Pergamon Atmospheric EnvironmentVol. 27B, No. 4, pp. 389-399, 1993.

Copyright © 1994 Elsevier Science Ltd

Printed in Great Britain. All rights reserved
0957-1272/93 $6.00+0.00

THE CAR MODEL: THE DUTCH METHOD TO DETERMINE

CITY STREET AIR QUALITY
H. C. EERENS
Laboratory for Air Research (LLO), RIVM, P.O. Box 1, 3720 BA Bilthoven, The Netherlands

C. J. SLIGGERS
Ministry of Housing, Physical Planning and Environment; Air Directorate, P.O. Box 450,
2260 MB Leidschendam, The Netherlands

and

K. D. VAN DEN HOUT

TNO Institute of Environmental Sciences, P.O. Box 6011, 2600 JA Delft, The Netherlands

(First received 22 February 1992 and in final form 30 April 1993)

Al~traet--The CAR model (Calculation of Air pollution from Road traffic) is a simple parameterized model
for the determination of air quality alongside roads (including street canyons) in cities. The calibration of
the model, using data from the Dutch National Air Quality Monitoring Network, is described and a limited
validation of the model is given. The model supports the implementation of air quality decrees under the Air
Pollution Act by provincial and municipal authorities. Various applications of the model, including
a scenario analysis for city street air quality in the Netherlands in the period 1989-2010, are presented.

Key word index: Air quality modelling, air quality standards, traffic pollution, environmental traffic maps.

1. INTRODUCTION by (non-scientific) local personnel, the model must

fulfil certain quality criteria (Dekker et al., 1990; NNI,
Approximately two billion people living in cities 1991) and must be easy to use. In 1985 such models
throughout the world are exposed to polluted air were not generally available; neither was there
every day (UNEP/WHO, 1988). The situation in the a recognized method to calculate NO2 concentrations
Netherlands is not different. Half of the Dutch popu- in the immediate vicinity of busy streets. The Dutch
lation lives in medium-sized and large cities. Traffic is Environmental Ministry therefore decided to pro-
the predominant source of pollution in most Dutch mote the development of a suitable model, resulting in
cities. It is estimated that the Dutch air quality stan- the model described in this paper--the CAR (Calcu-
dards (limit values) for carbon monoxide (CO), lead lation of Air Pollution from Road traffic) model.
(Pb), nitrogen dioxide (NOz), suspended particulates The CAR model is able to calculate annual percen-
and benzene are being violated along 1000-2500 tile values and average concentrations close to streets
streets in these cities (RIVM, 1988, 1991). In order to (e.g. at the kerb side) for non-reactive air pollutants
put an end to this undesirable situation, a policy has and NO2. The model is calibrated annually against
been developed which requires limit values to be met measurement results, taken from the Dutch National
in these cities by the year 2000 at the latest (Sliggers, Air Quality Monitoring Network (Elskamp, 1989).
1989). The limit values which must be met are set out The CAR model is now being used in the Nether-
in air quality decrees which have the force of law lands by municipalities not only in the implementa-
under the Air Pollution Act. tion of national air quality decrees but also for traffic
Since the municipalities are responsible for traffic and environmental planning with the use of instru-
within their boundaries, supervision of air quality ments such as Environmental Traffic Maps; and by
along roads in built-up areas and the institution of the central government for the evaluation of current
measures to prevent violation of the limit values have policy and the consequences of new policy proposals
been put in the hands of the municipalities. (environmental forecasting).
Because of the impracticability of carrying out rou- This paper provides a description of the principles
tine measurements in all busy Dutch streets, air pollu- of the CAR model, an explanantion of the calibration
tion in such streets is determined by means of model procedure and a discussion of model results; and
calculations. Because decisions based on the outcome shows how the model is used by municipalities and
of the models have considerable, financial, con- the central government. Those interested may obtain
sequences and because the model will often be applied the software on request from the authors.
389
390 H.C. EERENSet al.

