Improving the SDG energy poverty targets: residential cooling needs in the Global South

ContentslistsavailableatScienceDirect

Energy & Buildings

journalhomepage:www.elsevier.com/locate/enbuild

Improving the SDG energy poverty targets: Residential cooling needs in the Global South

Alessio Mastrucci

^a,∗

, Edward Byers

^a

, Shonali Pachauri

^a

, Narasimha D. Rao

^a,b

aInternational Institute for Applied Systems Analysis (IIASA) - Energy Program (ENE), Schlossplatz, 1 A-2361 Laxenburg, Austria

bYale University, School of Forestry and Environmental Studies, 06511 New Haven, CT, United States

a rt i c l e i n f o

Article history:

Received 31 July 2018 Revised 13 November 2018 Accepted 15 January 2019 Available online 22 January 2019 Keywords:

Sustainable development goals Heat stress

Developing countries Cooling

Decent housing Poverty Energy needs

a b s t r a c t

Withgrowing health risks fromrising temperaturesin theGlobal South, the lackofessential indoor coolingisincreasinglyseenasadimensionofenergypovertyand humanwell-being.Airconditioning (AC)isexpectedtoincreasesignificantlywithrisingincomes,butitislikelythatmanywhoneedACwill nothaveit.Weestimatethecurrentlocationandextentofpopulationspotentiallyexposedtoheatstress intheGlobalSouth. We applyavariable degreedays (VDD)methodonaglobal gridtoestimatethe energydemandrequiredtomeetthesecoolingneeds,accountingforspatiallyexplicitclimate,housing types,accesstoelectricityandACownership.

Ourresultsshowlargegapsinaccesstoessentialspacecooling,especiallyinIndia,South-EastAsia andsub-SaharanAfrica.Between1.8to4.1billion,dependingontherequiredindoortemperaturesand daysofexposure,mayneedACtoavoidheatrelatedstressesundercurrentclimateandsocio-economic conditions.ThisnumberfarexceedstheenergypovertygapindicatedbytheSustainableDevelopment Goal for electricityaccess (SDG7). Coveringthis cooling gap would entail a median energy demand growthof14% ofcurrentglobalresidentialelectricityconsumption,primarilyforAC.Solutionsbeyond improvedAC efficiency,such aspassive buildingand city design,innovativecooling technologies,and parsimonioususeofACwillbeneededtoensureessentialcoolingforallwithminimizedenvironmental damage.Meetingtheessentialcoolinggap,asestimatedbythisstudy,canhaveimportantinteractions withachievingseveraloftheSDGs.

© 2019TheAuthors.PublishedbyElsevierB.V.

ThisisanopenaccessarticleundertheCCBY-NC-NDlicense.

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

1. Introduction

Spacecoolingisthefastestgrowingenergyserviceinbuildings worldwide. Cooling-relatedenergydemandtripledsince 1990[1]. The demandfor airconditioning(AC)is expectedtodramatically increase in developingcountriesundersevere climaticconditions induced by climatic change, population and income growth, and demand forcomfort[2–4].However, manyhouseholdscannot af- fordcoolingsystemsorliveinbuildingswithpoordesign,result- inginpoorservicelevelsandhealthrisks[5].Significantheatwave relateddeathshavebeenreportedinrecentyears,inparticularin denselypopulatedcities[6,7].Accesstoadequateheatingandcool- ing is increasingly seen as a deprivation that demands attention frompolicymakers, asreflectedinrecenteffortsto improvemea- surementandtrackingofenergypoverty[8].

∗ Corresponding author.

E-mail address: mastrucc@iiasa.ac.at (A. Mastrucci).

Morethanonebillion peoplelackaccesstoelectricity[4].The goalofuniversalelectricity accessby 2030setbythe Sustainable DevelopmentGoal(SDG)7remainsanambitiouschallenge.Inthis paper,we showthattheenergypovertygapisevenbroaderthan indicatedby the SDG7electricity accessindicator whenthe need forspacecooling isalsoconsidered. However, ACis considereda luxury,duetoitshighcost,asreflectedinitsassignmenttothetop tierof electricityaccessin theWorld Bank’s multitierframework [8]. With only 8% of the 2.8 billion people living in the hottest regions of the world possessing air-conditioning (AC) [1], access tocooling isa majorequity issue[9].Typically themostaffluent in societycan afford AC,but it is often the aged and poor who are most vulnerable to heat stress, particularly in climates with hot summersand inurban areas [10]. Intropics andsub-tropics, ACisalsonecessarytomaintainacomfortablesleepenvironment [11],importantforallowingthebodytorecoverbetweenhotdays.

Morebroadly,gaps inaccessto modernenergyandtechnologies, poorhousingdesignquality,andclimatechange-relatedheatwaves

https://doi.org/10.1016/j.enbuild.2019.01.015

0378-7788/© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license. ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

exacerbatethisissue[10,12]by increasingtherisk ofheat-related mortalityinmanydevelopingcountries[13].

Ensuring electrificationand affordableaccessto ACsystemsis partofthe solutionto meeting thespacecooling energypoverty gap. This challenge also entails a massive increase in energy demand,grid issues in handlingpeak demand,and relatedenvi- ronmentalstresses, such ason climate change [1]. Policy strate- giesmayincludeincreasingtheefficiencyofACsystems,embrac- ingpassivedesignforbuildingsandcities,anddistrictcoolingand renewableenergy[9].

Previous studies have estimated the demand for cooling at a globalscale [3,14,15]. However, there is a lack ofresearch inde- velopingcountriesbeyondlargescalemodels[16,17].Whilstmany authorsestimatea generalincreaseinenergydemandduetothe futurespreadofACunits[18,19],nobodyhasexaminedwherethe combinationofadverse climaticconditions andpoverty coincide.

The characterization of heat stress in homes as a dimension of energypoverty has thus, scarcely been explored. The energy re- quirementstomeetthisgaparealsounexplored.Additionally,the qualityofexistinghousingstocks,andpotential ofenergysavings measuresareoftennotconsideredeveninconventionalestimates ofcooling energydemand. Identifyingregions where populations lackaccess to suitable indoor comfort andestimating energy re- quirementstocoverthisgapare thereforeessentialto directand prioritizeclimateadaptationeffortsandpoliciestoimplementthe SDGs[20].

