Environmental Modelling & Software 21 (2006) 454e460
www.elsevier.com/locate/envsoft
Traffic pollution modelling and emission data
R. Berkowicz*, M. Winther, M. Ketzel
National Environmental Research Institute, Frederiksborgvej 399, DK-4000 Roskilde, Denmark
Received 12 August 2003; received in revised form 8 June 2004; accepted 10 June 2004
Available online 25 March 2005
Abstract
Evaluation of traffic pollution in streets requires basically information on three main factors: traffic emissions, the meteorological
conditions and the street surroundings. Dispersion models exist with various degree of sophistication, which are able to properly
describe the dispersion conditions, and thus to predict the relationships between emissions and the concentration levels in the street.
However, for real-world applications, the model calculations must be based on ‘‘true’’ emission data, and their estimation is not
trivial. Significant uncertainty is still connected with emission data. Examining the relationships between model predictions and
measurements with respect to the meteorological conditions and inter-relationships between different pollution components allows
quantitative evaluation of the traffic emissions. This methodology is illustrated using the Danish Operational Street Pollution Model e
OSPM, and time series of traffic related pollutants. Street level concentrations of NOx and CO are calculated using OSPM as the
dispersion model and emission data estimated by the widely used COPERT methodology. Comparison with measurements shows
significant underestimation of the pollution concentrations and especially the CO/NOx ratio. An alternative set of traffic emission
factors, using a more simplified vehicle classification methodology, provides better agreement with the measured concentrations.
Evaluation of these results provides some guidance on the necessary modifications of the ‘‘real-world’’ traffic emission factors.
Ó 2004 Elsevier Ltd. All rights reserved.
Keywords: Traffic emissions; Pollution modelling; OSPM; COPERT
1. Introduction from laboratory measurements performed on selected
vehicles under different simulated driving conditions.
As with all other air pollution models, traffic One of the most extensive traffic emission modelling
pollution models require data on the source emissions. methods used within the European context is the
The traffic emissions can hardly be measured in real- COPERT model (Ntziachristos and Samaras, 2000).
world conditions, and as a rule, the emissions are Lots of laboratory data were processed statistically
calculated based on the traffic data and vehicle specific within the COPERT project resulting in vehicle emission
emission factors. These emission factors depend not factor expressions as function of the vehicle travel speed.
only on the vehicle type but also to a large degree on the Expressions are provided for different vehicle categories
driving conditions and thereby can vary between with differentiation between vehicle types (passenger
different street locations. cars, vans, trucks, etc.), fuel used, engine capacity or
Different methods exist for the determination of the weight and perhaps most important, the emission
vehicle emission factors, but the most common approach legislation category. Besides the basic emission factors,
is based on utilisation of experimental data obtained corrections are also provided accounting for cold starts
and degradation of the emission reduction equipment
with the age (mileage) of vehicles. The model is
* Corresponding author. extremely flexible and easy to use. The original aim of
E-mail address:
[email protected] (R. Berkowicz). the COPERT model was to provide a standardised
1364-8152/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2004.06.013
R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460 455
method for estimation of national emissions of traffic
related pollutants. However, experimental verification
of emission predictions on a national level is hardly
possible. An alternative way is to conduct a comparison
between measurements of traffic related pollutants at
a street site and model predictions using the recom-
mended vehicle emission factors.
Proper validation of pollution dispersion models is
a crucial prerequisite for their practical applications. In
Sahm et al., 2002 and Ketzel et al., 2002 results of
a comprehensive evaluation study of several advanced
local scale pollution models are presented. This study
was mainly focused on evaluation of the model
Fig. 1. Wind direction dependence of the measured NOx concen-
performance with respect to their ability to reproduce
trations in Jagtvej (street contribution only) and OSPM modelled
the meteorological and dispersion conditions under concentrations with unit emission. All data are averaged over 10 wind
different obstacle configurations. Although, very impor- direction sectors. Wind directions parallel with the street axis are
tant, this step is not sufficient to guaranty that models indicated by dashed lines.
are suitable for practical applications. This was very well
illustrated in a European-wide study, the so-called Only data from working days and working hours (8e18)
‘‘Podbielski exercise’’ (Lohmeyer et al., 2002), where are selected. OSPM can reproduce the observed
results of 24 models for the same street location were variation of the street level concentrations with the
compared and evaluated. Importance of all stages of wind direction. Wind direction and wind speed are the
model application, including emission modelling, was two most important meteorological factors governing
shown to be crucial for the final results. dispersion conditions in street canyons.
