A modelling system for predicting urban air pollution ...lib.tkk.fi/Diss/2001/isbn9512257599/article4.pdf · time series of concentrations. The modelling system contains a method,

*Corresponding author. Fax: #358-9-1929-5403.E-mail address: ari.karppinen@fmi." (A. Karppinen).

Atmospheric Environment 34 (2000) 3723}3733

A modelling system for predicting urban air pollution:model description and applications in the Helsinki

metropolitan area

A. Karppinen!,*, J. Kukkonen!, T. ElolaK hde", M. Konttinen!, T. Koskentalo",E. Rantakrans!

!Finnish Meteorological Institute, Air Quality Research, Sahaajankatu 20 E, 00810 Helsinki, Finland"Helsinki Metropolitan Area Council, Opastinsilta 6 A, 00520 Helsinki, Finland

Received 24 November 1999; accepted 11 December 1999

Abstract

We have developed a modelling system for evaluating the tra$c volumes, emissions from stationary and vehicularsources, and atmospheric dispersion of pollution in an urban area. The dispersion modelling is based on combinedapplication of the urban dispersion modelling system (UDM-FMI) and the road network dispersion model (CAR-FMI).The system includes also a meteorological pre-processing model and a statistical and graphical analysis of the computedtime series of concentrations. The modelling system contains a method, which allows for the chemical interaction ofpollutants, originating from a large number of urban sources. This paper presents an overview of the modelling systemand its application for estimating the NO

xand NO

2concentrations in the Helsinki metropolitan area in 1993.

A companion paper addresses comparison of model predictions with the results of an urban measurement network. Thismodelling system is an important regulatory assessment tool for the national environmental authorities. ( 2000Elsevier Science Ltd. All rights reserved.

Keywords: Model; Urban; Air pollution; Nitrogen oxides; Chemical transformation

1. Introduction

Urban scale dispersion modelling systems have beendeveloped in many European countries; these have re-cently been reviewed by Moussiopoulos et al. (1996). Thelatest generation of local scale models is used in combina-tion with meteorological pre-processing models, whichare based on scaling theories of the atmospheric bound-ary layer (ABL). Examples of such models are the DanishOML model (Olesen, 1995a), the UK-ADMS system ofthe United Kingdom (Carruthers et al., 1995) and themodels applied in this study, UDM-FMI (Karppinen etal., 1998c) and CAR-FMI (HaK rkoK nen et al., 1995, 1996).

On the other hand, various local scale Gaussian modelsusing the Pasquill (or equivalent) stability classes are stillwidely used in practical applications in many Europeancountries.

The urban scale modelling systems should be able toallow for the various local scale e!ects, for instance, thein#uence of buildings and obstacles, downwash phe-nomena and plume rise, together with chemical trans-formation and deposition (e.g., Kukkonen et al., 1997).The modelling systems also need information concerningthe hourly tra$c volumes, travel speeds and emissionsfrom various urban mobile pollution sources.

This paper describes an integrated urban pollutionmodelling system and discusses predicted concentra-tion distributions of nitrogen oxides in the Helsinkimetropolitan area in 1993. Some selected results havebeen published previously by Karppinen et al. (1997a,Karppinen et al. (1997a, 1998a, b). A companion paper

AEA=2815=Durai=Venkatachala=BG

1352-2310/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.PII: S 1 3 5 2 - 2 3 1 0 ( 0 0 ) 0 0 0 7 4 - 1

voigt

Reprinted from Atmospheric Environment 34 (2000), pp. 3723-3733. Copyright 2000, with permission from Elsevier Science.

Fig. 1. An overview of the modelling system.

addresses the testing of the model against the results ofan air quality monitoring network.

This atmospheric dispersion modelling is based ona combined application of the urban dispersion model-ling system UDM-FMI and the road network dispersionmodel CAR-FMI (contaminants in the air from a Road),both of these models have been developed at the FinnishMeteorological Institute (FMI). Both dispersion modelsinclude a treatment of the chemical transformation ofnitrogen oxides. The dispersion modelling system wasre"ned to take into account the chemical interaction ofpollutants from a large number of individual sources.This novel modelling method allows for the interdepen-dence of urban background NO, NO

2and O

3concentra-

tions, and NO and NO2

emissions from various sources.In a previous study by Valkonen et al. (1995, 1996), the

spatial distribution of concentrations and the statisticalparameters of CO, NO

2and SO

2concentrations were

numerically evaluated in the city of Espoo in southernFinland in 1990, with the UDM-FMI model. The com-bined dispersion modelling system UDM-FMI andCAR-FMI has been applied in air quality assessments,which have been conducted nationally in numerous cities(for instance, Pietarila et al., 1997).

2. The mathematical models

2.1. Comparison of various modelling approaches

An advantage of Gaussian modelling systems is thatthese can treat a large ensemble of emission sources,dispersion situations, and a receptor grid network, whichis su$ciently dense spatially (of the order of tens ofmeters). For instance, an hourly time series of one year isrequired, in order to evaluate statistical concentrationparameters, which have been de"ned in air quality guide-lines (for instance, various percentile values).

Computations of the present study contain approxim-ately 5000 line sources, a couple of hundred stationarysources, an hourly time series of one year (8760 meteoro-logical and emission situations) and about 10 000 recep-tor points. The dispersion equations therefore had tobe solved separately of the order of 5]103]104]104"

5]1011 times. The transformation chemistry computa-tions increased the computational time even further. Thecomputations therefore required 24 h CPU-time ona super-computer Cray C94. The corresponding compu-tations using an Eulerian grid model (for instance,Yamartino et al., 1992; Nikmo et al., 1999) would not benumerically possible using the presently available com-puter resources.