In this paperwe estimate spatially explicitresidential cooling needs in the Global South in combination with access to space conditioningtohighlightthelocation andsizeofpopulations po- tentiallyexposedtoheatstress andthegapincoolingenergyde- mand under current climate and socio-economic conditions. We applyavariabledegreedays(VDD)methodtoestimatespacecool- ingneedsatahighspatialresolution,accountingforclimate,hous- ingtypesandspaceconditioningtechnologies.Energygapsarees- timatedby combiningdataon populationdistribution andaccess tocoolingtechnologieswiththeestimatedspatiallyexplicitcooling needs,subjecttoa rangeofindoor temperaturesetpoints, users’

behaviourandbuildingdesignconditions.

The key outcomes of our study include an estimate of the numberofpeople currentlyaffected by the spacecooling energy povertygap,theirdistributionintheGlobalSouth,andtheenergy requiredtofillthisgap.

In the following,we first discussthe state ofthe artonther- malcomfortandwell-being,andglobalenergydemandmodelling (Section2).InSection3wedescribethemethodsweapplyinour estimations.Our key resultson cooling gapsand energyrequire- mentsarepresentedinSection4,withadeeperdiscussionofthese includedinSection5.Finally,inSection6weconclude.

2. Stateoftheart

2.1.Thermalcomfortandwell-being

Much isknownabouttheadverseimpacts ofextremeheaton human health and functioning [21,22]. In particular, heat stress has been linked with increased risk of morbidity and mortality, cognitiveimpairment, restrictedproductivityandeconomic losses [23–27].Differencesinvulnerabilitiesbasedonage,gender,infras- tructureandgeneralhealthstatusare alsowell recognized.How- ever, little consensus exists on temperature thresholds at which theseadverseimpactsareexperienced.Thisisbecauseheatstress depends on an interplay of environmental, contextual and be- haviouralfactorssuch aslocalclimate,buildingconstruction and, occupant clothing.Furthermore,acclimatization varies by geogra- phyandlocal climate.Forthesereasons, thereare differences in estimatedtemperaturethresholdsforheat-relatednegativehealth

outcomes across regions and climates [28,29]. A review ofsome oftheliteraturesuggeststhatheat-relatedhealthrisksriselargely between a band of 21°C to 36°C [30,31]. Most of these studies areforNorthernandWesternnations,butthereissomeevidence that heat thresholds may be slightly higher in tropical and sub- tropical regions. Typical thermal comfort standards, which have largelybeendevelopedforcommercialsettingsintheUKandUSA, uselower temperatures(typicallya basetemperatureof18–22°C has beenused in cooling degree days(CDD) calculations), which resultinlikelyexaggeratedestimatesofenergydemand[32].Inor- dertomoreconservativelyestimateenergydemandtoprovideba- sicthermalcomfortandreflectlocalclimatesintheGlobalSouth, we assumea relativelyconservative indoorset pointtemperature of26°Candexplorethesensitivityofourresultstothisvalue by varyingitbetween20–32°C[11].

2.2. Globalenergydemandmodelsforbuildings

In the last decade, a few globalmodels havebeen developed to estimate the energy demand of buildings and develop future energy scenarios. Some of these models rely on a considerable amount of statistical data to generate energy demand intensi- ties for different world regions, building types and vintages [3]. While statistical models provide sound estimations for the cur- rent building stocks, they are of limited use in predicting the impact of newtechnologies [33]. In contrast,physical simulation models have theability to overcome thisissue,but they rely on a number of assumptions and require a considerable amount of input data [34]. The Degree Day (DD) method is a steady-state methodbroadlyused forenergyanalyses forheatingandcooling services(seee.g.[2,14,35]).DDsaredefinedasthesumofdegrees each daythe outdoor temperature exceeds a reference tempera- ture(balancetemperature).Whilsttraditionallyusedatthebuild- ing andlocalscale,increasingly studies haveusedcooling degree days(CDD)fortheestimationofcoolingloadsforEurope[36–38], USA [35] andworldwide [2,14,15,34] using geographically exten- sivegriddedclimatedata.Studiesondevelopingcountriesarestill limited [5,39,40]. Most of the CDD studies covering large scales use arbitrarily fixed balance temperatures, limiting the ability of analysing the effect of building configuration and fabric on the results. The variable degree day (VDD) methodadvances the DD method,byanalyticallycalculatingthebalancetemperatureasthe outdoor temperatureatwhich neither heating, norcooling is re- quired[41,42].Thisdefinitionbetterrepresentsactualbalancetem- peratures of buildings, which depend on many factors, such as qualityofconstruction,thermalinsulation,internal andsolarheat gains,thermostat settings,and behaviour of occupants.However, VDDs arerarelyused forthe estimationofbuildingenergyneeds [43,44], anduntil now,have neverbeen applied toa globalgrid.

Moredetailedapproaches,i.e.dynamicmodels,wouldmoreaccu- rately account for dynamiceffects on cooling loads,but they re- quirean excessiveamount of input data andcomputational load for global scale applications, thus being typically used for case studiesandcountry-levelanalyses(seee.g.[45–47]).

3. Methodology

The energy demand for space cooling can be described as a function ofthreedrivers [3,14]: thetechnological intensitydriver, thestructuralintensitydriverandtheactivitydriver.Technological parameters includethe efficiencyforeach ofthe n spacecooling devicesconsidered(

η

i).Inthisstudyweconsidertwotechnologies toprovidethermalcomfort:ACunitsandfans(i=AC,fans).Struc- tural parameters include the conditionedfloor surfaceper capita (A_f/P)andbuilding-relatedfeaturesusedtocalculatetheusefulen- ergyintensityperfloorareaforeachcoolingdevice(E_c,u,i/A_f)under

contextualclimaticconditions.Activitydriversincludethepopula- tion(P), applianceavailability(a_i), andall otherdrivers relatedto servicelevel,behaviourandlifestyle.Thefinalenergydemandfor spacecooling(E_C)canthereforebeexpressedas:

EC=P·A_f P ·

n

i

Ec,u,i

A_f · 1

η

^i·ai

(1)

We calculate the space cooling needs using the VDD method andthepenetrationofACownership,appliedtoaspatialgridand usingmethods available inliterature[14].The spacecoolinggaps are then estimatedin termsof bothpopulation in needof space coolingbutwithoutaccesstoit,andtheassociatedenergydemand gap.Aparametric analysisisrunto estimatethe effectofseveral technologicalandbehaviouralparametersontheresults.Detailsof thecalculationsandinputparametersaredescribedinthefollow- ingsections.

