在线视频碳影响测算白皮书（Carbon impact of video streaming）VIP专享VIP免费

下载本文档

阅读 436
格式 pdf
大小 5.03 MB
约102页
2023-04-17
收藏
点赞(0)
海报
举报

Carbon

impact

of video

streaming

Carbon impact of video streaming Acknowledgments

We are extremely grateful to all of the following who

contributed their time and expertise into the process

of developing this white paper, and between them

provided a combination of valuable insights and

information, and supported the review process of

various drafts of the white paper.

Chloe Fletcher and Jigna Chandaria - BBC

Louise Krug and Penny Guarnay - BT

Christian Toennesen, Glynn Roberts and Catherine Van Loo -

Carnstone Partners

Damien Lieber and Tyler Marcus - Engie Impact

Jens Malmodin - Ericsson Research

George Kamiya - International Energy Agency (IEA)

Clare Phillips, Gareth Barr, Julia Giannini and Tim Davis - ITV

Jonathan Koomey - Koomey Analytics

Arman Shehabi - Lawrence Berkeley National Laboratory

(LBNL)

Emma Stewart, Juan Carlos Ivars and Sergio Vinay - Netix

Lena De Geer and Örjan Åberg - Nordic Entertainment

Group (NENT)

Anna Ma and Marianne Matthews - Sky

Emma Fryer - techUK

Daniel Schien and Paul Shabajee - University of Bristol

Eric Masanet - University of California, Santa Barbara (UCSB)

Acknowledgments

This white paper was made possible by seed funding from

Netix to DIMPACT with the opportunity for all DIMPACT

members to contribute. Extensive consultation with various

DIMPACT members (as listed in the acknowledgements

above) was undertaken, and very useful review and feedback

was provided. The Carbon Trust had full editorial control of

the white paper, and is responsible for the content. The views

and opinions expressed in this white paper are those of the

Carbon Trust and do not necessarily reect the views or

position of any of the reviewers or contributors.

Date

June 2021

Carbon impact of video streaming Acknowledgments

Carbon Trust team

Andie Stephens, Associate Director

Chloe Tremlett-Williams, Analyst

Liam Fitzpatrick, Analyst

Luca Acerini, Senior Analyst

Matt Anderson, Manager

Noor Crabbendam, Analyst

Design:

Flora Buchanan, Brand Designer

About the Carbon Trust

Established in 2001, the Carbon Trust works with

businesses, governments and institutions around the

world, helping them contribute to, and benet from, a more

sustainable future through carbon reduction, resource

eciency strategies, and commercialising low carbon

businesses, systems and technologies.

The Carbon Trust:

• works with corporates and governments, helping them to

align their strategies with climate science and meet the

goals of the Paris Agreement;

• provides expert advice and assurance, giving investors and

nancial institutions the condence that green nance will

have genuinely green outcomes; and

• supports the development of low carbon technologies and

solutions, building the foundations for the energy system

of the future.

Headquartered in London, the Carbon Trust has a global

team of over 200 staff, representing over 30 nationalities,

based across ve continents.

About DIMPACT

DIMPACT is a collaborative project, convened by Carnstone,

with world-class researchers from the University of

Bristol and thirteen of the world’s most innovative media

companies. DIMPACT participating companies are: BBC, BT,

Cambridge University Press, Channel 4, dentsu international,

Informa, ITV, Nordic Entertainment Group, Netix, Pearson,

RELX, Schibsted, and Sky. DIMPACT developed originally

out of the Responsible Media Forum in 2015, with a group

of companies that wanted to better understand the GHG

emissions of their digital media products and services.

DIMPACT was formed in 2019, and started working with

the University of Bristol Department of Computer Science

to develop a tool that provides assessment modules for the

GHG emissions of digital publishing, advertising services,

business intelligence, and video streaming.

