UrbanClimate23(2018)260–286 ContentslistsavailableatScienceDirect Urban Climate journalhomepage:http://www.elsevier.com/locate/uclim Evaluating the impacts of greening scenarios on thermal comfort and energy and water consumptions for adapting Paris city to climate change C.deMuncka,*,A.Lemonsua,V.Massona,J.LeBrasa,M.Bonhommeb aCentreNationaldeRecherchesMétéorologiques,GAME,Météo-France/CNRS,Toulouse,France bLaboratoiredeRechercheenArchitecture,Toulouse,France A R T I C L E I N F O A B S T R A C T Articlehistory: Recentclimateprojectionspredictanamplificationofglobalwarmingandmorefrequent Received31May2016 extremeeventssuchasheatwaves.Therefore,theadaptationofcitiestocounterbalance Receivedinrevisedform29November2016 theseadversechangesisurgent.Amongavailableadaptationstrategies,urbangreeningisa Accepted5January2017 measurethatisfrequentlyencouragedtoimprovethermalcomfortorenergydemand,but Availableonlinexxxx whoseimpactsarenotwellknownatthescaleofcities.Inthisstudyweevaluatetheeffects ofvariousurbangreeningscenariosbasedonurbanclimatesimulationsacrosstheParisarea. Keywords: The modelling relies on the Town Energy Balance model. The scenarios tested consist of Climatechangeadaptation anincreaseinground-basedvegetationoranimplementationofgreenroofsoncompatible Greeningstrategies buildings,orthetwocombined.Resultsshowthatincreasingthegroundcoverhasastronger Thermalcomfort coolingimpactthanimplementinggreenroofsonstreettemperatures,andevenmoreso Energyconsumption Waterresources whenthegreeningrateandtheproportionoftreesareimportant.Greenroofsarehowever themosteffectivewaytoreduceenergyconsumption,notonlyinsummerbutalsoonan annualbasis.Theeffectsthevariousgreeningmeasuresmayhaveoverdifferentseasonsis finallyaddressedinordertodrawupacomprehensiveinventoryoftheclimaticimpactsof suchstrategies. ©2017ElsevierB.V.Allrightsreserved. 1. Introduction Currentclimateprojectionsforecastanamplificationoftheglobalwarming(MeehlandTebaldi,2004;Stockeretal.,2013) andanincreaseinextremeevents(Giorgi,2006).Thoseglobaltrendsthattranslateatregionalandlocalscales,arelikelytohave significantimpactsincities,thatarehometothemajorityofthepopulation,namelyachangeintheintensityoftheurbanheat island(highlydependentonthesoilwaterresources,suchashighlightedbytheworkofLemonsuetal.,2013),increasedhealth risks(asattestedbythe15,000excessdeathsduetoextremeheatobservedinAugust2003,HémonandJougla,2004),notto mentiontheincreaseintheuseofairconditioningtomeetthecomfortneedsoftheinhabitants(180%between2003and2020 accordingtoAdnot,2003a,b).Theseperspectivesmakecitiesterritorieswherenotonlymitigationissueswillbethestrongest, butalsopotentiallyterritoriesmorevulnerabletoclimatechangethannaturalenvironments. *Correspondingauthor. E-mail address:[email protected](C.Munck). http://dx.doi.org/10.1016/j.uclim.2017.01.003 2212-0955/©2017ElsevierB.V.Allrightsreserved. C.Muncketal. /UrbanClimate23(2018)260–286 261 Therearedifferentformsofadaptationstrategiestoreducethisurbanvulnerability,involvingactionsonrequalificationof surfaces,urbanforms,innerandouterbuildingenvelopes,oronindividualbehaviors.Urbangreeningstrategies,morespecifi- cally,havemultipleinterests.Thankstovegetationcoolingpotential,theyarebothadaptationmeasurestoclimatechangeand mitigationmeasuresforurbanheatislandandair-conditioningdemand.Beyondclimaticconsiderations,urbangreeningalso deliversadditionalbenefitstotheurbanecosystem,suchasabettermanagementofstormwater(Ashleyetal.,2011;Stovin, 2010;Stovinetal.,2008),areductionofairpollution(CurrieandBass,2008;Hill,1971;Nowaketal.,2006;Pughetal.,2012)or anincreaseinurbanbiodiversity(GetterandRowe,2006;Madreetal.,2014;Mullaneyetal.,2015)forexample. This is by modifying its radiative, thermal, hydrological and aerodynamic properties that vegetation affects the urban environment.Thesemodificationsresultfromthreephysicalprocesses:directevaporationofthewaterheldinthesoiland interceptedbythefoliageofplants(daytimeandnighttime),thetranspirationofalltypesofplants(generallydaytime,butcan alsooccuratnighttimeforsomeplantspeciesorundercertainclimates),andfinallytheinterceptionofsolarradiationfortree species(daytime).Butthereisawiderangeofgreeningsolutionsthathavedifferenteffectsdependingonthescaleatwhich theyareimplemented(building,street,neighbourhood,city),ontheplantspeciesselected,onthelocalclimaticconditions,but alsoontheaspectsstudied(thermalcomfort,energetics,dynamics,hydrology,etc.).Thisisshownbythevastliteratureonthe