3.1. Spacecoolingdemandmodel

The space cooling demand calculation is based on the VDD method[41].ThismethodadvancesthesimpleDDmethodbyan- alyticallycalculating (asdescribedalreadyinSection 2.1),instead ofassuminga fixedvalueforthebalancetemperature.Ourcalcu- lationsaccount fortheuseofbothACandfans,by assumingtwo distinct threshold temperatures: fansswitch on afterthe balance temperature(T_bal) isexceededandtheACaftera maximumtem- perature(Tmax>T_bal)isexceeded.ThebalancetemperatureT_bal,m (°C)iscalculatedforeachmonthusingthefollowingequation:

Tbal,m=Tsp−gsol,m+gint

Htr+H_ve

(2)

where Tsp (°C)isthe desiredindoor setpoint temperature, g_sol,m (W) is the heat flow from solar heat sources for the month m, g_int (W) isthe heatflow frominternal heatsources, Htr (W/K)is theheattransfer coefficientbytransmissionandHve (W/K)isthe heattransfercoefficientbyventilation.Thisequationassumesthat ventilation is not introduced and all windows are closed. While Htr,Hve,andg_int dependsolelyonbuildingcharacteristicsandoc- cupants’ behaviour, g_sol,m also depends on the climate (solar ir- radiation) and therefore is variablein space andtime. Detailsof the calculationsfor heat transfer coefficientsand heat flow from heatsources arereportedintheSupplementaryMaterial,sections SM1.1–1.2.ThemaximumtemperatureT_max,m (°C)iscalculatedfor eachmonthassumingfan-poweredventilationandopenwindows:

Tmax,m=Tsp,max− g_sol,m+g_int Htr+H_ve,max

(3)

Where Tsp,max is the maximum indoor set point temperature whenfansareoperatingandHve,max(W/K)istheheattransferco- efficientbyventilationwhenwindowsareopen.Themonthlyvari- ablecoolingdegreedays(VDDc,m),basedonTmax,arecalculatedas follows:

VDD_c,m(Tmax)= Dm

d=1

_¯

T_out,d−Tmax,m

+

(4)

WhereT¯_out,d istheaverage dailyoutdoortemperature, andDm

isthenumberofdaysinthemonth.The+signindicatesthatonly positivevaluesareaccounted.TheannualfinalenergyE_c,_ACforAC isthencalculatedasfollows:

E_c,AC= 12

m=1

(

^Htr+Hve

)

^·fc·

VDDc,m(^Tmax)+

Tmax,m−T_bal,m

·Nc,m

η

^AC

(5)

WherefcisthedailyoperationtimefractionandNc,mthenum- berofdayspermonthwhencoolingisrequired(Tout,m > Tmax,m) forthem-thmonth,and

η

ACthe efficiencyoftheACsystem. The secondterminsquare bracketsaccountsforACoperationataset pointTsp(andnotTsp,max),eventhough ACisactivatedafterTmax

isexceeded.TheelectricityrequirementsE_fforfansarecalculated usingthefollowingequation:

E_{c,f ans}= f_f·P_f· 12

m=1

N_f,m (6)

Wheref_f is theoperation time fractionforfans, P_f (W)is the powerandN_f,m thenumberofdaysforthem-thmonthwhenfans areused(Tout,m >T_bal,m).

The equations are modelled on a flexible andscalable spatial grid,conformingto the resolutionofthe input data,allowing for analysesfromglobaltolocalscalesdependingondataavailability andapplication. Inthisstudy,VDDandcoolingenergyneedscal- culationsarerunacrosstheentireGlobalSouth.Resultswerecom- paredtodynamicbuildingsimulationforaseriesofselectedloca- tionstocheckforconsistency(seeSupplementaryMaterial,section SM2.5).

We use the observed historical weather datasets, EWEMBI (EartH2Observe, WFDEI and ERA-Interim data Merged and Bias- correctedforISIMIP),withglobalcoverage at0.5°grid resolution (approximately 50kmat the equator)and atdaily time stepbe- tween1979and2013. EWEMBIcombinesobservedglobalclimate datavariablesfromanumberofsources,consistentlydownscaled andbias-correctedforuseinclimateimpactsassessments[48].

Weusedailydata of30years(1980–2009)tocapturethefull variabilityoftherecentclimate.Inthisimplementationdataisag- gregatedtomonthlymeans,whilstmakinguseofthedailytemper- aturedatatocalculatenumberofdayspermonthrequiringcool- ing,Dm (Eq 4). The framework is predominantlyimplemented in Pythonusingxarray[49]andDask[50],toenableparallelizedpro- cessingofmultidimensionaldatasets. Monthlyhorizontalsolarir- radiation(fromEWEMBI)wasprocessedusingtheRpackage“so- laR”[51]tocalculateverticalsolarirradiationfordifferentexposi- tionsonthe0.5°grid.

3.2.Housingcharacterization

Theenergydemandofhousingis influencedby buildingchar- acteristics, such asgeometry and construction materials. A thor- ough characterization of the housing across different world re- gions is beyond the scope of this study. Instead, we rely on a limited set of housing archetypes transversal to different world regions andrepresentingprevailingconstruction practices (“refer- ence case”, see Table 1), similar to other global studies [3]. We then run sensitivities to account forranges of variability forkey technologicalandbehavioural parameters.Ruralhousingisrepre- sentedby asingle-familyhousewithbrickmasonry anda mixof tileand concrete roofing. Forurban housing,we consider an av- erage height of four storeys, concrete structure and roofing, and brick masonry. Main construction features and U-values are set based ona review ofrelevant literature for developingcountries (see Supplementary Material, section SM1.3–1.4). A window sur- face of 1/8thof the floor surfacearea andsingle-glazing are as- sumedforallarchetypes.Whileconsideringonlyarestrictednum- ber of archetypes is a limitationof this study, our methodologi- calframeworkallowsincludingadditionalregionalbuildingtypolo- gies,oncetheybecomeavailable,andestimatingthesensitivityto buildingstockheterogeneity.

Inthereferencecase(Table2,valuesinbold),weassumethat homesare equippedwithceilingfansof55W rating[52],andAC systemswithenergyefficiencyratio(EER)of2.9, closetothecur- rentaveragevaluefordevelopingcountries[1].Fanspenetrationis

Table 1

Housing archetypes (Reference case).

Description Envelope area (m ²/m ²floor area) Roof area (m ²/m ²floor area) Wall material Roof material U-value Envelope (W/m ²K)

Rural 3.70 1.00 Brick masonry Clay tiles - Concrete 2.11

Urban 1.61 0.25 Brick masonry Reinforced concrete slab 2.18

Table 2

Overview of the parametric analysis. Values for the reference case in bold.