Limitations of Gaussian urban modelling systems in-clude that these are based on so-called quasi-steady-stateassumptions (for instance, Seinfeld and Pandis, 1998). Itis assumed that pollutant concentrations can be treated

as though they resulted from a time sequence of di!erentsteady states (commonly taken as 1 h). However, thisassumption can be invalid particularly during peak con-centration episodes, caused by accumulation of air pollu-tion in an urban area. In contrast, in Eulerian grid orLagrangian models, meteorology, dispersion and chem-istry can be described (at least in principle) in real time,selecting a suitable numerical time step.

Clearly, modelling of chemical transformation and de-position processes is not so straightforward in Gaussianmodelling systems, compared with Eulerian grid modelsor Lagrangian models. However, some solutions havebeen presented (for instance, Benson, 1984, 1992; HaK r-koK nen et al., 1996).

2.2. The integrated modelling system

Fig. 1 shows the overall structure of the modellingsystem applied in this study. The system includes thefollowing models: (i) the estimation of tra$c volumes andtravel speeds with the EMME/2 transportation planningsystem (INRO, 1994), (ii) the computation of emissionsfrom vehicular sources, using the EMME/2 and LIISAsystems (MaK kelaK et al., 1996), (iii) the model for evaluat-ing the emissions from stationary sources, (iv) the me-teorological pre-processing model MPP-FMI, developedat the FMI (Karppinen et al., 1997b, 1998c), (v and vi) thedispersion models for stationary and mobile sources,UDM-FMI and CAR-FMI, and (vii) post-processingmodels, including a statistical and graphical analysis ofthe computed time series of concentrations.

The urban dispersion modelling system UDM-FMItakes into account all stationary sources and the roadnetwork dispersion model CAR-FMI all tra$c sources,respectively. The programs have been executed on theCray C94 supercomputer.

Aea=2815=Durai=VVC=BG

3724 A. Karppinen et al. / Atmospheric Environment 34 (2000) 3723}3733

Table 1A summary of the NO

xemissions in the Helsinki metropolitan

area in 1993

Source Emissions (as NO2)

t a~1 %

Public power generation 13 305 47.7Other point sources and residential

heating systems855 3.0

Road transport 12 090 43.3Aviation 437 1.6Marine tra$c and harbours 1223 4.4

Total 27 910 100

Fig. 2. The location of the cities of Helsinki, Vantaa, Espooand Kauniainen and the meteorological stations of Helsinki}Vantaa, Helsinki}Isosaari and Jokioinen.

2.3. Trazc volumes and emissions

We conducted an extensive emission inventory in theHelsinki metropolitan area in 1993, which included theemissions from various mobile sources (road tra$c, har-bours and marine tra$c, and aviation) and stationarysources (power plants, other point sources and residentialheating). Table 1 presents a summary of the total emis-sions of NO

x. Approximately, one half of the total NO

xemissions originates from stationary sources.

In the spatial concentration distributions presented inthis paper, we have neglected the emissions from marinetra$c and from aviation, as their in#uence is local and ingeneral small (about 6% of all the emissions of NO

xwithin the study area), compared to emissions from roadsand streets. The in#uence of marine tra$c and aviationon the NO

xand NO

2concentrations in the Helsinki

metropolitan area has been discussed in more detail byPesonen et al. (1996).

The emissions from vehicular tra$c were estimated asfollows. First, the so-called tra$c demand matrices(which describe the vehicle trips from every spatial zoneto all the other zones within the area considered) wereformed, based on a comprehensive inquiry in 1988. Thematrices were "rst evaluated for three speci"c times:morning and afternoon tra$c peak hours and an averagenon-peak hour. We then utilised the regression coe$-cients, which were based on tra$c countings performedby the city of Helsinki and the Finnish National RoadAdministration. We formed 24 new matrices: 10 forweekdays, seven for Saturdays and seven for Sundays.For a more detailed description, the reader is referred toElolaK hde and Koskentalo (1996).

The tra$c volumes and average travel speeds of eachtra$c link were computed using EMME/2 transporta-tion planning system (INRO, 1994), which assigns thetrips (from zone to zone) to links of a network model. Themodel allows for the diurnal and daily variations both intra$c volumes and speeds, and tra$c emissions.

The emission factors of cars in city tra$c are based ontra$c cycle measurements in Helsinki. The emission fac-tors of the LIISA system (MaK kelaK et al., 1996) were usedfor road tra$c and for heavy duty vehicles.

The stationary sources considered include energy pro-duction, industry and residential heating systems. Theseare considered as point or area sources. The computa-tions included 5000 line sources, 169 point sources, areasources and the regional background concentrations.

2.4. Atmospheric boundary-layer scaling

The relevant meteorological parameters for the modelsare evaluated using data produced by a meteorologicalpre-processing model (Karppinen et al., 1997b, 1998c).The model is based mainly on the energy budget methodof van Ulden and Holtslag (1985). The model utilisesmeteorological synoptic and sounding observations, andits output consists of estimates of the hourly time series ofthe relevant atmospheric turbulence parameters (theMonin-Obukhov length scale, the friction velocity andthe convective velocity scale) and the boundary layerheight.

