未來能源研究所-太陽能地球工程對溫度歸因死亡率的影響（英）

ImpactofSolar

Geoengineeringon

Temperature-Attributable

Mortality

AnthonyHarding,DavidKeith,WenchangYang,andGabrielVecchi

WorkingPaper23-23

May2023

ResourcesfortheFuturei

AbouttheAuthors

AnthonyHardingisapostdoctoralfellowresearchingtheintersectionofinnovativetechnologiesandclimatepolicy.HereceivedhisPhDineconomicsfromGeorgiaInstituteofTechnology,wherehisresearchfocusedonclimateandenergyeconomics,andearnedaBSfromRensselaerPolytechnicInstituteinmathandphysics.Harding’sresearchappliesbotheconometricsandeconomicmodellingtoevaluateclimatepolicyandclimateimpacts.Hismostrecentworkestimatesthedistributionofeconomicimpactsofsolargeoengineeringacrosscountriesandcomparesittotheimpactsofclimatechange.Hiscurrentinterestsincludethedesignofeffectiveinternationalclimategovernancestructuresandthemeasurementofthevalueofscientificlearning.

DavidKeithhasworkedneartheinterfacebetweenclimatescience,energytechnology,andpublicpolicyfortwenty-fiveyears.HetookfirstprizeinCanada'snationalphysicsprizeexam,wonMIT'sprizeforexcellenceinexperimentalphysics,andwasoneofTIMEmagazine's

HeroesoftheEnvironment

.KeithisProfessorofAppliedPhysicsattheHarvardSchoolofEngineeringandAppliedSciencesandProfessorofPublicPolicyattheHarvardKennedySchool,andfounderof

Carbon

Engineering

,acompanydevelopingtechnologytocaptureCO2fromambientairtomakecarbon-neutralhydrocarbonfuels.Bestknownforhisworkonthescience,technology,andpublicpolicyofsolargeoengineering,Keithledthedevelopmentof

Harvard’sSolarGeoengineeringResearchProgram

,aHarvard-wideinterfacultyresearchinitiative.

WenchangYangisanassociateresearchscholarintheDepartmentofGeosciencesatPrincetonUniversity,workinginthegroupofProf.GabrielVecchi.Hisresearchisfocusedonbetterunderstandingclimatevariabilityandchangeonbroadtimescalesfromsub-seasonstomillennia,aswellaswhythemeanclimateoftheplanetisthewayitis.

GabrielVecchiisaprofessorofgeosciencesat

TheHighMeadowsEnvironmental

Institute

,anddirectorof

CooperativeInstituteforModelingtheEarthSystem

atPrincetonUniversity.Hisresearchinterestsareclimatescience;extremeweatherevents;hurricanes;mechanismsofprecipitationvariabilityandchange;ocean-atmosphereinteraction;detectionandattribution.

ImpactofSolarGeoengineeringonTemperature-AttributableMortalityii

Acknowledgments

AnthonyHardingandDavidKeithacknowledgesupportfromtheLADClimateFund.GabrielVecchiandWenchangYangacknowledgesupportfromUSDepartmentofEnergyGrantDE-SC0021333.WethankKevinCromarandotherparticipantsinworkshopsatResourcesfortheFutureforusefulcommentsandfeedback.WethankSimoneTilmesforassistanceinaccessingGLENSsimulationdata.WethankAntonellaZanobettiandJoelSchwartzforhelpfulconversationsintheearlystagesoftheproject.

AboutRFF

ResourcesfortheFuture(RFF)isanindependent,nonprofitresearchinstitutioninWashington,DC.Itsmissionistoimproveenvironmental,energy,andnaturalresourcedecisionsthroughimpartialeconomicresearchandpolicyengagement.RFFiscommittedtobeingthemostwidelytrustedsourceofresearchinsightsandpolicysolutionsleadingtoahealthyenvironmentandathrivingeconomy.

Workingpapersareresearchmaterialscirculatedbytheirauthorsforpurposesofinformationanddiscussion.Theyhavenotnecessarilyundergoneformalpeerreview.TheviewsexpressedherearethoseoftheindividualauthorsandmaydifferfromthoseofotherRFFexperts,itsofficers,oritsdirectors.

