Hawkins et al Unvertainty.docx

Hawkins et al Unvertainty.docx

《Hawkins et al Unvertainty.docx》由会员分享，可在线阅读，更多相关《Hawkins et al Unvertainty.docx（24页珍藏版）》请在冰豆网上搜索。

HawkinsetalUnvertainty

UncertaintyintheMixed-UnitInput-OutputLifeCycleAssessmentModeloftheUSEconomy

TroyHawkinsa,ChrisHendricksonb,H.ScottMatthewsc

GreenDesignInstitute

CarnegieMellonUniversity

5000ForbesAvenue,Pittsburgh,PA15213USA

atrh@andrew.cmu.edu,bcth@andrew.cmu.edu,chsm@andrew.cmu.edu

Abstract

Bringinginput-outputbasedtechniquesforenvironmentalresearchtoabroaderaudiencerequiresbetterunderstandingandcommunicationoftheuncertaintyassociatedwiththeirresults.Herewediscussuncertaintiesininput-outputlifecycleassessmentmodelsbasedonourexperienceindevelopingtheMixed-UnitInput-OutputLifeCycleAssessment（MUIO-LCA）modelfortheUSeconomy.TheMUIO-LCAmodelextendsthe500sector1997USBenchmarkmakeandusetablesthroughtheadditionofcommoditiesandindustriestorepresenttheflowofcadmium,lead,nickel,andzincinmassunits.Thesesectorsallowexplicittrackingofmaterialflowsandforthecalculationofpollutantreleasesbasedonphysicalquantitiesratherthandollarvalues.UncertaintiesintheUSGeologicalSurveydatausedtocreatetheseaccountsarediscussed.TheeffectoflevelofaggregationontheusefulnessanduncertaintyofIO-LCAmodelsispresentedinthecontextofMUIO-LCA.GuidancerelatingtouncertaintyassociatedwiththeassumptionofaUStechnologymixforimportedmetalsisalsoprovided.UncertaintyintoxicreleasemultipliersbasedontheUSEPAToxicsReleaseInventoryispresentedaswellasadiscussionofthetreatmentofuncertaintyforasetofmaterialusemultipliersbasedonUSGeologicalSurveydata.OurexperiencewithuncertaintyinthedevelopmentoftheMUIO-LCAmodelprovidesguidancefortheinterpretationofIO-LCAmodelresultsandforimprovedtreatmentofuncertaintyinthenextgenerationofIO-LCAmodels.

Introduction

Input-outputtechniquesareincreasinglyusedforenvironmentalpolicyanalysisandenvironmentallifecycleassessment.ResearchersarerealizingthebenefitofIOmodelsinsimplifyingtheanalysisofsupplychainsandreducingthetruncationerrorassociatedwithprocess-basedanalysis.ImprovingtherobustnessoftheresultsofIObasedenvironmentalassessmentsrequiresimprovingourunderstandingofmodeluncertainty.WeofferanassessmentoftheuncertaintiesassociatedwithIOmodelsforenvironmentalassessmentbasedonourexperiencedevelopingtheMixed-UnitInput-OutputLifeCycleAssessment（MUIO-LCA）model.

LiketheEIO-LCAmodel,alsodevelopedthroughtheGreenDesignInstituteatCarnegieMellonUniversity,theMUIO-LCAmodelisbasedontheUSBenchmarkIOAccountscombinedwithadditionaldatarelatedtoreleasesofpollutants,energyconsumption,andmaterialuse.MUIO-LCAextendsthecapabilityofEIO-LCAbyaddingcommoditiesandindustriesrelatedtocadmium,lead,nickel,andzincflows.Metaloutputofthesesectorsaretrackedinmassunits.Theinclusionofadditionalsectorsallowsforexplicittrackingofmaterialflowsandcalculationofmetaluse.LikeEIO-LCA,MUIO-LCAallowsforthecalculationofpollutantreleasesandenergyusethroughoutthecompletesupplychainofanindustry.

Modelpredictionsarenevercertain.Understandinguncertaintyinamodelisimportanttointerpretingitsresults.Thisbecomesespeciallyimportantiftheoutcomestobecomparedarenearoneanotherinmagnitude.InterpretingtheresultsofanIO-LCAmodelisespeciallytrickyduetothelargeamountsofdataandmanyassumptionsonwhichtheresultsarebased.ThecommonguidancegiventothoseinterpretingresultsofEIO-LCAhasbeenthattheyshouldbeconsideredwithinanorderofmagnitudeofthetruevalues.ThroughoutdevelopmentoftheMUIO-LCAmodelwehaveattemptedtotracktheassumptions,errors,anduncertaintiesinvolvedinthemodel.HerewewillusethisexperiencetoprovideguidancerelatedtotheuncertaintyofMUIO-LCA.OurdiscussionalsohighlightsuncertaintiesinEIO-LCAandthe1997USBenchmarkAccountsonwhichMUIO-LCAisbased.

InTable1wepresentanoverviewofsourcesoferrorinIOLCAmodelspresentedinnoparticularorder.Weprovidebriefdescriptionsofthefirst7typesoferrorinthesectionthatfollows.ThefinalthreetypesoferroraredescribedinmoredetailwithspecificattentiontotheMUIO-LCAmodel.

SeveralsourcesoferrorinIOLCAmodelshavebeenillustratedinpreviouslypublishedworks.Itisnotourdesiretoprovideacomprehensivediscussionhere.Ratherwewillfocusoninstanceswhereourexperienceprovidesuniqueinsights.Lenzen（'01）providesamorecomprehensivediscussionoferrorinIOLCAmodelstowhichthereadercanrefer.