subject,whetherfromexperimentalornumericalstudies. Greeningthebuildingenvelopescanreducesurfacetemperaturescomparedtoartificialsurfaces,whichlimitstheamount ofheatthatcanpenetrateintobuildings(EumorfopoulouandKontoleon,2009;KontoleonandEumorfopoulou,2010;Onmura etal.,2001;Sternbergetal.,2011;TakebayashiandMoriyama,2007;Wongetal.,2010).Indoingso,theydiminishtheseasonal temperaturevariationsinsidebuildings(Castletonetal.,2010),generatingenergysavings,butwithouttotallyachievingthe efficiencyofconventionalinsulationaccordingtoEumorfopoulouandAravantinos(1998)andEumorfopoulouandKontoleon (2009).Urbangreeningtakesalsootherformsthroughstreettrees,urbanparksorforests.Onthebasisofameta-analysisofthe literatureonthesetypesofstrategies,Bowleretal.(2010)showthaturbanparksgenerateanaveragedaycoolingintherange of0.94◦C,withvariationswhichareexplainedinparticularbythesizeofthepark(Barradas,1991;Changetal.,2007)andits composition(tree/lawnratio,Potchteretal.,2006)butalsothesurroundingclimate.Theyalsohighlightthatthetemperatures under the trees are systematically cooler than the surroundings during the day, with variations depending on tree shading capacityShashua-BarandHoffman(2000).Atnighttime,however,treecanopycouldretainheat(Liangmeietal.,2008;Souch andSouch,1993;Tahaetal.,1991).Besides,astudybySouchandSouch(1993)suggeststhatthethermalperformanceoftrees canvarydependingonthetypeofsurfaceonwhichtheyareplanted(grassorconcrete,forexample).Theevapotranspiration potentialconditionedtowaterstatusofplantsandsoils,andirrigationpractices,isalsoacrucialpointwhichhasreceivedvery littleattentiontodate,exceptbyShashua-Baretal.(2009)whoproposeanefficiencycoefficientwhichlinkscoolingeffectand evaporation. Foroperationalpurposeswithintheframeworksofurbanpolicies,theanalysisandevaluationofsuchstrategiesneedtobe examinedatalargescale.Sofar,onlyafewnumericalstudiesinvestigatethecoolingpotentialoflarge-scalestrategiesofurban planning,suchastheimplementationofgreenandbluecorridorsforParisurbanarea(GroupeDescartes,2009),orofgreen roofsoverToronto(Bassetal.,2003)andNewYorkCity(Rosenzweigetal.,2009). Ourstudyaimstocomplementthescientificliterature,butalsobettermeettheexpectationsofurbanplannersanddeci- sionmakers,throughaphysically-basedanddetailedmodellingofurbanclimate,usingParisurbanareaasastudycase.The originalityoftheworklies(1)inthedevelopmentofrealisticurbangreeningscenarios,implementedlocallybutassessedat cityscale;(2)inthecomparisonofdifferentvegetationinstallationsinurbanenvironment,onroofs(withorwithoutirrigation) andontheground(lawns,mixedvegetation);(3)inthemulti-criteriaevaluationofscenarioswhichcoverstheissuesofheat exposureandthermalcomfort,energyconsumption,andwaterdemand;and(4)inthetimescalesthatareinvestigated(fora specificheatwave,aswellasatseasonalandannualscales). Thenextsection(Section2)presentsthebasisfortheconstructionofthedifferentscenariosthatwesimulated,thenumerical configurationusedaswellasitsevaluationbeforegreening.Thenvariousimpactsareassessed,bothinextremesummercon- ditions(heatwave,Section3)andfortherestoftheyear(Section4)toevaluatewhetherthesegreeningmeasuresdevisedto meetheatwaveissueshaveadverseeffectsontherestoftheyearornot.Theimpactsstudiedarethermalcomfortandenergy consumptionforuseofair-conditioning(AC)andheating,whicharegoodindicatorsofthevulnerabilityofcitiestoheatwaves, butarealsoaboutthemanagementofurbanwater(volumesrequiredforwateringthevegetationandgeneratedbysurface runoffthroughouttheyear).Thelastsection(Section5)drawstheconclusionsofthisworkandsuggestswaystodevelopthis approach. 2. Methodologyforevaluatinggreeningscenariosatcity-scale Theinitialaimofthisstudyistoevaluate,basedonmodelling,greeningscenarioswhichmayrepresentstrategiestoimprove futuresummerconditionsincities.AstheEuropeanheatwaveof2003isidentifiedbythescientificcommunitytoberep- resentativeinmeantemperaturesofsummersofthesecondhalfofthe21st century(Déqué,personalcommunication),the meteorologicalconditionsassociatedtothisheatwavearetakenasastudycasetoevaluategreeningscenarios.Thismethod- ology,focusingonPariscity,reliesonsimulatingtheurbanclimateofPariswiththeTownEnergyBalance(TEB)urbancanopy model(HamdiandMasson,2008;Masson,2000).ThefirststepconsistsindesigningforPariscitygreeningscenariosthatare atthesametimerealisticandpromisingandthatcanalsobetakenintoaccountbyourmodel.Inasecondstep,theurban climateofParisissimulatedatthespatialresolutionof1kmwithTEBinofflinemode(withmeteorologicalforcings)foreach 262 C.Muncketal. /UrbanClimate23(2018)260–286 greeningscenario,alongsidesustainablestrategiesforenergyandwaterusage.Thisisdoneundera10-yearmeteorological forcing(1999–2008)thatincludestheheatwaveeventof2003(8–13August2003).Thiswaybothcanbestudiedthebenefits ofgreeningscenariosinaheatwavecontextaswellastheirrespectiveimpactsatthescaleofotherseasons.Thismethodology allowsforacomprehensiveassessmentofgreeningstrategiestoberealised.Duringthesimulations,indicatorsidentifiedto berepresentativeofcityvulnerabilitytoclimatechangearecomputedsothattheimpactsoftherespectivescenarioscanbe quantifiedandcompared. 