AbouttheProject

TheResourcesfortheFutureSolarGeoengineeringresearchprojectappliestoolsfrommultiplesocialscienceresearchdisciplinestobetterunderstandtherisks,potentialbenefits,andsocietalimplicationsofsolargeoengineeringasapossibleapproachtohelpreduceclimateriskalongsideaggressiveandnecessarymitigationandadaptationefforts.Theprojectbeganin2020withaseriesofexpertworkshopsconvenedundertheSRMTrans-AtlanticDialogue.Thesemeetingsresultedina2021articleinSciencethatlaysoutasetofkeysocialscienceresearchquestionsassociatedwithsolargeoengineeringresearchandpotentialdeployment.TheProjectfollowedthiswithadditionalsponsoredresearch,includingacompetitivesolicitationdesignedtoaddressresearchareashighlightedintheSciencearticle.Thispaperisoneofeightresearchpapersresultingfromthatcompetitionandsupportedbytwoauthorworkshops.Akeygoalofthesolicitationandtheoverallprojectistoengagewithabroadersetofresearchersfromaroundtheglobe,agrowingnumberofinterestedstakeholders,andthepublic.

ResourcesfortheFutureiii

SharingOurWork

OurworkisavailableforsharingandadaptationunderanAttribution-NonCommercial-NoDerivatives4.0International(CCBY-NC-ND4.0)license.Youcancopyandredistributeourmaterialinanymediumorformat;youmustgiveappropriatecredit,providealinktothelicense,andindicateifchangesweremade,andyoumaynotapplyadditionalrestrictions.Youmaydosoinanyreasonablemanner,butnotinanywaythatsuggeststhelicensorendorsesyouoryouruse.Youmaynotusethematerialforcommercialpurposes.Ifyouremix,transform,orbuilduponthematerial,youmaynotdistributethemodifiedmaterial.Formoreinformation,visit

/licenses/by-nc-nd/4.0/

.

Abstract

Temperature-attributablemortalityisamajorriskofclimatechange.Weanalyzethecapacityofsolargeoengineering(SG)toreducethisriskandcompareittotheimpactofequivalentcoolingfromCO2emissionsreductions.WeusetheForecast-OrientedLowOceanResolutionmodeltosimulateclimateresponsetoSG.UsingempiricalestimatesofthehistoricalrelationshipbetweentemperatureandmortalityfromCarletonetal.(2022),weprojectglobalandregionaltemperature-attributablemortality,findthatSGreducesitglobally,andprovideevidencethatthisimpactislargerthanforequivalentcoolingfromemissionsreductions.Ataregionalscale,SGmoderatestheriskinamajorityofregionsbutnoteverywhere.Finally,wefindthatthebenefitsofreducedtemperature-attributablemortalityconsiderablyoutweighthedirecthumanmortalityriskofsulfateaerosolinjection.Thesefindingsarerobusttoavarietyofalternativeassumptionsaboutsocioeconomics,adaptation,andSGimplementation.

ImpactofSolarGeoengineeringonTemperature-AttributableMortalityiv

Contents

1.Introduction1

2.Results2

2.1.ClimateResponse2

2.2.EmpiricallyEstimatedImpact4

3.LimitationsandUncertainties6

3.1.EmpiricalEstimates6

3.2.ClimateSimulations9

4.TowardaRisk-RiskComparison11

5.Discussion11

6.Methods12

6.1.FLOR12

6.2.GLENS12

6.3.Normalization13

6.4.EmpiricalEstimatedImpact13

6.5.DownscalingandBiasCorrection14

6.6.WBTandWGBT14

7.References16

ResourcesfortheFuture1

1.Introduction

Climatemodelanalysesdemonstratewideaccordancethatsolargeoengineering(SG)applieduniformly(balancedacrosshemispheres)andmoderately(tooffsetlessthanhalfofthewarmingfromgreenhousegases)moderatessalientclimaterisks,suchasextremeweather(DagonandSchrag2017;IrvineandKeith2020),permafrostloss(Chenetal.2020,2022),andchangesincropyields(Fanetal.2021).Yet,italsointroducesnovelrisks.Theseincludethedirectrisksfromtheaerosolsused(Easthametal.2018),regionalexacerbationofclimatechanges(Moreno-Cruzetal.2012),andrapidandextremewarmingifsuddenlyterminated(ParkerandIrvine2018).DecisionsaboutSGshouldbeinformedbycomprehensiveandquantitativerisk–risk(Hardingetal.2022;Felgenhaueretal.2022;Parson2021)analysesthatweighSG’scapacitytomoderateclimaterisksagainsttherisksitsuseentails.