Parameters Unit Values Description

Technological parameters

Average U-value of the building envelope W/(m ²K) Rural: 1.08, 2.11 , 4.08 Estimate the effect of different building envelope materials.

Values from literature review (see Supplementary Material, section SM1.4)

Urban: 1.32, 2.18 , 4.23

EER – 2.9 ; 3.5; 4.0 Efficiency of AC units (adapted from [1] and own elaboration):

average for developing countries; high end of typically available in most countries; best available in most countries.

Behavioural parameters

S et point temperature (T sp) °C 20; 22; 24; 26 ; 28; 30; 32 Evaluate the effect of different indoor set point temperatures in the range 20–32 °C [11] .

Hours of AC operation (f c) Hours per day 4; 8 ; 12 Assess the effect of different user schedules for AC.

Hours of fans operation (f f) Hours per day 4; 8 ; 12 Assess the effect of different user schedules for fans.

Use of AC only (no fans) – AC and fans ; AC only. Evaluate the effect of AC in combination with fans. If only AC is used (no fans), AC operates when T out,d> T bal,m.

definedonacountrybasisdependingontheexistingnationallevel ofelectrification(datasource:WorldBank.SeeSupplementaryMa- terial,section SM1.6). A dedicatedmodel is usedto estimate the penetrationof AC(see Section 3.3). We considera wide rangeof indoorsetpointtemperaturesTsp (20°Cto32°C)andwarmdays ofexposurebeforeACadoption toinvestigatetheeffectofdiffer- ent thresholds foravoiding potential heat stress, on the basis of the literature review (Section 2.1). For the reference case, we il- lustrateresultsforan indoorset pointof26°C,beingthecentral valueofthe investigatedtemperaturerange.The allowed number ofwarm days of exposure over the maximum temperature Tmax

beforeadoptingACisalsoakeyparameterforcoolinggaps.Inthe referencecase, we exclude fromthecooling gaps those locations withlessthanfivedaysofexposureoverTmax.Weallow,accord- ing to theASHRAE 55–2013 standard [53], the operation ofceil- ingfans(airvelocityupto0.8m/s),whichenablesanextensionof thecomfortzoneandanincreaseinsetpointtemperatureby2°C (Tsp,max=Tsp+2°C)forachievingthesamecomfortlevels[54].AC isturnedonwhentheoutdoortemperatureexceedsT_maxandop- erates atTsp. Inthe reference case, ACand fan operation sched- ulesaresetto8hperdaybasedontypicalschedulesintropicand sub-tropicareas[11].We refer thereaders tothe Supplementary Material(sectionSM1)foradetaileddescriptionoftheinputdata.

An overview of the parametric analysis to investigatethe in- fluence of key technological and behavioural parameters on per- capitaenergyneedsisreportedinTable2.Wevarybuildingchar- acteristics andsystem efficiencystarting fromthe reference case toconsiderthevariabilityinhousingconstruction,coolingsystems and potential future improvements. We also vary the set point temperature,hoursofoperationofACandfansandcombineduse ofACand fansto account fordifferent servicelevels andprefer- ences connected to thermal comfort standards and health risks.

TheoptionofusingAConly,withoutfans,thoughmoreenergyin- tensive, may be suitable for areas where open windows are not recommended,e.g.duetooutdoorairpollutionormosquitos.

3.3.Air-conditioningpenetrationmodel

DataonthestockofACareavailableforonlyalimitednumber ofcountries.However, itis likelythat ACuseisprevalent among high-incomehouseholdsacrossthe globe,particularlyinsubtrop-

ical and tropical climates. Previous research has shown that AC ownership is driven by income, but that the climatic conditions determine the maximum penetration of ACs, even among high- incomehouseholds[14,18].Thesestudiesrely onanempiricales- timation ofthe maximumAC ownershipasa functionof cooling DDintheUnitedStates,whichspansmanyclimaticconditionsand wherearguablyformostincomeisnotaconstrainttoowningAC.

WeadoptthismodeltoestimateexistinglevelsofACownershipin countriesacrosstheworld,usingthepopulation-weightedaverage cooling DDs andGDP per capita asinputs foreach country. The results match reasonably well with the actual ownership shares forcountrieswhere dataare available, withfew exceptions,such asBrazil(10 percentvsthe predicted65 percentin2009,dueto thehighconcentrationofpopulationalongthecoastwherethecli- mateismilder).SeetheSupplementaryMaterialfordetailsofthe model (section SM1.8) and a comparison between predictedand actualpenetrationratesforselectcountries(sectionSM2.1).

3.4. Gapestimationandaggregation

The energy demand for AC is calculated for every housing archetype and then aggregated to the country scale and global scale using Eq. (1)separately for ruraland urban areas. A refer- encefloorsurfaceof 10m² percapita isassumed forthisanaly- sis,basedonaspacestandardidentified bypreviousstudies [54]. Whilefloorsurfacevalueper-capitamayvarygreatlyacrossdiffer- entworldregions,previous studiesshowthelinearityofdwelling size withoperational energy consumption,therefore allowing for an easy extension of our results to different floor space values.

Weuse ruralandurban populationdata(see Supplementary Ma- terial,section SM1.5) toupscale per-capitaenergydemand ofthe twohousingarchetypestothecountryandglobalscale.

Gapsare expressedboth interms ofspacecooling energyde- mandgapand populationneeding butlacking spacecooling. En- ergy gaps are calculated as the difference between the poten- tial energydemand with universal access to ACand fans, where needed(i.e.whereVDD>0andtheallowed numberofwarmdays ofexposure isexceeded),and thecurrentaccessto cooling tech- nologies,undertheassumptionthatACandfansareoperatedonly to achieve the selected indoor set point temperature. Access to coolingdevicesisestimatedatanationallevelbyapplyingtheAC

Fig. 1. Comparison of global population lacking different services related to hous- ing: space cooling, electricity access and slum conditions. Share on total population (%) reported on top of the bars (share of slum is on urban population). Electricity access and slum population data from World Bank [57] .

penetration modelandconsidering fanspenetrationequivalentto electricity access. Results are presented at thecountry level, and for six of the eleven world regions used in the global energy- economy integrated assessment model MESSAGE [55] that cover theGlobalSouth:sub-SaharanAfrica(AFR),CentrallyplannedAsia and China (CPA), Latin America andthe Caribbean (LAC), Middle East andNorth Africa (MEA), Other Pacific Asia (PAS), andSouth Asia (SAS). Sensitivityanalysis is conductedon two key parame- ters influencingthe magnitudeofthe energygapandpopulation affected: thetemperaturethresholdandset pointforAC,andthe allowed number of days over the temperature threshold before adoptingAC.

