Spatio‐temporally Resolved Methane Fluxes From the Los Angeles Megacity Vineet Yadav.pdf

该【Spatio‐temporally Resolved Methane Fluxes From the Los Angeles Megacity Vineet Yadav 】是由【宝钗文档】上传分享，文档一共【36】页，该文档可以免费在线阅读，需要了解更多关于【Spatio‐temporally Resolved Methane Fluxes From the Los Angeles Megacity Vineet Yadav 】的内容，可以使用淘豆网的站内搜索功能，选择自己适合的文档，以下文字是截取该文章内的部分文字，如需要获得完整电子版，请下载此文档到您的设备，方便您编辑和打印。:..YadavVineet(OrcidID:0000-0002-2805-3345)DurenRiley(OrcidID:0000-0003-4723-5280)MuellerKim,Lynn(OrcidID:0000-0002-3516-2259)VerhulstKristal,R.(OrcidID:0000-0001-5678-9678)NehrkornThomas(OrcidID:0000-0003-0637-3468)KimJooil(OrcidID:0000-0002-2610-4882)WeissRay,F(OrcidID:0000-0001-9551-7739)KeelingRalphF.,F.(OrcidID:0000-0002-9749-2253)SanderStanley(OrcidID:0000-0003-1424-3620)NewmanSally(OrcidID:0000-0003-0710-995X)FalkMatthias(OrcidID:0000-0002-8901-2838)HopkinsFrancesca(OrcidID:0000-0002-6110-7675)MillerCharles,E.(OrcidID:0000-0002-9380-4838)Spatio-temporallyresolvedmethanefluxesfromtheLosAngelesMegacityeptedforpublicationandundergonefullpeerreviewbuthasnotbeenthroughthecopyediting,typesetting,:?.:..Abstractworkofatmosphericmonitoringstationswithinversemodelingtouniquelyobtainspatio-temporal(3km,4day)estimatesofmethaneemissionsfromtheLosAngelesMegacityandthebroaderSouthCoastAirBasin(SoCAB)for2015--InvertedLagrangianmodelforSoCABandinnovativelyemployamodelresolutionmatrixbasedmetrictodisentanglethespatio-,,,ourfindingsshowlargeseasonalvariationsinCH4fluxeswithsignificantlyhigherfluxesparisontosummermonths,whichisconsistentwithnaturalgasdemandandanti-correlatedwithair-,thisisthefirststudythatutilizesinversionstodetectbothenhancement(AlisoCanyonLeak)andreduction(PuenteHills)inCH4fluxesduetotheunintendedeventsandpolicydecisionsandtherebydemonstratestheutilityofinversemodelingforidentifyingvariationsinfluxesatfinespatio-?Detectedandspatio-temporallyresolvedmajormethanefluxanomaliesintheLosAngelesMegacityDomain.?CharacterizedvariabilityinSouthCoastAirBasinmethanefluxes.?Evaluatedimpactoflossofobservationalsensitivityonfluxestimates.?.:..1IntroductionMethane(CH4)hemistry,airquality,andclimate(Ciaisetal.,2014).RegionalandurbanCH4budgetsremainhighlyuncertainandirreconcilablediscrepanciesexistbetweentop-downandbottom-upestimatesforeventhelargestcomponents(Dlugokenckyetal.,2009;Kirschkeetal.,2013).,“thereisanurgentneedtoimprovethemonitoringandmeasurementofmethaneemissionsfromthemajorsourcesinCalifornia”andinparticular“high-emissionmethanehotspotsintheState”.CH4CH4CH4CH4emissionCH4studies(Cuietal.,2015;Jeongetal.,2016;Peischletal.,2013;Santonietal.,2014;Wechtetal.,2014;Wennbergetal.,2012;Wongetal.,2016;Wunchetal.,2009,2016).CH4?.:..workof15in--temporalvariabilityofCH4emissionsatafinespatio-,webineitwithoutputfrommeteorologicalmodelstailoredforSoCABtoobtaininverseestimatesofCH4emissionsandtheiruncertaintyatfinespatio--basinestimatesofCH4emissionsandspatio-(nearreal-time)identifyingtheseanomaliesthroughinversemodelingthatdoesnotrelyontailoredmeteorologicaloutputispresentedinWareetal.,,binein-H4concentrationswithinversemodelingto:(1)characterizebasinandsub-basinscaletemporalvariabilityinfluxesincludingtheonsetanddisappearanceoflargeCH4sources,(2)identifythelocationsofthemajorsourcesofCH4emissionsinthebasin,and(3)worktoupdatefluxesandidentifyspatio-,ouranalysisperiodspansthemassiveAlisoCanyongasleakfromlateOctober2015throughmid-February2016(Conleyetal.,2016),.,,eedsclosureofthePuenteHillslandfillwhoseimpactonsub--temporalresolutionofinversionsCH4fluxesinthisworkareestimatedfortheSoCAB(Figure1)°spatial(1826grid-cells)and4-daytemporalresolutionfromJanuary28,2015toDecember24,-,formost4-dayinversionwindowswewantedtohaveanapproximateratioof1:10betweenobservationsandestimatedfluxes,whichwasachievedforthespatio-,inversionswereperformedfortwo4-dayperiodsatatimewhilemaintaininganoverlapofa4--workcurrentlyconsistsof15urbansites;howeverduetodataavailabilityandthesensitivityofsitestofluxesinthestudyareaonlynineintheurbandomain?.:..andonebackgroundsite(Figure1)-continuousmeasurementsofboundarylayerCH4concentrations(see,Verhulstetal.,2017).Figure2aillustratesthetime-seriesof4-dayaverageCH4enhancementsateachsiteusingbackgroundestimatedfrommeasurementsobtainedfromSanClementeIsland(Verhulstetal.,2017).ThisbackgroundwasdeterminedbyapplyingthetechniquedescribedinThoningetal.,1989thatlooksatthestabilitycriteria(withinhourandhour-to-hourvariability)-datamismatch(fordefinition,)(see;Figure4b,Figure4dandFigureS10inVerhulstetal.,2017).Toensurewell-mixedconditions,weonlyusedafternoonmeasurementsfrom12to4pmwhenwindspeedsweretypicallybetween2to9m/.