Air pollution exposure monitoring and estimation Part V.‡ TraYc exposure in adults† Alena Bartonova,* Jocelyne Clench-Aas, Frederick Gram, Knut Erik Grønskei, Cristina Guerreiro, Steinar Larssen, Dag A. Tønnesen and Sam-Erik Walker Norwegian Institute for Air Research, PO Box 100, N-2027 Kjeller, Norway. E-mail: alena.bartonova@nilu.no; fax:+47 63 89 80 50 Received 7th April 1999, Accepted 25th June 1999 In Oslo, traYc has been one of the dominating sources of air pollution in the last decade.In one part of the city where most traYc collects, two tunnels were built. A series of before and after studies was carried out in connection with the tunnels in use. Dispersion models were used as a basis for estimating exposure to nitrogen dioxide and particulate matter in two fractions.Exposure estimates were based on the results of the dispersion model providing estimates of outdoor pollutant concentrations on an hourly basis. The estimates represent concentrations in receptor points and in a square kilometre grid. The estimates were used to assess development of air pollution load in the area, compliance with air quality guidelines, and to provide a basis for quantifying exposure-eVect relationships in epidemiological studies.After both tunnels were taken in use, the pollution levels in the study area were lower than when the traYc was on the surface (a drop from 50 to 40 mg m-3). Compliance with air quality guidelines and other prescribed values has improved, even if high exposures still exist. The most important residential areas are now much less exposed, while areas around tunnel openings can be in periods exposed to high pollutant concentrations.The daily pattern of exposure shows smaller diVerences between peak and minimum concentrations than prior to the traYc changes. Exposures at home (in the investigation area) were reduced most, while exposures in other locations than at home showed only a small decrease.Highest hourly exposures are encountered in traYc. for comparison of air quality in the area before and after the 1. Introduction tunnels were built. In Oslo, traYc has been one of the dominating sources of air The investigations are designed as a sequence of a crosspollution in the last decade. The traYc authorities implemented sectional and a diary study.This method provides a powerful plans for reducing the impact of traYc on the urban environ- tool for describing patterns in exposure. The cross-sectional ment, especially in those parts of the city where most traYc study provides a detailed description of exposure in the collects. One of these parts is the Old Oslo/Oslo East, where residential area (study area proper), where the individuals the traYc was directed from the surface into tunnels.In spend most of their time. However, if the whole pattern of connection with taking the tunnels in use, three environmental exposure is to be established, information is needed about studies were carried out between 1987 and 1996 individual’s movements through all environments, including (Kolbenstvedt1). They served as before and after studies to home, at work, in traYc and other places.Only by evaluating two separate tunnel projects. Air pollution and noise were the the whole pattern of exposure, can eVective measures be central environmental indicators. Impact parameters that were suggested that would prioritise alleviating the situation least studied were annoyance, symptoms of reduced health and well advantageous from the point of view of human health or other being and chronic diseases.environmental concerns. The investigations aimed at assessing the impacts of traYc changes in a geographically small study area with substantial 2. Aim of the study pollution gradients. A method for estimating personal exposure had to provide a suYciently exact individual estimate of The study aimed at providing estimates of personal exposure pollution exposure to reflect the gradients well.It was therefore to nitrogen oxides and particulate matter for a representative natural to use dispersion models as a basis. sample of population in a restricted study area. These estimates In Oslo, traYc and spatial heating are the principal source were to reflect the pollution gradients in the study area.of air pollutants. The primary compounds of interest are Further, the estimates were to be used to provide a description therefore nitrogen oxides (NO2 and NOx) and fine and coarse of impacts of traYc alleviating measures, studied in a series particulate matter (PM). In Norway, no significant indoor of before and after studies. The estimates were to serve as a sources of these compounds