Monitoring and modelling carbon monoxide concentrations in a deep street canyon: application of a two-box model



AtmosphericPollutionResearch3(2012)311Ͳ316





Atmospheric Pollution Research 



www.atmospolres.com



Monitoring and modelling carbon monoxide concentrations in a deepstreetcanyon:applicationofatwoǦboxmodel FabioMurena P.leTecchio,80,80125Napoli,Italy

ABSTRACT



Carbonmonoxideconcentrationsweremonitoredat3,10and30 metersinadeepstreetcanyonwithanaspect ratio of (H/W)у5.8 and were modelled using a two–box model developed in a previous CFD study. The th th monitoringcampaignlasted5days,from11 to15 July,2011.Theturbulentkineticenergyattherooftoplevel and traffic flow was also measured in the same period. Experimental data were used to evaluate parameters (mass transfer velocities and overall mass transfer velocity) of the box model. The daily pattern shows a significantincreaseoftheoverallmasstransfervelocityfrom9:00to11:00andadecreaseuntil14:00.Turbulent kinetic energy measured at the rooftop level seems to play a major role with respect to wind velocity in determining the mass transfer between the canyon and the atmosphere above. The evaluation of the overall masstransfervelocitycontributestotheuseofoperationalstreetcanyondispersionmodelsinthecaseofdeep streetcanyons.

Keywords: Atmosphericpollution Deepstreetcanyon Operationalmodels Boxmodels Masstransfer

ArticleHistory: Received:15March2012 Revised:28May2012 Accepted:29May2012

CorrespondingAuthor: FabioMurena Tel:+39Ͳ081Ͳ7682272 Fax:+39Ͳ081Ͳ2391800 EͲmail:[email protected] ©Author(s)2012.ThisworkisdistributedundertheCreativeCommonsAttribution3.0License.

doi:10.5094/APR.2012.034

 

1.Introduction  Modelling air pollution in large urban areas is a challenging task that is engaging researchers in finding an effective solution. Deterministic models such as Computational Fluid Dynamic (CFD) modelscanprovideusefulinformationiffocusedonthestudyofa singleorafewstreets(Sinietal.,1996),buttheirapplicationover awholeurbanareaisnotfeasibleduetothelargeextensionand complexityofurbanareas,thusincurringhighcalculationtimeand costs.Operationalmodelsaremuchlesssophisticated,needfewer inputparametersandthemathematicsisoftenreducedtosimple algebraic equations. For this reason, operational models seem to be more appropriate if a large–scale urban problem must be addressed.Themostwell–knownoperationalmodels(e.g.,STREET, OSPM, and ADMS) are frequently used at local (single or few streets) or urban scale and perform reasonably well but not always,andnotunderalloperatingconditions(Vardoulakisetal., 2007).  The mass transfer between the canyon and the atmospheric flow above is one of the key parameters in operational models. High concentrations of vehicular pollutants can occur in street canyons at pedestrian height in the case of high vehicular emissions and ineffective mass transfer due to the flow regimes established.Flowregimesareclassifiedasisolatedroughnessflow, wakeinterferenceflowandskimmingflow(Oke,1987)depending on the aspect ratio (H/W). If (H/W)>1.6–2, the street canyon is classified as deep (Vardoulakis et al., 2003), and two counter–

rotating vortices may form (Sini et al., 1996), with the bottom vortexweakerthantheupper.  Manypapershavefocusedonmasstransferinstreetcanyons (BenthamandBritter,2003;HamlynandBritter,2005;Salizzoniet al., 2009). Mass transfer inside the canyon and between the canyon and the atmosphere above is influenced by several parameters,andtheconclusionsofdifferentpapersdonotalways agree. The parameters investigated include the following: the external wind velocity (Murena et al., 2011), the turbulent structures in the outer flow and the structures in the shear–layer interfacebetweentheouterflowandthecanyon(Salizzonietal., 2011), the turbulence inside the shear layer that forms between the canyon and the atmosphere above (Caton et al., 2003), the aspectratio(SolazzoandBritter,2007);thefrontalareadensityof buildings(Rattietal.,2002),andtheplanarareadensityorpacking density(BenthamandBritter,2003;HamlynandBritter,2005).  Once a certain understanding of the mass transfer phenomena inside the canyon and between the canyon and the atmosphere above is achieved, street scale operational models, also called mass balance or box models, can be developed (Salizzonietal.,2009;Murenaetal.,2011).  In a previous paper, Murena et al. (2011) proposed a box model to simulate mass transfer inside deep street canyons and between them and atmospheric flow above based on a CFD simulationstudyonideal(2D)deepstreetcanyons.Anidealstreet canyon is defined as a single road of infinite length delimited by