We have made use of the meteorological database ofour institute, which contains routine weather and sound-ing observations. Fig. 2 shows the location of the studyarea, the meteorological stations and the backgroundair quality measurement stations. We used a combina-tion of the data from the stations at Helsinki-Vantaaairport (about 15 km north of Helsinki downtown)and Helsinki}Isosaari (an island about 20 km south ofHelsinki). The mixing height of the atmospheric bound-ary layer was evaluated using the meteorological pre-processor, based on the sounding observations atJokioinen (90 km northwest) and the routine meteoro-logical observations.


A. Karppinen et al. / Atmospheric Environment 34 (2000) 3723}3733 3725

2.5. Road network dispersion model (CAR-FMI)

The dispersion from a road network is evaluated withthe Gaussian "nite-line source model CAR-FMI (Con-taminants in the Air from a Road; HaK rkoK nen et al., 1995,1996). The model includes an emission model, a disper-sion model and statistical analysis of the computed timeseries of concentrations.

The dispersion equation is based on an analytic solu-tion of the Gaussian di!usion equation for a "nite linesource (Luhar and Patil, 1989):

C"

Ql

2 J2p pzu sin h Cexp A!

(z!H)2

2p2zB

#exp A!(z#H)2

2p2zBDCerf A

sin h (p!y)!x cos h

J2 py

B#erf A

sin h (p#y)#x cos h

J2 py

BD, (1)

where C is the concentration, Ql

is the source strengthper unit length, u is the average wind speed, h is the anglebetween the wind direction and the road, x, y and z arethe coordinates, H is the e!ective source height, p is thehalf-length of the line source, erf is the error function andpz

and py

are the vertical and lateral dispersion para-meters, respectively. The solution (1) allows for any winddirection with respect to the road.

The dispersion parameters are modelled as function ofthe Monin}Obukhov length, the friction velocity and themixing height (Gryning et al., 1987). These quantities arecomputed by the meteorological pre-processing model(Karppinen et al., 1997b, 1998c). Tra$c-originated tur-bulence is modelled with a semi-empirical treatment(Petersen, 1980).

The model includes the basic reactions of nitrogenoxides, oxygen and ozone:

NO2#hl KR

P NO#O, (2a)

O#O2#MPO

3#M, (2b)

NO#O3

KFP NO

2#O

2, (2c)

where M is a molecule and KF

and KR

are reaction rateconstants, which are functions of ambient temperatureand solar radiation intensity, as presented by Hertel andBerkowicz (1989).

The in#uence of hydrocarbons on the transformationof nitrogen oxides is important in the regional and long-range transport scales. In the urban scale, their in#uenceis less signi"cant, due to short transport times. Theirin#uence in Northern European urban areas may be

substantial in episodic conditions, during prevailingstable atmospheric strati"cation and low wind speed.

The system of Equations (2a)}(c) can be solved analyti-cally (Benson, 1984, 1992). Appendix A presentsa method for including the chemical transformationmodule corresponding to these equations into a Gaus-sian line source dispersion model.

The advantage of using such fairly simple chemicalmodules is that dispersion and chemistry computationsare su$ciently e$cient numerically. For the number ofsources and receptor grid points applied in this study, thenumerical requirements of more complex numericalchemistry schemes would be unreasonable, even for nu-merical program execution on a super computer.

The predictions of the CAR-FMI model have beenpreviously compared with the results of two measure-ment campaigns, conducted near major roads (i) in a sub-urban area in the city of Espoo in 1994 (Walden et al.,1995; HaK rkoK nen et al., 1997) and in a rural area insouthern Finland in 1995. For both of these measure-ment campaigns, the predicted NO

xand NO

2concen-

trations agreed well with the experimental data in mostcases considered; however, there were some speci"c me-teorological and tra$c conditions, in which there wassubstantial disagreement.

2.6. Urban dispersion modelling system (UDM-FMI)

The urban dispersion modelling system (Karppinen etal., 1998c, 1997b) includes a multiple source Gaussianplume model and the meteorological pre-processor. Thedispersion model is an integrated urban-scale model,taking into account all source categories (point, line, areaand volume sources).

For the most general case involving volume sources, itis assumed that these are evenly distributed within thevolume *x *y *z, i.e. Mx, y, z D x 3 [x

1, x

2] ' y 3

[y1, y

2] ' z 3 [z

1, z

2]N. Assuming a total re#ection of

pollutants from the ground, the resulting ground-levelconcentration can be written as (Karppinen et al., 1998c):

C(x, y, 0)"Q

p *x *y *z u

]Py2

y1P

x2

x1P

z2

z1

exp (!y2/ 2p2y(x))

py(x)

]exp(!H2/2p2

z(x))

p2z(x)

dz dx dy, (3)

where Q is the total emission strength of the individualsources. If the sources are of equal height, the volumesource reduces to an area source, and Eq. (3) can bewritten in terms of error functions.

As previously, the dispersion parameters are modelledas function of the Monin}Obukhov length, the friction



Fig. 3. A schematic presentation of the model for evaluating thechemical interaction of pollutant plumes originating from vari-ous urban sources.

velocity and the mixing height (Hanna, 1985; Gryning etal., 1987). These quantities are computed by the meteoro-logical pre-processing model.

The transformation of NO}NO2

is described by thesimple relation (Janssen et al., 1988)

C(NO2)/C(NO

x)"B

1(1!exp (!b

1x)), (4)

where C(i) is the concentration of the species. B1

andb1

are parametrized in terms of local atmospheric condi-tions, ozone concentration, wind speed and season of theyear (Janssen et al., 1988). Eq. (4) has been derived frommeasurements over a period of ten years in the Nether-lands.

