Fine Particles and Coarse Particles: Concentration Relationships Relevant to Epidemiologic Studies

Transcription

1 Journal of the Air & Waste Management Association ISSN: (Print) (Online) Journal homepage: Fine Particles and Coarse Particles: Concentration Relationships Relevant to Epidemiologic Studies William E. Wilson & Helen H. Suh To cite this article: William E. Wilson & Helen H. Suh (1997) Fine Particles and Coarse Particles: Concentration Relationships Relevant to Epidemiologic Studies, Journal of the Air & Waste Management Association, 47:12, , DOI: / To link to this article: Published online: 01 Mar Submit your article to this journal Article views: 4586 View related articles Citing articles: 259 View citing articles Full Terms & Conditions of access and use can be found at Download by: [ ] Date: 18 November 2017, At: 03:13

2 TECHNICAL Wilson and Suh PAPER ISSN J. Air & Waste Manage. Assoc. 47: Copyright 1997 Air & Waste Management Association Fine Particles and Coarse Particles: Concentration Relationships Relevant to Epidemiologic Studies William E. Wilson U.S. Environmental Protection Agency, Research Triangle Park, North Carolina Helen H. Suh Harvard University, Boston, Massachusetts ABSTRACT Fine particles and coarse particles are defined in terms of the modal structure of particle size distributions typically observed in the atmosphere. Differences between the various modes are discussed. The fractions of fine and coarse particles collected in specific size ranges, such as total suspended particulate matter (TSP),,, are shown. Correlations of 24-h concentrations of, at the same site show that, in Philadelphia and St. Louis, is highly correlated with but poorly correlated with. Among sites distributed across these urban areas, the site-to-site correlations of 24-h PM concentrations are high for but not for. This indicates that a PM measurement at a central monitor can serve as a better indicator of the community-wide concentration of fine particles than of coarse particles. The fraction of ambient outdoor particles found suspended indoors is greater for fine particles than for coarse particles because of the difference in indoor lifetimes. Consideration of these relationships leads to the hypothesis that the statistical associations found between daily PM indicators and health outcomes may be the result of variations in the IMPLICATIONS Many epidemiologic studies have found statistical associations between health outcomes and PM concentrations, usually TSP or. However, fine particles are considered more likely than coarse particles to be responsible for respiratory health effects. 1,2 Concentration relationships among three PM fractions, (an indicator of fine particles) (an indicator of coarse particles) suggest that the association between health effects and may be the result of an underlying association of health effects with fine particles. Fine particles and coarse particles should be considered separate classes of pollutants. The extent to which a measurement at one site can represent the community concentration of either fine or coarse particles also needs to be determined. fine particle component of the atmospheric aerosol, not of variations in the coarse component. As a result, epidemiologic studies using or TSP may provide more useful information on the acute health effects of fine particles than coarse particles. Fine and coarse particles are separate classes of pollutants and should be measured separately in research and epidemiologic studies. are indicators or surrogates, but not measurements, of fine and coarse particles. INTRODUCTION Epidemiologic studies have reported statistical associations between day-to-day changes in health outcomes, such as daily mortality, and day-to-day variations of indicators of daily ambient particulate matter (PM) concentration, most frequently TSP (total suspended particulate matter) or (particles with an upper 50% cut point of 10 µm aerodynamic diameter). 3-6 However, PM is composed of many different components, and many different parameters may be used as indicators. To be most useful as a PM indicator for time-series epidemiology, an indicator should be variable on a day-to-day basis, be nearly uniform in concentration across the area providing health data, and have a high correlation between outdoor and indoor concentrations. As a first step, PM should be considered to consist of two separate classes of pollutants, fine particles and coarse particles. Fine and coarse particles differ in sources, formation mechanisms, composition, atmospheric lifetimes, spatial distribution, indoor outdoor ratios, and temporal variability, as well as size. They also may differ in biological effects. 7,8 This paper will analyze fine and coarse particle data collected in Philadelphia, PA, during the summers of 1979, 1992, and 1993, and the winter of The usefulness of fine and coarse PM concentration as indicators for time-series epidemiology will be examined in regard to day-to-day variability, uniformity across the urban area, and indoor outdoor particle concentration ratios Journal of the Air & Waste Management Association Volume 47 December 1997