2. MODEL DESCRIFrlON - - The recirculation contribution, i.e. the contribu-

tion of pollutant that reaches the receptor point after
2.1. Basic principles recirculation in the lee of a building or between build-
A literature review showed that most of the existing ings, was generally found to be much smaller than the
line source models assume a homogeneous terrain direct contribution arising from the relatively concen-
around the source (see, among others, Szepesi, 1989). trated plume blowing directly to the receptor point.
Application is thus limited to highways in relatively - - Rows of trees were found to have a variable
fiat terrain, e.g. the G M model (Chock, 1978) and effect on the concentration pattern, varying from
CALINE (Benson, 1979) and to those city streets slightly negative to very positive (100% rise).
where the effects of buildings and trees can be de-
scribed by aerodynamic terrain roughness, e.g. PAL These results were incorporated into a generally ap-
(Turner and Peterson, 1975). Buildings, however, are plicable, but complex, street dispersion model, the
known to affect the flow of air in their direct vicinity, T N O traffic model (van den Hout et al., 1989).
Since wind tunnel experiments are not suitable for
such models cannot be applied to most city streets.
For the special case of street canyons, the submodel of modelling the conversion of NO to NO2 in streets, an
the APRAC model is often employed (Johnson et al., NO2 module was derived from theoretical and em-
1973). Since the development of this street canyon pirical considerations. This NO2 module takes into
model, several other studies have been devoted to account the non-linear relation between the NO~, and
this particular configuration (e.g. Yamartino and NO2 concentrations, which is due to the chemical
Wiegand, 1986). At most locations near city traffic, reaction of NO with ozone. This relationship, which
including intersections of streets, the buildings and also contains a term for the direct emission of NO2, is
trees do not form a street canyon. Because the existing applied for each wind direction sector separately. The
models were not applicable to those situations, it was results of these investigations were also incorporated
decided to investigate especially these circumstances. in the TNO traffic model (van den Hout and Baars,
An analysis of measured concentrations in streets 1988).
(Den Tonkelaar and van den Hout, 1980) revealed From a sensitivity analysis of the model, the para-
that although the short-term (hourly) concentration meters with the largest influence on the calculated
values are strongly affected by the buildings in the concentrations were selected. An important choice
streets, the long-term (yearly) average concentration made was to distinguish five types of streets, see Fig. 1.
pattern is much less sensitive to these than is often From the wind tunnel experiments it became clear
assumed. Further, it was found that the ratio between that one of the highest average traffic pollutant contri-
the annual average values and the high percentiles of butions has to be expected in street type 4. In this
the frequency distribution of concentrations did not situation, the wind direction at street level is rather
vary very much from street to street. This is mainly insensitive to the wind direction above roof level, and
due to the fairly constant diurnal emission patterns is in most cases directed from the road to the building,
giving a high annual average contribution at receptor
and the linear source configuration, which strongly
reduce the variability of the concentration as a func- points between the road and the building. In Fig. 2
tion of wind direction. wind roses from the wind tunnel experiments and the
In an extensive programme of wind tunnel experi- measured, normalized wind roses for CO are given for
ments (van den Hout and Baars, 1988; van den Hout street types 2 and 4. As can be seen from the figure, the
wind roses from the windtunnel experiments agree
and Duijm, 1988; van den Hout et al., 1989) the
well with the measured windroses. Further, a list of
relation between emissions and concentrations was
investigated for 49 configuration sets which differed data that are readily available to the municipal
with respect to dimensions, distances and shapes of authorities was made. From this, an optimum input
streets, its buildings and its trees. Each configuration data set for a simplified model version (the CAR
contained one to five receptor points at a (full-scale) model) could be derived which would permit calcu-
height of 1.5 m. Since the variation of pollutant con- lations with an acceptable degree of accuracy.
centration with wind speed and stability is relatively
well known for most conditions, the experiments fo-
2.2. Model design
cused on the effects of wind direction on the concen-
trations. In the CAR model the calculations are performed in
The experiments showed that the observed pattern the following way:
could be fairly well understood by applying a few
basic concepts, of which the most prominent are sum- (1) Calculation of the city background level (Cb).
marized here: (2) Calculation of the local street emission (Et).
- - The most obvious influence of buildings was (3) Calculation of the local street traffic contribution
found to be the change of direction of the local street (C,)
level wind in the wake of buildings and between build- (4) Calculation of the annual average (pollutant) con-
ings. centration or percentile (C~O.
 The CAR model 391

Street types in the CAR-model

Type It: ~==0.75.10- 4 S2 - 0.70.10- 2 S + 0.17

Road in open terrain. No buildings or trees within
a dislance of 100 m (or so few thai roughness
length is not affected).

Type 2 : ~ , = 3 . 1 0 . 1 0 "4 S2 - 1.82.10 .2 S + 0.33

Basic street type, all roadsrot defined by 1,
2 3A, 3B or 4.
Type 3A: ~s=3.25.10-4 S 2 - 2.05.10 .2 S + 0.39
Street with buildings (at least 3 m high) on both
sides; no gaps larger than 25 m and at least 75 m
buildings per 1(30 m s0"net. The ratio between
height of buildings and dislance of these: [o road
3a axes is from 1.5 to 3 on one side and less than 3
on the olher.
LIlH1 <3 1.5 <12/H2 <3

H2 Type 3B: ~D=--4.88.10-4 S2 - 3.08.10 -2 S + 0.59

Special cas¢ of sa'ect type 3A, ih¢ aforemenlJoned
3b ralio is lessthan 1.5 for boll; sidesof [he road.
LI/H1 '<1.~ __~_~=,'1.42 ',at.6 Type 4 : ¢ = = 5 . 0 0 , 1 0 .4 S2 - 3.16.10- 2 S + 0.57
Street with buildings (at least 3 m high) on one
side., wilh the ratio betiding-height/road-axis-
distance less than 3; the other fide "has nua
buildings or a ratio much greater than 3
4 (preferably gtealer than 10).

LIIH1 <3 H- Heil~t building {m) l) Wilb excep0onof slxcetlype 2, ~ dclinilJon

L = Diltance road axle to building (m) Jsonly validif fl~ SEtterhas ~ sa~ oJilfigura-
lion overat least 100 m. S: distanceIo roadaxis(m)

F i g . l. S t r e e t t y p e s a n d t h e i r d i l u t i o n f u n c t i o n a s d e f i n e d in t h e C A R m o d e l .

cavity

T
receptor i receptor
!
tit O

(a) ) (a) (b)

F i g . 2. N o r m a l i z e d wind roses from CO measurements (a) a t s t r e e t s t a t i o n s (238 a n d 636) a n d w i n d t u n n e l e x p e r i m e n t s

(b).
392 H.C. EERENSet al.

