Comparing
future
patterns
of
energy
system
change
in
2
C
scenarios
with
historically
observed
rates
of
change
Mariësse
A.E.
van
Sluisveld
a,b,*
,
J.H.M.
Harmsen
a,b,
Nico
Bauer
c,
David
L.
McCollum
d,
Keywan
Riahi
d,
Massimo
Tavoni
e,j,
Detlef
P.
van
Vuuren
a,b,
Charlie
Wilson
f,
Bob
van
der
Zwaan
g,h,i aCopernicusInstituteofSustainableDevelopment,UtrechtUniversity,Heidelberglaan2,NL-3584CSUtrecht,TheNetherlands
b
PBLNetherlandsEnvironmentAssessmentAgency,POBox303,3720BABilthoven,TheNetherlands
cPotsdamInstituteforClimateImpactResearch(PIK),POBox601203,14412Potsdam,Germany dInternationalInstituteforAppliedSystemsAnalysis(IIASA),Laxenburg,Austria
e
FondazioneEniEnricoMattei(FEEM),Milan,Italy
f
TyndallCentreforClimateChangeResearch,UniversityofEastAnglia,Norwich,UK
g
ECNPolicyStudies,EnergyresearchCentreoftheNetherlands,Amsterdam,TheNetherlands
h
UniversityofAmsterdam,FacultyofScience,Amsterdam,TheNetherlands
i
JohnsHopkinsUniversity,SchoolofAdvancedInternationalStudies,Bologna,Italy
jPolitecnicodiMilano,DepartmentofManagementandEconomics,Milan,Italy
ARTICLE INFO
Articlehistory: Received9March2015
Receivedinrevisedform14September2015 Accepted23September2015
Availableonline7November2015
Keywords:
Integratedassessmentmodeling Energysystemchange Technologicalchange Modelvalidation 2degrees Feasibility
ABSTRACT
Thispapersystematicallycomparesmodeledratesofchangeprovidedbyglobalintegratedassessment modelsaimingforthe2Cobjectivetohistoricallyobservedratesofchange.Suchacomparisoncan provide insights intothedifficultyof achieving suchstringent climatestabilizationscenarios. The analysisfocusesspecificallyontheratesofchangefortechnologyexpansionanddiffusion,emissionsand energysupplyinvestments.Theassociatedindicatorsvaryintermsofsystemfocus(technology-specific orenergysystemwide),temporalscale(timescaleorlifetime),spatialscale(regionalorglobal)and normalization(accountingforentiresystemgrowthornot).Althoughnoneoftheindicatorsprovide conclusiveinsightsastotheachievabilityofscenarios,thisstudyfindsthatindicatorsthatlookinto absolutechangeremainwithintherangeofhistoricalgrowthfrontiersforthenextdecade,butincrease tounprecedentedlevelsbeforemid-century.Indicatorsthattakeintoaccountornormalizeforoverall systemgrowthfindfuturechangetobebroadlywithinhistoricalranges.Thisisparticularlythecasefor monetary-basednormalizationmetricslikeGDPcomparedtoenergy-basednormalizationmetricslike primaryenergy.Byapplyingadiversesetofindicatorsalternative,complementaryinsightsintohow scenarioscomparewithhistoricalobservationsareacquiredbuttheydonotprovidefurtherinsightson thepossibilityofachievingratesofchangethatarebeyondcurrentdaypractice.
ã2015ElsevierLtd.Allrightsreserved.
1.Introduction
Keeping temperatureincrease toless than 2C witha high likelihoodwillrequiresubstantialchangesinenergyandlanduse. Integrated assessment model (IAM) studies on mitigation scenarios can provide insights into the level of the required change over time. IAM-based studies often conclude that the required transition for reaching the 2C target is ‘technically
feasible’,dependingonthemodelset-upandassumptions.Inthe past,suchstudiesgenerallyconsideredratheridealizedconditions suchasfullparticipationofallregionsandsectorsinclimatepolicy. However, more recently, models have also studied the achiev-ability of the 2C target under less idealized circumstances assuminglimitsintechnologyavailabilityorreducedparticipation ininternationalclimatepolicy(Clarkeetal.,2009;Kriegleretal.,
2013b;Riahietal.,2015;WeyantandKriegler,2014).Eveninthose
cases,mostmodelsstillidentifyscenariosthatreduceemissionsin linewiththe2Ctarget.Itshould,however,benotedthatintheir assessment,IAMsmostlyaccountfortechnologicalandeconomic factorsthatcan beeasilyincludedin themodels.Thesefactors include,forexample,mitigationpotentials,capitalstockturnover
*Correspondingauthorat:PBLNetherlandsEnvironmentAssessmentAgency,PO Box303,3720BABilthoven,TheNetherlands.
E-mailaddresses:mariesse.vansluisveld@pbl.nl,m.a.e.vansluisveld@gmail.com
(M.A.E. vanSluisveld).
http://dx.doi.org/10.1016/j.gloenvcha.2015.09.019
0959-3780/ã2015ElsevierLtd.Allrightsreserved.
ContentslistsavailableatScienceDirect
Global
Environmental
Change
rates,mitigationcostsandinertiaininvestmentspatterns.Several otherfactorsare notincluded suchastherole of international negotiations,societalinertiaorthetimeassociatedwith decision-makingprocessesontheonehandandbehavioralchangesonthe other.Clearly,suchfactorscanhaveasubstantialinfluenceonthe probable(future)rateofchange.
Historically observed rates of change can be important referencepointsforassessingthedifficultyassociatedwithfuture ratesof change–providing possiblyalsoinsightsin realworld factorsnot coveredin themodels. In fact,several studieshave alreadytriedtocomparemodelresultsandhistoricaldatausing differentindicators(KramerandHaigh,2009;Loftusetal.,2014;
Riahietal.,2015;TavoniandvanderZwaan,2009;vanderZwaan
etal.,2013;VanVuurenandStehfest,2013;Wilsonetal.,2012).In
thesestudiesdifferentmethodsanddatasetshavebeenusedto confrontexistingscenarioswithhistoricalevidence,meaningthat their results and conclusions cannot be easily compared. For instance,VanVuurenandStehfest,2013;Riahietal.,2015lookedat overallchangeinemissionsoremissionintensity.Incontrast,the studiesofvanderZwaanetal.andWilsonetal.lookatabsolute and relative changes in the deployment of particular energy technologies. It should also be noted that model comparison projects have shown that models select different pathways in achieving similargoals, and that models can be‘diagnosed’ as beingmoreorlessresponsivetoclimatepolicy(Kriegleretal.,2015
).Inordertorepresentmodeluncertainty,itisthereforeimportant to compare the results of a diverse set of IAMs against a standardizedsetofhistoricalindicators.
In this light, the goal of this study is (1) to systematically compareseveralmethodsthatusehistoricalevidenceasabasisfor analyzingthedifficultyassociatedwithfutureenergytransitions and(2)tousethesemethodsforevaluatingmodelresults.Weuse the resultsof a multi-model study toprovide insight into the uncertainty resulting from a wide diversity of technology
trajectories that are consistent with the 2C target. Questions thatareaddressedare:
-How dohistorical ratesofchangecomparetofuturerates of changerequiredunderthe2Cclimateobjective?
-Dovariousindicatorsoftechnologychangedepictacoherent storyline?