Wetakeasmallsteptowardamorecomprehensiverisk–riskanalysisbyestimatingtheimpactofSGontemperature-attributablemortality—amajorriskofclimatechange.Arecentglobal-scalestudy(Carletonetal.2022)findsthatthemortalityriskofclimatechangeisaround85deathsper100,000bytheendofthe21stcenturyforahigh-warmingscenario(14foramoderate-warmingscenario).Inmonetaryterms,thestudyestimatesthatthemortalitycomponentofthesocialcostofcarbonis$37/tCO2($17/tCO2foramoderate-warmingscenario).Forcomparison,theInteragencyWorkingGrouponSocialCostofGreenhouseGases(2021)reviseditsestimatesofthetotalsocialcostofcarbonforuseinregulatoryimpactanalysisto$51/tCO2initsFebruary2021report.Anotherrecentestimateofthesocialcostofcarbonthatincorporatestemperature-attributablemortalityriskfindsthatitisabouthalfthatcost(Rennertetal.2020).

DespitetheimportanceofmortalityriskandgrowinginterestinSG,researchonitshumanmortalityimpactisscant.Easthametal.(2018)quantifytheimpactofstratosphericsulfateaerosolonground-levelozone,particulatematter,andground-levelUV-Bfluxandfindanetmortalityincreaseof26,000(95percentCI:-30,000–79,000)deathsperyeartoreduceglobalmeantemperaturesby1°C.Otherresearchanalyzesmortalityfromtheeffectonheatstress(Kuswantoetal.2022)andmalaria(Carlsonetal.2022)butdoesnotquantifyimpacts.

RoughlyuniformSGmightemployarangeofmethods,fromvariousstratosphericaerosolstocirrusthinningtospace-basedmethods.ThemosttechnicallyfeasiblemethodistoaddSO2tothestratosphere.Ourjudgmentisthatthemostpolicy-relevantquestionforSGishowamoderateamountofitmightsupplementinamoderateemissionsreductionscenario(e.g.,tograduallyachieveapeakreductionof1°Cinthelatterhalfofthecentury).Giventheuncertaintyabouttechnologyanddeploymentstrategy,ourprimaryclimatesimulationusessolarconstantreductionasaproxyforuniformSG.Weexploretheconsequencesofthischoicebyusinganalternativesimulationinwhichsulfateaerosolisinjectedwithacontrolalgorithmthattriestomaintainmultipletemperaturetargets.

ImpactofSolarGeoengineeringonTemperature-AttributableMortality2

WeusetheGeophysicalFluidDynamicsLaboratoryForecast-OrientedLowOceanResolution(FLOR)modeltosimulateclimatechanges(Vecchietal.2014)(seeMethods).Relativetoa1990scontrol,wecompareclimateresponseovera200-yearexperimentfordoublingCO2concentrations(2xCO2experiment)anddoublingCO2concentrationsoffsetwithasolarconstantreductionof1.7percent(2xCO2+SGexperiment).Thissolarconstantreductionapproximatelyoffsetsthechangeintop-of-atmospherenetradiativeforcingfromdoublingCO2(FigureS1).FLORhasaspatialresolutionofabout50kmforlandandatmosphereand1°x1°foroceanandice.Wechoseamodelwithahighspatialresolutionbecauseitimprovesaccuracyinrepresentingextremeweather(vanderWieletal.2016;Philipetal.2021),whichisparticularlyimportantinoursetting.Weusethefinal100yearsofeachexperimenttoallowtheclimatesystemtoequilibrate.

