three-dimensional pollutant concentration dispersion of a ... · and velocity are stochastic and...

ARTICLE IN PRESS

1352-2310/$ - se

doi:10.1016/j.at

�Correspondfax: +8522365

E-mail addr

Atmospheric Environment 40 (2006) 484–497

www.elsevier.com/locate/atmosenv

Three-dimensional pollutant concentration dispersion of avehicular exhaust plume in the real atmosphere

J.S. Wanga, T.L. Chana,�, C.S. Cheunga, C.W. Leunga, W.T. Hungb

aDepartment of Mechanical Engineering, Research Centre for Combustion and Pollution Control, The Hong Kong Polytechnic University,

Hung Hom, Kowloon, Hong KongbDepartment of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

Received 22 September 2004; received in revised form 6 July 2005; accepted 23 September 2005

Abstract

The pollutant dispersion process from the vehicular exhaust plume has a direct impact on human health, particularly on

vehicle drivers and passengers, bicyclists, motorcyclists, pedestrians and people working nearby. A three-dimensional

vehicular pollutant dispersion numerical model was developed based on the Reynolds-averaged Navier–Stokes equations

coupled with a k2e turbulence model to simulate the initial pollutant dispersion process of carbon monoxide, CO, from a

vehicular exhaust plume in the real atmospheric environment. Since the ambient wind direction and velocity are stochastic

and uncontrollable in the real atmospheric environment, a wind-direction–frequency-weighted (WDFW) approach was

used to obtain the real pollutant concentration dispersion along with the development of the vehicular exhaust plume.

Within the specified sampling period, the ambient windflow conditions are transformed into the corresponding frequencies

of wind directions and averaged magnitudes of wind velocities from directions N, E, S or W. Good agreement between the

calculated and measured data for two diesel-fuelled vehicles indicates that with the WDFW approach the initial dispersion

of pollutant concentration from a vehicular exhaust plume in the real atmospheric environment can be truly reflected. The

present study shows that the dispersion process in the near region for the relative concentration of CO, from RC ¼ 0:1 (or

10%) to 1 (or 100%), is less influenced by the ambient wind velocity than by the vehicular exhaust velocity, but it is vice

versa in the far region from RC ¼ 0 (or 0%) to 0.1 (or 10%). It implies that the effect of vehicular exhaust exit velocity on

the dispersion process is more pronounced than that of ambient wind velocity at the vicinity of the exhaust tailpipe exit,

while the effect of ambient wind velocity gradually shows a significant role for the dispersion process along with the

development of a vehicular exhaust plume.

r 2005 Elsevier Ltd. All rights reserved.

Keywords: Vehicular exhaust plume dispersion model; Carbon monoxide; Wind-direction–frequency-weighted (WDFW) approach;

Numerical simulation

e front matter r 2005 Elsevier Ltd. All rights reserved

mosenv.2005.09.046

ing author. Tel.: +852 2766 6656;

4703.

ess: [email protected] (T.L. Chan).

1. Introduction

Motor vehicles are the major source of airpollutants in most densely populated urban cities.There are substantial literatures describing themethods for modelling of vehicular exhausts in the

.

www.elsevier.com/locate/atmosenv

ARTICLE IN PRESS

Vehicle exhaust tailpipe

9

1.5m

21m 1.5m

3

5

4

0

8

7

6 1

1.5m

3m

1m

1m

Fig. 1. Configuration of a vehicular exhaust plume and experi-

mental set-up. (Points 0–8 are the CO sampling locations, and

point 9 is the meteorological and background CO sampling

location).

J.S. Wang et al. / Atmospheric Environment 40 (2006) 484–497 485

atmosphere (Sharma and Khare, 2001), modellingair quality in street canyons (Vardoulakis et al.,2003) and showing the comparison and evaluationof several mobile-source and line-source models(Marmur and Mamane, 2003). Among these differ-ent-scale environments, the dispersion process fromthe vehicular exhaust plume has a direct impact onhuman health, particularly on the vehicle driversand passengers, bicyclists, motorcyclists, pedes-trians and people working nearby in roadwaymicroenvironments. As the ambient wind directionand velocity are stochastic and uncontrollable in thereal atmospheric environment, the physics involvedin the pollutant concentration dispersion of avehicular exhaust plume in the vicinity of roadwaysbecomes more complex in nature. Although manyempirical models (e.g. the Gaussian and K theorydispersion models) have been widely used to assessthe pollutant concentrations of urban road environ-ments due to traffic vehicles, they can only providethe gross averaged predictions. Flow structures suchas the eddy variation or turbulence characteristicsare difficult to be revealed.

In order to reproduce the qualitative features ofairflow and pollutant concentrations from a vehi-cular exhaust plume near field region, limited two-or three-dimensional Reynolds-averaged Navier–Stokes (RANS) equations and pollutant transportequations including the turbulence models havebeen proposed. Fraigneau et al. (1995) simulated atwo-dimensional pollutant (nitrogen oxides, NOx)dispersion process near the motorway which wascoupled with the reaction of nitric oxides andambient ozone. The interaction between the turbu-lence and chemical reaction was calculated withinthe local scale of 200m in the vicinity of themotorway. A two-dimensional pollutant dispersionnumerical model was developed based on the joint-scalar probability density function (PDF) approachcoupled with a k– e turbulence model to simulate theinitial dispersion process of NOx, temperature andflow velocity distributions from a vehicular exhaustplume (Chan et al., 2001). A three-dimensionalnumerical model coupled with a k– e turbulentclosure was developed for simulating the dispersionof carbon dioxide, CO2, from a heavy-duty dieseltruck exhaust plume (Kim et al., 2001). Theprevious studies used a fixed/main dominant winddirection in their numerical calculations. In thepresent study, it is intended to apply the numericalmodels to the near-field region of initial dispersionbehaviour of the pollutant plume emitted from a

vehicular exhaust tailpipe. This type of pollutantdispersion not only has a direct impact on humanhealth, but also constitutes a major fraction of thetotal pollutant dispersion (Venkatram et al., 1999).