312 2

Murena–AtmospherricPollutionResearch3(2012)311 1Ͳ316

buildings of the same s constant height on botth sides of the road d with wind dirrection perpendicular to the street axis. The box and mo odel proposed by Murena ett al. (2011) asssumes that a deep streeet canyon can n be divided, in the vertical length, into tw wo or three well–mixed volumes exch hanging mass and a with the upper u volume exchangin ng mass with the t atmospherre above. The mass excchange rates were evaluaated through transient 2D D–CFD sim mulations. Theirr values are rep ported as a fun nction of the aspect a ratiio and characcteristic wind velocity (Murrena et al., 2011). 2 Beccause the masss transfer proccess from the bottom level in the can nyon to the atm mosphere abovve is made up of a series of first– ord der processes, an overall mass m transfer coefficient caan be deffined (Murena et al., 2011).. Once the ovverall mass traansfer coeefficientisknow wn,thepollutaantconcentratiioninsidethesstreet can nyonattheped destrianlevelccanbeeasilycaalculated.Thereefore, thee overall masss transfer coeffficient is a powerful param meter; unffortunately,itissnotconstantbecauseitdepeendsongeomeetrical parrametersandm meteorologicalcconditions.InttheSTREET(Joh hnson et al., a 1973) and OSPM (Hertel and Berkowiczz, 1989) modelss, the mass transfer velo ocity is assumeed proportionaal to a characteeristic velo ocityintheatm mosphericflow wabove.Howevver,thedepend dence of the mass tran nsfer velocity on o the atmosp pheric flow is more mplex, as repo orted by otherr authors (Hottchkiss and Haarlow, com 197 73;Solazzoand dBritter,2007;Salizzonietal.,2011).  Inthispaper,,theresultsoffamonitoring campaignofcaarbon mo onoxidemeasurredat3,10an nd30minadeeepstreetcanyyonin thee urban area of o Naples are interpreted with w the box model m preeviously propossed (Murena et e al., 2011), and a values of mass tran nsfer velocitiess and the oveerall mass tran nsfer coefficien nt are obttained. 



