
The influence of traffic on urban air quality  model predictions and their comparison to measurements Annual report 1994: Carbon monoxide concentrations in the Helsinki metropolitan area A. Karppinen, J. Kukkonen, E. Valkonen, J. Härkönen, E. Rantakrans, L. Jalkanen and S. Haarala Finnish Meteorological Institute, Air Quality Department Sahaajankatu 22 E, FIN00810, Helsinki, Finland T. Koskentalo, T. Elolähde and P. Aarnio Helsinki Metropolitan Area CouncilOpastinsilta 6 A,FIN00520,Helsinki,Finland J. Laurikko Technical Research Centre of Finland PL 1601,FIN02044 VTT, Helsinki, Finland Finnish Meteorological Institute, Air Quality Department 1995 This study is a cooperative project between the Air Quality Department of the Finnish Meteorological Institute (FMI), the Helsinki Metropolitan Area Council (YTV) and the Technical Research Centre of Finland (VTT). The traffic emissions were analysed by YTV and VTT, while the atmospheric dispersion of pollutants was computed by FMI. The main objective of this study has been to estimate air pollution concentrations in the Helsinki metropolitan area. This area consists of four cities  Helsinki, Espoo, Vantaa and Kauniainen, and has a total population of 850 000. This report describes the first part of this study, computations done in 1994. We address here only the results for the carbon monoxide (CO) concentrations, using the hourly emissions and meteorological data for one year, 1993. For evaluating the emissions of the road and street network, we have applied the transportation planning system EMME/2 (INRO, 1990), utilising the results by YTV (1990) and Spiess and Suter (1990). The emission factors were evaluated by the model LIISA 2.1 (Mäkelä et al., 1991), together with the results of the Ministry of transportation (1994) and Laurikko (1994). The atmospheric dispersion computations we made by the mainframe version of the CARFMI model (Härkönen et al., 1995). The computations included the emissions from mobile sources in the Helsinki metropolitan area. For carbon monoxide, almost all emissions in the study area originate from traffic (within an accuracy of few per cent). We computed the hourly time series of COconcentrations for the year 1993, for a set of receptor points in the area. The dispersion model also includes a computation of statistical parameters
(based on the concentration time series), which can be compared to air
quality standards or experimental measurements. We have compared these
computed parameters with the air quality measurements done by YTV at two
locations in the Helsinki downtown area.
2 Mathematical models 2.1 Emission models 2.1.1 Traffic volumes and average speed
The main problem in evaluating traffic emissions for a street network is a lack of measured data of hourly traffic volumes and speeds on a great number of street links. Most information therefore has to be obtained using average values, assumptions and calculations. As the emission factors depend on speed, average speeds on links are needed, in addition to traffic volumes. Traffic volumes and hourly average speeds were evaluated using the EMME/2 transportation planning system (INRO, 1990). The method is based on a comprehensive travel pattern interview in 1988. This has provided the information needed for the morning peak hour, offpeak hour and evening peak hour demand matrices (i.e trips from each of the zones to all the others) (YTV, 1990). The original network includes 117 traffic zones (i.e. subdivisions of the area considered) and 3400 directed links. This study was concentrated in the Helsinki inner city area. This network is more accurate in Helsinki area and has 396 zones and 4066 directed links. In order to evaluate the hourly emissions, regression model parameters were estimated and 24 hourly demand matrices were formed as a linear combination of those three basic matrices using the method of Spiess and Suter (1990). Hourly traffic counts from 48 sites of Helsinki Traffic Planning Division and 12 sites of Finnish National Road Administration in March 1993 were used; the traffic volumes were then updated for the year 1993. To reduce the calculating time, we selected 24 most representative hours from a total of 72 hours. The new matrices consisted of 10 average weekday matrices (average 23  05, 0506, 0708, 0809, average 09  14, 1415, 1516, 1718, 1920 and 2122), 7 Saturday matrices (average 2307, 0809, 1011, average 1216, 1718, 1920 and 2122) and 7 Sunday matrices (average 2207, 0708, 0910, 1112, average 1317,1819 and 2021). After the matrices had been formed, the hourly traffic volumes and average speeds on links were calculated using EMME/2 auto assignment (i.e network loading with the demand matrix). Transit volumes, i.e. buses on links, were evaluated separately using