Hence, the aim of the present study is to developa three-dimensional numerical model based on theRANS equations coupled with a k– e turbulencemodel using the commercial FLUENT code and awind-direction–frequency-weighted (WDFW) ap-proach to simulate the initial dispersion process ofCO distributions from a vehicular exhaust plume.This numerical model will then be validated bycomparing the calculated CO concentration withCO data measured in situations that represent thevehicle behaviour in the crowded roads of mostAsian cities. The simulated situations includevehicles at both low idle (when vehicles stop dueto the congested road traffic or at red traffic light)and high idle (when vehicles start moving inresponse to the change of traffic light from red togreen) conditions.

2. Experimental description

To validate the developed numerical model, twodiesel-fuelled vehicles, namely, a taxi (vehicle A) anda light-duty van (vehicle B), were used in the presentstudy. The experimental set-up is shown in Fig. 1 inan open environmental area. In the experiment, theselected test vehicles were not moving but theirengines were running at low and high idle condi-tions in order to simulate the situations of vehiclesstopping at the red traffic light in a low idlecondition during the congested traffic, and in ahigh idle condition during the change from red togreen traffic light. Location 0 represents the exitconditions of a vehicular exhaust tailpipe. Nine

ARTICLE IN PRESSJ.S. Wang et al. / Atmospheric Environment 40 (2006) 484–497486

sampling points (1–9) were chosen along thevehicular exhaust plume with the x-, y- and z-axesrepresenting the streamwise, lateral and transversaldirections, respectively, as shown in Fig. 1. Theconditions of the vehicular exhaust tailpipe exit aretaken as one of the inlet boundaries in the wholecomputational domain. The locations 1, 3, 5, 7 and9 approximately simulated the breathing level ofhuman body at 1.5m above the ground level. Thelocations 2, 4, 6 and 8 were at the same height alongthe centreline of a vehicular exhaust plume instreamwise and lateral directions. Location 9represents the meteorological and ambient back-ground conditions, which is far from the pollutantsof the vehicle exhaust plume. Measured values atthe vehicular exhaust tailpipe exit and ambientconditions are given in Table 1. In the present study,the CO pollutant was chosen for studying thepollutant concentration dispersion process from avehicular exhaust plume because its chemicaltransformation is relatively slow in the atmosphere.The corresponding concentrations of CO werecollected using sampling bags (points 1–9) andanalysed using the gas filter correction CO analyser(Model 48, Thermo Environmental InstrumentsInc., USA). The concentrations of CO at location0 were analysed using the CO analyser (Model 300NDIR, California Analyser Instruments Inc.,USA). The temperatures at location 0 were mea-sured using the thermocouple data logger (ModelTC-08, Pico Technology Company, UK). Thevelocities at location 0 were measured using theanemometer (Model AM5000, Airflow Develop-ment Ltd., UK). The local ambient weather condi-tions were monitored using the combined windvelocity and direction sensors, temperature andhumidity sensors and the data acquisition device(Models DNA022, DMA570 and BSA020.E, LSISpa, Italy) close to location 9. In the present study,the sampling data of local ambient wind velocityand direction, temperature and relative humidity

Table 1

Measured values at the vehicular exhaust tailpipe exit and ambient con

Diesel

vehicle

Tailpipe

diameter

(m)

Tailpipe

height

(m)

Idle

condition

Exhaust exit

velocity Ve

(m s�1)

A 0.048 0.32 Low 8.4

High 22.7

B 0.054 0.37 Low 15.2

High 24.5

were taken at every 10 s, 1min and 10min intervals,respectively. The total sampling period was 4minfor both low and high idle modes.

3. Mathematical model description

3.1. WDFW approach

The local ambient wind directions and velocitiesare complex in the real atmospheric environment.The general approach is either to use the windtunnel experiment to simulate the ambient windflowpattern or to assume the ambient windflow in afixed or a specified direction, which cannot trulyreflect the heterogeneous windflow characteristics(Kim et al., 2001; Chan et al., 2002a). However, it isvital that the pollutant dispersion along with thedevelopment of a vehicular exhaust plume on near-field region and its response to the local ambientflow field be known for representative windvelocities and directions for such short time anddistance scales. The vehicle-based pollution has notonly a direct impact on human health, but alsoconstitutes a major fraction of the total pollutantdispersion (Venkatram et al., 1999). Hence, atransformation approach of real local ambient winddirection is used in the present study. In general, theambient wind velocity could be transformed into 8or 16 directions for most empirical Gaussian modelsthat are used to predict the variation of long-term(i.e., month or year) averaged pollutant concentra-tion along the downwind distance according to themeteorological wind velocities and directions of thewind rose diagram. Niemeier and Schlunzen (1993)used a three-dimensional mesoscale model to studythe flow patterns around the island for eightdifferent geostrophic wind directions.