The latest Windows version of OSPM (WinOSPM,
http://ospm.dmu.dk) contains a fully integrated emis-
2. Traffic pollution model sion module. This module is based on the COPERT
methodology but the dependence of the emission factors
The Danish Operational Street Pollution Model, on the travel speed can be user defined. Also the split-up
OSPM (Berkowicz, 2000), is used here for modelling of the vehicle fleet in the different technology and
traffic pollution in streets. OSPM is a simple para- legislation categories can be specified by the user
meterised model. Parameterisation of flow and disper- according to the available data. The alternative set of
sion conditions in street canyons was deduced from emission factor expressions used in this work is based on
extensive analysis of experimental data and model tests. a study conducted in Denmark in the beginning of 1990s
The model requires little CPU-time. Therefore many (Hansen et al., 1995; Jensen, 1995). The average travel
model simulations for longer time periods can be easily speed is used here as the main driving pattern parameter
performed, and this makes the model suitable for and the classification of the vehicle fleet is determined
emission scenario studies and more general data only by the share of cars with and without the catalytic
analyses. A detailed discussion of the model concept converter. For heavy-duty traffic and buses the split-up
and also of advantages and disadvantages of this type of is according to the share of vehicles complying or not
simple models compared to more sophisticated fluid with the EURO criteria. No additional differentiation
dynamics models is given in Berkowicz (1998). with respect to EURO classes is made. Although very
The ability of OSPM to reproduce the dispersion simplified, this emission modelling procedure was
conditions in urban streets has been demonstrated in successfully used in many previous OSPM applications.
several model studies (Ketzel et al., 2000; Kukkonen
et al., 2001; Berkowicz et al., 2002). Fig. 1 shows
a comparison between the measured and modelled NOx
concentrations at a monitoring location in the street 3. The site and traffic data
Jagtvej, Copenhagen, Denmark (see Section 3). The
model predictions are shown in terms of calculated The data used in this study originate from a perma-
concentrations but with a unit emission. No comparison nent monitoring station at the street Jagtvej in Copen-
between the absolute values should thus be made but hagen, Denmark (Palmgren and Kemp, 2002). This site
only the shape of the curves concerning the variation is especially suitable for the purpose of this study
with the wind direction (roof level) should be taken into because continuous street pollution data, meteorological
account. The average values are calculated based on and urban background data are available since 1993 and
hourly data from 1994 using 10 wind sector averaging. the geometrical configuration of the street resembles
456 R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460
a regular street canyon very well. The street is 25 m wide Table 1
with 18 m high buildings on both sides. The street axis is Traffic volume and composition in the street Jagtvej
orientated 30 with respect to North and the monitoring Total traffic Passenger Vans Trucks Buses
station is located on the East side of the street. A map of (veh/day) cars (%) (%) (%) (%)
the site showing the location of both the street and the Working days 29,367 87.0 8.6 3.1 1.3
urban background station is given in Fig. 2. The urban Saturdays 24,635 89.6 8.4 1.3 0.7
Sundays 23,976 90.0 8.3 1.0 0.7
background measurements are conducted on the roof of
the Copenhagen University building (height ca. 20 m).