3.Th heModels

 In a previous study, s Murena et al. (2011) performed p 2D––CFD simu ulations of the mass transfer inside deep street canyons and betw ween the canyyons and the atmosphere above. Ideal street canyyonswith(H/W W)=3and5weereconsidered.Theresultsoffthe CFDsimulationssho owed(Murenaetal.,2011)th hatinthecase ofa et canyon witth (H/W)=3, two counter rotating vorttices stree form med, which is in accordancee with resultss reported in the literaature (e.g., Sin ni et al., 1996; Jeong and An ndrews, 2002). For (H/W W)=5, the sam me paper (Murena et al., 20 011) indicates the prese ence of three vortices. v This ffinding is less documented d in the literaaturebecaused deepstreetcan nyonsarerarelyyconsidered.M Most pape ers study streett canyons with h an aspect rattioу1, for whicch a single vortex is formed. Jeong an nd Andrews (20 002) and Sini et al. (1996), however, reported r the p presence of three vortices when w W)>3.CFDsimulationsshoweedthatCOconcentrationscan nbe (H/W assumedasuniform mineachvorteex(Murenaetaal.,2011).Withthis assumption, mass balance equations can be eassily written. In the hemodelisatw wo–boxmodeland caseof(H/W)=3(sseeFigure1),th them massbalanceequationsare:   dcb f Q Hb u bu cb  cu  e v  (1) dt W dc H u u u bu cb  cu  u ua cu  c a  (2) dt  where Hb and Hu are, a respectivelly, the height of o the bottom and uppe er boxes; cb an nd cu are the C CO concentratio ons in the bottom andupperboxes;caistheairconccentrationabovvetheroofs(ou utof m–1); theccanyon);feisaweightedaverrageCOemissiionfactor(gkm and Qv isthenumberofvehiclesp perhour.Finally,ubu(ms–1)issthe 2.EExperimentalMethodandAnalyticalAp pparatus massstransferveloccitybetweenth hebottomand dupperboxes, and  uua(ms–1)isthemaasstransfervelocitybetweentheupperboxand noxide concentrations were measured at via Carbon mon theaatmosphere.Th hemasstransfeervelocityisa spatiallyaveraaged Narrdones,adeep pstreetcanyon intheurbanaareaofNaples((Italy) veloccity that repreesents the net mass exchan nge between two witth the followin ng geometry: width Wу6m m, average building volum mes at uniform m concentratio on; it is also caalled the exchaange heightHу35m(H H/W)у5.8and dlengthLу315 5m.Onlyoneccross– rate(BenthamandBritter,2003)o orexchangevelocity(Hamlynand roaad and a sidee road are preesent throughout its length h. Via Britter, 2005). The mass transfer velocity takes into account both b Narrdones is a on ne–way uphill street with an average slop pe of instaantaneous(turb bulent)andmean(advective)ccontributions. app proximately5% %.Theorientatio onofthestreetaxisand,hence,of thee traffic flow is in the direcction 70–250° (i.e., from E–N NE to  W––SW).  COwasmeassuredusingano on–dispersiveinfraredphotom meter anaalyser (ML 9830B Monitor Eu urope Ltd. with a lower detecctable limit LDL=0.05pp pm). Calibration of the instru uments was cheecked eacch day during the monitoringg campaign, performing zero o and spaanoperations.O One–minuteavverageconcenttrationswerestored by theML9830B analyserandd downloadedonalaptopdurin ngthe cam mpaign. The saampling points were located at 3, 10 and 30m fromtheroadlevelandapproxim mately1mfromthebuilding walls on the south sidee of the streett. The samplingg line was made of Tefflon. Airflow was w turned from m one samplin ng point to another eveery20minutes operatingonaavalvesystem. Thelagtimefo orthe mo ostremotesam mplingpointfromtheanalyserrwasapproxim mately 1m minute. Therefo ore, the first 2minutes of each measureement were discarded to allow flushing of the sampling line.. The ons were calcu ulated and assu umed 20––minute averagge concentratio tob behourlyaveraageconcentratiionsforeachsaamplingpoint.  Figurre1.Schemeofth heboxmodels. wheel,  Traffic flow was manually measured by counting 4–w 2–w wheel and low––duty vehicles.. Two measureements of 5minutes If (H/W)=5, the model is a three–box model, and the were performed each hour from f 7:00 to 23:00. Some spot equaationsare: w performed during nigh ht hours. Turb bulent meeasurements were  kinetic energy was w measured at the roofttop level usin ng an dcb f Q Hb u bm cb  c m  e v  (3) ultrrasonicanemom meter(DeltaOhmHD2003),w whichwasplaceedon dt W a pole p (approxim mately 4m heigght) above thee roof of the same dc buildingwherethesamplingpointswerelocateedonthesouth hside H m m ubm cb  cm  u mu cm  cu  (4) oftthecanyon.Theemonitoringcaampaignlasted5days,from1 11thto dt th 15 July2011.