However, the interested computational domainin the present study for solving the relativeconcentration of CO distributions from the vehi-cular exhaust plume along the streamwise, x, lateral,

ditions

Exhaust

temperature Te

(K)

Ambient

temperature Ta

(K)

Exhaust CO

concentration Ce

(ppm)

355.1 304.0 280

488.5 302.0 385

341.4 303.0 119

548.0 303.4 435

ARTICLE IN PRESSJ.S. Wang et al. / Atmospheric Environment 40 (2006) 484–497 487

y, and transversal, z, directions were only within 10,8 and 4m, respectively. Hence, a WDFW approachwas used to obtain the real pollutant concentrationdispersion along with the development of a vehi-cular exhaust plume. In order to optimise thecomputational costs in terms of time and storagedata, and the large amount of input boundary datafor simulating the three-dimensional vehicularpollutant dispersion problem, the local arbitrarywind magnitude and direction within the specifiedsampling period were transformed into the fourmajor corresponding directions (i.e., N, E, S or W)and was input as one of the corresponding inflowboundary conditions in the present study. In Fig. 2,the magnitude and direction of wind, uP, withrespect to an arbitrary wind angle, a, for 0oaop=2can then be transformed into

uS ¼ uP cos a and uW ¼ uP sin a, (1)

where uP is the measured wind velocity in directionOP, and uS and uW are the transformations of windvelocity instantaneously from uP into the directionsS and W, respectively. Similarly, the magnitude anddirection of wind with respect to an arbitrary windangle for p=2papp, poao3p=2 and 3p=2pap2pcan also be transformed into its corresponding N, E,S or W direction.

The frequency of the transformed wind directionis then defined as

Fi ¼NiP

i

PjNij

, (2)

where Fi is the frequency of wind direction in the ithdirection coordinate, i ¼ 1, 2, 3 and 4 represent theN, E, S and W directions, respectively, Ni is thenumber of times corresponding to the ith directioncoordinate, and Nij is the number of times in respectto the corresponding ith direction coordinate for thespecified sampling period, j.

E

P α

S

W

N

O

Fig. 2. Transformation of wind direction.

The averaged ambient wind velocity is defined as

Ui ¼

Piui

Ni

, (3)

where Ui is the averaged ambient wind velocity inthe corresponding ith direction coordinate and ui isthe instantaneous ambient wind velocity in thecorresponding ith direction coordinate. Once theambient wind velocities and directions during thespecified sampling period are transformed andsummed up, the frequency of ambient wind direc-tions and the averaged magnitude of ambient windvelocities can be calculated. It is obvious that thetotal frequency of wind in the four directions duringthe specified sampling period should be 100%.Table 2 shows the frequency of correspondingambient wind directions and the averaged magni-tude of ambient wind velocity during the specifiedsampling period at the height of 1.5m above theground for vehicles A and B at low and high idleconditions. Based on the WDFW approach for thereal atmospheric environment, the averaged pollu-tant concentration during the specified samplingperiod can then be defined as

C ¼X

i

F iCi, (4)

where Fi and Ci are the obtained frequencies ofambient wind direction and the simulated pollutantconcentration within the computational domainunder the averaged ambient wind velocity in thecorresponding N, E, S or W directions, respectively.

3.2. Governing equations and numerical solvers

In general, vehicular pollutants are emitted froman exhaust tailpipe at a certain flow rate, then theymix with air and disperse along three directions intothe atmosphere as shown in Fig. 1. Considering thethree-dimensional steady flow problem in Fig. 1, thegoverning equations are described as follows:

Continuity equation:

qrui

qxi

¼ 0. (5)

Momentum equation:

qui

qtþ uj

qrui

qxj

¼ �qp

qxi

þqqxj

mquj

qxi

þqui

qxj

� �� ru0iu

0j

� �

þ giðr� r0Þ, ð6Þ

ARTICLE IN PRESS

Table 2

Corresponding boundary conditions of the vehicular exhaust pollutant dispersion model

Parameter Inflow Inlet Outflow Symmetry Ground

Velocity Specified profile: Uz Constant: Ve Zero gradient Zero gradient No slip condition

Temperature Constant: Ta Constant: Te Zero gradient Zero gradient Zero gradient

Concentration Constant: Cb Constant: Ce Zero gradient Zero gradient Zero gradient

J.S. Wang et al. / Atmospheric Environment 40 (2006) 484–497488

where

�rju0iu0j ¼ mt

qui

qxj

þquj

qxi

� ��

2

3rkdij . (7)

k and e transport equations in the standard k– eturbulence model:

qk

qtþ

qðrkuiÞ

qxi

¼qqxj

mþmt

sk

� �qk

qxj

� �þ Gk þ Gb � re, ð8Þ

qeqtþ

qðreuiÞ

qxi

¼qqxj

mþmt

se

� �qeqxj

� �

þ C1eekðGk þ C3eGbÞ � rC2e

e2

k, ð9Þ

where

Gb ¼ bgi

mt

Prt

qT

qxi

, (10)

b ¼ �1

rqrqT

� �p

, (11)

Gk ¼ mt

quj

qxi

qui

qxj

þquj

qxi

� �. (12)