/102

下载本文档

文本预览下载提示常见问题

Carbonimpactofvideostreaming2CarbonimpactofvideostreamingAcknowledgmentsWeareextremelygratefultoallofthefollowingwhocontributedtheirtimeandexpertiseintotheprocessofdevelopingthiswhitepaper,andbetweenthemprovidedacombinationofvaluableinsightsandinformation,andsupportedthereviewprocessofvariousdraftsofthewhitepaper.ChloeFletcherandJignaChandaria-BBCLouiseKrugandPennyGuarnay-BTChristianToennesen,GlynnRobertsandCatherineVanLoo-CarnstonePartnersDamienLieberandTylerMarcus-EngieImpactJensMalmodin-EricssonResearchGeorgeKamiya-InternationalEnergyAgency(IEA)ClarePhillips,GarethBarr,JuliaGianniniandTimDavis-ITVJonathanKoomey-KoomeyAnalyticsArmanShehabi-LawrenceBerkeleyNationalLaboratory(LBNL)EmmaStewart,JuanCarlosIvarsandSergioVinay-NetflixLenaDeGeerandÖrjanÅberg-NordicEntertainmentGroup(NENT)AnnaMaandMarianneMatthews-SkyEmmaFryer-techUKDanielSchienandPaulShabajee-UniversityofBristolEricMasanet-UniversityofCalifornia,SantaBarbara(UCSB)AcknowledgmentsThiswhitepaperwasmadepossiblebyseedfundingfromNetflixtoDIMPACTwiththeopportunityforallDIMPACTmemberstocontribute.ExtensiveconsultationwithvariousDIMPACTmembers(aslistedintheacknowledgementsabove)wasundertaken,andveryusefulreviewandfeedbackwasprovided.TheCarbonTrusthadfulleditorialcontrolofthewhitepaper,andisresponsibleforthecontent.TheviewsandopinionsexpressedinthiswhitepaperarethoseoftheCarbonTrustanddonotnecessarilyreflecttheviewsorpositionofanyofthereviewersorcontributors.DateJune20213CarbonimpactofvideostreamingAcknowledgmentsCarbonTrustteamAndieStephens,AssociateDirectorChloeTremlett-Williams,AnalystLiamFitzpatrick,AnalystLucaAcerini,SeniorAnalystMattAnderson,ManagerNoorCrabbendam,AnalystDesign:FloraBuchanan,BrandDesignerAbouttheCarbonTrustEstablishedin2001,theCarbonTrustworkswithbusinesses,governmentsandinstitutionsaroundtheworld,helpingthemcontributeto,andbenefitfrom,amoresustainablefuturethroughcarbonreduction,resourceefficiencystrategies,andcommercialisinglowcarbonbusinesses,systemsandtechnologies.TheCarbonTrust:•workswithcorporatesandgovernments,helpingthemtoaligntheirstrategieswithclimatescienceandmeetthegoalsoftheParisAgreement;•providesexpertadviceandassurance,givinginvestorsandfinancialinstitutionstheconfidencethatgreenfinancewillhavegenuinelygreenoutcomes;and•supportsthedevelopmentoflowcarbontechnologiesandsolutions,buildingthefoundationsfortheenergysystemofthefuture.HeadquarteredinLondon,theCarbonTrusthasaglobalteamofover200staff,representingover30nationalities,basedacrossfivecontinents.AboutDIMPACTDIMPACTisacollaborativeproject,convenedbyCarnstone,withworld-classresearchersfromtheUniversityofBristolandthirteenoftheworld’smostinnovativemediacompanies.DIMPACTparticipatingcompaniesare:BBC,BT,CambridgeUniversityPress,Channel4,dentsuinternational,Informa,ITV,NordicEntertainmentGroup,Netflix,Pearson,RELX,Schibsted,andSky.DIMPACTdevelopedoriginallyoutoftheResponsibleMediaForumin2015,withagroupofcompaniesthatwantedtobetterunderstandtheGHGemissionsoftheirdigitalmediaproductsandservices.DIMPACTwasformedin2019,andstartedworkingwiththeUniversityofBristolDepartmentofComputerSciencetodevelopatoolthatprovidesassessmentmodulesfortheGHGemissionsofdigitalpublishing,advertisingservices,businessintelligence,andvideostreaming.4CarbonimpactofvideostreamingContentsExecutivesummary61.Introduction101.1.Purposeofthewhitepaper111.2.Contextualbackground142.Background–globalenergyandcarbonimpactoftheE&MandICTsectors162.1.VideostreamingaspartoftheEntertainmentandMediasector172.2.DefiningICTandE&Msectorandboundary182.3.TheICTsector’scarbonfootprint–variationsinestimates202.4.CurrentandhistoricalcarbonfootprintofICT212.4.1Innovationenablesgreaterenergyefficiency212.5.ICT’scarbonfootprintin2020222.5.1HistoricalcarbonfootprintofICT242.6.CurrentandhistoricalemissionsofICTsectorcomponents262.6.1Datacentres262.6.2Networkdatatrafficandenergy282.6.3End-userdevices302.6.4EntertainmentandMedia(E&M)312.7.Growthofglobaldataconsumption322.8.FuturetrendsinICTcarbonemissionsfootprint363.Methodologyandapproach373.1.Overviewoftheconventionalapproach383.2.Approachboundary403.3.Datacentres403.4.Transmissionnetwork423.5.End-userdevices464.Resultssummary474.1.TheamountofcarbonemittedandenergyconsumedperhourofvideostreaminginEuropeissmall484.2.Thelocalelectricalgrid’scarbonintensityhasacriticaleffectonthecarbonimpactofvideostreaming504.3.Userdevicesdrivetheimpactofvideostreaming525.Discussion54Contents5CarbonimpactofvideostreamingContents5.1.Emergingresearchoffersnewinsightintotheshort-termmarginaleffectsofchangingviewingpatterns555.2.Malmodin’ssimplepowermodelprovidesacloserrepresentationofthedynamicsofinternettransmission555.3.Methodologyofthepowermodelapproach575.4.Overviewofthepowermodelapproach575.5.Approachboundary595.6.Transmissionnetwork595.7.Theshort-termmarginaleffectofvideostreamingqualityoncarbonemissions635.8.Furthervalidationandrefinementofthepowermodelisalogicalnextstep665.9.Videostreamingimpactcontextualised685.10.Forconsumers,deviceselectioncanreduceenvironmentalimpact,butsystemicapproachtoenergyefficientdevicesoffersamoremeaningfulopportunity705.11.Internetpeakcapacitydrivesenergyconsumption705.12.Consumptionofvideohasprogressedinacarbonefficientmanner725.13.Ashiftinbehaviourmayreduceemissions725.14.Predictingthefutureisdifficultandhasinherentuncertainty735.15.Ourunderstandingofnetworkenergyandemissionsislimitedbydetaileddataavailability745.16.Areasforfurtherinvestigation746.Policydevelopments766.1.Introduction–Policydevelopments776.2.Policydevelopments786.2.1Traceabilityandmonitoringframeworks786.2.2DataCentres786.2.3NetworkTransmission796.2.4End-userviewingdevices806.3.Industryinitiatives816.3.1Datacentres-Industry826.3.2Networksandtransmission-Industry826.3.3End-userdevices–Industry836.4.Policydevelopmentsandobservations837.Conclusions848.QuestionsandAnswers879.References9310.Appendix996CarbonimpactofvideostreamingExecutivesummaryExecutivesummary7CarbonimpactofvideostreamingExecutivesummaryThiswhitepaperisaboutthecarbonimpactofwatchingonehourofvideostreamingTwomethodologiesarepresented,withthekeydifferencebetweenthemethodsbeinghowthenetworkelectricityisallocatedtovideostreaming.(specificallyinreferencetoondemandstreaming,notlivestreaming).Itlooksatthisfromalifecycleperspective,andpresentstheresultsintermsofcarbonemissionsforonehourofvideostreaming.Itconsiderstheenergyuseofthedifferentcomponentsthatareinvolvedinthedistributionandviewingofvideocontent:datacentresandcontentdeliverynetworks(usedforencodingandstorage);internetnetworktransmission;homerouters;end-userviewingdevices(e.g.TVs,laptops,tablets,smartphones);andTVperipherals(e.g.set-topboxes),whererelevant.Theboundaryscopeincludesonlytheoperationalelectricityuseofthedifferentcomponentsnotthecontentcreationnortheembodiedemissionsoftheequipment.Thecarbonemissionsfromtheelectricityuseareallcalculatedbasedonnationalelectricitygridaverageemissionfactors,sodonotrecognisewheredatacentresandnetworkoperatorsdirectlyuserenewableelectricity.Thiswhitepaperexplainsthedetailsoftheassumptionsandmethodsusedanddiscussesthechallengesanduncertaintiesinvolvedinestimatingthecarbonimpactofvideostreaming.Theaimistocontributetotheunderstandingofthetopic,sothatfuturedecisionscanbebasedonaninformedunderstandingoftheissue,withaninsightofthemethods,uncertaintiesandvariabilitythatcanaffecttheestimates.Thefirstistheconventionalapproach,whichiswellestablishedandhasbeenusedinmostpreviousstudies,andfollowsanaverageallocationmethodology,wheretheinternetnetworkelectricityisallocatedusinganaverageenergyperdatavolumemetric[kWh/GB].Thiswhitepaperalsopresentsapowermodelapproach,whichusesamarginalallocationmethodology,whereabaseloadpowerisallocatedperuser,andamarginalenergycomponentisallocatedrelatedtothedatavolumeused.Thepowermodelapproachrecognisesthatthedynamicrelationofenergytodatavolumeinanetworkisveryflat–i.e.thereisahighfixedpowerbaseloadwhichdoesnotvaryinrelationtothedatavolume,withonlyasmallincreaseinpowerconsumptioninresponsetothedataconsumption.ThepowermodelapproachusesresearchpublishedinSeptember2020;beforethentherewasnotsufficientinformationpublishedforthepowermodelallocationapproachtobeappliedwidely.Whilethepowermodelapproachmorecloselyrepresentstheinstantaneoususeofenergyinthenetwork,theconventionalapproachrepresentstheaverageenergyuseandthereforeisgenerallyusedforreportingandaccountingpurposes.Thefollowinganalogyofabusnetworkishelpfulinillustratingthedifferencesbetweenthetwoallocationapproaches.Theenergy(i.e.fuel)usedbythebusesinthenetworkisfairlyfixed,withonlyamarginalincreaserelativetothenumberofpassengers.Thepowermodelapproachwillallocateafixedamountofenergyperuser,plusamarginalamountperkilometretravelled.Theconventionalapproachwillallocateanaverageamountofenergyperpassenger-km,(whichwouldbederivedfromthetotalannualfuelconsumption,andthetotalannualpassenger-kmtravelled).Thepowermodelapproachreflectstheimmediateimpactofwhetheryouusethebusornot,whiletheconventionalapproachreflectstheaverageimpactofusingabus(asitconsidersthetotalannualoperationalemissionsofthebusnetwork)andisusefulforaverageaccountingpurposes.8CarbonimpactofvideostreamingExecutivesummaryAswithmostcarbonfootprintassessmentsthereisinherentvariabilityanduncertaintyintheestimationofthecarbonimpactofvideostreaming,whichgivesrisetoarangeofresults.(Variabilityreferstovariationsduetofactorssuchastimeorplace,whileuncertaintyreferstothedegreeofprecisionofmeasurements.)Attheindividuallevel,thecarbonfootprintofviewingonehourofvideostreamingisverysmallcomparedtoothereverydayactivitiesTheEuropeanaveragefootprintestimatedinthiswhitepaperisapproximately55gCO2eperhourofvideostreamingfortheconventionalallocationapproach.(ThisestimateusesaEuropeanaveragegridemissionfactor,arepresentativemixofviewingdevices,andnetworkenergyintensityfiguresfor2020.)Forcomparison,theemissionsfrommicrowavingabagofpopcornforfourminutesisabout16gCO2e(alsousingaEuropeanaveragegridemissionfactor),whiledriving100metresinanaveragepetrolcaremitsaround22gCO2e.Thesefootprintfiguresforvideostreamingarecomparablewithsomeotherrecentestimates.However,therearealsosomepreviousstudieswithmuchhigherestimates,themainreasonforthedifferencebeingthatthosestudiesusedoldernetworkenergyintensityfigureswhicharesignificantlyhigherthanfiguresrelevantto2020.Theanalysisinthiswhitepaperalsoshowsthattheviewingdeviceistypicallyresponsibleforthelargestpartofthecarbonfootprint.Itshouldbenotedthatneitheroftheallocationmethodsreflectsthepeakdatausage,andthisisoneofthedriversofthelonger-termtotalnetworkenergyconsumption.Thevariabilityisduetotemporal,geographicalandtechnologicalfactors.Thebiggestvariabilityrelatestothecountry-specificelectricitygridemissionfactor–forexample,inEuropeGermany’sgridemissionfactorisapproximately30timesthatofSweden,whichtranslatesdirectlytoa30timesdifferenceintheoverallcarbonfootprint.Thesecondmostsignificantfactoraffectingthevariabilityinthecarbonfootprintistheviewingdeviceused–thefootprint(relatedspecificallytotheenergyoftheviewingdevice)ofwatchingona50-inchTVisroughly4.5timesthatofwatchingonalaptop,androughly90timesthatofwatchingonasmartphone.Theyearthatanestimationrelatestoisalsosignificant,asimprovementsintechnologymeanthattheenergyintensityofequipmentiscontinuallydecreasing,andseparatelytheelectricityemissionfactorsaredecreasingastheelectricitygridsdecarbonisethroughtheutilisationofgreaterproportionsofrenewables.Also,networkenergyintensityfactorswillvarybyoperatorandbycountry,duetofactorssuchasageofnetworkequipment,topologyofthenetwork,populationdensity,andevenclimaticfactorssuchasambienttemperatureandhumidity.TheEuropeanaveragefootprintisestimatedtobeapproximately55gCO2eperhourofvideostreamingThemostsignificanteffectoftheuncertaintyisrelatedtotheinternetnetworkcomponentofthefootprint.Theuncertaintyinthenetworkenergyarisesfromthedifferencesinallocationmethod,andfromthefactthatthereisalimitednumberofpubliclyavailabledatapointsfortheenergyintensityofnetworks.100mPOPCORN1/39CarbonimpactofvideostreamingExecutivesummaryThiswhitepaperaimstoimprovetheunderstandingofthecarbonfootprintofvideostreaminganditscomplexity,variabilityanduncertainty.Itshouldbeseenasawork-in-progress,astherearemanyopportunitiesforfurtherresearchtoimproveunderstandinginthisarea.Akeyareatoinvestigatefurtherisdifferentallocationmethods,asitiscriticallyimportantthatthemethodisappropriateforthequestionsbeingaskedandthedecisionstaken.Relatedtothisistheneedforunderstandingthekeydriversaroundincreaseddemandfordata,inparticularpeakdatademand.Tounderstandtheseissuesalsorequiresavailabilityofmoredetailedinformationfromnetworkoperatorsonenergyanddata.Asthecarbonintensityofelectricityreduces(andoperatorsuse100%renewableelectricity),thentheembodiedemissionsofequipmentanddeviceswillbecomemoresignificant,thereforethisshouldbethenextareatoassessinmoredetail.Otherareasthatwouldbenefitfrommoreresearchandmoredataare:theallocationofthehomerouterenergyandinformationondevicesandapplicationsusingthehomerouter;andinformationonthemixanduseofviewingdevices.WhatisclearisthatastrongunderstandingoftheimpactandcontextofvideostreamingisvitaltoinformfuturedecisionsaffectingtheuseofvideostreamingandtheuseofICTingeneral.AnalysisofCiscoforecastsshowthat,in2020,long-formvideostreaming(i.e.withanaverageviewingtimeofgreaterthanfiveminutes)accountedforabout45%oftotalinternettraffic.Dependingonhowtrendsinvideostreamingchangeinthefuture,theassociatedinternettrafficcouldhaveimpactsonthetotalenergydemandoftheinternet.Usingthepowermodelapproach,whichreflectstheinstantaneous(ormarginal)changesinenergy,demonstratesthatchangesinbitrate(duetodifferentresolutionsandothersettings)resultinonlyaverysmallchangeinthecarbonfootprint.Thisisbecausetheinternettransmissionandthehomerouterusemuchthesameenergywhateverthedatavolumesare,andtheviewingdevicesenergyconsumptionalsoonlychangesbysmallamountsdependingontheviewingresolution.Understandingthelonger-termimpactsofvideostreamingismorecomplicated.Thetotalnetworkenergyisprimarilydrivenbythetotalpeakdemandfordata.However,asnetworksarecontinuallyupgradedwithnewernetworkequipmentthatismoreenergyefficient,thegreaterdatavolumescanbehandledwithlessenergyconsumption.Whatisdrivingthepeakdemand?Isitdemandforservices,andwhichservices–thosethatusehigherdatavolumes,orthosethatrequirefasterresponsetimes(lowerlatency)?Orarethetechnologyimprovementsthatenablehigherbandwidthsdrivingnewapplicationsandservicesthatcantakeadvantageoftheseimprovements?InthiswhitepaperwealsonotethatactionsandtrendsintheICTsectoraredrivingdownthecarbonintensityofICTservicesincludingvideostreaming.Thelargedatacentrecloudprovidersareincreasinglypurchasingrenewableelectricity,manywith100%renewabletargets,withsomealreadyat100%.Similarly,anumberofmajortelecomsnetworkoperatorshave100%renewabletargets,andanincreasingnumberaresettingapproved1.5°Ccompatiblescience-basedtargets.Theend-userviewingdevicesarealsobecomingmoreenergyefficientduetoamixoftechnologyadvances,regulationandstandards(e.g.aroundstandbypowerandmaximumpowerthresholds).ThismaybefurtherhelpedthroughthetrendofusingsmallerdevicessuchastabletsandlaptopsforviewingratherthanTVs,althoughitisnotclear;a)howmuchthisispurelysubstitutionofadeviceratherthanadditionalviewing;andb)towhatextentcommunalviewingoflargerscreensoffsetstheadditionalenergyrequirements.However,thereisatrendforTVstohavelargerscreensizes,andthereforepotentiallyhigherenergyconsumption(albeitasnotedabove,technologyisimprovingtheenergyefficiencyoftelevisionsandotherdevices,andscreensizescannotcontinuetoincreaseinsizeindefinitely),togetherwithanoverallincreaseinnumberofviewinghours.CarbonimpactofvideostreamingIntroduction101.Introduction11CarbonimpactofvideostreamingIntroduction1.1.PurposeofthewhitepaperThepurposeofthiswhitepaperistocontributefurthertodevelopknowledgearoundthemeasurementoftheenergyandcarbonimpactofondemandvideostreaming(VoD).Thewhitepaperpresentsthecurrentmagnitudefortheelectricityconsumptionandoperationalcarbonemissionsofvideostreaming.Subsequently,thewhitepaperoutlinescurrentpolicytrends,bothongovernmentalandcompanylevels.Thewhitepaperdiscussesthecomplexitiesofvideostreamingandmoreoverthecomplexitiesofmeasuringitscarbonfootprint.Toprovideaproperunderstandingofthiscomplexity,thewhitepaperexploresthedifferentcomponentsinvolvedinvideostreaming(asshowninthediagrambelow)fromalife-cycleperspective:DataCentres(fororiginatingandencodingofvideocontent),ContentDeliveryNetwork(CDN-fortemporarystorageanddelivery),InternetNetworkTransmission,HomeTerminalsandRouters,HomePeripherals(e.g.set-top-boxes),andEnd-Userdevices(screens).Toestimatethevideostreamingcarbonfootprint,thispaperoutlineswhereinthelife-cycletheemissionstakeplaceandwhatmagnitudestheyhave.Figure1.ProcessmapshowingboundaryscopeofvideostreamingCorenetworksHometerminalsandroutersWiredaccessnetworksCellularaccessnetworksCloudstorageandencodingSubscriberpremisestransmissionPeripheralsScreensInternettransmissionContentDeliveryNetworkUsermediadeviceDatacentresTransmission12CarbonimpactofvideostreamingIntroduction12CarbonimpactofvideostreamingIntroductionThescopeincludedisthatrelatedtothedistributionandviewingofvideostreaming,asshowninFigure1.Contentcreationisnotincluded.Theemissionsrelatetotheoperationalelectricalenergyuseofthedifferentcomponents,anddonotincludetheembodiedemissionsoftheequipment.Thecarbonemissionsfromtheelectricityuseareallcalculatedbasedonnationalelectricitygridaverageemissionfactors,sodonotrecognisewheredatacentresandnetworkoperatorsdirectlyuserenewableelectricity.Thiswhitepaperhasbeendevelopedtoprovideanexplanationofthecalculationsbehindthenumbersfortheemissionsofvideostreaming,andexplainwhatfactorsandassumptionsareused,andhowtheseaffectthecalculatednumbers.Thewhitepaperpresentstwodifferentmethodsforallocationofenergytovideostreamingservices,discusseswhythisisimportant,andwhyallocatinginternetnetworkenergypurelyonadatavolumebasis,cangivemisleadingresults,ifnotproperlyunderstood.TheconventionalapproachTheconventionalapproachusesanaverageallocationmethodology,wheretheinternetnetworkelectricityisallocatedusinganaverageenergyperdatavolumemetric[kWh/GB].Thetwomethodologiespresentedinthiswhitepaperare:ThepowermodelapproachThepowermodelapproachusesamarginalallocationmethodology,wherenetworkbaseloadpowerisallocatedperuser/subscriber,andamarginalnetworkenergycomponentisallocatedrelatedtothedatavolumeused.1.2.13CarbonimpactofvideostreamingIntroductionThetwoapproachesalsousedifferentallocationsforthehomerouterenergy.Theconventionalapproachusesanaverageallocationrelatedtothedatavolume,whilethepowermodelapproachallocatesrouterenergyconsideringnumberofusers,numberofconnecteddevicesperuser,andalsoallocatestheenergyusedbytherouterwhenitisinanidlestate.Theconventionalapproachisthewell-establishedapproachthatrepresentstheaverageenergyuseandthereforeisverysuitableforreportingandaccountingpurposes.Thepowermodelapproachmorecloselyrepresentstheinstantaneoususeofenergyinthenetwork,andisthereforeusefulforunderstandingtheshort-termmarginalchangeofenergyconsumptioninresponsetochangesinviewingpatterns.ForamoredetaileddescriptionoftheconventionalapproachmethodologypleaseseetheMethodologysection,forapresentationoftheimpactsseetheResults,andforadescriptionofthepowermodelapproachandfurthercommentontheimpactanditsimplications,seetheDiscussionsection.ThemethodologiesforcalculatingtheemissionsimpactofvideostreamingpresentedinthiswhitepaperarebasedontheapproachesusedinmodelsdevelopedbyDIMPACTandNetflix.TheCarbonTrustwasapproachedbyDIMPACTtoreviewbothmodelsandtointerviewthecompaniesusingthem.DIMPACT,acollaborativeprojectconvenedbyCarnstone,withresearchersfromtheUniversityofBristoland13globalentertainmentandmediacompanies,hasoverthelastyearsdevelopedanonlinetoolforreportingemissions.DIMPACT’sonlinetoolisdesignedfortheDIMPACTmembercompaniestousewiththeirspecificdatatoestimatethecarbonemissionsofthevaluechainoftheirdigitalmediaservices,excludingcontentcreation.Itisusedprimarilyfororganisationalreporting(forexample,reportingofScope3emissions),butalsotosupporttheidentificationofopportunitiesfordesigninglowercarbonservices.Thetoolprovidesmodulesforavarietyofdigitalservicesincludingvideostreaming.Independently,NetflixdevelopedamodelinpartnershipwithEngieImpactandwiththeadviceofaninternationalpanelofacademicexperts,toestimatetheemissionsimpactofanhour’sworthofvideostreamingfromalife-cycleperspective.Bothmodelsusetheconventionalapproachwithsimilarmethodologyandassumptions,andtheresults(whenexpressedasemissionsforanhourofvideostreaming)showclosealignment.TheNetflixmodeladditionallyhasanoptiontousethepowermodelapproach,whichwasdevelopedbasedonpublishedresearchonnetworkpowermodels(Malmodin2020b)andwithfurtherinputfromacademicexperts.TheCarbonTrusthascriticallyreviewedthestructureandassumptionsinbothmodels,andwherepossiblewehavecrosscheckedsomeofthekeyinputsandassumptionsagainstotherdata.Inthiswhitepaper,wealsodiscussthemeritsanddisadvantagesofdifferentassumptionsandapproaches.ThiswhitepaperthenconsiderscurrentandfuturepolicyrelatedtoboththeenvironmentalimpactofstreamingandoftheICTsectoringeneral,inthecontextoftheseestimates.Acomprehensivesetoflegislativeandnon-legislativeinitiativesareinplacetoworktowardsclimateneutralityinEuropeby2050.FortheICTsectorspecifically,theEuropeanCommissionsetsoutadigitalstrategy.ThemainfocusofthisanalysiswillbeonEuropeanUnionPolicyandspecificrelevantnationalpoliciesthatarecurrentlyinplaceorinadevelopmentstage.OnaEuropeanlevelthemainfocuswillbetheEuropeanGreenDeal.Forthepurposeofthispaper,thefocuswillbeonpolicyrelatingtoenvironmentalreportingandclimateactionforthestreamingandICTsector.Thewhitepaperalsodiscussesindustry-ledinitiatives,focusingonexistinginitiativesbeingundertakenbythesectoritselftoeithermeasure,reportorreducecarbonemissionsassociatedwithvideostreaming.Giventhedevelopingnatureofresearchandmethodologyforthemeasurementoftheemissionsofvideostreaming,thispaperhighlightsopportunitiesforfurtherresearchintoimprovingthemethodsandcriteriaforevaluatingvideostreamingemissions.14CarbonimpactofvideostreamingIntroduction1.2.ContextualbackgroundThereisgrowingawarenessandconcernexpressedinthemediaoverthecarbonandenergyimpactoftheICTsector,includingontheimpactofvideostreaming.Avarietyofdifferentestimatesofthecarbonimpactofvideostreaminghavebeenpublished(SeeTable1).Thistableillustratesthevariabilityofrecentlypublishedestimates.Companiesandacademicshavebeenworkingtobetterunderstandtheimpactofvideostreamingandtoimprovetheestimatesofthecarbonimpact.Table1.EstimatesofthecarbonimpactofvideostreamingEstimateYearrelatestoReferenceCarbonintensity[gCO2e/streamedhour]PurdueUniversityestimate2020Obringer,2021440IEAglobalestimate(Revisedestimate,December2020)2019IEA,2020c36IEAglobalestimate(Originalestimate,February2020)2019IEA,2020c82BITKOM:globalestimatefor2018720p65"TV2018Bitkom,2020130BITKOM:globalestimatefor20184K65"TV2018Bitkom,2020610BITKOM:globalestimatefor2018720pSmartphoneFixednetworks2018Bitkom,202030ShiftProjectupdatedglobalestimate2018TheShiftProject,2020394ShiftProjectglobalestimate(fromAFPinterview)2018TheShiftProject,2019bFrance24,20193,200BBCiPlayerestimate2016BBC,202098LBNL/NUestimatefortheU.S.2011Shehabi,2014360BBCestimatefortheUKfor2011STB+TVSD(480p)2011BBC,20117615CarbonimpactofvideostreamingIntroductionThereareanumberoffactorsthatcanhaveasignificantimpactontheresultsandexplainsomeofthevariabilityintheresults.OneisthefactthatICTtechnologyiscontinuallybeingupdatedandimprovinginenergyefficiency,thustheyeartowhichtheresultsrelateissignificant,andusingoutdatedenergyintensity(andhencecarbonintensity)figureswillover-estimatetheresult.Thesecondfactorisrelatedtothemethodofallocatingtheenergy(andhenceemissions)ofsharedresources(suchastheinternettransmission).Athirdsignificantfactoristhatthecarbonintensityofelectricity(i.e.theelectricitygridemissionfactors)varysignificantlyfromcountrytocountry,andalsohavebeenreducingovertimeduetodecarbonisationoftheelectricitysupply.TheCOVID-19pandemichighlightedsociety’srelianceonICT,andparticularlytheinternetinfrastructure.Datatrafficincreasedduetodemandsfromhomeworking,homeeducation,andhomeentertainment.ThisincreasedawarenessoftheimpactofICTingeneral,includingvideostreaming.Interestingly,althoughtherewasasignificantincreaseindatatraffic,thisdidnothaveasimilarimpactintermsofenergyuse.Telecomnetworkoperatorsreportedonlymarginal(lessthan1%)increasesinenergyconsumption,despiteincreasesindatatrafficofupto50%(GSMA,2020).Thiseffecthasalsobeenreflectedforthewholeof2020.Telefonicareporteda45%increaseindatatrafficin2020,duetoCOVID-19,yetreportedaslightdeclineinenterpriseenergyuse(notingthatthenetworksaccountfor~90%oftheirenterpriseenergyuse)(Telefonica,2020).Similarly,Cogent,alargeoperatoroffibre-opticbackbonenetworks,reporteda38%increaseindatatrafficfor2020,however,itsoverallnetworkenergyusedecreased(Cogent,2020a;2020b).Suchup-to-datereportingbynetworkoperatorsrefutestheassumptionthatenergyuseisdirectlyproportionaltodatavolumes,anddemonstratesthatincreaseddatatrafficdoesnotautomaticallyresultinmorenetworkenergyuse.ThisrelationshipisevenmoresignificantinayearwheretheeffectsofCOVID-19hasresultedinlargesurgesindatatraffic,duetohomeworking.CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors162.Background–globalenergyandcarbonimpactoftheE&MandICTsectors17CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.1.VideostreamingaspartoftheEntertainmentandMediasectorVideostreamingisanentertainmentservicedeliveredovertheinternet.Assuch,itisheavilydependentonelementsoftheICTsector.Tounderstandtheemissionsofstreamingitishelpfultorealisehowvideoentertainment(andthemediaindustryingeneral)hasbeendigitising.ICThasbecomeanintrinsicpartofeverydayworkandsociallife,withconstantconnection,instantaneousmediaandsocialmedia.MultipleeconomicsectorsrelyonICT,withICTprovidingahorizontallayerthatcutsacrossavastnumberofindustries.Allofthesestagesconsumeelectricityandhencegeneraterelatedcarbonemissions.ThisthereforerequiresanunderstandingofthewidercontextofICTandE&M.Thedevelopmentofvideoentertainmenthasbeenrapid.VideostorerentalswerereplacedwithDVDpostaldelivery,whichinturnhasbeenreplacedbyonlinestreaming.Likeothersectors,theEntertainmentandMedia(E&M)sectorhasgraduallyshiftedtowardsdigitalisationanddematerialisationofitsservices.VideostreamingisreliantontheICTandE&Msectorstodelivercontentintothehome,andwithintheICTsystemtherearemultipletouchpoints.Thekeyonesbeing:TheoriginatingandencodingofvideocontentisperformedinDataCentresThehomeisconnectedtotheinternetusinghometerminalsandroutersThetransmissionofvideofromthedatacentrestotheCDNtothehomeoccursoverthetelecommunicationsnetworkscomprisingtheinternetAndfinally,watchingthevideousesanend-userdevicesuchasalaptop,tablet,smartphone,orTV.Videocontentisstoredonedgeserversclosetotheend-userforbetterqualitystreamingusingContentDeliveryNetworks(CDN)Somevideoservicesusehomeperipherals(e.g.set-topboxes)toenableselectionoftheservices18CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.2.DefiningICTandE&MsectorandboundaryTheremainderofthissectionlooksattheglobalcarbonfootprintofICTandE&M,includinghistoricandfuturetrends.Westartwithsomedefinitions.TheInformationandCommunicationTechnology(ICT)sectorisdefinedbytheOECDas“acombinationofmanufacturingandservicesindustriesthatcapture,transmitanddisplaydataandinformationelectronically”1.ForreportingofGHGemissionsthisusuallyrelatestotheemissionsofthesethreecomponents:•datacentresthatstoreandprocessdata,•telecommunicationsnetworks(includingbothmobileandfixed)thattransmitdata,and•end-userdevicesthatfurtherprocessanddisplaydata.1http://www.oecd.org/digital/ieconomy/2771153.pdfDefiningtheboundaryoftheICTandE&Msectorsisofcriticalimportancewhenestimatingthesector’scarbonemissions,recognisingthepointsofcross-sectoroverlapandconvergence.AssessmentsoftheglobalemissionsimpactoftheICTandE&MsectorstypicallyusequitespecificboundarydefinitionsofwhatequipmentisincludedinICTvswhatisincludedinE&M.Thesearepurelyforthepurposesofestimatingtheglobalemissionsimpact,andmaynotcompletelyalignwithmoregeneralperceptionsoftheICTandE&Msectors.Thus,lookingattheglobalemissionsimpact,wefollowtheboundarydefinitionsfromMalmodin&Lundén,(Malmodin,2018a).ICTemissionsboundaryTheInformationandCommunicationTechnology(ICT)sectorisbroadlycategorisedasITservicesandtelecommunicationsnetworks,andiscategorisedbythreesub-components:datacentresthatstoreandprocessdata,networks(includingbothmobileandfixed)thattransmitdataandend-userdevices(excludingdevicesincludedintheE&Mboundary).E&Memissionsboundary‘TheEntertainmentandMedia(E&M)sectorcomprisesallelectronicequipmentutilisedformediaandentertainmentpurposes,including:TVs,cameras,andotherE&Mconsumerelectronics,aswellasphysicalpapermediaandprinting.19CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsICTemissionsboundaryE&MemissionsboundaryCorenetworksHometerminalsandroutersWiredaccessnetworksCellularaccessnetworksCloudstorageandencodingSubscriberpremisestransmissionPeripheralsScreensInternettransmissionContentDeliveryNetworkUsingtheMalmodin&Lundéndefinition,thefollowingFigure2relatesthecomponentsofvideostreamingtotheICTandE&Memissionsboundaries.Figure2.ProcessmapshowingICTandE&MboundariesUsermediadeviceDatacentresTransmission20CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.3.TheICTsector’scarbonfootprint–variationsinestimatesThereisinherentcomplexityanduncertaintywhencalculatingthefootprintofICT.Asaresult,estimatingthecarbonemissionsofICThas,historically,provenquitechallenging.PreviousestimationsofICT’scarbonandenergyfootprintvarysubstantially.Thisvariationbetweenfootprintestimationsresultsfromdifferencesinthescope,methodologyandboundarydefinitionfortheICTsector,thatrangessubstantiallybetweenpreviousstudies(Freitagetal.,2020).Firstly,definingtheboundaryoftheICTsectorcanimpacttheresultsofcalculation.TheICTsector,unlikeothermorefinitesectorsismoredifficulttodefineintermsofitsboundary.TheICTsectoractuallyconsistsofavarietyofverydifferentsub-sectors(e.g.ICTequipmentmanufacturing,componentmanufacturing,datacentreoperations,telecommunicationnetworkoperations,software,ITservices).Whereas,forexample,globalsteelproductionhasawell-definedboundarywithafinitenumberofmanufacturingoperations/plantsandvaluechainthatcoversitssupply.Additionally,manysteelcompaniesreporttheiremissionsdata,viatheWorldSteelAssociation,enablingmoreaccuratedeterminationofthesteelindustry’stotalfootprint.ThisleveloftransparencyhasnotyetbeenachievedwithintheICTsector.Therefore,definingICT’ssectorboundaryispotentiallymorecomplexthanotherindustries.Theinternetreferstoaglobalintegratednetworkofnetworksconnectingmillionsofdifferentusersanddevices,thatareeachcapableofsending,receivingandprocessingdataallofthetime.Thus,itishardertoagreeonauniversaldefinitionofitsboundary.Assuch,previousstudiesmayusedifferingdefinitions.Importantconsiderationsincludethescopeoftechnologiesusedforthecalculation,suchastheinclusionofTVsortypesofIoTdevices.Also,theinclusionofembodiedemissionsofICThardwareandequipmentisanotherkeyconsideration.Studiesvarydependingontheirallocationofthesefulllifecycleemissions.Namely,whethertheirestimationaccountsforend-of-lifeemissionsaswellasupstreamproductionandmaterialextractionemissions(Freitagetal.,2020).ThecarbonfootprintofICTservicesisalsodependentontheelectricitymix,thatvariesbetweendifferentcountries.Thus,whetherthecalculationaccountsforrenewableenergyportionofelectricityusedwillsignificantlyaffecttheestimation.Secondly,asaresultoftheICTsector’scomplexity,themethodsandassumptionsusedforcarbonfootprintcalculationsdifferbetweenstudies.Someadoptatop-downapproachofglobalICTenergyestimates.Thisapproachreliesontheextrapolationofhistoricestimatesofcarbonintensityandestimatingfuturetrendsfrommodelprojections.Alternatively,systematicbottom-upapproacheshavebeenapplied,usingreal-worlddatatoestimatethefootprintofeachICTsectorcomponentsuchasglobaldatacentreservers(Masanet,2020a)ornetworkoperatoremissions(Malmodin&Lundén,2018a;Malmodin2020a).Thesetendtoproducemorerobustestimationsastheyarebasedondetaileddatapointsthatcanbescaled-uptoprovideaglobalfigure.Inaddition,somestudiesareactuallyscenarioanalyses–modellingwhattheimpactforenergyandemissionswouldbe,basedondifferinggrowthassumptions.Then,thesescenariosareoftenoversimplifiedandreportedinthemediatothepublicaseitherconcreteprojectionsorasfacts,whentheyweresimplyansweringaseriesof“whatif”scenarioquestions.SomestudieshaveproducederroneousestimatesbyrelyingonpreviouslypublishedestimatesoftheICTsectoremissions(thatmaythemselvesbefiveto10yearsold),thenextrapolatingthemtothecurrentdateusinghistoricalgrowthprojections(whichmayalsobefiveto10yearsold).Becausethetechnologychangessorapidlyitisnotreliabletosimplyuseextrapolationsfromhistoricdata.Themorerobustestimatesrecalculatetheemissionsbyusingthelatestavailableindustrydata:fordatacentresusingindustrydataonthenumberofserversandserverenergyconsumption;fortelecomsnetworksusingactualreportedemissionsfromnetworkoperators;andforend-userdevicesusingindustrydataonnumbersofdevicessoldandtheenergyconsumptionofdifferentdevicecategories.Anadditionalcomplexityisthattheemissionsintensityofelectricitygenerationvariesovertimeandbylocation,andthesevariationsarebigenoughtomatter(morethananorderofmagnitudeinsomecases).ItisthereforedifficulttocompareindependentestimatesofICTemissionsdirectlywithoutsupportinginformationontheenergyintensityofICTtechnologyandtheemissionsintensityofelectricitygeneration.21CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.4.CurrentandhistoricalcarbonfootprintofICT2.4.1InnovationenablesgreaterenergyefficiencyThepaceofICTtechnologicalinnovationhasenabledcontinuedimprovementsinprocessingpowerandgreaterenergyefficiencyacrossthesector’sentirevalue-chain.ItiswidelyacknowledgedthatenergyefficiencyofICTandcomputingequipmenthashistoricallydoubledeveryonetothreeyears(Kamiya,2020;Koomeyetal.,2011a,KoomeyandNaffziger,2016),andsimilarlyenergyefficiencyofnetworkshashistoricallydoubledapproximatelyeverytwoyears(Aslanetal.,2018),seeboxbelow.ThissustainedefficiencyhashelpedtostabiliseICT’scarbonfootprint,evenasthesectorhascontinuedtoexpand.Moore’sLawandKoomey’sLawIn1965,GordonMooreobservedthatcomputermicroprocessorsdoubledintransistordensityeveryyear(modifiedtoeverytwoyearsin1975),increasingthenumberoftransistorsperunitareaandthusimprovingperformance.Subsequentanalysisidentifiedarelatedtrend,adoublingofefficiencyroughlyevery1.6yearsforcomputinghardwarerunningatfulloutput(Koomeyetal.,2011a),atrendthatisoftenknownasKoomey’sLaw.Subsequentanalysisshowedthatefficiencyimprovementsforcomputersatpeakoutputslowedafter2000forreasonsassociatedwithsemiconductorphysicsbutcontinuedtodoubleevery2.6years(KoomeyandNaffziger,2016).Thesecontinualimprovementsinmicroprocessorchipdesign,manufacturing,andsoftwaredroveenergyefficiencyimprovementsthathavetosomeextentoffsetrisingdemandsforcomputinganddataservices.Aslan’sruleSimilarly,analysisofestimatesfortheaverageelectricityintensityoffixed-lineinternettransmissionnetworksfordatatransfersfrom2000to2015concludedthatelectricityintensity(inkWh/GB)decreasedbyhalfapproximatelyeverytwoyearsoverthattimeperiod(Aslanetal.,2018).22CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.5.ICT’scarbonfootprintin2020TheITUTelecommunicationStandardizationSector(ITU-T)initsrecommendationL.1470estimatedtheICTsector’scarbonfootprintfor2015at740MtCO2e,includingembodiedemissions.Thisequatestoapproximately1.3%ofglobalgreenhousegas(GHG)emissions(ITU,2020).Thisisalmostfivetimessmallerthantheglobalfootprintoftheironandsteelsector,andsmallerthanmanyotherlargeindustries(seeFigure3).Figure3.Globalgreenhousegasemissions(MtCO2e)byindustry,2014(ITIF,2020)Withinthisfootprint,end-userdevicesaccountforthegreatestportionofemissions(401MtCO2e),followedbynetworks(198MtCO2e)anddatacentres(141MtCO2e).ThebreakdownofICT’scarbonfootprintbythesector’sdifferentcomponentsisshowninFigure4.Iron&steelChemicals&plasticsCementAluminiumRefiningMachineryPulp&paperCeramicsICTsectorFood&tobacco06947307548369379501,1092,5453,3473,487DatacentersDatatransmissionnetworksUserdevices2,0003,0004,00023CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsFigure4.CarbonfootprintofICT(2015)(AdaptedfromITU-TL.1470)TheITU-TL.1470recommendationsuggeststhatthesector’s2020carbonfootprintwouldremainatasimilarleveltothe2015value,basedontheavailabledatain2019andusingthesamebottom-upmethodologyasinMalmodin(2018a).Thisfindingsupportstheviewthatefficiencyimprovementsandreductionsinemissionsintensityofelectricitycontinuetoeffectivelystabilisetheemissionsofthesectorevenascomputingservicedemandsrise.Enterprisenetworks2%Mobilenetworks16%Fixednetworks9%Datacentres19%Userdevices54%24CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors1200100080060040020002.5.1HistoricalcarbonfootprintofICTThecarbonemissionsoftheICTsectorincreasedfromtheearly1990sto2010(GeSI,2008;Malmodinetal.,2013;Malmodin,2018a).However,thisemissionstrendhaslargelyplateaued,remainingrelativelystableoverthelastdecade,despitenetworkdatavolumescontinuingtogrowyearonyear.Malmodin(2020a)showsthattheICTemissionscurvehasflattenedandactuallydroppedfrom1.5%to1.3%ofglobalcarbonemissionsoverthepastdecade(Figure5),whiletheabsoluteemissionsofICThavefallenslightlyfromapeakof730MtCO2ein2015to710MtCO2ein2018,andtoabout690MtCO2ein2020.SomeotherstudieshaveoverestimatedtheglobalGHGfootprintoftheICTsector,particularlywhenprojectingintothefuture.Thisisoftenduetoacombinationofusinghistoricaldataandprojectingthatforwardusingassumedgrowthfigures.Forexample,Andrae&Edler(2015)modelledthreedifferentscenariosofICTenergyusingprojectionsforIPdatatrafficgrowthandenergyefficiencyimprovementtrends,withdifferentparametersforeachscenario.Thisshowedsignificantvariancebetweenthescenariosandsensitivitytotheparameters,butallthescenariosassumedanexponentialincreaseinenergyconsumptions,usingeffectivelyfixedcompoundannualgrowthrates(CAGRs).Insubsequentyears,Andraeupdatedthismodellingwithnewparameters,reflectingchangesintechnology(Andrae2017;2019;and2020).Eachtimethenewparametersresultedinatleast50%lowerprojectedenergyconsumptionthanthepreviousmodelling.ThisdemonstratesthedifficultyofmakingpredictionsforICTenergyandGHGemissions,andthedangersofrelyingonoldestimatesandapplyinggrowthratesbasedonhistoricaltrends.TheShiftProject(2019a)usedassumptionsbasedonAndrae&Edler(2015)resultinginanoverestimationoftheglobalGHGfootprintoftheICTsector.Belkhir&Elmeligi(2018)basedassumptionsonolderstudiesandextrapolatedtheseforwardsatfixedgrowthratestodetermineacurrentvalue,andthenextrapolatedthesefurtherforwardsatthesameratewellintothefuture.Forexample,theestimatesfordatacentreemissionsassumedafixed12%perannumgrowthrateextrapolatedto2040.AnextremeexampleofthisisHuber&Mills(1999)study,whichclaimedthat‘halfoftheelectricgridwillbepoweringthedigital-interneteconomywithinthenextdecade’.Clearly,20yearslaterthatisnotthecase.Figure5.HistoricalcarbonfootprintofICTsectorSource:(Malmodin,2020a)%ofglobalcarbonfootprintICTelectricity0.811.21.51.41.3ICTcarbonfootprintICTelectricity(TWh)ICTcarbonfootprint(MtCO2e)12001000800600400200025CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsThereissignificantevidencethatenergyandGHGemissionsarenotdirectlylinkedtodatatrafficgrowth.Thiscomesbothfromacademicstudies(suchasMalmodin,2020b;Kamiya,2020;Masanet,2020;Stobbeetal.,2015;Stobbeetal.,2021),andfromannualandsustainabilityreportsofvarioustelecommunicationnetworkoperatorsshowingyear-on-yeardecreasesinnetworkenergyintensity(e.g.Cogent,Telefónica,Vodafone).Overall,itisclearthatinternetdatatrafficalongwithdatacentredemandshavegrownsteadilyinthepastdecade.However,thisgrowthhasnotresultedinaproportionalgrowthintheenergyconsumptionofICT(Malmodin,2020a;Kamiya,2020).ThisdecouplingofenergyfromdatavolumesisillustratedbyFigure6below.Figure6.Globaltrendsininternetdatatraffic,datacentreworkloadsandenergyuse(2010-2019)Source:(IEA,2020a)IEAanalysisbasedon:Cisco,2015;Cisco,2018b;Cisco,2019b;Masanetetal.,2020a(AllFiguresindexedto1for2010)1412108642020102012201420162018InternettrafficDatacentreworkloadsDatacentreenergyuse26CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.6.1DatacentresDataCentreprocessingandenergyDatacentreworkloadsandenergyusehaveriseninthepastdecade,however,thereissomeuncertaintyaroundthemagnitudeofthistrendandthetotaloperationalcarbonfootprintofdatacentrestoday.Thesedifferencesinestimatesaredependentonthechoiceofmethodologyandboundaryforcalculations.Masanetetal.(2020a)estimatesthatsince2010,globaldatacentreenergyusehasonlyincreasedby6%since2010to205TWhin2018.Where,thissmallriseinenergyuseoccurreddespitelargeexpansionsofdatacentreworkloadsandcomputeinstancesgrowingby550%.Thestudyfollowedabottom-upapproachtoestimateglobalserverenergyuseincludingtraditional,cloudandhyperscalecentreswithinitsscope,butexcludednon-CPUcomputing,suchascryptocurrencydatamining,initsreporting.Malmodin(2020)estimatedthetotalelectricityuseofdatacentresat208TWhin2018,or0.9%offinalglobalelectricitydemand,equatingtoapproximately0.2%ofglobalcarbonfootprint.ThesestudiesalongwiththeITUallestimatesimilarvaluesfortheenergyconsumptionofdatacentres,rangedbetween200–208TWhoverthepastthreeyears.Datacentresactascentralisedhubsforprocessingandstoringdatausedforallinternetactivitiesorservices,includingvideostreamingbutalsocorporateandgovernmentdatabases,weatherforecasting,banking,websitesusedforcommerceandinformation.Thiscentralisedprocessingensuresgreaterefficiencyandbetterdistributionofinformation.Thegrowthofinternetconnectivityanddatatrafficvolumesthroughdatacentreshasdrivenuptheirworkloads,thatcontinuetoriseyear-on-year(Kamiya,2020).Conversely,Hintemann(2018)estimatedthatglobaldatacentreenergyconsumptionincreasedbyathirdbetween2010to2015,reaching287TWhin2015.Thistrendacceleratedfurtherinthesubsequenttwoyears,estimatedat350TWhby2017.ThisreportidentifiedthegrowingroleofBitcoindataminingindrivinguphigherenergydemandsonglobaldataprocessing.However,itisnotclearwhetherthiswasincludedwithinthescopeofthesecalculations(Hintemann,2018;HintemannandHinterholzer,2019).OneofthereasonsforthedifferencesinestimatesbetweenHintermannandMasanetistheassumptionsontheratethatmoreenergyefficienthyperscaleandcloudcomputinghasreplacedlessefficienttraditionaldatacentres.Thismaywellvaryregionally,andwouldbeaddressedbymorereliabledatasourcesfordifferentdatacentretypesandperformance(seeconclusionsinMasanet2020b).Risingdemandbeyond2020Thereiscommonagreementthatdatatrafficdemandsondatacentresareprojectedtocontinue,particularlyforclouddatacentretraffic,withCiscoprojectingthatthiscouldreach95%oftotaldatacentretrafficby2021,whichwouldrepresenta3.3-foldgrowthofcloudtraffic(CiscoGlobalCloudIndex,2018a).Thequestionishowwilltheseincreaseddataandcomputeworkloadstranslateintodatacentreenergyconsumptioninthefuture.Thereissignificantuncertaintyaroundthis.Oneofthekeyfactorsisbyhowmuchandhowquicklyclouddatacentresarereplacingtraditionallessenergyefficienton-premisesdatacentres.ThisuncertaintyisillustratedbytworecentreportsfortheEUCommission.AreportforDGEnergyontheimpactofICT(EUCommission,2020a)estimateddatacentreelectricityconsumptionoftheEU27memberstatesat40TWhin2020,risingto43TWhby2025.WhileareportforDGConnectoncloudcomputingtechnologies(EUCommission,2020b)expectsenergyconsumptionofdatacentresintheEU28toincreasefrom77TWhin2018to93TWhin2025.2.6.CurrentandhistoricalemissionsofICTsectorcomponentsThecurrentandhistoricalemissionsarenowconsideredinmoredetailbythesub-sectorswithintheICTsector’sboundary.Asmentionedabove,theseare:datacentres,networks,end-userdevicesandE&M.27CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsThesedifferencesindatacentreenergyuseprojectionsvaryduetobothboundarydefinitionsandtheinherentuncertaintyofpredictingfutureemissionsandtheimpactsofemergingtechnologies.Specifically,studiesvaryintheirassessmentofenergyefficiencytrendsandtowhatextenttheywillcontinuetonegatetheenvironmentalimpactsofgrowingdataworkloads.EnergyefficiencyofdatacentresAsmentionedpreviously,innovationinICThasenabledmoredatatobeprocessedandtransmittedwiththesameamountofenergyacrosstheentirevalue-chainofICT(servers,datacentresandnetworkequipment).Specifically,fordatacentres,theconsolidationandvirtualisationofdatacentreworkloadsintothecloudhasenabledsignificantefficiencygains(Masanet,2020a).Additionalefficiencygainshavebeenachievedthroughtrendsinend-userdevices,suchasshiftingfromdesktopstousemoreenergyefficientlaptops,tabletsandsmartphones(Malmodin&Lundén,2018a),orthroughchangesinscreendisplaytechnologiesthathaveenabledsignificantreductionsinpowerconsumption.TheseimprovementsinICThardwareefficiencyhavethereforecompensatedforthegrowthofnetworkdatavolumesandtheincreasednumberofconnecteddevices(Malmodin,2016;Malmodin2020a).Somestudiessuggestdatacentreefficiencyimprovementsarebeginningtoslow,asPUE(PowerUsageEffectiveness)hasplateauedsince2013(surveybytheUptimeInstitute,2019).However,thesefindingswerebasedonanunweightedaverageofsurveyeddatacentres,thereforerelyingonthescopeofthesurvey,andnotrecognisingthelargerimpactthathyperscaledatacentreswillhavecomparedtoasmall-scaledatacentre.Also,PUEperformanceonlymeasuresfacilityenergyuse,sodoesnotaccountforserverefficiency.Thiscanbemoreaccuratelymeasuredbymonitoringthetotalenergyuseofdatacentres(UptimeInstitute,2019).Conversely,industryexpertssuggestthatthegrowthinenergydemandthathashistoricallybeencompensatedbyefficiencyimprovements,mightcontinuetoholdinthefuturefordatacentres(Kamiya,2020;Masanet,2020a;Shehabi,2018;Koomey,2011b).Masanet(2020a)indicatestheremaybecapacityforfurthertechnologicalandinfrastructuralefficiencyimprovementsofdatacentres,thathavepreviouslyenabledthegrowthofinternetserviceswithonlyarelativelysmallincreaseinassociatedenergyconsumption.Intandem,datacentrescouldcontinuetoenhanceserverefficiencies(throughvirtualisationanduseofadvancedcoolingsystems)andfurthershifttowardsmoreefficientcloudcomputing.Additionally,ageneralshiftfrominefficient,smaller-scaledatacentrestohyperscaleandcloudcentreshasenabledgreaterefficiencyofscale,utilisingthemostadvancedcoolingandpowersavingtechnologytoreducetheirPUEandinfrastructurefootprint(RoyalSociety,2020;Masanet,2020a).Infrastructureenergysavingscouldcontinueinthefuture,ifhyperscaledatacentrescontinuetoreplacesmallerscaledatacentres.Thistrendissuggestedtocontinue,asthenumberofhyperscaledatacentreswereprojectedtonearlydoublefrom2016-2021,toaccountfor53%ofglobalserversby2021(EUCommission,2020a;Cisco,2018a).However,thisshiftindatacentreinfrastructuretohyperscalemaybeapproachingitslimitinthenextdecade,sotheaddedefficiencygainsmaynotcontinueindefinitely.Conversely,otherstudiessuggestthatenergyefficiencytrendsfordatacentreshaveslowedandmaynotholdbeyond2020(RoyalSociety,2020).Waldrop(2016)suggeststhatMoore’slawcouldbeimpededbythetechnical/physicallimitationsoftransistortechnology.However,Malmodin&Lundén(2018a)associatesthisdecelerationofefficiencywithatimelagbetweenresearchfindingsof,andtheimpactsof,efficiencygainsondataconsumption.Meanwhile,otherstudiessuggestthisefficiencybalancingactcouldbecomedisplacedbyelectricityconsumption(Lange,Pohl,andSantarius,2020;Bordage,2019).Itisalsoimportanttoconsidersoftware’spotentialforincreasingtherateofefficiencyimprovements.Onlyconsideringhardwareimprovementsgivesanincompletepicture.Althoughtechnologymaybeapproachingthephysicallimitsoftransistordensity,softwarecanprovideadditionalenergyefficiencyimprovementsthatwillcompensateforthehardwareenergyusetosomedegree(Leisersonetal.,2020).Otherstudiesalsosuggestthatthetrendsinimprovingenergyefficiencyfordatacentrescouldcontinue,Malmodin(2020a)claiming‘dataisafunctionoftechnology(efficiency)’.Specifically,identifyingopportunitiesforfurtherefficiencygainsthroughmobilenetworksexpansionshiftto5G(Malmodin,2020a)aswellasthecapacityforfurthertechnologicalandinfrastructuralefficiencyimprovementsofdatacentres(Masanet,2020b).Clearly,themagnitudeandlongevityoftheseenergyefficiencytrendsremainuncertainandanyprojectionsshouldbeunderstoodcarefullyinthelightoftheirassumptions.28CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsAgain,thereiscommonagreementthatnetworkdatatransmissionisgrowingrapidly.Malmodin(2020a)estimatedthatinternetdatatrafficgrewbyafactorof12inthepastdecade.While,Cisco(2019b)projectedathree-foldincreaseintotalIPtrafficbetween2017and2022,orannualgrowthrateof26%,withthetotalannualtrafficsurpassing3Zetabytesby2020.Despitethesignificantgrowthinnetworkdatatraffic,theenergyconsumptionassociatedwithnetworkstransmissionhasonlyslightlyincreasedinthelastfiveyears,(withsomenetworkoperatorsholdingenergyconsumptionataconstantlevel),andemissionsfallingduetoincreaseduseofrenewableelectricity(bothasaveragegridemissionsintensitiesfall,andasnetworkoperatorspurchaseincreasingamountsofrenewablesthroughPPAsandrenewabletariffs).2.6.2NetworkdatatrafficandenergyFigure7.Fixednetworkenergyintensity(LOGscale)Reproducedfrom:Aslanetal.,2018;Aslan,2020.Note:Datapoints‘Aslan(1-6)’arefromAslanetal.(2018);datapoint‘AslanEngD.’isfromAslan(2020).Solidlineshowstheregressionfordatapoints‘Aslan(1-6)’foryears2000to2015.Dashedlineshowsextrapolationoftheregressionlineto2019.Majornetworkoperatorsregularlypublishtheirenergyandemissionsdataintheirannualandsustainabilityreports,whichprovidesevidenceforthistrend.(Forexample,AT&T,BT,Cogent,Sprint,Telefónica,T-Mobile,Vodafone–seeFigure8andFigure9).Theelectricityintensityofdatatransmission,definedastheenergyconsumptionperdatavolume(kWh/GB)iscontinuingtofallyear-on-yearduetochangesintechnology.Aslanetal.(2018)projectedthattheelectricityintensityofdatatransmissionhalvedapproximatelyeverytwoyearsfrom2000to2015acrossICT-maturecountries(seeFigure7).ThisisinagreementwithMalmodin&Lundén(2018a)thatenergyintensitycanbedecoupledfromdatagrowth,becauseofthetechnologicalimprovementsofICTequipmentandthesubsequentenergyefficiencygains.Energyintensity10.001.000.100.01199820002002200420062008201020122014201620182020(kWh/GB)Aslan(1),6.80Aslan(2),0.68Aslan(3),0.16Aslan(4),0.14Aslan(5),0.023Aslan(6),0.06AslanEngD.,00129CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2016Figure8.TelefónicaenergyintensitykWh/GBFigure9.Cogentnetworkelectricityintensityindexbytraffic(indexedto2016)Source:datapointsfromTelefónica(2019);Telefónica(2020)Note:Telefónicaisamobilenetworkoperator,andtheenergyintensityisexpectedtobehigherrelativetofixed-linenetworks.Thischartispresentedtoshowthetrendindecreasingenergyintensity.Source:datapointsfromCogent(2020b)Note:Cogentreportthisinformationindexedrelativeto2016.kWh/GB0.4500.4000.3500.3000.2500.2000.1500.1000.0500.0002015201620172018201920201.00.80.60.40.20.0201720182019202030CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.6.3End-userdevicesUserdevicesmakeupthelargestportionofcarbonemissionsofICT.Theemissionsofend-userdevicesmakingupjustover50%ofICT’soverallfootprint(shownbyFigure10below)withthelargestportioncomingfromPCsandlaptops(Malmodin2020a).Malmodin(2018a)suggestedthattheemissionscontributionofuserdevicesdecreasedbetween2010-2015.Thisisduetotechnologyefficiencytrendsinuserdevices,andthetrendtousesmallerdevices(e.g.fromPCsandlaptopstotabletsandsmartphones).Changesintypesofuserdevices:ThereisanincreasingshiftinICTuserdevicepreferencesfromlargerPCsandlaptopstomobiledevices,supportedbyapps.Thisshifthasacceleratedduetothetechnologicaladvancesinmobiletechnology,enablingsmallerdevicestosupportversatilefunctionsandtomaintainhigherbitratespeedstostreamvideowithahigherquality.Thishasalsoledtoenergyefficiencyimprovementsasdesktopsarereplacedbylaptops,andlaptopsarereplacedbylessenergyintensivemobileandtabletdevices.Thistechnologyshiftisexpectedtocontinue,withwifiandmobiledevicesexpectedtoconstitute70%oftotalIPtrafficby2022,upfromaround50%in2019(Cisco,2019b).Figure10.BreakdownofICTsectoremissionsbycomponentReproducedfrom:Malmodin(2020a)0100200MtCO2e300400DatacentresNetworksUserdevicesEmbodiedcarbonfootprintUsephasecarbonfootprint31CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.6.4EntertainmentandMedia(E&M)WediscussheretheglobalcarbonfootprintofE&M,usingthenarrowemissionsboundarydefinitiongivenearlier(E&Mcomprisesallelectronicequipmentutilisedformediaandentertainmentpurposes,including:TVs,cameras,andotherE&Mconsumerelectronics,aswellasphysicalpapermediaandprinting).Theseemissionsareestimatedat640MtCO2egloballyfor2015(Malmodin,2018a),withthemajorityoftheseduetotheemissionsofTVs.TheseE&Memissionshave,historically,followedsimilarenergytrendstothoseofICT,drivenlargelybyincreasingenergyefficiencyofTVs.PreviousstudiessuggestedthatinSweden,by2010,bothICTandE&Memissionshadpeakedandremainedindependent/decoupledfromrisingdatatrafficvolumes,thatcontinuedbeyondthispoint(Malmodin&Lundén,2016).Malmodin&Lundén’s(2018a)follow-upstudyestimatedthattheglobalelectricityconsumptionofE&Mhaddecreasedto585TWhby2015,fallingby30%from2010levels(Figure11).E&M’scarbonfootprintreductionshavebeenenabledthroughenergyefficiencyimprovementsinhardwareandalsothroughashifttowardsdigitisedmedia,awayfrompapermediaproductionandotherphysicalmediasuchasdiscs,tapes,andharddrives.Inparticular,energyefficiencygainshavebeenachievedthroughimprovementsindevicedisplaytechnology.Specifically,televisiondisplaysconsumelessenergypersurfacearea,soaremoreenergyefficient,offsettingthecontinuedgrowthofaverageTVpanelsizes.Thisshifttomoreefficientdeviceshasresultedindecreasedenergyuseofhouseholds,asindicatedbytheFraunhoferInstitutestudyofenergyuseforconsumerdevicesinUShomes(Singhetal.,2019).However,thefuturetotalenergyuseforTVsislikelytoleveloff,andmayincreasewithincreasingnumberofTVsperhouseholdandincreasingscreensize.Thefuturetrendsmayvarysignificantlybycountry,withdifferenttrendsemergingbetweenEuropeandtheUSA.Whiletheoperationalelectricityenergyuseofpapermediadevices(printers,faxes,photocopiersetc)halvedbetween2010-2015,asaresultofreducedpaperuseinofficesandfewerdevicesrequiredfornormaloperations(Malmodin&Lundén,2018a).Figure11.CarbonfootprintofICT(includingE&Msector)(2015)(AdaptedfromMalmodin&Lundén,2018a)ICTnetworks13%Datacentres&enterprisenetworks12%Userdevices29%E&M46%TherehavebeennocomprehensiveglobalstudiesofthecarbonfootprintofotheraspectsoftheE&Msector,thusthenarrowdefinitionusedabovefordefiningtheglobalemissionsofE&M,whichexcludescinemas,theatres,andotherarenasorphysicalsiteevents(e.g.,sports),andcontentcreationsuchasfilmandTVproduction.Althoughthereexistinitiativesforthecarbonfootprintingofindividualevents,andoffilmproduction,mostnotablythatofAlbert.22https://wearealbert.org/production-handbook/production-tools/32CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors100,00010,00010001001010.10.011990199520002005201020152020Globaldatatraffic(Gbps)2.7.GrowthofglobaldataconsumptionTheexpansionofglobalinternetnetworks,connecteddevicesanduserconsumptionhasledtoasteadygrowthofdatavolumes.Thistrendhasbeenapparentsincetheearly1990s.Between1990–2000,globaldatatrafficroseexponentially,slowingslightlybetween2010–2020,showinga12-foldincreaseindata.Itisprojectedtocontinuegrowthbutatalowerrate,estimatedbetweenafactoroffiveto10overthenextdecade(Malmodin,2020a).Source:Malmodin&Lundén(2018a)Figure12.Totaldatatraffic1990-20151995-20153xenergy10xusers1,000,000xdatatraffic(10,000xwithalsovoicedata)ICTsectoroperationalelectricityconsumption(TWh)Operationalelectricityconsumptionper“user”(kWh/user)Numberof“users”(connectionsorsubscriptions)ThishasfuelledconcernsthatglobalICTemissionsandenergyuseisfollowingasimilartrend.However,asmentionedpreviously,increaseddatausagedoesnotcorrelatedirectlywithincreasedenergyuseoftheICTsector(asshownbyFigure12below).Overalltrendsindatatrafficareprojectedtocontinueincreasing:totalIPtraffic(Figure13)andmobiledatatraffic(Figure14).33CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsFigure14.Globalmobiledatatrafficprojections(2020–2026)Source:Cisco(2019b)Exabytespermonth12215620125431939620172018201920202021202205010015020025030035040026%CAGR2017-2022Figure13.CiscoVNIGlobalIPtrafficforecast(2017–2022)Reproducedfrom:Ericsson(2020)Globalmobiledatatraffic(EBpermonth)250200150100500201520162017201820192020202120222023202420255G2G/3G/4G34CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsIncreaseinvideostreamingAlongwithashiftindevicepreference,therehasbeenashiftinthenatureofdevicedatausageandactivity.Inparticular,thedemandforvideostreaming,gamingandtelecommunicationsisgrowingsubstantially.CiscoregularlypublishesitsVNIforecastsoftotalIPdatatrafficprojections,withthelatestanalysispublishedforthefive-yearperiod2017-2022(Cisco,2018c;Cisco,2019b).CiscodefinesInternetTrafficas‘allIPtrafficthatcrossesaninternetbackbone’,andtotalIPtrafficincludesInternetTrafficplusManagedIPTraffic,whichisdefinedas‘corporateIPWANtrafficandIPtransportofTVandVoD’.ThebreakdownofglobalIPtrafficisshowninFigure15,andthecompoundannualgrowthratesfortheseparatedatatypesisshowninTable2.