2.Methodology
2.1.Comparinghistoricalandfutureratesofchange
Historicalobservationsprovideanimportant referencepoint for therequiredlevelof efforttoachievefuture energy system changes associatedwithambitiousclimatepolicyobjectives.To date, differentindicators havebeen used tocompare historical trendswithfutureratesofchange,varyingintermsofsystemfocus (technology-specific or energy system wide), temporal scale (timescale or lifetime), spatial scale (regional or global) and normalization(accountingforentiresystemgrowth).Inorderto gainamoreholisticinsightfromtheseanalyseswecombineand harmonizethemethodstoencompassanoverallsimilarscopeof research. In the following paragraphs the various methods are describedfirstfollowedbyhowtheyareinterpretedinthecurrent study.Table1andTable2providesummariesofthemetricsused and scope of study. Fig. 1 provides a visual example of the introducedmethods.
2.1.1.Averageannualcapacityaddition
VanderZwaanetal.investigatedhistoricalandfuturecapacity growthbycomparingtheaverageannualcapacityadditions (in GW/yr)inamulti-modelcontextforlow-carbontechnologiesfor theshort-term(2010–2030)andmedium-term(2030–2050)(van
Table1
Overviewoftechnologychangeindicatorsincludedforstudy.
Indicator Variations Reference Metric
(a)Averageannualcapacityaddition Averageannualcapacityaddition Eq.(1) GW/yr Normalizedaverageannualcapacityaddition Eq.(2) GW/yr/$
(b)Technologydiffusion Normalizedextentandduration(Dt) Eq.(3) GW/EJ/yr (c)Averageannualemissiondeclinerates Averageannualemissiondeclinerate Eq.(4) %/yr
Normalizedaverageannualemissiondeclinerates Eq.(5) %/yr
(d)Averageannualsupply-sideinvestments Averageannualsupply-sideinvestments Eq.(6) $/yr Normalizedaverageannualsupply-side Eq.(7) %/yr
Table2
Overviewofmethodologiesandthescopeofthisstudy.
Indicator Systemfocus Temporalscale Spatialscale Normalization(Metric)b
(a)Averageannualcapacityaddition Technologyspecific Annuala Global No
Technologyspecific Annuala
Global Yes(GDP)
(b)Technologydiffusion Technologyspecific Lifetime Global Yes(PrimaryEnergy)c
(c)Averageannualemissiondeclinerate Energysystem Annuala Global/National No
Energysystem Annuala Global/National Yes(GDP)
(d)Averageannualsupply-sideinvestments Energysystem Annuala
Global No
Energysystem Annuala
Global Yes(GDP)
a
Onaverageoveraselectedperiodoftime.
b
ThisstudydepictsGDPthroughouttheresultsasthemeasureofgrowth,othermetricsofgrowtharefurtherdiscussedinSection4.2.
c
derZwaanetal.,2013).Thestudyfocusedontheabsoluterateof change required to reach the 2C target compared to rates experienced during historical periods of rapid expansion for established technologies (e.g., natural gas power) and newer technologies(e.g.,solar power). The comparison provided easy interpretableinsightsintotheexpansion rateforfuture deploy-mentversushistoricalfigurespublishedin literatureandonline databases(e.g.,EPIA,2014;Platt’s,2013;USEIA,2014).
The averagecapacity additionoveraselectedperiodof time (wheret0 andtnrepresentrespectivelythestartingand ending
pointofthetimeframeunderstudy)isdefinedas(seeEq.(1)): Averageannualcapacityaddition
¼ Xtn
t¼t0
newlyinstalledcapacity
ð Þt
tnt0 ðin
GW=yrÞ ð1Þ Usingthisapproach,vanderZwaanetal.(2013)concludedthat future global expansion rates need to increase significantly, reachingexpansionrateswellbeyondthoseobservedhistorically inordertostaybelowthe2Ctarget.Inparticulartheexpansionof renewableenergytechnologieswould needtobeseveral times largerthanthehistoricalrate.
Thecomparisonofabsolutefuturerateswithhistorical rates doesnotcorrectforthegeneralgrowthinthesizeoftheenergyor electricitysystem.Itispossibletoaccountfortheoverallgrowthby normalizing the absolute indicators with metrics representing totalsystemgrowthaspresentedinEq.(2).Normalizationmetrics thatcanbeusedtorepresenttotalsystemgrowthare,forexample, globalGDP(inT$),global primaryenergydemand(in EJ),total electricity generationcapacity(in GW)and totalcapital invest-ments in the energy system (in billion USD$). For reading considerations,wepredominantlyuseglobalGDPinthepresented outcomesinthisstudy,butdiscusstheothermetricsinSection4.2. Normalizedaverageannualcapacityaddition
¼ Xtn
tn¼t0
Newlyinstalledcapacity Normalizationmetric
t
tnt0 ðin
GW=metricunit=yearÞ ð2Þ AsimilaranalysishasbeendonebyLoftusetal.(2014),who normalizedelectricitycapacitydeploymentratesinvariousglobal decarbonizationscenariosusing globalGDP.Intheirstudythey found that the rates of change are broadly consistent with historical experience. Only specific decarbonization scenarios with imposed restrictions onthe implementationof clean and
Fig.1.Conceptualoverviewofthemethodologiesandkeyindicators.Panel(a)and(b)representcumulativecapacityofcoalwithoutCCS.Althoughthefiguredemonstrates future(modeled)trendstheanalysisissimilarforhistoricaltrends.
carbonsequestrationtechnologies wouldleadtounprecedented rates of change for the remaining eligable low-carbon energy technologies(Loftusetal.,2014).
2.1.2.Technologydiffusion
Technologygrowthdynamicsaregenerallycharacterizedby S-shapedcurvesthatshowaninitial‘formative’phase,followedby rapiddiffusioninan ‘upscaling’phase and finishintoa mature ‘growth’phase(Grübleretal.,1999;Wilson,2012).Growthrates varyoverthistechnologylifecycle,beginningslowlyuntilalift-off point is reached and growth accelerates. After some time, an inflectionpointispassedandgrowthratesleveloffandeventually saturate,reducinggrowthtozero.
Inthislight,Wilsonetal.(2012)comparedhistoricalandfuture dynamicsoftechnologydiffusionintheenergysystembyfitting logisticgrowth curves(withaR-squaredfitof98%orhigher)to cumulativecapacitytimeseries.Theadvantageofusing cumula-tivecapacityoverthetechnology’slifecycle,asopposedtoinstalled capacity orgrowth ratesduring particulartime periods,is that short-term volatility and potential selection biases towards specificperiodsofgrowthareavoided.
The(symmetrical) logistic functionused in this approach is portrayedbyEq.(3).Theparametersdefiningthelogisticgrowth curveareof particularinterest,as each parameterrepresents a specificgrowthcharacteristicusedinthiscomparisonapproach. Forexample,theparameterdefiningthesteepnessofthecurve represents the diffusion rate, whereas the parameter defining growthbetween10%to90%ofsaturationrepresentstheduration ofdiffusion(alsodepictedas
D
t)andtheparameterdescribingthe theoreticalasymptoterepresentstheextentofgrowthor satura-tionpointofatechnology(seeEq.(3),Fig.1andSupplementary information).Toaccountforthegrowingsizeoftheenergysystem,Wilsonetal.(2012)normalizedtheextentofdiffusionbythesize
oftheenergysystem(expressedinprimaryenergy)atthemidway pointofeachtechnology’slifecycle(theinflectionpointtm).The
normalizedextentand
D
tcreatetheextent-durationrelationship forthenumberoftechnologiesincluded.Technologydiffusion¼
Extent Normalizationmetrictm
ð1þediffusionrateDtÞ ð3Þ
The maindisadvantage of this methodology isthat it is not readily comparable to recent observations or to maximum or frontier growth rates over short time periods. Moreover, only historical and future technologies that expose S-curve growth behaviorsareincludedintheanalysis.Thisexcludes,forexample, windandsolarpowertechnologieswhichremaintohavearapid growthrateinanexpandingenergysystemandthereforedonot conformtoS-curvebehaviorwithinthetimehorizonofthemodel. The results from the methodology applied by Wilsonet al.