morning peak and offpeak bus timetables. We appied the network of YTV,
having 282 zones and 3442 transit links. Regression model parameters were
estimated to calculate bus volumes of 24 average weekday, 24 Saturday and
24 Sunday hours.
2.1.2 Emission factors The applied emission factors of CO for cars have been shown in Figure 1. For passenger cars with the average speed up to 50 km/h, we applied new emission factors specially developed for the Helsinki city traffic (Ministry of transportation, 1994). The emission factors are based on driving cycles driven in different types of streets in Helsinki, and the engine maps evaluated for test vehicles. These emission factors depend strongly on the average speed, growing rapidly when the speed decreases below 20 km/h. This is especially the case for cars without a catalytic converter. For CO emissions of passenger cars with a catalytic converter, we applied a constant value of 8 g/km (Laurikko 1994), instead of the city traffic emission function. For passenger cars with average speeds above 70 km/h, we applied the emission factors given by the LIISA 2.1 model (Mäkelä et al., 1991). In the transition stage, with average speed ranging from 50 to 70 km/h, we applied a simple linear interpolation. The applied emission factors of CO for trucks have been shown in Figure 2. For trucks with or without a trailer, we have applied similar threepart emission factors based on the LIISA 2.1 model: city traffic factors for average speeds below 50 km/h, road traffic factors for average speeds above 70 km/h and a linear interpolation. For all other vehicles (for instance, buses and diesel vans), we have applied the emission factors (constants) of the LIISA 2.1 system. In winter a substantial number of cars are driving with cold engines,
causing increase in the emissions of CO and HC. We have therefore used
the emission factors derived from the socalled ECE (United Nations Economic
Commission for Europe) 15 cycle, determined at the temperature of 7 °C
(Laurikko, 1994).
2.1.3 Traffic emissions The emissions of CO, NO_{x} and HC were computed link by link for the given 24 hours using EMME/2, and the emissions for the remaining 48 hours were interpolated or copied. The emission values for each link were attached with coordinates of the start and end points of the link. This way the emissions for each hour during an average weekday, Saturday and Sunday in March 1993 were obtained. The emissions for other hours within the year were computed from these values, according to the seasonal variation in traffic volumes. The emission results are most accurate in the Helsinki inner city area. The model EMME/2 computes the combined traffic volumes of passenger cars and vans and the additional volumes of truck traffic. The average fractions of different vehicle types were based on the traffic count results from Helsinki. Based on results on the engine temperatures (Mäkelä 1994), the fraction of cars driving with cold engines was assumed to be 15 % of the gasoline driven cars. The emission factors for cold engines were used when the average temperature was below 0 °C. A scenario was computed, where all gasoline driven passenger cars are
equipped with catalytic converters. Traffic volumes and speeds as well
as the fractions of all other vehicle types (diesel cars and vans, trucks)
were supposed to be the same as in the base case calculations.
2.2 Atmospheric dispersion models and boundary layer scaling The CARFMI (Contaminants in the Air from a Road, by FMI) model describes dispersion from a number of line sources. The model estimates hourly time series of the concentrations of carbon monoxide, and can be used on a Cray XMP supercomputer. The mainframe version of the model is vectorized and parallellized, which enables computation with a large number of line sources and receptor points. It also computes statistical parameters from the time series, which can be compared to air quality standards and measurements. Dispersion parameters for the model are evaluated using stability data produced by a meteorological preprocessing model. This model has been developed at our institute and is described in detail by Nordlund et al. (1994). The meteorological parametrization is based mainly on the energy budget method by Van Ulden and Holtslag (1985). The model estimates the hourly time series of the relevant atmospheric turbulence parameters, as the MoninObukhov length scale (L), the friction velocity (u*), temperature scale (q*) and mixing height (h) from the data of routine meteorological and sounding observations. 3 Numerical results and discussion We have evaluated the emissions from the street and road network, and the resulting distributions of carbon monoxide concentrations. In the interpretation of these results, one has to bear in mind the inherent limitations of the modelling approach. For instance, the emission computations do not include the specific vehicle emissions at the street and road junctions, and in serious traffic jams. In a major junction, the emissions may be substantially higher than predicted. The dispersion model does not allow for the detailed structure of buildings and street canyons, and the computed concentration distributions should therefore be intepreted as regionally averaged values. The actual concentrations inside a street canyon may vary by as much as by a factor of ten. All the presented predicted concentrations are groundlevel values. 3.1 Emissions from mobile sources For computational reasons, the emissions from mobile sources have been discretised into line sources. Figure 3 shows the emissions of carbon monoxide in the Helsinki metropolitan area during a typical workday. The emissions from buses not included in the Figure. The emissions are represented as the total daily emission at each linear section of a street or a road (kg/d). The width of each rectangle is proportional to the total emission (as shown in the legend at the upper lefthand corner of the Figure). 