3 DEFINITIONS The definition of fine and coarse particles is an operational one based on observations, beginning in the early 1970s, that measurements of size distribution, when appropriately plotted, usually yielded bimodal distributions of particle mass, with a minimum diameter between 0.7 and 3.0 µm. 2,9 An idealized distribution, showing the normally observed division of ambient aerosols into fine and coarse particles, is presented in Figure 1. Aerosol refers to a suspension of solid or liquid particles in air; however, aerosol is sometimes used to refer to the particles only. Fine and coarse particles may overlap in the intermodal region between 1 and 3 µm. 10 Existing samplers have been designed to collect a specific portion of the distribution. For example, TSP is defined by the design of the High Volume Sampler (HiVol), which collects all of the fine particles but only part of the coarse particles. The upper cut-off size of the HiVol is undefined, except by the design of the sampler, and varies with wind speed and direction. Extraordinary measures, such as were undertaken with the Wide Area Aerosol Classifier (WRAC), are required to collect the entire coarse mode. 11 To focus regulatory concern on particles small enough to enter the lower (i.e., thoracic) regions of the respiratory tract, the indicator for the National Ambient Air Quality Standard for PM was changed in 1987 from TSP, as measured by the HiVol, to. 12 samplers collect all of the fine particles and part of the coarse particles. The upper cut point is defined as having a 50% collection efficiency at 10 ± 0.5 mm diameter. The slope of the collection efficiency curve also is defined. 13 Samplers with upper cut points of 3.5, 2.5, 2.1, and 1.0 µm are also in use. Dichotomous samplers split the particles into smaller and larger fractions that may be collected on separate filters. Detailed information on the design and use of particle samplers may be found in the review by Chow. 14 Interest in fine and coarse particles, as distinct components of the atmospheric aerosol, began in the early 1970s, largely because of size distribution studies by Whitby and co-workers. 9,15 The need to measure fine and coarse particles separately was endorsed by U.S. Environmental Protection Agency (EPA) scientists in Based largely on existing technology, they recommended 15 µm for the upper cut point and 2.5 µm for the cut point between fine and coarse particles. The sample contains all of the fine particles, but, especially in dry areas or during dry conditions, may collect a small but significant fraction of the coarse particles. 10 In 1979, EPA initiated the Inhalable Particle Network (IPN), 17 which included measurements of fine and coarse particles using dichotomous samplers with cut points of 15 and 2.5 µm. The Regional Air Pollution Study in St. Louis, MO, 18 and the Harvard Six Cities Study 19 included similar measurements of fine and coarse particles. The 15-µm upper cut point subsequently was changed to 10 µm in order to simulate the thoracic collection efficiency of the human respiratory system. An historical account of the development of size-selective samplers for collecting size fractions that reach various portions of the respiratory system is given by Lippmann. 20 Over the last 15 years, a significant body of data has been collected using samplers with an upper 50% cut point of 2.5 µm. As a result, some authors use fine as a substitute for. Some authors use coarse to describe that portion of PM between and. In this paper, samples specified by relatively sharp cuts will be described numerically (i.e.,, ). Because fine and coarse have lost the precise meaning intended by Whitby, 9 the addition of mode is sometimes used to emphasize reference to the fine- or coarse-mode particles shown in the distributions in Figures 1 and 2. Thus, may be considered an indicator of fine-mode particle mass; Figure 1. An idealized size distribution of ambient particulate matter showing fine and coarse modes and the portions collected in various samples. Figure 2. A particulate matter size distribution collected in traffic showing formation mechanism for nuclei and fine and coarse modes. Volume 47 December 1997 Journal of the Air & Waste Management Association 1239