(2012) showed that the full lifecycles of advanced power
generation technologies as modeled in many future scenarios havelongerdurationsthanthefulllifecyclesofenergy technolo-gies that have diffused historically. In other words, there is evidencethatdeepdecarbonizationscenariosmaybesomewhat conservative in their long-term technology growth dynamics. However,theauthorsacknowledgedseveralcaveats,includingthe possibilitythatcomparinglong-runhistoricalgrowthwith long-runfuturegrowthinthiswayisproblematic.Thiswasspecifically thecasefortheanalysisofcoalornuclearpower,whichcombined historical and future growth dynamics in the logistic fitting procedure.
2.1.3.Averageannualemissiondeclinerate
An indicator often used to gain insight into economy-wide changesistheaverageannualemissionsdeclinerate(Riahietal.,
2015;VanVuurenandStehfest,2013).Wedefinethisindicatoras
giveninEq.(4).Similartotheannualcapacityadditions(described inSection2.1.1)weconsidertheaverageannualdeclinerateovera selectedperiodoftime(where‘Emissions’describethetotalCO2
emissionsand t0 and tnrepresentrespectively thestarting and
endingpointofthetimeframeunderstudy). Averageannualemissiondeclinerate
¼ 1 Emissionstn
Emissionst0
1
tn t0!
100ðin%=yrÞ ð4Þ Toaccountforsystemgrowthchangesthisstudyalsoconsiders thenormalizedversionoftheaverageannualemissiondeclinerate (which is also known as either the intensity decline rate or decarbonization rate ifGDP is consideredas thenormalization
Table3
Keymodelcharacteristics.
Name Timehorizon Modelcategory IntertemporalSolutionMethodology
IMAGE 2100 Partialequilibrium Recursivedynamic
MESSAGE 2100 Generalequilibrium Intertemporaloptimization REMIND 2100 Generalequilibrium Intertemporaloptimization TIAM-ECN 2100 Partialequilibrium Intertemporaloptimization WITCH 2100 Generalequilibrium Intertemporaloptimization
Table4
Overviewofselectedhistoricaltimeframesperindicatorandtheuseddatabases
Indicator Technology Historicalreference Source
(a)Averageannualcapacityaddition PV 2003–2013 EPIA(2014)
Wind 2003–2013 GWEC(2014)
Nuclear 1980–1990 Platt’s(2013)
Biomass 2005–2011 USEIA(2014)
Fossil 2003–2012 Platt’s(2013)
CCS – –
(b)Technologydiffusion PV 1970s Wilsonetal.,(2012)
Wind 1970s
Nuclear 1950s
Fossil Early1900s
(c)Averageannualemissiondeclinerate System 1970s–2000 Riahietal.,(2015)
metric)(seeEq.(5)).
Normalizedaverageannualemissiondeclinerate
¼ 1 Emissionstn NormalizationMetrictn Emissionst0 NormalizationMetrict0 0 B @ 1 C A 1 tn t0 0 B B @ 1 C C A 100ðin%=yrÞ ð5Þ The disadvantageofthis genericdescriptiveindicatoris that details on underlying drivers of emissions are not visible. Moreover,asemission reductionand emissionintensitydecline rateshavenotbeenmajorpolicygoalsinthemoredistantpast,a comparison against the long-term historical record can be regardedashavinglimitedrelevance.Nevertheless,thestudyby
VanVuuren and Stehfest(2013)used historicalcomparisons to
concludethatemissionreductionsaswellasdecarbonizationrates forscenariosconsistentwiththe2Ctargetcanberegardedas extremelyrapidcomparedtohistoricalratesofchange.
2.1.4.Averageannualsupply-sideinvestments
Structural changesin theenergysystem areassociated with increasing supply-side investments. As investments are also neededtoachieve othersocial andeconomic goals there could beconstraintsintherequiredpaceofchange.Therefore,welook intotheglobalinvestmentsintoelectricitygenerationandsupply (includingelectricity storageand transmissionanddistribution, butnotinvestmentsintothefossilfuelextractionsectornorthe bio-energy fuel supply costs) to assess the efforts needed to mobilizeanenergysystemtransformationthatisinlinewiththe 2C objective. Demand-side investments are not taken into accountassuchestimatesaresubjecttoconsiderableuncertainty duetoalackofreliablestatisticsanddefinitionalissues(McCollum etal.,2013).ThegeneralapproachisdescribedbyEq.(6)forwhich theaverageannualsupply-sideinvestmentsarecalculated(where t0andtnrepresentrespectivelythestartingandendingpointofthe
timeframeunderstudy):
Averageannualsupplysideinvestments ¼
Xtn
t¼t0
supplysideinvestments
ð Þt
tnt0 ðin
$=yrÞ ð6Þ Asthetotalamountofinvestmentsiscoupledtothesizeofthe economy,wenormalizetheannualsupply-sideinvestments(see Eq.(7)).IfglobalGDPisconsideredasthenormalizationmetricthis createsanindicatorreflectinginvestmentsaspercentageintotal GDP.
Normalizedaverageannualsupplysideinvestments ¼
Xtn
t¼t0
supplysideinvestments Normalizationmetric
t
tnt0 ðin
$=metricunit=yearÞÞ ð7Þ
McCollumetal.(2013)examinedabsoluteratesofchangefor
investments in more detail, concluding that future investment levelsremainconsistentontheshorttermalthough significant increases in investments in both developed and developing
countrieswillbenecessaryoverthenextdecadesunderthe2C objective.
2.2.Comparingfuturetechnologicalchangetohistoricalreferences 2.2.1.Futureratesofchange
Wedemonstratetheindicatorsbyusingthreescenariosfroma five-modelstudywithvaryingassumptionsonlong-term interna-tional climate policy. A marked advantageof the multi-model approachisthatitinherentlyaccountsfortechnologybiasesand preferencesamongindividualmodels.Thestudyhere,however,is notamodelcomparison:weonlyincludethemodelrangeasan indicationoftheuncertaintyinmodelresults.Wethereforedonot discusstheresultsofindividualmodelsinanydetail.Thefocusin thefiguresisalsoonthemedianoftherangeofmodelresults.
The fiveglobal energy-environment modelsincludedin this study are: REMIND:(Bauer et al., 2013; Luderer et al., 2013); MESSAGE: (Messner and Strubegger,1995);IMAGE: (Bouwman etal.,2006);WITCH:(Bosettietal.,2006)andTIAM-ECN:(Keppo
and van der Zwaan, 2011)) (see Table 3). These five models
representadiversearrayofdifferentsolutionframeworks(general equilibrium, partial equilibrium, dynamic recursive, perfect foresight and systems engineering) and differ in a variety of model characteristics, such as coverage of sectors and their disaggregationandintechnologicalandsocio-economic assump-tionsthatdeterminetechnologydiffusion.
Thethreescenarios thatareusedin thisstudyarebasedon differentpolicyassumptionsforlong-terminternationalclimate policy and have been developed as part of the LIMITS project
(Kriegleretal.,2013a).
(1) Thebaseline(Baseline)scenarioaddressesthefutureenergy systemandemissiondevelopmentsintheabsenceofclimate policy.Thisscenarioisabestreferenceforhistoricalratesof changeasnoclimatepolicyisinvolved.