(Note,Ciscousesthecriteriafor“long-formvideo”as‘videositeswhoseaverageviewingtimeislongerthanfiveminutes’).ItcanbeseenfromTable2thatlong-forminternetvideo-on-demandhasaCAGRof33%for2017-2022,andin2020wasforecastatabout40%oftotalIPtraffic,orabout45%oftotalinternettraffic.Whilefortotalvideotraffic,CiscoforecaststhatglobalIPvideotrafficwillbe82%ofallIPtrafficin2022,upfrom75%in2017,ataCAGRof29%for2017-2022(Cisco,2019b).NotethattheseCiscoforecastsarefromanalysiscarriedoutin2018,soarenotconsideringanyimpactofincreasedvideostreamingorinternettrafficfromCOVID.Figure15.GlobalIPdatatrafficbytypeSource:CarbonTrustanalysisofCiscoVNIGlobalIPTrafficForecast,2017–2022(Cisco2018c)450.0400.0350.0300.0250.0200.0150.0100.050.00.0201720182019202020212022GamingFilesharingWeb/DataIPVOD/ManagedIPVODVideosurveillanceLiveinternetvideoLong-forminternetVoDShort-forminternetVoDExabytespermonth35CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectorsTable2.Compoundannualgrowthrates(CAGR)forIPTrafficdatatypes(2017-2022)DatatypeCAGR(2017-2022)Short-forminternetVoD20%Long-forminternetVoD33%LiveinternetVideo75%Videosurveillance45%IPVOD/ManagedIPVOD12%Web/Data20%Filesharing-2%Gaming55%TOTAL26%Source:CarbonTrustanalysisofCiscoVNIGlobalIPTrafficForecast,2017–2022(Cisco2018c;andCisco2019b)Note:somefigureshavebeenrounded.OtherestimatesoftheproportionoftrafficthatisduetovideostreamingmayuseslightlydifferentdefinitionscomparedtotheCiscoanalysis.Sandvinereportsthatin2020allvideostreamingaccountedforover65%ofallcellularnetworktrafficbyvolumeand58%oftotalinternettraffic(Sandvine,2020a;Sandvine,2020b).Concernoverincreasingvideostreaminghasgeneratedalotofmediaattention,incorrectlyassertingthatincreaseddatatraffictrendsforvideostreamingarecausinganincreaseintheICTsector’scarbonfootprint.Internetaccess&IoTtrendsTheICTsectorhasundergonehugeexpansionsincetheearly1990s,increasingitsapplicationacrossmanysectorsandindustries.Thisshifttodigitalisationhasbroughthugebenefitstosocietysuchasuniversalaccessofinformation,easeofcommunication,efficiencyofprocessaswellasenablingemissionssavingsacrossothersectors.Assuch,globalinternetaccessibilityanduserconnectivityhasgrown.Thistrendisprojectedtocontinue,with66%oftheglobalpopulationexpectedtohaveaccesstotheinternet,and70%willhavemobileconnectivityby2023.Additionally,theglobalaverageconnecteddevicesperpersoncouldreach3.6by2023,upfrom2.4in2018(Cisco,2020).WhiletheapplicationoftheInternetofThings(IoT)/M2Mconnectionsisprojectedtoreach4.4billionby2023(representingafour-foldannualincreasefrom2018)(Cisco,2020)andgrowexponentiallyto2026ataCAGRof23%forcellularIoTconnections(Ericsson,2020).Additionally,thegrowthofconnectedIoTdevicessuchassmartmetersorvideosurveillanceareprojectedtoaccountforoverhalfofallconnecteddevicesby2023(Cisco,2020).ThesetrendswillinevitablyplacegreaterenergyanddatademandsontheICTsector.However,studiessuggestthatIoTdevicesuseverysmallamountsofenergy,thustheadditionofextraconnectionpointsanddeviceswouldonlymarginallyincreasetheoverallcarbonfootprintoftheICTandE&Msectorsby2020(Malmodin&Lundén,2018a)andby2025(Das&Mao,2020).36CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors36CarbonimpactofvideostreamingBackground–globalenergyandcarbonimpactoftheE&MandICTsectors2.8.FuturetrendsinICTcarbonemissionsfootprintForecastingfutureICTemissionsisfraughtwithuncertainty.Emissionsprojectionestimatesrelyontheextrapolationoffiguresfrompasttrends,usinghistoricalICTdata.However,thespeedofICTtechnologicaladvancementandgainsinenergyefficiencyusuallyoutpacetheseestimates,renderingassumptionsoutdated.Asmentionedpreviously,Andrae&Edler(2015)paper’sprojectedworstcasescenariohashadtoberevisedinAndraeetal.(2017),Andrae(2019),andinAndrae(2020)inordertoaccountforthisever-changingenvironment.Therefore,theaccuracyofICTfutureemissionsprojectionsisthesubjectofuncertaintyandscrutiny.ItisdifficulttoprojectfarintothefutureoftheICTsector,duetotherapidnatureoftechnologyadvancementsandimprovementsinenergyefficiency.Thereisdebateaboutwhetherthistrendwillcontinuetohold.Somestudiessuggestthatcomputingefficienciesmayreachtheirtechnologicallimits(Waldrop,2016),however,todate,thishasyettobeproven.While,otherstudiespositthatenergyefficiencytrendswilleventuallypeak,andbeyondthispoint,theenergyimpactsofincreasingdatatrafficandconnecteddeviceswillnolongerbenegatedbyincreasedenergyefficiency.EmergingtrendsinICT,beyond2020,addtotheuncertaintyofforecastingfutureemissions.Thesetrendsincludemachinelearning,blockchainandcryptocurrencies,IoTand5Gmobilenetworks(IEA,2020a).Thewiderimpactsofthesetrendsonenergydemandareunclear.Theywillcontributesubstantiallytodatacentreworkloadsandinternetdatatrafficbutalsopresentopportunitiesfortechnologyefficiencygains.Forexample,energyimprovementsarepredictedforfuturemobilenetworks,asmoreefficient4Gnetworksbegintodominatethemarket.Thistrendcouldacceleratefurtherby2025,as5Gnetworksemerge(Cisco,2020),withleadingoperatorssuggesting5Gcouldbebetweentento20-foldmoreenergyefficientperbytethan4G(Huawei,2019).Whetherenergyefficiencyimprovementswilloutpaceincreasesindatatrafficisunclear–anumberofmobilenetworkoperatorsexpectthattherolloutof5Gwillincreasetheirtotalenergyconsumption.Evenifthereareoverallincreasesinenergyconsumption,thetotalemissionsoftheICTsectorarelikelytofallsignificantly,aslargernumbersofnetworkanddatacentreoperatorsmovetousing100%renewableelectricity,andthenationalelectricitygridscontinuetodecarbonise.Torelatethisbacktothesubjectofthiswhitepaper–videostreaminghastouchpointswithICTalongthevideodeliveryprocess,andisthereforedependentonICT.However,videostreamingisonlyoneoftheservicesthatreliesontheICTinfrastructure,andhasdifferentdemandsandtrendsfromotherrapidlydevelopingareassuchasIoTandautomation.CarbonimpactofvideostreamingMethodologyandapproach373.MethodologyandapproachThissectionprovidesasummaryofthemethodologicalapproachusedtoestimatethecarbonemissionsimpactofvideostreaming.Anoverviewofthisapproach,referredtoastheconventionalapproachwithinthispaper,isdiscussedincludingthecharacteristicsoftheconventionalapproach,theboundaryusedtodefinethefootprintandabriefexplanationofallocationanditsroleinthisestimationtechnique.Followingthisoverview,thekeyparametersusedinthisestimationapproachandahigh-levelexplanationoftheirderivationisprovided.38CarbonimpactofvideostreamingMethodologyandapproach3.1.OverviewoftheconventionalapproachCharacteristics•Utilisesaveragetransmissionnetworkenergyintensityestimatesderivedfromacademicliterature•Reliesonanallocationapproachtoattributetheenergyconsumptionofasharednetworktotheservicesthatusethenetwork•Transmissionnetworkenergyconsumptionisallocatedtostreamingbasedondatavolumetransmittedforonehourofstreaming•DatacentreandusermediadevicesenergyconsumptionisallocatedbasedonviewingdurationIntendeduse•Organisationalfoot-printingforvideostreamingserviceproviders•Network/systemlevelcarbonfootprintestimationStrengths9Wellunderstoodandacceptedbytheindustry9Accountsforallenergyconsumptionacrossthenetworkwhenscaledtonetworklevel9StraightforwardaccountingandallocationapproachLimitationsXRepresentativeofaparticularnetworkandperiodoftimeXSensitivetocharacteristicsofthetransmissionnetworkconsideredforestimation.Thesenetworkcharacteristicsincludenetworkequipmentefficiency,quantityofnetworkusersanddatatrafficXUnsuitableforestimationofthemarginalcarbonimpactduetoachangeinlevelofservice,suchasachangeinvideoquality39CarbonimpactofvideostreamingMethodologyandapproachTherearesensitivitiesandlimitationstotheconventionalapproach,whichshouldbeunderstoodaswell.Theconventionalapproachisrepresentativeofaparticularnetworkandperiodoftime.Forthisstudy,theparametersusedarecharacteristicofmaturetransmissionnetworksin2020suchasthosefoundinWesternandNorthernEurope.Asnetworksarecontinuallyevolving,theparametersusedinthisapproachshouldbeupdatedtoreflectthecharacteristicsofthenetworkduringtheperiodoftimebeingevaluated.Transmissionnetworkenergyintensitiesinparticulararequicklyoutdated,asdiscussedintheprecedingsectionsofthispaper.Additionally,whiletheconventionalapproachissuitablefororganisationalfoot-printingpurposes,itisunsuitableforestimationandanalysistoassessthemarginalcarbonimpactduetoachangeinlevelofservice,suchascomparingthecarbonemissionsofstreamingindifferentresolutions.Thisisprimarilyduetotheconventionalapproach’srelianceonaveragetransmissionintensities,whichdonotreflectthedynamicsofnetworktransmissionequipmentasnetworkloadchanges.Theconventionalapproachmaybedescribedasanaverageenergyintensityapproach,asitutilisesaveragetransmissionnetworkenergyintensity3toestimatethenetworkenergyconsumptionattributabletovideostreaming.Furthermore,asthetransmissionnetworkisaninterconnectedsystemofnetworkusersandservices,theenergyconsumedbythenetworkissharedamongthevarioususersandservicesthataccessthenetwork.Inordertodividethenetworkenergyconsumptionamongthevarioususersandservices,anallocationapproachisemployed.Inthiscase,thetransmissionnetworkenergyconsumptionisallocatedbasedonthevolumeofdatatransmittedacrossthenetworkbythevideostreamingservice,hencetheutilisationoftransmissionnetworkenergyintensity.Animplicationofthisisthatenergyconsumptionofidlenetworkequipmentisallocatedbasedontheinternetservice’stransmitteddatavolume,withhigherbitrateserviceslikevideostreamingreceivingalargershareofidleenergy.Otherallocationapproachesareconceivable,suchasnetworkenergyallocationbasedonshareofpeaknetworktrafficorbasedondurationofuse,butallocationbasedondatavolumeisthetypicalfocusinacademicstudiesandthereforethemostwidelyunderstood.Astheenergyintensityisderivedatthenetworklevel,theconventionalapproachiswidelyacceptedasanappropriateestimationmethodologyfororganisationalfoot-printingfororganisationsthatprovideinternetservices,suchasvideostreamingproviders,andifemployedacrossanentiretransmissionnetwork,wouldaccountforalloftheenergyconsumed.Thisapproachisalsoappropriatetouseforanetworkorsystemlevelestimationofvideostreaming’scarbonfootprint.Forthisstudy,theconventionalapproachhasbeenadaptedtoestimatethecarbonimpactforonehourofvideostreaming.3Transmissionnetworkenergyintensityrelatesthenetworkenergyconsumedtoametric,inthiscasethedatavolumetransmittedthroughthenetwork.TheunitofmeasurefortransmissionnetworkenergyintensityiskWh/GB,wherekWhrepresentstheenergyconsumedbythetransmissionnetworkandGBrepresentsthedatatransmittedoverthenetwork,ingigabytes.Transmissionnetworkenergyintensityestimatesaregenerallyderivedandpublishedthroughacademicliteratureusingatop-downevaluationofenergyconsumptionofatransmissionnetworkandthevolumeofdatathatistransmittedacrossthenetworkoveraspecifiedperiodoftime.40CarbonimpactofvideostreamingMethodologyandapproach3.2.ApproachboundaryTheconventionalapproachdrawsitsboundaryaroundthecomponentstagesofthevideostreamingprocess:datacentres,transmissionandend-userdevices,asshowninFigure16below.Thelifecycleboundaryforthisapproachincludesonlythein-useelectricityconsumptionofeachvideostreamingprocesscomponentandexcludestheembodiedcarbonandend-of-lifeemissionsofdatacentres,networkequipmentandend-userdevices.Thedatacentresprocesscomponentencapsulatesboththecloudstorageandencodingservicesandcontentdeliverynetworks(CDN).Insimplisticterms,datacentresarewherethevideodataarestoredandretrievedwhenauserstreamsvideooverthenetwork.CDNsserveaparticularregionbystoringmoreproximatecopiesoftheoriginalvideodatawhichhelpstoreducecongestiononthenetworkandimprovetransmissiontimes.Withintheconventionalapproachmethodology,theenergyconsumptionrelatedtodatacentresisestimatedwithanenergyintensityperviewinghourof1.3Wh/hour,whichisderivedfromaselectionofDIMPACTmembersbasedonmeasureddatain2020.Theenergyconsumptionrelatedtodatacentresisestimatedasshowninequation(1).3.3.DatacentresWhereEDCisthedatacentreenergyconsumption,IDCistheenergyintensityofdatacentresandDisthedurationofvideostreaming,inthiscaseonehour.ThesecomponentstagesarediscussedinfurtherdetailbelowandassumptionsrelatedtoeachcomponentarepresentedinTable3.EDC=IDC×D(1)41CarbonimpactofvideostreamingMethodologyandapproachGlobalparametersEV=energyconsumptionofthevideostreamingprocessCV=carbonemissionsofthevideostreamingprocessEFg,n=electricalgridemissionfactorforregion,nD=durationofvideostreamingR=datatransmissionrateDataCentres&ContentDeliveryNetworkEDC=datacentreenergyconsumptionIDC=energyintensityofdatacentresTransmission–Network(Core&Access)EFN=fixednetworkenergyconsumptionIFN=energyintensityoffixednetworktransmissionPFN,i=proportionofviewingtimeoverfixednetworkatdatatransmissionrateRirelativetotheentirestreamingserviceEMN=mobilenetworkenergyconsumptionIMN=energyintensityofmobilenetworktransmissionPMN,i=proportionofviewingtimeovermobilenetworkatdatatransmissionrateRirelativetotheentirestreamingserviceTransmission-HomerouterEHR=homerouterenergyconsumptionIHR=energyintensityofhomeroutertransmissionEnd-UserdevicesEVD=viewingdeviceenergyconsumptionWs=averagepowerconsumptionofscreensWp=averagepowerconsumptionofperipheralsPs,i=proportionofviewingtimeatWs,irelativetotheentirestreamingservicePp,i=proportionofviewingtimeatWp,irelativetotheentirestreamingserviceParameterdefinitions42CarbonimpactofvideostreamingMethodologyandapproach3.4.TransmissionnetworkThenextcomponentofthevideostreamingprocessisthetransmissionnetwork,whichincludesenergyconsumptionoverbothfixedandmobilenetworks,whichcanbefurtherbrokendownintocorenetworktransmission,accessnetworktransmissionandsubscriberpremisesequipment(e.g.homewifirouters).Figure16.Videostreamingprocessmapindicatingthecomponentsthatmakeupthevideostreamingprocess.TheassociatedenergyintensitiesofthetransmissionnetworkelementsareincludedCorenetworksHometerminalsandroutersWiredaccessnetworksCellularaccessnetworksCloudstorageandencodingSubscriberpremisestransmissionPeripheralsScreensInternettransmissionContentDeliveryNetworkUsermediadeviceDatacentresTransmissionConventionalapproachEnergyallocatedbasedontotalviewinghoursCoreandaccessnetworksFixednetworkenergy=0.0065kWh/GBMobilenetworkenergy=0.1kWh/GBHomerouters0.025kWh/GB(onlyappliestofixednetworks)EnergyallocatedforonehourofstreamingFortheconventionalapproach,transmissionnetworkenergyintensitiesarederivedfromacademicliteratureforfixedandmobilenetworks.Theseestimatesincludeboththecoreandaccessnetworkelementsinanaggregatedfigure.43CarbonimpactofvideostreamingMethodologyandapproachTable3.ConventionalapproachassumptionsbystreamingprocesscomponentStreamingcomponentConventionalapproachassumptionsDatacentres&ContentDeliveryNetworkEnergyintensity,IDC(2020)=1.3Wh/hrDerivedfromaselectionofDIMPACTmembers,basedonmeasureddatain2020Transmission–Network(Core&Access)Fixednetwork•Energyintensity,IFN(2020)=0.0065kWh/GB•DerivedfromAslanet.al,2018usingtheregressionanalysispresentedinthepaperMobilenetwork•Energyintensity,IMN(2020)=0.1kWh/GB•SourcedfromPihkolaet.al,2018Transmission-Homerouter•Energyintensity,IHR(2019)=0.025kWh/GB•Basedon10Whomerouter(onlyusedwithfixednetworks)andaveragehouseholdfixednetworkdataconsumptionderivedfromOfcomfigures•OnlyapplicabletofixednetworkviewingDatatransmissionratesFixednetwork•Standarddefinition(SD):2.22Mbps(1GB/hr)•Fullhighdefinition(FHDorHD):6.67Mbps(3GB/hr)•Ultra-highdefinition(UHDor4K):15.56Mbps(7GB/hr)Mobilenetwork•Savedatasetting:0.37Mbps(0.17GB/hr)•Automaticdatasetting:0.56Mbps(0.25GB/hr)•Maximumdatasetting:6.67Mbps(3GB/hr)ThesefiguresarederivedfrompublishedNetflixfiguresondatausage(Netflix,2021)End-userdevices•Reasonableestimatesofaveragepower(W)forspecificdevices,seeAppendixfordetails•Thestandbytimeofend-userdeviceswasnotincludedinthisanalysis44CarbonimpactofvideostreamingMethodologyandapproachThefixednetworkenergyintensityusedintheconventionalapproachis0.0065kWh/GB,whichisrepresentativeoffixednetworkenergyintensityin2020.Thisfigureisderivedthroughacademicliterature(Aslanet.al,2018),wheretransmissionnetworkenergyintensitieswereevaluatedandconsolidatedfromanumberofacademicstudiestogenerateestimatedfixednetworkenergyintensityfrom2000to2015.Aregressionanalysiswasperformedandthisregressionwasusedtoextrapolatefixednetworkenergyintensityto2020,resultinginthefigureusedintheconventionalapproach.Theenergyconsumptionfromfixednetworktransmissionisrepresentedbelowbyequation(2)forasingleviewingscenarioandbyequation(3)forastreamingservicewithamixofbitrates.EFN=IFN×D×R(2)EMN=IMN×D×R(4)EFN=IFN×D×Ri×PFN,in∑i=1(3)EMN=IMN×D×Ri×PMN,i(5)WhereEFNisthefixednetworkenergyconsumption,IFNistheenergyintensityoffixednetworktransmission,RisthedatatransmissionrateandPFN,iistheproportionofviewingtimeoverfixednetworkatdatatransmissionrateRirelativetotheentirestreamingservice.Themobilenetworkenergyintensityusedintheconventionalapproachis0.1kWh/GBandisrepresentativeofmobilenetworkenergyintensityinFinlandin2020.Thisfigureissourcedfromacademicliterature(Pihkolaetal.,2018),wheremobilenetworkenergyintensitywasestimatedusingpubliclyreportedenergyconsumptionfiguresfrommobilenetworkoperatorsanddatatrafficfiguresfromtheFinnishCommunicationsRegulatoryAgency(FICORA).Aregressionanalysiswasthenperformedtoestimatethemobilenetworkenergyintensityin2020.Theenergyconsumptionfrommobilenetworktransmissionisrepresentedbelowbyequation(4)forasingleviewingscenarioandbyequation(5)forastreamingservicewithamixofbitrates.WhereEMNisthemobilenetworkenergyconsumption,IMNistheenergyintensityofmobilenetworktransmissionandPMN,iistheproportionofviewingtimeovermobilenetworkatdatatransmissionrateRirelativetotheentirestreamingservice.n∑i=145CarbonimpactofvideostreamingMethodologyandapproachTheremainingelementofthetransmissionnetworkisthesubscriberpremisesequipment,whichisrepresentedbyhomeroutersintheconventionalapproachandonlyappliestovideostreamingoverfixednetwork.Thehomerouterenergyintensityusedintheconventionalapproachis0.025kWh/GBwhichisderivedbasedontheannualenergyconsumptionofa10Whomerouterandfixednetworkdataconsumptionpercapitafiguresrelatingto2019andpublishedbyOfcomin2020(Ofcom,2020a).Theenergyconsumptionofhomeroutersisrepresentedbelowbyequation(6)forasingleviewingscenarioandbyequation(7)forastreamingservicewithamixofbitrates.EHR=IHR×D×R(6)EHR=IHR×D×Ri×PFN,i(7)WhereEHRisthehomerouterenergyconsumptionandIHRistheenergyintensityofhomeroutertransmission.n∑i=146CarbonimpactofvideostreamingMethodologyandapproachEVD=(Ws+Wp)×DCV,n=EV×EFg,nEV=EDC+EFN+EMN+EHR+EVD(8)(10)EVD=(Ws,i×Ps,i+Wp,i×Pp,i)×Dn∑i=1n∑i=1(9)(11)3.5.End-userdevicesThefinalcomponentofthevideostreamingprocessincludedintheconventionalapproachisend-userdevices,whichincludesscreens(e.g.TVs,laptops,smartphones)andperipherals(e.g.set-topboxesandgamingconsoles).Aselectionofend-userdeviceswasresearchedandareasonableestimateofaveragehourlyenergyconsumptionwasassociatedwitheachdevice,seeAppendixfordetailedinformation.Theenergyconsumptionofuserdevicesisrepresentedbelowbyequation(8)forasingleviewingscenarioandbyequation(9)forastreamingservicewithamixofdevices.Intotal,theenergyconsumptionofvideostreamingusingtheconventionalapproachisthesumoftheenergyconsumptionofthevideostreamingprocesscomponents,asshowninequation(10)belowandtheemissionsareestimatedbyapplyinganemissionfactorforgridelectricity,asshowninequation(11).WhereEVistheenergyconsumptionofthevideostreamingprocess,CVisthecarbonemissionsofthevideostreamingprocessandEFg,nistheelectricalgridemissionfactorforregion,n.WhereEVDistheviewingdeviceenergyconsumption,Wsistheaveragepowerconsumptionofscreens,Wpistheaveragepowerconsumptionofperipherals,Ps,iistheproportionofviewingtimeatWs,irelativetotheentirestreamingserviceandPp,iistheproportionofviewingtimeatWp,irelativetotheentirestreamingservice.TheconventionalapproachasdescribedaboveisusedtoestimatetheenergyandemissionsofvideostreamingandtheanalysisispresentedintheResultssection.Arepresentativemixofdeviceswasdevelopedwhichdefinestheproportionsintheprecedingequationsandfacilitatestheestimationofvideostreaming’scarbonimpactatthesystemlevel.Thedevicemixdefinestheproportionofviewingtimeforeachcombinationofnetworktype,datatransmissionrate,screenandperipheralandcanbefoundintheAppendix.CarbonimpactofvideostreamingResultssummary474.Resultssummary48CarbonimpactofvideostreamingResultssummaryThissectionpresentstheresultsoftheassessmentofthecarbonimpactperhourofvideostreaminginEuropein2020alongsideanalysisillustratingthemaindriversofvideostreaming’scarbonimpact.EuropewasselectedastheprimaryregionofanalysisasthemodellingparametersusedarerepresentativeofdevelopednetworksinWesternandNorthernEurope.Topresenttheanalysisinthissection,includingthequantityofcarbonemittedperhourofvideostreaminginEurope,theconventionalapproachisused.Insummary,thissectionhighlightsthefollowingkeyinsightsfromtheanalysisperformedforthiswhitepaper:•TheamountofcarbonemittedperhourofvideostreaminginEuropeissmall•Theelectricalgrid’scarbonintensityhasacriticaleffectonthecarbonimpactofvideostreaming•EmissionsfromuserdevicesareanimportantconsiderationinthevideostreamingprocessThisaveragefigurereflectstheEuropeanaveragegridintensity(2020IEAgridfactors,gridyear2018(IEA,2020b)),anassumedrepresentativemixofend-userdevices(seeAppendix),andthemodelledaveragebitrate(6.40Mbpsor2.88GB/hour)basedontherepresentativedevicemix.TheequivalentenergyconsumptionisaEuropeanaverageof188Whperhourofvideostreamingusingtheconventionalapproach(Figure18).TheEuropeanaveragecarbonemissionsperhourofvideostreamingfortheyear2020hasbeenestimatedtobe56gCO2e/hourvideostreamingusingtheconventionalapproach,asshowninFigure17.ViewingdeviceTVperipheralHomerouterNetworktransmissionDatacentresViewingdeviceTVperipheralHomerouterNetworktransmissionDatacentres6050403020100gCO2e/hourAveragemixFigure17.Estimatedemissionsfromonehourofvideostreaming(Europeanaveragein2020)Figure18.Estimatedenergyconsumptionfromonehourofvideostreaming(Europeanaveragein2020)4.1.TheamountofcarbonemittedandenergyconsumedperhourofvideostreaminginEuropeissmallAveragemix200180160140120100806040200Wh/hour49CarbonimpactofvideostreamingResultssummaryThebreakdownofemissionsandenergyperhourofstreamingfromtheresultsaboveyieldsthefollowing:Datacentres(includinghosting,encodingandCDNs),accountforlessthan1gCO2e/hourandapproximately1Wh/hour,representingroughly1%oftotalemissionsandenergy.Networktransmission(coreandaccess)accountsfor6gCO2e/hourand20Wh/hour(10%oftotalemissionsandenergy).Homeroutersaccountfor21gCO2e/hourand71Wh/hour(38%oftotalemissionsandenergy).Finally,end-userdevicesaccountfor28gCO2e/hour(25gCO2e/hourfromviewingdevicesand3gCO2e/hourfromperipherals)and96Wh/hour(86Wh/hourfromviewingdevicesand10Wh/hourfromperipherals),whichmakesuptheremaining51%oftheemissionsandenergyfromvideostreaming.Theseresultsshowthattheemissionsandenergyconsumptionfromonehourofvideostreamingaresmall.However,itshouldbenotedthatthesefiguresmustbeusedwithcare,andarenotdesignedtobeusedasrepresentativefiguresforanygivenscenario.Thesefiguresarederivedusingtheconventionalapproach,withitsassociatedintendedusesandlimitationsasdescribedintheMethodologysection.Astheinternetoperatesasanetwork,itsenergyconsumptionisinherentlysharedbetweenawiderangeofservicesandend-users.Therefore,toarriveatafigurequantifyingtheimpactofonehourofstreaming,energyconsumptionofthenetworkmustbe‘allocated’tovideostreamingusingsomeallocationapproach.Theallocationapproachisnotanexactscience,andthuswillnotbeatotallyaccuraterepresentationoftrueenergyconsumption.ThesefiguresarealsobasedonaspecificsetofparametersthatmodelarepresentativescenarioinEurope.Theparametersusedtomodelthenetworktransmissionarecharacteristicofthemostdevelopedandefficientnetworks,suchasthosefoundintheUKandNorthernandWesternEurope.Finally,themodellingparametersusedinthesefigures,representasnapshotintimeandanyprojectionsusingthesefiguresshouldbedonesowiththisunderstandingandcare.50CarbonimpactofvideostreamingResultssummary4.2.Thelocalelectricalgrid’scarbonintensityhasacriticaleffectonthecarbonimpactofvideostreamingThegeographicallocationofvideostreamingconsumptionhasacriticalinfluenceonthecalculatedcarbonimpactofanhourofvideostreaming.Thecarbonimpactofanhourofvideostreamingshowsconsiderablevariabilityfromcountrytocountry,duetothecalculationofvideostreamingemissionsusingcountry-specificelectricalgridemissionfactors.ThisisillustratedclearlyinFigure19,whereemissionsperhourofstreamingestimatedusingtheconventionalapproachwitharepresentativemixofend-userdevicesinFrance,Sweden,GermanyandtheUnitedKingdomarepresentedalongsidetheEuropeanaverage.Ofthecountriesincludedinthisanalysis,thehighestfigureisinGermanywith76gCO2e/hourandlowestfigureisinSwedenwith3gCO2e/hour.Incomparison,theEuropeanaveragecarbonimpactofvideostreamingisestimatedas56gCO2e/hour.ThisgroupofcountrieshasbeenchosentocomparewiththeEuropeanaverage,astheyfitthenetworkcharacteristicsrepresentedbytheconventionalapproachinthispaper,andrepresentarangeofgridintensitiesinEurope,whilstalsohavinglargepopulations.Theseresultsdemonstratethecriticaleffectadecarbonisedelectricalgridhasonemissionsfromvideostreaming.Energyconsumptionperhourofstreamingisalreadylowasaresultofthedistributednatureofthetransmissionnetworkbetweenmillionsofusersandwhencoupledwithanefficient,decarbonisedelectricalgrid,theemissionsimpactisverylow,suchasinthecaseofSwedenandFrance.80706050403020100gCO2eperhourofstreamingViewingdeviceTVperipheralHomerouterNetworktransmissionDatacentresFranceAveragemixSwedenAveragemixGermanyAveragemixUnitedKingdomAveragemixEuropeAveragemixFigure19.Emissionsfromvideostreamingbyregionin2020Thisdemonstrateshowimportantitisforgovernmentstocontinuetodrivethedecarbonisationoftheelectricitygrids.10376485651CarbonimpactofvideostreamingResultssummaryComparedtotheEuropeanaverage,Indiarepresentsacountrywithsignificantlydifferentcharacteristics,whichmaycontributetoamuchdifferentassessmentofthecarbonandenergyimpactsofanhourofvideostreaming.