2.1. Simulationconfigurationandevaluationofthereferencesimulation 2.1.1. Simulationconfiguration Forthisstudy,thesimulationswereperformedwiththeSURFEXsurfacescheme(Massonetal.,2013),andmorespecifically withTEB,theTownEnergyBalancemodel(HamdiandMasson,2008;Masson,2000)builtinSURFEXforurbanandartificial surfaces,andISBA,theInteractionbetweenSoil-Biosphere-Atmospheremodel(Boone,2000;NoilhanandMahfouf,1996)for naturalsurfaces(soilandvegetation).TheconfigurationchosenforthesimulationfollowsthatsetupintheMUSCADEproject (Massonetal.,2014b),whichthisstudyispartof.ThisconfigurationisdetailedinAppendixA.Inthisconfigurationthatinvolves aplatformofmodels,theinputdataforSURFEXareprovidedbyanarchitecturalmodel,theGENeratorofInteractiveUrban blockS(GENIUS,Bonhomme,2013;Bonhommeetal.,2013).ThismodelprovidesSURFEXwithlandcoverinformationatthe resolutionof250m(includingexistingurbangreenspaces),andmorespecificallyprovidesTEBwithamapdescribingurban morphologybasedoneighttypesofurbanblocks(Continuouspavilion,Discontinuouspavilion,Continuousblock,Discontinuous block,High-risetower,Ancientcentre,Industrialbuilding)withspecificgeometricfeaturesandbuildingdates.Then,TEBtrans- latestheseeighturbanblocksintofivebuildingtypes(Haussmannianbuilding,Individualhousing,Densecollectivehousing, Officetower,Warehouse)andaggregatesthemattheresolutionof1kmtoreducecomputationalcosts.Fivetypesofbuildinguse (Residential,Office,Industrial,Agricultural,Commercial)completethisrepresentation.Theradiativeandthermalcharacteristics ofbuildingsareprescribedonthebasisofbuildingtypeandage(characteristicsprovidedinAppendixA),buildingequipment characteristics(air-conditioning,heating,ventilation)onthebasisofbuildinguse(seeAppendixA)andscenariohypothesis.For example,inoursimulations,virtuoustemperaturesetpointsforair-conditioningandheatingarechosentobeinlinewiththe sustainabilityofthegreeningscenarios:19◦Cforheatingand26◦Cforair-conditioning.Heatingandair-conditioningareactive all-yearroundandtriggeredassoonastheirrespectivetemperaturesetpointsarereachedwhatevertheseason. InthisconfigurationTEBisruninofflinemode(i.e.with1Dmeteorologicalforcingsasopposedtocoupledtoanatmospheric model),atthespatialresolutionof1km,andwithmostofitsparameterizationsactivatedtoallowformorephysicalprocesses andinteractionstobeaccountedfor:verticalturbulentdiffusionwithinmultilayerurbancanyons(HamdiandMasson,2008; MassonandSeity,2009),buildingenergetics(Buenoetal.,2012;Pigeonetal.,2014),urbanvegetationontheground(Lemonsu etal.,2012),greenroofs(deMuncketal.,2013a),aswellasamoduleforcomputingthermalcomfortindexes(Fialaetal.,2012; Pigeon, 2011). In addition, the urban weather generator (UWG) implemented in SURFEX is applied in order to derive a 2D temperatureforcingfroma1Dforcingandcityfeatures(LeBrasandMasson,2015).Despiteeffortstoimprovethemodel,some physicalprocessesarenotyetrepresentedintheversionusedinthisstudy,suchastheshadingeffectoftrees,aswellasthe radiativetrappingbytreecanopies. Underthisconfiguration,themodificationstothesurfaceinducedbythegroundgreeningscenariostranslatewithinTEB intothemodificationoftherespectivefractionsofroadsandgreenspaces,wherevergreeningispossibleandaccordingtothe threegreeningratiostested.Inthecaseofscenarioswithroofgreening,thethermalandradiativecharacteristicsofthegreened artificialroofs(thoseofwarehousesandindividualanddensecollectivebuildings)arereplacedbythoseofgreenroofs.Thetotal densityofurbanvegetationinducedbythedifferentscenariosispresentedinFig.1.Additionalsurfacecharacteristics(building fractionandtype,usageandheightofbuildings)whichmaybehelpfulfortheanalysisofsimulationresultsareprovidedin AppendixA. 