4 an indicator of the thoracic component (the fraction of PM that enters the thorax); an indicator of thoracic coarse-mode particles (i.e., that portion of the coarse-mode particle mass that reaches the thoracic compartment). However, because of the overlap of fine- and coarse-mode particles in the intermodal region (1-3 µm), is only an approximation of fine-mode particles is only an approximation of thoracic coarse-mode particles. In this paper, fine particles and coarse particles will refer to fine-mode particles and thoracic coarse-mode particles, respectively, not to the approximations given by or. Also, diameter normally will refer to the aerodynamic diameter. The development of experimental evidence and theoretical understanding of particle formation and growth led to a model of a fine particle mode composed of a nuclei mode and an accumulation mode. 9 Because very small particles rapidly combine to form larger ones, the nuclei mode can be seen as a completely separate mode only near sources of nuclei mode particles. The size distribution measurement in Figure 2, made in automobile traffic at the General Motors Proving Grounds, shows such a separation. 21 Measurements of particles in the background air, made before and after measurements in simulated freeway traffic, showed that all of the nuclei mode and a small fraction of the coarse mode were generated by the traffic, but that the accumulation mode and most of the coarse mode were present in the background air. Such studies have led to a better understanding of the formation mechanisms for fine and coarse particles. In the simplest terms, coarse particles are formed by breaking up bigger particles into smaller particles. The coarse particle mode is sometimes called the dispersion, mechanically generated, or comminution mode. However, as particles become smaller and smaller, more and more energy is required to break them into smaller units. This establishes a lower limit of approximately 1 µm for coarse particles. 22 The major sources of coarse particles are windblown dust from soil, unpaved roads, piles of material containing coarse dust, etc., and dust reentrained by turbulent air generated by traffic on paved or unpaved roads. Coarse particles also are generated by demolition of buildings and evaporation of sea spray. Pollen, mold spores, and parts of plants and insects also are found in the coarse-particle-size range. Fine particles usually are formed from gases. The processes of nucleation and growth are shown in Figure 2. Nucleation involves the formation of very small particles from gases. Substances with low saturation vapor pressures are generated in the gas phase by high-temperature vaporization or through chemical reaction in the atmosphere. Growth of particles in the nuclei mode occurs through two processes: (1) coagulation, in which two small particles combine to form a larger particle, and (2) condensation, in which gas or vapor molecules condense onto existing particles. The rate of coagulation depends on particle number and velocity; the rate of condensation depends on surface area. These parameters decrease rapidly as the particle size approaches 1 µm. As a result, particles normally do not grow by these processes to diameters above approximately 1 µm. Because particles tend to accumulate in the size range from 0.1 to 1 µm, this range is called the accumulation mode. Secondary fine particles are formed by the atmospheric conversion of gases into particles. In one mechanism, a gas is converted into the vapor of a material with a low saturation vapor pressure. An example is the oxidation of sulfur dioxide (SO 2 ) to sulfuric acid (H 2 SO 4 ), which forms fine particles by nucleation, followed by coagulation. Newly formed H 2 SO 4 molecules also may condense on existing particles. In a second mechanism, a gas is converted into a different gas that can further react to form a substance with a low saturation vapor pressure. An example is the oxidation of nitrogen dioxide to nitric acid (HNO 3 ), which can react with ammonia to form fine particles of ammonium nitrate. Acid gases in the atmosphere, such as SO 2 and HNO 3, also may react with coarse particles (such as calcium carbonate [CaCO 3 ] from soil, construction, or demolition, or sodium chloride [NaCl] from sea spray or road salt) to form salts. These would still be considered coarse particles. Metallic vapor (e.g., formed during smelting or high temperature combustion) and organic vapor (e.g., formed during cooking or low temperature combustion) that coagulate or condense without chemical reaction form primary fine particles. Fine PM contains particles resulting from human activities and particles of natural origin (e.g., atmospheric reaction products of gases, such as terpenes emitted by plants). Thus, the organic component of fine particles may contain both natural and anthropogenic material and both primary and secondary material. Combustion products from the burning of gasoline and diesel fuel form fine particles. However, combustion of coal and heavy fuel oil yields both fine particles from material vaporized during combustion, and coarse particles (i.e., fly ash, from noncombustible material). In recent years, it has been realized that when relative humidity is near 100%, as in fog or clouds, certain types of particles can grow by absorption of water, attaining diameters up to five times their dry size. 23 Size distributions of particles subjected to such conditions sometimes show the accumulation mode split into a condensation mode dry or less hygroscopic and a droplet mode wet or more hygroscopic. 24,25 Some of the growth of the droplet mode may come from gases that dissolve and react in the wet particles. 24, Journal of the Air & Waste Management Association Volume 47 December 1997