The species (pollutants) transport and energyequations

qrh

qtþ

qruih

qxi

¼qqxi

ki þmt

Prt

� �qT

qxi

� �

�qqxi

Xj

ðhjJiÞ þDp

Dtþ tik

qui

qxk

, ð13Þ

qrCi

qtþ

qruiCi

qxi

¼qqxi

rDi;m þmt

Sct

� �qCi

qxi

, (14)

where �u0iu0j is the Reynolds stress, r is the fluid

density, ui and uj are the mean velocity componentsin ith and jth direction coordinates, respectively, p is

the pressure, mt ¼ rCmk2=e is the turbulent viscosity,m is the laminar viscosity, gi is the gravitationalacceleration in the ith direction coordinate,(0,0,�g), sk and se are the turbulent Prandtlnumbers for k and e, respectively, Gk is thegeneration of turbulence kinetic energy due to themean velocity gradients, Gb is the generation ofturbulence kinetic energy due to buoyancy, b is thecoefficient of thermal expansion, empirical con-stants of C1e, C2e, C3e, Cm, sk and se for the standardk– e turbulence model are 1.44, 1.92, 1.44, 0.09, 1.0and 1.3, respectively, Ci is the mean chemical species(pollutants) concentration, T is the temperature, Prtis the turbulent Prandtl number, 0.85, Sct is theturbulent Schmidt number, 0.7, h is the staticenthalpy, ki is the molecular conductivity, Ji is thediffusion flux of the ith species due to theconcentration gradients, tik is the deviatoric stresstensor, and Di,m is the diffusion coefficient for ithspecies in the mixture. Based on the presentexperimental set-up in an open environmental area,a horizontal homogeneous meteorology was used.Vehicle movement, the vehicle- and building-in-duced turbulence were not taken into account in thenumerical simulation.

Fig. 3 shows the three-dimensional computa-tional domain behind a vehicular exhaust tailpipe.The simulated physical domain is 10m long, 8mwide and 4m high. Fine meshes were used near theexit of the vehicular exhaust tailpipe. Extensive testsof the independence of the meshes were carried outwith increasing mesh numbers until further refine-ment is shown to be less significant. Mixedhexahedral and tetrahedral meshes used for thepresent calculation were about 250,000 in order tocover the computational domain. The computa-tional domain for the vehicular exhaust pollutant onnear-field dispersion is shown in Fig. 3. One of theinflow boundary conditions is the local ambientwind with specified velocity and direction andanother inlet boundary condition is the vehicularexhaust jet plume from the tailpipe exit. The


symmetrical boundaries and the solid boundary arethe top and both sides of the computational domain(i.e., parallel to the windflow direction) and theground face, respectively. The local arbitrary windmagnitude and direction within the specified sam-pling period were transformed into the four majorcorresponding directions (i.e., N, E, S or W). Thecorresponding boundary conditions of a vehicularexhaust pollutant dispersion model are shown inFig. 4 and listed in Table 2.

A power law for the profile of wind velocity wasused to describe the fully developed vertical profileof the horizontal wind velocity under a neutral

Fig. 4. Boundary conditions of a vehicular exhaust pollutant

0

2

4

6

802

46

810

2

4

Vehicle exhaustplume inlet

z

x (E)

y (N)

Fig. 3. Geometry and mesh of computational domain.

Table 3

Transformation of ambient wind directions and velocities for vehicles

Wind direction Low idle condition

Vehicle A Vehicle B

Fi Ui (m s�1) Fi Ui (m

N 0.250 0.871 0.250 0.900

E 0.458 1.017 0.354 1.236

S 0.250 0.665 0.250 0.876

W 0.042 0.426 0.146 0.917

stability condition as follows:

Uz

U ref¼

z

zref

� �p

, (15)

where Uz and Uref are the mean wind velocities atthe heights of z and zref, respectively, z and zref arethe specified and reference (1.5m) heights, respec-tively, and p is the vertical wind profile power index,0.25. In the present study, Uref is equal to Ui atzref ¼ 1.5m, as shown in Table 3.

The initial conditions were given according to theambient and vehicle exhaust jet conditions for thepresent computational study. All calculations wereperformed using second-order accurate upwindschemes. The well-known SIMPLE algorithm wasused for solving the applied numerical model. Thesegregated, implicit and steady solvers were set toobtain the numerical results based on the simulta-neous under-relaxation factors, convergence criter-ion and iteration numbers. All these settings andcalculations were carried out by the operationenvironment of commercial FLUENT code exceptthe final averaged pollutant concentration fromEq. (4).

The relative concentration (RC) of carbon mon-oxide is defined (Kim et al., 2001) as follows:

RC ¼Ck � Cb

Ce � Cb, (16)

where Ck is the simulated concentration at thespecified point k.

dispersion model for different ambient wind directions.