The model also includes a treatment of dry and wetdeposition for nitrogen oxides and SO

2, plume rise,

downwash phenomena, the dispersion of inert particlesand the in#uence of a "nite mixing height. The systemcomputes an hourly time series of concentrations andstatistical parameters, which can be directly compared toair quality guidelines.

The model predictions have been compared with thetracer experiments of Kincaid, Copenhagen and Lilles-troK m, presented by Olesen (1995b). The predictions werewell in agreement with the Kincaid data; somewhatlarger di!erences were found for the Copenhagen andLillestroK m data.

2.7. Chemical interaction of pollutants originating fromvarious sources

Both dispersion models applied allow for the chemicaltransformation of nitrogen oxides, by consideringa single plume in a background air with a uniformconstitution. However, the plumes originating from vari-ous sources interact chemically also with each other. Forinstance, it is not uncommon that urban backgroundO

3concentration vanishes, caused by the depletion of

O3

in the oxidation of NO into NO2. However, many

regulatory modelling systems assume for simplicity thatthe urban background O

3concentration is equal to the

regional O3

background concentration.We have therefore developed a modelling system,

which allows for the chemical interdependence of theNO

xconcentrations originating from various sources

and the O3

concentrations. Fig. 3 illustrates the mainprinciples of the modelling system.

(i) First, the modelling system evaluates the time vari-ation of the regional background concentrations.The regional background concentrations are basedon the data from the monitoring station of Luukki,situated in the North-Eastern part of the Helsinkimetropolitan area. In order to "lter out temporallyhigh episodic concentrations, originating from localsources, we computed diurnal hourly average con-

centrations for each month. This procedure produc-ed a matrix of 12 times 24 regional backgroundconcentration values, for each pollutant and eachmeasurement station.

(ii) Secondly, we summed the spatial pollutant distrib-ution from all urban stationary sources to theregional background. This simpli"cation is reas-onable, as most of the emissions of stationary so-urces are released from higher altitudes. Theirchemical transformation therefore mostly takesplace before substantial interaction with pollutantplumes from other urban sources. This procedureproduces a so-called "rst-order spatial backgrounddistribution for each pollutant (NO, NO

2and O

3).

(iii) Thirdly, the system evaluates the contribution onthe background caused by the mobile sources. Theurban mobile sources and the receptor points aresorted out in terms of their location with respect tothe wind direction. The plume from the most up-wind mobile source (a link of road or street) isallowed to interact chemically with the "rst-orderspatial background distribution. This producesa second-order spatial background distribution.Correspondingly, the second most upwind mobilesource is allowed to interact with the second-orderspatial background distribution. All the mobilesources are then treated consecutively.

The system properly takes into account, for instance,the depletion of O

3in the urban area. The in#uence of

the chemical interactions of the sources is largest duringthe rush hours and air quality episodes, in which atmo-spheric di!usion conditions are unfavourable. We havetested this assumption with numerical computations, us-ing the modelling system (i) including the above-men-tioned modelling method, and (ii) excluding it from the



Fig. 4. NOx

emissions from mobile (upper "gure) (lg m~2 s~1)and stationary (lower "gure) (t a~1) sources in the Helsinkimetropolitan area in 1993.

computations. For the Helsinki Metropolitan Area in1993, the in#uence of the chemical plume interactions onthe hourly NO

2concentrations during rush hours ex-

ceeds 2 and 5% at 10 and 2% of the grid points, respec-tively. For O

3concentrations, the relative in#uence

of the chemical plume interactions can be substantiallylarger.

3. The numerical results

Helsinki metropolitan area is situated by the Baltic Seaat the latitude of 603N. The climate is relatively mildercompared with many other areas in the same latitudes,largely because the Gulf Stream and the prevailing globalatmospheric circulation have a warming e!ect. The Hel-sinki metropolitan area comprises four cities: Helsinki,Espoo, Vantaa and Kauniainen. The population of theHelsinki metropolitan area is 850 000 and the area covers743 km2.

The dispersion models applied are Gaussian and thesedo not generally take into account the in#uence of indi-vidual buildings on the atmospheric dispersion (althoughthe UDM-FMI model allows for the in#uence of thesource building itself on the dispersion). However, theterrain in the area is relatively #at and the average heightof the buildings is fairly low (most buildings are lowerthan 15}20 m). There is only a moderate number of streetcanyons in the area. We have used the roughness lengthof 0.7 m in the numerical computations.

We have computed the concentrations of nitrogenoxides (NO

x) and nitrogen dioxide (NO

2) in the Helsinki

metropolitan area for one year, 1993. The concentrationtime series were computed on a receptor grid, whichcontains approximately 10 000 receptor points. The re-ceptor point network covers the whole area, and thelargest grid intervals are 500 m. A more densely spacedgrid was applied in the Helsinki downtown area, the gridinterval being 100 m. In the vicinity of the major roads inthe area, the smallest grid interval was 50 m. The variablereceptor grid is required in order to evaluate isoconcen-tration curves with adequate accuracy from the com-puted data.

The NO2

concentrations in Helsinki are generallycomparable with those in the major Central Europeancities (Jol and Kielland, 1997; Kukkonen et al., 1999).In other Finnish cities the NO

2concentration levels

are usually somewhat lower than those in the capital(Kukkonen et al., 1999).