2.1.2. Evaluationofthereferencesimulation Thereferencesimulation,whichtakesintoaccounttheexistingurbanvegetationandisthereforetheclosesttorealityin termsoflandcover,isevaluatedtoappreciatetheperformanceofthemodelconfigurationsetupinthisstudy,bothduring theheatwave(8–13August2003)andthedecade(1999–2008)studied.Thisevaluationexerciseisbasedonthe2-mairtem- peraturesrecordedateightoperationalweatherstationsthatarelocatedwithinthespatialdomainstudied(seeAppendixB). Theyprovidecompleteseriesofdailyminimumandmaximumtemperaturesoverthe10-yearperiod1999–2008,andhourly temperaturesduringthesixdaysofthe2003heatwavearealsoavailableforsevenofthem.Ata1-kmhorizontalresolution, themodelcannotaccuratelyreproducethesub-gridheterogeneityinthesurfacecharacteristics.Thereforeitismoresuitable tocomparemodelestimatesandobservationsforgroupsofstationshavingsimilarsurfacecharacteristicsratherthantocon- ductameshbymeshcomparison.Consequently,theevaluationexerciseisperformedfortwotypesofstations,thoseurban andthoserural(seeAppendixB),dependingontheurbanisedfractionofthemodelmeshtheybelongto,onaverage81±16% and21±16%respectively.Whilesimulatedandobservedhourlytemperaturesarecomparedforthe2003heatwave,minimum andmaximumtemperaturesarecomparedintermsofthemonthlyclimatologyoverthe10-yearperiodstudied.AppendixB providesgraphsshowingtheevolutionofsimulatedandobserved2-mtemperaturesduringtheheatwaveaswellasthatof monthlymeansofsimulatedandobserved2-mminimumandmaximumtemperatures(referredtothereafterasTNmeanand TXmean).TheassociatedstatisticalscoresaresummarizedinTable1. C.Muncketal. /UrbanClimate23(2018)260–286 263 Fig.1. Urbanvegetationdensityineach1-km2gridmesh.RepresentsgroundvegetationforLV25,LHV25,LV50,LHV50,LV75,andLHV75scenarios,androoftop vegetationforGRandIGRscenarios(greysquaresshowthelocationofbuildingswithoutgreenroof,i.e.Haussmannianbuildingsandofficebuildings). Duringtheheatwave,althoughthemodelcorrectlysimulatesthedailydynamicsoftemperatures(seeAppendixB),thebias valuesobtained(+1.33◦Cforurbanstationsand+1.14◦Cforruralones)suggestthatthemodeltendstosimulatetemperatures alittlehigherthanthoseobserved,whateverthelocationofthestation.Withstandarddeviationsforsimulatedtemperatures veryclosetothoseobservedatstations,5.22◦Cversus5.06◦Crespectivelyaturbanstationsand6.03◦Cversus5.51◦Catrural stations,themodeliscapableofcapturingtheslightlyhighertemperaturevariabilityobservedatruralstations.Runningat the spatial resolution of 1km, an average error on the six days of the heat wave of about 1.9◦C is made on the simulated temperatures,whetherinurbanorruralareas.Thiserrorseemstobeduetoanoverestimationbythemodelofthetemperatures 264 C.Muncketal. /UrbanClimate23(2018)260–286 Table1 Rootmeansquareerrors(RMSE)andmeanbiaserrors(MBEModel-Obs)associatedwith2-mairtemperatureforthe referencesimulationduringthe2003heatwaveandforthemeanclimatologyofthe1999–2008decade. 8–13August2003 1999–2008decade T(hourly) TNmean TXmean Urban Rural Urban Rural Urban Rural RMSE(◦C) 1.93 1.90 1.26 0.77 0.73 0.86 MBE(◦C) +1.33 +1.14 +1.25 +0.76 +0.53 +0.71 duringthefirstthreedaysoftheheatwave(seeAppendixB),whichisprobablytheconsequenceofaninsufficientrepresentation ofadvectionprocessesinthesimulationconfigurationused(discussedfurtherinAppendixB). Onaverageoverthedecadesimulated,RMSEandMBEvaluesobtainedshowthatatthespatialresolutionof1km,beitat nightorduringtheday,themodelconfigurationweusetendtooverestimate2-mairtemperatures.Atnight,themodelpresenta RMSEof1.26◦CandaMBEof+1.25◦CforurbanstationsandaRMSEof0.77◦CandaMBEof+0.76◦Cforruralstations.Whatever theseason,errorsontheTNmeanarelowerforruralthanforurbanstations(onaverageby0.5◦Cforbothscores),suggesting thatthemodelfurtheroverestimatestemperaturesatnightinurbansettings.Thistendencyofthemodeltooverestimate2-m temperaturesismuchlesspronouncedatdaytimethannighttime,withRMSEandMBEvaluestwotimeslessthoseatnight (RMSEof0.73◦CandMBEof+0.53◦C).Contrarytonighttimeperiod,themodelpresentsaslighttendencytooverestimate moremaximumtemperaturesforruralthanforurbanstationsduringtheday.Asfarastemperaturevariabilityisconcerned, thecalculationofthestandarddeviationsforthemonthlyclimatologyshowedthattheydonotdiffermuchbetweenurbanand ruralstations:fortheTNmean,theyareonaverage4.77◦Cforobservationsand4.63◦Cformodelestimations,andfortheTX meanrespectively6.79◦Cand6.48◦C.Suchstandarddeviationvaluesthatareequivalenttoslightlylessthanforobservations demonstratethatthemodelisquitegoodatcapturingthevariabilityfoundintheobservationsforthedecadestudied. OurresultsareinlinewithpreviousnumericalstudiesrunningwiththeTEBmodelthatarebasedonmoresophisticated simulationconfigurationsthantheoneusedinourstudy.Whensimulatingthesamesix-dayperiodofthe2003heatwavewith theTEBmodelcoupledtoamesoscaleatmosphericmodel,deMuncketal.