The annual average pollutant concentrations or the traffic category; the second part of the equation is an
percentiles are calculated at a height of 1.5 m above empirical relation between the NO2 and the NOx
pavements for distances from the receptor point to the concentrations at a certain ozone concentration level.
road axis of between 5 and 30 m. In this paragraph the The factor fl in this equation is a street-type depend-
design of the CAR model will be discussed, based on ent factor describing the fraction of the background
the above four steps. ozone that is available for the reaction of ozone with
(1) The city background level. Cb is calculated as NO. K is an empirical parameter obtained from
the sum of the regional background concentration or measurements.
percentile outside the city, Cr and the city specific (4) The annual averaoe concentration or percen-
contribution, Co: tile. Cpot is calculated as the sum of the city back-
ground level, Cb, and the street specific contribution,
Cb=Cr+Cc. (1.I) C1:
The city specific contribution C¢ is assumed to be
Cpol= Px* C,+Cb, (1.6)
proportional to the city's (effective) radius (calculated
from the city surface area): where P~ is the ratio between the traffic contribution
to the percentile concentration to be calculated and
Cc = ~t• Re, (1.2)
the contribution to the annual average concentration.
where ct is a parameter (gg m - 3 kin- 1) to be estimated Px depends on the street type and is determined an-
from measurements at city background monitoring nually from measurements at street monitoring sta-
sites (Section 3) and Re is the average distance from tions; Cpo, is the calculated annual average concentra-
the centre of a city to the border of its built-up area in tion or percentile.
km.
(2) The local street emission. Et(#g s- l m - 1) is cal-
culated from speed-class-dependent emission factors 3. YEARLY UPDATE AND CALIBRATION OF THE CAR
in each of two traffic categories (passenger vehicles MODEL

and light-duty traffic in one, and heavy-duty traffic

The expressions in the CAR model contain para-
(>3500 kg) and buses in the other):
meters that vary from year to year. Therefore, the
Et=(1-Fv)*N*Ep+Fv*N*Ev, (1.3) model is annually updated and calibrated using meas-
ured data. In Table I the CAR parameters for the
where Fv is the fraction of heavy-duty traffic and buses
street stations of the Dutch National Air Quality
in a street; N is the total number of vehicles in a street
Monitoring Network (LML) are presented. The street
per day; Ep and Ev are the speed-dependent emission
stations are carefully sited in order to cover all para-
factors for the two traffic categories.
meters used in the CAR model. The parameters to be
(3) The street-specific contribution. Ct is calculated
revised can be divided into two categories:
from the estimated traffic emission, Et, and street type
specific dispersion parameters: (1) default values for input data (Fr and Cr)
(2) (calibrated) values for the internal model para-
C,=Et*Os * F~* Fo, (1.4)
meters (Ep, Ev, Fr~o~, a, Px, K and fl).
where q),(s m - z ) is a street-type (see Fig. I) depend-
In Sections 3.1 and 3.2 the estimation procedure for
ent dilution function in the form of an empirical
the above categories will be explained, using N O 2 as
polynomial, using the distance of the receptor point to
an example. In Section 3.3 the CAR model will be
the road axis as a variable; F~ accounts for the differ-
compared with measured values at street stations. An
ence between the actual regional average wind speed
extensive overview of the calibration procedure is
and the national potential average wind speed [a
given by Eerens and Schokkin (1991).
reference value derived from the average wind speed
measured at 10 m height near the airport of Amster- 3.1. Annual update of the external model parameters
dam (Schiphol) converted to an open field value with
In the Dutch National Air Quality Monitoring
roughness length Zo=0.15m]; Fo accounts for the
Network, NOx (NO plus NOz) and ozone are meas-
effects of trees on the wind velocity, as derived from
ured at 43 and 37 monitoring stations, respectively
wind tunnel measurements, in the street.
(see Elskamp, 1989; Buisman 1990 for network config-
For the calculation of the NO2 concentration a uration), of which 25 and 21 are regional background
module was derived from the more elaborate TNO
stations, 5 and 4 are city background stations, and 13
traffic model (Section 2.1). This module uses the rela- and 12 are street stations, respectively. Based on long-
tionship: term measurements (1978-1987), 14 different NO2
3 * Cb.o~ * C,.~o~ regions are distinguished in the Netherlands, where
CcNO2~-FNo2*Ct.NOar , (1.5) the 1990 regional 98 percentile 1-hour NO2 concen-
K + Ct.NO x
trations (Cr) vary from 61 to 98/~gm -3.
where FNo~ is the fraction of total NO~ emission For the determination of the average wind speed
directly emitted as NO2: it is dependent on speed and (for the estimation of the meteorological correction
 The CAR model 393

Table 1. Description of the street stations in the Dutch national air quality measuring network (as of 1990)
Street- Station Vehicles St V~t Ratio ~"
type Street names no. City* per day (m) F,~" (kmh- l) Fo L/H't CO/NOx
2
Amsterdamsevaart~: 537 Haarlem 70,000 14/30 0.04 13 1.0 variable 8.6
Kardinaal de Jongweg§ 636 Utrecht 28,500 10/28 0.04 44 1.25 7/10 5.3
Loolaan 727 Apeidoorn 11,000 8 0.04 22 1.5 4 II 6.4
Arnhemseweg 729 Apeldoorn 14,000 6 0.04 22 1.0 3 [J 5.6
3A
Vieutenseweg 638 Utrecht 30,000 15 0.06 22 1.0 2.3 6.2
Constant Enzeijstraat 639 Utrecht 12,000 6 0.04 22 1.0 2/1.7 4.9
Stationstraat 728 Apeldoorn 15,000 I0 0.07 22 1.0 1.8 6.3
3B
Wittevrouwenstraat 637 Utrecht 14,000 5 0.10 13 1.0 0.5/0.8 5.8
4
Genovevalaan 236 Eindhoven 12,000 6 0.04 22 1.0 1.0 6.8
Noord-Brabantlaan 237 Eindhoven 20,200 8 0.04 44 1.0 1.0 5.5
Piuslaan 238 Eindhoven 30,200 9 0.04 44 1.0 2.3 5.4