A and B under low and high idle conditions

High idle condition

Vehicle A Vehicle B

s�1) Fi Ui (m s�1) Fi Ui (m s�1)

0.111 0.630 0.333 0.886

0.167 1.479 0.139 0.860

0.389 0.844 0.167 1.085

0.333 1.057 0.361 1.053


4. Results and discussion

4.1. Calculated results for different ambient windflow

conditions using the WDFW approach

Fig. 5 shows the calculated relative CO concentra-tion, RC, distributions of a vehicular exhaust plumeat y ¼ 4m along the streamwise, x, and transversal,z, directions for vehicle A under the low idlecondition using different ambient wind directionsand the WDFW approach. It shows clearly that thepollutant concentration dispersion process in theatmosphere from a vehicular exhaust plume alongthe streamwise distance is highly dependent on thecorresponding frequencies of wind directions andmagnitudes of wind velocities as shown in Table 3.When the ambient wind flows from direction E (i.e.,an opposite flow direction of the vehicular exhaustplume), the exhaust plume is dispersed closer to thevehicle tailpipe exit along the streamwise distancedue to the reversed flow resistance, but is dispersedfurther in the transversal distance, as shown inFig. 5(a). On the other hand, the vehicular exhaustplume is dispersed further along the streamwisedistance, but is dispersed closer to the vehicle tailpipeexit in the transversal distance when the ambientwind flows from direction W (i.e., the same flowdirection of the vehicular exhaust plume), as shownin Fig. 5(c). In Table 3, the frequencies of ambientwindflows from directions S and N (i.e., the flowdirection perpendicular to the exit flow of vehicularexhaust plume) are the same (Fi ¼ 0.250) and themagnitude of wind velocities is only slightly different(in S direction, Ui ¼ 0.871m s�1, and in N direction,Ui ¼ 0.665m s�1). Hence, the magnitude of pollutantconcentration distributions of CO should be quitesimilar, as shown in Figs. 5(b) and (d). Fig. 5(e)shows that the relative concentration distributions ofCO along the streamwise and transversal distances ofa vehicular exhaust plume using the WDFWapproach is similar in shape to Fig. 5(a). The relativeconcentrations show the influence of the combinedeffect and weight of the mean wind velocities andfrequencies in all the considered directions. This isbecause the frequency of ambient windflows in Edirection, Fi ¼ 0.458, is 10 times more than thefrequency of ambient windflows from direction W,Fi ¼ 0.042, and the averaged ambient wind velocityfrom direction E, Ui ¼ 1.017m s�1, is also relativelyhigher than the averaged ambient wind velocity fromdirection W, Ui ¼ 0.426m s�1. The values fromdirections S and N have already been mentioned.

Fig. 6 shows the calculated relative CO concen-tration, RC, distributions of a vehicular exhaustplume at z ¼ 0.32m along the streamwise, x, andlateral, y, directions for vehicle A under the low idlecondition using different ambient wind directionsand the WDFW approach. It also shows clearly theimpact of the frequencies of wind directions andmagnitudes of wind velocities on the pollutantconcentration dispersion process in the atmospherefrom a vehicular exhaust plume along the lateraldistance. If the ambient windflows are dominantfrom direction E, the vehicular exhaust plume isdispersed closer to the vehicle tailpipe exit along thestreamwise distance due to the reversed flowresistance, but is dispersed further along the lateraldistance, as shown in Fig. 6(a) and vice versa inFig. 6(c). In Figs. 6(b) and (d), the vehicular exhaustplumes are dispersed further along the lateraldistance in the opposite way if the correspondingambient wind flows from direction S or N. Again, asimilar vehicular pollutant concentration dispersionpattern along the lateral distance is observedbecause the frequencies of ambient windflows fromdirections S and N are equal, as shown in Table 3.Fig. 6(e) shows the relative CO concentrationdistributions of a vehicular exhaust plume alongthe streamwise and lateral distances using theWDFW approach. The vehicular exhaust plume isdispersed closer along the streamwise distance, but isdispersed almost equally along the lateral distance.

Figs. 7–10 show the calculated relative COconcentration, RC, distributions of a vehicular exhaustplume at z ¼ 0.32 and 0.37m along the streamwise, x,and lateral, y, directions for vehicles A and B underlow and high idle conditions using the WDFWapproach. It is observed again that the pollutantconcentration dispersion process in the atmospheredepends on the frequencies of wind directions andmagnitudes of wind velocities from a vehicularexhaust plume along the streamwise, x, and lateral,y, distances. From Fig. 7, the vehicular exhaust plumeis dispersed closer along the streamwise distance, butis dispersed almost equally along the lateral distance.In Fig. 8, the dispersion of the vehicular exhaustplume from vehicle A under a high idle condition isenhanced along the streamwise distance at whichRC ¼ 0.02 (or 2%) at x ¼ 2.2m, but is slightly towardthe directions N and E along the lateral distance.This is because the frequencies of ambient windflow,Fi ¼ 0.389, from direction S and Fi ¼ 0.333 fromdirection W are much higher than the frequencies ofambient windflows from direction N, Fi ¼ 0.111,

ARTICLE IN PRESS

12345791114

1234568101

x (m)

z (m

)

0 0.25 0.5 0.75 1 1.25

x (m)

0 0.25 0.5 0.75 1 1.25

x (m)

0 0.25 0.5 0.75 1 1.25

0 0.25 0.5 0.75 1 1.250

0.2

0.4

0.6

0.8

z (m

)

0

0.2

0.4

0.6

0.8

z (m

)

0

0.2

0.4

0.6

0.8

z (m

)

0

0.2

0.4

0.6

0.8

z (m

)

0

0.2

0.4

0.6

0.8

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

123581114

x (m)

0 0.25 0.5 0.75 1 1.25

x (m)

1235791114

45679144

(a)

(c)

(b)

(d)

(e)