5 Table 1. Comparison of ambient fine- and coarse-mode particles. Fine-Mode Particles Coarse-Mode Particles Formed from: Gases. Large solids/droplets. Formed by: Chemical reaction or Mechanical disruption vaporization. Nucleation, condensation on nuclei, and coagulation. Evaporation of fog and cloud droplets in which gases have dissolved and reacted. (crushing, grinding, abrasion of surfaces, etc.). Evaporation of sprays. Suspension of dusts. Composed of: Sulfate, nitrate, ammonium, Resuspended dust (soil dust, and hydrogen ions. Elemental carbon. Organic compounds (e.g., polyaromatic hydrocarbons). Metals (e.g., lead, cadmium, vanadium, nickel, copper, zinc, manganese, iron). Particle-bound water. street dust). Coal and oil fly ash. Fine particles and coarse particles differ in many aspects other than size. Some of these differences are summarized in Table 1. USE OF FINE AND COARSE PARTICLE MASS IN TIME-SERIES EPIDEMIOLOGY One class of epidemiologic studies seeks statistical associations between daily, nontrauma deaths and daily indicators of exposure to air pollutants. 3-6 Only large urban areas have enough daily deaths to provide adequate statistical power. The use of a concentration measured at one site to infer the average community concentration places certain restrictions and requirements such as temporal variability and spatial uniformity on the indicators used. (The requirements for use in acute timeseries studies within a single community and in comparisons among communities are different. In acute time-series studies, the day-to-day differences are important; in studies among various communities, the average value is important.) Oxides of crustal elements, (silicon, aluminum, titanium, and iron). CaCO 3, NaCl, sea salt. Pollen, mold, fungal spores. Plant/animal fragments. Tire wear debris. Solubility: Largely soluble, hygroscopic, Largely insoluble and deliquescent. and nonhygroscopic. Sources: Combustion of coal, oil, Resuspension of industrial dust Atmospheric gasoline, diesel fuel, and wood. Atmospheric transformation products of nitrogen oxides, SO 2, and organic compounds, including biogenic organic species (e.g., terpenes). High-temperature processes, smelters, steel mills, etc. and soil tracked onto roads and streets. Suspension from disturbed soil (e.g., farming, mining, unpaved roads). Biological sources. Construction and demolition. Coal and oil combustion. Ocean spray. Half-Life: Days to weeks. Minutes to hours. Travel distance: 100s to 1000s of km. <1 to 10s of km Variability in Time. The concentration of the indicator must vary on a day-to-day basis. If the indicator were constant, no test of a relationship would be possible (except among different communities). The greater the variability, the easier to extract the signal from the noise. If the indicator is composed of several pollutants, it must express the variability of the component responsible for the health outcome. Uniformity in Space. It is desirable that the concentration of the indicator be constant across the community from which the deaths or other health outcomes were recorded. However, determining a community-wide average may be possible if site-tosite correlations within the community are reasonably high and statistically significant, even if the absolute concentrations vary. Because of the large differences in chemical and physical properties, it might be expected that fine and coarse particles would have different biological properties and health effects. Epidemiologic studies seeking statistical associations between time series of daily ambient PM concentration parameters and time series of daily mortality usually use measurements of TSP or as the indicators of particulate pollution, primarily because these measurements are available from regulatory monitoring networks. Unfortunately, few long time series containing both fine- and coarse-particle mass are available for use in epidemiologic studies. In this paper, a time series of 24-h concentrations of and collected in Philadelphia during the summers of 1992 and will be examined. Paired comparisons and correlations between, will be used to determine how well these components of PM satisfy the above requirements for use as indicators in epidemiologic correlations. Although the measurements provide only two independent data sets, analyzing, as independent data sets is useful. Same-Site Correlations Paired relationships between, at individual sites will be examined first. A time series of, based on measurements of and taken with Harvard impactors at the Presbyterian Home (PBY) site in Philadelphia during the summers of 1992 and 1993, 27 is shown in Figure 3. Sometimes increase and decrease together. At other times, Volume 47 December 1997 Journal of the Air & Waste Management Association 1241