* The city's radius, Re, is 3 km for the mentioned cities.

tS: distance of the street station (inlet point) to the road axis;
Fv: fraction heavy duty traffic and buses in the street;
Vx: speed class;
L/H: the ratio between the distance from the road axis to lateral buildings and the height of the said buildings;
Ratio: ratio of the 98 l-h values, NO x expressed in pgNO 2m -3.
~tAt the cross point of two busy roads, with an intensity of respectively 31,000 and 39,000 vehicles per day.
§The two sides of the road are separated by a central reservation of more than 3 m (in this case 10 m) and are therefore
considered for the model as two separate roads, each with a traffic intensity of 14,000 vehicles per day.
[IThe lateral buildings are covering about 50% of street-side.

factor, F,) 33 meteorological stations are available in g,

the network. Six different windspeed regions are dis-
tinguished in the Netherlands; during 1987-1990 the .~ 40
Y=2.3+4.5X R = 0.97 1990
parameter F, ranged from 0.88 to 1.39 in the six 35
Z
regions.
..~ 30
3.2. Annual update of the internal model parameters 25
Emission data: Ep and E~. The emission factor for ~t 2o
NOx is reported annually by the Dutch Central / A Utrecht
Bureau of Statistics (cf. CBS, 1991) for passenger cars, ~+ o Den Haag
light-duty and heavy-duty vehicles. For the different o / • Rotterdam
components used in the CAR model, the emission as 5 f • Amsterdam
a function of the average speed of a vehicle is known I I I I I I
1 2 3 4 5 6 7
and used to calculate the emission factor for three
Urban area radius (kin)
speed classes in the city:
1. Average speed 13 km h - l ; a street with a traffic Fig. 3. Determination of • by plotting the urban
jam occurring regularly for at least 50% of the cars radius (km) against the city background contribu-
tion of NO2.
using the street each day.
2. Average speed 22 km h - 1; a street with a speed
with an average population density of 5800 inhabit-
limit of 50 km h - ~.
ants per km 2 and an average traffic density of
3. Average speed 44 km h - 1; a street with a speed
75,000km(day.km) -z. For NO2, with a relatively
limit of 70 km h - 1.
small city background contribution, one average
Parameter ~. For all city background stations the value for a for all cities is used. For components with
city background contribution (Co) is determined by a large contribution from the city background, e.g.
subtracting the measured 98 percentile regional back- benzene, a range of values of ~, dependent on the
ground concentration from the 98 percentile concen- traffic density, which varies between 30,000 and
tration measured at the city background stations. The 100,000 kin(day" kin) -2, is used.
resulting values are plotted against the radius of the Parameter Px. Px is calculated directly for each
city; ct is determined from the regression line (see street type and component as the ratio of the street
Fig. 3). The estimated u is representative for cities in contributions to the desired percentile and to the
the Netherlands with 40,000 to 1,000,000 inhabitants, annual average concentration, respectively.
394 H.C. EERENSet al.