2.Results

2.1.ClimateResponse

Ifglobalaveragetemperatureswerethesoledeterminantofclimateimpacts,thenSGmightperfectlycompensatefortheclimateimpactsofCO2.ButimpactsdependonlocalclimatechangesthatcannotbeeliminatedbySG.AcentraltechnicalquestionaboutSGishowmuchitexacerbateslocalclimatechanges—increasingtheirdeviationsfrompreindustrial.BecausetheamountofSGcoolingisapolicychoice,itisoftenmostusefultocomparetheeffectstothoseofidenticalglobalaveragecoolingfromreducedCO2concentrations.WeintroducearatiometricTXthatmeasurestheeffectofSGonavariableXnormalizedperdegreeglobalmeancoolingrelativetotheeffectofreducingCO2concentrationsnormalizedperdegreeglobalmeancooling(seeMethods),whichcouldrepresenttheeffectofemissionsreductionsrelativetoahigherconcentrationscounterfactualortheeffectofdirectcarbonremoval.Intheremainderofthepaper,werefertothisasthe“effectofemissionsreductions.”AnTX>1indicatesthattheresponsetoSGisgreaterthantheresponsetoemissionsreductions,andTX<1indicatestheresponsetoSGislessthantheresponsetoemissionsreductions.

WeanalyzetheclimateresponsetoSGfordry-bulbtemperature,theclimateinputvariableforourempiricallybasedanalysisoftemperature-attributablemortality.Weconsidertheresponseoftheannualmeansoftemperature(),intensityofconsecutivehotdays(THW),andintensityofconsecutivecolddays(TCW)atthegrid-celllevel.Theintensityofcold(heat)extremesaremeasuredasthe10th(90th)percentileoftherollingfive-daymaximum(minimum)dailytemperaturesannually.TheSupplementaryMaterials(FiguresS2andS3)presenttheresultsforotherpercentiles.

ResourcesfortheFuture3

Figure1.TemperatureResponsetoSolarGeoengineering(SG)RelativetoEmissionsReductions

Left-handsubpanelsshowtheratiooftheresponseof(a)annualmeantemperature,(c)heatwaveintensity,and(e)coldwaveintensityperdegreeofcoolingfromSGrelativetotheresponseperdegreeofcoolingfromemissionsreductions.Displayedvaluesarethemedianover100climatesimulationyears.Blue(red)gridcellsindicateSGreducestemperaturesmore(less)thanemissionsreductions.Crosshatchesindicatestatisticalsignificanceat95percentconfidencelevelusingaWilcoxonsigned-ranktestcorrectedfollowingthefalsediscoveryrateprocedure.Right-handsubpanelsshowthezonalaverageoftheleft-handsubpanelsforthreeweightingschemes.

ImpactofSolarGeoengineeringonTemperature-AttributableMortality4

Figure1showstheregionaldistributionoftheestimatedratiometricsT,TTHw,andTTCw.Perhapsthemoststrikingcharacteristicisthatcomparedtoemissionsreductions,perdegreeofglobalmeancooling,SGcoolsequatorialregionsmoreandpolarregionsless.Thisdampeningoftheequator-to-polegradient(oranovercoolingofthetropicsandundercoolingofthepoles)isawell-documentedeffectofgloballyuniformSG(GovindasamyandCaldeira2000;Ban-WeissandCaldeira2010;Irvine2016).ItcanbemoderatedthroughnonuniformSGthatadjuststhelatitudinaldistributionofaerosolinjection(Kravitzetal.2019).Despiteconcernabouttropicalovercooling,uniformtemperaturereductionisobjectivelyneverthecorrectgoalforSG.Giventhatalargefractionoftheglobalpopulationlivesinequatorialregions,andthehealthandproductivityimpactsofadditionalwarmingarestrongestinhotregions,utilitarianorjusticeconcernsprovideanargumentforconcentratingcoolinginthetropics.

Foreachtemperaturemetric,wecalculatetheglobalpopulation-weightedmean,usingpopulationweightsbecauseweareconcernedaboutchangesrelevanttohumanmortality.Forannualaveragetemperatures,themedianglobalpopulation-weightedmeanTis1.087(95percentCI:1.066–1.11).Thisindicatesthatforequivalentaverageglobalcooling,SGcoolsannualmeantemperaturesbyaround8.7percentmorethanemissionsreductionsintheplacespeoplelive.Thisisdrivenbyanovercoolinginthelatitudinalbandsof30°Nto40°S,whichiswherealargefractionoftheglobalpopulationresides.