(Notethatforanextendedperiodin2016observationsfromonlysixsiteswereavailable.)Toavoidintroducingapositivebiasinthefluxestimates,(GIM)approach(.,Fangetal.,2014;Gourdjietal.,2008;Michalaketal.,2004;Shigaetal.,2014),theobjectivefunctionforwhichcanbegivenas:1???11???1????,??=(?????)??(?????)+(???????)(???????)(1)22where??(,1)arehourlyCH4measurements,(,)isaJacobianmatrixrepresentingthesensitivityofmeasurementstounderlyingflux,??(,1)aretheCH4fluxes,??(,)isthemodel-datamismatcherrorcovariancematrix,??(,??)isamatrixofcovariates,??(??,1)arethecoefficientsorweightsofindividualcovariatesand(,)istheerrorcovariancematrix(akapriorcovariance)thatdescribesthedeviationsof??from????.Asinearlierstudies,??and??areestimatedaspartoftheinversion,.,???and???.isobtainedfromtheWeatherResearchForecasting(WRF)-StochasticTime-InvertedLagrangianTransport(STILT)model(seesection4fordetails)andparametersfor??andarecomputedthroughrestrictedmaximumlikelihood(seeGourdjietal.,2010;Michalaketal.,2004).Inthiswork,(.,Figure1)aninventoryofCH4emissions(Jeongetal.,2012)wasprescribedasacovariatein??.putedforeachtowerin??anda?.:..putedforwhoseoff-diagonalentrieswerezeroanddiagonalentriescontainedCALGEMemissionestimatesforeachgrid-cell(.,Yadavetal.,2016).WeusedLagrangemultiplierstoimposeanon-negativityconstraintoninverseestimatesofCH4fluxestoavoidnon-physicalnegativefluxes(seeMilleretal.,2014).Imposingnon-negativityputationuntenable;therefore,weadoptedasimulation-??,and??,onlyparametersof??,andregulatetheestimateof???and???,toestimategrid-scaleuncertaintyon???,firstweuserestrictedmaximumlikelihoodtoobtain??andparametersandtheirposteriorerrorcovariancefromtheinverseoftheobservedFisherinformationmatrix(.,Kitanidis,1995,1996).Inthenextstep,wegenerate1000samplesof??position(seeMichalaketal.,eneratingtheserealizations)andusethemtocompute1000realizationsof???fromthestandardGIMequationsgiveninMichalaketal.,,weimposenon-negativityonthese1000realizationsof???usingLagrangemultipliersandthestandarddeviationofthese1000realizationsisconsideredasposterioruncertaintyongrid--scaleposterioruncertaintyisthebasin-,thisprocessofestimatinguncertaintyresultsinmoreconservativeestimatesincomparisontoothermethodsincludingMarkovChainMonteCarlo(seeJeongetal.,2012),analyticalandconditionalsimulations(seeequationsinMichalaketal.,2004).Computationally,fora4-dayperiod,putethese1000realizationsbutasmostoftheserunswereputertheuncertaintyestimateforall1754-:(1)correlationandRMSE(RootMeanSquaredError)betweenobservations(??)andconvolutionofestimatedfluxeswithtransport(???),(2)time-seriesofthetraceorthediagonalofthemodelresolutionmatrixscaled()bythecountofobservations(??),and(3)reducedchi-,,.,themodelresolutionmatrixoranaveragingkernelmatrix(fordetailssee,Asteretal.,2018;Tarantola,2004):(1)hasadefinitelowerandupperbound(rangeof0-1),(2)passesinformationabouttheposterioruncertaintyontheestimatedfluxes,(3)conveysinformationregardingthespatio-work,?.:..and(4)leadstoidentificationofthegrid-(.,Yadavetal.,2016)canbegivenas:???=??(2)whereisasdefinedearlierand??(,)canbewrittenas:????????1?????1(3)??=??(??????)????+?????1??????1??(?????????1??)?1?????????1and??=T+??(4)eofinversionif???(,),????(???)(Notethesymbol????denotesthetraceofamatrix),????(???)≤????(???)=??,itisunlikelythatfluxofaparticulargrid-.,ithasavalueofonealongthediagonalof???.Thishappensduetothepresenceofuncertaintyassociatedwithobservations,transportandprior,whichpreventsperfectresolutionofthetruegrid-.,grid-,theresolutionisspreadovermultiplegrid-cellsindicatedbytheoffdiagonalentriesin??,asuniquelyemployedinthisstudy,athresholdbasedonthediagonalentriesof???canstillbedefinedtoidentifyregionsoflargeemissions,,puting??2statisticanddividingitbydegreesoffreedom?