4. Results

Ourresultssuggestthatuptofourbillionpeoplearepotentially at risk of exposure to heat stress in the Global South, mostlyin India, South-East Asiaand sub-Saharan Africa.The following sec- tions report ourresults ofthe space cooling gapestimation, and per-capitaenergyneedstofillthegapundervariationofkeytech- nologicalandbehaviouralparameters.

4.1. Residentialcoolinggaps

4.1.1. Populationexposedtopotentialheatstress

Weestimate coolinggapsacrosstheGlobalSouthusingdiffer- entsetpointtemperaturesandreporthereresultsfor26°Cindoor set point with a five-day allowance above the set point thresh- old before ACadoption isrequired (full results of the sensitivity are in Section 4.1.4). Fig. 1 showsthe population potentially af- fected by cooling gaps in different world regions in comparison with gaps in electricity access andextent of slum population as a proxy for lack ofaccessto proper housing (see Supplementary Material,sectionsSM1.6–1.7forelectricityaccessandslumpopula- tiondata).ThecoolinggapisparticularlysevereintheSASregion withalmost1.5billonpeople(92%ofthetotalpopulation)poten- tiallyaffected,duetoa combinationofsevereclimaticconditions andlow accesstoAC(seeSupplementary Material,section SM2.1 for detailed AC accessresults). Other regions, such as AFR, MEA andPAS, alsoexhibit ahighshareofpopulation exposed (70% or more),eventhoughtheabsolutenumbersofpeoplearelower.The cooling gapisfurtherexacerbatedbylarge gapsinelectrification, especiallyfortheAFRregion(seeSupplementaryMaterial,section

Fig. 2. Comparison of the space cooling energy gaps and basic electricity access gaps, assuming for the latter a Tier 2 threshold of household electricity supply (200 Wh/day) [8] .

SM1.6).Inthiscase,theprovisionoffansorACswouldrequirethe preconditionof electrification ofregions currentlylacking access.

Thespace coolinggap isalso relatedtothe lackof properhous- ing,whichcanbeexpectedtoresultinmuchpoorercomfortcon- ditionsandhigher energydemands sincefans andACswouldbe lesseffectivewithpoorconstructionquality.Theaverageshareof populationlivinginurbanslumsrangesbetween10%and20%,de- pendingontheregion.However,thehousinggapisexpectedtobe muchlargerinmanycountriesduetopoorqualityhousinginthe countryside,overcrowdingandhomelesspopulations[56].Closing thecoolinggapinthissettingwouldrequireasignificantimprove- menttohousingconditionsasafirststeptoprovidingpropercool- ingdevices.

4.1.2. Coolingenergygap

Fig.2showsthespacecoolingenergygap,assuminganindoor setpointtemperatureof26°Cfordifferentworldregions.Theen- ergycoolinggapisdominatedby theSAS regionwithatotalen- ergygap of428TWh/y which,to putinto context,is 70percent higher than India’s current total residential electricity consump- tion [58]. The total energy gapis also high in AFR (135 TWh/y) andPAS(89TWh/y)regions.Otherregions,inparticularCPA,have a relatively lower energygap incomparison withthe population exposed(Fig.1)duetothecombinationofhighpopulationdensity butwiderelectricityaccessandmilderclimaticconditions.

Thecomparisonofgapsshowsthattheenergyneededtobridge thespacecoolinggapwouldbemuch higherthan theenergyre- quired forproviding all with the basic services such aslighting, television and radio, assuming a Tier 2 threshold of 200Wh/day percapita[8],andcurrentunelectrifiedpopulation.Thishighlights the importance of considering the space cooling gap alongside otherbasicenergyneeds.

4.1.3. Spatialdistributionofthecoolinggap

Sourcesofspatialvariation inthecooling gapincludepopula- tion distribution, climatic conditions andaccess to cooling tech- nologies. While the diffusion of regional housing typologies and constructionmaterialsareexpectedtoinfluencethespatialdistri- butionofcooling gapsto some degree,inthisstudywe consider homogeneousbuildingcharacteristicsacrosstheGlobalSouthand distinguishrural andurban housing,thereforeonly partially cap- turingtheeffectofbuildingcharacteristics.

The map of population exposed to potential heat stress (Fig. 3) showsthat thelargest gapsare inareas characterized by

Fig. 3. Space cooling gaps: population without access to AC where needed.

concurrenthighpopulationdensity,severeclimaticconditionsand lowACownership(see SupplementaryMaterial,Fig.SM1forspa- tialdistribution of population density, Fig. SM9-10 for VDD, and Fig.SM4forACaccess).LargegapscovermostpartsofIndiaand parts of Pakistan andBangladesh. Other significant gaps exist in South-East Asia, notably in Indonesia, China, sub-Saharan Africa andthe Nile valley. Latin America and Southern African regions exhibit lower gaps dueto both milderclimatic conditions inthe populatedareasandlower populationdensityoverall.Other areas suchastheMiddleEast,areaffectedbythegaptoalowerextent, despitetheir severeclimate,asa consequenceofhigherGDPand ACownership.Whileinmostregionsthecoolinggapisduetolack ofaccesstoAC,forspecificareasofAFRandSAS,includingNige- ria,Uganda, Burkina FasoandBangladesh,the gapiseven bigger duetolackof accesstoelectricity, andconsequently tofans(see SupplementaryMaterial,sectionSM2.4).

Onlythreecountries– India,ChinaandIndonesia– covermore than 50% ofthe population potentially exposed to heat stress in theGlobal South (see Supplementary Material, Table SM6). India dominatesby farthe listoftop countriesaffectedby thecooling gapwith1.1billion peoplepotentially exposed toheatstress and anenergycoolinggapofalmost335TWh/yforanindoorsetpoint of 26°C (see Supplementary Material, Table SM7). China comes secondwitharoundhalfasmanypeopleaffected.However,dueto themilderclimate,theenergyrequirementsareanorderofmagni- tudelowerthanthat ofIndia.Indonesia, Pakistan,Bangladeshand Nigeriarankamongthetopcountriesforbothenergygapandpo- tentialpopulationaffected.