TheresponseofheatextremesTTHwis1.086(95percentCI:1.010–1.194),similartothatofannualmeantemperaturesbothregionallyandglobally.Thus,SGtendstoovercooltheheatextremesinplacespeoplelivemorethanemissionsreductionsdoes.Thisisconsistentwithanalysisofitseffectonheatwaves(DagonandSchrag2017).ThechangeforcoldextremesTTCwis1.045(95percentCI0.967–1.167).WedonotfindastatisticallysignificantdifferenceintheglobalmeanresponsetoSGandemissionsreductions.TherespectivezonalsubplotinFigure1showsthatSGleadstoarelativeovercoolingofcoldextremesinamuchsmallerlatitudinalbandrangeandundercoolsinmoreareasrelativetoannualmeansandheatextremes.Takentogether,wefindthatSGreducestheintra-annualvariabilityoftemperaturesinmostpopulatedregions(FigureS4).Astemperature-attributablemortalityisparticularlysensitivetoextremes,thisisanimportantmechanismthroughwhichSGmaydifferfromequivalentglobalcoolingfromemissionsreductions.

2.2.EmpiricallyEstimatedImpact

WeapplyempiricalestimatesofthehistoricalrelationshipbetweentemperatureandmortalitytoquantifythepotentialimpactofSGataglobalscale.Carletonetal.(2022)estimatetherelationshipbetweentemperatureandmortalityratesusingsubnationaldatafor40countriesandcaptureanonlinearexposure–responsefunctionwithheterogeneityacrossagegroups(<5,5–64,>65)andacrossregionsbasedontheirhistoricalclimateandincome(FigureED1).

ResourcesfortheFuture5

WefollowthemethodologyoutlinedinCarletonetal.(2022)toextrapolatetheirempiricalestimatesandprojecttemperature-attributablemortalityfor24,378regionsspanningtheglobe,eacharoundthesizeofaUScounty(seeMethods).Forourbenchmarkestimates,weassumeincomelevelsareconsistentwithSharedSocioeconomicPathway3(SSP3)in2015andthatpeopleareadaptedto1990sclimate.WeaccountforuncertaintythroughMonteCarlosimulation,samplingacrossclimatevariability,andstatisticaluncertainty(seeMethods).

Figures2a–cshowourestimateoftheimpactofSGonannualtemperature-attributablemortalityrisk.Weshowthedifferencebetweentemperature-attributablemortalityinthe2xCO2experimentandthe2xCO2+SGexperimentnormalizedbythechangeinglobalmeantemperatures.Theriskispooledacrossagegroupswithinregions.Globally,wefindthatSGreducestemperature-attributablemortalitybyanaverageof17deathsper100,000peryearper1°C.Forcontext,thisisaround2.1and1.4percentofthe2019andprojectedend-of-centuryglobalall-causemortalityrate,respectively(UN2022).FigureED2ashowsuncertaintyinthisestimateacrossMonteCarlosimulations;themajorityofitisdrivenbyuncertaintyintheeconometricestimates.

Figure2.ImpactofSolarGeoengineering(SG)onTemperature-Attributable

Mortality

Impactofsolargeoengineeringontemperature-attributablemortalityratesby(a)populationand(c)area.Differenceintemperature-attributablemortalityperdegreeofglobalmeancoolingfromemissionsreductionsandSGby(b)populationand(d)area.PlotsdisplaymedianestimateacrossMonteCarlosimulation;(e)fractionofglobalpopulationwithpositiveornegativeimpact.