6 Figure 3. Comparison of mass concentrations of measured at the PBY site in Philadelphia, summer they move in opposite directions. Furthermore, it appears that dominated both the mass and the day-today variability in mass. The variability of, can be examined mathematically by comparing the coefficient of variation (CV; standard deviation [SD] divided by the mean)., all show significant variability; however, the variability of is lower than that of either component. The contribution of each component to the variability of, if the other component were constant, may be examined by dividing the SD of each component by the mean. PM (10-2.5) was found to have a higher variability than. However, because the actual concentration of is significantly lower than that of, the variability of is dominated by the variability of. CV data from all sites, averaged over both summers, are shown in Figure 4. On the average, the CV is only 87% of the CV and only 77% of the CV of, showing a reduction in variability by combining. On the average, contributes 2.75 times as more variability to than. The determination of a statistical association between air pollution exposure and health outcomes depends on Figure 5. Average concentration of particle fractions for all sites in Philadelphia, summer 1992 and the existence of a gradient in each. In going from TSP to and from to or, the gradient in exposure will increase and the ability to identify statistical relationships will be enhanced. Further demonstration that the variability in reflects the variability in more than the variability in is provided by statistical analyses. For each of the time series, as shown in Figure 3, the correlation between, as indicated by the coefficient of determination (R 2 ), was low (average R 2 = 0.11). However, the correlation between and was high (average R 2 = 0.90) and the correlation between was moderate (average R 2 = 0.35). The average concentrations for the three particle fractions are shown in Figure 5. The R 2 s for the relationships among the three fractions are shown in Figure 6. Similar data from St. Louis are given in Table 2. Measurements of, made with dichotomous samplers, for 212 city-year data sets were Figure 4. Coefficient of variation for particle fractions for all sites in Philadelphia, summer 1992 and Figure 6. Coefficient of determination (R 2 ) for the relationship among particle fractions for all sites in Philadelphia, summer 1992 and Journal of the Air & Waste Management Association Volume 47 December 1997

7 Table 2. Coefficient of determination (R 2 ) tables for St. Louis,. Community Sites Source Site Year Site Average 7001 vs vs. vs retrieved from EPA s Aerometric Information Retrieval System (AIRS). 28 The distribution of R 2 is shown in Figure 7. For 75% of the data sets, the R 2 s between are less than For these data sets, the R 2 s between and were relatively high. Therefore, for these city-year situations, the variability in is dominated by the variability in. However, the R 2 s between are greater than those between and in 30% of the data sets. Most of these data sets were from dry western cities with high coarse particle concentrations. In many cases, the average concentration of exceeded that of the particles. In other cases, the sampler may have been placed near a coarse particle source rather than in a location representative of the community-wide concentration. This analysis suggests that in most cities where the fine particle concentration exceeds the thoracic coarse particle concentration, the variation in particle concentration at any individual site characteristic of the community will be dominated by the variability in the fine particle component. In cities where fine particles drive the variations in, the statistical associations of health outcomes with TSP and may be the result of the variability in fine particles, not of the variability in coarse particles. This possibility is supported by 75% of the data sets in AIRS. However, in cities with high levels of coarse particles and low levels of fine particles, the coarse component may drive the variations in. In cities where fine and coarse PM mass are comparable, the variability of will reflect the sum of the fine and coarse particle variability. However, unless fine and coarse particles are highly correlated, which is rarely the case, the variability of will be less than that of either component. Therefore, separate measurements of fine and coarse particles would allow a better assessment of the statistical correlation between PM parameters and health outcomes. Figure 7. Distribution of values of R 2 for the paired relationships between,, calculated from data in EPA s AIRS. concentrations in from the Regional Air Monitoring System network in St. Louis were reported to have site-to-site correlation coefficients of greater than Sulfate and respirable particle mass were reported to show little variation across sites in five of the six Harvard Six Cities Study cities. 