Parameter FNov For the determination of FNo~, 25

the annual average NO 2 and 0 3 concentrations, Cf~O2 =(O.6*Cc.o~*CtNo)/(Ct.NOx+130) 1990
measured at street stations, are summed ( N O 2 + 20
+
el
Oa = O~) and plotted against the NOx concentration
as measured at these stations. The regression coeffic-
ient of the line through these datapoints is defined as
ENOs •
] lo
Parameters K and ft. K and fl are determined by
~ 5
plotting Cd.o~ against Ct.Nox. Here, Cd.o3 is the differ-
ence between the (calculated) annual average city
I I 1 I I I
background ozone concentration and the measured 0 50 100 150 200 250 300 350
concentration of ozone in the street; this is considered Traffic contributionCLNOx(~g m"3NO~
an approximation for the NO2 (Cf.No~) formed in the Fig. 4. Determination ofthe parameters K and fl accord-
street by the reaction of ozone with the NO emitted ing to Equation 2 with data from the Dutch National Air
by cars in the street. Ct.Nox is the difference of the city Quality Measuring Network, solid line indicates best fit
NO~ background concentration (in #g NO 2 m - 3) and with K = 130/~gm -3 and fl=0.6.
the measured NO~ in the street. Values for the para-
meters K and fl are obtained from best-fit approxima-
tions (see Fig. 4) in Equation 2: between calculated and measured values is therefore
only an indication of the minimum error that can be
/~ * Cb.o3 * C,.NOx expected for calculations with the CAR model in
C d . o 3 = C f . NO2 "~- (2) other streets. Better methods for the estimation of the
K + Ct.Nox
expected errors are either comparison with indepen-
In Fig. 4 the results for the estimation of fl (0.6) and dent measurements or comparison with another,
K (130) in 1990 are given. The correlation coefficient is validated, model. Heida et al. (1989) compared CAR
moderate. A somewhat better fit could be found if the model results with measurements in 10 streets in
amount of effectively available ozone (determined by Amsterdam. They found differences, averaged over
fl) is considered as dependent on the distance of the the 10 streets, between CAR results and measured
receptor/measuring point to the road axis; part of the values of -10___ 12% for the 98 percentile of CO and
scatter in Fig. 4 can thus be explained. As an example, 6 + 10% for the 98 percentile of NO2.
in Fig. 4 a data point, labelled 638, is measured at It can be concluded that the CAR model is a suffi-
a street station at a distance of 15 m from the road ciently reliable and feasible method for the calculation
axis, while the other street stations are located at of traffic-exhaust-related air pollutants in urban
a distance of 5-10 m. As can be clearly seen in the streets and that the accuracy of the model is well
figure, the data point is above the calculated regres- within the limits as set by the Air Quality decrees
sion line, indicating a relative high contribution (fl) (30% at 70% confidence level).
from the city ozone background concentration to the
measured NO2 concentration. Further investigation
of this problem is being undertaken and may result,
4. APPLICATION OF THE CAR MODEL
for distances greater than 10 m, in a revision of the
IN THE NETHERLANDS
fl parameter.
4.1. Air quality decrees
3.3. Comparison of the CAR model results with values In 1987 the CAR model was introduced as a tool for
obtained from street measurement stations the implementation of air quality decrees. CAR has
The CAR-calculated values and the measured con- been made available on diskette for use on MS-DOS
centrations from the Dutch National Air Quality PCs. Data files and parameter updates have been
Monitoring Network (LML) street stations are given provided annually since 1988 by the Dutch National
in Table 3.2 for CO (annual 98 percentile of the 8-h Institute for Public Health and Environmental Pro-
averages), NOx (annual average) and NO 2 (98 percen- tection (RIVM) for, among others, all 73 municip-
tile of the 1-h values). The calculations are corrected alities with more than 40,000 inhabitants, the 12
for differences in height (the CAR model calculates the provinces and the nine regional environmental
concentration at 1.5 m, the street stations measure at inspectorates.
2.5 m) by multiplying the calculated street contribu- At national workshops, held in 1987, the CAR
tions for CO and NOx by 0.91, as derived from model was introduced to these groups as an instru-
measurement campaigns near these street stations. ment to identify those streets that were at risk of
The average, relative, differences between the cal- exceeding the limit values set in the air quality decrees.
culated and measured values are -3-1-9% for CO, From Dutch National Air Quality Monitoring
8_+19% for NOx, and 6_+9% for NO2. Network (LML) measurements, it was estimated that
The measurement results in Table 2 have been used streets with calculated long-term 98 percentile
for the calibration of the CAR model. The difference concentrations higher than 80% or 90% of the limit
 The CAR model 395

Table 2. Comparison of measured and calculated concentrations of NO~, NO 2 and CO at 11 street

stations in the Netherlands for 1990
2 3A 3B 4
Street type:
Component\station: 537 636 727 729 638 639 728 637 236 237 238

CO
Traffic contr. (Ct.co) 2900 1050 1450 1400 1850 1350 1300 3750 1750 1550 2150
P98-8h:CAR* 4400 2550 2950 2900 3350 2850 2800 5250 3250 3050 3650
Measured 4300 2800 2700 2800 3200 3100 3000 6400 3700 3400 3400
Difference(%) 2 -9 9 4 5 -8 - 7 -18 -1 -10 7
NO~
Traffic contr. (Ct.Nox) 135 117 86 77 122 95 95 270 112 188 266
Yearly av: CAR* 199 179 144 135 184 157 153 332 178 248 328
Measured 182 163 120 122 184 207 148 401 165 190 226
Difference (%) 9 9 17 11 0 -24 3 -17 8 31 45
NOz
Traffic contr. (C~.No~) 27 26 22 20 27 23 23 46 26 35 42
P98-1 h: CAR* 131 123 108 106 124 120 109 143 118 127 134
Measured 132 110 97 91 134 121 108 151 105 109 118
Difference (%) -1 12 11 16 -7 -1 1 -5 12 17 14

*City background concentration for CO: 1500#gin-3; NO~: 58-64#g NO2m-3; NO2:86-104
/~gm-3.

values for CO and NO2, respectively, should be important advantages of this approach is that
considered at risk. For these streets, exceedance is solutions that merely move environmental problems
expected at least once every 5 years, and annual from one street to another can be foreseen and
concentration calculations are obligatory. avoided.
A questionnaire distributed for evaluation of the
practical use of the model revealed that the CAR
model is well accepted by users, who also rated its 4.3. Scenario studies for policy evaluation by the Dutch
practical usability as good. government
Every 2 years a national environmental survey is
4.2. Environmental Traffic Maps made by the RIVM for the Dutch government
Environmental Traffic Maps have been developed (RIVM, 1988, 1991). In the 1990 survey (RIVM, 1991),
as tools to assess environmental quality in cities and the effect of the Dutch clean car policy was evaluated
to support coherent traffic and environmental policies with respect to the air quality near busy streets in
by municipalities. Environmental Traffic Maps are Dutch cities. Environmental traffic maps, as described
based on three (sub)-models: above, were used to estimate the number of street
kilometres where exceedances of reference values for
1. A traffic model incorporating the road network
seven pollutants (cf. Table 3) had occurred (period
and the origins and destinations of the traffic, with
1970-1989) and what could be expected (period
tools to calculate the effects of changes in traffic flow
1989-2010). The results of this study are summarized
according to projections.
here, with the main focus on the use of the CAR
2. A dispersion model - - the CAR model - - to
model.
calculate air quality along streets.
Period 1970-1989. In Table 4 the calculated ex-
3. A model to calculate the noise level caused by
ceedances are given for the reference values given in
traffic.
Table 3. From Table 4 it clear that exceedances for the
Colours on these maps indicate the extent to which pollutants carbon monoxide, black smoke, lead, ben-
limit values, such as those set out in air quality decrees zene and benz(a)pyrene in Dutch cities have de-
or (noise) nuisance regulations, are being exceeded. creased since 1970. Exceedances for nitrogen dioxide
Figure 5 is an example of such a map for the city of and fine particles, on the other hand, have increased in
Maastricht. In addition to maps describing the cur- this period.
rent situation, prognostic maps can be made for vari- Period 1994-2010. For the 1994-2010 scenario,
ous projected situations. Municipalities using the the basis is provided by the recently published Dutch
models for traffic prognoses, (a)under the status quo government environment plan, N M P + (VROM,
situation and (b) to achieve an improvement over the 1990); emission factor reductions for 1994, 2000 and
present environmental situation, are subsidised, for 2010 consequent on this plan's aims are listed in
75% of the costs, by the central government. Table 5. The reduction percentages given in Table 5
In this way, the environmental problems related to also take into account limitations in catalyst function,
traffic in a city and their possible solutions are com- such as those caused by ageing, lead "poisoning" or
bined in a coherent, integrated way. One of the low operating temperature at the start of a journey
396 H.C. EERENS et al.