•IndiahasamuchhighergridintensityemissionfactorthanEurope:~2.5xEuropeanaverage•Betteraccesstofastinternetspeedsthroughmobilenetworksasopposedtofixed•Indiasurpassed500millioninternetconnectionsin2018,themajorityofwhicharemobile/dongleconnections(TechwireAsia,2020)•Morecarbonintensivemobilenetwork,duetogreaterusageofdieselgeneratorsandlesscleanenergyusedfornetworkequipmentatcellsites(CNBCTV18,2020)•India’senergyandcarbonimpactisthereforelikelytodiffersignificantlycomparedtotheaverageEuropeanimpact,drivenbyamuchmoremobilefocusedaveragemixofdevices,highergridintensityandlikelymoreenergyandemissionsintensivemobilenetwork•Encouragingly,telecommunicationcompaniesinIndia,suchasAirtel(Airtel,2016),arerecognisingthesecharacteristicsandtakingactiontoreducetheirrelianceondieselgeneratorsandtoincreasetheiruptakeofrenewableenergyAcasestudyforIndiaFurthermore,thisanalysisshowslittlevariationintheenergyconsumptionassociatedwithanhourofvideostreamingfromcountrytocountry,asthemodellingparametersfornetworktransmissionarerepresentativeofefficientEuropeannetworksgenerallyandarenotgranularenoughtorepresentthedifferencesinnetworktransmissionefficienciesbetweenEuropeannations.Inreality,networkrelatedenergyconsumptionislikelytovarydependingongeographicallocation,astheefficiencyofequipmentinthenetworkwillvary.Forexample,somenetworksindevelopedeconomieswillstillcontainamixoflegacyandnewerequipment,butmayhavegenerallymoreefficientequipmentcomparedtolessdevelopedeconomies(Coworker.com,2019).However,thisisnotreflectedinthesemodelledresults,asnetworkenergyconsumptionismodelledusingaveragenetworkenergyintensity.52CarbonimpactofvideostreamingResultssummary4.3.UserdevicesdrivetheimpactofvideostreamingEnd-userdevices,includingscreens,TVs,laptops,smartphonesandperipherals,likeset-topboxes,arethelargestcontributingcomponenttotheenergyandcarbonimpactofanhourofvideostreaming,accountingfor51%ofthetotalEuropeanaveragemixemissionsimpactusingtheconventionalapproach(seeFigure17andTable4).Withtheinclusionofhomerouters,thedevicesinthehomerepresent89%ofthetotalEuropeanaveragemixemissionsimpactperhourofvideostreaming.Attheperuserlevel,thecarbonemissionsassociatedwithdatacentresandnetworktransmissionequipmentissmall,duetotheallocationofthenetworklevelenergyconsumptionacrossasignificantpopulationofmillionsofvideostreamingusers.Thisresultsinlowemissionsandenergyuserelativetothatofend-userdeviceswhenassessingtheimpactofonehourofstreaming.Withtheexceptionofvideostreamingwithasmartphone,end-userdevices(screensandperipherals)accountforalargeportionofthecarbonandenergy-relatedimpactfromanhourofvideostreamingacrossarangeofusecases(Figure20andFigure21).Table4.BreakdownofemissionsandenergyconsumptionbyvideostreamingprocesscomponentforEuropein2020VideostreamingcomponentstageEmissions(gCO2e/hourstreaming)Energyconsumption(Wh/hourstreaming)%oftotalDataCentres<111%TransmissionNetwork62010%HomeRouter217138%TVPeripheral3105%Screens258646%Total56188100%Figure20.Emissionsfromvideostreamingbyviewingdevice(Europeanaveragesin2020)ViewingdeviceTVperipheralHomerouterNetworktransmissionDatacentres706050403020100gCO2e/hour5881656iPhoneCellularnetworkAutomaticdatasetting50”SmartTVFHD(1080p)LaptopSD(480p)Averagemix53CarbonimpactofvideostreamingResultssummaryFigure21.Proportionofemissionsbystreamingprocesscomponentbyviewingdevice(Europeanaveragesin2020)ViewingdeviceTVperipheralHomerouterNetworktransmissionDatacentresThisphenomenonismoreprominentinthecaseofamorepowerhungryend-userdevice,suchasthe50insmartTVscenario(streaminginFHDoverfixednetwork)presentedabove,whereend-userdevicesaccountfor51%ofemissionsandthehomerouteraccountsforanadditional38%ofemissions.Forthelaptopscenario(streaminginSDoverfixednetwork),thehomerouterislargestcontributortototalemissionsat45%,followedcloselybyemissionsfromthelaptopitselfatanestimated40%.Incomparison,device-relatedemissionswhenusinganiPhoneoveramobilenetworkconnectionaccountfor4%oftotalemissions.Theaveragemixscenariobroadlyreflectsthebreakdownofemissionsofthe50insmartTVscenario,astheaveragedevicemixassumesthat70%ofviewingoccursonTVs.100%90%80%70%60%50%40%30%20%10%0%Notably,contentencodingandhostingindatacentresandCDNshasarelativelysmallenergyandcarbonimpact.ThisisprimarilyduetothefactthatdatacentresandContentDeliveryNetworksarehighlyefficient,particularlyasdatacentresareincreasinglytrendingtowardshyperscalecapacity,andthusgainingsignificantefficiencyadvantages(GlobeNewsWire,2019).CDNsenablecloserproximitybetweentheend-userandthevideocontenttoachieveareductioninlatencyandarethereforecomputationallylight,onlyreadingandwritingdatacopiedfromthecorecontentstack.iPhoneCellularnetworkAutomaticdatasetting50”SmartTVFHD(1080p)LaptopSD(480p)AveragemixCarbonimpactofvideostreamingDiscussion545.DiscussionThissectionwilldiscussandcontextualisethefindingsofthiswhitepaper’sanalysis.Wewillintroduceemergingresearchsurroundingthemodellingofnetworktransmissionenergyconsumptionandexplorehowthisnewinformationcanbeappliedtoapowermodelapproachthatattemptstoestimatetheshort-termmarginaleffectsofachangeinviewingpatternsrelatedtovideoquality.TherewillalsobediscussionrelatedtotheanalysispresentedintheResultssection,uncertaintyaroundthefutureofvideostreamingandpointsforfurtherresearch.55CarbonimpactofvideostreamingDiscussion5.1.Emergingresearchoffersnewinsightintotheshort-termmarginaleffectsofchangingviewingpatternsRecently,publicawarenessoftheimpactofourdigitalliveshasgrownandmanyinternetusersareinterestedinunderstandingthemostmeaningfulactionstheycantaketoreducetheirdigitalcarbonfootprint.Videostreaminghascomeintofocusinthiscontextduetothelargevolumeofdatathatistransmittedthroughthenetworkandinturntheestimatesofvideostreaming’scarbonimpactwhichtypicallyrelyonaveragenetworkenergyintensityfigures.AsdiscussedintheMethodologysection,theconventionalapproachisnotwellsuitedtoassessingtheshort-termmarginaleffectofchangingviewingpatternsi.e.howchangingvideoqualityfromSDtoHDorHDto4Kaffectscarbonemissions.Thisisduetotheutilisationofaveragetransmissionnetworkenergyintensities,whichrelatenetworkenergyconsumptiontodata,andisparticularlytrueforhighbitrateapplicationslikevideostreaming.Estimationapproachesutilisingaveragetransmissionnetworkenergyintensitiesworkwellfororganisationalfoot-printingandsystemlevelestimates,butarenotsufficientlygranulartoreflectthepowerconsumptiondynamicsoftransmissionnetworkequipmentrelativetocomputationalloadandnetworktraffic.RecentworkbyJensMalmodin(Malmodin,2020b)proposesasimpletransmissionnetworkpowermodelthatoffersapromisingapproachtoestimatingtheshort-termmarginaleffectsofchangingviewingpatternsinvideostreamingquality.Abriefoverviewofthisresearchispresentedbelowandfollowingthisoverview,thepowermodelapproachtoestimatingthemarginalcarbonimpactofstreamingatvaryingvideoqualitiesispresented,whichincorporatesthesimpletransmissionnetworkpowermodel.5.2.Malmodin’ssimplepowermodelprovidesacloserrepresentationofthedynamicsofinternettransmissionMalmodin’sproposedpowermodeloffersamoreaccuratereflectionoftheinstantaneousandshort-termeffectofvideostreamingonnetworkenergyconsumption.Thepowermodelmakesuseofabaseloadanddynamiccomponentpowermodelinaccountingfortheinternettransmission-relatedenergyofnetworksandmorecloselyrepresentsthealwaysonstateofnetworkequipmentandenergyconsumption,comparedtoaconventionalaveragekWh/GBfactor.Inmodernnetworkequipment,thepowerdrawisnotonlyafunctionofcomputationalloadordatatraffic,butitisalsodrivenbythebaseload(oridlepower)consumptionoftheequipment’soperation.Networkandcorecomputingequipmentoperates24/7ataconstantbaselineofpowerconsumptionwhenidle,i.e.minimalload/datademand.Thisconstantoperationisarequirementofnetworkequipmentsothatitisreadytolistenandrespondinstantlytoincomingsignals.Whendataloadincreases,asillustratedformobileradioequipmentinFigure22(Malmodin,2020b),thepowerconsumptionoftheequipmentexperiencesonlyamarginalincreaseofupto30%relativetobaseloadpowerconsumption.Thisisbecausethebaseloadpowerconsumptionisprovisionedtocopewithpeakdemand.Figure22.Poweranddatamodelforasuburban4Gradiounit/basestation(Malmodin,2020b)56CarbonimpactofvideostreamingDiscussionFigure23.Power/dataandpower/timemodelsforafixedBBaccesslineinahousehold.Thisincludeshomerouterequipment(Malmodin,2020b)Figure23(Malmodin,2020b)illustrateswhatthismeansinahomeforafixedline,wherehomerouterpowerdrawisaconstantwattage24/7.Consequently,thisillustratesthatadditionaldataconsumptionandusageloadappliedtonetworkequipmentfrommediumtohighbandwidthserviceslikevideostreamingdoesnotleadtoasignificantshort-termincreaseinnetworkdeviceenergyconsumption.Instead,theshort-termimpactofhighbitratevideostreamingisonlymarginal.Thehigherbitraterequiredtowatchanhourofvideostreamingin4K(2160p)comparedtoFullHD(1080p)requiresadditionalelectricalpower,butonlymarginallysorelativetobaseloadpowerconsumption.UsingMalmodin’sproposedpowermodelallowsustocapturethisaspectofvideostreamingenergyandemissionsimpactinamorerepresentativemanner.Theconventionalapproachontheotherhandisnotasuitabletoolforthisparticularapplicationasitdoesnotreflectthisbehaviourofnetworkcomputing,insteadsimplifyingtherepresentationofnetworkstoasingleaveragekWh/GBfactorderivedfromenergyanddatavolumefiguresatthenetworklevel.Therearestill,however,openquestionsanduncertaintysurroundingthepowermodelapproach,includingvalidationthatitcanscaletothenetworklevelandaccountfor100%ofnetworkenergyconsumptionandhowbesttoaddressallocationofidleenergyconsumption.ThisisexpandedonlaterintheDiscussionsection.57CarbonimpactofvideostreamingDiscussion5.3.MethodologyofthepowermodelapproachThissectionprovidesasummaryofthemethodologicalapproachusedtoestimatetheshort-termmarginalcarbonemissionsimpactofvideostreamingatvaryingvideoqualities.Anoverviewofthisapproach,referredtoasthepowermodelapproachwithinthispaper,isdiscussedincludingthecharacteristicsofthepowermodelapproach,theboundaryusedtodefinethefootprintandabriefexplanationofallocationanditsroleinthisestimationapproach.Followingthisoverview,thekeyparametersusedinthisestimationapproachandahigh-levelexplanationoftheirderivationisprovided.Characteristics•UtilisesMalmodin’stransmissionnetworkpowermodel(Malmodin,2020b)•Reliesonanallocationapproachtoattributetheenergyconsumptionofasharednetworktotheservicesthatusethenetwork•Transmissionnetworkenergyconsumptionisallocatedtovideostreamingintwoways:baseloadenergyisallocatedbyviewingdurationandactivedevicesanddynamicenergyisallocatedperdata•Energyconsumptionofdatacentresandend-userdevicesisallocatedbasedonviewingduration•Mobilenetworkpowermodelisrepresentativeof4GnetworksonlyIntendeduse•Assessmentoftheshort-termmarginalimpactofchangeinvideoqualityStrengths9Powermodelprovidesacloserrepresentationofthedynamicsofinternettransmissionenablingassessmentofthemarginalimpactofvaryingviewingquality9Datacentresandend-userdevicesareestimatedinananalogousmannertotheconventionalapproachLimitationsXRepresentativeofaparticularnetworkandperiodoftimeXSensitivetocharacteristicsofthetransmissionnetworkconsideredforestimation.Thesenetworkcharacteristicsincludenetworkequipmentefficiency,quantityofnetworkusersanddatatraffic.XRelativelynascent,withopenquestionsanduncertaintythatcurrentlyrestrictitfrombeingusedasanalternativetotheconventionalapproachfororganisationalfoot-printing5.4.Overviewofthepowermodelapproach58CarbonimpactofvideostreamingDiscussionThepowermodelapproachbuildsonthefoundationoftheconventionalapproachandreplacestheaverageenergyintensityapproachusedforestimatingenergyandemissionsofnetworktransmissionwithMalmodin’spowermodel.ThisisrepresentedinFigure24below,wherethedifferencefromtheconventionalapproachisboxedinorange,alongsidethepowermodelparametersthatdefinethefixedandmobilenetworkestimation.Asthedatacentresandend-userdevicescomponentsofthepowermodelapproachareconsistentwiththeconventionalapproach,theywillnotbediscussedinthissection.Fordetailonthesecomponents,refertotheMethodologysectionfortheconventionalapproach.ThegoverningparametersandassumptionsofthepowermodelapproacharedetailedinTable5.Figure24.Videostreamingprocessmapindicatingthecomponentsthatmakeupthevideostreamingprocess.Theassociatedparametersthatdefinethenetworktransmissionmodelofthepowermodelapproachareboxedinorange.Theremainingcomponentsareconsistentwiththeconventionalapproach.CorenetworksHometerminalsandroutersWiredaccessnetworksCellularaccessnetworksCloudstorageandencodingSubscriberpremisestransmissionPeripheralsScreensInternettransmissionContentDeliveryNetworkUsermediadeviceDatacentresTransmissionPowermodelapproachEnergyallocatedbasedontotalviewinghoursCorenetworksFixedBBpower=1.5W+0.03W/MbpsMobileBB(4G)power=0.2W+0.03W/MbpsAccessnetworksFixedBBpower=5W+0.02W/bitrate%MobileBB(4G)power=1W+1.5W/MbpsAllocatedrouterpowerconsumptionperuser(onlyappliestofixednetworks)Energyallocatedforonehourofstreaming59CarbonimpactofvideostreamingDiscussionThisapproachmaybedescribedasatransmissionnetworkpowermodelbasedapproach,asitutilisesMalmodin’sproposedpowermodeltoestimatethetransmissionnetworkenergyattributabletovideostreaming.Allocationisanimportantconsiderationforthepowermodelapproach,whichallocatestransmissionnetworkenergydifferentlyforthebaseloadanddynamiccomponentsofthepowermodel.Thisisdiscussedfurtherinthissectionwhentheequationsgoverningthepowermodelarepresented.AsthisapproachutilisesMalmodin’srecenttransmissionnetworkpowermodel,itisarelativelynascentapproachtoestimatingthecarbonimpactofvideostreamingandhasopenquestionsanduncertainty,particularlysurroundingvalidationthatthepowermodelcanappropriatelyscaletothenetworklevel,andhowbesttoaddresstheallocationofauser’sidlenetworkconnection.Therefore,inthispaperweusethepowermodelapproachtoassesstheshort-termmarginalimpactofachangeinvideoquality,ratherthanasanalternativetotheconventionalapproach.5.5.ApproachboundaryInananalogousmannertotheconventionalapproach,thepowermodelapproachdrawsitsboundaryaroundthecomponentstagesofthevideostreamingprocess:datacentres,transmissionandend-userdevices,asshownFigure24.Thelifecycleboundaryforthisapproachincludesonlythein-useelectricityconsumptionofeachvideostreamingprocesscomponentandexcludestheembodiedcarbonandend-of-lifeemissionsofdatacentres,networkequipmentandend-userdevices.ThesecomponentstagesarediscussedinfurtherdetailbelowandassumptionsrelatedtoeachcomponentandmodelparameterarepresentedinTable5.Thegoverningequationsofthepowermodelapproach’stransmissioncomponentsincludemultipleelements,butcanbesplitintoafixednetworkmodelandamobilenetworkmodel,bothwithdistinctcoreandaccessnetworkcomponents.Withineachofthecoreandaccesscomponentsisabaseloadanddynamiccomponent,wherethebaseloadisfixedandrepresentsthenetworkequipment’sidlepowerconsumption,andthedynamicelementvariesinproportiontobitrate.Thisisdemonstratedbelowbyequation(12)forthefixednetwork.Thebaseloadpowerelementsrepresenttheidlepowerconsumptionofnetworkequipmentperfixedlineconnection,whereafixedlineconnectiontypicallyservesasinglehousehold.Asthefixedlineconnectionisasharedservice,weallocatethebaseloadamongtheusersoftheconnection,whichisdoneperactivedevice.Thus,thefixedlineconnectionisdividedbythenumberofactivedevicesaccessingthefixedline,representedbyequation(13),whichisderivedfromthequantityofusersperfixedline,thequantityofconnecteddevicesperuserandtheaverageproportionofactivelyconnecteddevicesatanygiventime.Furthermore,asthebaseloadofthefixedconnectionisdrawingpowercontinuously,thepowermodelonlycapturesthepowerdrawnwhileactivelyvideostreaming.Inpractice,thereareperiodsofthedayandnightwherethefixedlineconnectionisnotbeingactivelyusedbyanyinternetservice,butisdrawingpowernonetheless.Toaccountforthisidleenergy,weuseanidletimeallocationfactortoattributeaportionoftheidleconnectiontovideostreaming.Finally,thedynamicpowercomponentsareestimated,wherethedynamiccorenetworkpowercomponentisproportionaltobitrateofthevideostreamandthedynamicaccessnetworkpowercomponentisproportionaltothebitrate%ofthevideostream.Inotherwords,thebitrate%isthepercentageofthefixedlineconnection’sbandwidththatisbeingutilisedbythevideostream.5.6.Transmissionnetwork60CarbonimpactofvideostreamingDiscussionGlobalparametersDRFQAQUQDA=durationofvideostreaming=datatransmissionrate=idletimeallocationfactor=quantityofactivedevicesperfixedlineconnection=quantityofusersperfixedline=quantityofdevicesperuser=activedevicefactorasaproportionoftotaldevicesTransmission–FixedNetworkEFNBFN,CBFN,AVFN,CVFN,AS=energyconsumptionoverthefixednetwork=baseloadfixedcorenetworkelementperfixedlineconnection=baseloadfixedaccessnetworkelementperfixedlineconnection=dynamicfixedcorenetworkcomponent=dynamicfixedaccessnetworkcomponent=bandwidthofthefixedlineconnectionTransmission–MobileNetworkEMNBMN,CBMN,AVMN,CVMN,A=energyconsumptionoverthemobilenetwork=baseloadmobilecorenetworkelementpersubscriber=baseloadmobileaccessnetworkelementpersubscriber=dynamicmobilecorenetworkdynamicelement=dynamicmobileaccessnetworkdynamicelementTransmission-HomerouterEHRBHR=energyconsumptionofthehomerouter=baseloadpowerconsumptionofthehomerouterParameterdefinitions61CarbonimpactofvideostreamingDiscussionTheresultingtotalpowerdrawismultipliedbythedurationofthevideostreamtoestimateenergyconsumption.(14)(15)(12)QA=QU×QD×A(13)WhereEFNistheenergyconsumptionoverthefixednetwork,BFN,CandBFN,Aarethebaseloadelementsperfixedlineconnectionforcoreandaccessnetworks,respectively,QAisthequantityofactivedevicesperfixedlineconnection,Fistheidletimeallocationfactor,VFN,CandVFN,Aarethedynamiccomponentsforfixedcoreandaccessnetworks,Sisthebandwidthofthefixedlineconnection,RisthedatatransmissionrateandDistheviewingduration.QUrepresentsthequantityofusersperfixedline,QDisthequantityofdevicesperuserandAistheactivedevicefactorasaproportionoftotaldevices.Inprinciple,themobilenetworkpowermodelfollowsasimilarstructure,however,thepowermodelisderivedpersubscriber,sothereisnofurtherallocationtotheperuserlevel.Representedbyequation14,themobilenetworkenergyconsumptionisestimatedviabaseloadpowerelementsforthecoreandaccessnetworksinadditiontodynamicpowerelementsofthecoreandaccessnetworks,whichareproportionaltobitrate.Theresultingtotalpowerdrawismultipliedbytheviewingdurationtodetermineenergyconsumptionoverthemobilenetwork.WhereEMNistheenergyconsumptionoverthemobilenetwork,BMN,CandBMN,Aarethebaseloadelementspersubscriberforcoreandaccessnetworks,respectively,andVMN,CandVMN,Aarethedynamiccomponentsformobilecoreandaccessnetworks,respectively.Thefinalcomponentoftransmissionoverafixednetworkisthehomerouter,whereenergyconsumptionisestimatedasshowninequation15Ashomeroutersaretypicallyalwaysonanddrawpoweratanearconstantrate,theenergyconsumptionofthehomerouterisestimatedsimplyusingabaseloadpowerconsumption,whichisdividedbythenumberofactivedevicesaccessingtherouter.Inananalogousmannertothecoreandaccessbaseloadelements,anidletimeallocationfactorisappliedandtheproductismultipliedbyviewingdurationtodetermineenergyconsumption.EMN=(BMN,C+BMN,A+(VMN,C+WMN,A)×R)×DEHR=×F×DIncontrasttothefixednetworkpowermodel,thereisnoallocationofidleenergyinthemobilenetworkmodelbecausemobilenetworkequipmentiscontinuouslyinteractingwithconnecteddevicestoservereferenceandsyncdata(Malmodin,2020b).Asaresult,subscribersmaybeconsideredtohaveaconstantactiveconnectiontothemobilenetwork.WhereBHRisthebaseloadpowerconsumptionofthehomerouter.EFN=(()×F+(VFN,C+×100)×R)×DBFN,C+BFN,AQAVFN,ASBHRQA62CarbonimpactofvideostreamingDiscussionTable5.PowermodelapproachassumptionsbystreamingprocesscomponentStreamingcomponentConventionalapproachassumptionsDatacentres&ContentDeliveryNetwork•Energyintensity,IDC(2020)=1.3Wh/hr•DerivedfromaselectionofDIMPACTmembers,basedonmeasureddatain2020.Transmission–Network(Core&Access)Fixednetwork•Baseloadelements,BFN,C=1.5W/lineandBFN,A=5W/line•Dynamicelements,VFN,C=0.03W/MbpsandVFN,A=0.02W/bitrate%•Usersperline,QU,sourcedfromPopulationReferenceBureau(PopulationReferenceBureau,2020),seeAppendix•Devicesperuser,QD,sourcedfromCiscoAnnualInternetReport(Cisco,2020),seeAppendix•Idletimeallocationfactor,F=3•Activedevicefactor,Aassumedtobe0.5•Fixedlineconnectionbandwidth,S=75MbpsassumptionbasedondiscussionwithJ.MalmodinandverifiedwithspeedteststatisticsMobilenetwork(4G)•Baseloadelements,BMN,C=0.2W/subscriberandBMN,A=1W/subscriber•Dynamicelements,VMN,C=0.03W/MbpsandVMN,A=1.5W/MbpsBaseloadanddynamicelementsforfixedandmobilenetworksaresourcedfromMalmodin,2020b,page94Transmission-Homerouter•Baseloadelement,BHR=10W/lineDatatransmissionratesFixednetwork•Standarddefinition(SD):2.22Mbps(1GB/hr)•Fullhighdefinition(FHDorHD):6.67Mbps(3GB/hr)•Ultra-highdefinition(UHDor4K):15.56Mbps(7GB/hr)Mobilenetwork•Savedatasetting:0.37Mbps(0.17GB/hr)•Automaticdatasetting:0.56Mbps(0.25GB/hr)•Maximumdatasetting:6.67Mbps(3GB/hr)ThesefiguresarederivedfrompublishedNetflixfiguresondatausage(Netflix,2021)End-userdevices•Reasonableestimatesofaveragepower(W)forspecificdevices,seeAppendixfordetails•Thestandbytimeofend-userdeviceswasnotincludedinthisanalysis63CarbonimpactofvideostreamingDiscussion5.7.Theshort-termmarginaleffectofvideostreamingqualityoncarbonemissionsWithanunderstandingofMalmodin’sproposedtransmissionnetworkpowermodelandhowithasbeenappliedtodevelopthepowermodelapproachusedinthiswhitepaper,weestimatetheshort-termmarginaleffectofvideostreamingqualityoncarbonemissions.Inthiscontext,short-termeffectreferstothemarginaleffectonnetworkemissionsasitrespondstoachangeindemand,givenafixednetworkcapacity.Overalongerperiodoftime,networkinfrastructureisupdatedwithnewnetworkequipmenttechnologiesandadditionalcapacityisaddedtoaddressmediumandlong-termchangesinnetworkdemand,particularlyinresponsetoconsistentlyelevatedlevelsofpeakdemand.Thishasknock-oneffectstototalenergyconsumptionofthenetwork.Thepowermodelapproachandtheresultinganalysispresentedheredonotattempttomodelthesemediumandlong-termeffects,nordotheymodeltheeffectsondatacentresandend-userdevicesasvideoqualitychanges.Toassesstheshort-termmarginaleffectofvideostreamingqualityoncarbonimpact,weusedarepresentativefixednetworkscenarioandarepresentativemobilenetworkscenarioforEuropein2020.Therepresentativefixednetworkscenarioismodelledwitha50insmartTV,whichrequiresnoadditionalperipheralstoconnecttothefixednetworkviaahomerouter,whiletherepresentativemobilenetworkscenarioismodelledwithaniPhone11.Figure25andFigure26demonstratethemarginalimpactofbitrateonaggregatedcoreandaccesstransmissionemissionsoverthefixednetworkandmobilenetwork,respectively.Forclarity,thefixednetworkemissionsshowndonotincludeemissionsfromthehomerouter.Overthefixednetwork,threevideoqualitiesareevaluated:SD,HDand4K,withbitratescorrespondingto2.22Mbps,6.67Mbpsand15.56Mbps,respectively.Overthemobilenetwork,Netflix’susermobiledatasettingsareusedwhichhavecorrespondingdatausagelimits(inGBperhour)whichhavebeentranslatedtoanaveragebitrateinMbps.Thesesettingsare:savedatasetting,automaticdatasettingandmaximumdatasettingwithbitratescorrespondingto0.37Mbps,0.56Mbpsand6.67Mbps,respectively.TransmissiondynamicTransmissionbaseloadFigure25.Marginalimpactofbitrateoncoreandaccesstransmissionemissions(fixednetwork)4.03.53.02.52.01.51.00.50.0gCO2e/hourSD(2.22Mbps)HD(6.67Mbps)4K(15.56Mbps)64CarbonimpactofvideostreamingDiscussionTransmissiondynamicTransmissionbaseload4.03.53.02.52.01.51.00.50.0gCO2e/hourSaveDataSetting(0.37Mbps)AutomaticDataSetting(0.56Mbps)MaximumDataSetting(6.67Mbps)Figure26.Marginalimpactofbitrateoncoreandaccesstransmissionemissions(4Gmobilenetwork)Themarginalimpactoverfixednetworkasresolutionandbitrateincrease,isrelativelysmall,wherethetransmissionemissionsgrowfromjustunder1gCO2e/hourtojustover1gCO2e/hourbetweenSDand4K.Thisrepresentsanincreaseof26%inaggregatedcoreandaccesstransmissionemissionsfromSDupto4K.Thebaseloademissions,whichremainconstant,areestimatedasapproximately0.8gCO2e/hourandthereforemakeupthelargestproportionoffixednetworktransmissionemissionsacrossthespectrumofvideoqualityassessedhere.Inessence,thisdemonstratesthefixednetwork’slowelasticityinrelationtodatatransmissionrate.Themobilenetworkhasadifferentresponse,demonstratingahigherelasticityinrelationtodatatransmissionrate.Fromthesavedatasettingtotheautomaticdatasetting,thereisanincreaseintransmissionemissionsof16%,thoughbothsettingsstillresultinestimatedtransmissionemissionsoflessthan1gCO2e/hour.Comparingthemaximumdatasettingtothesavedatasetting,transmissionemissionsincreaseby545%,growingtonearly3.5gCO2e/hour.Theconstantbaseloademissionsareestimatedaslessthan0.5gCO2e/hour.Whilethemobilenetworkdemonstratesahigherelasticityrelatedtodatatransmissionrate,thetransmissionemissionsarestillrelativelysmallevenathigherbitrates.TheseresultsarecontextualisedinFigure27andFigure28,wherethetransmissionemissionsarepresentedalongsidethecorrespondingtotalemissionsfromthevideostreamingprocess.Again,thelowelasticityofthefixednetworkisdemonstratedwithtotalemissionsgrowingonly1%fromSDto4Kviewing.Coreandaccesstransmissionemissionsmakeupapproximately3%ofthetotalvideostreamingemissionsforthefixednetworkscenarioacrossthethreevideoqualitiesevaluatedandend-userdevicesmakeupthemajorityoftotalemissions.Incomparison,themobilenetworktransmissionemissionsdriveanincreaseintotalemissionsof7%betweenthesavedataandautomaticdatasettingandanincreaseintotalemissionsof231%fromthesavedatatothemaximumdatasetting.Thetransmissionemissionsaccountfor42%oftotalemissionsinthesavedatasettingupto83%oftotalemissionswhenusingthemaximumdatasetting.65CarbonimpactofvideostreamingDiscussion35302520151050gCO2e/hourTransmissiondynamicTransmissionbaseloadFigure27.Marginalimpactofbitrateontotalemissions(fixednetwork)Figure28.Marginalimpactofbitrateontotalemissions(4Gmobilenetwork)TotalvideostreamingTransmissiondynamicTransmissionbaseloadTotalvideostreaming4.54.03.53.02.52.01.51.00.50.0gCO2e/hourSaveDataSetting(0.37Mbps)AutomaticDataSetting(0.56Mbps)MaximumDataSetting(6.67Mbps)66CarbonimpactofvideostreamingDiscussionThekeyinsightsderivedfromthisanalysisarethatthelowelasticityofthefixednetworkinresponsetobitrateimplythatthefixednetworktransmissionemissionsaremorecloselylinkedtonetworkcapacitythantheyarelinkedtotheshort-termresponsetoincreasedtraffic.Forthemobilenetwork,increasedbitratehasademonstrableeffectonshort-termtransmissionemissions,primarilyasaresultofthepowerconsumptioncharacteristicsofmobilebasestations(Malmodin,2020b).Forusers,themosteffectivewaytoreducetheircarbonimpactofstreamingdependsonthenetworkbeingused.Overafixednetwork,usingamoreenergyefficientorsmallerviewingdevicehasafargreaterimpactthanchangingvideoquality,particularlyasend-userdevicesdominatethetotalemissionsfromstreamingoverfixednetwork.Infact,thisanalysisdemonstratesthatchangingvideoqualitywhilestreamingoverafixednetworkhasanegligibleshort-termimpact.Foratypicaluserstreamingoveramobilenetworkwithasmartphone,emissionsaresmallevenathigherbitrates,estimatedatlessthan5gCO2e/hourinthisscenariousingthepowermodelapproach.However,utilisingdatausagesettingsthatminimisebitrateoffersanopportunitytoreduceemissions.5.8.FurthervalidationandrefinementofthepowermodelisalogicalnextstepThepowermodelapproachprovidesanallocationapproachthatmorecloselyrepresentsthedynamicsofinternettransmissionasitincorporatesapowermodelderivedfromthestudyofpowerconsumptionprofilesoffixedandmobilenetworks.Furthermore,itpartiallydecouplesenergyconsumptionfromdatavolumesthroughuseofthebaseloadelementandweunderstandthatenergyconsumptioninnetworksisnotlinearlyproportionaltodatavolume,asevidencedbyreportedenergyconsumptionfiguresinrelationtoincreasedinternettrafficduringtheCOVID-19pandemic(GSMA,2020).Asthepowermodelapproachoffersanewmethodologyforallocatingnetworkenergyandemissionstointernetservices,duringthecourseofthisstudywehaveperformedaninitialsensecheckofthenumbersusedinthepowermodeltodeterminetheirrepresentativenessforalimitednumberofleadingtelecommunicationsnetworkoperators(telcos).Thetotalannualtransmissionnetworkenergyforeachoperatorwasestimatedusingthepowermodel,whereannualbaseloadenergyconsumptionwasestimatedwiththequantityoffixedlinesubscribersandquantityofmobilesubscribersforeachtelcomultipliedbythebaseloadelementofthepowermodelandmultipliedbyhoursperyear.Theannualdynamicenergyconsumptionwasestimatedusingtheannualquantityofdatatransmittedoverthenetworktoderiveanaverageannualnetworkbitrate,whichwasthenmultipliedbythedynamicelementofthepowermodelandmultipliedbyhoursperyear.Theresultingannualenergyconsumptionfiguresforfixedandmobilenetworkswerethencomparedtotheactualtotalreportedenergyfigures,totesttheaccuracyofthepowermodelapproach.Thisindicatedthatusingthepowermodelsforfixednetworksaccountedforapproximately60%ofthetotalmeasuredenergyexpected,whilethemobilepowermodelsaccountedforapproximately70%ofthemeasuredenergyexpected(seeTable6).Thisinitialcomparisondemonstratesthatthepowermodelapproachprovidesareasonableapproximationacrossarangeofnetworkoperatorswhencomparedagainsttheirmostrecentenergyconsumptionfiguresforoperationoftheirnetwork(typicallythesewere2019figures).Whileoursensecheckindicatesthepowermodelisareasonableapproximation,thefiguresinTable6belowdemonstratethatthepowermodelgivesaslightlylowestimateofnetworkenergyconsumption.However,asdiscussedlaterintheDiscussionsection,limitedavailabilityofdetailedandgranularnetworkdatamakesthevalidationprocesschallengingandtosomeextent,incomplete.67CarbonimpactofvideostreamingDiscussionNetworktypeTelco1Telco2Telco3Telco4Telco5Mobile76%N/A33%82%125%Fixed64%N/AN/AN/A62%Total73%56%N/AN/A63%Table6.PercentageofnetworkenergyestimatedbythepowermodelrelativetoreportednetworkenergyconsumptionofselectedtelcosFurthervalidationofthepowermodelrequiresamoredetailedunderstandingofoperationalenergyconsumptionsplitbytypeofnetworki.e.fixedbroadband,2Gmobilenetwork,3Gmobilenetwork,4Gmobilenetworkand5Gmobilenetwork.Furthermore,wemustknowthenumberofsubscribersandthetotaldatavolumetransmitted(orideallyameasureofaveragebitrate)throughbothfixedandmobilenetworks,measuredusingconsistentmethods.Initscurrentform,themobilenetworkpowermodelisrepresentativeof4Gnetworks.However,networkoperatorsoperatearangeofgenerationalmobilenetworktechnologiesfromlegacy2Gand3Gequipment,tomodern4Gandcutting-edge5Gandifnetworkdataispublished,itisnotdisaggregatedbynetworktechnology,andoftennotevendisaggregatedbetweenfixedandmobile.WhilethepowermodelisrepresentativeofthecharacteristicsofefficientEuropeannetworks,refinementofthepowermodelislogicaltorepresentabroaderrangeoftechnologies(i.e.3Gand5Gmobilenetworks)andregions(i.e.IndiaandtheUnitedStates).Furthermore,networkoperatorsthemselvesmaywishtodeveloptheirownbaseloadanddynamiccoefficientsforuseinthepowermodeltospecificallyrepresenttheiruniquenetworkcharacteristics.Thereisalsoaneedforrefinementoftheidlenetworkenergyallocationmethodologyinthepowermodelapproach.Thisrefinementiscurrentlylimitedbyalackofdataaroundkeyassumptionsandconsensusonthemostappropriateallocationapproach.Therefinementoftheseassumptionscouldbesupportedthroughgreaterdatafromnetworkandserviceproviders,withaccesstogranularlevelconsumerdata(althoughthiswouldneedtorespectpersonaldataprivacy),aswellasbyutilisingconsumersurveysanddatathatprovidesinsightintotheinternetservicesthataredrivingpeaknetworkdemand.Thepowermodelapproachoffersanewperspectiveontheallocationofnetworkenergyandemissions.Consideringthisnewperspective,furtherdiscussion,consensusandstandardisationofallocationapproachesamongindustryplayersisalogicalnextstep.68CarbonimpactofvideostreamingDiscussion5.9.VideostreamingimpactcontextualisedInordertoputtheemissionsimpactofanhourofvideostreamingintocontext,somecomparativescenariosarepresented,comparingtheEuropeanaverageimpactofanhourofvideowithday-to-dayactivitiesofregularconsumers.POPCORNPOPCORNPOPCORNPOPCORNAnhourofvideostreamingvs.microwavingabagofpopcornAnhourofvideostreamingvs.driving100mTheEuropeanaverageemissionsofanhourofvideostreamingusingtheconventionalapproach=56gCO2e.Theaverageemissionsofmicrowavingabagofpopcornforfourminutesinatypical800Wdomesticmicrowaveoven(assumingEuropeanaveragegridintensity)=16gCO2e.TheEuropeanaverageemissionsofanhourofvideostreamingusingtheconventionalapproach=56gCO2e.Theaverageemissionsofdrivingadistanceof100m=22gCO2e.Theemissionsofanhourofvideostreamingareroughly3.5xthatofmicrowavingabagofpopcornTheemissionsofanhourofvideostreamingareapproximately2.5Xthoseoftheaverageemissionsofdrivingadistanceof100m.100m69CarbonimpactofvideostreamingDiscussionComparingtheenergyofanhourofvideostreamingwithsomehouseholddevicesTheEuropeanaveragetotalenergyconsumptionofanhourofvideo,usingtheconventionalapproach,of188Wh/hourhasbeencomparedtotheenergy(Wh)usedbysometypicalhouseholddevices.Table7belowrepresentstheuseofseveraldomesticappliancesincertainusecases,intermsofequivalentminutesofvideostreaming,e.g.oneboilofthekettle=19minutesofvideostreaming.18008008.51060538.510191733Power(W)Energy(Wh)EquivalentminutesofEUaveragevideostreaming(conventionalapproach)SourceofdevicepowerDeviceusecaseHomerouter(1hourontime)D+RInternational,2020Frequencycast,2021Eartheasy,2021TheSpruce,2021(Assumed600-1200W)Electrickettle(2minuteboil)LEDlamp(1hourontime)POPCORNMicrowave(4minutescookingpopcorn)Table7.Comparisonoftheenergyofonehourofvideostreamingwithsomehouseholddevices70CarbonimpactofvideostreamingDiscussion5.10Forconsumers,deviceselectioncanreduceenvironmentalimpact,butsystemicapproachtoenergyefficientdevicesoffersamoremeaningfulopportunity5.11InternetpeakcapacitydrivesenergyconsumptionTheresultsoftheanalysisinthiswhitepaperindicatethatconsumerdeviceselectionhasamuchgreaterimpactonvideostreaming-relatedemissionsthanthechoiceofvideostreamingquality.Thispaperacknowledgesthepotentialimpactconsumerinfluencecanplaythroughtheirconsumptionhabits.However,whenend-userdevicesareconsideredinaggregate,thereisgreatsystemicopportunityfordevicemanufacturerstodrivedownenergyconsumptionofviewingdevicesthroughacontinuedfocusonenergyefficiencyimprovements.Voluntaryagreementshavebeenusedinthiscontexttoencourageimprovementstoenergyefficiencyinsmallnetworkequipmentsuchashomerouters(D+RInternational,2020).InWesternEuropealone,end-userdevicesmaynumberinthehundredsofmillions,basedonanestimatedsixconnecteddevicesperperson(Cisco,2020).Energyefficiencyimprovementsforend-userdevicesoutoftheboxwilltranslatetoenergyandemissionsreductionsfortheconsumer.Theanalysispresentedinthispaperfocusesontheusephaseenergyconsumptionandemissionsofthevideostreamingprocess.Devicesalsohaveassociatedemissionsrelatedtothefulldevicelifecycle,fromrawmaterials,tomanufacturingandend-of-life.Forsmalldeviceslikesmartphones,materialsandmanufacturingmakeupthelargestproportionoflifecycleemissions,asevidencedbytheiPhone12(Apple,2020),whereproduction-relatedemissionsaccountfor83%oflifecycleemissions.Therefore,designimprovementsthatenableconsumerstoincreasetheperiodoftimebetweendeviceupgradesareanimportantconsiderationforreducingenvironmentalimpact.Whereconsumerswanttoinfluenceandreducevideostreamingimpacts,deviceselectionhasthegreatestpotentialtoenablethemtodoso.Viewinganhourofvideostreamingcontentonsmallermobiledevices,suchassmartphones,tabletsandlaptops,willhaveasignificantlysmallercarbonandenergyimpactcomparedtowatchingonanenergyintensivedevicesuchas50insmartTV.Thecarbonandenergyperhourofvideostreamingbyusingsmallerdevices,isapproximately15%-30%oftheemissionsfromtheuseoflargesmartTVs.Inapracticalsense,therearesimplebehaviouralchangesthatcanoptimisetheenergyusefromvideostreamingsuchasusingamobiledevice(~1W)whenwatchingvideowhilemultitaskingorbystreamingdirectlyfromasmartTV(~100W)orwiththeuseofastreamingsticksuchasChromecastorRoku(~2W)insteadofstreamingviavideogameconsole(~90W)inconjunctionwithaTV.Crucially,whetheraconsumerwatchesinSD,fullHDor4K,theenergyandcarbonimpactsofanhourofvideostreamingwillonlybemarginallygreaterforthehigherqualitycontent.Asystemicapproach,whichincludesdevicemanufacturersfocusingoncontinuallyimprovingtheenergyefficiencyandextendingthelifetimeofthedevicestheyproduce,offersamoremeaningfulopportunityforenergyandemissionsreductionthanbehaviouralchangeattheconsumer-levelalone.Inordertounderstandtheimpactofvideostreamingconsumption,itisimportanttounderstandthedynamicsoftheinternetinrelationtoenergyconsumption,andtheimpactofdataconsumptiononenergyandemissionsatthelevelofindividualinfrastructurecomponentsandthesystemlevel.Anappropriateanalogytoillustratethisisabustransportnetwork.Thebusnetworkruns24hoursaday,sevendaysaweek,withbusesregularlymovingaroundthenetworktotransportpassengers.Evenwhenabsentofpassengers,thebusconsumesfueltomoveitselfandaspassengersgetonthebus,thebuswillonlyconsumeasmallamountofadditionalfuelinordertotransportthesepassengers.Atthenetworklevel,thebiggestfactorintotalfuelconsumptionofthenetworkisthecapacityofpassengersthatitcansupport.Asdemandforthebusservicegrows,additionalbuses,withtheirownfixedamountoffuelconsumption,areaddedtotheroutesthatcomprisethenetwork.Similarly,theinternetisinconstantoperation,andlikethebusnetwork,currentlyhasanalmostconstantconsumptionofenergytopoweritsnetworkequipment.Asdatatrafficincreaseswithinthecapacityofthenetwork,theadditionalenergyrequiredtotransmitthisdataisonlymarginalcomparedtotheidleenergyoftheinternettransmissionnetworkthatisconstantlyoperating.71CarbonimpactofvideostreamingDiscussionAsvideostreamingandotherinternetservices’demandfortheinternet’stransmissionnetworkincreases,thecapacityofthenetworkmustincreasetosupporttheadditionaldemand.Whenthebusisfull,capacityisincreased,effectivelyincreasingthenumberofbusesinthenetwork.Thesameoccurswhendemandforinternettransmissionnearspeakcapacity,thecapacityisincreasedbyaddingtothenetwork.Thisanalogyeffectivelydemonstratesthatvideostreaming’sdemandforinternettransmissioncapacityhasbothshort-termandmedium-termeffects,whereintheshort-termthereisonlyamarginalincreaseinenergyconsumptionandcarbonemissionswhichdoesnothaveameaningfulimpactonthesystemlevelenergyandemissions.Inthemedium-term,additionalnetworkinfrastructureisaddedtosupporttheincreaseddemandwhichcontributestoincreasedembodiedcarbonfrommanufactureofthenetworkequipmentandpotentialforanincreaseinbaseloadnetworkenergyconsumption.Whileincreaseddatausagefromserviceslikevideostreamingmaynotbesignificantlyimpactingthetotalenergyandemissionsofinternettransmissionatthesystemlevel,itisimportanttounderstandwhatdeterminesthehighbaseloadpowerconsumptioninthefirstplace.Asillustratedbythebusanalogy,peakdemanddrivesthesystemlevelenergyandemissions.Whendemandreachespeakcapacity,infrastructureexpansionisrequired,resultinginmoreenergyconsumptionandthereforeemissions.Wheninfrastructureexpansionisrequired,thismaybeduetothesameusersmakingmoretrips,butalsoduetonewusersshiftingfromothertransportationmodesorfromgeneralpopulationgrowth.Itishoweversimplistictoassumethatthesamelegacybuseswillbeaddedtothesameroutestoadapttothisdemand;instead,largerbusesmaybeprocuredwithhigherfuelefficiencyorthatusealternativecleanerfuelsources,androuting/dispatchmaybefurtheroptimised.Asaresult,thebusnetwork’stotalemissionsmayincreasebutnotlinearlyfrompasttrends,andpartoftheincreaseisduetoshiftingcarbonfromonemodeoftransportationtoanother.Inasimilarmanner,networkcapacityexpansionmayberequiredduetothecurrentusersstreamingmorevideo,fromnewuserswhoareshiftingfromtraditionalformsofTVviewingorfromanincreaseinthenumberofuserswithinternetconnection.Networkcapacitywillexpandtomeetthisadditionaldemandthroughacombinationofmeans,whichmayresultinanincreaseintotalnetworkenergy,butisunlikelytodosolinearly.Internetcapacityessentiallyrevolvesaroundafeedbackloopofincreasingdemandwhichresultsinadditionalinfrastructureandcapacityandthereforeenergyandemissions.Theexistenceofthisincreasedcapacityinturnenablesincreaseddemandforthenetwork’sservicesandthecyclecontinues(Figure29)(Preistetal.,2016).Figure29.Feedbackloopofnetworkinfrastructure(Preistetal.,2016)InfrastructurecapacityDemandServicesenablesthedesignofnewobservedandanticipateincreasesdrivegrowthoffersgreateraffordances,whichstimulatesR72CarbonimpactofvideostreamingDiscussion5.12Consumptionofvideohasprogressedinacarbonefficientmanner5.13AshiftinbehaviourmayreduceemissionsAlothaschangedinthelasttwodecadesinrelationtohowweconsumevisualmediacontent.Untiltheearly-2000s,videorentalsathometraditionallyrequiredatriptobrickandmortarstores,suchasBlockbuster,andrentingphysicalcopiesofvideocontent,beforereturninghomeandwatchingamoviethroughaTVandVCR.Aroundthistime,digitalvideorecorders(DVRs)integratedwithinset-topboxeswereintroducedtothemarketandallowedrecordingofvideodirectlyfromtheTVforon-demandconsumptionatalatertime.Bythemid-to-late-2000svideocontentwasincreasinglyconsumedthroughmonthlysubscriptionservicesofferingDVDrentalsdeliveredthroughthepostalsystem.Thismodelnolongerrequiredconsumerstotravelinordertopurchaseorrenttheirvideocontent,andsawthegrowthinpopularityofamonthly-subscriptionpackage.CompaniessuchasNetflixandAmazonwereresponsibleforthemajorgrowthofthisconsumptionmechanism(West,2014).Bytheearly2010s,videoconsumptionhadfinallyshiftedtowardsonlinestreamingservices(TheGuardian,2013).Thisdevelopmentofhowweconsumevideocontentovertimehasonecleartrend,ithasbecomeincreasinglycarbonefficientanddematerialised.Fromthedaysofaconsumerhavingtogetintheircaranddrivetoavideorentalstoretoaccessseveralhours’worthofphysicalmediacontentatatime,tonowbeingabletostreaminstantaneouslyavastamountofcontentfromwithintheirhome,theemissionsassociatedwithvideoconsumptionhavebecomemoreefficient.Withvideostreaming,operationalemissionsarevirtuallyallelectricandaswecontinuetoprogresstowardsafutureofelectricalgridswithzerooperationalemissionsthroughtheuseofrenewableelectricity,emissionsfromvideostreamingwillcontinuetofall.Aspreviouslyhighlighted,thechoiceofend-userdevicesonwhichvideoisstreamed,canhavealargeimpactoncarbonemissionsandenergyconsumption.Therefore,ashiftinconsumerbehaviour,shiftingfromwatchingvideoonlargeTVs,tosmallermobiledevices,mayhavethepotentialtoreducevideostreamingrelatedemissionssignificantly.Additionally,whereconsumersswitchtorenewableelectricitytariffs,thiswillhaveanevengreaterimpactontheoverallcarbonimpact.Consumerelectronicsanditsassociatedtrendsevolverapidly,likewiththeswiftdevelopmentandpenetrationofsmartphonesintheglobalconsumerelectronicsmarket.AtrendseenintheUnitedStates,forexample,highlightsthepotentialimpactthatshiftingconsumerbehaviourcouldmakeonvideostreamingemissions.Between2013and2017,thenumberofinstalledTVsintheUSfellbyapproximately5%,adecliningtrendthathasbeeninmotionsince2009(Urbanetal.,2019)(Figure30).Itisnotclearifthistrendhascontinuedsince2017,andcertainlybroadcastersinEuropehaveseenanincreaseinTVviewing,andatrendforlargerscreenTVs.IntheUK,figuresfromOfcomindicatethatconsumersincreasinglyprefermobileandportabledevicesforaccessingtheinternet(Ofcom,2020b),whereagrowingproportionofadultsnolongeruseacomputerforgoingonlineandinsteadrelyonsmallerdeviceslikesmartphonesandtablets.Ashiftinconsumerbehaviourtowardsuseofsmallerdevicesmayrepresentanopportunityforanincreaseintheproportionofvideostreamingcontentviewedthroughthesesmallermoreefficientdevices,suchassmartphones,tabletsandlaptops.ShouldconsumerbehaviourshiftawayfromTVconsumption,theenergyandcarbonsavingsthatcouldbeenabledarepotentiallysignificant.ThecounterargumenttothisisthatthetrendtowardstheuseofsmallerdevicesisnotdirectlyreplacingviewingonTVs,butisinadditiontoTVviewing,thereforegivingrisetoanincreaseintotalvideostreamingconsumption.Figure30.InstalledbaseofTVsintheUnitedStates(Urbanetal.,2019)191199520052015installedbase(millions)35328473CarbonimpactofvideostreamingDiscussionFurthermore,withthedevelopmentandrolloutof5Gmobilenetworkequipmentglobally,andthepotentialthisholdsforlowlatency,high-qualityconnectivity,wemaywellexpecttoseeashifttowardsincreasedmobilenetworkdeviceconsumptionofvideostreaming.5Ghasthepotentialtounlockasignificantincreaseinmobileconsumption,beingabletoprovidefasterstreamingservicesovermobilenetworks.However,consumerhouseholdsarestilllikelytohaveahomerouter,eitherafixedlinerouterora5Ghomebroadbandhubreplacingafixedlinerouter(orpossiblyboth),andthereforesavingswillnotbedrivenbyavoidinghomeroutertransmissionsrelatedenergyuse.5.14PredictingthefutureisdifficultandhasinherentuncertaintyItisinherentlydifficulttopredictwhatthefuturemayholdforvideostreaminggoingforward.Severalkeyaspectsoffuturedevelopmentshouldatleastbeconsideredwhendiscussingthefutureofvideostreaming.Theseincludetheuncertaintysurroundingfuturebehaviouralpatternsofconsumption,whatdevicesareusedtostreamvideo,theeffectof5G,andthedifficultyinprojectingmodellednetworkenergy.Withtherapidnatureofend-userdeviceproductlifecyclesquicklychangingduetoproductinnovation,anddynamicconsumerbehaviourtrends,itishardtosayhowvideostreaminghabitsmaychange.Herearejustafewconceivablescenarios.WhilethedecliningtrendoftelevisionsinhomesdemonstratedintheUSmaycontinue,wemayseemoreofashifttowardsviewingofvideostreamingonmobiledevices.However,thereissignificantuncertaintysurroundingthesetrends.Unknownssurroundingthefutureofproductdevelopmentmakepredictingconsumerdeviceusedifficult.5Gmayplayabigroleinthefutureofvideostreaming,withtherolloutof5Gmobilenetworkequipmentincreasingaroundtheworld.Asmobilevideotrafficandmobiledevicesbegintoovertakethegrowthofotherdatatrafficandfixeddevices,internetserviceprovidersmayattempttoproposenewmobileinfrastructuresandsolutionsforhighperformancevideostreamingservices,providinghigh-quality,high-efficiencystreaming(Voetal.,2017).Should5Gunlockavastpotentialofhigh-qualitystreamingcapability,wemayseeaheightenedshifttowardsvideostreamingcontentviewedonmobilenetworkdevices,andsmallermoreefficientdevices,comparedtotraditionalfixednetworkdevicesinthehome.Thismaypotentiallyleadtoenergyandcarbon-relatedsavingsassociatedwithvideostreaming.However,alotremainstobeseeninthedevelopmentof5G–whetheritwillachievethepotentialcapabilitytodeliverhigh-qualitycontentandwhetheritmaydisplacesomeoftheuseoffixednetworksandhomeroutersforvideostreaming.Fromamodellingofvideostreamingemissionsandenergyimpactperspective,thereisalsosignificantdifficultyinunderstandinghowthefutureassessmentofinternettransmission-relatedenergypervolumeofdatamightchangegoingforward.Themodellingofnetworkenergy,thathasbeenusedinassessingtheenergyassociatedwiththetransmissioncomponentofvideostreaming,isreflectiveofaspecifictimeframe.Consequently,thenetworkmodelsmustbecontinuallyupdatedgoingforward,inordertoreflectthemostrecenttimeperiodthatisbeingassessed.Simpleextrapolationsusingthefiguresandapproachesinthispaperwillhaveasignificantamountofuncertaintyassociatedwiththem,astheywillnotreflectthenetworkcharacteristicsoffuturescenarios.However,despitethisuncertainty,wecanatleastexpectthatnetworkswillonlygetmoreenergyefficientovertime.Historically,theelectricityintensityofnetworkdatatransmissionhashalvedroughlyeverytwoyearssince2000(Aslanetal.,2018).Asalsodemonstratedbyreportingofenergyintensitybynetworkoperators–forexample,seedatareportedbyTelefónicaandCogentinsection2.6.2(Figure8andFigure9).Thefuturemayalsoholdpotentialreboundeffects,resultingfromtheincreasedeaseofaccessoflow,fixedmonthlycostvideostreamingservices.Asvideostreaminghasbecomemoreaccessibleandcheaperforconsumers,consumersareviewingmoreandmorevideostreamingcontent.Asvideostreamingconsumptionincreasesinthefuturetheremaywellbeareboundeffectofincreasedenergyuseandcarbonimpact,potentiallytothepointwhereefficiencygainsinnetworktransmissionandend-userdevicesareoutweighedbytheincreasedconsumptionofstreamedvideocontent.Thishighlightstheimportanceofthecontinueduptakeofrenewableelectricitytopowertransmissionnetworks,sothatincreasednetworkenergyconsumptiondoesnottranslatedirectlytoincreasedemissions.74CarbonimpactofvideostreamingDiscussion5.15Ourunderstandingofnetworkenergyandemissionsislimitedbydetaileddataavailability5.16AreasforfurtherinvestigationWhatisclearfromthisresearch,isthatalackofpubliclyavailabledataandtransparencyaroundenergyconsumptionforallstagesofthelifecycleofvideostreamingislimitingtheaccuracyandrobustnessofassessmentsofvideostreamingemissions.Thisisapplicableforboththeconventionalandpowermodelapproaches.Shoulddetaileddataofnetworkenergyconsumptionateachstageofthenetworktransmissionprocessandnetworkdatatrafficbeeasilyaccessibletostudiessuchasthis,itwouldbeinvaluableinhelpingtoperformassessmentsoftheenvironmentalimpactofvideostreamingandotherinternetservices.Fromanaccountinganddiagnosisperspective,thiswouldalsoallowforagreaterunderstandingoftheaccuracyofbothapproachescurrentlyandenablemoreprogressivethinkingonthedeterminationofappropriateallocationapproaches.Thecurrentreportingbynetworkprovidersoftheirenergydataistypicallyatanaggregatelevel.Energyordatatrafficfiguresinannualreportsarenotnecessarilyconsistentlydefined,andofinsufficientgranularitytohelpvalidateassessmentslikethese.Ifnetworkprovidersweretodisclosenetworkrelatedenergyanddatatrafficconsumptionatamoredetailedgranularlevel,thenthedeterminationofnetworkenergyconsumptionwouldbeenhancedsignificantly,andfutureassessmentsofvideostreamingemissionswouldbemoreeasilyvalidatedandmoreaccurate.However,itisrecognisedthatmostofthemajornetworkoperatorsreporttheircarbonemissionsinannualreportsandCDPsubmissions,andthattelecommunicationsisoneoftheleadingsectorsincommittingtosciencebasedtargetsandinuseofrenewableelectricity.Thereisalsoaneedforgreaterdatatransparencyfromvideostreamingserviceproviders,whichwillinformtheassumptionsthatwillhaveasignificantimpactontheallocationofenergyintransmissionofvideostreaming.Detailedmeasureddatafromserviceprovidersaroundcustomernumbers,andhowtheyinteractwiththeirserviceswouldgreatlyenhancethevalidationofthismethodology.Dataonthedevicemixusedbyconsumersanddataonstreamingactivityitselfwouldhelpimprovetheaccuracyofenergyallocationtoend-userdeviceswhenvideostreaming.Thisneedfordatatobecollectedandreportediscrucialforfutureambitionstovalidateandrefinethisapproach.Theseassumptionshaveasignificantimpactontheenergyconsumptionallocatedtovideostreaming,andthustobeabletoimprovetheestimationoftheenergyandcarbonimpactfromvideostreamingrequiresthatdatacollectionandsurveyingofthiskindofdatabecomesavailableinthefuture.Certainareasforfurtherresearchhavebeenidentified,whichmaybenefitthefuturedevelopmentofvideostreamingimpactassessmentsandrefiningthecalculations.Firstly,moreresearchcouldbeundertakentounderstandpeakcapacityoftheinternet,andhowthiswillimpactenergyandemissionsgoingforward,aspeakcapacitygrows.Thisisimportanttounderstandhowfutureincreasesinvideostreamingdemandwillimpactpeakcapacityoftheinternet,andhowtheenergyandemissionsimplicationsofthatwillchange.Aspreviouslymentioned,itispeakdemandwhichultimatelydrivesthebaselineenergyconsumptionoftheinternettransmissionnetwork.Moreresearchisneededtounderstandwhatisdrivingthepeakcapacitydemandthatinturndrivestheprovisionedcapacityfortransmissionequipment(andwhetheradvancesintechnologyaresimplyenablinggreatercapacitythatisthenusedbynewservices).Improvedunderstandingofpeaknetworkdemandwillalsoinformdecisionmakingsurroundingallocationapproacheslikethoseusedintheanalysisofthispaper.Ifwearetodeterminethatitispeakdemandwhichisdrivingthebaselineconsumptionofinternetenergy,provisionedtomeetpeakdemand,thenwemustunderstandwhoisdrivingtheneedforpeakdemand.Midtohighbitrateservices,suchasvideostreamingorstreamedgaming,maybethenaturalfirstguessatwhattopointto.Serviceslikethese,whichoftenresultinsignificantincreasesindemandinasingleperiod,maybethedriversoftheneedforgreaterpowerconsumptionandinfrastructuretomeetdemand.Ifthiswerethecase,thenitraisesthequestion,isthepowermodelapproachafairwaytoallocateemissionstotheseservices?Thepowermodelapproachmayeffectivelybeallocatingtheseserviceproviderslessemissionsbasedontheirmarginalimpactoninternettransmissionconsumption,whereasinreality,bydrivingpeakdemandandthereforebaselinepowerconsumptiontosupportthis,theyshouldinfactbeallocatedmoreemissionsrelatedtothis.Furtherresearchinthisareaandabetterunderstandingofhownetworkcapacityisprovisioned,isimportanttohelpunderstandthefuturetrendsofcapacityandthepotentialtrajectoryofemissionsandenergyconsumptionfrominternettransmission.75CarbonimpactofvideostreamingDiscussionByunderstandingwhothekeyplayersareindecisionmakingaroundpeakcapacitywecanfurtherunderstandhowtoallocatethebaselineenergyconsumptionoftransmissionnetworksmoreappropriately,andcontinuetorefinethepowermodelapproach.Byconductingfutureresearchintothedriversanddecisionmakersofpeakcapacity,wewillbeabletodevelopanimprovedallocationmethodologyfornetworkenergyandemissions,accountingmoreaccuratelyforthedemand-drivenbaseloadofenergyconsumptioninthenetworkstothekeyplayersdrivingthisdemand.Anotherareawherefutureresearchshouldbefocusedisaroundtheenergyperdata[kWh/GB]metricsforfixedandmobilenetworks,andhowthesemaychangeinthefuture,andwhatisanappropriateglobalaveragetouseintheconventionalapproachasoutlinedinthiswhitepaper.Forinstance,the0.0065kWh/GBfigureusedintheconventionalmethodologyforfixednetworkenergyallocationmaynotbearepresentativeglobalaverage,butmaymorelikelyreflectalowerEuropeanaveragewhereefficiencyinthenetworksisgreaterthaninlessdevelopedareasoftheworld.Greaterresearchandvalidationoffiguresliketheseisrequiredtoenableaconventionalapproachthatissoundenoughtoprovideahigh-leveltotalsystemboundaryanalysisofvideostreamingimpactsandorganisationalfoot-printing.Furtherresearchandinvestigationshouldbeconductedtobuildupthisdatabaseandimprovetherobustnessandvalidityofsuchfiguresusedinanaverageglobalscenario.76CarbonimpactofvideostreamingPolicydevelopments6.Policydevelopments77CarbonimpactofvideostreamingPolicydevelopments6.1.Introduction–PolicydevelopmentsThispolicysectionofthewhitepaperfollowsthemaincomponentsofthevaluechainofvideostreaming,focusingongovernmentalpolicyrelatedtodatacentres,networksandend-userdevicesbothonregionalandnationallevel.ThegeographicalscopeforthisisprimarilyEuropewheretherehavebeensignificantdevelopmentsrecently,butweacknowledgethatfurthercomparisonandanalysiswithotherregulatoryeffortsshouldbeconducted.Subsequently,thissectionlooksatwhattheindustryiscurrentlydoingeithercollectivelyoratacompanylevel.Similarly,thevaluechainofvideostreaming(datacentres,networksandend-userdevices)willbeconsideredfortheseindustryinitiatives.Lastly,thereisanassessmentofthegapsandopportunitieswithingovernmentalpolicyandcorporateaction.Asmentionedintheintroduction,thereisanincreasingawarenessamongpolicymakersthatICThasboththecapabilitytodelivertechnologytoreducecarbonemissions,butitalsohasitsownsignificantcarbonfootprint.Therefore,policymakersareconcernedbothaboutreducingthedirectcarbonimpactfromICT,andalsoaboutencouraginginnovationthatcanenableemissionreductionswithinthesectorandinothersectors.Particularlyin2020,thisawarenesshasbeenreflectedinpublicpolicyandcorporateactionontacklingemissionsfromthesector(includingvideostreamingspecifically).WithastrongrecommendationfromthesectorthatfundingforICTinnovationispartofagreenrecovery,decisionmakersarelookingtoimprovetheconditionsforcoverageandconnectivitybyimprovingtheinvestmentenvironmentforICT.ThisisillustratedbytheexamplesofexplicitinclusionofICTintheEUTaxonomyandtheEuropeanGreenDeal.OnaEuropeanlevel,concerningtheICTsectoranddigitalisation,theEuropeanGreenDealincludesadigitalstrategytitled‘ShapingEurope’sDigitalFuture’,publishedinFebruary2020.TheEuropeanGreenDealisatthecoreofdefiningasustainabilityframeworkforEuropeandislayingthegroundworkfortheEuropeanCommission’ssustainabilityroadmap.Thiscomprehensivesetofpoliciesisprovidingaframeworktomeettheclimatetargetsfor2030and2050.Firstly,Europewantstobethefirstcontinenttohavenonetemissionsby2050.TheaimofthedigitalstrategyistoprepareEuropeforanewdigitalageandsimultaneouslyachievethe2050climateneutralitytargetsproposedbytheEuropeanCommission.ThemaincomponentswithinthedigitalstrategyareonArtificialIntelligence,Europeandatastrategy,Europeanindustrialstrategy,HighPerformanceComputing(HPC),DigitalMarketsAct,DigitalServicesAct,Cybersecurity,Digitalskillsandconnectivity.Theobjectivesareformulatedundertheumbrellaofthreepillars:(I)Technologythatworksforpeople;II)Afairandcompetitivedigitaleconomy,andIII)Adigitalandsustainablesociety.UnderlyingthisstrategyistheobjectiveforEuropetobeoneofthelargestdigitalplayersglobally.A‘digitalandsustainablesociety’isdefinedunderthestrategyastheICTsector‘contributingtoasustainable,climate-neutralandresource-efficienteconomy’.AsoutlinedintheIntroductionandBackgroundsectionsthereisuncertaintyontheexactcontributionoftheICTsectortoglobalenergyuseandemissions,andthisisalsoaffectedbywhatisincludedwithinthedefinitionofICT.TheEuropeanCommission’sdigitalstrategyclaimsthattheICTsectoraccountsfor5-9%oftheglobalelectricityuseandover2%ofglobalgreenhousegasemissions(EuropeanCommission,DGCONNECT,2020c).Despitetheuncertaintyoverthesenumbers,thestrategydoesclearlyacknowledgethattechnologyandinnovationsfromthatverysamesectorcouldalsohelpreduceglobalemissions,morethanthesectoritselfemits.Forexample,theGermangovernmentismakingstrideswithpresentinganenvironmentaldigitalagendawithmorethan70measurestomakethetechsectormoresustainable(GermanMinistryforEnvironment,2020a).ThegovernmentclaimsthatthisdigitalagendaisthefirststrategyinEuropethatcombinesdigitisationandenvironmentalprotectioninsuchaconsistentmanner.TheGermanenvironmentministryplacesthefootnotethat"Ifunchecked,digitalisationwillbecomeaproblemfortheclimate”,butsimultaneouslystressingthepotentialthatdigitalisationhasforcontainingclimatechange.Addressingthisparadoxisattheheartofthemanynationalpolicies.(ibid.)InMarch2021,theEuropeanCommissionannouncedtheGreenDigitalCoalition(EuropeanCommission,2021c).ThisspecificallyacknowledgesboththeemissionsoftheICTsectorandthepotentialforreducingemissionsinothersectors.Itincludesacommitmentto‘developmethodsandtoolstomeasurethenetimpactofgreendigitaltechnologies’,andasofmid-April2021hadCEOsof26companiesthathadsignedthedeclaration.ThebalancebetweenbothnegativeandpositiveimpactsisdependentonimprovingenergyefficiencyandcirculareconomyperformanceoftheICTsector,fromdatacentrestobroadbandnetworkstoend-userdevices.78CarbonimpactofvideostreamingPolicydevelopments6.2.Policydevelopments6.2.1TraceabilityandmonitoringframeworksCurrently,consistentdataontheenergyandcarbonemissionsfromICTisnotavailable.Manystudiesarebasedonacademicmodelling,ratherthanbottom-updatareportedbyICTcompanies.DatareportedbyICTcompaniesisvariable,withsomecompaniesprovidingdetailedcomprehensivedata,whileothersprovideverylittledataonenergyandemissions.Goodpolicyrequiresgooddata.Tosupportthis,governmentsandinternationalorganisationsgloballyareexploringdifferentmeasures,onenergyefficiencyandcirculareconomy.OnaEuropeanUnionleveltheobjectiveofpropermonitoringcanbefound,amongothers,inapublicconsultationon‘‘Environmentalmanagement&performance–sectoralreferencedocumentfortelecommunications/ICTservices”.TherequestforconsultationdescribesthattheICTservicessectorshould‘setoutbestenvironmentalmanagementpracticeforalltelecommunicationsandICTservicesprovidersincludingtelecommunicationoperators,ICTconsultancyfirms,dataprocessingandhostingcompanies,softwaredevelopersandpublishers,broadcastersandinstallersofICTequipmentandsites.Specificenvironmentalperformanceindicatorsandbenchmarksofexcellenceforaparticularbestenvironmentalmanagementpracticeshouldalsobegivenwheneverpossibleandmeaningful’.