exist, except for smoking (a source basis for establishing if air quality guidelines were satisfied, of particulate matter, Braathen2).It is therefore reasonable to and were to be used to quantify exposure-eVect relationships assume that personal exposure reflects well outdoor concen- between air pollution and a variety of eVect variables related trations.Estimates of outdoor concentrations can be used to to health and well being. construct suitable environmental indicators. The same air pollution estimates can be used both as indicators of personal 3. Methods exposure in a health and annoyance study, and as descriptors The three consecutive environmental surveys carried out in 1987, 1994 and 1996 have been performed as cross-sectional †Presented at AIRMON ’99, Geilo, Norway, February 10–14, 1999.‡For Part IV, see ref. 15. epidemiological studies, with a subsequent diary study. The J. Environ. Monit., 1999, 1, 337–340 337surveys were carried out using almost identical design and The traYc measures were taking in use the newly built tunnels, restricting or closing roads for through-traYc, lower- methods, and the data were pooled for the final analysis.Each individual was assigned an estimate of air pollution exposure ing of speed limits on some road links, and increasing speed limit on the main throughway. and an estimate of exposure to noise, both assessed outside the individual’s home. In addition, the collected data include 3.3 Dispersion calculations and exposure estimates individual personal information about health, well being and perception of the environmental quality in the survey area, The dispersion model EPISODE (Grønskei et al.,6 Larssen and about mobility within the area and in Oslo.et al.,7 Walker8 and Walker et al.9) was used in this study. The model is a combination Lagrangian–Eulerian urban scale 3.1 Design of the study model, and provides estimates of concentration fields or recep- The surveys were designed as two-step epidemiological studies.tor point concentrations on an hourly basis for NO2, NOx , In the first step, a cross-sectional design was used to provide PM2.5 and PM10. Inputs to the model are meteorological data, a snapshot of the situation in the investigation time-point emission inventories and background pollution concentrations.(Klæboe et al.3 and Clench-Aas et al.4). For each of the three Walker et al.10 describe the model validation. consecutive years, a sample of over 1100 individuals was In order to improve the emission database for traYc in the selected for an interview carried out either by personal visit main study area, traYc counts were performed and traYc (1987), or over the phone (1994, 1996).The sample was density was estimated for each road link in the 2×3 km geographically stratified. investigation area (Hanssen and Grue11 and Hanssen12). In the second step, a sub-sample of volunteers was recruited TraYc in the area is then described as line sources. Outside from the participants in the cross-sectional study.They were the study area, a lesser-resolution traYc database for Oslo asked to fill out a diary with hourly resolution, for a period was used, with main roads as line sources, and smaller roads of 2–3 weeks. The diary contained information about the summarised as area sources. Other emission inventories include participant’s whereabouts, their activity and their well being.ship and air traYc and spatial heating defined as area sources, For each individual in the study, the home address and each and individual larger point sources. Detailed emission invenaddress indicated in the diary were recorded and coded into a tories with hourly resolution were available for NOx, PM2.5 co-ordinate system. Within the study area proper, the position and PM10 for each year.Actual meteorological data were also of the centre of the building’s fac�ade defined a receptor point. available from a combination of measurements and models, This coding was performed with a 5 m resolution using a map- on an hourly basis for the modelling periods. based GIS program. The air pollution and traYc noise estimates The exposure estimates used for the cross-sectional investiwere calculated for each such receptor point (Kolbenstvedt1 and gation (STINEX model, Clench-Aas et al.5), were for each Clench-Aas et al.5).For the addresses indicated in the diary, receptor point and year based on estimated hourly concenthat were outside the study area proper, the addresses were trations of air pollutants over a period September through coded into the co-ordinate system with a precision down to December, thus including the interview period (usually