1

23

32

1

Fig. 5. Relative concentration, RC, of CO distributions on x– z plane along the streamwise distance of a vehicular exhaust plume at

y ¼ 4m for vehicle A under a low idle and different ambient windflow conditions: (a) E direction, (b) S direction, (c) W direction, (d) N

direction and (e) WDFW approach used.


and from direction E, Fi ¼ 0.167, and the averagedambient wind velocities from direction S, Ui ¼

0.844m s�1, and Ui ¼ 1.479m s�1, from direction Eare also slightly higher than the averaged ambientwind velocities, Ui ¼ 0.630m s�1, from direction N

and Ui ¼ 1.057m s�1 from direction W, as shown inTable 3. On the other hand, the dispersion processof a vehicular exhaust plume for vehicle A at a lowidle condition is reduced along the streamwisedistance at which RC ¼ 0:02 (or 2%) at x ¼ 1.1m,

ARTICLE IN PRESS

1

234567913

x (m)

y (m

)

0 0.25 0.5 0.75 1 1.253.6

3.8

4

4.2

4.4

15 0.95014 0.88313 0.81612 0.74911 0.68110 0.6149 0.5478 0.4807 0.4136 0.3465 0.2794 0.2113 0.1442 0.0771 0.010

Rc

0.9500.8830.8160.7490.6810.6140.5470.4800.4130.3460.2790.2110.1440.0770.010

Rc0.9500.8830.8160.7490.6810.6140.5470.4800.4130.3460.2790.2110.1440.0770.010

Rc

0.9500.8830.8160.7490.6810.6140.5470.4800.4130.3460.2790.2110.1440.0770.010

Rc

32

1

2

1

456781013

y (m

)

3.6

3.8

4

4.2

4.4

151413121110987654321

2

3

45714

1

x (m)

y (m

)

0 0.25 0.5 0.75 1 1.253.8

4

4.2

4.4

4.6

4.8

5

151413121110987654321

1234681013

x (m)

y (m

)

0 0.25 0.5 0.75 1 1.253.4

3.6

3.8

4

4.2

4.4

4.6

151413121110987654321

2

3

45

111

x (m)

y (m

)

0 0.25 0.5 0.75 1 1.25x (m)

0 0.25 0.5 0.75 1 1.253.2

3.4

3.6

3.8

4

4.2

151413121110987654321

0.9500.8830.8160.7490.6810.6140.5470.4800.4130.3460.2790.2110.1440.0770.010

Rc

(a) (b)

(c) (d)

(e)

Fig. 6. Relative concentration, RC, of CO distributions on x– y plane along the streamwise distance of a vehicular exhaust plume at

z ¼ 0.32m for vehicle A under a low idle and different ambient windflow conditions: (a) E direction, (b) S direction, (c) W direction, (d) N

direction and (e) WDFW approach used.


ARTICLE IN PRESS

1234610

x (m)

y (m

)

0 0.5 1 15 2 2.5 33

3.5

4

4.5

5

15 0.95014 0.88413 0.81712 0.75111 0.68410 0.6189 0.5518 0.4857 0.4196 0.3525 0.2864 0.2193 0.1532 0.0861 0.020

Rc

813

Fig. 7. Relative concentration, RC, of CO distributions on x– y

plane along the streamwise distance of a vehicular exhaust plume

at z ¼ 0.32m for vehicle A under a low idle condition using the

WDFW approach.

12

35

6812

x (m)

y (m

)

0 0.5 1 1.5 2 2.5 33

3.5

4

4.5

5

15 0.95014 0.88413 0.81712 0.75111 0.68410 0.6189 0.5518 0.4857 0.4196 0.3525 0.2864 0.2193 0.1532 0.0861 0.020

Rc



at z ¼ 0.32m for vehicle A under a high idle condition using the

WDFW approach.

1

2

3456812

x (m)

y (m

)

0 0.5 1 1.5 2 2.5 33

3.5

4

4.5

50.9500.8840.8170.7510.6840.6180.5510.4850.4190.3520.2860.2190.1530.0860.020

Rc151413121110987654321



at z ¼ 0.37m for vehicle B under a low idle condition using the

WDFW approach.

1234

7

11

x (m)

y (m

)

0 0.5 1 1.5 2 2.5 33

3.5

4

4.5

5

15 0.95014 0.88413 0.81712 0.75111 0.68410 0.6189 0.5518 0.4857 0.4196 0.3525 0.2864 0.2193 0.1532 0.0861 0.020

Rc



at z ¼ 0.37m for vehicle B under a high idle condition using the

WDFW approach.


but is equally distributed to both directions S and Nalong the lateral distance, as shown in Fig. 7. This isdue to different wind velocities and directionsoccurring under this sampling period. Similarly,the dispersion of the vehicular exhaust plume fromvehicle B at a low idle condition is enhanced alongthe streamwise distance at which RC ¼ 0:02 (or 2%)at x ¼ 2.2m, but is almost equally distributed toboth S and N directions along the lateral distance,as shown in Fig. 9. This is because the frequency ofambient windflows from direction E, Fi ¼ 0.458, is10 times greater than Fi ¼ 0.042 from direction W,and the averaged ambient wind velocity from

direction E, Ui ¼ 1.017m s�1, is 2 times higher thanthe averaged ambient wind velocity from directionW, Ui ¼ 0.426m s�1. In addition, the frequency ofambient windflow conditions from directions Sand N are equal to Fi ¼ 0.25, and the averagedambient wind velocity from direction S, Ui ¼

0.665m s�1, is only slightly lower than the averagedambient wind velocity from direction N, Ui ¼

0.871m s�1. On the other hand, the exhaust plumeof vehicle B at a high idle condition is dispersedrapidly along the streamwise distance at whichRC ¼ 0.02 (or 2%) at x ¼ 2.6m, but is slightlytoward the directions S and E along the lateral


distance, as shown in Fig. 10. The details ofindividual frequencies of ambient windflow direc-tion and velocities are presented in Table 3.

