Human exposure to urban air pollution.

This study deals with some methods of making human exposure estimates, aimed at describing the human exposure for selected air pollutants in Sweden that are suspected carcinogens. Nitrogen oxides (NOx) have been chosen as an indicator substance for estimating the concentration of the urban plume. Earlier investigations have shown that the traffic in Swedish cities contributes around 85% to the measured NOx concentrations, and that most of the mutagenicity in urban air originates from traffic. The first section of this paper describes measurements in Stockholm of some unregulated light hydrocarbons, such as ethene, ethyne, propane, propene, butane, and isobutane. In addition, measurements of some volatile aromatic hydrocarbons are presented. Simultaneous measurements of carbon monoxide (CO) were made. The ratios between CO and the individual specific compounds were determined by linear regression analysis. By analysis of relationships between CO and NOx, NOx concentrations can be used as a tracer to describe the exposure for these specific compounds. NOx are considered to be a better tracer than CO, because NOx or NO2 values exist for many places over a long time, while CO is measured mostly in streets with high concentrations. At low concentrations, instruments that measure normal CO levels give no detectable signals. Through use of atmospheric dispersion models and models that describe how people live and work in urban areas it has been possible to describe the average exposure to NOx in cities of different sizes. The exposure to NOx for people living in the countryside has also been estimated. In this way, it has been possible to calculate the average exposure dose for NOx for the Swedish population. This figure is 23 micrograms/m3. By use of the relationships between NOx and specific compounds the average dose has been calculated for the following compounds: polyaromatic compounds (PAH); ethene, propene, and butadiene; benzene, toluene, and xylene; formaldehyde and actaldehyde; nickel, chromium (VI), arsenic, and cadmium; asbestos; and silicon.


Introduction
Human exposure to urban air pollution can in principle be estimated in three ways: a) the concentration of hazardous substances in the urban air in different environments can be recorded or modeled together with movements of individuals or population groups; b) personal samplers can be used; and c) the target dose in man can be determined, e.g., blood.
Determining human exposure is a very complex task, and most attempts to estimate it in risk assessment articles rely on literature data on normal pollutant concentration. This probably results in overestimations of average exposures, since measurements are made mostly in areas where there is a reason to survey the air quality. The study reported here also uses commonly available data and makes corrections for such biases.
The steps from emission to human exposure consist of source emissions, transport and transformation, accumulation in the environ-This paper was presented at the Symposium on Risk Assessment of Urban Air ment (indirect uptake via food), and human exposure.
A complex task is the measurement of volatile organic compounds and polyaromatic compounds in urban air, the latter both in the gas phase and the solid phase. These analyses are expensive and therefore are made to a limited extent. However, if major air pollutants, such as CO and nitrogen oxides (NOX) are measured in a more continuous way, and these specific pollutants are measured during a limited time, it is possible to get quotas between NOX and specific pollutants like ethene.
In this estimate, the exposure of the Swedish population to NOx and carcinogens in urban air has been calculated using NOx as an indicator substance. The results will form the basis ofcancer risk assessment.

Air Quality Data
Air quality data in Sweden are known from point measurements and from dispersion modeling of more or less well known emissions. Measured data exist for sulfur dioxide, soot (black smoke), and nitrogen oxides in air, but data on the genotoxic substances in which we are interested are rare. It is still possible to estimate the concentrations of the genotoxic components by assuming that the ratios between the substances are nearly constant. This is not unreasonable, considering the relatively large contribution from car traffic and the moderate dispersion time and room scales involved. The ratio between different components in car exhausts is not likely to change notably during its residence time in the city air (about one hr). Most atmospheric reactions are too slow to alter the ratio more than marginally.
Despite the negative correlation between black smoke and NOx during a drive cycle of an individual car, the mixed pollution from many cars being run at various conditions may still show a positive correlation between black smoke and NOX (1).