19 Kotchmar et al. 30 examined the adequacy of using a single monitoring site for defining mean outdoor concentrations of particle mass fractions. Measurements were made of, PM 15, and TSP at three sites in each of five demarcated residential communities. Most site pairs were between 1.6 and 6 km apart (although two pairs were 8 and 9 km apart, respectively). Using analysis of variation to test for comparability within each set of three sites, the authors concluded that was distributed more uniformly spatially than PM (15-2.5). However, few analyses of the distribution of particle mass across large Site-to-Site Correlations Earlier studies have reported that sulfate is distributed evenly across St. Louis, and that sulfate and respirable particles are evenly distributed across small cities. Sulfate Figure 8. Correlation coefficient (r) for the paired relationships between,, plotted as a function of the distance between sites for eight sites in Philadelphia, summer 1992 and Volume 47 December 1997 Journal of the Air & Waste Management Association 1243

8 Figure 9. Map of Philadelphia showing location of sampling sites. urban areas or of differences in correlations among the various components of have been reported. For situations in which concurrent measurements are available at two or more sites in an urban area, was distributed more evenly than throughout the urban area. This can be shown by examining site-to-site correlations of,. In Figure 8, the correlation coefficients (r) are plotted as a function of distance between sites for eight sites in Philadelphia. These also are based on summer 1992 and 1993 measurements of and, using Harvard impactors. 27 The locations of the sites are shown in Figure 9. The correlation coefficient between particles at different pairs of sites varied from 0.80 to 0.96, with an average of However, for, r ranged from 0.14 to 0.63, with an average of The r range for was 0.79 to 0.96, with an average of Therefore, as expected for this situation, values were more variable across the city than the fraction, but less variable than. Earlier studies in Philadelphia 31 report concentrations at sites chosen to be characteristic of the community and Figure 10. Correlation coefficient (r) for the paired relationships between, PM 15, PM (15-2.5), and TSP, plotted as a function of the distance between sites for four community sites from the IPN study in Philadelphia, summer Figure 11. Correlation coefficient (r) for the paired relationships between, PM 15, PM (15-2.5), and TSP, plotted as a function of the distance between sites for five source-oriented sites from the IPN study in Philadelphia, October 1979 to February sites chosen to document the impact of specific, identified local sources. Measurements were made with HiVols (TSP) and dichotomous samplers (with PM 15 and cut points). Correlation coefficients for site pairs, taken from the reference, are plotted as a function of distance between sites for each pair for the community sites in Figure 10 and for the local source impact sites in Figure 11. For the community sites, all site pairs showed high correlations for and TSP, slightly lower correlations for PM 15, but much lower correlations for PM (15-2.5). For the local source impact sites, the site pair correlations are still high for the component, although, as expected, the correlations are lower for the other components. Similar results from St. Louis are shown in Table 3. These observations are in agreement with the current understanding of particle formation and transport. Because fine particles travel long distances and undergo extensive atmospheric mixing, they should be distributed evenly over urban or larger areas. Measurement of fine particles at one site, therefore, should give a good measurement of the concentration of fine particles across the entire city. Also, the day-to-day variability in fine particles at one site should be a good measure of the average variability across the city. The sources and area of impact of coarse particles, however, are more local and may be quite variable from place-to-place. Therefore, as indicated by the low site-to-site correlations, a measurement of coarse particles may be an indicator of concentrations only in the immediate neighborhood. An alternative explanation for the lack of correlation of particles at different sites across Philadelphia has been suggested by Warren White. 32 The particles could be distributed evenly across Philadelphia, but the precision of the measurement might be low enough to cause the low and variable correlations observed. This possibility cannot be ruled out. The concentrations in Philadelphia were 1244 Journal of the Air & Waste Management Association Volume 47 December 1997