4.1.4. Sensitivitytosetpointtemperaturesandwarmdaysof exposure

The set point temperature and the number of days of expo- sureto temperatures above thisthreshold strongly influence the affected population andenergy demand. Foran indoor set point temperature between 20°C and 32°C (Fig. 4), the total popula- tionpotentially exposedtoheatstressrangesbetween1.8and4.1 billion people (3.7 billion people for 26°C), and the energy gap from100 to 2014 TWh/y (786 TWh/y for 26°C). In comparison, thecurrentelectricity gap (forbasic serviceslike lighting, televi- sion and radio) affects one billion people and requires an addi- tional73 TWh/y, assuming a thresholdof 200Wh/dayper capita [8]. The population affected by the gap rapidly increases when movingfrom32to26°C,duetoaprogressiveexpansionofthere- gionsrequiringcooling.Furtherreducing thesetpointfrom26°C

Fig. 4. The sensitivity of exposed population and energy requirements to indoor temperature set point (T sp).

to20°C entailslessofan increaseintheaffectedpopulation,but energy gap increasing proportionately withthe required thermal comfortlevel.ForIndiaalone,thecooling energygapcouldrange between59–748TWh/ywitha potentially exposed populationof 1.00–1.15billion.

There are a very limited sets of conditions under which the number of warm days of exposure over a given temperature threshold hasa sharpeffect on the estimated atrisk population (Fig. 5). Forlessthan fivedays ofexposure overa threshold,the exposedpopulationincreasessteeplywithincreasinglevelsofthe temperaturethresholdandfewerdaysofexposure.Otherwise,be- yond fivedays, the at risk population is considerably lesssensi- tivetothenumberofdaysofexposureforanygiventemperature threshold. Naturally, with higher threshold temperatures, fewer peopleinneedofcoolingareestimatedtobeatrisk.

4.2. Coolingenergyneedstofillthegap

4.2.1. Residentialcoolingdemandindifferentworldregions

Fig.6illustratestheaverage,population-weightedfinal cooling demandintensityestimatedfordifferentworldregionsanddiffer- ent housing types.Cooling energy demand intensityis higher in SouthAsia(SAS), OtherPacific Asia(PAS),andsub-Saharan Africa (AFR), wheretemperatures are on average warmer.Urbanhomes havelowerdemandintensitythanruralhomes,duetotheirmore

Fig. 5. The sensitivity of exposed population to warm days of exposure and indoor set point temperature (T sp).

compact shapeandloweraverage roofU-values limitingtheheat load.EnergydemandforACdominatesthatforfansinallregions, comprising63–79%(80–88%)ofthetotaldemandforrural(urban), despite themuchlower hours ofuse.The shareofACdemandis higherforregionswithmoresevereclimates(SASandMEA).The spatial distributionof energyintensities (see Supplementary Ma- terial, Fig.SM12-13) is mainly driven by climatic conditions, and to aminorextentby theshareofurban andruralbuildings(due topopulationweighting).Whileweassumedhomogeneouscharac- teristicsofthehousingstockacrossdifferentworldregions,differ- entbuildingtypesmightresultinadditionalenergyintensityvari- ationsonaregionallevel(seeSection4.2.2).

4.2.2. Sensitivitytotechnologicalandbehaviouralparameters Fig.7illustrates theinfluenceofvarious technologicalandbe- havioural parameters on the per capita energy needs for space cooling. Varying the thermal properties of the buildingenvelope (U-value) entails major variations of the energy needs. Building envelopeswithpoorthermalquality(high U-values)entail anin-

Fig. 7. Sensitivity of technological and behavioural parameters to per capita energy needs for space cooling. Indoor set point temperature (T sp) was varied by ±2 °C, assuming 26 °C as a reference.

creaseinenergyneedsover40%,duetoadditionalheatgains.Con- versely,insulatingthebuildingenvelope(lowU-values)hasaben- eficialeffectontheenergyneeds(−28%).Increasingtheefficiency ofAC(referenceEER 3.9)contributesto reduce theenergyneeds from13%(EER4.5)to22%(EER5.0).

Behavioural parameters appear to have a large effect on the energygapvariation.Varyingthereferenceindoorset pointtem- perature(26°C)by2°C higherorlowerentailsadifferenceofre- spectively−33%and+38%.Similarly,doublingorhalvingthedaily numberof hours ofAC usage,has an impact ofalmost ±40% on theresults. Varyingthe numberof dailyhours of fansusage has amuch lowerinfluence ontheestimatedtotal gap(±10%).Using ACwithoutfans toensure thesame setpoint of26°Centailsan increase of 25% inthe energy needs, relative to the conditionof windowspermanentlyclosed,e.g.againstexternalairpollutionor mosquitos.

Combiningallinputvariationssimultaneouslyresultsinarange ofenergyneeds variationfrom −78% to +200%compared to the reference case. While this is a very broad range of variation, it

Fig. 6. Per capita energy demand for space cooling of urban and rural housing in different world regions (indoor set point 26 °C).

is in line withprevious studies on developing countries [54,59]. Thelargesensitivityofenergydemandtouser-relatedparameters highlightstheimportanceofcarefullyconsideringthelifestyleand behaviouraldimension in spacecooling assessments. The hetero- geneityofbuildingenvelopecharacteristicsacrosstheGlobalSouth isalso an importantdriver of energy needs variation. Future re- searchshouldfocusonidentifyingregionalbuildingtypologiesand recurrentmaterialstoimprovetheaccuracyofthemodelonare- gionallevel,asdiscussedfurtherinSection5.4.

5. Discussion

5.1.Vulnerabilityduetolackofaccesstocooling

This studyconfirmsthe importance ofinadequate space cool- ing,includingthelackofAC,asadimensionofenergypovertybe- yond that currently within the ambit of SDG7. We estimate that between1.8and4.1billionpeopleintheGlobalSouth,withame- dianof3.7billionfor26°Csetpointthresholdandatleast5days ofannualexposure,arepotentiallyexposed toheatstressintheir homes.This population includes almost 1 billion without access toelectricity, thereby highlighting thataccounting forthelack of coolingsignificantly increases the energypoverty gap definedby SDG7.Additionally,thegapincludesthe populationwithelectric- ityaccessbutlackingACs, mostlikelydueto their beingtooex- pensiveorunsuited to people’s housing conditions.These results areconsistentwitha recentreportby theUN SustainableEnergy for All (SE4All), which estimates that 3.4 billion people face is- sueswithadequatecoolingaccess[9].TheUN reportisbased on acruderassessmentofpotentialheatstressbutexaminesawider set ofchannels that could lead to such stress, such asthe need forcoldstoragefor foodandvaccines.In contrast,thisstudyhas twoprimarynovelties:first,thespacecoolingaccessgapsarees- timated based on a detailed, spatially explicit assessment of cli- mate,housing conditionsandAC ownership, fora wide range of target temperatures and days of exposures that could result in heatstress;second,itprovidesafirstestimationoftheenergyde- mandassociated withmeeting thishousehold spacecooling gap.