3.1. The emissions

Figs. 4a}b show the evaluated NOx

emissions of mo-bile and major stationary sources in the Helsinki metro-politan area in 1993. The size of the depicted area is35 km]25 km and the legend in the top left-hand corner

of the "gures shows the absolute values of the pollutantemissions. In Fig. 4b, the height of each rectangle isproportional to the stack height, and its width is logarith-mically proportional to the emission rate. The dottedlines indicate the city borders.

The emissions from mobile sources are mainly depen-dent on tra$c densities and driving speeds. The emis-sions of NO

xincrease continuously with increase in

vehicle travel speed in the speed range from 40 to120 km h~1 (MaK kelaK et al., 1996), caused by the enhancedoxidation of air-originated nitrogen with the increasingcombustion temperature. The largest emission densitiesoccur in the Helsinki city centre area, and along RingRoad 1, situated at a distance of 8}10 km from the citycentre. There are also substantial emissions along themajor roads leading to the Helsinki city centre, and atmajor crossroads.

All the largest stationary sources are coal-"red powerplants, also using heavy fuel oil as an additive fuel. Thestack heights of the three largest power plants are about



Fig. 5. Predicted spatial distribution of the yearly means ofNO

x(upper "gure) and NO

2(lower "gure) concentrations

(lg m~3) in the Helsinki metropolitan area in 1993. The locationof the monitoring stations of YTV (ToK oK loK , Vallila, LeppaK vaaraand Tikkurila) has also been indicated. The size of the depictedarea is 35 km]25 km. The lower "gure shows an outline of theHelsinki downtown area, presented in Fig. 6.

Fig. 6. Predicted spatial distribution of the yearly mean concen-tration of NO

2(lg m~3) in the Helsinki downtown area in 1993.

150 m; the emissions of the largest individual unit(`Hanasaari Ba) are approximately 6200 tonnes NO

x(as

NO2) annually. The contribution of minor residential

heating plants and industrial plants (not shown in Fig.4b) on the total NO

xemissions is only a few per cent.

3.2. The annual average spatial concentration distributions

Figs. 5a}b show the computed annual means of NOx

and NO2

concentrations at the ground level in the Hel-sinki metropolitan area in 1993. The legend in the topleft-hand corner shows the absolute values of the pollu-tant concentration.

Clearly, the tra$c emissions have a larger relativein#uence on the ground-level concentrations, comparedto the stationary emissions, which are mostly released

from higher altitudes. Although the contribution of traf-"c on the total emissions is slightly less than a half,approximately 80}95% of the ground level NO

xconcen-

trations originate from tra$c sources. The modest con-tribution of stationary sources is caused by the strongconcentration of the NO

xemissions to large power

plants, which have substantial stack heights, within thestudy area. The concentrations of NO

xand NO

2are the

highest in the vicinity of the main roads and streets, andin the downtown area of Helsinki. The "gures show thedistinct in#uence of the ring roads (situated at the distan-ces of about 8 and 15 km from the city centre), the majorroads leading to the Helsinki city centre, and the junc-tions of major roads and streets. The concentrations ofNO

2are strongly distributed along major roads, with

higher vehicle travel speeds, due to the increase of theNO

xemissions with the travel speed, and in the down-

town area, caused by the largest tra$c volumes withinthe area.

Fig. 6 illustrates the NO2

concentrations at the groundlevel in the Helsinki downtown area in 1993. Again, thein#uence of the main tra$c routes is clearly visible.



Fig. 7. Predicted spatial distribution of the second highest dailymean NO

2concentration (lg m~3) (upper "gure) and of the

highest 1-h 99-percentile NO2

concentration (lg m~3) (lower"gure) in the Helsinki metropolitan area in 1993.

3.3. The chemical transformation

The ratio of NO2}NO

xis approximately 10% in traf-

"c emissions, and somewhat smaller in power plant emis-sions. Most of NO released from tra$c is oxidised toNO

2within a time scale of a few minutes (e.g., HaK rkoK nen

et al., 1996), while for stationary releases this time scale isof the order of 10}20 min (Janssen et al., 1988). Far fromthe sources, an approximate photochemical equilibriumis reached, resulting in an NO

2/NO

xratio of about 90%

in typical ambient conditions.The predicted concentration ratio NO

2/NO

xvaries

from approximately 50% in the vicinity of busy roads toapproximately 90% at the outer edges of the computa-tional regime. As expected, the NO

2/NO

xconcentration

ratios are substantially higher than the correspondingratios in the emissions, even close to the tra$c emissionsources. This is caused by the in#uence of the more`ageda air masses from urban background and station-ary sources, and partly also by the numerical averaging

caused by the computational procedure (with grid sizesof 50}100 m).

3.4. The spatial distribution of the statistical concentrationparameters

Figs. 7a}b show the predicted distribution of statisticalparameters for the highest hourly 99-percentile and thesecond highest daily mean NO

2concentrations in 1993.

These parameters can be compared to the correspondinghourly (150 lg m~3) and daily (70 lg m~3) nationalhealth-based air quality guidelines. The guideline valuesfor the daily NO

2concentrations were exceeded fairly

extensively, in the vicinity of the main roads and in theHelsinki downtown area. The corresponding guidelinevalues for the hourly NO

2concentrations were exceeded

only at a few limited areas.

4. Conclusions

The modelling approach adopted has certain inherentlimitations, both concerning the evaluation of emissionsand atmospheric dispersion. Gaussian dispersion model-ling does not allow for the detailed structure of buildingsand obstacles. However, the terrain in the area is #at andthe average height of the buildings is fairly low (mostbuildings are lower than 15}20 m). The computed con-centrations should be interpreted as spatially averagedvalues (on the scale of the grid spacing, varying from 50to 500 m), while for instance, inside a street canyon theactual concentrations can vary substantially on the scaleof tens of meters.