Murena–AtmospherricPollutionResearch3(2012)311 1Ͳ316

dcu u mu cm  cu  uua cu  ca  (5) dt    The notation n is the same as in Equatio ons(1) and (2). The sub bscriptmcorresspondstothem middleboxinth hecanyon.  Starting from m the models above, a it is po ossible to defin ne an oveerallmasstranssfervelocitycoeefficientUov(M Murenaetal.,20 011):  1 U oov  1 1 (6)  ubu uua   inthecaseoffthetwo–box modelor,incaaseofthethreee–box mo odel:  1 U ov  o 1 1 1 (7)   ubm umu uua   masstransfercoefficientcan beusedtoevaaluate Theoverallm CO concentration ns at the pedestrian level (M Murena et al. 2011), 2 assumingsteadysstateconditionss:  f Q (8) cb ca  e v   WU ov  Equation(8) has practicaal applicationss other than the app parently strongg steady–statee assumption. In fact, if a time inteervalofoneho ourisconsidereed,asteadystaatecanbeassu umed, and dhourlyaveraggevaluesofcb canbecalculatedbyEquatio on(8). Onee hour is thee minimum averaging time of all air quality legislationsinthe world.Therefo ore,hourlyaveerageconcentraations quation(8) is similar of pollutants are of great practical interest. Eq quations of op perational mod dels such as OSPM. O to the solving eq Con nsidering the simple case when the vo ortex is completely imm mersed inside the canyon (according to asssumptions maade in thee OSPM, this is the case when (H/W)ш1, the pollutant con ncentrationcan nbewrittenasffollows(Berkow wiczetal.,1997 7):  Q  c (9) WV wt   where Q is thee CO emission n rate in thee street (gm–1s–1) oᐦՂܳ˜inEquatio on(8)],Wisth hecanyonwidthand [correspondingto ʍwt isthecanyon ventilationvelo ocity.If(H/W)шш1,thenʍwtу 0.1ut (Beerkowicz et al. 1997), where ut is wind speeed at the top of o the can nyon.Theoveraallmasstransfeercoefficientin nEquations(6)––(8)is giveenasafunction nofaspectratioandwindvelocity(Murenaetal., 201 11).  Hu

4.R Results  onditions durin ng the Before discusssing the resultts, the wind co mo onitoring campaaign are described. The averaage hourly valu ues of win nd velocity and d wind direction are reported d in Figures2 and a 3. Thee average hourly values of wind w direction were calculateed by dividing the wind d arising in 16ssectors and avveraging the daata of thee sector with the higheest frequency of observations. Meeteorological measurementss were obttained from the meeteorological station of Naples Cap podichino, located app proximately 5kkm NE of the monitoring site. The data were com mpared with sp pot measurements from the sonic anemom meter, and dtheywerein goodagreement.Thewindp pattern,asexpeected, is that t of a breezze regime. Thee wind velocityy is low (a5km mh–1) durring the night and starts to increases at 10 1 a.m., reach hing a

313

plate eau value (у15 5kmh–1) at no oon and remain ning constant until u 7p.m m., after which h it decreases. Additionally, the t wind direcction follows the breezee regime. Durin ng the day, staarting from 9aa.m., the sea s breeze brings the wind ffrom the S–SW W direction. Du uring the night, the dirrection is morre variable, bu ut it comes most m  frequ uentlyfromtheenorthernsectors(NW–NE). 

Figurre2.Averagehou urlyvaluesofwindvelocity. 

Figurre3.Averagehou urlyvaluesofwin nddirection.Ang glesbetweenthe grey lines representcondittionsofwindmaiinlyparalleltoth hestreetaxis(i.e.,50Ͳ 90°a and250Ͳ290°). 

Considering th hat via Nardon nes is oriented d in the direcction 70–2 250°, during most m of the m monitoring cam mpaign, the angle betw ween the wind direction and the street axiss was greater than t 20° (Figure3). Theerefore, the wind componentt perpendicular to nd the flow inside the can nyon the street axis is significant, an assumes a helicoidal pattern (YYamartino and Wiegand, 19 986). Therrefore, the mon nitoring campaaign was perforrmed in conditions apprroachingthoseofanidealstreeetcanyon(win ndperpendiculaarto thesstreetaxisandinfinitestreetleength).  The hourly average a CO co oncentrations measured at via Nard donesarereportedinFiguress4and5.Thecconcentrationsare –3 inmgm normalisedatT=20°C andP=1atm.Bothfiguressh how that the CO concentration at thee pedestrian level (cH=3) is higgher than the CO at 10m (cH=10) and 30m (cH=30 ). In fact, the H conccentrationiscH==3>cH=10уcH=300andsometime escH=30>cH=10. The average values du uring the who ole monitoringg campaign were w 1.65 mgmͲ3atH=3 3m,0.89mgm m–3atH=10mand0.65mgm mͲ3at H=3 30m. A hypoth hesis test show ws that the diffference in averrage value esofCOconccentrationatH=10mand dH=30mis not

314 4

Murena–AtmospherricPollutionResearch3(2012)311 1Ͳ316



where ߂੓ܾ and ߂੓‫ ݑ‬are the differences of CO C concentrations betw ween the jth ho our and the j––1th hour. All other variablees in Equaations(10) and d (11) are constant (Hb, Hu, W), are assum med consstant (ᐦՂ) or aree hourly averagge values (c an nd u). The averrage weigghtedemission factor(ᐦՂ)wasscalculatedthrroughtheform mula, ᐦՂ ൌ σܺ݅ᐦ݅ where the subscript refers to the e class of vehiicles measured (two–strroke; four–stroke and light–duty vehicles). ܺ݅ is theffractionofvehiclesofclassipassinginthecaanyonand f i issthe

Figu ure4.HourlyaveerageCOconcenttrationsmeasured datH=3,10and d30m th th from m11 to15 Julyy2011.