(2013b)obtainatthespatialresolutionof1.25kma similarMBE(+1.0◦C)andahigherRMSE(2.6◦C)valuesthanthoseobtainedwithoursimulationconfigurationfortheprediction of2-mtemperatures.It’sonlybyrunningatahigherspatialresolution(250m)andwithaverydetaileddescriptionofsurface characteristicsthattheperformanceofthemodelisimproved,withaMBEof+0.2◦CandaRMSEof1.5◦C,yetcomparableto theoneobtainedwithoursimulationconfiguration.SimilarscoresareobtainedbyLemonsuetal.(2015)whosimulatethe2003 heatwaveatthespatialresolutionof1kmwithachainofmodelsincludingTEB:aMBEonnighttimetemperaturesranging from+0.41to+3.72◦Cbetweenthemostandtheleasturbanisedlandcovers,witherrorsrangingfrom0.83to3.83◦C,and atdaytime,anaverageMBEof−0.52◦Canderrorsrangingfrom0.78to2.06◦C.Inordertominimizetheimpactofthebiases anderrorshighlightedinthereferencesimulation,theanalysisoftheimpactsofthegreeningscenariosinthepresentstudyis carriedoutinrelativeterms(relativetothereferencesimulationbeforegreening). 2.2. DesigningrealisticgreeningscenariosforPariscity Thegreeningstrategiesweevaluateconsistingreeningthecitywhereverpossiblebypartiallytransformingartificialurban surfacesavailableintogreenspaces.Weassumeaspartlyavailableforgreening,theurbansurfacesatgroundlevelsuchas pavements,squares,roundaboutsandcarparks(surfacesthatareneitherbuiltasbuildingsorroads,oralreadyvegetated), aswellastheroofsofsomebuildingtypes.Atgroundlevel,acertainpercentageoftheurbansurfaceavailableforgreening Table2 Characteristicsofthescenariossimulated. Scenarioacronym Conversiontovegetation Additionalgreenareas(km2) Vegetationtype Irrigation GR Roofsofsuitablebuildingsa 251(+28%) Sedums Noirrigation IGR Roofsofsuitablebuildingsa 251(+28%) Sedums Dripirrigation LV25 25%ofgroundavailableb 100(+11%) Lowvegetation:grassandsmallshrubs Sprinklersystems LV50 50%ofgroundavailableb 199(+23%) LV75 75%ofgroundavailableb 298(+34%) LHV25 25%ofgroundavailableb 100(+11%) Lowandhighvegetation:60%grassand Sprinklersystems smallshrubs40%ofdeciduoustrees LHV50 50%ofgroundavailableb 199(+23%) LHV75 75%ofgroundavailableb 298(+34%) LHV75-IGR 75%ofgroundavailableb 549(+62%) Lowandhighvegetation Sprinklersystems Roofsofsuitablebuildingsa Sedums Dripirrigation a “Suitablebuildings”fortheimplementationofextensivegreenroofsrefertobuildingtypeswithflatroofsforwhichtheenergeticalperformancehasbeen demonstratedintheliterature,i.e.individualanddensehousingandwarehouses. b “Groundavailable”referstogroundsurfaceswhichareneithernatural,norbuiltasbuildingsorroads.Inarealcityitcorrespondstopavements,carparks, squares,roundabouts,etc. C.Muncketal. /UrbanClimate23(2018)260–286 265 is converted to green space, either 25, or 50, or 75%. This greening is realised with either low vegetation (grass and small shrubs)oramixoflowandhigh(10m)vegetationcomposedof40%ofdeciduoustrees(oftenreferredtohereafterasmixed wooded vegetation). At roof level, the surfaces available for greening are those of roofs of low-rise buildings (following the conclusionsofNgetal.,2012)forwhicharchitectureorregulationwouldallowfortheirinstallation,whichmeansinTEB,roofs ofIndividualandDensecollectivebuildingsandWarehouses(seeAppendixA).RoofsofHaussmannianbuildingsareexcluded forarchitecturalandregulatoryreasonsandroofsofofficetowersfortheirinefficiency.Ontheroofsofthethreetypesofsuitable buildingsareimplementedextensivegreenroofs(plantedwithsedum),whicharerecognizedfortheirenergeticalperformance. ThecharacteristicssimulatedinTEBforthistypeofgreenroofaredescribedindeMuncketal.(2013a).Followingprevious studieshighlightingthatvegetationisaneffectiveadaptationmeasureonlyifitiswatered(Danieletal.,2016;Kounkou-Arnaud etal.,2013),wechosetowatergroundvegetationinsummerwhateverthescenario.Inthecaseofgreenroofstwoscenariosare tested,withorwithoutirrigationinsummer.Alastscenario,consistingincombiningimplementationofirrigatedgreenroofs andgreeningof75%withamixedwoodedvegetationisalsosimulated. Forsummerirrigation,thechoiceofwateringvolumeandscheduleisbasedonasensitivityanalysisundertakenwithTEB underthemeteorologicalconditionsofthe2003heatwave(deMunck,2013),whichaimwastooptimizeplantwaterusefor anon-restrictedevapotranspiration,henceaircoolingeffect.Withregardtovolumes,wechoseforgreenroofs25Lperm2 perweek,avaluederivedaftertheguidelinesprovidedbySOPREMA(internationalmanufacturerspecializedinwaterproofing