If we thus have determined the mean exposure of the population to NOx, we can calculate the mean exposure of the population to another component by simply multiplying the NOx value with the mean ratio between the component and NON. The reason for selecting NOq as a tracer for genotoxic substances is that some 80 to 90% of the NOX in urban air is generated from traffic and that most of the NOX and genotoxic substances in urban air originate from sources within the own urban complex. For other substances, such as SO2, a large part may originate from long-range transport.
Environmental Health Perspectives 39 The concentration of NO. in urban areas is seldom measured in small-and medium-sized cities. Only in a few of the larger cities are there data series from longer periods. However, in about 60 Swedish cities, the winter mean of NO2 has been determined in the city centers over a 5-year period using a sodium arsenite method (2), and on five regional background stations as winter and yearly means (

Results and Discussion
Our results are based on 1-hr mean values. The investigation shows that the heavy traffic sites had an increased concentration of all hydrocarbons, compared to the low traffic sites, i.e., 5 to 10 times. Markedly higher levels of the unsaturated hydrocarbons, ethene, ethyne, and propene were found here, compared with the saturated compounds. The contribution of traffic to various hydrocarbons was investigated using CO as an indicator substance for the traffic.
The average values for the measured unsaturated compounds in the heavy traffic sites were in the magnitude of 10 to 20 ppb, and for the saturated compounds 1 to 10 ppb. A strong correlation indicates origin from the same source. Strongest correlation (r) was found at the high-traffic sites, the unsaturated HC having stronger correlation than the saturated HC. Propane was the only HC in which the regression equation had a significant positive intercept, i.e., a positive concentration at zero CO. This indicates that propane occurs in the background air of Stockholm. The intercept was highest during the winter, lower in spring, and not significant  Abbreviations: C, central (C+) and noncentral (C-); D, high-diesel (D+) and low-diesel (D-) location; T, high-traffic (T+) and low-traffic (T-) flow location; rev, revertant.
Environmental Health Perspectives in the summer. The lower intercept during the warmer period of the year may reflect the higher rate of photochemical degradation and a better air mixture than in the winter. The divergent observations of HC were found in the time series by study of the difference between the actual value and the value given by the regression equation, with CO as input data. In this way, the observations could be extracted from vehicular influence. At sites near dense traffic, all HC showed a strong correlation to CO, which can be assigned solely to the vehicle traffic. The good covariance in time, hour for hour, is exemplified in Figure 1 with ethene and CO over 10 days.

Polycydic Aromatic Compounds
Among the several thousand compounds present in ambient air are the polycyclic aromatic compounds (PAC), of which some are known carcinogens and suspected mutagens (3). The most important sources in Swedish cities are mobile-light-and heavy-duty vehicles emitting exhaust containing PAC. Results presented below emanate from samples collected at several locations in the Gothenburg urban area. Sampling, chemical analysis, and results are presented in detail in a report published in Agurell et al. (1). The results from Gothenburg are from areas with traffic densities denoted as central (C+) and noncentral (C-) locations, high-diesel (D+) and low-diesel (D-) locations, and high-traffic (T+) and low-traffic (T-) flow locations. In Table 1  The T-sampling series (high-traffic-flow location) was measured for 18 months with a time resolution of 1 month. In Figure 2, the variation of PAH concentrations is presented, from which it can be seen that during the winter an increased PAH concentration can be measured a factor ofapproximately three to five times that of the summer. A possible explanation for the increase is that during winter automobile engines are running on richer fuel-air mixtures, and contributions are coming from domestic heating and meteorologic conditions. In addition to PAH measurements, 1nitropyrene also was measured; it showed a larger variation than PAH (Figure 2), indi-  cating a significant impact from atmospheric transformation on measured 1-nitropyrene concentrations.

Summary
A seasonal variation of a factor three to five has been determined for the PAH concentration in Gothenburg ambient air samples, i.e., winter/ summer. Average concentration determined in Gothenburg ambient air for PAH was 20 ng/m3, 1-nitropyrene 10 ng/i In central Stockholm, the vehicle traffic is an important source of some light nonmethane hydrocarbons (LNMHC). The investigation shows that carbon monoxide can be used as a tracer in the urban air for LNMHC originating from vehicle traffic. The quantitative relationship between CO and individual HC was rather stable for unsaturated HC, independent of traffic intensity and time of year. Saturated HC showed greater seasonal variations.