(EuropeanCommission,2021d).Itshouldbementionedthatanopportunityforpublicconsultationisnotaguaranteeforconcretelegislationorenactment.However,itdoesreflectthecurrentawarenessregardingthesectoranditsenvironmentalefforts.WealsoseethatinvestinginbettermeasurementisgainingnationalpoliticaltractioninsomeEuropeangovernments.ArecentFrenchlegislationdocumentstates:‘Weneedprecise,clear,objectivedataandconsensusmethodologiesontherealimpactofdigitaltechnologyontheenvironment’(FrenchSenate,2020).Theneedforconsensusonmethodologyhasbeenechoedbytheindustryandhavingnationalpolicyonthiswillenhancetrueaction.ThetelecomandenvironmentauthorityinFrancehadestablishedaroadmaptoworktogetherwiththeindustryonimprovingthemethodstobettermeasuretheimpactonICT.InSeptember2020,theUnitedKingdomreleasedtheir‘SustainableICTanddigitalservicesstrategy:targetsfor2020-2025’(UKDepartmentofEnvironment,Food&RuralAffairs,2020).ThestrategyquiteclearlyrevealsguidelinesfortheICTsectorthatincludesclimatetargets.TheICTsectorshouldreducegreenhousegasemissionsandworktowardsnetzerotargets,usingscience-basedtargets.Existingsuppliersshouldworkwiththegovernmenttomeetlegallybinding,orexisting/emergingsectoraltargets.Similartotheothernationalandregionalstrategies,technologyanddigitalinnovationsareconsideredessentialdriversforsustainablesolutions,suchasreducingtravel,energytransitionandreducingwaste.TheUnitedKingdomstrategyisalsospecificonmaterialstraceabilityfortheICTsector.Oneofthegoalsis‘100%traceabilityofICTatend-of-life(mapping)and100%complianceandtransparencyinsupplychains’.ThegoalsonthetransparencyarefortheICTsectortopublishanaccurateICTfootprintbasedontheservicesconsumed,ofestatesandwithsuppliers,encompassingembodied/embeddedcarbon.Secondly,forthesectortomapandaccountforallICTatend-of-life.Thirdly,thereisaneedforcollaborationandsettingupagovernmentalsustainabilitysteeringgrouptoincreasetransparencyandsubsequentlyaccountability.6.2.2DataCentresConsideringthevaluechainofvideostreaming,datacentreshavesofarbeenthemostprioritisedtargetofregulatoryinitiatives.Nationalandregionalpolicyrelatedtothegrowingnumberofdatacentresandtheconcernovertheirenergyconsumption,arewellestablished.AnexampleoftheseinitiativesondatacentresfromtheCommissionistheEUCodeofConductforDataCentres,firstintroducedin2008(EuropeanCommission,2008).Itsaimistoinformandstimulatedatacentreoperatorsandownerstoreduceenergyconsumptioninacost-effectivemannerwithouthamperingmissioncriticalfunctions.TheCodeofConductaimstoachievethisbyimprovingunderstandingofenergydemandwithinthedatacentre,raisingawareness,andrecommendingenergyefficientbestpracticesandtargets.TheCodeofConductisavoluntaryinitiativeandpartnersthataresignedupareexpectedtofollowtheintentandcommitments.ItshouldbenotedthattheCodeofConducthasbeenregularlyupdatedsinceitsoriginaldevelopment,andhasbeenalignedtoEN50600.79CarbonimpactofvideostreamingPolicydevelopmentsThereisalsoanarrayofotherindustrysustainabilitystandardsandmetricswhichareincreasinglybeingreferencedinpolicy(whichthesectorwelcomes).ForinformationonDataCentreenergyefficiencyandothersustainabilitymetrics,andrelatedDataCentrestandardssee(techUK,2017a)and(techUK2017b).Althoughstillused,theCodeofConductisnowrelativelyoldanduntil2020wasonlyoriginallyaframework.Tobemoreinlinewiththelatestsustainabilitystrategy,theEuropeanCommissionrecentlyreleasedthe‘2021BestPracticeGuidelines’(EuropeanCommission,2021a),whichisasupplementtotheCodeofConductasaneducationandreferencedocumenttoassistdatacentreoperatorsinidentifyingandimplementingmeasurestoimproveenergyefficiency.Underbestpracticesconcerningassociatedcarbonimpacts,therearespecificguidelineson‘EnergyUseandEnvironmentalMeasurement’,‘EnergyUseandEnvironmentalCollectionandLogging’,and‘EnergyUseandEnvironmentalReporting’.Dependingonthelevelofcontrolofthedatacentrethatanindividualorganisationhas,thegeneralpolicyisthatallactorsshould‘IntroduceaplanforEnvironmentalManagementinaccordancewithemergingEUguidelinesandinternationallystandardisedmethodologies’and‘IntroduceaplanforEnergyManagementinaccordancewithemergingEUguidelinesandinternationallystandardisedmethodologies’.Also,backin2008theUnitedStatesestablishedthevoluntaryNationalDataCenterEnergyEfficiencyInformationProgram.Theprogramengagesnumerousindustrystakeholderswhoaredevelopinganddeployingavarietyoftoolsandinformationalresourcestoassistdatacentreoperatorsintheireffortstoreduceenergyconsumptionintheirfacilities(USEPA2008).Sincethenwehaveseenothervoluntaryandself-regulatoryinitiativesemerge.AnexampleoftheincreasingfocusonDataCentresbynationalgovernmentsistheproposaloftheGermanFederalEnvironmentAgencytocreatearegisterfordatacentrestomonitorfutureefficiencytargets.Theagencyisthereforepreparingauniformstatisticalsurveyofdatacentrestocreatearegisterandtoserveasabasisforeffectivesectorcoupling.Sectorcouplingistheideaofinterconnecting(integrating)theenergyconsumingsectors–buildings(heatingandcooling),transport,andindustry–withthepowerproducingsector.Thedigitalstrategyalsospecificallyaddressesstreamingservicesproviders,encouragingthemtooperatewithdatacentreswith100%greenelectricityandtomake‘sensibleuse’ofwasteheat.InFrance,boththeFrenchgovernmentandthesenatereleasedrecommendationsin2020onthe'greendigitaltransition'.Thesenatepresentedadraftlegislation,mentioninglimitingtheimpactofvideostreamingandimprovingenergyefficiencyindatacentres.Mentioningthelimitingofinfinitescrolling,atechniquethatloadscontentcontinuouslyastheuserscrollsdownthepage.Also,adaptingthequalityofthedownloadedvideotothemaximumresolutionoftheterminal.Ondatacentresspecifically,therecommendationswouldrequiredatacentrestosubscribetobindingmulti-yearcommitmentstoreducetheirenvironmentalimpacts(monitoredbyARCEP)andbysubjectingtaxbenefitstoenvironmentalperformance.6.2.3NetworkTransmissionPolicymakersandindustryhaveastrongfocusonimprovingtheenergyefficiencyofnetworktechnologies.ThisisreflectedbyGermanEnvironmentMinisterSchulze,commentingonresearchintotheCO2emissionsfromvideostreamingcommissionedbytheGermanFederalEnvironmentAgency:"Todate,thedataavailableonhowdigitalinfrastructureaffectstheclimatehasbeenextremelysparse.Thisiswhyweareworkingtobridgetheexistinggapsinourknowledgewithsolidresearch.Afterall,goodpolicyneedstobebasedongooddata.Themostrecentfindingsnowshowusthatitispossibletostreamdatawithoutnegativelyimpactingtheclimateifyoudoitrightandchoosetherightmethodfordatatransmission.Fromanenvironmentalperspective,itwouldbeagoodideatosetupmorepublicwifihotspots,asthisismoreclimatefriendlythanstreaminginmobilenetworks.Theclimatebenefitofworkingfromhomeandvideoconferencingcanevenincreasewiththerighttransmissionmethodsandmoreefficientdatacentres.MygoalistocapitaliseontheGermanEUCouncilPresidencytoreachacommonpositiononenvironmentallyfriendlydigitalisationbecausethebestapproachwouldbetosetgoodstandardsthroughoutEurope."(GermanMinistryforEnvironment,2020b)DirkMessner,PresidentoftheFederalEnvironmentAgency,furthercommented:"Thisisgoodnewsforpeoplewholiketowatchmoviesandseries.YoucanusestreamingservicesathomewithafibreopticcableorVDSLwithouthavingtofeelguiltyabouttheclimate.Butthevolumesofdataallarounduswillgrowsteadilyoverthenextfewyears,beitintheformofnetworkedvehicles,homecinemaorvideoconferencing.Thisiswhyitisimportanttofindclimatefriendlytransmissionchannels.Ourresearchshowsthatweshouldstepupinvestmentsinexpandingourfibreopticnetworks.Thenew5Gtransmissiontechnologyisalsopromisingintermsofclimatechangemitigation."(ibid.)80CarbonimpactofvideostreamingPolicydevelopmentsThe‘Environment,climatechangeandcirculareconomy’workinggroupfromTheInternationalTelecommunicationsUnion,aspecialisedbodyfromtheUnitedNations,hasidentifiedstandardisationrequirementsforthesustainableuseanddeploymentofICTsanddevelopinginternationalstandards.Formulatedasthe‘ITU-TRecommendationsonmethodologiesandguidelinesthatassesstheenvironmentalimpactsofdifferentICTapplications’.TheserecommendationscoverspecificICTrelatedfunctions,productsandservices,including,forexample:ICTsupportingequipmentandfacilities,installationactivities(suchasonradiosites),andnetworksandotherservices.Additionally,aframeworkadoptedbyITUanditsmemberstatesistheConnect2030Agenda.ThepurposeoftheagendaistoshapethefutureoftheICTsectorbyworkingtowardsfourdistinctgoals;Growth,Inclusiveness,Sustainability,InnovationandPartnership.Also,thesameworkinggrouphasdevelopedasetofinternationalstandards(ITU-TRecommendations)thatassesstheenvironmentalimpactsof5Gsystemsincludingtheelectromagneticcompatibility(EMC)aspects,theelectromagneticfields(EMF)aspects,energyefficiencyin5Gsystemsandtheirresistibilitytolightningandpowerfaultevents(ITU,2019).AsdiscussedintheResultssection,theend-userdevicesaccountforthegreatestportionofemissioninvideostreamingfootprint.Reductionstrategieslaywithinenergyefficiencyoftheend-userdevices(Malmodin&Lundén,2018a),andthroughchangesinscreendisplaytechnologiesthathavethepossibilitytoenablesubstantialreductionsinpowerconsumption.TohelpEUconsumerscuttheirenergybillsandcarbonfootprint,anewversionofthewidely-recognisedEUenergylabelwasintroducedinallshopsandonlineretailersfromMonday,1March2021.Thenewlabelswillinitiallyapplytofourproductcategories;fridgesandfreezers,dishwashers,washingmachines,andtelevisionsetsandotherexternalmonitors.Newlabelsforlightbulbsandlampswithfixedlightsourceswillfollowon1September2021,andotherproductswillfollowinthecomingyears.IntheUSA,theEnergyStarprogramhasexistedsince1992,managedbytheEnvironmentalProtectionAgency(EPA),earlyonitsetminimumstandardsofenergyefficiencyforcomputersandservers,andhasovertimebeenextendedtoawiderangeofproducts.6.2.4End-userviewingdevicesRegulationsrelatedtostandbymodepowersettingshavebeenestablishedformanyyears,forexampleEUregulationNo642/2009,andIEC801/2013.InFrance,aspartofthedigitalstrategyinearly2021,theEcologicalTransitionMinisterBarbaraPompiliandtheSecretaryofStateforDigital,CédricO,presentedconcreteplanstobringenvironmentalanddigitalissuestogether.Theplansareoutlinedunderthreemainpillars:‘developknowledgeofthedigitalenvironmentalfootprint’;‘supportamoresoberdigitalenvironment’;and‘makedigitaltechnologyaleverfortheecologicalandsolidaritytransition’.Therebyalsodirectlyaddressingtheconsumerintheirdigitalbehaviour.SimilartotheGermanstrategythereisapartonaddressingtheconsumerandthefactthattheymustbeempoweredandeducatedtocommittoanenvironmentallyconscioususeofdigitaltechnologies.Regardingthereuseorrecyclingofthedeviceitself,theEuropeanUnionisexaminingthebenefitsof‘take-back’schemesfordevices.Supportingatake-backschemeshouldincentiviseconsumerstoreturndevicesthatarenolongerneeded,withthehopeofhigherlevelsofrecycling.Also,the‘righttorepair’isontheagendaandisreceivingmorepublicattention.The‘righttorepair’canapplytoallconsumergoods,butespeciallyintroducedtoreducee-waste.InFebruary2021theEuropeanParliamentvotedinfavourofthe‘righttorepair’,aspartoftheCircularEconomyActionPlan.Legislationiscurrentlynotinplaceyet,butitsendsastrongsignalontacklingembodiedemissionsande-waste.81CarbonimpactofvideostreamingPolicydevelopments6.3.IndustryinitiativesAsmentioned,theICTandE&Msectorsfacevariouschallengestodecarbonise.However,throughthepowersectorandpurchasesofrenewableelectricity,ICTiswellpositionedtokeeppacewithfuturetargets.Inthepast,theICTsectorhasrespondedproactivelytoaddressitsemissionschallengeandunlockopportunitiesforgreaterenergyefficiencyacrossthesector.LeadingplayerswithintheICTsectorhaveincreasedtheirclimateambition.Inthelast10years,thesectorhasbeentakingstepstodecarbonise,throughaccountingandreportingofgreenhousegasemissions,implementingenergyefficiencyprogrammes,andincorporatingtheuseofrenewablesintotheenergygrid(GSMA,ClimateActionHandbook,2019).ThisGSMAhandbookis‘designedtobeahighlevelguidetoclimatechangeforanyoneworkinginorwiththemobileindustry.Itexplainstheneedfortimelyanddecisiveaction,howemissionsarecategorisedandtherelatedterminology,beforefocusingonhowthemobileindustryisrespondingandpotentialnextsteps'.TheICTsectorismakingpositivestridestoaddressitsenergyconsumption.ICTcompanies,specificallydatacentreoperatorssuchasGoogle,Facebook,Microsoft,AmazonWebServices(AWS),Equinix,alongwithnetworkoperators,continuetoincreasetheirshareofrenewableelectricity,throughtheprocurementofPowerPurchaseAgreements(PPAs).Leadingbyexample,ICTcompanieshavebeenresponsibleforover50%ofcorporaterenewablesprocurement,globally,inthepastfiveyears(Kamiya,2020;BloombergNEF,2020;FinancialTimes,2021).Overall,theICTsectorhasthemeansatitsdisposaltoachievedeepdecarbonisation,throughenergyefficiencyanduseofrenewableelectricity.However,thereisaneedforindustry-ledreportingandincreasedtransparencyofICTcompanies’energyandcarbonimpacts.Thefollowingsectionwilloutlineseveralindustrychallengesandopportunitiestotackletheassociatedemissionsforvideostreamingofdatacentres,networksandend-userdevices.82CarbonimpactofvideostreamingPolicydevelopmentsTheIEA(2020a)rankeddatacentresandnetworksasasector“ontrack”toachievedeepdecarbonisation.Thisissupportedbybothcorporateandpoliticalplayerswithinthesectorthatarerespondingtothechallengesofclimatechange,inordertoreducethesector’scarbonfootprint.TheUnitedNationsspecialisedagencyforICTs,theITU,hasbeenworkingwiththeindustrytominimisethecarbonfootprintofICTs,developinginternationalstandards(ITU-TRecommendations),forexample,inareasasdiverseassmartcities,datacentresande-wastemanagement.Fordatacentresspecifically,theyhighlighttheexponentialgrowthofdatacentresworldwideandsubsequentlytheneedfordatacentrestomovetowardsutilisingrenewableenergysourcesintheiroperations.Theythereforehighlightthatinvestmentandresearchingreeningdatacentresiscrucial.Recently,over40organisationsannouncedthecreationoftheClimateNeutralDataCentrePact,asetofself-regulatorymeasuresdevelopedwiththeEuropeanCommissiontomakedatacentresclimateneutralby2030.ThepactwasinitiatedinJanuary2021andwasdescribedasa‘sectorcollaborationtoensuredatacentresareanintegralpartofthesustainablefutureofEurope’.Inthepact,datacentreoperatorsandtradeassociationsagreetomakedatacentresclimateneutralby2030,inlinewiththeEuropeanGreendeal.Thesignatories,includingGoogle,Microsoft,OVHCloud,AWSandAtos,committovarioustargetsfor2025and2030toimproveenergyefficiencythroughwaterconservation,heatrecyclinganduseofrenewableenergy.ThepactisopentocompaniesthatownoroperatedatacentreswithintheEU.Specifically,forenergyefficiency,thepactstatesthatthedatacentreswillmeethighstandards,whichwillbedemonstratedthroughaggressivepoweruseeffectiveness(PUE)targets.Anothernotablepointonenergyefficiencyisthecreationofanewdatacentreefficiencymetricandinlinewiththis,theneedtostandardisemeasurementmethodologiesforfuturereporting.For‘cleanenergy’thepactstatesthatthesignatorydatacentresshouldmatchtheirelectricitysupplythroughthepurchaseofrenewableenergy.Specifically,ittargetsthatdatacentreelectricitydemandshouldbematchedby75%renewableenergyorhourlycarbon-freeenergyby2025and100%by2030(ClimateNeutralDataCentrePact,2021).Largetechcompaniesindividuallyarealsosettingambitiouspledgestoreducetheiremissions.Intermsofrenewableenergyuseforoperations,Googleachieved100%renewablesin2017,Applein2018,Facebooksetatargettouse100%renewablesbyendof2020,andMicrosoftandAWSbothhave2025targets.6.3.1Datacentres-IndustryGooglehasnowannouncedatargettogomuchfurtherthanmatching100%renewablesonanannualbasis,andby2030willrunoncarbon-freeenergy24/7everywhereatalltimes.Thiswillmatchdatacentreenergyusewithrenewableproductiononanhourlybasisacrosstheglobe.Theirstrategyintheir24/7carbon-freefuturewhitepaper(Google2020)isverycomprehensive,andincludeshourlymonitoringandcontrolstoshiftdatacentreworkloadsintimethroughoutthedayandgeographicallyfromonedatacentretoanother,inordertosynchronisetheadaptableworkloadswithrenewablesgenerationthroughitsPPAagreements.InJanuary2020,Microsoftsetanambitiousgoal,pledgingtobeacarbonnegativecompanyby2030(significantlyincludingitsScope3supplychainemissions),andstrivingtoremoveallofthecompany’shistoricalcarbonemissionsby2050,aswellassettingupa$1billionclimateinnovationfund(Microsoft,2020).Thisrepresentsoneofthemostprogressivetargetssetbyanyprivatecompanytoaddressitscarbonfootprint(Reuters,2020).6.3.2Networksandtransmission-IndustryAcollaborationoftheITU,GeSIandtheGSMAdevelopedadecarbonisationpathwayfortheICTsector,asthebasisforsettingscience-basedtargets(SBT)guidanceforcompanieswithinsub-sectorsofICT.Thisguidance,approvedbytheScienceBasedTargetsInitiative(SBTi)inFebruary2020,recommendsthefirsteverscience-basedpathwayestablishedbytheindustrytocutgreenhousegasemissionsintheICTsector.Table8.SBTiguidanceforICTsectorcompaniesSub-sector%GHGreduction(2020-2030)Mobilenetworkoperators45%Fixednetworkoperators62%Datacentreoperators53%83CarbonimpactofvideostreamingPolicydevelopmentsThekeyrecommendationoftheSBTGuidanceisfortheICTsectortodecarboniseinalignmentwitha1.5°Ctrajectory(equivalenttoapproximatelya50%reductioninGGHemissionsovertheperiod2020–2030)andmorespecifically,totargetemissionsreductionsof45%formobilenetworkoperators,62%forfixednetworkoperatorsand53%fordatacentreoperatorsby2030(asshowninTable8)(SBTi,2020).Someofthemeasuresoutlinedtohelpachievethesetargetsincludethecontinuedimplementationofenergyefficiencyplans,switchingtorenewable/lowcarbonelectricitysupply,encouragementofgreatercarbonconsciousnessamongend-users.AsofMarch2021,41telecommunicationscompanieshaveeithercommittedtoorhavesetscience-basedtargets,ofwhich23companieshaveseta1.5°Ctarget(SBTi,2021).Therefore,signallingacleardirectionfortheICTsectortofollowa1.5°Ccompatibleemissionsreductiontrajectory,andachievenetzeroemissionsby2050.6.3.3End-userdevices–IndustryAsdiscussedintheResultssection,theend-userdeviceisresponsibleforthelargestpartofthevideostreamingfootprint.Toreducethispartofthefootprinttherearesomeopportunitiesthroughinnovationsandincreasedenergyefficiency.Generally,energyefficiencyofend-userdeviceshasbeenimprovingduetoamixoftechnologyenhancementsandpowerthresholdsbeingsetforstandbyandoperation.Thehistoricalshifttosmallerdevices(e.g.PCstolaptopstotablets)hassignificantlycontributedtoreductionsintotalenergyconsumptionfromend-userdevices.Also,historically,improvementsinTVdesign,shiftingfromCRTtoLCDandthenLCDwithbacklitLEDhasenabledsignificantenergysavings.Forgamingconsoles,theindustrydevelopedtheVoluntaryAgreementundertheEUEcodesignDirectiveforgamesconsoles,toachieve1TWhofenergysavingsperyearby2020acrossEurope(EuropeanCommission,2015).Thisvoluntaryagreementencompassesgamesconsolesmanufacturedbythethreemajormanufacturers;SonyInteractiveEntertainmentInc,MicrosoftandNintendo,accountingfor100%ofthemarket.Assuch,gamesconsolesmustalsocomplywiththeregulationssetoutinIEC801/2013forstandbyandnetworkedstandbypowerconsumption,includingsettingmaximumpowerlimitsaccordingtoeachconsoletypesoldwithintheEUandalsoprovideinstructionsforminimisingenergyuse(EfficientGaming,2018).6.4.PolicydevelopmentsandobservationsThereiscurrentlysignificantactivityinthepolicyarenarelatedtotheICTsector,withmuchinprogress.TheEuropeanGreenDealisstillindevelopmentandconcretelegislation,thatalignswiththeframework,willberolledoutthroughout2021.ButthereisaclearerunderstandingfrompolicymakersthattheICTsectorhasthepowertoreducetheothersectors’emissions.Thereisaclearneedtoimprovereporting,bothoncollectingconsistentandreliableempiricaldataandstreamliningreportingmechanisms.BoththeICTsectorandgovernmentalinstitutionsplayanimportantroleinthis.Currently,policymakersdonothavedetailedandcomprehensiveinformationoftheimpactoftheICTsector.Bothcompanyandgovernmentpoliciescouldbenefitfromimproveddataandstreamliningofreporting.Industryagreementisneededonclearersub-sectorboundarydefinitions,andconsistencyinmeasurementofenergyefficiencymetrics.Basedonabetterunderstandingofthefullvaluechainandtheassociatedemissions,policyshouldaddressthecomponentsthatwillhavethemostmaterialimpactalongthelifecycle.Thereisanunderstandingthatend-userdevicesareresponsibleforthelargestportionofthevideostreamingemissions,however,concretepolicytoreducetheseemissionsseemslimited.Policymakersshouldensurethatthedigitalandenergytransitionwillbeasgreenaspossible.Therefore,policyalsoneedstorecogniseexistinginitiativesintheindustry,andexistingtrendsintechnology.Anexampleofsupportingthisunderstanding,istheITU,theUnitedNationsICTspecialisedgroup,developingasetofmethodologiesandkeyperformanceindicators(KPIs)toassesstheenvironmentalimpactincludingmeasuringcarbonfootprint,energyperformanceandefficiencyacrossICTnetworks,goodsandservices.ThemethodologiesandKPIsaredevelopedwiththeinputofthesectoritself,andcanprovideaframeworkforgovernmentalpolicy.CarbonimpactofvideostreamingConclusions847.Conclusions85CarbonimpactofvideostreamingConclusionsThekeyconclusionsfromtheanalysispresentedinthiswhitepaperare:•Thecarbonfootprintofviewingvideo-on-demandstreamingisrelativelysmallincomparisontootherhumanactivities•Themarginalchangeofenergyconsumptioninresponsetochangesinviewingpatternsissmall•ThereisinherentvariabilityanduncertaintyintheestimationofthecarbonimpactofvideostreamingTheseconclusionsarenowdiscussedinabitmoredetailtoprovidesomecontextandexplanation.Thecarbonfootprintofviewingonehourofvideostreamingissmallcomparedtootherpotentialhumanactivities.TheEuropeanaveragefootprintestimatedinthiswhitepaperisapproximately55gCO2eperhourofvideostreamingfortheconventionalallocationapproach.(ThisestimateusesaEuropeanaveragegridemissionfactorof0.295gCO2e/Wh,arepresentativemixofviewingdevices,andnetworkenergyintensityfiguresfor2020).Forcomparison,theemissionsfrommicrowavingabagofpopcornforfourminutesisabout16gCO2e,boilingakettlefortwominutesis18gCO2e(usingthesameEuropeanaveragegridemissionfactor),whiledriving100metresinanaveragepetrolcaremitsaround22gCO2e(usinganaveragecaremissionfactorof0.216kgCO2e/km).Theanalysisinthiswhitepaperalsoshowsthattheviewingdeviceistypicallythesourceofthelargestpartofthecarbonfootprint.Theinstantaneous(ormarginal)changesinenergyinresponsetochangesinbitrate(duetodifferentresolutionsandothersettings)resultinonlyaverysmallchangeinthecarbonfootprint.Thisisbecausetheenergyconsumptionofmostdevicesandofthenetworkequipmentchangeverylittleinresponsetodynamicchangesindatavolumes,astheyhaveafairlyhighconstantbaseloadofenergyconsumption.Thisiswellillustratedbythepowermodelmethodology,whichreflectsthedynamicpowerprofilesofthenetworkequipment.Aswithmostcarbonfootprintassessmentsthereisaninherentvariabilityanduncertaintyintheestimationofthecarbonimpactofvideostreaming,whichgivesrisetoarangeofresults.(Variabilityreferstovariationsduetofactorssuchastimeorplace,whileuncertaintyreferstothedegreeofprecisionofmeasurements).Thiswhitepaper,andotherinformedresearch,shouldhelptoreducetheuncertaintyinthemeasurements,andprovideabetterunderstandingofthevariability.Thereareanumberofreasonsforvariabilityintheresults.Oneofthemostsignificantisthelocation.Thecarbonintensityofelectricity(measuredastheelectricitygridemissionfactorinkgCO2e/kWh)variessignificantlyfromcountrytocountry.Forexample,Germany’sgridemissionfactorisapproximately30timesthatofSweden.Therearealsovariationsbycountrythatcanaffectthenetworkemissions(suchasnetworktechnology,topologyandambienttemperature),andviewingpatternsmayvarybycountry.Thetypeofviewingdevicehasasignificantimpactonthetotalcarbonfootprint–thefootprint(relatedspecificallytotheenergyoftheviewingdevice)ofwatchingona50inTVisroughly4.5timesthatofwatchingonalaptop,androughly90timesthatofwatchingonasmartphone.Theyearthatanestimationrelatestoisalsosignificant,asimprovementsintechnologymeanthattheenergyintensityofequipmentiscontinuallydecreasing,andseparatelytheelectricityemissionfactorsaredecreasingastheelectricitygridsdecarbonisethroughtheutilisationofgreaterproportionsofrenewables.Relatedtotheuseofrenewables,itshouldbenotedthattheresultsinthiswhitepaperusethecountrygridaverageelectricityemissionfactors,andthereforedonotrecogniseanyadditionaluseofrenewablesbydatacentreornetworkoperators(orindeedbydomesticuseofrenewabletariffs).ThisissimilartotheGHGProtocolScope2Guidance‘location-based’accountingmethodforelectricity.Whereasthe‘market-based’accountingmethodwouldrecogniseuseofspecificrenewableelectricitypurchases.86CarbonimpactofvideostreamingConclusionsThisisanimportantdistinction,asanincreasingnumberofbothdatacentreoperatorsandnetworkoperatorshavemadesignificantstepsinmovingto100%useofrenewableelectricity.Followingthemarket-basedaccountingmethodwouldresultinaloweroverallcarbonfootprintofvideostreamingformanycountriesinEurope.Themostsignificantsourceofuncertaintywouldseemtoberelatedtothenetworkenergycomponent.Asthecomparisonbetweenthetwomethodsshows,theallocationapproachhasasignificantinfluenceonthetotalfootprint,andhighlightstheneedtounderstandthedifferentapproachesandtheirapplicationwheninterpretingandusingtheresults.Thisisdiscussedfurtherbelow.Additionally,thereareonlyalimitednumberofconsistentlymeasured,publiclyavailabledatapointsfortheenergyintensityofnetworks,whichalsocontributestotheuncertaintyinthenetworkenergy.Thiswhitepaperpresentsanewmethodforthecarbonfoot-printingofvideostreaming.Thepowermodelapproachisdifferentfromtheconventionalapproachinthewaythatenergyisallocatedfromthesharedcomponentsofthenetworkandthehomerouter.Itisverycommoninproductcarbonfoot-printingandinlifecycleanalysistouseallocationapproaches.Often,theremaynotbeanobviousmethodtouse,ratheravarietyofapproachesthatreflectdifferentrealities.Thiswhitepaperisnotintendedtoendorseonemethodovertheother,buttohighlightthatthetwomethodsaresuitablefordifferentpurposesanddifferenttypesofanalysis.Bothwouldbenefitfromfurtherrefinement,validationandresearch.Returningtotheanalogyofabusnetworkthathasbeenusedthroughthiswhitepaper,theconventionalapproachreflectsanaverageemissionfactorperpassenger-km,(whichwouldbederivedfromthetotalannualbusnetworkfuelconsumption,andthetotalannualpassenger-kmtravelled).Whereasthepowermodelapproachreflectstheinstantaneousmarginalchangeinemissionsbasedondynamicchangesinnumberofpassengers.Therefore,tounderstandtheimpactofadecisiontotakeabusornot,thetwoapproacheswillgivedifferentanswers–theoutcomefromtheconventionalapproachwillbethatemissionsarereducedbynottakingthebus,whiletheoutcomefromthepowermodelapproachwillbethatthereisonlyasmallmarginalreductioninemissions,becausethebusisrunninganyway.However,foracompanyreportingitsannualbusinesstravelemissions,itmakessensetousetheconventional(averageemissions)approach.Toextendtheanalogyabitfurther,verydifferentresultsariseifconsideringtravelbycar–thenthemarginalandaverageemissionsaremuchmoresimilartoeachother,andindeedthemarginalemissionsoftravellingbycarwouldbehigherthantheaveragewhentravellingalone(astheaveragewouldassumeanaveragenumberofpassengersgreaterthanone).Thisreflectsthefactthatacarisnotahighlysharedresource,unlikethepublictransportnetworkortheinternetnetwork.Oneareaforfurtherinvestigationofthepowermodelapproachishowtoappropriatelyallocatethebaseloadenergy.Thepowermodelapproachinthiswhitepaperallocatestheenergyperuserandtimeofuse,howeveritisdifficulttoestablishwhatisanappropriateallocationoftime–isitconnectiontime,ordownloadtime,andhowisidletimebestallocated?Indeed,otherallocationmethodscanalsobeconsidered,suchasconsideringtheutilityorvalueoftheservicebeingused,orthecontributiontopeakdatademand.However,ultimatelyusingamethodthatistransparentandpracticalisalsoimportant.Theuseoftheallocationmethodshouldalsobeconsistentforotherservicesthatusethenetwork,otherwisealloftheenergyandcarbonmaynotbefullyallocatedandaccountedfor.This,then,raisesthequestionofwhataretheimpactsandconsequencesforotherservices?Finally,thiswhitepaperhasidentifiedareasforfurtherinvestigationandimproveddata.Inordertoimproveunderstandingandmeasurement,andtoinformdecisionmaking,moregranularmethodologiesandmoregranulardataisneeded.Inadditiontofurtherworkonrefiningtheallocationmethodologiesdiscussedabove,thereisaneedformoredatainordertoprovidegreaterinsightsthanpresentedinthiswhitepaper.Specifically,moredetailedandconsistentdataonnetworkenergyandcarbonintensitywouldbehelpfultounderstandthevariabilityinnetworkenergyintensity,reducetheuncertainty,provideregularupdatedfigures,andhelpinvalidationofthepowermodelcoefficientsthatareusedinthepowermodelapproach.Ideally,energyintensityfiguresfordifferentnetworktechnologies(e.g.fixedADSL,fixedfibre,2G,3G,4G,5G)wouldbeveryuseful.However,itisrecognisedthatnetworkoperatorsmaynotwishtopublishthislevelofdetail,andanonymisedaggregateddatacouldbecollatedthroughorganisationssuchasGSMA,ETNOorITU(similartotherolethattheWorldSteelAssociationundertakesforthesteelindustryortheIMOfortheshippingindustry).Similarly,itwouldbeusefultohavemoreconsistentandcomprehensiveinformationondatacentreenergyandemissions(although,asnotedinthiswhitepaper,thedatacentrecomponentmakesonlyasmallcontributiontotheoverallfootprintofvideostreaming).Greaterinformationonuserbehaviour,intermsoftypesofviewingdevices,andmixofservicesandconnecteddevices,wouldagainhelptoimprovetheanalysisofthefootprintofvideostreaming,andidentifyanylongertermtrends.Twootherareasforfurtherinvestigationarebetterunderstandingofthefactorsthatdrivepeaknetworkdatademand,andmoredetailedanalysisoftheimpactoftheembodiedemissionsofdevicesandequipmentonthecarbonfootprintofvideostreaming.CarbonimpactofvideostreamingQuestionsandAnswers878.QuestionsandAnswersThepurposeofthisQandAsectionistohighlightsomeofthequestionsthatmayberaisedfromthiswhitepaper,andsummarisesomeofthepointsthatarecoveredinmoredetailinthewhitepaper.88CarbonimpactofvideostreamingQuestionsandAnswersWhatisacarbonfootprint?Acarbonfootprintmeasuresthetotalgreenhousegasemissionscauseddirectlyandindirectlybyaperson,organisation,serviceorproduct.Itismeasuredintonnesorkgofcarbondioxideequivalent(CO2e),combiningtheimpactofdifferentgreenhousegasesintoonefigureequivalenttoifitwereallCO2,basedontheirwarmingpotential.WhatisaCDN?AContentDeliveryNetwork(CDN)actsasalocalstore(orcache)fordigitalcontentontheinternet.TheCDNcontentisonlyupdatedfromtheoriginserverwhenthereisnewcontentoranewversion.Particularlyforvideostreamingthisisveryuseful,sowhenyouarewatchingavideoyouwillbeaccessingthelocalversionofthevideoratherthanthehostedversionattheoriginserver(whichmaywellbeinanothercountry).Thisgivesyouabetterviewingexperience,itreduceslatency(thetimetakenforthedatatogettoyouacrosstheinternetfromwhereitisstored),meanslesswaitingandlessbuffering.CDNsalsosignificantlyreducethetotaldatatrafficacrosstheinternet,particularlyforinternationaltraffic,andsubmarinecabletraffic,becausetheyavoidtheneedtotransmitlargedatavolumesfromtheoriginserverdirectlytoyoueverytimeyouwatchvideostreaming.ManyvideostreamingservicesuseoneofthethirdpartyCDNs,whichwillhavepresenceacrosstheglobe.Particularlyinlargercountries,aCDNwillhavemultiplepointsofpresence,locatedclosetothelargerpopulationcentres.SomevideostreamingservicesoperatetheirownCDN.NetflixoperatesitsownCDN,OpenConnect,whichhasapresenceinnearlyallcountriesthatitoperatesin.Whatistheinternet?Theinternetisthenetworkinfrastructurethatconnectsalldevicessothattheycanexchangedata.EverydeviceontheinternethasanIPaddresssothateachdeviceknowswheretosenddata.Theinternetisoperatedbynetworkoperators(telecommunicationscompaniesandInternetServiceProviders–ISPs),andthenetworkoperators’networksareconnectedtoothernetworkoperators’networks.Colloquially,theinternetcanalsomeananythingthatyoucandoorlookupontheinternet.So,youcouldchecktheweatherforecastontheinternet.Todothisyouwouldconnectyourdevice(PC,laptop,orsmartphone)overtheinternetnetworktoawebsiteonaserverinadatacentre,whichwouldholdasummaryoftheweatherforecastinformation.Theweatherforecastisgeneratedinahigh-performancecomputingdatacentrethatisalsoconnectedtotheinternetnetwork.So,colloquially,“theinternet”canalsoincludeallthedatacentresthatareconnectedtotheinternet.Whenyouviewawebpageontheinternet,youareusingawebbrowseronyourdevice(e.g.smartphone,tabletorlaptop)toviewwebpagesandcontenthostedonserverslocatedindatacentres,whichareconnectedtoyoubytheinternetnetwork.89CarbonimpactofvideostreamingQuestionsandAnswersHomerouter–thehomerouterconnectsyoutotheinternetandison24hoursaday,usingenergyallthetime.Therouterisbeingusedbyvariousdevicesandfunctionsthroughouttheday(e.g.internetbrowsing,emails,work,videocalls,videostreaming).Theamountofenergyusedisfairlyconstantnomatterhowmuchdataisbeingdownloaded.Therearetwoobviousallocationmethods:1.Bydataconsumed;2.Bytimeconsumed.Thedataconsumedallocationmethodassumesanaverageamountofdailydataused,andforonehourofvideoallocatestheenergybasedonthedatausedforonehourofvideo.Theconventionalmethodinthiswhitepaperusesanaveragehouseholddatausageof294GB/month,whichisabout10GB/day.HDvideostreamingatabitrateof6.67Mbpsisequivalentto3GBforonehour.So,onehourvideostreaminguses3/10ofthedailyaveragedata(orjustunderathird).Thisisthenmultipliedbythedailyenergyuseofthehomerouter(10Wx24h),whichgivesabout70Whforonehourofvideostreaming.Youcanseehowthisallocationmethodwouldallocatemorethan100%oftheenergyifyouwatchedfourhoursofvideoinoneday.Thetimeconsumedallocationmethodsimplyassumesthatyouareusingtherouterforonehour,outofthe24hoursintheday.So,thiswouldallocate10Whforonehourofvideostreaming.Themethodusedinthepowermodelapproachinthiswhitepaperalsoallocatesidletime-relatedenergy(whennodataisbeingconsumed),andalsoconsidersthatmultipleusersordevicesmaybeusingtherouterinahousehold(whichvariesbycountry).Thisresultsinanallocationoftheenergyofabout3Whforonehourofvideostreaming.Thisexampleservestoillustratethedifferencesandtheimportanceofallocationapproaches.Bothapproachesarevalidandbothusereasonableassumptions,itislikelythatthepowermodelapproachrepresentsanunderallocationandtheconventionalapproachanoverallocation.Allocation–whatisitandwhyisitimportant?Measuringthecarbonfootprintofaproductorserviceoveritslifecyclerequirescalculatingtheuseofenergyandotherresourcesthatcausegreenhousegasemissionsforeachofthedifferentlifecyclestages.Inmanycasesaparticularproductwillshareuseofresourceswithotherproducts,andsotheresourcesneedtobeallocatedbetweenthedifferentproducts.Forexample,afactorymaymake12differentproducts;ifyouknowthetotalenergyusedbythefactoryhowdoyouallocatethistothe12products?Youcouldallocateitequallybasedonthetotalnumberofproductsproduced,butdifferentproductsmayneedverydifferentamountsofenergytomanufacturethem,anditmaybedifficulttogetaccurateinformationontheenergyperproduct.Forvideostreamingtherearemultiplestageswhereallocationisimportant,andthedifferentallocationmethodscangivedifferentresults.Allocationapproachesforeachstageareexplainedfurtherbelow,startingwithyoutheviewer,andworkingbackthroughthelifecycletowherethevideocontentoriginates.Lifecyclestagesandallocationapproaches:•TV•Homerouter•Internettransmission•DatacentresandCDNSWatchingonaTV–whentheTVitisswitchedon,andyouarewatchingavideoforanhour,thenalloftheTV’senergyforthathourcanbeallocated100%towatchingthevideo.WeareassumingthattheTVisswitchedoff(orinlowpowerstandbymode)whenyouarenotwatching,andthestandbypowerisverylow,andisthereforeexcludedfromtheallocationmethodusedinthiswhitepaper.