/'~ "" 4 isl - 16o

t.......~..~.. ". I"""" 5 >160
.~..t

Fig. 5. Environmental Traffic Map (l-h 98 percentile NO2) of the city of Maastricht (116,000 inhabit-
ants).

Table 3. Limit and reference values for traffic-relevant substances (/~gm -3) (RIVM,
1991)

Annual average 98 percentile Average time

Co mponent (t~g m - 31 ( #g m - a) (h)

Carbon monoxide (CO) -- 6000 8

Nitrogen dioxide ( N O , ) -- 135 1
Lead (Pb) 0.5 2 24
Black smoke -- 90 24
Formaldehyde -- 40 24
Benzene 10 -- --
Benz(a) pyrene 0.001 -- --
Sulphur dioxide -- 250 24
Fine particles (PM-10) 40 - - - -

Table 4. (Expected) exceedance of reference values along busy roads in the Netherlands in the period
1970-2010

Exceedances (km street length)

Component 1970 1980 1985 1989 1994 2000 2000* 2010

Carbon monoxide (CO) > 1000 260 70 50 9 2 1-2 0

Nitrogen dioxide (NO2) 40 290 280 380 200 40 30-50 10
Black smoke > 1200 210 540 450 380 40 20-55 0
Lead (Pb) > 2100 > 2100 > 2200 240 2 0 0-0 0
Benzene > 1200 680 480 470 130 65 50-80 25
Fine particles (PM-10) >2100 >2300 >2200 >2400 >2300 800 700-930 130
Benz(a)pyrene >2100 > 2000 > 1600 > 1900 > 1800 720 610-820 330
Streets ( > 500 mtv/day)(km) 2100 2300 2500 2700 2900 3000 3000 3100

* Expected exceedance ranges, calculated for a favourable and an unfavourable meteorological year.
 The CAR model 397

Table 5. Emission and concentration reductions for the period 1989-2010

Fine
Component CO NOx/NO~' Black smoke Benzene particles

Emission (g km- z)l" (1989)

passenger/light-duty vehicles~ 12 1.5 0.13 0.13 0.24
heavy-duty vehicles/buses 16 17 0.95 0.08 2.8
Emission from 1989
1994 (passenger/light-duty) 68 61 84 59 84
2000 (passenger/light-duty) 56 36 37 47 37
2010 (passenger/light-duty) 45 21 26 36 26
1994 (heavy-duty/buses) 96 98 104 100 104
2000 (heavy-duty/buses) 69 67 76 70 76
2010 (heavy-duty/buses) 47 51 52 50 52
City background§ (/zgm- 3) 98P-8 h 98P-1h 98P-24h annual av. annual av.
1989 2300 120 80 5.2 50
1994 2000 110 80 4.2 48
2000 1500 95 70 2.8 40
2010 1200 90 40 2.1 30

* Emission factor given as NOt (in NO 2 equivalents),concentrations are only given for NOz.
f At an average speed of 22 km h-1 (speed class 2 of the CAR model).
~/Weighted average emission of passenger cars (gasoline, LPG and diesel) and light duty traffic
( < 3500 kg).
§City background for a city of 750,000 inhabitants.

(cf. Rijkeboer et al., 1990). The scenario assumes a 3% Further developments

annual growth in traffic, made up of a 2% increase in - - The more detailed TNO street dispersion model
streets in built-up areas and a 1% increase in traffic (Section 2.1) will be issued as a software package
intensity per street. The calculations, see Table 4, under the name CAR-SPECIAL. At present
show for all pollutants for the period 1989-2010 a de- a verification study is being carried out. The model
crease in exceedances. However, the calculations also is to be used in situations where various measures
indicate that some exceedances will still occur in the are being discussed for reducing the level of air
years 2000 and 2010. pollution. In these situations, when mostly costly
measures are being discussed, it is advisable to use
a model that describes the current situation as
5. CONCLUSIONS AND FURTHER DEVELOPMENTS accurately as possible.
- - The approach described in this paper can be used
Conclusions
for related air pollution problems. A similar
- - Based on wind tunnel experiments, theoretical model, CAR-petrol, is presently under develop-
considerations and measurements, a dispersion ment for calculating benzene concentrations close
model has been developed that calculates average to petrol delivery stations in the Netherlands.
concentrations and concentration percentiles in
streets for inert pollutants as well as NO2, using
readily available input data. Acknowledgements--Thanks are due to James Burn, Roel
van Aalst and Dirk Onderdelinden for their comments o n
- - The results indicate that the CAR model is a suffi- early drafts of the paper.
ciently reliable and feasible method for the calcu-
lation of traffic-exhaust-related air pollutants in
urban streets, and that the accuracy of the model is
well within the limits set by Dutch air quality R E F E ~ N C E S