From these calculated results, it is also found thatthe effect of vehicular exhaust exit velocity on thedispersion process is more significant than that ofambient wind velocity in the vicinity of the exhausttailpipe exit, while the effect of ambient windvelocity gradually shows a significant role in thedispersion process in the periphery of the vehicularexhaust plume. The present study shows that thepollutant concentration dispersion process is lessinfluenced by the ambient wind velocity in the nearregion for RC ¼ 0.1–1 (or 10–100%), than by thevehicular exhaust exit velocity, but it is vice versa inthe far region for RC ¼ 0–0.1 (or 0–10%).

4.2. Effect on pollutant concentration dispersion

using the WDFW approach

To validate the developed three-dimensionalnumerical model coupled with a k– e turbulencemodel and the WDFW approach, two diesel-fuelledvehicles, A and B, were used to obtain the COconcentration distributions of a vehicular exhaustplume. Figs. 11–18 show the comparison betweenthe measured and calculated relative concentrationsof CO of a vehicular exhaust plume in x– z and y– z

planes for both vehicles. The results show a good

Downstream distance alongexhaust plume (m)

Rel

ativ

e co

ncen

trat

ion

0 2 4 6 8 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Calculated (low idle)Calculated (high idle)Measured (low idle)Measured (high idle)

Fig. 11. Comparison of relative concentration, RC, between the

measured and calculated CO data (at points 0, 2 and 8) of a

vehicular exhaust plume at z ¼ 0.32m for vehicle A under high

and low idle conditions using the WDFW approach.

prediction using the present three-dimensionalnumerical model coupled with the WDFW ap-proach. In Figs. 11 and 15, the relative concentra-tions of CO decrease exponentially along thecentreline of the vehicular exhaust plume fordifferent idle conditions due to the initial dispersionprocess. The relative pollutant concentrations re-duce rapidly close to ambient concentration valuesin a short streamwise distance of 4m. In Figs. 12and 16, the relative concentrations of CO at 1.5mabove ground level mix rapidly with the surround-ing ambient air along the lateral distance. Therelative pollutant concentration at location 7 isfound to be slightly higher than at location 1. Itimplies that the combined effects of entrainmentand buoyancy of the vehicular exhaust thermalplume have taken place with the surroundingambient air. Details on heat transfer and flowcharacteristics of a thermal jet can be found in Chanet al. (2002b). In Figs. 13 and 17, the relativeconcentrations of CO at 1.5m above grounddecrease with increasing lateral distances fromlocation 2 for different idle conditions. The sym-metric relative pollutant concentration distributionsfor a low idle condition in both vehicles are due tosimilarities between the frequencies of wind in the Nand S directions and their velocities as shown inTable 3. On the other hand, the less symmetricrelative pollutant concentration distributions for a

Downstream distance alongexhaust plume (m)

Rel

ativ

e co

ncen

trat

ion

0 2 4 6 8 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


0 2 4 6 8 100

0.005

0.01

Fig. 12. Comparison of relative concentration, RC, of CO

between the measured and calculated CO data (at points 1 and

7) of a vehicular exhaust plume at z ¼ 1.5m for vehicle A under

high and low idle conditions using the WDFW approach.

ARTICLE IN PRESS

Transverse distance alongexhaust plume (m)

Rel

ativ

e co

ncen

trat

ion

0 2 4 6 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1



between the measured and calculated data (at points 2, 4 and




Rel

ativ

e co

ncen

trat

ion

0 2 4 6 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


0 2 4 6 8

0

0.005

0.01





Downstream distancealong exhaust plume (m)

Rel

ativ

e co

ncen

trat

ion

0 2 4 6 8 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1




8) of a vehicular exhaust plume at z ¼ 0.37m for vehicle B under



Rel

ativ

e co

ncen

trat

ion

0 2 4 6 8 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


0 2 4 6 8 100

0.005

0.01


between the measured and calculated data (at points 1 and 7)

of a vehicular exhaust plume at z ¼ 1.5m for vehicle B under high

and low idle conditions using the WDFW approach.


high idle condition in both vehicles result from thefact that the frequency of wind in N and S directionsis quite different and their wind velocities are alsoslightly different from each other. In Figs. 8 and 10,the effects of wind direction and velocity on thevehicular pollutant concentration dispersion are in

N and S directions, respectively. Here, more realisticqualitative features of airflow and pollutant con-centrations of the vehicular exhaust plume in near-field scale can be reproduced using the developedthree-dimensional numerical model coupled with ak– e turbulence model and the WDFW approach. In

ARTICLE IN PRESS


Rel

ativ

e co

ncen

trat

ion

0 2 4 6 8

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Calculated (low idle)Calcualed (high idle)Measured (low idle)Measured (high idle)






Rel

ativ

e co

ncen

trat

ion

0 1 2 3 4 5 6 7 8

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


0 1 2 3 4 5 6 7 8

0

0.005

0.01






Figs. 14 and 18, the relative concentrations of COabove the ground level of 1.5m decrease withincreasing lateral distance from the exhaust tailpipeexit at location 1 for different idle conditions.