Previousenergydemandstudies haveestimatedACusebasedon standardCDDcalculationsusingmuchlower(typically18.3°C)set pointtemperatures.Theseestimatestherefore,go beyondtheen- ergyneededto avoidheat stressandextendtothat forproviding comfort.

Notably,ourestimateof1.8–4.1billionrepresentspeoplepoten- tiallyatriskbasedontheiravoidingheatstressrelativetoarange ofassumedsetpointthresholds forindoorcomfortandexposure daysperyear.Inreality,thespecificexposureconditions,acclima- tizationandacceptablecomfortthresholds,mightvaryacrossdif- ferent subpopulations, so that different set point thresholds and standardsmaybe applicableto each.Furthermore,farfewerthan thoseestimatedto be potentiallyat risk arelikely tobe exposed to conditions that pose serious health risks (very high tempera- turesoverconsecutive,evenifforonlyafew,days).Addedtothis fact,sincewedonotfilterthepopulation byincomelevel,ageor physiologicalconditions,thevulnerable amongthepotentially ex- posedarealsolikelytobefewer.Ourresults,thusincludealarge populationwhomightfaceonlyfunctioningimpairments.

An importantarea forfutureworkistounderstandhowmany ofthe1.8–4.1 billion atrisk would adoptACsandmitigate these risksfromincomegrowthalone.TheInternationalEnergyAgency (IEA)estimatesthat by 2050, 2billion residential ACswill be in- stalledinIndiaandChinaalone[1].Butitisnotclearhowmany ofthese2billionunitswillbeinstalledinhomesthataidinavoid- ingheatstress,andconsequentlyhowmanyoftheestimatedpo- tentiallyexposed,andinwhatlocations,wouldstillremainatrisk.

5.2. Energyrequirements

We estimate the median energy demand required to fill the cooling access gap to be 786 TWh per year, around 14% of cur- rent globalresidential electricity consumption [60], for the 26°C setpointandatleast5daysofannualexposure.Seventy-fiveper- centof thisdemand isfromIndia, Africa andEastAsia.This can be compared tothe IEA estimatesof2800 TWhgrowthbetween nowand2050ofresidentialcoolingdemandfromincomegrowth.

Ourenergyestimateisconservative,sinceweassumethatACfunc- tions onlyto achieve ourmoderaterisk alleviationthresholds. As withthepopulationatrisk, itisnotpossible toknowhowmuch of this projected energy growth would serve to fill the ‘cooling energy poverty gap’. This is particularly a challenge in emerging economieswithhotclimateandhighlevelsofbothincomegrowth andpoverty,suchasIndia,China,Indonesia,andNigeria.Theutil- ityofthisstudyistoprovidealowerboundofcoolingenergyde- mandforbasicwell-being,anda startingpoint toidentifyitsge- ographicdistribution.Inadditiontothespatialvariationofat-risk population,thepercapitaenergydemandtomeetthecoolingen- ergypovertygapvariessignificantly,duetoclimateconditionsand income,forthesamesetpoint.Thepercapitaenergyrequirement ishighestinruralpartsofSouthandEastAsia,andlowestinChina andLatinAmerica.

5.3. Policyoptionstowardsreducingthecoolinggap

Filling the cooling energy poverty gap requires integrated strategies,beyond simply providingaccess to affordableandeffi- cient AC. Populationsin many regions still lack electricity access anddecent housing, requiringa profound rethinkin energysup- ply,buildingdesign,andurban planningasa wholetoeffectively addressadaptationchallengestoheatstress. Interconnectionsand potential synergies betweenfillingthe cooling gapsand reaching other SDGsarealso ofutmostimportance. Inaddition totheob- viousimplicationsforSDG7,theinteractions withSDG1(poverty), SDG3 (health),SDG9 (infrastructure), SDG10 (inequalities),SDG11 (cities)and SDG13 (climate)are ofparticular significance,asdis- cussedfurtherbelow.

Accesstoelectricityisstilllimitedinmanyregionswherespace coolingisrequired,inparticularsub-SaharanAfricaandSouthAsia, as shown by our results. Achieving universal access to electric- ity(SDG7) is a prerequisite to access space cooling technologies.

Off-gridsolarsolutions,suchassolarhomesystemsorsolarmini- grids, may contribute to completing electrification in rural areas [4].Morerenewable-basedACoptionsmightalsoneedtobecon- sideredtomeetcoolinggaps.

With more than one billion people currently living in slum conditions[61], providingaccessto adequate, safeandaffordable houses is a challenge (SDG11). Poor-quality informal houses not onlyraiseissuesofsafetyanddecentliving,butalsoprovideinad- equate protection against climate-related extremeevents (SDG1), increasing the risk of temperature-related mortality(SDG3) [62]. Provisionofnewhomesandslum-upgrading(SDG9)aretherefore essentialtofillthecoolinggapsinmanycountries.Low-energyand sustainable housing design further contributes to reduced space coolingrequirementsanddoesnotnecessarilyentailhigherinvest- mentcosts[46].

Promoting energy-efficient and low-emitting AC systems has beenidentifiedasaneededmitigationmeasuretoreduce theen- ergyrequirementsassociatedwithclosingthecoolinggap[19,63]. Potentialsolutions include,usingbetter coolantsandrefrigerants, more efficient cooling units,centralized systems formulti-family buildingsand,whereviable,districtcoolingnetworks(e.g.indense urban areas). However, cooling demandis alsosubject tothe re- boundeffect[64].Ourresultsshowedthatthebehaviourofoccu-

pants havea large influence and should be addressedby policy, e.g. by limitingtheset pointtemperatures forAC,asalreadyim- plementedinJapanforcommercialbuildings[65].

ThehighdemandforACintheGlobalSouthposesbothachal- lenge and opportunity for mitigating climate change. Efforts to switch out ofusing climatewarming refrigerantsas requiredun- dertheKigaliamendmentwillbeimportanttolimitingtheclimate impactsofAC(SDG13).Inaddition,timelypoliciestomakehigher- efficiencyACs affordable,andtoimprovethe designofcities and buildingstoreduceheat-islandeffectscould beawin-winforcli- mateanddevelopment.