in 100 m.This was because less exact information about these October). In addition, this period is usually fairly representalocations was available, and in addition, the input data for the tive for calculating annual concentrations. Actual meteorologiexposure models were less detailed. For these additional receptor cal data were used for each hour.The exposure estimate did points, exposure to air pollution was calculated. not take into account the floor on which the actual home is located, or information on orientation of the main living and 3.2 The main study area and traYc measures sleeping quarters relative to the neighbouring streets, despite The approximately 2×3 km core study area in central Oslo, the availability of this information. where the participants in the environmental survey lived, For the diary study, exposure was estimated for each collects a substantial part of through traYc towards the individual at each indicated location for each hour (DINEX suburbs east of the city and to the north of the country (see model, Clench-Aas et al.5).For locations within the study Fig. 1). Within the area, 8 sub-areas in 1987 and 14 sub-areas area proper, the estimate was the value at the receptor point. in 1994 and 1996 were identified on the basis of the planned For locations in Oslo, but outside the study area proper, the traYc changes. The sub-areas included background areas with estimate represented average concentration in the appropriate not much expected traYc changes, areas where traYc was to grid square, for a given hour.For estimates of exposure while drop dramatically, areas where the traYc was to increase, and in traYc and while shopping, representative grid squares were areas situated by the tunnel openings. selected, describing squares with low, medium and high traYc density.If an individual was at several locations for a given hour, the exposure was represented by a weighted average of estimated outdoor concentrations at each of the locations. 3.4 Selected indicator compounds In both the cross-sectional and the diary study, the main pollutants of interest are nitrogen dioxide (NO2) and particulate matter (PM). In Oslo, particulate matter is emitted from point and area sources connected to spatial heating, and from exhaust of the vehicular traYc.In addition, the use of studded tyres also generated particulate pollution in the winter half year. In this study, particulate matter was considered from all the sources. The individual sources contribute by diVerent amounts to the total mass in the two fractions, PM2.5 and PM10.A special feature of the model made is that it is also possible to single out the studded tyres contribution. In the cross-sectional study, four compounds were estimated: NO2, Fig. 1 Map of Oslo with an enhanced view of the study area with PM2.5, PM10 and PM10–2.5. In the diary study, results are sub-areas. Scattered line marks the main roads from 1987, the broken line with two dots indicates the tunnels.shown here only for nitrogen dioxide. 338 J. Environ. Monit., 1999, 1, 337–3403.5 Air quality indicators Air quality indicators in this study may be defined quite freely, as hourly pollution estimates by a dispersion model are available for each receptor point. For the cross-sectional study, indicators included for each receptor point are the arithmetic mean, 98th percentile and maximum concentration during the calculation period.These statistics were for NO2 based on hourly concentrations, and for particles, on daily concentrations (although the model provides hourly data). In this way, the average concentration is a suitable indicator for the epidemiological study, the maximum provides a basis to compare with Norwegian legislation, and the 98th percentile provides a robust indicator of changes in the occurrence of high concentrations.Fig. 2 Cumulative distribution functions of individual period exposure 4. Results to NO2 in 1987, 1994 and 1996. 