From this investigation we learned that a simple way to grasp the pollution pattern from traffic is to study intensively the pollutants directly near dense traffic during a few days and nights. The time resolution should be at least one hr, to detect, among other factors, the rush hours. The covariance in time between different compounds gives valuable information on the stability of the pollution pattern and also on the occurrence at other sources than traffic. If CO is selected as tracer for traffic, the quantitative relationship toward the individual compound is made by linear regression analysis. Exposure doses for citizens to the different components can then be made indirectly with the use of representative data for CO doses.

Method of Estimating Human Exposure
An overview of methods used today to estimate human exposure to air contaminants was made by Omstedt and Szego (6). In most cases, except when directly measuring the biologic dose, the methods are based on the identification of a number of microenvironments, with similar exposure conditions occurring within each microenvironment. The total dose is calculated on the basis of how long people stay in different microenvironments. An example of time-activity patterns in the western world is shown in Table 2.
Exposure data are estimated by the use of measurements and/or models. Measurements are done either by exposure measurements or by measurement of the biologic dose. Exposure measurements may be done by personal monitors or by stationary monitors. In the latter case, one has to know the activity patterns of the population studied.
The models used can be divided into three classes: physical, statistical, and physical-statistical. The borders between these classes are not clear and the classification is based more on typical conditions than on the absence or presence of physical or statistical elements in the models. In our study, we have used a physical-statistical model together with air quality data from stationary sources. The mean exposure in a city is determined from the expression: C=KXK2XK3xK4X C(N02) [1] where K1 is the ratio of the mutagenic substance and NOX; K2 is the ratio of winter and year means; K3 is the ratio of population exposure mean and central city mean; K4 is the ratio of NO. and NO2.
The NOX concentration is thus chosen as a tracer for the general pollution level in the urban plume. The reason for this is explained in the section, "Air Quality Data," but it may also be expressed as an attempt to decrease the overall error in estimation of the average exposure using available data. The correlation between the various parameters mentioned in the formula above may be low, but the precision in estimating the mean ratio (K1 to K4) increases considerably with the number of observations available.
One may also regard the method of estimating average exposure as a method of estimating a correction factor that transforms the average concentration from a limited number of measurements of a substance to the grand population average. This correction factor is the ratio between the grand average of population NOX exposure and the average of NOX measured parallel to the substance average.
When estimating risks that are linearly dependent on exposure and when estimating risks for large groups, the individual exposures do not have to be determined. This means that the population exposure can be determined by multiplying the number of people in a certain microenvironment by the time period and the concentration, and successively adding all such exposure data from different microenvironments to get the total exposure in a city or part of the country, without knowing how people move between the different microenvironments.
Each city or rural part of the country is regarded as a microenvironment. For the two largest cities, Stockholm and Gothenburg, transportation is also considered a microenvironment. Each city with an urban population more than 20,000 is treated individually, while urban communities with a population less than 20,000 are treated in groups of 20,000 to 10,000, 10,000 to 5,000, 5,000 to 1,000, and 1,000 to 200, respectively. An urban area is not an ideal microenvironment since the concentration varies within the area, a circumstance that has not been possible to consider in detail except for the city centers. For the rest we have introduced a correction factor for the concentration measured or estimated for the central parts. The correction factor is the ratio between population-weighted mean concentration and central concentration. This factor has been deter-mined by Svanberg (8). The correction fa( for Gothenburg for only its own and 0.68 for the regional back Helsingborg the correction fac and for Orebro 0.61 regarding pollution. A population-weigh these cities would thus be 0.55 f own air pollution, and 0.7 i] regional background. The re ground for Helsingborg and Or assumed to be the same as the ground in Sweden, which is aboi To find out the NOC conce responding to the measured r tration, one has to consider the chemistry of NOX.  (7), Photochemical equilibrium can approxisingborg and mately be expected to occur in rural areas. In ctor was 0.56 urban areas, however, the time necessary to air pollution, mix NO-rich air with background, ozone-rich ground. For air is also important. tor was 0.43, Reaction 3 is dependent on square of the the cities own concentration of NO and is most important Lted mean for where the concentration is high, i.e., close to for each cities' the source. Besides the reactions above, there ncluding the is a certain primary production of NO2 in the gional backsource in the order of a few percent. With a *ebro has been simple box-model including atmospheric e mean back-chemistry and mixing processes designed by ut 20% (2).