(2) The second (Reference) scenario reflects current (unilateral) climatepolicyimplementationbasedonnationalenergyand climatetargetsfor2020formulatedasunconditional Copen-hagenpledges.Thesetargetsarethenextrapolatedpost-2020 byassuming similarlevels of stringency in the subsequent decades.Thisscenariorepresentsthecurrentdaysituationand imposesnoadditional(technological)restrictions.
(3) The third (2 Degrees) scenario is a cost-optimal mitigation scenariothatassumesimmediateglobalcooperationtoward thelong-termtargetof2C.Thisscenariorepresentsthemost optimistic view on technology availability, availability of carbon sinksand (bio-) resourcestoattain the2C climate
target.
Differencesarecreatedduetothevaryingassumptionson long-terminternational climatepolicy, allother factors,suchas the penetrationandexpansionratesoftechnologies,aretreatedthe sameacrossallscenarios.
The methods and indicators set out in Section 2.1 are comparativelyappliedonthis setof threescenarios. As timing
Table5
Overviewofnormalizationmetrics,availablehistoricaltimeframeandsource.
Method Metric (Historical)timeframe Source
Normalization GDP 1980–2012 TheWorldBank(2015)
PrimaryEnergy 1980–2012 USEIA(2014)
Investments 2000–2013 IEA(2014)
ofchangeisimportantthisstudyhasrestrictedtheanalysistothe timeperiodbetween2010and2050becauseitisconsideredmost relevantforcurrentpolicyanddecisionmaking.
2.2.2.Historicalreferences
For the average annual capacity addition indicator, we reconstruct a similar analysis tothat of van der Zwaan et al.
(2013)bycomparingmodeledaverageannualratesofchangein
Fig.2.Averageannualcapacityadditions(overthe2010–2030and2030–2050period)forvariouselectricity-generationtechnologiesunderdifferentclimatepolicy assumptions.Thehorizontallinesindicatethetechnology-specificpeakormaximumvalueobservedhistorically(solidlines)andthepeakvalueacrossalltechnologieswhich isgivenbycoalwithoutCCS(dottedlines).Thegreen,blueandredareasindicatewhetherahistoricalbenchmarkhasbeenexceeded(redforall-technologypeak,bluefor technology-specificpeak)ornot(green).Thebarsindicatetherangeofmodeledratesofchangewiththemedianvaluehighlightedinblackinsidethebars.(Forinterpretation ofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)
totalnewinstalledcapacitytohistoricalaverageannualratesof change. Several databases provided historical data on various technologies(seeTable4)ofwhich thedecadewiththelargest absolutegrowthincapacityisselectedforfurtheranalysis.
Forthetechnologydiffusionindicator,similarlogisticgrowth curvesareconstructedasinWilsonetal.(2012)onbothhistorical
(if applicable)and future timeseries. Thehistorical time series beginasfarbackastheearly1900s(naturalgasandcoalpower), the1950s(nuclearpower),the1970s(windpowerandsolarPV),or startnosoonerthanthe2020sorlater(CCS—thusfullybasedon modeleddataonly).
Fig.3. Averageannualcapacityadditions(GW/yr),normalizedusingGDP(intrillionUS$2005)forbothhistoricaldataaswellasscenarioprojections.Thehorizontallines indicatethetechnologyspecificmaximumvalueandthemaximumvalueofanytechnologyinthepast.Thegreen,blueandredareasindicatewhetherahistoricalbenchmark hasbeenexceeded(greenbelowtechnologyspecificrate;blueabovetechnologyspecificrate;redabovethehistoricalrateofanytechnology).Thebarsindicatetherangeof modeledratesofchangewiththemedianvaluehighlightedinblackinsidethebars.(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredto thewebversionofthisarticle).
For the average annual emission decline rate indicator, we depictaverageCO2emissionandCO2intensityreductionratesand
comparethemtohistoricalnationaleventsthatledtoemission (intensity) reductions (such as oil crises, collapse of political regime)(Riahietal.,2015;VanVuurenandStehfest,2013).
Fortheaverageannualsupply-sideinvestmentsindicator,we showtheaverageannualinvestmentsortheshareinGDPoverthe 2010–2050 timeframe and compare them to the historical investments (or share in GDP) over the 2000–2013 timeframe (IEA,2014).
In order to normalize the absolute indicators, taking into account relative changes in the size of the energy system or economy, we use GDP, primary energy, total energy system investments and total capacity as normalization metrics. The historicalperiodtakenintoconsiderationisthe1980–2012period asmostmetricshaveannualdataavailableinpublicsourceswith investments as an exception (see Table 5). The analysis will predominantlyfocusonglobalGDPasthemain systemgrowth factor,othermetricswillbediscussedinSection4.2.
3.Results
In the results below, we show the results of each of the indicatorspresentedinSection2forthethreeLIMITSscenarios andall5modelsaswellasthehistoricalreferenceperiod. 3.1.Averageannualcapacityaddition
Themodeledannualcapacityadditions(inGW)forthe2010– 2030periodareonaverageconsistentwiththehistoricalreference acrossallthreescenarios.IntheBaselinescenario,theexpansion rates from 2010–2030 are broadly consistent with historical observations(seeFig.2).CoalwithoutCCSmaintainsaconstant annualexpansionratewhereasgaswithoutCCSwillnearlydouble itscurrentannualexpansionrate,matchingandovertakingcoal withoutCCS.Underclimateconstraints,wefindashiftawayfrom fossilfuelseithershiftingtoalesscarbon-intensesubstitute(gas) orshiftingtonon-fossilresources.ForsolarPV,windandbiomass theexpansionratesstaywithinhistoricalpeakobservations.The projectionsofnuclearpowercapacitygrowtharealsoconsistent with historically observed expansion rates. However, currently plannedadditionalnuclearcapacitybetween2015–2019(World
Nuclear Association, 2014)indicatesthat the expansion rate of
nuclearenergy willmost likelynot exceed the3GW/yr. Hence,
given the long inertia in nuclear power plant planning and constructionprocess,theactualexpansionratesofnuclearpower mightcontinuetobebelowthedeploymentratesasdepictedin someofthehighscenarios(Table6).
In the 2030–2050 timeframe, the modeled rates of annual capacityadditionsincreasebeyondtechnology-specificexpansion ratesobservedhistorically.Someevenventureintoterritorythat goesbeyondoverallbestsystemachievementfromthepast.Under Baselineassumptions,thisisthecaseforbothcoalandgaswithout CCS,whichwillexpandtheirgrowthtounprecedentedlevelsas fossilfuelsremainthefuelofchoice.Underthe2Cobjective(2 Degrees), it will bethegrowthof solar and wind capacity that becomesparticularlyrapid,showingdeploymentratesabovethe historicalpeakvalueofoverallsystemachievements.
The outcomes change if overall system growth between historical and modeled periods is taken into account (by normalizing the average annual capacity indicatorusing global GDP growth). On the short-term the modeled average annual capacity additions showto remainconsistentwith technology-specificexpansionratesofthepast(seeFig.3).However,although sometechnologies(windandsolarinparticular)willexceedtheir technology-specific historical reference on the mid-term, all remain in linewith the overall bestsystem achievementfrom thepast.
3.2.Technologydiffusion
Iftheextent-durationrelationshipsforallelectricitygeneration technologiesandscenariosareassessed(seeFig.4)wefindthatall technologies followthehistoricallyobservedpatterns.However, under Baseline assumptions the diffusion durations (
D
t) are generallylonger(furthertotheright)andaneventualsaturation point(extent)isreachedbeyondthetimehorizonofthemodels involved(presentedasadurationthatisbiggerthana100yearsinFig.4).