The meteorological measurements originate from the
create an average diurnal profile with split-up between
same location using an instrumented 10 m high mast
working days, Saturdays and Sundays. The same
raised on the roof of the building.
average diurnal profile is used for all the model
The traffic in the street is dominated by passenger
calculations presented here.
cars with about 29,000 veh/day on the working days and
somewhat less on the weekends. The heavy-duty traffic
is quite moderate (about 4.4%) and its share is
significantly smaller during the weekends. 4. Emissions
Table 1 shows the details of the traffic volume and
composition. All traffic data originate from a combina- Figs. 3 and 4 show the diurnal variation of traffic
tion of automatic and manual traffic counts performed emissions in Jagtvej estimated using the COPERT
over several years. The automatic counts use the model and the alternative method based on the Danish
Autoscope SoloÒ Wide Area Video Vehicle Detection study. Results are presented for two years: 1994 and
System (by Image Sensing Systems, Inc, http://www. 2000. The share of cars with the catalytic converter has
autoscope.com). These automatic counts were supple- increased during these years from about 25% in 1994 to
mented by manual counts in order to get a more reliable about 65% in 2000. The NOx and CO emissions are
data on traffic composition. For this study, the traffic shown as aggregated data for two vehicle classes only:
counts of the last two years (2002/2003) were used to passenger cars and diesel vehicles. Passenger cars in
Fig. 2. Map showing the location of the monitoring sites used in this study. The urban background station is located on the roof the Copenhagen
University building. A 10 m high meteorological mast is situated at the same location.
R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460 457
8000 8000
NOx emissions (g/km/h) Jagtvej, 1994 Total NOx emissions (g/km/h) Jagtvej, 1994 Total
Passenger cars COPERT Passenger cars
Diesel Diesel
6000 6000
4000 4000
2000 2000
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
8000 8000
NOx emissions (g/km/h) Jagtvej, 2000 Total NOx emissions (g/km/h) Jagtvej, 2000 Total
Passenger cars COPERT Passenger cars
Diesel Diesel
6000 6000
4000 4000
2000 2000
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
Fig. 3. Diurnal variation of the NOx emissions in Jagtvej in the years 1994 and 2000 estimated by the two different emission models.
Denmark are mainly gasoline cars, with only small Comparing results from the two emission models it is
percentage of diesel cars (ca. 7% in the year 2000). On obvious that the main difference in the NOx emissions
the contrary, the heavy-duty traffic and vans are can be attributed to the difference in the emissions from
predominantly diesel vehicles. heavy-duty vehicles (the ‘‘Diesel’’ class). COPERT
80000 CO emissions (g/km/h) Jagtvej, 1994 80000 CO emissions (g/km/h) Jagtvej, 1994
Total
COPERT Total
Passenger cars Passenger cars
Diesel Diesel
60000 60000
40000 40000
20000 20000
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
80000 CO emissions (g/km/h) Jagtvej, 2000 80000 CO emissions (g/km/h) Jagtvej, 2000
Total COPERT
Passenger cars Total
Diesel Passenger cars
Diesel
60000 60000
40000 40000
20000 20000
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
Fig. 4. Diurnal variation of the CO emissions in Jagtvej in the years 1994 and 2000 estimated by the two different emission models.
458 R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460
emissions are here about two times smaller than processes applied in the model. However, the ratio of
predicted by the alternative method. The difference in concentrations of different pollutants is independent of
the CO emissions is significantly larger. For passenger the dispersion parameterisation, and depends only on
cars, which totally dominate the CO emissions, this the applied emissions. The ratios of measured and
difference amounts to almost a factor of 3. modelled NOx and CO concentrations are shown in
Fig. 7. The CO/NOx ratio is significantly underestimated
when emissions are calculated using the COPERT
5. Street concentrations model. A similar pronounced underestimation of the
CO/NOx ratio is also reported by Dommen et al. (2003)
The differences in the emissions clearly affect the based on a comparison of emission inventory for the
results of model calculations for NOx and CO concen- Lombardy region, Italy, and air pollution measurements
trations in the street (Figs. 5 and 6). All concentrations at two locations in the Milan area. This indicates that
are presented here as a ‘‘street contribution’’ only, i.e. using COPERT emission factors can lead to underesti-
measured urban background concentrations are sub- mation of emissions from a whole region, and this can
tracted from the total street concentrations. also have implications for large-scale model calculations.