Hertel and Bercowics (9), Omstedt has calcu-.ntration cor-lated the NO2 concentration as a function of°4 2 concen-the NO concentration (10). Regional backatmospheric ground concentrations were taken from st important Dahlberg et al. (10) and Larsen (11), and sun radiation data were taken from the Swedish Metorological and Hydrological Institute [Reaction 1] (12); 5% of the NOX have been assumed to exist as NO2 at the point of emission. The [Reaction 2] result is shown in Figure 3. Figure 3 shows two curves. The lower rep-[Reaction 3] resents photochemical equilibrium. In the higher, mixing time in urban areas has been ictions occur, considered. The points are measured data by ut their influa chemiluminescent method from central en NO2 and Gothenburg during 1985. The measurering the first ments were carried out at roof-level. When the average calculating NOX values from NO2 values in Reactions 1 urban areas, the NO2 concentration is multinutes there is plied by a factor of 2.5. In rural areas, this k1 and J are factor is assumed to be 1. 1. nd 2), which Since most of the NO2 data were measured during the winter, it is necessary to know the ratio between the yearly mean and the winter mean. In Gothenburg this [Reaction 4] ratio is 0.6, and this ratio is used for the other cities until further statistics exist. For the rural areas, measured data represent the can be deterentire year (3).

3), k1andj
p. The fact that people spend most of their time indoors must also be considered. No systematic determination of this factor has been made for Sweden, but measurements in United States show indoor/ outdoor ratios of about 0.8 for lead (27). This indicates that a factor of one may be used as a first approximation for most inert substances. However, substances that occur in large particles (in this case silicon) are given a factor of 0.5, representing its penetration into buildings.
As described above, population exposure to NOX can be determined for the general ambient air conditions in urban and rural areas. However, as shown in Table 2 (7). The figures are approximately in agreement, assuming that the number of vehicles in the peak hours is about 10% of the number per day.
Through assistance by the local community planning office, the expression illustrated above was used on all major streets in Gothenburg to calculate the exposure dose to the travelers (5). The exposure was calculated for peak hours, and the rest of the day was divided in work hours, evening hours, and night hours.
For Stockholm, the exposure was estimated by the use of figures from a Stockholm country council investigation of traveling habits during 1990. According to these data, the mean traveling time in cars to work and back is 440,000 hr/day. There are no measurements of the mean concentrations within the cars, but there is no reason to assume a lower concentration than in Gothenburg. The concentrations have thus been assumed to be 0.7 pg/ m3 as a mean. The ratio between the exposure-dose from traveling to and from  work during peak hours and from all traveling in the traffic system during the day was 0.27 in Gothenburg. This is also assumed to be the case in Stockholm.
The exposure-dose in other cities has been neglected. This is thought to introduce only a minor error to the overall exposure estimate since the dose is proportional to the square of the traffic intensity and since the traveling time is much higher in large cities than in small-and medium-sized ones.
The method we have used to calculate the mean exposure thus starts with measurements of winter mean concentration of NO2 in the central parts of 60 Swedish cities, and estimating the concentration in the remaining cities by regression analysis. Then the population-weighted mean exposure of NO. is determined for each city by multiplying with a number of correction factors. Finally, a population-weighted mean for all the country is calculated for NOX and this value is multiplied with various factors expressing the mean ratio between different mutagenic substances and NO, to give the mean exposure to these substances.

Calculations and Results
Mean Exposure to NOX The number of inhabitants, the winter mean of NOX, the corrected population mean exposure of NOX and the total population exposure for each city larger than 20,000 inhabitants and for groups of communities with less than 20,000 inhabitants has been calculated by Steen (12). The corresponding figures are also given for rural areas.  Table 3.