ImpactofSolarGeoengineeringonTemperature-AttributableMortality6

Impactsacrossbothregionsandagegroupsareheterogeneous.Someregionsbenefitfromcooling,butothersareharmed.ManyregionsoftheGlobalSouthbenefit,butmortalityriskincreasesinmanyregionsoftheGlobalNorth.Forexample,perdegreeofcoolingwithSG,mortalityriskinBoston,USincreasesbyanaverageof3deathsper100,000peryearwhilemortalityriskinMumbai,Indiadecreasesbyanaverageof12deathsper100,000peryear.TakingthemedianacrossMonteCarlosimulations,wefindthat68percentoftheglobalpopulationbenefitfromareductioninmortalityriskwithsolargeoengineeringand32percentoftheglobalpopulationexperienceanincreaseinmortalityrisk.Thisbenefitisstrongestforthepoorestregions(FigureED3).Becausetheoldestagegroup(65+)isthemostsensitivetotemperature-attributablemortality(FigureED1),itisthemajorityofbothglobalandregionalimpacts,followedbythelowestagegroup(<5).

Figures2b–dshowthenormalizedimpactofSGontemperature-attributablemortalityrelativetothatofemissionsreductions;wecalculatetheglobalratiometricTMasthisratioandestimateitas1.12.ThisindicatesthatSGreducestemperature-attributablemortalitybyaround12percentmorethanemissionsreductionsperdegreeofglobalmeancooling.Thisratioisconsistentacrossagegroups,although,again,mostoftheleveldifferenceinriskisintheoldestagegroup.FigureED2bshowsuncertaintyinthisestimateacrossMonteCarlosimulations.

MuchoftheregionalheterogeneityinSG’simpactrelativetoemissionsreductionscanbeexplainedbytheheterogeneityinclimateresponseandwhetherregionsaremadebetterorworseoffwithglobalcooling.Thisismadeclearbydecomposingmortalityriskimpactsintoheat-andcold-attributablemortality.Generally,equatorialregionsthatbenefitfromcoolingarebetteroffwithSGbecauseitovercoolsthemrelativetoemissionsreductions(FigureED4a).IntheGlobalNorth,regionsaregenerallyworseoffwithcoolingfromSGbecauseitovercoolsthemrelativetoemissionsreductions(FigureED4b).Equatorialregionsseeanincreaseincold-attributablemortalityriskwithSG,butthebenefitsfromreducedheat-attributablemortalityriskarestronger.Fortemperature-attributablemortality,wefindthat80%oftheglobalpopulationbenefitmorefromcoolingwithSGwhile20percentbenefitmorefromcoolingwithemissionsreductions.

3.LimitationsandUncertainties

Ouranalysisisonlyascredibleasourassumptions.Webreakdownthelimitationsanduncertaintiesinouranalysisintothosestemmingfromtheempiricalestimatesandtheirapplicationandthosefromtheclimatesimulations.

3.1.EmpiricalEstimates

Section7ofCarletonetal.(2022)outlinesthelimitationsoftheirmethodologyindetail.Arguablythemostrelevantlimitationforoursettingisthatdry-bulbsurfacetemperatureistheironlyclimatevariableinput.Fromaphysiologicalperspective,itiswellknownthatfactorsotherthanambienttemperatures—suchashumidityandradiation—arealsoimportantdeterminantsofhumanhealth,particularlyforheatstress

ResourcesfortheFuture7

(BuzanandHuber2020).Thesefactorswillbecapturedinthemodeltotheextentthattheycorrelatewithdry-bulbtemperature,butthesecorrelationswillchangewithSGorclimatechange(Hardingetal.2020).Forexample,SGoptimizedtorestoretemperaturechangeswillstronglyreduceprecipitationchanges(Moreno-Cruzetal.2012).

SGcouldbelesseffectiveatreducingtemperature-attributablemortalitythanwefindhereif,forexample,changesinotherrelevantenvironmentalconditionsincreasedphysiologicalstressinawayweareunabletocapture.Lackinganempiricalmodeltoquantifytheseeffects,weanalyzetheclimateresponsetoSGforbothwet-bulbtemperature(WBT)andwet-bulbglobetemperature(WBGT)atthesurface,twocompositemeasuresofheatstressthatconsiderhumidity,windspeed,andradiativeflux(seeMethods).

Figure3.Wet-bulbandwet-bulbglobetemperatureresponsetosolargeoengineering(SG)relativetoemissionscuts

Theleft-handandmiddlecolumnsofsubpanelsshowther