UncertaintyinIO-LCA

SourceDataUncertainty

Sourcedatauncertaintyreferstouncertaintyintheunderlyingdataonwhichthemakeandusetablesarebased.Inthecaseofthe1997BenchmarkAccountstatisticaltechniquesareappliedtoalargeamountofdatafromtheEconomicCensus,ForeignTradeDatabase,andCommodityFlowSurveytoestimatetheentriesinthemakeandusetables.ResponsestotheEconomicCensusarenotalwaysaccurate.Althoughadjustmentsaremadetoaccountforthis,someamountofuncertaintypropagatesthroughthemodel.Uncertaintyisalsointroducedbysampling,estimations,anddatamanipulation.

EstimationofTransactions

Estimationoftransactionsreferstouncertaintyintroducedbytheestimationofmakeandusetableentries.Thisuncertaintyisstronglyrelatedtosourcedata.Incaseswheresourcedataisverylimited,simplifyingassumptionsmustbemadetoallowtheestimationofinter-industrytransactions.Entriesinthe1997USBenchmarkmakeandusetablesarealsoadjustedtoreallocateproductionofsomesecondaryproductstotheirprimaryindustryandtobalancethetotaloutputsofthemakeandusetables.Commodityproductionandconsumptionarereallocatedfromtoreducetheamountofsecondaryproductsproducedbyindustries.Productionofcertaincommoditiesismovedtotheprimaryindustryandtheconsumptionmixisadjustedaccordingly.Tablesarebalancedbyadjustingtheentriesuntilthetotalindustryoutputandtotalcommodityoutputcalculatedasthesumsofrowsandcolumnsofthemakeandusetablesbalance.Thesequantitiesoftendonotmatchinitiallyduetomisreported,erroneous,ormissingdataaswellasthetimelagbetweenthepurchaseofinputsandtheproductionofgoods.BalancingwasperformedbytheBEAbasedonexpertopinionandcomparisontothe1992account.Remainingdifferencesarecorrectedbyadjustingentriesinothervalueadded（Lawson'02）.

ProportionalityAssumption

IOmodelsestimatesupplychainaffectsunderanassumptionofproportionality.Large-scalechangeswhicheffectavailabilityofsupply,augmentationofinfrastructure,orpricesarenotwellrepresentedintypicalIOmodelsdescribedhere.Generallytheimpactoflarge-scalechangesisunderestimatedbyIO-LCAmodels.

Cradle-to-GateTruncation

IO-LCAmodelscaptureonlycradletogateimpactsofaproduct.Thatistheimpactoccurringfrommaterialextractionthroughmanufacturingtothepointofsale.Additionalinformationisneededtoestimatetheuseandend-of-lifephasesoftheproductlifecycle.Thisshouldnotintroduceuncertaintyintoresultsaslongastheuserunderstandstheproperuseofthemodel.Oftenhowever,IOmodelresultsaremisrepresentedastheentireimpactofaproduct.

ChangesinTechnologyorProductionMixOverTime

ChangesintechnologyorproductionmixovertimeareoftennotwellcharacterizedbyIO-LCAaccountswhichrepresentasnapshotofaneconomy.Allofthedatausedarefromaspecificpointintime,1997inthecaseofthe1997USBenchmarkAccounts.Changesaffectingthetechnologystructureoccurevenoveraoneyeartimeperiod.Beyondthis,theresultsofIOmodelsareoftenextrapolatedtorepresentfutureyears.TheUSEconomicCensusisperformedevery5years.TheUSBEArequiresanother5yearstoconstructthemakeandusetables.Thusthemostrecentmodelavailableisoftenbasedondatafrom5to10yearsearlier.Properlyinterpretingmodelpredictionsoftheconsequencesofcurrentdecisionsshouldinvolveconsiderationoftheinfluenceofchangesintheeconomyoverthepast5-10yearsonmodelpredictions.

ModelInputUncertainty

UsersoftheEIO-LCAmodelareofteninterestedintheproductionofacertainamountofagoodsuchasabarrelofoil,alead-acidbattery,oranautomobile.Usingthemodelrequirestransformingthefunctionalunittoadollaramountoffinaldemandinthemostcloselyrelatedsector.Inputsmustalsobeadjustedtoreflectproducer’spricesforgoods（UNDESA'99）.Marginsanddeliverycostsshouldbeinputtothemodelasfinaldemandsforretailtrade（4A0000）,wholesaletrade（420000）,trucktransportation（484000）,railtransportation（482000）,watertransportation（483000）,airtransportation（481000）,etc.Allfinaldemandinputsmustalsobeinflatedordeflatedtoreflect1997dollars.

Generallymodelusersaremorefamiliarwiththevaluesofgoodsincurrentpurchaser’sprices.DevelopersofIOLCAmodelsshouldtakethisintoconsiderationwhendesigningtheiruserinterfaceanddocumentation.Ideallyuserswouldbepromptedwithinformationabouthowthemodelinputshouldbedetermined.Consumerpriceindices（CPI）areavailableforinflating/deflatingpricesto1997dollars,howeverthecalculationofCPIitselfintroduceserror.Adjustingafinaldemandinpurchaserpricetoreflectproducerprice,margins,anddeliverycanbedonewiththeuseofatransformationmatrixbasedontheaveragemarginsforacommodity.Thepurchaser-producerpricetransformationmatrixcanbecalculatedusinginformationprovidedintheUSBenchmarkAccountsbasedonintermediateorf