3.3 Concentration distributions in 1993
Figures 6 and 7 show the highest annual hourly COconcentrations in the Helsinki metropolitan area and in downtown Helsinki in 1993. The national air quality standard (from the year 1984) for the hourly maximum COconcentration is 30 mg/m^{3}, and the new proposed national air quality standard (1994) is 20 mg/m^{3}. According to the computations these standard values have been exceeded at a few spatially limited areas, mainly in the downtown area. Figure 7 shows concentrations above 40 mg/m^{3 }in the areas of Kamppi, Töölö and Sörnäinen. The highest value in the area is about 80 mg/m^{3}, located in Sörnäinen, in the northeastern downtown area (near the upper righthand corner of Figure 7).
3.4 Concentration distributions for the assumed scenario
We have computed the CO concentrations for one assumed emission scenario. It was assumed that all the passenger cars have been equipped with a catalytic converter. In all other respects, the emissions, the street network and the meteorological conditions were assumed to be the same as for the above mentioned base case in 1993.
This scenario therefore describes the potential influence of the introduction of the catalytic converters only. It neglects all the changes in emissions due to, for instance, the changes in the fuel content and the improvements in the engine technology. Further, this scenario does not allow for the influence of the actual weather conditions of the year considered. However, at least the present experience shows that the influence of the introduction of catalytic converters on the traffic emissions is substantially larger than the respective influence of the changes in the fuel content and the engine technology.
The emissions of CO for cars decrease substantially, with the introduction of the catalytic converter (see Figure 1). For the assumed scenario, the emissions of busses and other heavy vehicles are therefore much more important, compared to the base case computations.
Figures 8  11 present the corresponding catalysatorscenario concentration distributions. The concentration levels are clearly lower and also the distribution of concentration differs now clearly from the earlier calculations. The growing importance of the emissions from buses can particularly clearly be seen at the Kamppi area, where the main busstation is located. The highest yearly mean concentration of 110 mg/m^{3} is located in this area, while it is not specifically distinguishable in the 1993 calculations.
3.5 Comparison of the concentration distributions with air quality standards
Table 1 presents a comparison of the computed maximum hourly concentrations with the national air quality standards, both for the base case and for the assumed scenario.








The measurement network and the methods have been described by Aarnio
et. al. (1993). Table 2 shows basic information on these two stations.









(i) The measurement heights range from 3.5 to 5 m, and the predicted concentrations considered are groundlevel values. The measured and predicted concentrations are therefore not strictly comparable.
(ii) Both measurement sites are located near densely trafficked street junctions. This tends to cause an underprediction of the emissions.
(ii) The immediate environment of both measurement sites is densely built. The dispersion model does not allow for the detailed structure of buildings and street canyons. The environment of the measurement sites tends to lead to a reduced dispersion, compared to dispersion in spatially uniform terrain. The computed concentrations are therefore expected to be underpredictions, even if the emissions are known accurately.
The monthly averaged predicted and measured concentrations have been presented in Figure 12. The measured concentrations are clearly higher at both stations, and the differences are largest at the station in Töölö. This is to be expected, when one takes into account the location of this station at a densely trafficked junction. However, the annual variation of the measured and predicted concentrations is very similar.
Figure 13 shows the correlation coefficients of the model predictions and the measurements, computed separately for each month in 1993. During the winter season (from October to March) the correlation between the predicted concentrations and the measurements is qood; correlation coefficients range from 0.4 to 0.6. However, during the summer season (from April to September), the correlation is fairly poor.
This behaviour is at least partly caused by the modelling of the emissions. The applied trafficplanning model applies March as the socalled reference month, resulting in more realistic emission predictions during the winter season.
5. Conclusions
We have made an emission inventory of the mobile sources in the Helsinki metropolitan area. Atmospheric dispersion was subsequently evaluated, resulting in hourly time series of concentrations of carbon monoxide in 1993.
The predicted results show that the highest CO concentrations occur in the Helsinki downtown area and at the junctions of the main streets and roads. Comparison of the model predictions with the national air quality standards shows that the standard values for the hourly concentrations are exceeded at a few spatially limited areas, mainly in the downtown area. According to the computations concentrations above 40 mg/m^{3 }occurred in the areas of Kamppi, Töölö and Sörnäinen. The highest hourly concentration value in the area was about 80 mg/m^{3}, located in Sörnäinen, in the northeastern downtown area.