decrees.
- - The CAR model has proved, in itself as well as in Benson P. E. (1979) CALINE-3-a versatile dispersion model
combination with the Environmental Traffic for predicting air pollutant levels near highways and
Maps, to be a valuable tool for local authorities in arterial streets, Report No. FHWA/CA/TL-79/23
Buisman E. (1990) Measurement activities of the National
describing the current air quality in their streets. Air Quality Monitoring Network in 1990 (in Dutch),
- - It has been shown that the model can successfully RIVM, report 222101005, Bilthoverh the Netherlands
be used to provide a general survey of air quality Chock D. P. (1978) A simple line-sourcemodel for pollutant
in busy streets in the Netherlands and can he used dispersion near roadways. Atmospheric Environment 12,
823-829.
to calculate the consequences of various scenarios, Davenport A. G. (1965) Proc. N.P.L. Symposium No. 16.
such as may be proposed by the central govern- Wind Effects and Building Structures, pp. 54-102.
ment. H.M.S.O., London.

A[{8) 27:4-0
398 H.C. EERENS et al.

Dekker C. M., Groenendijk A., Sliggers C. J. and Verboom evaluation of simple models for the flow, turbulence and
G. K, (1990, Dutch, 1991, English) Quality Criteria for pollutant concentration fields within an urban street can-
Models to Calculate Air Pollution. Publication Series Air, yon. Atmospheric Environment 20, 2137-2156.
90, Staatsuitgeverij/DOP, 's-Gravenhage.
Den Tonkelaar W. A. M. and Hout K, D. van den (1980)
System for Calculating Air Pollution part X, Air pollution
by traffic, analysis of measurement data (in Dutch),
CMP-TNO, report CMP 80/12, Delft, the Netherlands. APPENDIX
Eerens H. C. and Schokkin G. J. H. (1991) CAR-model: DEVELOPMENT OF THE TNO TRAFFIC MODEL
annual update procedures, RIVM, report 228475015,
Bilthoven, the Netherlands (in preparation), A1. Wind tunnel measurements
Elskamp H. J. (Ed.) (1989) National Air Quality Monitoring
Network--Technical description, RIVM, report Strategy. As a general rule, it was attempted to select
228702017, Bilthoven, the Netherlands. parameter variations that gave the least predictable results,
Heida H., Jong A. L. de and Huygen C. (1989) Model in order to obtain the maximum amount of information. For
calculations of street-air concentrations for carbon the meteorological parameters only the influence of wind
monoxide and nitrogen dioxide in Amsterdam. In Man direction was retained. The effect of stability is relatively
and his Ecosystem. Proc. 8th Worm Clean Air Congress. small near traffic in the vicinity of buildings. The concentra-
(edited by Brasser L. J. and Mulder W. C.), Vol. 3, pp. tion is inversely proportional to the wind speed when there is
233-238. Elsevier, Amsterdam. no influence of the motion of traffic. Not simulated were the
Johnson W. B., Ludwig F. L., Dabbart W. F. and Allen R. J. chemical formation of NO2, the spatial variation of emis-
(1973) An urban diffusion simulation model for carbon sions along the street and the additional dispersion induced
monoxide. J. Air Poll. Control Ass. 23, 490-493. by traffic motion.
NNI (1991) Air Quality-requirements for the description of In total, 49 different configurations were studied, each
models to calculate the dispersion of air pollutants, Draft with several receptors. Variations were made regarding the
NEN 2828 (Standard of the Netherlands), Dutch distances between the line source and building blocks, re-
Standardization Institute (NNI), Delft, the Netherlands. garding building dimensions and shape and regarding the
Pasquill F. (1974) Atmospheric Diffusion. John Wiley and presence of trees. The set included "infinitely" long buildings,
Sons, Chinchester. shorter buildings and blocks separated by open spaces.
Rijkeboer R. C., Sloten P. van, Haagen M. F. van der (1990) Experimental setup. The atmospheric boundary layer was
Control Program--Restricted Number of Emission simulated on a scale of 1:250. A turning disk of diameter
Measurements of Cars; Yearly Report 1989-90. Publica- 2.3 m contained the line source, up to 12 receptors and the
tion Series Air, 89, Staatsuitgeverij/DOP, 's-Gravenhage. building blocks and trees when present. Beyond 75 m (full
RIVM (1988) Concern for tomorrow: a national environ- scale) from the line source blocks were placed on the turning
mental survey 1985-2010; publication of the National disk and further upwind to simulate the roughness of city or
Institute of Public Health and Environmental Protection of open country when desired. The building models were in
(RIVM), P.O. 1, 3720 BA Bilthoven, the Netherlands. most cases 15 m high and 10 m deep.
RIVM (1991) National environmental survey 2: 1990-
2010; publication of the National Institute of Public A2. Results of wind tunnel measurements
Health and Environmental Protection (RIVM), P.O. 1, A detailed description of the results is given in van den
3720 BA Bilthoven, the Netherlands (English version in Hout and Baars (1988) and van den Hout and Duijm (1989).
preparation). A selection of results is shown in Fig. A1.
Sliggers C. J. (1989) The implementation of ambient air In the analysis of the results, two contributions were
quality standards in urban areas in the Netherlands. In distinguished. The so-called direct contribution arises from
Man and his Ecosystem. Proc. 8th World Clean Air Con- pollution that reaches the receptor without first being
gress (edited by Brasser L. J. and Mulder W. C.), Vol. 4, pp. brought upward in the wake of a building. The other contri-
117-122. Elsevier, Amsterdam. bution, called the recirculation contribution, is caused by
Szepesi D. J. (1989) Compendium of Regulatory Air Quality pollution that has first been transported upward in the lee
Simulation Models. Akademiai Kiado, Budapest. vortex behind a building or between two buildings before
Turner D. B. and Peterson W. B. (1975) A gaussian plume reaching the receptor. In most cases the recirculation contri-
algorithm for point, area and line sources, 6th CCMS bution is present on both sides of the source and does not
International Technical Meeting on Air Pollution exhibit strong gradients near the source, as opposed to the
Modelling, Batelle Inst., Frankfurt/Main. direct contribution.
UNEP/WHO (1988) Assessment of urban air quality world Ground level wind. When the wind above roof level passes
wide, WHO, Geneva. over the roof of a building, a wake vortex is formed at the
van den Hout K.D. and Baars H. P. (1988) Development of leeside of the building in which the wind at ground level is
two models for the dispersion of air pollution by traffic: directed back towards the building. This is illustrated in
the TNO-traffic model and the CAR-model (in Dutch), Fig. AI: receptor c shows a pattern in which the lobes are
MT-TNO, report R88/192, Delft, the Netherlands. roughly inverted with respect to receptor b (no near build-
van den Hout K. D. and Duijm N. J. (1988) The dispersion of ings). See also the fourth configuration where for winds from
traffic emissions: the effect of recirculation near buildings the top of the page the pollution does not arrive at receptor
and the influence of trees (in Dutch), MT-TNO, report i but at h.
R88/447, Delft, the Netherlands. Recirculation contribution. For almost all configurations
van den Hout K.D., Baars H. P. and Duijm N. J. (1989) the recirculation contribution was found to be small com-
Effects of buildings and trees on air pollution by road pared to the direct contribution at receptors close to the
traffic. Proc. 8th Worm Clean Air Congress (edited source. In Fig. A1, for receptor c the lobes directed towards
by Brasser L. J. and Mulder W. C.) Vol. 4, Elsevier, the source are seen to represent relatively low concentra-
Amsterdam. tions.
VROM (1990) National Environmental Policy Plan--plus Vortices near the building face. Close to the face of build-
(in Dutch), Ministry of housing, physical planning and ings, both at the windwardside and the lee side, a vortex can
environment, report 21.137 nr 20-21, Staatsuitgeverij, form, which lowers the concentration. At receptor h, at winds
's-Gravenhage, the Netherlands. from the bottom of the page, the concentrations are similar
Yamartino R. J. and Wiegand G. (1986) Development and to those of receptor b, while for receptor g the concentrations
 The CAR model 399