(Amoredetailedanalysiswouldalsoconsidertheaveragestandbyenergyandallocateittothedifferentviewingtimesduringtheday.Asthestandbypowerissmallitwillhaveaminimalimpactontheoverallestimation).90CarbonimpactofvideostreamingQuestionsandAnswersInternetdatatransmission–theinternetconsistsofhundredsofthousandsofnetworkroutersthatareallconnectedtoeachotherandmanagethetransmissionofdatafromthesourcetoyourhomerouter.Similartothehomerouter,theseroutersarecontinuouslyon,usinganearlyconstantamountofenergyallthetime,varyingonlyslightlydependingontheamountofdatatraffic.Theinternetisprovidingconnectivitytomultipledevicesformultiplepurposes,soagaintherearepotentiallymanydifferentwaysthattheenergycouldbeallocated.Inthiswhitepapertheconventionalapproachusesanaverageenergyperdatatrafficvalue(inkWh/GB),whilethepowermodelapproachusesabaseloadpowerelement(independentofthedatatraffic),andadynamicpowerelement(relatedtothedatatraffic),whichmorecloselyrepresentshowthenetworkenergyuseactuallyrespondsinrealtimetodatatrafficvolumes.Thepowermodelapproachthereforeallocatesamuchloweramountofenergyforvideostreamingthantheconventionalapproachdoes–thisisbecauseitassumesthatthereiseffectivelyanenergycostforbeingconnectedtotheinternet,nomatterhowmuchdataisbeingused.Thismeansthatotherusersandserviceswillbeallocatedahigheramountofenergyunderthepowermodelapproachthantheconventionalapproach.Theconventionalapproachallocatesanaverageamountofenergybasedondatausage,whereasthepowermodelapproachallocatesamarginalamountofenergybasedondatausage,plusafixedamountofenergyforbeingconnected.DatacentresandCDNs–thesehostthevideocontent,encodeandprepareitforvideostreaming,andstorealocalversionofthevideoforstreamingtotheend-user.Thus,theseresourcesaresharedwithalltheusersofthevideostreamingservice.Inthiswhitepaper,boththeconventionalandpowermodelapproachesusethesamemethod,whichistakingthetotalenergyusedbythedatacentresandCDNforvideostreaming,anddividingbythetotalhoursofvideostreaming.HowcanIreducemycarbonfootprintfromvideostreaming?Well,actually,thecarbonfootprintofwatchinganhourofvideostreamingisnotverymuch.Aboutthesameasboilingthekettletomakeacupoftea,ormicrowavingabagofpopcorn.So,youarenotgoingtosavetheplanetbychangingyourviewinghabits.ProbablythemostusefulthingyoucaneasilydoistoswitchoffyourTVwhenyouhavefinishedwatching.TheenergyusedbytheTVisprobablythemostcarbonintensiveaspectofthevideostreaminglifecycle.Asmallerdevicelikeatabletorasmartphonewillusemuchlessenergy.Ofcourse,thenumberofpeopleviewingeachdevicewillalsohaveanimpact–fourpeoplewatchingthesameTVtogetherwilluseaboutthesameamountofdeviceenergyasifthosefourpeoplewerewatchingonseparatelaptops,butaboutfivetimesmoredeviceenergythanthefourpeopleeachwatchingonseparatetablets.Butasthecarbonfootprintisnotverymuchtostartwiththatshouldnotbethemainreasonforchoiceofviewingdevice.Asaconsumer,youcaninfluencethecarbonfootprintbycheckingthatyouhavechosenanenergyefficientTVwhenyoubuyanewTV,andtheotherthingwouldbetoswitchtoadomesticrenewableenergytariff.Themanufacturersofdevicesandoperatorsofnetworksanddatacentrescananddoreducethecarbonfootprintoftheirproductsbymanufacturingenergyefficientdevices,andusingrenewableenergyintheirmanufacturing,andbyusingrenewableelectricitytooperatethenetworksanddatacentres.91CarbonimpactofvideostreamingQuestionsandAnswersDoesHDusedoubletheenergyofSD?Theshortanswerisno.Formostofthestagesinvideostreaming,theenergyusedoesnotinstantaneouslyvarysignificantlywiththeamountofdataused.So,thereisonlyamarginaldifferenceintheamountofenergyusedbetweenstreaminginHD(highdefinition)andSD(standarddefinition).However,iftheaccountingfortheenergyisdonepurelyonanaverageenergyintensityperdatavolume(kWh/GB),thenitwouldshowthatsomeofthestagesusetwicetheamountofenergyinHDasSD,butthisdoesnotreflecttheimmediateenergyuse.Thecaseissimilarforvideoconferences–regardlessofwhetheryouhaveyourvideoonornot,ithasonlyamarginalimpactonthetotalenergyused.Fortheinternetnetworkstageananalogywouldbeabusnetwork.Abuswillusealmostthesameamountoffuelwhetherthereare20peopleor40peopleonthebus(orindeediftherearenopeopletravellingonthebus).Thereisafixedamountofenergyrequiredsimplytoprovidetheservice,irrespectiveoftheamountofusage.However,fromanaccountingperspectivethebusemissionsmaybeallocatedtogiveanaverageperpassenger(orperpassenger-km)emissionfactor.Thisillustratestheimportantdifferencebetweenadynamic(ormarginalapproach)forallocation,andanaverageallocationapproach.Bothareusefulfordifferentpurposes,withthemarginalapproachbeingmorerelevantfordecisionmakingandreflectingshort-termactions(e.g.shouldIcatchthebus,orwalkortakethecar),whiletheaverageallocationapproachismorerelevantforaccountingpurposes,andlongertermdecisions(e.g.howmanybusesareneededforthenetwork).Forafurtherdiscussionofthisseealsothequestiononallocation,above.So,whatisdrivingthetotalenergyrequirementsoftheinternetnetwork?Thisfollowsonfromthepreviousquestion,however,thisonedoesnothaveasimpleanswer,andwouldbenefitfromfurtherresearch.ThisquestionisalsocoveredintheDiscussionsectionofthewhitepaper.Theinternetnetworkisdesignedtomanageapeakdatatrafficload,andasthenetworkisbeingcontinuallyupgradeditisactuallybuilttohandleexpectedfuturepeakdemand.Thetotaldemanddeterminesthecapacityofthenetworkandthereforetheenergyrequiredtorunthenetwork.So,toensurethatthereisnotcongestionthenetworkneedstobeabletohandlepeakdemand.Thetotalenergydemandwillberelatedtothepeakcapacityratherthantheaveragedemand.Thispictureiscomplicatedmorebythefactthatnewnetworkequipmentwillbemoreenergyefficient–i.e.willbeabletohandlemoredatatrafficforthesameamountofenergy.Thiscanbeillustratedbyreturningtoourbusanalogy.Ifthebushasamaximumcapacityof50people,andthereare55peoplewaitingforthebus–thenfivepeoplewillnotbeabletogetonthebus.Withtheinternetnetworkthereisthesamedifficultyifatpeaktimesthedemandonthenetworkisgreaterthanthecapacity,thisresultsincongestion,withdatapacketsbeing“dropped”,resultingineitherbufferingorlossofpicturequalityifyouarewatchingvideostreaming.Thedifferenceisthatforthebusnetwork,itmaybepossibletoputonextrabusesatpeaktimes(andhavelessbusesatoff-peaktimes),whereasfortheinternetnetworkthepeakcapacityisbasicallyfixed,andcannotbevariedonadailybasis–itrequiresinvestmentinnewnetworkequipmentandtechnologytoincreaseit,sofromthatperspectiveitismoresimilartotheroadnetworkthatthebusesrunon.Theotherdifferenceisthattechnologyimprovementswithnetworkshasmeantthatnetworkcapacitycangrowtoaccommodateannualincreasesofover25%indatatraffic.HD92CarbonimpactofvideostreamingQuestionsandAnswersDoeswatchingtwohoursofvideostreamingusetwicetheenergyofonehour?Onaverageyes,however,asdiscussedabove,theinternetnetworkusesmuchthesameenergywhateverthetotaldatatrafficis.So,themainmarginaldifferencebetweenwatchingtwohoursandonehourofvideostreamingisintheenergyusedbytheend-userviewingdevice(i.e.theTV,laptoportablet).CarbonimpactofvideostreamingReferences939.References94CarbonimpactofvideostreamingReferencesAccessPartnership,2020,TechPolicyTrends2in2020(link)Airtel2016,‘BhartiAirtel'sfocusonGreennetworkinitiativestransforms40,000telecomtowersacrossIndia’,(link)Amazon,TheClimatePledge(link)Andrae,A.S.G.,&Edler,T.,2015:Onglobalelectricityusageofcommunicationtechnology-trendsto2030.Challenges,6,117-157(link)Andrae,A.S.G.,Hu,L.,Liu,L.,Spear,J.,&Rubel,K.(2017):DeliveringtangiblecarbonemissionandcostreductionthroughtheICTsupplychain.InternationalJournalofGreenTechnology,3,1-10(link)Andrae,A.S.G.,2019:ComparisonofSeveralSimplisticHigh-LevelApproachesforestimatingtheglobalenergyandElectricityuseofICTnetworksanddatacenters.InternationalJournalofGreenTechnology(link)Andrae,A.S.G.,2020:Newperspectivesoninternetelectricityusein2030.Sweden(link)Apple,2019(link)Apple,2020,‘ProductEnvironmentalReportiPhone12’(link)Apple,2021a:GeneralBatteryInformation(link)Apple,2021b:AppleandBeatsProductInformationSheet(link)Aslan,J.,Mayers,K.,Koomey,J.G.andFrance,C.,2018,‘ElectricityintensityofInternetdatatransmission:Untanglingtheestimates’.JournalofIndustrialEcology,22(4),pp.785-798(link)Aslan,Joshua,2020,ClimateChangeImplicationsofGamingProductsandServices,DoctoraldissertationsubmittedforthedegreeofEngineeringDoctorateinSustainabilityforEngineeringandEnergySystems(link)BBC,2011,JignaChandaria,JeffHunter,andAdrianWilliams,‘Thecarbonfootprintofwatchingtelevision,comparingdigitalterrestrialtelevisionwithvideo-on-demand’,Proceedings2011IEEEInternationalSymposiumonSustainableSystemsandTechnology(ISSST),Volume:1,Pages:1-6(link)BBC,2020,UsingBehaviouralDatatoAssesstheEnvironmentalImpactofElectricityConsumptionofAlternateTelevisionServiceDistributionPlatforms(link)BelkhirL.,ElmeligiA.,2018,AssessingICTglobalemissionsfootprint:Trendsto2040&recommendations,JournalofCleanerProduction177(2018)448-463Bitkom,2020,‘NachhaltigkeitvonStreaming&Co.,EnergiebedarfundCO2-AusstoßderVideonutzungimNetz’(link)BloombergNEF,2020:CorporatePPADealTracker,March2020(link)Bordage,Frédéric,2019,Theenvironmentalfootprintofthedigitalworld(link)Cisco,2015:TheHistoryandFutureofInternetTraffic(link)Cisco,2018a:CiscoGlobalCloudIndex-ForecastandMethodology,2016-2021(link)Cisco2018b,‘CiscoAnnualInternetReport-CiscoAnnualInternetReportHighlightsTool‘(link)Cisco2018c,‘CiscoVisualNetworkingIndex(VNI)CompleteForecastUpdate,2017–2022,Americas/EMEARCiscoKnowledgeNetwork(CKN)Presentation’(link)Cisco2019a:CiscoVisualNetworkingIndex-GlobalMobileDataTrafficForecastUpdate,2017-2022(link)Cisco,2019b:CiscoVisualNetworkingIndex:ForecastandTrends,2017-2022WhitePaper,CiscoForecastandMethodology(link)Cisco2020:CiscoAnnualInternetReport(2018-2023),WhitePaper(link)ClimateNeutralDataCentrePact,2021(link)CNBCTV18,2020,‘COVID-19lockdown:India'sdiesel-runtelecomtowerscouldrundry’(link)Cogent,2020a:Corporateresponsibility(link)Cogent,2020b:OurEnvironment(link)95CarbonimpactofvideostreamingReferencesComscore,2014,TheUSTotalVideoReport(link)Coworker.com2019,‘RankingtheWorld’sFastest(&Slowest)InternetSpeeds(link)D+RInternational,2020:2019AnnualReportVoluntaryAgreementforOngoingImprovementtotheEnergyEfficiencyofSmallNetworkEquipment(link)Das,S.&Mao,E.,2020:TheglobalenergyfootprintofinformationandcommunicationtechnologyelectronicsinconnectedInternet-of-Thingsdevices(link)DepartmentofEnvironment,Food&RuralAffairs,2020,SustainableICTanddigitalservicesstrategy:targetsfor2020-2025’(link)Eartheasy,2021(link)EfficientGaming,2018(link)Ericsson,2020:MobilityReport(link)EuropeanCommissionandJointResearchCouncil,2008,CodeofConductforEnergyEfficiencyinDataCentreshttps://ec.europa.eu/jrc/en/energy-efficiency/code-conduct/datacentresEuropeanCommission,2009,ImplementingDirective2005/32/ECoftheEuropeanParliamentandoftheCouncilwithregardtoecodesignrequirementsfortelevisions(link)EuropeanCommission,2015,VoluntaryagreementforGameConsoles(link)EuropeanCommission,2019,AEuropeanGreenDeal2019(link)EuropeanCommission,2020a,ICTImpactstudy–Finalreport(link)EuropeanCommission,2020b,EnergyefficientCloudComputingTechnologiesandPoliciesforanEco-friendlyCloudMarket(link)EuropeanCommission,2020c,Directorate‑GeneralforCommunicationsNetworks,ContentandTechnology,TheEuropeanDigitalStrategy,2020(link)EuropeanCommission,2020d,SupportingaGreenTransition(link)EuropeanCommissionandJointResearchCentre,2021a,2021BestPracticeGuidelinesfortheEUCodeofConductonDataCentreEnergyEfficiency(link)EuropeanCommission,2021b,NewEUenergylabelsapplicable(link)EuropeanCommission,2021c,EuropeanGreenDigitalCoalition(link)EuropeanCommission,2021d,Environmentalmanagement&performance–sectoralreferencedocumentfortelecommunications/ICTservices(link)Eurostat,2021:Averagehouseholdsize-EU-SILCsurvey(link)FinancialTimes,10February2021,HowTechcompanieswentbigongreenenergy(link)France24,2019,“ChillyourNetflixhabit,climateexpertssay”(link)Freitag,Charlotte,MikeBerners-Lee,KellyWiddicks,BranKnowles,GordonBlairandAdrianFriday,2020,TheclimateimpactofICT:Areviewofestimates,trendsandregulations(link)FrenchSenatedraftlegislation,2020(link)Frequencycast,2021(link)GermanFederalMinistryforEnvironment,2020a,Germangovernmentenvironmentaldigitalagenda,2020(link)GermanFederalMinistryforEnvironment,2020b,10September2020,PressRelease,Videostreaming:datatransmissiontechnologycrucialforclimatefootprint(link)GeSI,2008,SMART2020,Enablingthelow-carboneconomyintheinformationage(link)GeSI,2020,DigitalSolutionsforSupportingtheAchievementoftheNDC(link)GlobeNewsWire2019,‘HyperscaleDataCenterCountPassesthe500MilestoneinQ3(link)96CarbonimpactofvideostreamingReferencesGoogle,2020,24/7by2030:RealizingaCarbon-freeFuture(link)GreenElectronicsCouncil2019,‘Stateofsustainabilityresearchfornetworkequipment,SmallNetworkEquipment,FinalReport’,FraunhoferIZM(link)GSMA,2019,ClimateActionHandbook(link)GSMA,2020:COVID-19NetworkTrafficSurgeIsn’tImpactingEnvironmentConfirmTelecomOperators(link)(AccessedMarch2021)Hintemann,2018:DataCentres2017-BoomisleadingtoasignificantincreaseintheenergydemandofdatacentersinGermanyin2017.BorderstepInstititue,Berlin(link)Hintemann,R.andHinterholzer,S.,2019:Energyconsumptionofdatacentresworldwide.BorderstepInstititue,Berlin(link)Huawei,2019,Bringingthebenefitsoftechnologytoeveryone,OpeningremarksbyChairmanLiangHuaatthelaunchofHuawei's2018SustainabilityReport(link)HuberP.,MillsMP.1999.Digmorecoal—thePCsarecoming.Forbes(May31):70–72IEA2020a,DataCentresandDataTransmissionNetworks(link)IEA,2020b,EmissionFactors2020(link)IEA,2020c,‘Thecarbonfootprintofstreamingvideo:fact-checkingtheheadlines’(link)ITIF,2020:BeyondtheEnergyTechlash:TheRealClimateImpactsofInformationTechnology(link)ITUPublications,2019,Turningdigitaltechnologyinnovationintoclimateaction(link)ITU,2020:RecommendationITU-TL.1470-GreenhousegasemissionstrajectoriesfortheinformationandcommunicationtechnologysectorcompatiblewiththeUNFCCCParisAgreement(link)Kamiya,G.,2020:GlobalICTenergytrends.PresentationatthetechUKconferenceinLondonTechWeekon‘DecarbonisingData’,2020(link)(AccessedFebruary2020)Koomey,J.,Berard,S.,Sanchez,M.andWong,H.,2010.Implicationsofhistoricaltrendsintheelectricalefficiencyofcomputing.IEEEAnnalsoftheHistoryofComputing,33(3),pp.46-54.Koomey,JonathanG.,StephenBerard,MarlaSanchez,andHenryWong.2011a."ImplicationsofHistoricalTrendsinTheElectricalEfficiencyofComputing."IEEEAnnalsoftheHistoryofComputing.vol.33,no.3.July-September.pp.46-54(link)Koomey,Jonathan.2011b.Growthindatacenterelectricityuse2005to2010.Oakland,CA:AnalyticsPress.August1(link)Koomey,Jonathan,andSamuelNaffziger.2015."Efficiency'sbriefreprieve:Moore'sLawslowdownhitsperformancemorethanenergyefficiency."InIEEESpectrum.April(link)Koomey,Jonathan,andSamuelNaffziger.2016."Energyefficiencyofcomputing:What'snext?"InElectronicDesign.November28(link)Lange,Steffen&Pohl,Johanna&Santarius,Tilman,2020."Digitalizationandenergyconsumption.DoesICTreduceenergydemand?,"EcologicalEconomics,Elsevier,vol.176(C)(link)Leiserson,CharlesE.,NeilC.Thompson,JoelS.Emer,BradleyC.Kuszmaul,ButlerW.Lampson,DanielSanchez,andTaoB.Schardl.2020."There’splentyofroomatthetop:WhatwilldrivecomputerperformanceafterMoore’slaw?"Science.vol.368,no.6495.pp.eaam9744(link)Malmodinetal.,2013:ThefuturecarbonfootprintoftheICTandE&Msectors.InProceedingsoftheICTforSustainability(ICT4S),Zurich,Switzerland,9–12February2013.Malmodin,J.andLundén,D,2016:TheEnergyandCarbonFootprintoftheICTandE&MSectorinSweden1990–2015andBeyond.InProceedingsofthe4thInternationalConferenceonICTforSustainability(ICT4S),Amsterdam,TheNetherlands,29August–1September2016.97CarbonimpactofvideostreamingReferencesMalmodin,J.andLundén,D.,2018a:TheenergyandcarbonfootprintoftheglobalICTandE&Msectors2010–2015.Sustainability,10(9):p.3027.109(link)Malmodin,J.andLundén,D.,2018b:TheelectricityconsumptionandoperationalcarbonemissionsofICTnetworkoperators2010-2015.ReportfromtheKTHCentreforSustainableCommunicationsStockholm,Sweden(link)Malmodin,J.,2019:EnergyconsumptionandcarbonemissionsoftheICTsector.PresentationgivenatEnergimyndigheten,Stockholm,19Dec2019Malmodin,J.,2020a:TheICTsector’scarbonfootprint.PresentationatthetechUKconferenceinLondonTechWeekon‘DecarbonisingData’,2020(link)(AccessedFebruary2020)Malmodin,J.,2020b:Thepowerconsumptionofmobileandfixednetworkdataservices-Thecaseofstreamingvideoanddownloadinglargefiles.ElectricityGoesGreen,Berlin,1September2020(link)Masanet,E.etal.2020a,Recalibratingglobaldatacenterenergy-useestimates,Science,367(6481),984-986(link)Masanet2020b,EricMasanetandNuoaLei,Howmuchenergydodatacentersreallyuse?,Aspenglobalchangeinstituteenergyproject,March2020QuarterlyResearchReview(link)Microsoft,2020,Microsoftwillbecarbonnegativeby2030(link)MinistryofEconomicAffairsandClimatePolicy,2020DutchDigitalisationStrategy(link)Mills,N.andMills,E.,2015:TamingtheEnergyUseofGamingComputers(link)NENT2020,‘AlookattheclimateimpactofNENTGroup’sstreamingservices’(link)Netflix2020,‘EnvironmentalSocialGovernance,2019SustainabilityAccountingStandardsBoard(SASB)Report’(link)Netflix2021,‘HowtocontrolhowmuchdataNetflixuses’(link)NRDC(NaturalResourcesDefenceCouncil),2014,‘TheLatest-GenerationVideoGameConsoles,HowMuchEnergyDoTheyWasteWhenYou’reNotPlaying?’(Link)Obringer,R.etal.,2021,TheoverlookedenvironmentalfootprintofincreasingInternetuse,Resources,ConservationandRecycling,Volume167,April2021,105389.https://doi.org/10.1016/j.resconrec.2020.105389Ofcom2020a,‘InternationalBroadbandScorecard2020:interactivedata’(link)Ofcom2020b,‘Adults’MediaUse&Attitudesreport2020’(link)Pihkola,H.,Hongisto,M.,Apilo,O.andLasanen,M.,2018,‘Evaluatingtheenergyconsumptionofmobiledatatransfer—fromtechnologydevelopmenttoconsumerbehaviourandlifecyclethinking’.Sustainability,10(7),p.2494.PopulationReferenceBureau,2020,accessed4March2021(link)Preist,C.,Schien,D.andBlevis,E.,2016,May.Understandingandmitigatingtheeffectsofdeviceandcloudservicedesigndecisionsontheenvironmentalfootprintofdigitalinfrastructure.InProceedingsofthe2016CHIConferenceonHumanFactorsinComputingSystems(pp.1324-1337).Reuters,2020,Article:Microsofttoeraseitscarbonfootprint,pastandfuture,inclimatepush(link)RoyalSociety,2020,Digitaltechnologyandtheplanet,Harnessingcomputingtoachievenetzero(link)Sandvine,2020a,TheMobileInternetPhenomenaReport(link)Sandvine,2020b,TheGlobalInternetPhenomenaReport,COVID-19Spotlight(link)ScienceBasedTargetsInititative,ICTsectorguidance,2020(link)ScienceBasedTargetsInitiative,website,CompaniesTakingAction,accessed4March2021(link)Shehabi,A.,Walker,B.,&Masanet,E.(2014).TheenergyandgreenhousegasimplicationsofinternetvideostreamingintheUnitedStates.EnvironmentalResearchLetters,9(5),054007.98CarbonimpactofvideostreamingReferencesShehabi,A.,Smith,S.J.,Masanet,E.,&Koomey,J.(2018).DatacentergrowthintheUnitedStates:decouplingthedemandforservicesfromelectricityuse.EnvironmentalResearchLetters,13(12),124030.Singhetal.,2019:ResidentialConsumerElectronicsEnergyConsumptionintheUnitedStatesin2017.FraunhoferUSACenterforSustainableEnergySystems(link)Statcounter,2021,MobileOperatingSystemMarketshareWorldwideApr2020-Apr2021(link)Stobbeetal.,2015:DevelopmentofICT-relatedelectricitydemandinGermany.BorderstepInstitute,Berlin(link)Stobbeetal.,2021:MethodologyforModellingtheEnergyandMaterialFootprintofFutureTelecommunicationNetworks.In:KishitaY.,MatsumotoM.,InoueM.,FukushigeS.EcoDesignandSustainability(EDS)II.SustainableProduction,LifeCycleEngineeringandManagement.Springer,Singapore(link)techUK,2017a,MeasuringDataCentrePerformance(link)techUK,2017b,DataCentrestandardsmap(link)TechwireAsia,2020,‘India’sinternetconnectivitydoubledinjustfouryears’(link)Telefónica,2019:ConsolidatedManagementReport,2019(link)Telefónica,2020:ConsolidatedManagementReport,2020(link)TheGuardian2013,‘AhistoryofmediastreamingandthefutureofconnectedTV’(link)TheShiftProject,2019a:LeanICT-TowardsDigitalSobriety(link)TheShiftProject,2019b,“ClimateCrisis:theunsustainableuseofonlinevideo”(link)TheShiftProject,2020,‘DidTheShiftProjectreallyoverestimatethecarbonfootprintofonlinevideo?’(Link)TheSpruce,2021(link)UnitedStatesCensusBureau(link)UnitedStatesEnvironmentalProtectionAgency,2008,FactSheetonNationalDataCenterEnergyEfficiencyInformationProgramUptimeInstituteglobaldatacentersurvey,2019(link)Urban,B.,Roth,K.,Singh,M.andHowes,D.2019,‘ResidentialConsumerElectronicsEnergyConsumptionintheUnitedStatesin2017’(link)Vo,N.S.,Duong,T.Q.,Tuan,H.D.andKortun,A.,2017.Optimalvideostreamingindense5GnetworkswithD2Dcommunications.IEEEAccess,6,pp.209-223.Vox,2018,YoucanwatchNetflixonanyscreenyouwant,butyou’reprobablywatchingitonaTV(link)WaldropM.,2016,MorethanMoore,Nature,Vol530,11February2016,pp144-147(link)West,D.2014,‘TheEvolutionofVideoStreamingandDigitalContentDelivery’(link)CarbonimpactofvideostreamingAppendix9910.Appendix100CarbonimpactofvideostreamingAppendixAppendix1:AdditionalmodellingparametersTable9.RepresentativedevicemixforstreamingservicesDeviceNetworkVideoQualityShareofTotalDevicesAndroidCellularnetworkSavedatasetting1.2%AndroidFixednetworkSD(480p)1.2%AndroidCellularnetworkAutomaticdatasetting0.9%AndroidFixednetworkFHD(1080p)0.9%AndroidCellularnetworkMaximumdatasetting0.1%AndroidFixednetwork4K(2160p)0.1%iPhoneCellularnetworkSavedatasetting0.5%iPhoneFixednetworkSD(480p)0.5%iPhoneCellularnetworkautomaticdatasetting0.3%iPhoneFixednetworkFHD(1080p)0.3%iPhoneCellularnetworkMaximumdatasetting<0.1%iPhoneFixednetwork4K(2160p)<0.1%ComputerFixednetworkFHD(1080p)11.9%50”SmartTVFixednetworkFHD(1080p)27.0%50”SmartTVFixednetwork4K(2160p)1.2%50”TVw/STBFixednetworkFHD(1080p)36.6%50”TVw/STBFixednetwork4K(2160p)1.7%50”TVw/GamingConsoleFixednetworkFHD(1080p)3.3%50”TVw/GamingConsoleFixednetwork4K(2160p)0.2%LaptopFixednetworkSD(480p)6.6%LaptopFixednetworkHD(1080p)5.3%Assumptions:•TVsaccountfor70%ofviewinghours(Vox,2018)•ShareofTVswithperipheralsisanestimate,partiallybasedoninformationavailablefromDIMPACTmembers.Assumed40%ofTVviewingoccursonsmartTVswhichrequirenoperipheral,55%onTVsviaset-topboxandtheremaining5%onTVsviagamingconsole.•Non-TVviewingassumedtobe6%smartphones,12%desktopsand12%laptopsbasedonComscore’sTheUSTotalVideoReport(Comscore,2014).•AndroidandiPhoneshareofsmartphonesestimatedbasedonStatcounter’sMobileOperatingSystemMarketshareWorldwide(Statcounter,2021).Androidassumedtobe72%andiPhone28%.•Simpleassumptionthat50%ofsmartphoneviewingoccursonfixednetworksand50%onmobilenetworks.•ResolutionsharebyviewinghoursisadaptedfromCisco’sVisualNetworkingIndexestimationofGlobalUHDIPvideotrafficfor2021(Cisco,2019b).ShareofSDviewinghoursonTVsisconservativelyassumedtobefullHDviewingasweonlymodel50inTVs.ResultingshareofviewingresolutiononTVsis96%HDviewingand4%4K.DesktopcomputersassumedtoonlystreaminHD.AssumptionthatlaptopsstreamineitherSD(55%share)orHD(45%share).Smartphonesassumedtostream55%inSD,40%inHDand3%in4K.101CarbonimpactofvideostreamingAppendixTable10.EstimatedpowerconsumptionofdevicesforstreamingDevicetypePowerDraw(W)CommentSourceAndroidPhone1SamsungGalaxyS9,has11.55Whbatterywithupto16hourslifetimewhenwatchingvideosiFixitforbatteryWh,SamsungforbatterylifetimeiPhone1Assuming~11Whbatteryand10hofbatterylifeifstreamingApple,2021a,Apple,2021bDesktopcomputer115DesktopandmonitorSinghetal.,2019Laptop22PortablecomputerSinghetal.,2019NormalTV100Conservativeestimatefor2020offering,butreasonableconsideringolderTVsareusedtowatchNetflixENGIEanalysisusingBestBuybestsellers(Singhetal.,2019)SmartTV100Conservativeestimatefor2020offeringbutreasonableconsideringolderTVsareusedtowatchNetflixENGIEanalysisusingBestBuybestsellers(Singhetal.,2019)GamingConsole89SonyPlayStation4Mills,2015,NRDC,2014andSinghet.al,2019SetTopBox18STBswithDVRusingreportedinstalledbaseofallSTBsintheUSD+RInternational,2020Table11.RegionalmodellingparametersRegionAverageHouseholdSizeAverageperCapitaDevicesandConnectionsElectricalGridEmissionFactorSource(GridYear2018)France2.26.1IEA,2020bGermany2.06.4IEA,2020bSweden1.96.9IEA,2020bUnitedKingdom2.36.9IEA,2020bEurope2.36.2IEA,2020bAveragehouseholdsizerelatestofiguresfor2019sourcedfromPopulationReferenceBureau2020WorldPopulationDataSheet(PopulationReferenceBureau,2020),exceptforEuropewhichissourcedfromEurostat(Eurostat,2021).Averagepercapitadevicesandconnectionsrelateto2020andareinterpolatedfrom2018and2023figuresfromCiscoAnnualInternetReport(Cisco,2020).EuropeisassumedequivalenttoWesternEurope.ElectricalgridemissionfactorsaresourcedfromtheIEA’s2020publicationofemissionsperkWhofelectricityforgridyear2018.EuropegridfactorissourcedfromMemo:Europe(UN)withintheIEAdataset.Whilstreasonablestepshavebeentakentoensurethattheinformationcontainedwithinthispublicationiscorrect,theauthors,theCarbonTrust,itsagents,contractorsandsub-contractorsgivenowarrantyandmakenorepresentationastoitsaccuracyandacceptnoliabilityforanyerrorsoromissions.Alltrademarks,servicemarksandlogosinthispublication,andcopyrightinit,arethepropertyoftheCarbonTrust(oritslicensors).Nothinginthispublicationshallbeconstruedasgrantinganylicenceorrighttouseorreproduceanyofthetrademarks,servicesmarks,logos,copyrightoranyproprietaryinformationinanywaywithouttheCarbonTrust’spriorwrittenpermission.TheCarbonTrustenforcesinfringementsofitsintellectualpropertyrightstothefullextentpermittedbylaw.TheCarbonTrustisacompanylimitedbyguaranteeandregisteredinEnglandandWalesundercompanynumber4190230withitsregisteredofficeat4thFloorDorsetHouse,StamfordStreet,LondonSE19NT.©TheCarbonTrust2021.Allrightsreserved.

1、当您付费下载文档后，您只拥有了使用权限，并不意味着购买了版权，文档只能用于自身使用，不得用于其他商业用途（如 [转卖]进行直接盈利或[编辑后售卖]进行间接盈利）。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。
3、如文档内容存在违规，或者侵犯商业秘密、侵犯著作权等，请点击“违规举报”。

碎片内容

碳中和的最新文档