On the other hand, the use of fairly simple dispersionmodels facilitates the evaluation of the hourly time seriesof meteorological and emission conditions for one year,which is required for the computation of statistical con-centration parameters, de"ned in national health-basedair quality guidelines. It is not at present possible toconduct such an analysis for an agglomeration of citiesusing, for instance, street canyon dispersion models orcomputational #uid dynamics models.

It was also possible to include emissions from a largenumber of sources (this study included 5000 line sources,169 point sources and area sources), a substantial num-ber of receptor points (10 000), and to use a su$cientlydense computational grid. We also allowed for the essen-tial chemical transformation processes in the computa-tions. The central processing (CPU) time required for thecomputations on the Cray C94 supercomputer was ap-proximately 20 h.

Estimation of emissions also contains inherent limita-tions. Near the junctions of roads and streets there isacceleration and deceleration of the tra$c #ow, as well asstops and occasional congestion, which causes increasedemissions. The emission modelling takes properly into



account the in#uence of vehicle acceleration and deceler-ation on the emissions. The emissions, however, are as-sumed to be distributed evenly along each line source inthe numerical computations, although these can bestrongly concentrated in the immediate vicinity of thejunctions. This e!ect can cause an underprediction oftra$c emissions near major junctions.

The modelling system contains a novel method, whichallows for the chemical interaction of pollutants, origin-ating from a large number of urban sources. The systemproperly takes into account, for instance, the depletion ofO

3in the urban area. This phenomenon can have a sub-

stantial in#uence on the computed results particularly inepisodic conditions, in which the atmospheric di!usionconditions are unfavourable. However, the state-of-the-art modelling systems do not take these dependenciesinto account.

It can be shown numerically that although the contri-bution of tra$c on the total emissions is slightly less thana half, approximately 80}95% of the ground level NO

xconcentrations originate from tra$c sources. The modestcontribution of stationary sources on ground level con-centrations is caused by the strong concentration of theNO

xemissions to large power plants within the study

area.The concentrations of NO

xand NO

2are strongly

distributed in the vicinity of the main roads and streets,and in their junctions. The NO

xemissions increase with

the vehicle travel speed, which are substantially higher atmajor roads, compared with minor roads and streets.The concentrations of NO

xand NO

2were also higher in

the downtown area of Helsinki, caused by the large tra$cvolumes. The national air quality guidelines of the dailyNO

2concentrations were exceeded fairly extensively, in

the vicinity of the main roads and in the Helsinki down-town area.

The annual average of the NO2/NO

xconcentration

ratio varies from approximately 50% in the vicinity ofbusy roads to approximately 90% at the outer edges ofthe computational regime. The latter limit correspondsapproximately to photochemical equilibrium conditions.The NO

2/NO

xratio is substantially higher in the pre-

dicted concentrations, compared with the correspondingratio in the emissions, even near the sources. This iscaused by the in#uence of the more &aged' air masses fromurban background and stationary sources, and partlyalso by the numerical averaging caused by the computa-tional procedure.

The modelling system developed has been an impor-tant assessment tool for the local environmental authori-ties. The system has been applied in order to evaluate thecompliance of air quality with the guidelines and limitvalues (together with the measured concentrations), thein#uence of various emission categories on air quality,and the representativity of the urban monitoring sta-tions. The modelling system has also been used in the

environmental impact assessment of the di!erent trans-portation system scenarios in the Helsinki MetropolitanArea (HaK mekoski and Sihto, 1996).

Acknowledgements

This study is by the cooperation between the FinnishMeteorological Institute (FMI), the Helsinki Metropoli-tan Area Council (YTV) and the Technical ResearchCentre of Finland (VTT). We would like to thank ourcoworkers in this study, Ms. PaK ivi Aarnio (Helsinki met-ropolitan Area Council) and Mr. Juhani Laurikko (VTTEnergy). We also wish to acknowledge Mr. HarriPietarila, Mr. Risto Pesonen, Mr. Jari HaK rkoK nen and Mr.Esko Valkonen (FMI) for their cooperation. The studyhas been part of the national research programme `MO-BILE } Energy and the environment in transportationa.The funding from the Technology Development Centre(TEKES) is gratefully acknowledged.

Appendix A. Inclusion of the chemical transformationmodule into a Gaussian line source dispersion model.

The chemical transofrmation equations are included tothe dispersion model based on a modi"ed version of thediscrete parcel method (DPM); the original version hasbeen presented by Benson (1984, 1992). This methodconsiders air parcels, in which the emissions and back-ground air are assumed to be instantaneously uniformlymixed. The chemical reactions in each parcel are thenassumed to proceed independently of the dispersion pro-cess.

In this model, the initial mixing zone concentrationfrom a line source can be written as

Ci"C

i!#

siQ

u2H0

. (A.1)

where i refers to the species, Ci!

is the ambient back-ground concentration, s

iis the mass fraction of the

species i in the emissions, Q is the emission strength perunit length of the source (kg/(ms)), u is the wind velocityat the height H

0and 2H

0is the source height. The mass

fraction si

is needed, for instance, to take into accountthe fraction of NO

2in the total NO

xemissions.

The size of the reaction volume is therefore determinedby the height of the source and the wind velocity. Thechemical reactions are then allowed to proceed duringthe travel time, i.e., the time of transport from the sourceto the receptor.