Since the population mean exposure for NO. is 23 pg/m3, the corres3ponding mean 3 PAH exposure was 0.2 x 10x 23 pg/mi which is 4.6 ng/m 3. This is exposure in traffic-dominated areas. In most small cities and villages the concentration of black smoke is higher compared to the NOC concentration, probably because of an increased use of small-scale burning of oil and wood, which also generates some extra PAHs. The ratio of winter means of locally generated concentrations of black smoke and NOx correlate nega-  tively with the number of inhabitants in the different cities (2). There are not enough data, however, to determine this exposure in a direct way. Since we can expect that the contribution from wood burning is significant, an attempt will be made to estimate it from the increased black-smoke concentrations and estimation of the blacksmoke/ PAH ratio. Black smoke is measured and defined by an Organization for Economic Cooperation and Development (OECD) method (28).
According to Larssen, the PAH emission and inorganic or elemental carbon emissions are as listed in Table 4 (14) (elemental carbon is the main constituent in black smoke). The concentration of black smoke compared to NOX is about twice the relative concentration in Gothenburg. If the extra black smoke were to come only from small-scale wood burning it would contain: (40/ 2,000) / ((8 x 0.32 + 2 x 0.68) / (30,000 x 0.32 + 50 x 0.68) = 5.1 times as much PAH per black-smoke unit as in Gothenburg. Of the total fuel consumption for transportation, the use of diesel was 32% and gasoline 68%. If the extra black smoke were to come only from small-scale burning of light oil, the increase would be 1.7 times and for heavy oil in large incinerators 0.0005 times as much.
To weight these figures together, the fuel consumption in Sweden during 1989 is used ( Mean Exposure to Ethene. The mass concentration ratio ethene/CO has been determined by Persson and Almen in Stockholm at Sveavagen to 5.7 x 10-3 and at Homsgatan to 6.1 x 103, and to 8.6 x 103 and 14.9 x 103 in car exhausts (2). Another test on a gasolinepowered car (15) shows a ratio of 12 x 10-. The ratio of CO/NON concentrations has been determined in Gothenburg to be 11.2 (16).
This will give an ethene /NOx ratio of 6.6 x 10using the mean values from Homsgatan and Sveavagen. The ratio ethene /NOC in the urban plume from Gothenburg may be found from measurements made outside Gothenburg by Lindskog (17). It was the 6.3 x 10-2. The ethene /NOX ratio in car exhausts is 3 x 10-2 using the data shown in Table 3.
Using the mean value for the ratios found in ambient air, the mean population exposure may be determined to be 1.5 pg /m3. However, this figure does not indude contributions from woodburning. If we use the same type of reasoning as for PAH and use the ratio of ethene/PAH in emissions from small-scale burning ofwood (about 23) there will be an additional ethene concentration of 0.3 pg/m3 to the mean exposure concentration, i.e., totally 1.8 pg/m3. This population mean exposure concentration may seem low if compared with earlier estimates made in Canada by the National Research Council of Canada (NRCC) (18). NRCC has made an extensive literature study and reports concentrations between 0.5 and 3 8000 pg/m . In a report from Reid and Watson (19), values of 14 pg/m3 are claimed to be typical for forest and agricultural areas outside Calgary, Alberta, in Canada. In Sweden, we find only some tenth of a microgram at Rorvik, outside Gothenburg.
Some of these discrepancies may be explained by the skewed selection of data that comes out of a literature study. For different reasons, it is more interesting to make studies during circumstances with high concentrations than with low or normal ones. But it cannot be excluded that there is a significant difference in ethylene exposure between Canada and Sweden, or eventually North America and Europe.
We can impose two types of quality control on our data. The first is to compare measurements from different laboratories. As mentioned above, the ethene / NOX ratio determined by different laboratories differed only slightly (5%). The second involves comparing the ratios with emissions ratios. Even if there is some difference, the data of the ratio in emissions are of the same order of magnitude and the maximum deviation is a factor of two. The ethene data obtained at Rorvik have been further obtained and reported within the European TOR project, where they have been accepted as reasonable.