solutionsforurbaninfrastructuresincludinggreenroofsandwalls,SOPREMA,2011)forFranceandindroughtconditions.For groundvegetation,wetestedthreevolumes(25,50and75Lperm2 perweek),aswellasthreewateringschedules(in3,6 or8h),allthreeatnight,inordertooptimizebothplantwateruseandthenighttimeaircoolingeffect.Night-timewatering waschosentoimprovenight’srestandreducenighttimethermalstress,amajorfactorrelatedtoexcessmortalityduringthe 2003heatwaveinFrance(Doussetetal.,2011;Ledransetal.,2005).Weconcludedthatauniquewateringschedule,overeight hoursbetween21:00and05:00localtime,providingtheequivalentof25Lm−2 week−1 (i.e.410−5kgofwaterm−2 s−1) wouldbesuitableforbothgroundandroofvegetation,theonlydifferencebeingthewateringsystem,withsprinklersforground vegetationanddripirrigationforgreenroofs.ThewateringwithsprinklersisparameterizedinTEBasrain,whichisrealisticfor lowvegetationbutmayslightlyoverestimatetheamountsofwaterinterceptedinthecaseoftrees. Forpracticalreasons,acronymsweregiventothosescenarios.TheyareprovidedinTable2alongsideasummaryoftheir characteristics,andusedthereafter.Inallcases,scenariosincreasedthedensityofexistingurbanvegetation(representedbythe REFscenario)intheproportionswhichareprovidedinTable2andillustratedbythemapsofFig.1.Overtheentiresimulation domain,greening25,50of75%ofavailableground-basedurbansurfacescorrespondstoincreasingexistinggreenspacedensity byrespectively11,23and34%.Also,ascanbeseenonFig.1,becausethegroundsurfaceavailableforgreeningisquitelow inthecitycentre,themorewegreentheground(from25to75%),themorethegreeningisonlypossibleontheoutskirtsof Paris.Onthecontrarygreeningsuitableroofs(GRandIGRscenarios)allowswithinParistocompensateforthelackofground space.Theincreaseinurbanvegetationdensityobtainedthroughthesetwoscenariosisfarfromnegligibleatthescaleofthe simulationdomain(+28%)andishigherlocallywithinPariscity(between+30to+40%).Thescenariocombiningthegreening ofbothgroundandroofs(LHV75-IGR)presentsthemaximumvegetationdensityincreasewith+62%. 2.3. Choiceofindicators Theimpactsofgreeningstrategiesareanalyzedbycomparingthreetypesofindicatorsofinterestwhicharecomputedby TEB:levelsofthermalheatstressreached,energyconsumptionsforcoolingandheatingbuildings,waterresourcesrequiredfor summerwateringaswellasurbansurfacerunoff(fromartificialandnaturalurbansurfaces)allyearround. The 2-m air temperature is compared between scenarios as a first indicator of greening impact. Then levels of thermal comfortorheatstress(andtimespentateachlevel)areevaluatedthroughtheUniversalThermalClimateIndex(UTCI,Fiala etal.,2012)andthethresholdsestablishedbytheGlossary ofTermsforThermalPhysiology(TheCommissionforThermal PhysiologyoftheInternationalUnionofPhysiologicalSciences,2003)andindicatedinTable3.TheUTCIiscalculatedinTEB takingintoaccounttheclimaticconditionsforapersonplacedinthemiddleoftheurbancanyon(airtemperatureandhumidity, windspeed,directandreflectedsolarradiationandinfraredsolarradiationemittedbythewallsofthecanyon).TheUTCIcan becalculatedforapersonwhostandsinthesunorintheshade.Inthelattercase,thecalculationoftheUTCIwillnottakeinto accountthedirectsolarradiation. Tocomparetheenergeticalperformancesofthegreeningscenariosinthecontextofthe2003heatwave,wechoosefor indicatorstheenergyconsumptioncumulatedaswellasthepeakofenergydemandduringtheheatwave(air-conditioning). Table3 AssessmentscaleofUTCIforhightemperatures.Source:www.utci.org. UTCI(◦C) Thermalstresslevel Above+46 Extremeheatstress(EXS) +38to+46 Verystrongheatstress(VSHS) +32to+38 Strongheatstress(SHS) +26to+32 Moderateheatstress(MHS) +9to+26 Thermalcomfort 266 C.Muncketal. /UrbanClimate23(2018)260–286 Fig.2. Boxplotsofaverageimpactsofgreeningscenariosonminimumandmaximumdailyurban2m-temperaturesovertheentiresimulationdomainduring the2003heatwaveforeachscenarioandcomparedtothesimulationbeforegreening(REF). Fortheseasonalanalysis,wecomparethe10-yearmeanenergyconsumptioncumulatedduringeachofthefourseasons(air- conditioningandheating). 3. Impactsinthecontextofheatwave TheimpactsofgreeningscenariosareestimatedfortheheatwaveofAugust2003,andmorespecificallyforasix-dayperiod, extendingfromthe8thtothe13thofAugust,chosentobethehottestperiodofthisexceptionalheatwavefortheregionIle-de- France.Thesesixdaysweremarkedbysettledanticyclonicconditions,withdryandstableairmassesinthelowerlayersofthe atmosphere,whichgotwarmerandwarmeruntilthe14thofAugust,markingtheendoftheheatwave. 3.1. Thermalimpacts 3.1.1. 