Onceclimatepoliciesareintroduced(e.g.,theReferenceand2 Degreesscenarios),theextent-durationrelationshipschange.All technologiesshowtoshifttotheleft(shorterdiffusiondurations). ForfossilwithoutCCStechnologiesthisimpliesalowercapacity saturationlevel,ashorterlifecycle,andsomecapacityreductionin the year in which maximum growth is achieved (see the supplementary material online). For clean technologies (fossil withCCS,CO2neutralandrenewableenergytechnologies)onthe
otherhand,greaterextentsofgrowthareachievedwithshorter
Baseline Reference 2 Degrees
1e+00 1e+01 1e+02 1e+03 1e+04 1e+05 1e+06 50 100 50 100 50 100
Duration (years from 10% to 90% saturation)
Nor m aliz e d e x tent (MW/EJ)
Historical Fossil Fossil + CCS CO2Neutral Renewables
Fig.4.Capacitygrowthofenergytechnologiesin3futurescenariosofthe21stcentury:extentvs.durationofgrowthusingfittedlogisticfunctionparameters.Blackdots representhistoricalextent-durationrelationshipsofvariousenergy-supplytechnologies(suchasnuclear,coalandgaswithoutCCS,hydroandrefineries(FCC).
diffusion durations. However, despite the shorter diffusion duration, the rates remain above the historically observed reference.In thatsensethis studyisinagreement withWilson
etal.(2012)concludingthatthemodeleddiffusionratesappearto
beconservativecomparedtohistoricallysuccessfultechnologies.
3.3.Averageannualemissiondeclinerate
Fig.5showstheaverageannualemissiondeclinerateandthe decline rate normalizedusing GDP(creating a carbon intensity decline rate or decarbonization rate). Up till today, only rare historicaloccurrencesonanationallevelhaveledtosignificantly higherreductionratesthantheglobalaverage,whichhavebeen negative(0,8%per yearonaveragethroughoutthe1970–2010 period)owingtocontinuouslygrowingemissionsworldwide.For example,fairlyswiftemissionreductionrateswereobservedin Sweden from 1974 to 2000 as a result of policy impulses on greeningtheSwedishenergysystemaftertheoilcrisisin1973(2– 3%peryear).Anotherexampleistheemissiondeclinerateof2–4%
peryearforEasternEuropeanandformerSovietUnioncountries afterthecollapseoftheSovietUnion(Riahietal.,2015).Tostayin linewiththe2C objectivea sustainedglobal carbonemission reductionrateofabout1%till2030isrequired,remainingwithin theearlierdiscussedregionalhistoricalboundaries.However,after 2030 the models depict a sustained global carbon emission reductionrateof5%whichgoesbeyondbothglobalandregional historicalachievements.
Similarly, the global decarbonization rate (average annual emissiondeclineratenormalizedwithGDP)hasbeenaround0.5% over theperiod1900–2010and around1%over the1970–2010 (drivenbytechnologicalchangeandsectoralshifts)(VanVuuren
andStehfest,2013).Ifcomparedtothemodeleddecarbonization
rates,wefindrangesof2-3%underReferencescenarioassumptions whereas themarginsexpand to6–10%by 2050if 2C is tobe attainedattheend ofthecentury. Theseratesareconsiderably higherthantheglobalaveragerateexperiencedinthepast.Atthe regionallevel,historicallyfasterratescanbeobservedthanthe global average: some Asian regions have managed to achieve
Fig.5.Averageannualemissionsdeclinerates(top)andnormalizedaverageannualemissiondeclinerates(bottom).Negativenumbersindicateemissionsincrease.Green areaimpliesconsistencywithhistoricalevidenceforglobalrates,bluerepresentsvalueswithinhistoricalboundsofthefastestregionalreductionaddressedandredimplies beyondhistoricalreferenceforeitherconsideredspatialscale.Thebarsindicatetherangeofmodeledratesofchangewiththemedianvaluehighlightedinblackinsidethe bars.(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle).
decarbonizationratesof3–5%peryearduringthelate1980sand early1990s.Thiswouldimplythattheglobalratewouldneedto increasesignificantly, butalsogobeyond themostrapid(local) decarbonizationrateexperiencedintherecentpastandmaintain thisrate(globally)forseveraldecades.
3.4.Averageannualsupply-sideinvestments
Rapid transitions in the energy system are associated with increasinginvestmentflowscomparedtothestatusquo,whichis reflectedinFig.6.Bothcurrentclimatepolicy(Reference)aswellas the2Cpathway(2Degrees)wouldrequiregreaterinvestments than the business-as-usual case (Baseline), climbing up on the shortterm toabout 1.5 trillion USD per year which is slightly greater than observed historically. Under 2C ambitions these
investmentlevelsaremodeledtonearlydoubleforthesubsequent decade, increasing up to 2.5 trillion USD per year on average. Upscalinginvestmentstotheselevelsmightposeseveraldif ficul-tiesas two-thirdof thetotalsumisleviedbydevelopingareas
(McCollumetal.,2013)whichrequirefinancemechanismsother
thantheirowndomesticfunds(Bowenetal.,2014).
If total supply-sideinvestments are expressedas a share in globalGDP,itshowsthattheratioremainswithintheboundsof historical experience. However, by looking into global rates it potentially masks the large differences between regions. The averageinvestmentintensityofdevelopingeconomieswasaround
3.5%, whereas it was just 1.3% in industrialized countries (McColumnetal.,2014).
4.Discussion
4.1.Comparativeoverviewofindicatorsandresults
This study uses a diverse set of indicators that assess the consistencyofmodeledfutureratesofchangewiththehistorical record.Thestudyyieldsambiguousinsightsintotheconsistencyof modeledratesofchangewithhistoricalobservations(seeTable6). Absoluteandnear-term(2010–2030)ratesofchangevaryintheir consistencywithhistoricalobservationsforthethreescenarios, although these are mostly within the range of overall system achievements(blueshadedareasonthegraphs).Bynormalizing the indicators to account for system growth shows an overall consistency with historical records. Over the longer term the
Fig.6.Averageannualsupply-sideinvestments(top)andaverageannualsupply-sideinvestmentsinGDP(bottom).Barsrepresenttherangeofmodeloutcomesof respectivelyBaseline,Referenceand2Degrees.Thebarsindicatetherangeofmodeledratesofchangewiththemedianvaluehighlightedinblackinsidethebars.1
1
FortheBaselinescenario,thenumbersarerecalculated,asthey werenot includedinthestudyofMcCollumetal.(2013).Duetodataavailability,onlyresults forIMAGEandMESSAGEareshownhere.The2Degreesscenarioincludesunilateral climatepolicytargetstill2020,suspendingimmediateglobalaction,andtherefore deviatingfromthe2Degreesscenarioaspresentedinothergraphs.AstheReference and2Degreesscenariostarttodeviateonlyafter2020thetimeperiodsareamended to2020–2035and2035–2050.Thehistoricalobservationconsistsofcumulative energysupplyinvestmentsandcumulativetotalGDPfrom2000–2013.
indicators create a near similar picture for the Baseline and Reference scenarios. However various significant differences emergeunderthe2-degreeobjective(2 Degrees),specifically in terms of (absolute) average annual capacity expansion rates, (absolute)averageannualenergy-supplyinvestmentsand (abso-luteandnormalized)averageannualemissiondeclinerates. 4.2.Methodologicaldiversityandissues
Theindicatorsusedvaryinfocusandscope.Inthissectionwe furtherdiscusstheinfluencesandsensitivitiesofthestudydesign ontheoutcome.