We suggest a revised version of the discrete parcelmethod, in which the reaction volume is de"ned separ-ately for each receptor location, instead of the source



location. This method is named receptor-oriented DPM(R-DPM); we call the original method source-orientedDPM (S-DPM).

The relevant length scales of the reaction volume in theR-DPM are determined by the length of the "nite linesource, and the lateral and vertical dispersion parameterspy

and pz. We de"ne the vertical length-scale simply as

H7(x)H

0#kp

z, where k is a constant (selected here as

k"3) and x is the distance fro the source to the receptor.The horizontal length scale is de"ned as H

)(x)"2

(p sin h#k py), where p is the half-lengtht of the line

source and h is the angle between the wind direction andthe road.

The initial concentration in the reaction volume in theR-DPM is equal to the source mass #ux (2 ps

iQ) divided

by the volume #ux perpendicular to the line source at thereceptor location, H

7(x) H

)(c) (u sin h), i.e.,

Ci(t)"C

i!#

siQ

p(H#kp

2)(p sin h#kp

y)(u sin h)

. (A.2)

The e%uents are then assumed to be instantaneouslymixed into the reaction box, and the chemical reactionsare assumed to proceed during the travel time.

In bothe versions of the model, the time available forreactions is a function of the transport distance. How-ever, for longer transport distances, more background airis available for the reactions. This is allowed for in theR-DPM, as the reaction volume is a function of thetransport distance, while in the S-DPM, the reaction boxvolume is determined only by source properties. This isthe main physical di!erence between the two models.

Bothe DPMs do not explicitly allow for the interactionof the chemical reactions and the physical mixing pro-cesses, or the in#uence of the pollutant concentrationpro"les within the plume on the chemical transforma-tion. In order to allow for these e!ects, a substantiallymore complex numerical transformation model would berequired (for instance, Kerminen and Wexler, 1996).

For a more detailed discussion of these models, thereader is referred to HaK rkoK nen et al. (1996).

References

Benson, P.E., 1984. CALINE4 } a dispersion model for predic-ting air pollutant concentrations near roadways.FHWA/CA/TL-84/15, California Department of Transpor-tation, Sacramento, CA.

Benson, P.E., 1992. A review of the development and applicationof the Caline3 and 4 Models. Atmospheric Environment26B, 379}390.

Carruthers, D.J., Edmunds, H.A., Ellis, K.L., McHugh, C.A.,1995. The atmospheric dispersion modelling system(ADMS): comparisons with data from the Kincaid experi-

ment. International Journal of Environment and Pollution 5,382}400.

ElolaK hde, T., Koskentalo, T., 1996. Calculating emissions ofa transport network. Magazine of the Finnish Air PollutionPrevention Society 5}6, 26}28.

Gryning, S.E., Holtslag, A.A.M., Irwin, J.S., Sievertsen, B., 1987.Applied dispersion modelling based on meteorological scal-ing parameters. Atmospheric Environment 21, 79}89.

Hanna, S.R., 1985. Air quality modelling over short distances.In: Houghton, D.D. (Ed.), Handbook of Applied Meteoro-logy. University of Wisconsin, Wisconsin (Chapter 25).

HaK mekoski, K., Sihto, S., 1996. Transportation System 2020 andAir quality in the Helsinki metropolitan area, Finland. In:Baldasano Recio, J.M., Sucharov, L.J. (Eds.), Urban Trans-port and the Environment II. Computational MechanicsPublications, Southampton.

HaK rkoK nen, J., Valkonen, E., Kukkonen, J., Rantakrans, E., Jal-kanen, L., Lahtinen, K., 1995. An operational dispersionmodel for predicting pollution from a road. InternationalJournal of Environment and Pollution 5, 602}610.

HaK rkoK nen, J., Valkonen, E., Kukkonen, J., Rantakrans, E.,Lahtinen, K., Karppinen, A., Jalkanen, L., 1996. A model forthe dispersion of pollution from a road network. FinnishMeteorological Institute, Publications on Air Quality 23.Helsinki, 34 pp.

HaK rkoK nen, J., Walden, J., Kukkonen, J., 1997. Comparison ofmodel predictions and measurements near a major road inan urban area. International Journal of Environment andPollution 8 (3}6), 761}768.

Hertel, O., Berkowicz, R., 1989. Modelling NO2

concentrationsin a street canyon. DMU Luft-A131. National Environ-mental Research Institute, Roskilde, 31pp.

INRO, 1994. EMME/2 User's Manual. INRO Consultants Inc.,MontreH al, Canada.

Janssen, L., van Wakeren, J., van Duuren, H., Elshout, A., 1988.A classi"cation of NO oxidation rates in power plant plumesbased on atmospheric conditions. Atmospheric Environ-ment 22, 43}53.

Jol, A., Kielland, G. (Eds.), 1997. Air Pollution in Europe 1997.EEA Environmental Monograph No. 4. European Environ-ment Agency, Copenhagen, 107 pp.

Karppinen, A., Kukkonen, J., Konttinen, M., Rantakrans, E.,Valkonen, E., HaK rkoK nen, J., Koskentalo, T., ElolaK hde, T.,1997a. Comparison of dispersion model predictions and theresults from an urban air quality measurement network. In:Power, H., Tirabassi, T., Brebbia, C.A. (Eds.), Air Pollution.V. CMP, Southampton, pp. 405}411.

Karppinen, A., Jo!re, S., Vaajama, P., 1997b. Boundarylayer parametrization for Finnish regulatory dispersionmodels. International Journal of Environment and Pollution8(3}6), 557}564.