In general, good agreement between the calcu-lated and measured data for two diesel-fuelled

vehicles demonstrates that the developed three-dimensional vehicular pollutant dispersion numer-ical model based on the RANS equations coupledwith a k– e turbulence model and the WDFWapproach provides a good insight into the initialdispersion of pollutant concentration from a vehi-cular exhaust plume in the real atmosphericenvironment. The larger differences between themeasured and calculated data are observed atlocation(s) 2 and/or 4 for both vehicles A and B ata low idle condition. It may be due to the effects ofcomplex local ambient wind field in the realatmosphere, and the wake generation from theinteraction between the vehicle geometry, andarbitrary ambient wind magnitude and directionon near-field vehicular exhaust plume dispersion.

5. Summary and conclusions

A three-dimensional vehicular pollutant dis-persion numerical model was developed basedon the Reynolds-averaged Navier–Stokes equationscoupled with a k– e turbulence model to simulate theinitial pollutant dispersion process of carbon mon-oxide, CO, from a vehicular exhaust plume in thereal atmospheric environment. Since ambient winddirection and velocity are stochastic and uncontrol-lable in the real atmospheric environment, a wind-direction–frequency-weighted (WDFW) approachwas used to obtain the real pollutant concentrationdispersion along with the development of vehicularexhaust plume. Within the specified samplingperiod, the local ambient windflow conditions aretransformed into the corresponding frequencies ofwind directions and averaged magnitude of windvelocities from directions N, E, S or W. Based onthe present experimental set-up in an open environ-mental area, the horizontal homogeneous meteor-ology, vehicle movement, vehicle- and building-induced turbulence have not been taken intoaccount in the numerical calculation. Good agree-ment between the calculated and measured data fortwo diesel-fuelled vehicles indicates that the WDFWapproach can truly reflect the initial dispersion ofpollutant concentration from a vehicular exhaustplume in the real atmospheric environment. Thepresent study shows that the dispersion process inthe near region for the relative concentration of CO,from RC ¼ 0.1 (or 10%) to 1 (or 100%) is lessinfluenced by the ambient wind velocity than thevehicular exhaust exit velocity, but it is vice versa inthe far region, from RC ¼ 0 (or 0%) to 0.1 (or


10%). It implies that the effect of vehicular exhaustexit velocity on the dispersion process is morepronounced than the effect of ambient wind velocityat the vicinity of the exhaust tailpipe exit, while theeffect of ambient wind velocity gradually shows asignificant role in the dispersion process along withthe development of the vehicular exhaust plume.

Acknowledgements

This work was supported by grants from theResearch Grants Council of the Hong Kong SpecialAdministrative Region, China (RGC Project no.PolyU 5154/01E) and the Central Research Grantsof The Hong Kong Polytechnic University (Projectno. B-Q497).

References

Chan, T.L., Dong, G., Cheung, C.S., Leung, C.W., Wong, C.P.,

Hung, W.T., 2001. Monte Carlo simulation of nitrogen oxides

dispersion from a vehicular exhaust plume and its sensitivity

studies. Atmospheric Environment 35, 6117–6127.

Chan, T.L., Dong, G., Leung, C.W., Chung, C.S., Hung, W.T.,

2002a. Validation of a two-dimensional pollutant dispersion

model in an isolated street canyon. Atmospheric Environment

36, 861–872.

Chan, T.L., Leung, C.W., Jambunathan, K., Ashforth-Frost, S.,

Zhou, Y., Liu, M.H., 2002b. Heat transfer characteristics

of a slot jet impinging on a semi-circular convex surface.

International Journal of Heat and Mass Transfer 45,

993–1006.

Fraigneau, Y.C., Gonzalez, M., Coppalle, A., 1995. Dispersion

and chemical-reaction of a pollutant near a motorway.

Science of the Total Environment 169, 83–91.

Kim, D.H., Gautam, M., Gera, D., 2001. On the prediction of

concentration variations in a dispersing heavy-duty truck

exhaust plume using k– e turbulence closure. Atmospheric

Environment 35, 5267–5275.

Marmur, A., Mamane, Y., 2003. Comparison and evaluation of

several mobile-source and line source models in Israel.

Transportation Research Part D 8, 249–265.

Niemeier, U., Schlunzen, K.H., 1993. Modelling steep terrain

influences on flow patterns at the Isle of Helgoland. Beitraege

zur Physik der Atmosphare 66, 45–62.

Sharma, P., Khare, M., 2001. Modelling of vehicular exhausts—a

review. Transportation Research Part D 6, 179–198.

Vardoulakis, S., Fisher, B.E.A., Pericleous, K., Gonzalez-Flesca,

N., 2003. Modelling air quality in street canyons: a review.

Atmospheric Environment 37, 155–182.

Venkatram, A., Fitz, D., Bumiller, K., Du, S.M., Boeck, M.,

Ganguly, C., 1999. Using a dispersion model to estimate

emission rates of particulate matter from paved road.

Atmospheric Environment 33, 1093–1102.

three-dimensional pollutant concentration dispersion of a ... · and velocity are stochastic and...

Documents