(1)Systemfocus:Modelsareinherently limitedin their repre-sentation of energy-economy dynamics, and are highly dependent on their technological resolution (number of technologies included), underlying assumptions (on e.g., capital replacement or learning rates) as well as model structureandsolutionframeworks.Inthatrespect technolo-gy-specificindicatorsarepotentiallymoresensitivetospecific model behaviorthan system-wideindicators.However,in a multi-model set-upthese sensitivitiesare more-or-less bal-anced out and in that case, as depictedin Table 6, system indicatorsarenotconsistentlymoreorless likelytoremain consistent with historical observations than technology-specificones;
(2)Temporalscale:Indicatorsthatfocusonaspecifictimeframe (e.g.,theaverageannualcapacityadditionsoraverageannual emissiondeclinerates)canbesensitivetotheselectedtime period under study. This is especially the case if rapid expansion ordeclines ratesare nestedin certainperiodsof time,whichcanbeeitherhighlightedornumbeddowninthe longer-termaverage.
Focusing onthefull technologylifecycle,however, canalso influence theresults. For example, theWilsonet al. (2012)
methodologyissensitivetotechnologyprojectionswithaclear logisticgrowthprofile,suchasmaturehistoricaltechnologies
forwhich long time-seriesdataare available. Asrenewable technologiesaregenerallystillintheirearlydeploymentphase theseare not expectedtosaturate in thetimeframe of the modelandwillthereforenotappearaslogisticgrowthprofiles intheWilsonetal.methodology.Hence,somemodeledratesof change will not find application in the extent-duration analyses. The conservatism in the extent-duration curves could thus be an outcome of the overrepresentation of incumbenttechnologies;
(3) Spatialscale:Byfocusingonglobaloutcomesanindicatormay potentiallymaskthelargedifferencesbetweenregions.Inthis lighttheindicatorprovidesonlylimitedinsightsintotheactual challengesthatarefacedtoreachsuchratesofchange. Inthecasewhereaglobalbenchmarkisabsent(suchisthecase for(normalized)emissiondeclinerates)weselecteda more local(contemporary)achievedpeakvalue.Suchacomparison inherentlyincludesselectionbiasasfrontierreduction rates have specifically been selected. However although these regionalbenchmarks only lasted fora shortperiod of time andemergedunderrarecircumstances(suchasoilcrisesand regimechanges),thesespecificallyunderlinethedifficultyof achievingtheneededratesofchange;
(4) Normalization:Thenormalizationapproachisvisibly sensi-tive tothetype of system growth metric used(see Fig. 7). Monetary-basednormalizationmetrics(GDP,investmentsand capacitytosomedegreeaswell)resultinmoreconservative rates of change than energy-based normalization metrics (primary energy). As a result, rates of change that are normalizedby usingmonetary-basednormalization metrics arelesslikelytoexceedhistoricalratesthanthosenormalized using energy-based metrics. This is in particular true for indicators that experience rapid rates of change (for both technology-specifyandsystem-focusindicators).
Choosingtheappropriatenormalizationmetricisimportant– asthechoiceforaspecificmetriccouldrenderfutureratesof change(in)consistentwithhistoricalrates.Thechoicedepends accordingtotheauthorson(a)thevariablebeingnormalized,
Table6
Summaryofcomparisonsbetweenhistoricalobservationsandthreemodeledscenariosusingadiversesetofindicators.Thefossilandnon-fossiltechnologiesaregrouped— thetableconsidersthehighestrateofchangeinthegroupperscenario.
and(b)thequestionbeingasked.Forexample,ifthemodeled variableisannualcapacityadditions,then(a)suggestsusing historical primary energy or capacity as the normalization metric,unless(b)thespecificquestioniswhetherinvestment requirements in new capacity are in line with historical observations.
Insum,theresultsoftheindicatorsdiscussedinthisstudyare associatedwithseveralmethodologicalconsiderations.Applyinga widesetofindicatorsthereforeoffersalternative,complementary insightsintohowscenarioscomparewithhistoricalobservations ontwodifferentscales(e.g.,technology-specificandsystem-wide indicatorsandthechoicefornormalization).Althoughnoneofthe indicatorsprovide conclusiveinsights astothe achievabilityof
scenariostheyareusefulwaystocontributetoscenarioevaluation andprovokecriticalinterpretationofresults.
4.3.Expandingthescopeofresearch
By applying a diverse set of indicators one can gain more holistic insights into how scenarios compare with historical observations.Furtherresearchinlinewiththisstudycouldfocus on:
(1) Fine-tuningandextendingthescopeofcurrentindicators: Twofundamentalregularitiesofsuccessfultechnology diffu-sion patterns are described in Kramer and Haigh (2009). Accordingtotheirstudy,thebuildrateof newand existing energytechnologiesfollowtwo‘laws’whichhavebeenfairly
Fig.7.Deviationofthemedianmodelvaluefromthemaximumpeakbenchmarkperindicatorforeachconsiderednormalizationmetric.Positivevaluesindicatethatthe indicatorexceededhistoricalexperiencewhereasnegativevaluesimplyconsistencywithhistoricalobservations.Forplottingconveniencetheannualcapacityadditionsare limitedtonuclear,solarPVandwindtechnologies.Moreover,theinvestmentsindicatorisplottedonthe2010–2030and2030–2050timeframebuttheserepresentthe timeframesasdepictedinSection3.4.Thepicturefocusesonthe2Cobjective.
consistentacrossenergytechnologiesinthepast.Thefirstlaw describes how technologies grow quickly for the first two decades atexponential rates(26%/yr)until ‘materiality’ is reached,definedasa1%shareoftheglobalenergysystem. Thesecondlawstatesthataftermateriality,growthrateslevel downtoan eventual equilibriumor constant marketshare. Althoughtheexpansionphaseandthematuringgrowthphase characterizedbyWilsonetal.(2012)broadlycorrespondwith these‘fundamentallaws’,thiscouldbeembeddedmoreclearly within the historical comparison methods. Moreover, addi-tionalinsightsmayalsobeacquiredbydistinguishingbetween expanding systems (adding new capacity) and stabilizing systems(substitutingexistingcapacity);
(2)Introducingadditionalcomparisonmethods:Modeledrates ofchangecouldbecomparedagainstactualtrendsoverthe sameperiodoftime,forinstanceadecadeaftertheoriginal projectionwasmade.Anexampleofsuchanexerciseisfound
in Van Vuuren and O’Neill (2006). If short-term model
trajectories are significantly inconsistent with historical trajectories, it could expose conservatism in the long-term scenariologicandtheassumptionsonthedrivingforces.This methodology is, however, only useful if historical trends include similar climate policies as included in the model projections;
(3)Including demand-side indicators: Historical and future emissionsandtheirdrivingforceshavealsopreviouslybeen studiedbyapplyingtheKaya-identity(Kaya,1990).The Kaya-decomposition analysisis appliedin numerousstudies (i.e.,
Steckel et al., 2011; Zhang et al., 2009) to examine the
implications ofchanges in totalCO2 emissionsonaffluence
(representinggrowthofeconomicactivities),changeinenergy intensity (i.e., total primary energy over GPD reflecting efficiencyandconsumptionpatterns)andthecarbonintensity (i.e.,totalCO2emissionsovertotalprimaryenergy).Thethree
components could be assessed in tandem or as separate indicators in comparative work of prospective studies and historical records. This study has a greater energy supply orientation as all indicators focus on either energy supply technologies, investments orthe carbonintensityof energy supply but future work could also include demand side indicatorssuchasenergyintensityandaffluence;
(4)Goingbeyondthehistoricalbenchmark:Thisstudyconsiders historyasanimportantbenchmark,thoughhistoryprovides only limited information when looking at innovation. For example the results provide no further information about, amongst others, the drivers of technological change, (per-ceived) risks,scalability,structure of theindustryorrole of institutions.Expertelicitationcouldexpandtheknowledgeon criticalimplementationbarriersandfurthertestthefeasibility ofprospectivestudies.Severalprospectivestudieson technol-ogydevelopmentuseexpertelicitationprotocolsasaresearch tooltoassessthefeasibilityofemerging(carbon-free)energy technologies(seeforexampleBosettietal.,2012;Jennietal.,
2013;Fioreseetal.,2014).Expertscangobeyondthehistorical
benchmark by providing probabilistic information on the likelihoodthattechnologieswillovercomeparticularhurdles and estimate the overall probability of success for each technology(Bakeretal.,2009).