Figu ure5.Averageho ourlyvaluesofCO OconcentrationssmeasuredatH==3,10 th th and d30mfrom11 tto15 July2011atviaNardones.

 sign nificantataconfidenceintervvalCI=95%.Th hisobservation leads tottheconclusionthatalthoughttherealcanyon nhasanaspecttratio (H//W)>5, it can be modelleed by a two o–box model. The disccrepancy with the findings of the CFD simu ulations (Mureena et al., 2011), that indicate the presence of three vorticess and nsequentlythe needofathreee–boxmodelw when(H/W)>5,may con be due to the idealisation off the street canyon c in thee CFD sim mulations,wherretheeffectof lateralstreets wasnotconsid dered norrwastheroughnessofthereealcanyondueetothepresen nceof balconies. Both effects e could modify the flo ow field insidee the can nyon, determining the fusion of the middle and upper vorrtices. Theerefore, only the t two–box model m will be considered in n the following, assuming that the heeights H=10 m m and H=30m both belongtotheupp pervolumeandthatheightH==3mbelongsttothe botttomvolume(seeFigure1).  Differential Equations(1) and (2) weree than discreetised assuming a time interval ȴt of o 1hour and were rewritteen as follows:  'cb f Q Hb ubu cb  cu  e v  (10) 't W 'c H u u ubu cb  cu  uua cu  ca (11) 't 

average emission factor of classs I calculated d by the COP PERT procedure based on o the compo osition of the vehicular fleeet in Naples(Murenaet al.,2011).ܺ݅w wouldchangeh hourlyasthetraaffic becausethetraafficcompositio onis compositionchanges.However,b ht,itseffectonthe consstantduringtheedayandonlychangesatnigh value e of the averagge emission facctor is minor and a was negleccted. Therrefore,ᐦՂisconstant.IftheCO Oconcentration nsareknownffrom expe erimental data, then the only unknowns in Equations(10) and (11) are mass tran nsfer velocitiess, and ubu and d uua and can n be calcu ulatedthrougharegressionprrocedure.  Thehourlyaveeragevaluesoffubuanduuawe erethenevaluaated byEq quations(10)aand(11)assumiingashourlyavveragevalueso ofcb and cutheexperimentalvaluesattH=3mandthemeanofvalues atH =10mandH==30m,respectively.FortheC COconcentratio onin theaatmosphereabo ovetherooftop plevel,ca=0w wasassumed.  The other inp put parameters are Hb=3m m, Hu=32m and ᐦՂ=0 0.05gkm–1, as reported in M Murena et al. (2011). With th hese value es, it was obseerved that mass transfer velocities (ubu and uua) were e in many casees negative; in n fact, they mu ust be positivee by defin nition[Equation ns(1)and(2)]. Itwasimmediaatelyclearthattthe reaso onforthisresultwasthelow wvaluethatthe egenerationteerm § f eQ v · eters ¨ ¸  in Equattion10 assumees with the input parame © W ¹ adop pted. In fact, the t generation n term was ge enerally lower and 'c · § often nnegligiblewitthrespecttoth heaccumulationterm ¨ H b b ¸ . 't ¹ © Conssequently, everry time the acccumulation term was positivee (cb atho ourj>cbathou urj–1),asitisaalways(seeFiggure1)cb>cu,tthen ubum mustbenegativve[seeEquation(10)].  ntermcannot benegligiblew withrespecttothe Thegeneration accumulationterm becauseweexxperimentallyo observed(Figurres4 and 5) that it is cH=3m>cH=10mуcH=30m then cb>cu. We must m concclude that thee generation tterm is underestimated. In the gene eration term, th he only parameter that could d be affected by b a significanterrorinttheevaluationistheaverageweightedemission orᐦՂ.InthecalcculationofᐦՂ(M Murenaetal.,2 2011),theeffecctof facto coldstartsandpendencywereneglectedbecaussetheywereno otof real interestinthattcase.Tsangettal.(2011)anaalysedtheon–rroad emisssions of a Eurro–4 petrol carr driven on fou ur urban routees in Hongg Kong. They observed o that when the enggine is in fuel––rich operrationwithanaair/fuelratiolesssthan14.7,whichusuallyoccurs when a vehicle is accelerating a or when the veh hicle starts to drive d uphill,theCOemissionsaredrasticallyhigher.W Withroadgradeesof 10.5% %,7.0%and0% %,theemission nfactorsofCO wereestimatedas 20.5 gkm–1,2.05g km–1and0.43gkm–1,respecttively.Thereforre,it seem ms evident thaat, neglecting the effect of road pendency, ᐦՂ=0 0.05gkm–1 is an a underestimation of the actual value of the averageweightedeemissionfactor.Assumingasaaweightedaverrage emisssion factor that ᐦՂ=0.5gkm m–1 (ten times higher than that initiaally assumed), the generation n term was gen nerally higher than t the accumulation term, and the ubu values calculated by Equaation(10) werre positive. Once ubu was w obtained by Equaation(10), uua was calculated d by Equation(11). The averrage hourrlyvaluesofmasstransferveelocitiesubuand duuacalculated d by Equaations(10)and(11)arereporttedinFigure6. 