Karppinen, A., Kukkonen, J., Konttinen, M., HaK rkoK nen, J.,Valkonen, E., Rantakrans, E., Koskentalo, T., ElolaK hde, T.,1998a. The emissions, dispersion and chemical transforma-tion of tra$c-originated nitrogen oxides at the Helsinkimetropolitan area. International Journal of Vehicle Design20 (1}4), 131}136.

Karppinen, A., Kukkonen, J., Konttinen, M., HaK rkoK nen,J., Valkonen, E., Koskentalo, T., ElolaK hde, T., 1998b. De-velopment and veri"cation of a modelling system forpredicting urban NO

2concentrations. In: Gryning, S.-E.,



Chaumerliac, N. (Eds.), Air Pollution Modelling and itsApplication XXII, Vol 22. NATO, Challenges of ModernSociety, New York, pp. 567}574.

Karppinen, A., Kukkonen, J., Nordlund, G., Rantakrans, E.,Valkama, I., 1998c. A dispersion modelling system for urbanair pollution. Finnish Meteorological Institute, Publicationson Air Quality 28. Helsinki, 58 p.

Kerminen, V.-M., Wexler, A.S., 1996. The occurrence of sulfuricacid}water nucleation in plumes: urban environment. TellusB 48, 65}82.

Kukkonen, J., HaK rkoK nen, J., Valkonen, E., Karppinen, A.,Rantakrans, E., 1997. Regulatory dispersion modelling inFinland. International Journal of Environment and Pollu-tion 8 (3}6), 782}788.

Kukkonen, J., Salmi, T., Saari, H., Konttinen, M., KartastenpaK aK ,R., 1999. Review of urban air quality in Finland. BorealEnvironment Research 4 (1), 55}65.

Luhar, A.K., Patil, R.S., 1989. A General "nite line source modelfor vehicular pollution prediction. Atmospheric Environ-ment 23, 555}562.

Moussiopoulos, N., Berge, E., Bohler, T., de Leeuw, F., Gron-skei, K.-E., Mylona, S., Tombrou, M., 1996. Ambient airquality, pollutant dispersion and transport models. Euro-pean Environment Agency, Topic Report 19, Air Quality.Copenhagen, 94 pp.

MaK kelaK , K., Kanner, H., Laurikko, J., 1996. Road tra$c exhaustgas emissions in Finland } LIISA 95 calculation software. (inFinnish). VTT Communities and Infrastructure, TransportResearch, Research Notes 1772, Technical Research Centerof Finland, Espoo, 45 pp.# App. 51 pp.

Nikmo, J., Tuovinen, J.-P., Kukkonen, J., Valkama, I., 1999.A hybrid plume model for local-scale atmospheric disper-sion. Atmospheric Environment 33, 4389}4399.

Olesen, H.R., 1995a. Regulatory dispersion modelling in De-nmark. International Journal of Environment and Pollution5, 412}417.

Olesen, H.R., 1995b. Datasets and protocol for model valida-tion. International Journal of Environment and Pollution 5,693}701.

Pietarila H., Pesonen R., Rasila, T., Rantakrans, E., 1997. Airquality in the surroundings of the city of Turku (in Finnish).Finnish Meteorological Institute, Air Quality, Report. Hel-sinki, 81 pp.#A pp.

Pesonen, R., Rantakrans, E., Pietarila, H., Karppinen, A., 1996.Computations of the dispersion of nitrogen oxides in theHelsinki metropolitan area, (in Finnish). Finnish Meteoro-logical Institute, Air Quality, Report. Helsinki, 61 pp.#A pp.

Petersen, W., 1980. User's guide for HIWAY2, a highwayair pollution model. EPA-600/8-80-018, U.S. Environ-mental Protection Agency, Research Triangle Park, NC,69 pp.

Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry andPhysics. ISBN 0-471-17815-2. Wiley, New York, USA,pp. 945}947.

Valkonen, E., HaK rkoK nen, J., Kukkonen, J., Rantakrans, E.,Jalkanen, L., Haarala, S., 1995. Application of dispersionmodels for evaluating the in#uence of urban air pollution onhuman health. International Journal of Environment andPollution 5, 557}566.

Valkonen, E., HaK rkoK nen, J., Kukkonen, J., Rantakrans, E.,Jalkanen, L., Haarala, S., 1996. Modelling urban air pollu-tion in Espoo, Finland. The Science of the Total Environ-ment 189/190, 205}211.

van Ulden, A., Holtslag, A., 1985. Estimation of atmosphericboundary layer parameters for di!usion applications.Journal of Climate and Applied Meteorology 24,1196}1207.

Walden, J., HaK rkoK nen, J., Pohjola, V., Kukkonen, J., Kartasten-paK aK , R., 1995. Vertical concentration pro"les in urban condi-tions } comparison of measurements and model predictions.In: Anttila, P., et al. (Ed.), Proceedings of the 10th WorldClean Air Congress, Espoo, Finland, May 28}June 2, 1995,Vol. 2. The Finnish Air Pollution Prevention Society,Helsinki, pp. 268 (4 p.).

Yamartino, R.J., Scire, J.S., Carmichael, G.R., Chang, Y.S., 1992.The CALGRID mesoscale photochemical grid model } I.Model formulation. Atmospheric Environment A (26), (8),1493}1512.



Documents

A modelling system for predicting urban air pollution ...lib.tkk.fi/Diss/2001/isbn9512257599/article4.pdf · time series of concentrations. The modelling system contains a method,