5.Conclusions
Inthisstudywehavecomparedindicatorsofchangeinfuture scenariostohistoricaltrendsforvariousdegreesofclimatepolicy. Theanalysisconfronts scenariodatafromtheLIMITSproject to fourmethodologiesthatfocusondifferentindicatorsoftechnology
change,suchastheaverageannualcapacityadditions,technology diffusionandchangesinemissiontrendsorinvestments.Themain conclusionsofthisanalysisare:
Theachievabilityoffutureratesofchangedependsonthe indicatorused
Inthisstudy,weassessedavarietyofindicatorstolookatthe rateoffuturechangeversushistoricallyachievedratesofchange. Thiscomparisonprovidessomeinsightintotheeffortinvolvedin achievingthesescenariosbutishighlydependenton:(1)selecting thehistoricalbenchmark,(2)normalization,(3)dataavailabilityas wellasthe(4) underlyingeconomic andtechnological assump-tions, model structures and the includedlevel of technological detail in the models. Although none of the indicators provide conclusiveinsights astotheachievability ofscenarios they are useful ways to contribute to scenario evaluation and provoke criticalinterpretationofresults.
Indicators highlight that absolute rates of change in scenariosachievingthe2degreetargetarerapidinthemedium termcomparedtohistoricallyachievedratesofchange
Inabsolutetermswehaveobservedthatprojectionsare more-or-lessinlinewithreportedachievementsontheshort-term,but theseincrease tounprecedentedlevelsbymid-century. Speci fi-callytheaverageannualcapacityadditionratesforsolarandwind and the required energy-supply investments are particularly strongunder2Cconstraints,showingratesabovethehistorical
peakvalueofoverallsystemachievementsby2030.
Methodsthatlookatrelativeratesofchangebycomparing thechangetooverallsystemchangeconcludethatfuturerates of change are generally within the range of successful transitionsinthepast
Indicatorsthat accountfor thegrowthintheoverallsystem showthatthemodeledratesofchangeinthescenariosarelower compared to the rates of change in the past. We find that monetary-basednormalizationmetrics(GDP,investmentsandto somedegreecapacity)resultinlessconservativenormalization thanenergy-basednormalizationmetrics(primaryenergy).Thisis in particular true for indicators that experience rapid rates of change(forbothtechnology-specificandsystem-focusindicators). Acknowledgements
Theresearchleadingtotheseresultshasreceivedfundingfrom theEuropean UnionSeventhFrameworkProgramme FP7/2007-2013undergrantagreementno.282846(LIMITS)andno.603942 (PATHWAYS). The authors are indebted to Peter Kolp at the International Institutefor Applied Systems Analysis (IIASA)for providinginvaluabletechnicalsupport.
AppendixA.Supplementarydata
Supplementarydataassociatedwiththisarticlecanbefound, in the online version, at http://dx.doi.org/10.1016/j.gloenvcha.
2015.09.019.
References
Baker,E.,Chon,H.,Keisler,J.,2009.AdvancedsolarR&D:Combiningeconomic analysiswithexpertelicitationstoinformclimatepolicy.EnergyEcon.31,S37– S49.doi:http://dx.doi.org/10.1016/j.eneco.2007.10.008.
Bauer,N.,Mouratiadou,I.,Luderer,G.,Baumstark,L.,Brecha,R.J.,Edenhofer,O., Kriegler,E.,2013.2013.Globalfossilenergymarketsandclimatechange mitigation—ananalysiswithREMIND.Clim.Change doi:http://dx.doi.org/ 10.1007/s10584-013-0901-6.
Bosetti,V.,Carraro,C.,Galeotti,M.,Massetti,E.,Tavoni,M.,2006.WITCH:AWorld InducedTechnicalChangeHybridModel.EnergyJ.SpecialIssue:Hybrid ModelingofEnergy-EnvironmentPolicies:ReconcilingBottom-upand Top-down.
Bosetti,V.,Catenacci,M.,Fiorese,G.,Verdolini,E.,2012.ThefutureprospectofPV andCSPsolartechnologies:Anexpertelicitationsurvey.EnergyPolicy49, 308–317.doi:http://dx.doi.org/10.1016/j.enpol.2012.06.024.
Bouwman,A.F.,Kram,T.,KleinGoldewijk,K.,2006.IntegratedModellingofGlobal EnvironmentalChange.AnOverviewofIMAGE2.4.NetherlandsEnvironmental AssessmentAgency,Bilthoven,TheNetherlands.www.mnp.nlen.
Bowen,A.,Campiglio,E.,Tavoni,M.,2014.Amacroeconomicperspectiveonclimate changemitigation:meetingthefinancingchallenge.Clim.Chang.Econ.05, 1440005.doi:http://dx.doi.org/10.1142/S2010007814400053.
Clarke,L.,Edmonds,J.,Krey,V.,Richels,R.,Rose,S.,Tavoni,M.,2009.International climatepolicyarchitectures:OverviewoftheEMF22InternationalScenarios. EnergyEcon.31,S64–S81.doi:http://dx.doi.org/10.1016/j.eneco.2009.10.013. EPIA(EuropeanPhotovoltaicIndustryAssociation)(2014)GlobalMarketOutlook
forPhotovoltaics2014–2018.http://www.cleanenergybusinesscouncil.com/ site/resources/files/reports/
EPIA_Global_Market_Outlook_for_Photovoltaics_2014-2018_-_Medium_Res. pdf.
Fiorese,G.,Catenacci,M.,Bosetti,V.,Verdolini,E.,2014.Thepowerofbiomass: expertsdisclosethepotentialforsuccessofbioenergytechnologies.Energy Policy65,94–114.doi:http://dx.doi.org/10.1016/j.enpol.2013.10.015. GWEC(GlobalWindEnergyCouncil)(2014)GlobalWindStatistics2013.http://
www.gwec.net/wp-content/uploads/2014/02/GWEC-PRstats-2013_EN.pdf.
Grübler,A.,Nakicenovic,N.,Victor,D.G.,1999.Dynamicsofenergytechnologiesand globalchange.EnergyPolicy27,247–280. doi:http://dx.doi.org/10.1016/S0301-4215(98)00067-6.
IEA(InternationalEnergyAgency)(2014)WorldEnergyInvestmentOutlook2014 SpecialReporthttps://www.iea.org/publications/freepublications/publication/ WEIO2014.pdf.
Jenni,K.E.,Baker,E.D.,Nemet,G.F.,2013.Expertelicitationsofenergypenaltiesfor carboncapturetechnologies.Int.J.GreenhouseGasControl12,136–145.doi: http://dx.doi.org/10.1016/j.ijggc.2012.11.022.