Murena–AtmospherricPollutionResearch3(2012)311 1Ͳ316

A comparison of mass traansfer velocitiees obtained byy CFD sim mulations (Murena et al., 2011) with thosee obtained thrrough mo odelling (Figuree6) is reported d in Table1. It is evident thaat the masstransfervelo ocitiescalculateedintheCFDssimulations(Mu urena per. etaal.,2011)arehiigherthanthosseobtainedintthepresentpap 

315

ured average hourly vallues of traffic flow ܳ˜, whicch were measu ng the campaaign (Spano, 2 2011). Uov wass calculated from f durin Equaation(12) witth following parameters: a=0.0025m ms–1, –1 b=0 0.006ms  and d ʍ=1.5h. A comparison of cb from the mo odel [Equation(8)] with real data is rreported in Figgures8 and 9. The corre elation with th he hourly averaage values (Figgure8) is R2=0 0.64. Theaaveragehourlyvalues(Figure9)areadequattelymodelled. 

Figu ure6.Averageho ourlyvaluesofubu dline)anduua(bo oxand b (circleandsolid dashedline).

 Tab ble1.Comparisonofmasstransferrvelocitiesobtain nedbyCFDsimullations (Mu urenaetal.,2011 1)andinpresentp paper  Win ndvelocity Ͳ1 ( (kmh ) 5 15

CFDsimulatio on ubu uua Ͳ1 Ͳ1 ( (ms ) (m ms ) 0.06 0 0.025 0.15 0 0.06

Modelling uua ubu u Ͳ1 Ͳ1 (m ms ) (ms ) 0.005 0.0 005 0.01 0.01Ͳ0 0.025

 Ifthemasstrransfervelocitieesubuanduuaaareknown,theenthe oveerall mass transfer coefficieent Uov can be b calculated from Equ uation(6). Thee average hou urly value–day values of Uov o  are rep ported in Figuree7 as box sym mbols. A maximum of Uov occu urs at 11a.m. An increaase of Uov is co orrelated with a a similar increaase of turbulent kineticc energy (TKEE) measured with w an ultraasonic aneemometeratth herooftopleveelduringthemonitoringcamp paign. Inffact,anincreasseinTKEisobsserved(Spano, 2011)starting from 9:0 00a.m.(Figure7),thenaplateeauisreached atnoon.TheTTKEis theen constant until 3:00p.m. The TKE measurements in Figgure7 aree the averagevvalues obtained d during the monitoring camp paign. Thee increase of TKE T can be exp plained both with w the increaase of win nd velocity and the heating caused by so olar irradiation n that increasestheatmosphericturbulenceandmakesthemasstraansfer mo oreefficient.Th hepatternofU Uovintheaftern noonislessclear,as sho owninFigure7,,althoughadecreasefromthemaximumvalueat 11:00a.m. is evid dent. From the data reported in Figure7, th he Uov aveerage hourly vaalues have beeen modelled asssuming a miniimum con nstant value plus p a Gaussian n function to describe the more effiicientmasstran nsferinthemorningfrom9a.m.to13a.m.  Themodelfunctionisthen:  ª § h  11 ·2 º U ov a  bexp «  ¨ (12) o ¸ » «¬ © V ¹ »¼    wherehisthecurrenthour(fro om0to23).Parametersa,baandʍ willbedetermined dthroughabesst–fittingproceedure.  Defining the daily pattern of overall maass transfer veelocity through Equation n(12), the CO concentration n at the pedestrian leveelcanbecalculatedbyEquatiion(8),eventh houghitisderivvedin steady state con nditions. The real conditions are only slightly unssteady; in fact, the geneeration term prevails overr the acccumulationterm m,whichismin nor.Theother inputparametersin Equ uation(8) are the emission factor (ᐦՂ=0 0.5gkm–1) and d the