Mean Exposure to Propene. The concentration ofpropene in Stockholm is about onethird that of ethene, according to Persson (2). 3 This gives a mean exposure of 0.5 pg/mr. A few measurements from burning of wood indicate a six-fold propene emission compared to ethene, which would give a contribution to the mean exposure of 1.8 pg/m3. The total 3 propene exposure would thus be 2.3 pg/mr. Mean Exposure to Benzene, Toluene, and Xylene. Most of the benzene, toluene, and xylene in urban air originates from traffic; but local contributions from the oil industry, wood burning, and solvent use may also contribute. The ratio between the mass concentrations of benzene and CO measured in Stockholm was 0.0142. Using normal CO/NOX ratios a benzene /NOX ratio of 0.16 may be found, corresponding to an 3 average benzene exposure of 3.7 pg/mr The concentrations of toluene and xylene in Sweden are normally two to three times as high as the benzene concentration. During Environmental Health Perspectives our measurements in Gothenburg, the ratio to benzene was 3.4 and 4.0, which is higher than normal. The ratio 3 is used here for the entire country, giving the mean exposure to 11 pg/m3 both for xylenes (o, m, and p) and toluene.
Mean Exposure to Formaldehyde. The concentration of formaldehyde at Hornsgatan in Stockholm is 0.77 ppb/ppm CO according to Persson and Almen (2). Using the ratio between CO and NOX we obtain a formaldehyde/NOX ratio of 0.009. The mean exposure ofprimary formaldehyde will therefore be 0.2 pg/m3. Because formaldehyde also is a secondary pollutant through photochemical reactions, the total exposure should be somewhat higher. A few measurements outside Gothenburg at Rorvik indicate concentrations in the order of 1 pg/mi3 (23). According to Shah and Singh (20), the mean is determined by a few episodes and is about twice as high as the median values. The mean value at Rorvik is therefore estimated to be about 2 pg/m3. Further, Rorvik has about double the concentrations of regional background as the demographic mean of Sweden, where the overall mean of the population exposure is estimated to be 1.2 pg/m . Mean Exposure to Acetic Aldehyde.
During measurements made in Sweden the concentrations of acetic aldehyde have been on the level ofor somewhat lower than that of formaldehyde (21). Shan and Singh report concentrations that give a ratio between acetic and formaldehyde of 0.9 on weight basis (20). This means that the mean exposure in Sweden should be about 1 pg/m3.
The use of ethanol in car fuel has given an acetic aldehyde concentration above that of formaldehyde in Brazil (22). Mean Exposure to Nickel. A mean Ni concentration of 5 ng/m3 was measured in central Orebro (23). The concentration of NOX was then about 50 pg/m3. The ratio 103 Ni/NO was then 0.1 x 10-3 and the mean exposure will be 2.3 ng/m3. Measurements in Almhult (8,000 inhabitants, middle of southern Sweden) show concentrations of this order of magnitude (24). Mean Exposure to Hexavalent Chromium.
A mean Cr concentration of 7 ng/m3 was found in Orebro during the same measurements referred to in "Mean Exposure to Nickel." This corresponds to a mean exposure of 3 ng/m3. The percentage of the Cr that is hexavalent is not known. The only information we have is that hexavalent chromium was 10% of the total in a smelter's plume during measurements made 15 years ago. Ifwe assume that most of the Cr in normal air is in mineral partides and only a small portion is water-soluble, perhaps 1% of the Cr may be assumed to be hexavalent until further information is available. The mean exposure of hexavalent Cr is therefore estimated to be 0.3 ng/m3. Mean Exposure to Arsenic. At the measurement in Orebro mentioned above, the mean concentration of As was 6 ng/m3, indi-3 cating a mean exposure of 2.8 ng/m3. During our measurements in Gothenburg, we found an As/benzene ratio of3 x 10-3. With the benzene/NOx ratio earlier determined at 0.16, we obtained an As / NO x ratio of 0.5 x 10 3, which in turn indicates a mean exposure of 11 ng/m3. The mean value 7 ng/m3 is used as an estimate for the whole population.