Not only does the simple emission model provide However, street level measurements and modelling
results that are much closer to the measured levels than provide more direct indication of the quality of the used
the COPERT emissions, but also the diurnal and weekly emission data because it is easier to isolate the pollution
variation is reproduced better. This is most obvious for source than in the case of urban or regional studies.
NOx concentrations for which the proportion between Taking into account that the ‘‘alternative’’ emission
passenger cars’ emissions and the heavy-duty traffic data were to some extent adapted to conditions in the
emissions is affecting the diurnal variation in the street Jagtvej, an independent verification of the data is
concentration levels. For CO, the model results using required too. For this purpose OSPM calculations with
the COPERT emissions are always much lower than the the same emission factors as used for Jagtvej are also
measured values. A similar conclusion was also pro- performed for another street site. This site is the street
vided by Vogel et al. (2000) based on measurements near Albanigade in a Danish provincial town Odense. The
a motorway in Germany. traffic in this street is about 21,000 veh/day on working
The concentrations calculated by a dispersion model days. The heavy-duty traffic constitutes about 7%, i.e.
depend of course on parameterisation of dispersion substantially more than in Jagtvej. Results for NOx are
200 200
NOx (ppb) Jagtvej,1994 measured NOx (ppb) Jagtvej,1994 measured
modelled COPERT emissions modelled
160 background 160 background
120 120
80 80
40 40
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
200 200
NOx (ppb) Jagtvej, 2000 measured NOx (ppb) Jagtvej, 2000 measured
modelled COPERT emissions modelled
160 background 160 background
120 120
80 80
40 40
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
Fig. 5. Average diurnal variation of the measured and modelled NOx concentrations (street contribution only) in the years 1994 and 2000 calculated
using OSPM and the two different emission models. The urban background concentration is also shown.
R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460 459
3 CO (ppm) Jagtvej, 1994 3 CO (ppm) Jagtvej, 1994
measured measured
modelled COPERT emissions modelled
background background
2 2
1 1
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
3 3 CO (ppm) Jagtvej, 2000
CO (ppm) Jagtvej, 2000 measured measured
modelled COPERT emissions modelled
background background
2 2
1 1
0 0
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
Mon-Fri Sat Sun Mon-Fri Sat Sun
Fig. 6. Average diurnal variation of the measured and modelled CO concentrations (street contribution only) in the years 1994 and 2000 calculated
using OSPM and the two different emission models. The urban background concentration is also shown.
shown in Fig. 8. Calculation results with COPERT travel speed and percentage of cold starts have significant
emissions are not presented here but an analogous influence on the emission values. These parameters are
underestimation as for Jagtvej is observed here as well. difficult to quantify with sufficient accuracy and more
Unfortunately, lack of urban background data does not work is needed on this subject. However, the general
permit a similar comparison for CO. tendency is that application of COPERT emission factors
leads to a significant underestimation of street level
pollution concentrations. For the street Jagtvej, with the
6. Conclusions share of heavy-duty traffic of about 5%, the NOx was
under estimated by about 30%. This underestimation
It might be too premature to draw any final results probably from too low emissions attributed to
conclusions from this study. Several factors, such as heavy diesel traffic. An even larger underestimation, of
Fig. 7. The ratio between NOx and CO concentrations (street contribution only). (a) As measured, (b) calculated by OSPM with the alternative
emissions, (c) calculated by OSPM with COPERT emissions.
460 R. Berkowicz et al. / Environmental Modelling & Software 21 (2006) 454e460
200 NOx (ppb) Berkowicz, R., 2000. OSPM e a parameterised street pollution model.
measured
Albanigade,1998 modelled Environmental Monitoring and Assessment 65, 323e331.