Figurre 7. Average hourly values of ovverall mass transsfer velocity obta ained byEq quation6(boxes)andofturbulenttkineticenergy(ttriangle)measureedat thero oofͲtoplevelatvviaNardones.Theesolidlineisam modelcurve(Equa ation 12).TTheTKEvaluesarreontherightsca ale.



Figurre 8. Correlation of measured an nd modelled averrage hourly valuees of COco oncentrationsatH H=3. 

Figurre 9. Average ho ourly values of C CO concentration ns at H=3 meassured (box)andmodelled(so olidline).

316

Murena–AtmosphericPollutionResearch3(2012)311Ͳ316

A sensitivity test was performed on the effect of the dimensionofthetwovolumesinwhichthestreethasbeendivided (Hb and Hu). Hb could assume values in the range 3чHb<10 because,asshowninFigures4and5,thebottomboxmustinclude the 3m height and exclude the 10m height. Correspondingly, Hu=H–Hb. The variation of Hb in the range 3чHb<10m did not haveasignificanteffectontheresultsofthemodel. 

5.Conclusions  Thispapershowsthatthecarbonmonoxideconcentrationsin a deep street canyon can be effectively modelled by assuming a twoͲboxmodel(Murenaetal.,2011).ThetwoͲboxmodeldefines an overall mass transfer coefficient that, once known and with a few other input parameters, allows the evaluation of the concentrationatthepedestrianlevel,whichisofparticularinterest for environmental and health impact assessment studies. The overallmasstransfercoefficientdependsonmanyvariables,both geometrical (street geometry) and meteorological (wind speed, winddirection,atmosphericturbulence,temperature).Theresults of the present paper show that, at least in meteorological conditions occurring during the monitoring campaign (11thͲ15th July,2011)inNaplescharacterisedbyabreezeregime,theoverall mass transfer coefficient is quite constant during the 24Ͳh period apartfromasignificantincreaseinthemorningfrom9:00to11:00 followedbyadecreaseuntil14:00.Agoodcorrelationofrealand modelleddatawasobtainedbymodellingtheUovdailypatternasa constantplusaGaussianfunction.  Duringthemonitoringcampaign,thewinddirectionwasrarely paralleltothestreetaxis,sotheresultsofthispaperarerelatedto a wind direction mainly perpendicular to the street axis. In these conditions,theresultsseemtoindicatethatwindvelocityisnota significantparameter,whiletheturbulentkineticenergymeasured attherooflevelseemstoplayamorerelevantroleindetermining the mass transfer rate. These findings have been evidenced by a simulationstudy(Salizzonietal.,2011).  Theresultsofthispaperareofpracticalinterestbecausethey can be used to improve the performance of operational models, especiallyifappliedatdeepstreetcanyons,forwhichfewdataare reportedinliterature.Additionally,theresultsofthemostpopular operational models, such as OSPM and ASDM, are unreliable because they were developed and validated for street canyons withanaspectratioa1.  Futureresearchwillaimatavalidationofthemodelthrougha more prolonged monitoring campaign performed in different seasons and in street canyons with different aspect ratios. The correctevaluationoftheCOemissionrateduetovehiclespassing in the street canyon will be a critical issue in the validation procedure. 