2-mairtemperatures Tohaveinmindtheorderofmagnitudesbetweentheimpactsofgreeningstrategiesintermsof2-mairtemperaturesand theindicatorspresentedpreviously,wecomparedtheminimumandmaximum2-mairtemperaturesthroughtheanalysisof thetemperatureanomaliesbetweeneachgreeningscenarioandthesimulationbeforegreening(namedREF). Basedonoursimulations,ourresults(Figs.2,3and4)suggestthatgreeningsuitableroofsdoesnothaveanimpacton2-m airtemperaturesunlessthoseareirrigatedandthatthisimpactonlyreallyappearsatdaytime.Andeveninthiscase,theaverage daytimecoolinginducedduringthesixdaysoftheheatwaveatthescaleoftheentiresimulationdomainisrelativelylimited, withamediancoolingof−0.17◦C(Fig.2).Lookingatthespatialdistributionofthiscooling,thepotentialofirrigatedgreen roofsforcoolingmaximumtemperaturesgenerallyrangesfrom−0.25to−0.5◦Cdependingonbuildingtype,respectivelyfor individualanddensecollectivehousing.ItisnoticeableeveninthecentreofParisandcanreachupto−0.67◦Clocally(Fig.4). Themaximumcoolinginducedbyirrigatedgreenroofsinthesimulationdomaininthecourseoftheheatwaveis−0.97◦C. Thetemperatureanomaliesgeneratedbytheincreaseingroundvegetation(systematicallywatered)areshownonFig.2. Thesescenariosshow,asexpected,agreatercoolingof2-mtemperaturesthanthatgeneratedbyrooflevelvegetationwith threesignificantresults: 1. Foragiventypeofvegetation(LVorLHV)nightanddaytimecoolingdonothavethesamemagnitude(Fig.2).Amaximum effectisobservedatdaytimeforscenarioswiththeadditionoflowvegetation,withamediancoolingof−0.19◦C,−0.37◦C C.Muncketal. /UrbanClimate23(2018)260–286 267 Fig.3. Spatialdistributionoftheimpactsofgreeningoverminimumdailyurban2m-temperaturesforintermediategroundgreening(LV50andLHV50),roof greening(IGR)andmaximumgreening(LHV75-IGR),expressedastheaverageimpactsduringthesixdaysoftheheatwave. and−0.55◦CrespectivelyforLV25,LV50andLV75(Fig.2).Conversely,theimplementationofamixedwoodedvegetation leadstoamaximumeffectatnighttime,withamediancoolingof−0.38◦C,−0.45◦Cand−0.50◦CrespectivelyforLV25,LV50 andLV75(Fig.2).Whileitislogicalforthecoolingeffecttobeoptimumatdaytimewhenthevegetationisphotosynthetically active(andthereforeevapotranspirationratesattheirmaximum),asisthecaseforlowvegetationscenarios,weobtain ahighercoolingpotentialatnightforthemixedwoodedvegetationscenariosthatisexplainedbythewateringschedule andthevegetationtype.Indeed,whenthetreesarewateredatnight,sprinklersbeingsimulatedinTEBasartificialrain, alargepartofthewaterinterceptedbythefoliageandthesoilisdirectlyevaporatedandisnolongeravailableduring theday,generatinglesscoolingatthattimeofday.Thisoccursalsoforlowvegetation,buttoalesserextentduetoits smallerinterceptionsurfacesandlowertransferturbulentcoefficients.Notethatfortrees,thisphenomenonisprobably morepronouncedthaninrealityduetosprinklerwateringsincegreateramountsofwaterareinterceptedbytreeleaves andaredirectlyevaporated,hencearenotstoredintheground.Thisshowsthelimitsofthecurrentparametrization forthewateringofgroundvegetation(thatwasduetoconstraintsinthecodestructureoftheISBAscheme),whichis undifferentiatedforgrassesandtrees.Thesetimingandmagnitudedifferencesinthecoolinggeneratedbylowandmixed woodedvegetationscenariosarewellvisibleonthemapsofFigs.3and4whencomparingLV50andLHV50scenarios. Duringthenight,thestrongestcoolingisgeneratedbytheLHV50scenarioandrangesonaverageduringtheheatwave between−0.25and−2◦C,withamaximumcoolingpotentialobservedinareasofdensecollectivehousing.Meanwhile, theLV50scenarioonlygeneratedacoolingofaround−0.25to−0.5◦C.Inthecourseoftheheatwave,thegreatestcooling ofminimumtemperaturesgeneratedbytheLVH50andLV50scenariosamountsrespectivelyto−2.2and−1.1◦C. 2. Whenlookingateachtypeofgroundvegetationsimulatedseparately,thehigherthesurfacegreeningrate,thegreater the cooling generated. When greening 25, 50 or 75% of available surfaces with low vegetation, this corresponds to a daytimecoolingof−0.19,−0.37and−0.55◦Candanighttimecoolingof−0.04,−0.08and−0.12◦C(medianvaluesof boxplotsonFig.2).Similarly,foramixedvegetation,thiscorrespondstoadaytimecoolingof−0.08,−0.31and−0.49◦C andanighttimecoolingof−0.38,−0.45and−0.50◦C.Lookinglocally,iftakingLVscenariosasanexample,thegreatest 268 C.Muncketal. /UrbanClimate23(2018)260–286 Fig.4. Spatialdistributionoftheimpactsofgreeningovermaximumdailyurban2m-temperaturesforintermediategroundgreening(LV50andLHV50),roof greening(IGR)andmaximumgreening(LHV75-IGR),expressedastheaverageimpactsduringthesixdaysoftheheatwave. reductioninmaximumtemperaturesestimatedatthescaleofthesimulationdomainamountstoabout−0.5◦CforLV25, −1◦CforLV50and−2◦CforLV75. 