5.4. Limitationsandfurtherresearch

Inestablishingthisglobal,scalablemodellingframework,there are several caveats to our conclusions anda numberof possible further refinements are possible. First, we characterize the hous- ing stock witha limited set ofbuilding archetypes that describe common building configurations in the Global South, and run a sensitivity analysis to account for key variationsin building pa- rameters.Yet, thesearchetypescannotcomprehensively represent thefull varietyofmaterials andconstruction techniquesavailable acrossworldregions,rathertheyindicatedifferentrecurrenthous- ingtypes.Withcontinuedcollectionofdatafromnationalstudies, futureeffortscanextendthisworktoincorporateawidervariety ofmaterialsandhousingcharacteristicswithinthesamemodelling framework. Whilethespatialdistributionofcoolinggapsanden- ergy requirements is largely driven by climaticconditions, popu- lationandACaccess, accountingforgreatervariabilityinbuilding characteristicsmaycontributetomorepreciseresultsonaregional level.

Whilstpreviousstudies atthisscalehaveusedDDs toanalyse trendsandcomparealternatives[41],ouruseoftheVDDmethod explicitly considers building characteristics, user behaviour, and temperatureset points.However, thismethodhas limitations,for example:coolingloadsestimationcanbeimpactedbythelimited capacity to account forthe dynamics ofthe building,in particu- lar due to thermal inertia; average daily temperatures are used, thus not fully accounting for diurnal fluctuations; latent loads, which become more significant at higher outdoor temperatures [66], are ignored. The characteristics of user behaviour we con- sider inthisworkare onlythosewithrespecttotemperatureset points and reductions in use of fans and AC from varying user schedules.

The AC adoption model is derived from data on AC owner- ship data in just one country, the United States. Future model development ought to integrate a broader baseof data on own- ership, income and climate. Humidity is also likely to influence AC adoption, but has not been assessed in previous models of AC penetration.Estimations ofboth fansandACadoption in this workaredoneatthenationallevelandneglectwithinnationdif- ferences in affordability and penetration across rural and urban areas.

Further research is expectedto overcome partof theseissues by using intra-day temperaturedata, allowing for more accurate scheduling, provide a better description of the building stock in thedifferentregionsbyusingmicro-surveydata,andincludingla- tent loads in the calculation. Future works could also focus fur- ther on behavioural and lifestyle aspects which, we show, have a majorinfluence onenergyrequirements,andrelatedpolicyop- tions to avoid overuse of AC.The current studyaimed at giving a snapshot of the gaps in the current situation. Rapid changes in the housing stock, socio-economics and access to cooling de- vices are undergoing in many developing countries, and require theinclusionofatemporaldimensiontothecoolinggapsanalysis.

The developmentof such future scenarios, including the dynam- icsofsocio-economicsandclimatechange,isenvisagedforfuture research.

6. Conclusions

Thisstudyestimatedspatiallyexplicitresidentialcoolingneeds inthe Global South in combinationwith accessto space cooling technologiestohighlightthelocationofpopulationspotentiallyat riskofheatstressandtoquantifythecoolingenergygap.Weap- pliedamethodologybasedonVDDtoestimate spacecoolingata highspatialresolution,accountingforclimate,housingtypes,and spaceconditioningtechnologies.

Ourresultsshow that a totalof 1.8–4.1 billionpeople are po- tentiallyexposed to heat stress dueto lack of access to cooling, mostly located in India, South-East Asia and sub-Saharan Africa.

Theseresultssuggest much largerenergypoverty gapscompared tothecurrentdefinitioninSDG7,whenconsideringlackofaccess toessential spacecooling. Coveringthisgapcould leadto asub- stantialmedianincreaseinenergyrequirementsof786TWh/y,14%

ofcurrentglobalresidentialelectricityconsumption[60],primarily forrunningACs.SolutionsbeyondimprovedACefficiencyandfan use,such aspassivebuildingandcity designandinnovative cool- ingtechnologieswillbe neededtoensureessential coolingforall that minimize environmental damage. The large influence ofbe- haviouralaspects onenergyrequirements,suggeststhat parsimo- nious use ofAC andmoderate set point temperatures should be promoted.

Thisstudyhascontributedtodevelopingamorecomprehensive measureofenergy access,by introducing thedimension ofspace cooling,which has beenlargely overlooked so far, buthas broad implicationforhumanhealthandfunctioning.Meetingtheessen- tial cooling gap, as estimated by this study, can have important interactionswithachievingseveraloftheSDGs.

Acknowledgments

Thiswork ismadepossible by theEuropean ResearchCouncil StartingGrant[ERC-StG-2014,No.637462],fortheprojectentitled:

‘DecentLivingEnergy:energyandemissionsthresholdsforprovid- ingdecentlivingstandardstoall’.

Supplementarymaterial

Supplementary material associated with this article can be found,intheonlineversion,atdoi:10.1016/j.enbuild.2019.01.015.

References

[1] International Energy Agency (IEA), The Future of Cooling, 2018. https://www.

iea.org/.

[2] M. Santamouris, Cooling the buildings – past, present and future, Energy Build.

128 (2016) 617–638, doi: 10.1016/j.enbuild.2016.07.034 .

[3] D. Ürge-Vorsatz, L.F. Cabeza, S. Serrano, C. Barreneche, K. Petrichenko, Heating and cooling energy trends and drivers in buildings, Renew. Sustain. Energy Rev.

41 (2015) 85–98, doi: 10.1016/j.rser.2014.08.039 .

[4] World Bank, Tracking SDG7: the Energy Progress Report, 2018. http:

//trackingsdg7.esmap.org/data/files/download-documents/tracking _ sdg7-the _ energy _ progress _ report _ full _ report.pdf .

[5] E. Akpinar-Ferrand, A. Singh, Modeling increased demand of energy for air conditioners and consequent CO2emissions to minimize health risks due to climate change in India, Environ. Sci. Policy 13 (2010) 702–712, doi: 10.1016/j.

envsci.2010.09.009 .

[6] K. Knowlton, S.P. Kulkarni, G.S. Azhar, D. Mavalankar, A. Jaiswal, M. Connolly, A . Nori-Sarma, A . Rajiva, P. Dutta, B. Deol, L. Sanchez, R. Khosla, P.J. Webster, V.E. Toma, P. Sheffield, J.J. Hess, Development and implementation of South Asia’s first heat-health action plan in Ahmedabad (Gujarat, India), Int. J. Envi- ron. Res. Public Health 11 (2014) 3473–3492, doi: 10.3390/ijerph110403473 .