4.1 Cross-sectional study In the three cross-sectional investigations, the 3193 individuals air quality guideline (100 mg m-3 hourly NO2, 70mg m-3 that participated have lived at 488 distinct addresses divided daily PM10). into 14 sub-areas.The sub-areas were chosen along specified Fig. 3 shows that a certain improvement in compliance has road links, based on expected traYc development at these been observed. Unlike the description of the development in links. The sub-areas included areas with expected massive personal exposure above, this assessment is based on the improvement in traYc burden, areas with no expected change development of receptor points, using all existing receptor and areas with expected worsening of the situation.points encountered in the three years, and not only the receptor In the study area as a whole, the number of vehicle- points present at a given year. kilometres has increased from 639 000 in 1987 to 854 000 in 1996 and the traYc speed has somewhat increased, as the 4.2 Diary study main throughway has an increased speed limit.In the whole Individual exposure may be described in detail using a diary of Oslo, however, the traYc volume seems to be unchanged method. Here, information is recorded every hour as to where (ref. 13). the individual has been.From this account, detailed patterns The average exposure to NO2 (Table 1) has decreased from can be constructed, but most importantly, it can be assessed 51 to 40 mg m-3. A similar decrease is seen for particulate matter in both fractions, however, the fraction PM10–2.5 attributed to studded tyres did not decrease. In order to judge reasons for this development, it is necessary to consider meteorological conditions in some detail, but this information can not be easily evaluated. In general, the most favourable dispersion conditions were in 1994, while the least favourable were in 1987.The dispersion of the PM10–2.5 fraction is mainly influenced by wetness conditions of the road surface, and diVerences in the wetness explain the development in this indicator.Indicators of high but less frequent exposure, the 98th percentiles, show a similar decrease as for the average values. However, Fig. 2 shows that highly exposed homes still exist in the latter years. These are now mostly located near the tunnel outlets as opposed to the situation in 1987, when the highest exposures occurred more often and along the most traYcked road segments.Fig. 3 Receptor points (%) above the action limit (ACT), the mapping The estimated exposure at a home address may be used to limit (MAP) and the national air quality guideline (NGL). assess compliance with air quality guidelines and other reference values. In Norway, at the time of the study, the following air quality limits were set: Action limit (300 mg m-3 hourly NO2, 350 mg m-3 daily PM10), mapping limit (200 mg m-3 hourly NO2, 300 mg m-3 daily PM10), and suggested national Table 1 Development of exposure at home addresses.Numbers represent averagesver the 3 month period of calculation 1987 1994 1996 Exposure parameter mg m-3 mg m-3 mg m-3 Average hourly NO2 51 43 40 98th percentile NO2 136 97 101 Average daily PM10 30 18 20 98th percentile PM10 121 61 75 Average daily PM2.5 21 13 12 Fig. 4 Frequency distribution of hourly exposure to NO2 in a diary 98th percentile PM2.5 64 29 45 study, 1987 and 1994.J. Environ. Monit., 1999, 1, 337–340 339Table 2 Average exposure value for the two years with diary and therefore the source of the changes was not identified with average exposure duration at five types of locations, 1987 and 1994.certainty. Based on diary study The estimates of exposure outside individual homes and for each type of location, provided in this study, are an important Estimated average hourly NO2 Average input to epidemiological studies. This complex information concentration hours spent (%) can provide a basis for a socio–economic assessment of impacts 1987 1994 1987 and 1994 of large road projects, or other important changes.In addition, Home 23 19 70.8 the results provide a good approximation of outdoor pollution At work 21 36 14.2 levels, so that the eVect of changes can be evaluated against At school 31 14 2.8 the current legislation. The methodology provides a result that Other places 31 15 8.5 can be used flexibly to address diVerent scientific and regulat- While in traYc 70 40 3.8 ory needs.The study was funded by the Royal Norwegian Council for Scientific and Industrial Research/Norwegian Research Council, The Directorate of Public Roads, and the Norwegian Institute for Air Research. The study was co-ordinated by the Institute of Transport Economics in Oslo, Norway. References 1 M. Kolbenstvedt, Environmental consequences of main road diversion in Oslo East.Summary of studies 1987–1996 (in Norwegian) Institute of Transport Economics, Oslo, (TØI report 405/1998), 1998. 2 O. A. Braathen, Results of indoor/outdoor measurements, in NILU/NIPH. Air pollution and short term health eVects in an industrialised area in Norway, main report, Norwegian Institute for Air Research (NILU OR 81/91), Lillestrøm, 1991.Fig. 5 Daily pattern of exposure to NO2, diary study 1987 and 1994. 