References  Bentham, T., Britter, R., 2003. Spatially averaged flow within obstacle arrays.AtmosphericEnvironment37,2037–2043. Berkowicz, R., Hertel, O., Larsen, S.E., Sorensen, N.N., Nielsen, M., 1997. Modelling Traffic Pollution in Streets. Ministry of Environment and Energy,NationalEnvironmentalResearchInstitute,Roskilde,55pp.        



Caton,F.,Britter,R.E.,Dalziel,S.,2003.Dispersionmechanismsinastreet canyon.AtmosphericEnvironment37,693Ͳ702. Hamlyn, D., Britter, R., 2005. A numerical study of the flow field and exchange processes within a canopy of urbanͲtype roughness. AtmosphericEnvironment39,3243Ͳ3254. Hertel,O.,Berkowicz,R.,1989.ModellingPollutionfromTrafficInAStreet Canyon.EvaluationofDataandModelDevelopment.TechnicalReport, DMULuftAͲ129,NationalEnvironmentalResearchInstitute. Hotchkiss, R.S., Harlow, F.H., 1973. Air Pollution Transport in Street Canyons.TechnicalReport.U.S.EnvironmentalProtectionAgency,EPAͲ R4Ͳ73Ͳ029,Washington,D.C.,78pp. Jeong,S.J.,Andrews,M.J.,2002.Applicationofthek–ɸturbulencemodelto the high Reynolds number skimming flow field of an urban street canyon.AtmosphericEnvironment36,1137–1145. Johnson, W.B., Ludwig, F.L., Dabberdt, W.F., Allen, R.J., 1973. An urban diffusion simulation model for carbon monoxide. Journal of the Air PollutionControlAssociation23,490Ͳ498. Murena,F.,DiBenedetto,A.,D'Onofrio,M.,Vitiello,G.,2011.Masstransfer velocity and momentum vertical exchange in simulated deep street canyons.BoundaryͲLayerMeteorology140,125Ͳ142. nd

Oke, T.R., 1987. Boundary Layer Climates, 2  Edition, Methuen, London, 435pp. Ratti,C.,DiSabatino,S.,Britter,R.,Brown,M.,Caton,E.,Burian,S.,2002. Analysisof3ͲDurbandatabaseswithrespecttopollutiondispersionfor a number of European and American cities. Water, Air, and Soil Pollution:Focus2,459Ͳ469. Salizzoni, P., Marro, M., Soulhac, L., Grosjean, N., Perkins, R.J., 2011. Turbulent transfer between street canyons and the overlying atmospheric boundary layer. BoundaryͲLayer Meteorology 141, 393Ͳ 414. Salizzoni, P., Soulhac, L., Mejean, P., 2009. Street canyon ventilation and atmosphericturbulence.AtmosphericEnvironment43,5056Ͳ5067. Sini, J.F., Anquetin, S., Mestayer, P.G., 1996. Pollutant dispersion and thermaleffectsinurbanstreetcanyons.AtmosphericEnvironment30, 2659Ͳ2677. Solazzo, E., Britter, R.E., 2007. Transfer processes in a simulated urban streetcanyon.BoundaryͲLayerMeteorology124,43Ͳ60. Spano, E., 2011. Profili Verticali Di CO In DeepͲStreet Canyon. Chemical EngineeringThesis,UniversityofNaplesFedericoII. Tsang,K.S.,Hung,W.T.,Cheung,C.S.2011.Emissionsandfuelconsumption of a Euro 4 car operating along different routes in Hong Kong. Transportation Research Part D: Transport and Environment 16, 415Ͳ 422. Vardoulakis,S.,Valiantis,M.,Milner,J.,ApSimon,H.,2007.Operationalair pollution modelling in the UK Ͳ street canyon applications and challenges.AtmosphericEnvironment41,4622Ͳ4637. Vardoulakis, S., Fisher, B.E.A., Pericleous, K., GonzalezͲFlesca, N., 2003. Modelling air quality in street canyons: a review. Atmospheric Environment37,155Ͳ182. Yamartino,R.J.,Wiegand,G.,1986.Developmentandevaluationofsimple models for the flow, turbulence and pollutant concentration fields within an urban street canyon. Atmospheric Environment 20, 2137Ͳ 2156.