Mean Exposure to Cadmium. The mean concentration of Cd in Orebro was 0.65 ng/m3 during the same measurements as above, indicating an overall mean exposure 3 of 0.2 ng/m.
Mean Exposure to Asbestos. The only measurement known to us in Sweden ofasbestos in ambient air was made in Stockholm in 1981. It showed that the number of fibers detectable with an optical microscope was 3,000/m3 or lower. If the analysis was made in an electronic microscope, the number was 9,000 to 10,000 fibers//m3. Considering the relatively high NOX concentrations in Stockholm compared to the mean exposure concentration, the mean exposure concentration of asbestos fibers is estimated to be 200 optically detectable fibers/mi3 and 5,000 electro-microscopically detectable fibers/m3.
Mean Exposure to Silicon. The concentration of silicon (Si) was measured within this project by energidispersive X-ray fluorescence on samples collected on membrane filters. The mean mass fraction of Si in the partides was 9%. The air sample inlet was an inverted funnel giving an approximate cutoff at 10 to 20 pm, thus collecting inhalable partides (IP). The ratio IP/TSP (total 100 80 60 40 20 0 suspended particles) was determined to be 0.65 as an average in an American study (25). In our study, the TSP/ benzene ratio was 7.1, which together with the ratio of benzene/NOX (0.16) gives a ratio of Si/NOx of0.067 and a mean exposure of 1.5 /m3 in ambient air.
Because Si to a large extent exists in large particles, a great part of them will not penetrate into the indoor air. Fifty percent are assumed to penetrate, resulting in a mean population exposure of 0.7 pg/mi3.
Mean Exposure to Butadiene. Butadiene has not been measured in Sweden, but data in recent years focusing on its genotoxicity call for some attempt to estimate its average concentration. Shan and Singh give mean values of butadiene and benzene that indicate a ratio of 0.2, and thus a mean population exposure of 0.7 pg/mi3 (20).

Trends
Direct measurements of mutagenic substances in air are seldom made. Therefore, the mean exposure cannot be estimated except by indirect methods. Some indication of the trends in PAH levels, and possibly in ethene and propene levels can be obtained from trends in black-smoke concentration. In the central parts of Gothenburg, the concentrations decreased significantly from 1960 to 1990, probably due to changes in the heating systems, such as introduction of district heating, industrial waste heat recovery, and use of heat pumps ( Figure 5). At Fagelroskolan, a school in the northwest part of Gothenburg, and further to the west at the Volvo car production plant in Torslanda, which could be characterized as a peripheral and an industrial suburban area, respectively, the level is lower and the decrease is less. The contribution from the regional background and  some improvements to be expected in the emission of mutagenic substances from modern cars, but the contributions from malfunctioning cars, regional background levels, and increasing traffic may counteract the effect of these improvements.

Error Analysis
As mentioned earlier, we lack many of the data for estimating exposure to mutagenic substances, and of course, this is also true when trying to estimate the errors in the estimations. Some information is available, however, from the variance in the data used to determine the various ratios that we use in calculation of mean exposures. As mentioned in the section, "Method of Estimating Human Exposure," the mean exposure in a city is determined from the expression: C=K1xK2xK3xK4x CNO2 [3] where K, is the ratio of the mutagenic substance and NOx; K2 is the ratio of winter and year means; K3 is the ratio of population exposure mean and central city mean; and K4 is the ratio of NOx and NO2.
Ifwe divide the standard deviation in ratio data with the square root of the number of ratio values, we may get an estimation of the error in K-values, as seen in Table 6.
