pages.
NHTSA(NationalHighwayTrafficSafetyAdministration),2006.Vehicle
SurvivabilityandTravelMileageSchedules,ReportNo.DOTHS809
952,Washington,37pages.
Ntziachristos,L.,Gkatzoflias,D.,Kouridis,C.,Samaras,Z.,2009.COPERT:a
Europeanroadtransportemissioninventorymodel.Information
TechnologiesinEnvironmentalEngineering,491–504.
Ntziachristos,L.,Mellios,G.,Kouridis,C.,Papageorgiou,T.,
Theodosopoulou,M.,Samaras,Z.,Zierock,K.H.,Kouvaritakis,N.,
Panos,E.,Karkatsoulis,P.,Schilling,S.,Meretei,T.,AladarBodor,P.,
Damjanovic,S.,Petit,A.,2008.EuropeanDatabaseofVehicleStockfor
theCalculationandForecastofPollutantandGreenhouseGases
EmissionswithTREMOVEandCOPERT,LATReportNo.08.RE.0009.V2,
Thessaloniki,Greece,260pages.
Sawyer,R.F.,Harley,R.A.,Cadle,S.H.,Norbeck,J.M.,Slott,R.,Bravo,H.A.,
2000.Mobilesourcescriticalreview:1998NARSTOassessment.
AtmosphericEnvironment34,2161–2181.
Smit,R.,Smokers,R.,Rabe,E.,2007.Anewmodellingapproachforroad
trafficemissions:VERSIT+.TransportationResearchPartD:Transport
andEnvironment12,414–422.
Uherek,E.,Halenka,T.,Borken–Kleefeld,J.,Balkanski,Y.,Berntsen,T.,
Borrego,C.,Gauss,M.,Hoor,P.,Juda–Rezler,K.,Lelieveld,J.,Melas,D.,
Rypdal,K.,Schmid,S.,2010.Transportimpactsonatmosphereand
climate:landtransport.AtmosphericEnvironment44,4772–4816.
Washburn,S.,Seet,J.,Mannering,F.,2001.Statisticalmodelingofvehicle
emissionsfrominspection/maintenancetestingdata:anexploratory
analysis.TransportationResearchPartD:TransportandEnvironment6,
21–36.