o, lL
1 source

I[ .

)' / / : / ) r0.65 I ~ 0.78

ii-'----m, °"

V//S Jl
Fig. A1. Selection of wind tunnel results. Receptors are indicated by a, b, c. . . . . The wind roses represent the
concentration divided by the source strength and by the wind speed corresponding tollO m height and open
terrain, as a function of the direction of the wind above roof level. The average over all wind directions of this
value (in m - 1) is indicated for all receptors.

are lower than those at a. When an obstruction (another street, an increase in concentration by about a factor of two
building, trees) is placed at not too large a distance from the was found for receptors not strongly influenced by the wind-
building face the effect of the vortex was found to vanish. ward or secondary vortex.
Corners, interceptions. The shape of the concentration Incorporation of the results in computer model. The results
windroses near corners of buildings e.g. at street intersec- were incorporated in an earlier developed model of the
tions, proved to be difficult to explain in detail. An example Gaussian plume type, in many respects similar to many of its
is given Fig. 1, receptors d, e and f. The most important kind. The traffic is represented by line sources divided into
deviations from the expected concentration patterns were series of small point sources. The vertical wind profile is
the occurrence of peaks in unexpected wind directions. The described according to Davenport (1965), and the vertical
cause of these deviations can be qualitatively understood: the dispersion parameters are those of Pasquill (1974), with an
vortices near the corners of buildings have a complicated initial vertical dispersion due to traffic-induced dispersion
structure, causing the exhaust gases to be dispersed in direc- (taken as 1.5 m in cities and 2.5 m on highways). The model
tions that vary quite strongly spatially. Despite these devia- calculates the annual average concentration as a function of
tions, the concentration at the receptors averaged over all wind direction at receptors near roads. Further, the model
wind directions could be reasonably well predicted with the uses the dependence of the concentration on the wind direc-
above mentioned concepts. tion to estimate the shape of the frequency distribution.
Trees. For not too dense rows of trees a usually slight A semi-empirical relation is used to calculate NOz concen-
increase in concentration could be noticed at receptors that trations from the calculated local contribution to the N O ,
were not too close to buildings. For dense rows of trees contribution and estimated background ozone concentra-
standing at both sides of a street, and covering the entire tions.