160 background Berkowicz, R., Ketzel, M., Vachon, G., Louka, P., Rosant, J.-M.,
Mestayer, P.G., Sini, J.-F., 2002. Examination of traffic pollution
120 distribution in a street canyon using the Nantes’99 experimental
data and comparison with model results. Water, Air and Soil
80 Pollution: Focus 2 (5), 311e324.
Dommen, J., Prevot, A.S.H., Baertsch-Ritter, N., Maffeis, G.,
40 Longoni, M.G., Grüebler, F.C., Thielmann, A., 2003. High-
resolution emission inventory of the Lombardy region: development
0 and comparison with measurements. Atmospheric Environment 37,
3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24 3 6 9 12 15 18 21 24
4149e4161.
Mon-Fri Sat Sun Hansen, J.Q., Winther, M., Sorenson, S.C., 1995. The influence of
driving patterns on petrol passenger car emissions. The Science of
Fig. 8. Average diurnal variation of the measured and modelled NOx
the Total Environment 169, 129e139.
concentrations (street contribution only) in Albanigade, Odense. The
Jensen, S.S., 1995. Driving patterns and emissions from different
urban background concentration is also shown.
types of roads. The Science of the Total Environment 169,
123e128.
about 60%, is observed for CO, since the CO/NOx ratio Ketzel, M., Berkowicz, R., Lohmeyer, A., 2000. Comparison of
numerical street dispersion models with results from wind tunnel
deduced from the calculated emissions is much smaller
and field measurements. Environmental Monitoring and Assess-
compared to street pollution measurements. ment 65, 363e370.
Reliable data on total traffic emissions are obviously Ketzel, M., Louka, P., Sahm, P., Sini, J.-F., Moussiopoulos, N., 2002.
crucial for the calculated street level concentrations. Intercomparison of numerical urban dispersion models e Part II:
However, even more important is the right proportion street canyon in Hannover, Germany. Water, Air and Soil
Pollution: Focus 2 (5), 603e613.
of traffic emissions attributed to different vehicle
Kukkonen, J., Valkonen, E., Walden, J., Koskentalo, T., Aarnio, P.,
categories. Evaluation of the effectiveness of the future Karppinen, A., Berkowicz, R., Kartastenpää, R., 2001. A
emission reduction strategies might be significantly measurement campaign in a street canyon in Helsinki and
biased if the proportion of emissions from e.g. passenger comparison of results with predictions of OSPM model. Atmo-
cars and the heavy-duty traffic is incorrect. spheric Environment 35, 231e243.
Lohmeyer, A., Mueller, W.J., Baechlin, W., 2002. A comparison of
Valuable information on traffic emission factors can
street canyon concentration predictions by different modellers: final
be obtained by comparison of model results with results now available from the Podbi-exercise. Atmospheric
measurements from a street location. Availability of Environment 36, 157e158.
both NOx and CO measurements is very useful because Ntziachristos, L., Samaras, Z., 2000. COPERT III: computer pro-
their emissions can be mainly attributed to diesel and gramme to calculate emissions from road transport-methodology
and emission factors. European topic centre on air emissions, EEA,
gasoline vehicles, respectively. Similar studies for other
Copenhagen, Denmark.
street locations with different traffic composition and Palmgren, F., Kemp, K., 2002. The Danish air quality monitoring
driving pattern should provide additional information programme. Annual Report No. 407, NERI, Roskilde, Denmark.
to improve the emission models. Sahm, P., Louka, P., Ketzel, M., Guilloteau, E., Sini, J.-F., 2002.
Intercomparison of numerical urban dispersion models e Part I:
street canyon and single building configurations. Water, Air and
Soil Pollution: Focus 2 (5), 587e601.
References Vogel, B., Corsmeier, U., Vogel, H., Fiedler, F., Kühlwein, J.,
Friedrich, R., Obermeier, A., Weppner, J., Kalthoff, N.,
Berkowicz, R., 1998. Street scale models. In: Fenger, J., Hertel, O., Bäumer, D., Bitzer, A., Jay, K., 2000. Comparison of measured
Palmgren, F. (Eds.), Urban Air Pollution e European Aspects. and calculated motorway emission data. Atmospheric Environment
Kluwer Academic Publishers, pp. 223e251. 34, 2437e2450.