Using the method of least squares, we obtain a total error in terms of twice the standard deviation of 57% for PAH in traffic environments. In our estimation, for PAH in areas with a large emission contribution from small-scale burning of wood and oil, the error will increase. When estimating the exposure in these areas, we were forced to use an indirect method where the error is assumed to be a factor of two, or 100%. The error interval is obtained by dividing and multiplying the given error value with a factor 1 + error %/100. (The concentrations in ambient air are usually log normally distributed.) For ethene we have only two observations of the NOX ratio, but they differ by only a few percent. However, the contribution from wood burning is made in an indirect way for PAH, and the error is estimated to be the same, i.e., 100%. The error for the estimation of propene is larger, possibly 200%. For benzene, toluene, and xylene, the standard deviation of the K1 values is on the order of 20%. Measurements of benzene, toluene, and xylene occur rather frequently in traffic environments, which means that the number of K1 is large, resulting in a total error estimate of 50% for traffic environments. Because of exposure from other sources in ambient air, such as burning wood and evaporating solvents, the total error is estimated to increase by an extra 20 to 70%. The possible error in estimating the exposure of formaldehyde and acetic aldehyde is estimated to be of the same order of magnitude as that of PAH, i.e., 100%. For Ni, As, Cd, and Si, there are only a few data from which the NO, ratio can be deduced. Furthermore, we do not know that traffic or common city heating systems are predominant sources of these substances. A total error of 200% may be possible. For asbestos and hexavalent Cr the error may be very large. A guess is a factor of 5 for asbestos and a factor of 10 for hexavalent Cr.
The need for estimation of butadiene exposure was not identified until measurements by the Swedish Urban Project were made. The data used to calculate the NOX ratio are from a North American database. If we assume that the standard deviation of the NOX ratio are of the same order of magnitude as for other volatile organic components in this study, an error of 60% is indicated. However the limited quantity of Swedish data call for a somewhat higher error estimate, say 100%.

Discussion
Emissions measurements of light HC, using CO as a tracer, showed that the quantitative relationship between CO and individual HC was rather stable for unsaturated HC. The relative contributions from diesel-fueled cars (buses and trucks) become more important in the future. The emission of CO from diesel cars is quite low and the emission from passenger cars will be lowered by catalytic cleaning. Thus, the relationship between CO /NO, and HC will change with time.
It is important to consider the meteorologic parameters when measuring HC, and above all, PAH. The variation for PAH during the year is great-a factor of three to five when comparing winter and summer values.
To make risk assessments, it is necessary to have access to data on the concentration of genotoxic substances. Ambient air monitoring for these substances should be increased in urban areas and in areas with small-scale burning of oil and wood. Since these measurements are very complicated, work should be done to determine which tracer substances or genotoxic substances to monitor in these areas, as well as to harmonize measurement methods and reporting to enable intercomparison of the measurement results.
Through the methods described in this chapter, it is possible only to determine exposure via inhalation. Another important factor for exposure is indirect intake via food. Particulates containing PAH deposit on leafy vegetables and other crops. In this way, cereals can also be an important source for PAH exposure (26). It is an important task to find out in greater detail what the extent of this exposure is, and the main sources of deposited particulates.

Conclusions
The investigations performed show that it is possible to obtain quantitative realtionships between CO and unsaturated HC. Saturated HC and PAH-HC showed greater seasonal variations. In general, it is important to make more measurements of genotoxic substances in urban areas in a systematic way to get a better basis for risk estimations. Exposure from wood firing and the indirect uptake of PAH via food seem to be important, and more research is needed in those areas.
It is obvious from this investigation that the exposure of mutagenic air pollutants is not a problem only for large cities. Halfof the pop-ulation exposure-dose of NO. is obtained in cities and urban areas with a population ofless than 50,000. If the influence of wood firing and small-scale oil firing is as it appears from the limited data we have, the "median" city is even smaller.
Through use of meteorologic models, data concerning the pattern of human movement and whereabouts, measurement data of NO2 in urban areas and NOx as a tracer substance, it has been possible to calculate the mean exposure for a number of genotixic substances.
Relationships between NO, and some genotoxic substances have been determined. The mean exposure of nitrogen oxides for the Swedish population (8.5 million inhabitants) was calculated to be 23 pm/m3 as a yearly mean. The results for the mean exposure values to some genotoxic substances in Sweden are summarized in Table 7.