3 R. Klæboe, M. Kolbenstvedt, J. Clench-Aas and A. Bartonova, A holistic approach to assess traYc measures, in 8th International Symposium on Transport and Air Pollution, ed. P. J. Sturm, Technical University Graz, Graz, Report of the Institute for as to where to most eVectively direct measures to reduce Internal Combustion Engines and Thermodynamics, 1999, impact of outdoor air pollution. vol. 76, pp. 7–14. 4 J. Clench-Aas, A. Bartonova, M. Kolbenstvedt and R. Klæboe, In 1987 and 1994, 114 and 118 persons under 60 years Quantifying eVect of traYc measures using individual exposure mod- participated in the diary study, with a typical participation of elling, in 8th International Symposium on Transport and Air between 2 and 3 weeks.Similarly as in the cross-sectional part, Pollution, ed. P. J. Sturm, Technical University Graz, Graz, exposure to NO2 was lower in 1994 than in 1987, with a lower Report of the Institute for Internal Combustion Engines and percentage of hours registered with high exposures (Fig. 4). Thermodynamics, 1999, vol. 76. 5 J. Clench-Aas, A.Bartonova, T. Bøhler, K. E. Grønskei and The exposure may be classified as to where the participant S. Larssen, J. Environ. Monit., 1999, 1, 313. is situated: at home, at work/school, in traYc and other places. 6 K. E. Grønskei, S. E.Walker and F. Gram, Atmos. Environ., 1993, Exposure at a home address and in traYc was greatly reduced, 27B, 105. while for this particular population, exposure at work (i.e. 7 S.Larssen, K. E. Grønskei, F. Gram, L. O. Hagen and S. E. outdoor concentrations at a work address) increased somewhat Walker, Verification of urban scale time dependent dispersion model with subgrid elements in Oslo, Norway, in Air pollution modelling (see Table 2). While the work addresses in the two populations and its application X, ed. S.E. Gryning and M. M. Millan, Plenum may not be comparable, the home addresses are from the Press, New York, 1994, pp. 91–99. same general area, and the traYc exposure was estimated 8 S. E.Walker, The EPISODE air pollution dispersion model, version using the same reference traYc points. Thus, the results 2.2. Users Guide, Norwegian Institute for Air Research (NILU indicate a real decrease in the exposures.TR 10/97) Kjeller, 1997. 9 S. E. Walker, L. H. Slørdal, C. Guerreiro and K. E. Grønskei, The exposure can also be investigated for daily patterns. Development and evaluation of the urban dispersion model This is important for building epidemiological models, but EPISODE used in evaluating traYc diversion measures in Oslo, 8th also gives information about development in exposure in International Symposium on Transport and Air Pollution, general.Fig. 5 shows that the peak exposures were reduced, Technical University Graz, Graz, Report of the Institute for and that the diVerence between peak and saddle exposure, Internal Combustion Engines and Thermodynamics, 1999, vol. 76, pp. 25–32. including the night time exposure, are lessened in 1994. 10 S. E. Walker, L. H. Slørdal, C. Guerreiro, F. Gram and K. E. Grønskei, J. Environ. Monit., 1999, 1, 321. 5. Conclusions 11 J. U. Hanssen and B. Grue, Environmental studies Ekeberg/Old Oslo 1994. TraYc system, traYc registering and road links register. The results show the information that can be provided about (in Norwegian), Institute of Transport Economics (TØI note individual personal exposure to outdoor air pollution, when 1055/1996), Oslo, 1995. employing diVerent study designs. The results also illustrate 12 J. U. Hanssen, After studies Ekeberg tunnel 1996. TraYc system, traYc registering and road links register (in Norwegian), Institute the possibilities that lie in using dispersion models as a basis of Transport Economics (TØI note 993/1995), Oslo, 1996. for personal exposure assessment. 13 Information body for traYc, Opplysningsra°det for trafikk AS: Car It has been illustrated here that the expected changes in and road statistics 1996 (in Norwegian), Oslo, (Publ. Nr. traYc burden have occurred, and that despite an increase in 1000–96), 1996. traYc volume, the pollution levels locally in the area have 14 S. Larssen and L. O. Hagen, Air quality in Norwegian cities. Development, reasons, measures, future (in Norwegian), dropped. The diary shows also that levels in Oslo generally Norwegian Institute for Air Research (NILU OR 69/98), Kjeller, have been decreased, and this is confirmed by measurements 1999. (Larssen and Hagen14). 15 J. Clench-Aas, A. Bartonova, K. E. Grønskei and S.-E. Walker, The method in principle allows one to investigate to what J. Environ. Monit., 1999, 1, 333. degree the individual factors influence the result. The model has been run for each year using the actual meteorology, and Paper 9/02780G 340 J. Environ. Monit., 1999, 1, 337–340