3. Nighttimeresultssuggestthatforsimilargroundsurfacegreeningrates,thecoolingpotentialofamixedvegetationcom- posedof40%oftreesismoreeffectivethanthatofavegetationcomposedonlyofgrassesandsmallshrubs.Forthetime beingitishardertodrawconclusionsfromdaytimeresultssincethemodeldoesnotyetaccountfortreeshadingeffects. AsforthemaximumgreeningscenariocombiningLHV75andIGRhypothesis,itsimpacton2-mtemperaturesisquitesimilar tothatoftheLHV75sincethecoolingpotentialofIGRonstreetleveltemperaturesislimited(Fig.2).Itisonlyperforming sensiblybetteratdaytimeinareasaroundPariscitycentrewithdensecollectivehousingwhereIGRmanagetofurtherenhance thecooling,resultinginadomainmedianimpactof−0.65◦C. 3.1.2. UHIintensity Alongthecross-sectiondefinedonthemapofFig.5,the2-mnighttimeUHIintensitybetweenPariscitycentreandthe50- mawaycountrysidewasestimatedbeforegreeningataround5◦Cwiththestandardspatialstructurerepresentedbythegrey line.Asexpectedfromprevioustemperatureresults,thegreeningofroofsdoesnotmodifytheUHIintensityatallsinceIGR andREFtemperatureprofilesareidentical.ToevaluatethemaximumimpactofurbangroundgreeningontheUHI,previous temperatureprofileswerecomparedtothescenarioswiththegreatestgreeningrate(75%,i.e.LV75,LHV75andLHV75-IGR).As expectedfrompreviousresults,thescenariowithamixedwoodedvegetation(LHV75)inducedthegreatestreductioninUHI intensity(−1.2◦C),whichlowerstheUHIto3.8◦C.TheimpactoflowvegetationonUHIisrelativelylowincomparison(inthe rangeof0.25◦C).Asforirrigatedgreenroofs,theydonotallowforafurtherreductionoftheUHIintensityinducedbygreening thegrounds(LHV75-IGR). C.Muncketal. /UrbanClimate23(2018)260–286 269 Fig.5. 2-mmeannighttimeairtemperatureprofilesduringthesixdaysoftheheatwaveforthewest-to-eastcrosssectionidentifiedonthemapoftempera- turespresentedfortheREFsimulation,estimatedacrossthreehourlyterms-2,3and4UTC.Theprofileforthebaselinesimulation(REF)showsaheatisland intensityofaround5◦C. 3.1.3. Thermalcomfort Toanalyzetherespectiveimpactsofthegreeningscenariosintermsofoutdoorthermalcomfortorstressexperiencedduring theheatwaveviatheUTCI,wefirstcomputedtheUTCIforeveryhourofthesixdaysofthe2003heatwave(seeSection2.3). TheUTCIvaluesobtainedwerethensortedinthefiveheatstressclassesdescribedearlier(Table3)tofinallycomputethemean timespentdailyineachheatstressclass(overthe6:00to22:00timeslot)forapersonoutsideinthesunorintheshade. Consideringlookingforshelterintheshadeafairlynaturalreflexinheatwaveconditions,wefocusedtheanalysisofour greeningscenariosonthedailyimpactsthattheywouldhaveonthetimespentatleastinverystrongheatstress(referredto asVSHSandcorrespondingtoUTCI>38◦C)byapersonintheshade,acrosstheentiresimulationdomainandacrossthethree maintypesofurbanfabric.TheseresultsarepresentedinTable4. Asexpectedfrompreviousimpactsonstreetleveltemperatures,non-wateredgreenroofs(GR)donotimproveoutdoor thermalcomfort.Nevertheless,withoutmodifyingthetotaltimespentinthermalheatstress,theirwatering(IGR)allowsfor thetimespentatleastinVSHStobemarginallydecreased(by15minonaverageacrossthewholecity).Notethatwatered greenroofsappearmoreeffectivetoimprovestreetlevelthermalcomfortinthedensecollectivehousingfabricthaninany otherurbanfabrics(22minlessspentatleastinVSHScomparedto11and9minlessrespectivelyintheHaussmannianand individualhousingfabrics).Thisisexplainedbythehighroofgreeningrate(Fig.1)andtherathersmallstreetaspectratiosin thisurbanfabric(seeAppendixA). Onthecontrary,increasinggroundvegetationinthecityimprovesoutdoorthermalcomfort,bybothincreasingthetime spentinthermalcomfortandbyreducingthatspentinthevariousheatstresslevels,withthemostmarkedeffectonVSHS.And Table4 Meantime(inhoursandminutes)spentdailyduringthe2003heatwaveatleastinverystrongheatstress(UTCI>38◦C)forapersonintheshadebefore greening(REF)andtimereductionsimulatedforeachgreeningscenarios.Meansarecalculatedoverdaytimehours(06:00–22:00LST),integratedacrossthe entireurbanzoneandacrossthethreemainhousingtypesandweightedbypopulationdensity(estimatedfromnationalandpublicdatabasesbyBonhomme, 2013). Scenario Town Haussmannianbuildings Densecollectivehousing Individualhousing Meantimespentatleastinverystrongheatstress REF 5:37 5:15 6:34 4:55 Impactofscenario GR – −0:01 +0:02 – IGR −0:15 −0:11 −0:22 −0:09 LV25 −0:20 −0:18 −0:26 −0:15 LV50 −0:39 −0:39 −0:49 −0:30 LV75 −0:55 −0:43 −1:10 −0:45 LHV25 −0:15 −0:21 −0:19 −0:10 LHV50 −0:37 −0:37 −0:49 −0:28 LHV75 −0:56 −0:51 −1:12 −0:43 LHV75-IGR −1:04 −0:53 −1:24 −0:50