We have computed the CO concentrations for one assumed emission scenario. It was assumed that all the passenger cars have been equipped with a catalytic converter. The emissions of CO for cars decrease substantially, with the introduction of the catalytic converter. For the assumed scenario, the emissions of busses and other heavy vehicles were therefore much more important, compared to the base case computations. The concentration levels were clearly lower for this scenario, compared to the base case in 1993, and the distribution of concentration was also clearly different from the base case computations. The concentrations in this scenario were predominantly below the current air quality standards.
We have compared the model predictions against the air quality measurements at two stations of the YTV in the Helsinki downtown area. These comparisons are preliminary, and the results should be considered mainly qualitatively. Both measurement sites are located near densely trafficked street junctions. This tends to cause an underprediction of the emissions. The immediate environment of both measurement sites is densely built. The dispersion model does not allow for the detailed structure of buildings and street canyons. The computed concentrations are therefore expected to be underpredictions, even if the emissions were known accurately.
The measured monthly averaged concentrations were clearly higher at both stations, compared to the predicted concentrations. However, the annual variation of the measured and predicted concentrations was very similar. During the winter season (from October to March) the correlation between the hourly predicted concentrations and the measurements was qood, while during the summer season (from April to September), this correlation was fairly poor. This behaviour can at least partly be explained by the inaccuracies in the modelling of the emissions.
We have considered here only one gaseous compound, carbon monoxide.
In the next stage of this project, we will address the concentrations of
nitrogen oxides. We have also planned to conduct more detailed dispersion
computations using street canyon dispersion models, in the vicinity of
the measuring stations. The results presented here will then provide the
background concentrations for these nested model computations.
References
Aarnio, P., Hämekoski, K. and Koskentalo, T. (1993) , Ilmanlaatu pääkaupunkiseudulla vuonna 1992, PJS C 1993:8, YTV,Helsinki.
Härkönen, J., Valkonen, E., Kukkonen, J., Rantakrans, E., Jalkanen, L. and Lahtinen, K., 1995. An operational dispersion model for predicting pollution from a road. International Journal of Environment and Pollution, Vol. 5, Nos. 46, 602  610.
Laurikko, J., 1994, Valtion Teknillinen Tutkimuskeskus, Oral information.
Liikenneministeriö, 1994, Moottoriajoneuvojen kulutus ja päästöt liikenteessä, Liikenneministeriön julkaisuja 42/94.
Mäkelä, K.; Anila, M. & Kuusola, J., 1994, Henkilöautojen kylmäkäyttö, Valtion Teknillinen Tutkimuskeskus, Tie, geo ja liikennetekniikan laboratorio, Tutkimusraportti 192.
Mäkelä, K.; Kanner, H.; Himanen, V., Laurikko, J. & Anila, M., 1991, Tieliikenteen pakokaasupäästöt, Ennustava tietojärjestelmä LIISA 2.1, Valtion Teknillinen Tutkimuskeskus, Tie, geo ja liikennetekniikan laboratorio, Tutkimusraportti 67.
Mäkelä, K., Kanner, H., Himanen, V., Laurikko, J., Salusjärvi, H. and Anila, M. (1990) ‘The emissions from traffic: A mathematical model LIISA II (in Finnish)’, VTT / Traffic laboratory, Research report 826, Technical Research Center of Finland, Espoo, 57 p.
Luhar, A. and Patil, R., 1989. A General Finite Line Source Model for Vehicular Pollution Dispersion. Atmos. Environ. 23, p. 555  562.
Nordlund, G., Rantakrans, E., Karppinen, A. and Valkama, I. (1994) ‘The short range air pollution dispersion calculation model used by the Finnish Meteorological Institute’, Finnish Meteorological Institute, Report, Helsinki.
Petersen, W., 1980. User's guide for HIWAY2, a highway air pollution model. EPA600/880018, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina, 69 p.
Spiess, H. & Suter, D, 1990, Modelling the daily traffic flows on an hourly basis, PTRC Transport and Planning Summer Annual Meeting, University of Sussex, England, September 1990, p. 301.
Van Ulden, A. and Holtslag, A. (1985) ‘Estimation of atmospheric boundary layer parameters for diffusion applications’, J. Climate Appl. Meteor. 24, p. 11961207.
Yamartino, R. and Wiegand, G. (1986) ‘Development and evaluation of simple models for the flow, turbulence and pollutant concentration fields within an urban street canyon’, Atmos. Environ. 20:11, p. 21372156.
YTV, 1990, Liikennemallit, Pääkaupunkiseudun liikennetutkimus,
Pääkaupunkiseudun julkaisusarja B 1990:15.