Kaya,Y.,1990.ImpactofcarbondioxideemissioncontrolonGNPgrowth: interpretationofproposedscenarios.PaperPresentedattheIPCCEnergyand IndustrySubgroup,ResponseStrategiesWorkingGroup,Paris,France.
Keppo,I.,vanderZwaan,B.,2011.Theimpactofuncertaintyinclimatetargetsand CO2storageavailabilityonlong-termemissionsabatement.Environ.Model. Assess.17,177–191.doi:http://dx.doi.org/10.1007/s10666-011-9283-1.
Kramer,G.J.,Haigh,M.,2009.Noquickswitchtolow-carbonenergy.Nature462, 568–569.doi:http://dx.doi.org/10.1038/462568a.
Kriegler,E.,Tavoni,M.,Aboumahboub,T.,Luderer,G.,Demaere,G.,Krey,V.,Riahi,K., Rosler,H.,2013a.Canwestillmeet2Cwithglobalclimateaction?TheLIMITS studyonimplicationsofDurbanActionPlatformscenarios1.Clim.Chang.Econ. 282846.
Kriegler,E.,Weyant,J.P.,Blanford,G.J.,Krey,V.,Clarke,L.,Edmonds,J.,Fawcett,A., Luderer,G.,Riahi,K.,Richels,R.,Rose,S.K.,Tavoni,M.,Vuuren,D.P.,2013b.The roleoftechnologyforachievingclimatepolicyobjectives:overviewoftheEMF 27studyonglobaltechnologyandclimatepolicystrategies.Clim.Changedoi: http://dx.doi.org/10.1007/s10584-013-0953-7.
Kriegler,E.,Petermann,N.,Krey,V.,Schwanitz,V.J.,Luderer,G.,Ashina,S.,Bosetti,V., Eom,J.,Kitous,A.,Méjean,A.,Paroussos,L.,Sano,F.,Turton,H.,Wilson,C.,Van Vuuren,D.P.,2015.Diagnosticindicatorsforintegratedassessmentmodelsof climatepolicy.Technol.ForecastingSocialChange90,45–61.doi:http://dx.doi. org/10.1016/j.techfore.2013.09.020.
Loftus,P.J.,Cohen,A.M.,Long,J.C.S.,Jenkins,J.D.,2014.Acriticalreviewofglobal decarbonizationscenarios:whatdotheytellusaboutfeasibility?Wiley Interdiscip.Rev.Clim.Change doi:http://dx.doi.org/10.1002/wcc.324n/a-n/a.
Luderer,G.,Leimbach,M.,Bauer,N.,Kriegler,E.,Aboumahboub,T.,ArroyoCurras,T., Baumstark,L.,BertramC.,Giannousakis,A.,Hilaire,J.,Klein,D.,Mouratiadou,I., Pietzcker,R.,Piontek,F.,Roming,N.,Schultes,A.,Schwanitz,V.J.,Strefler,J.,2013. DescriptionoftheREMINDmodel(Version1.5).https://www.pik-potsdam.de/ research/sustainable-solutions/models/remind/description-of-remind-v1.5.
McCollum,D.,Nagai,Y.,Riahi,K.,Marangoni,G.,Calvin,K.,Pietzcker,R.,VanVliet,J., vanderZwaan,B.,2013.Energyinvestmentsunderclimatepolicy:acomparison ofglobalmodels.Clim.ChangeEcon.04, 1340010.doi:http://dx.doi.org/10.1142/ s2010007813400101.
Messner,S.,Strubegger,M.,1995.User’sguideforMESSAGEIII.WorkingPaper WP-95-069.InternationalInstituteforAppliedSystemsAnalysis(IIASA). Platt’s,2013.WorldPowerPlantDatabase,ScalingDynamicsofEnergyTechnologies
Model.
Riahi,K.,Kriegler,E.,Johnson,N.,Bertram,C.,denElzen,M.,Eom,J.,Schaeffer,M., Edmonds,J.,Isaac,M.,Krey,V.,Longden,T.,Luderer,G.,Méjean,A.,McCollum,D. L.,Mima,S.,Turton,H.,vanVuuren,D.P.,Wada,K.,Bosetti,V.,Capros,P.,Criqui, P.,Hamdi-Cherif,M.,Kainuma,M.,Edenhofer,O.,2015.LockedintoCopenhagen pledgeséImplicationsofshort-termemissiontargetsforthecostandfeasibility oflong-termclimategoals.Technol.ForecastingSocialChange90,8–23.doi: http://dx.doi.org/10.1016/j.techfore.2013.09.016.
Steckel,J.,Jakob,M.,Marschinski,R.,Luderer,G.,2011.Fromcarbonizationto decarbonization?—PasttrendsandfuturescenariosforChina’sCO2emissions. EnergyPolicy39,3443–3455.doi:http://dx.doi.org/10.1016/j.
enpol.2011.03.042.
Tavoni,M.,vanderZwaan,B.,2009.NuclearversuscoalplusCCS:Acomparisonof twocompetitivebase-loadclimatecontroloptions.Environ.Model.Assess.16, 431–440.doi:http://dx.doi.org/10.2139/ssrn.1516211.
TheWorldBank,2015.GDP(currentUS$)|Data|Table[WWWDocument].URL http://data.worldbank.org/indicator/NY.GDP.MKTP.CD/countries?
display=default(accessed7.10.15).
USEIA,2014.InternationalEnergyStatistics[WWWDocument].URLhttp://www. eia.gov/cfapps/ipdbproject/iedindex3.cfm?tid=2&pid=28&aid=7&cid=CG6,CG5, &syid=2005&eyid=2011&unit=MK(accessed9.5.14).
vanderZwaan,B.C.C.,Rösler,H.,Kober,T.,Aboumahboub,T.,Calvin,K.V.,Gernaat,D. E.H.J.,Marangoni,G.,McCollum,D.,2013.ACross-modelcomparisonofglobal long-termtechnologydiffusionundera2Cclimatechangecontroltarget.Clim. ChangeEcon.1–17.
VanVuuren,D.P.,O’Neill,B.C.,2006.TheconsistencyofIPCC’sSRESScenariosto 1990–2000trendsandrecentprojections.Clim.Change75,9–46.doi:http://dx. doi.org/10.1007/s10584-005-9031-0.
VanVuuren,D.P.,Stehfest,E.,2013.Ifclimateactionbecomesurgent:the importanceofresponsetimesforvariousclimatestrategies.Clim.Changedoi: http://dx.doi.org/10.1007/s10584-013-0769-5.
Weyant,J.,Kriegler,E.,2014.PrefaceandintroductiontoEMF27.Clim.Change123, 345–352.doi:http://dx.doi.org/10.1007/s10584-014-1102-7.
Wilson,C.,Grubler,A.,Bauer,N.,Krey,V.,Riahi,K.,2012.Futurecapacitygrowthof energytechnologies:arescenariosconsistentwithhistoricalevidence?Clim. Change118,381–395.doi:http://dx.doi.org/10.1007/s10584-012-0618-y.
Wilson,C.,2012.Up-scaling,formativephases,andlearninginthehistorical diffusionofenergytechnologies.EnergyPolicy50,81–94.doi:http://dx.doi.org/ 10.1016/j.enpol.2012.04.077.
WorldNuclearAssociation,2014.WNAReactorDatabase[WWWDocument].URL
http://world-nuclear.org/nucleardatabase/rdResults.aspx?id=27569(accessed 9.21.14).
Zhang,M.,Mu,H.,Ning,Y.,2009.Accountingforenergy-relatedCO2emissionin China,1991-2006.EnergyPolicy37,767–773.doi:http://dx.doi.org/10.1016/j. enpol.2008.11.025.