9 Table 3. Correlation coefficient (r) tables for St. Louis. a vs. Site vs. Site vs. Site a 2003, 5001, and 7002 are community sites; 7001 is a source site. derived by subtracting from. As a result, the precision of the measurements in Philadelphia (CV = 40%) is significantly lower than that of (CV = 9.2%) or (CV = 11%). 24 The precision of the dichotomous sampler is better for and comparable for and. The CV for five cities, as reported by Rodes et al., 33 ranged from 2.6 to 18.3% for, 2.2 to 9.2% for, and 3.6 to 11.4% for, improving with the experience of the operators and the concentration of PM. Measurements made with dichotomous samplers in St. Louis (Table 3) also show a lower site-to-site correlation for than for or, although not as much lower as in the Philadelphia data set. Kotchmar et al. 30 report, in a study of three monitoring sites in each of five cities, that one site provides an adequate measurement of average but not of average PM (15-2.5). These two examples support the idea that coarse particle concentrations vary across a city. However, current experimental data are not adequate to determine whether the lower site-to-site correlation of coarse particle indicators is the result of the variation of coarse particle concentrations across the community, a lack of precision of the monitoring technique, or both. Whatever the reason, a measure of coarse particles at one site will not necessarily be a good indicator of the concentration or variability of coarse particles across the community. The same considerations should hold for thoracic coarse particles as for coarse particles. This analysis is based largely on data from two summers in Philadelphia. Clearly such analyses need to be extended to other seasons, other locations, and other components of the atmospheric aerosol. The correlation of sulfate and strong particle acidity across Philadelphia and regional areas has been reported. 34,35 Unfortunately, daily indicators of the fine and coarse fractions of are rarely available, even for two sites in the same city. Indoor/Outdoor Relationships A significant effort has been devoted to studying the relationships among indoor, outdoor, and personal concentrations of. 19,36-51 In general, because of the large indoor sources of particles (fine particles from smoking, cooking, and wood burning 19,39-41 and coarse particles from cleaning and general indoor activity ), a significant correlation rarely is found between daily outdoor and indoor or outdoor and personal PM concentrations Better correlations are observed for subjects in homes with minimal indoor sources 48 and for extended measurements of the same home. 46,49 The relevant information on indoor, outdoor, and personal measurements of PM mass has been reviewed. 46,47 The indoor outdoor ratios of total PM mass, however, are not the relevant parameters for community epidemiologic studies because, as would be expected, the emissions of PM from indoor sources are poorly correlated with the PM measured at a community monitoring site. For example, a smoker does not smoke more cigarettes on a high ambient pollution day than on a low ambient pollution day. The relevant epidemiologic parameter is the concentration of the ambient (or outdoor) particles that have penetrated into the indoor microenvironment and remain suspended. The ratio of this quantity to the ambient (or outdoor) particle concentration will be called the infiltration factor. 39 When multiplied by the ambient concentration, it gives the equilibrium quantity of outdoor particles that have penetrated indoors and that remain suspended indoors during a period of constant ambient PM concentration. Less research has been done on infiltration factors than on indoor outdoor ratios. Indoor measurements combine particles from indoor and outdoor sources, and it is usually difficult to identify the source of an arbitrary particle found within an indoor microenvironment. However, a high correlation has been found between indoor and outdoor sulfate with an infiltration factor for sulfate of 0.88 (R 2 = 0.84) for non-air-conditioned homes. 50 For air-conditioned homes, Suh et al. 50 found an infiltration factor of 0.44 (R 2 = 0.57). The lower infiltration ratio for the air-conditioned homes may be due to a lower air exchange rate or to the removal of PM by the filtration of return air to the air-conditioning unit. Both types of homes had no known indoor sources of sulfate other than the possible resuspension by human activity of ambient sulfate that previously had deposited on carpets, furniture, and other surfaces. Less information is available on infiltration factors for other particle parameters, such as total PM mass or number, which have both ambient and indoor sources. Some information on the differences in infiltration factor for fine and coarse particles can be obtained from Volume 47 December 1997 Journal of the Air & Waste Management Association 1245

10 the analysis of the infiltration process. Several studies have discussed models of the infiltration factor. 40,41,46 At equilibrium, for a home without auxiliary PM removal devices (such as an electronic air cleaner or an in-line filter for recirculating air), the infiltration factor equals Pa/(a + k), where a is the air exchange rate, P is the penetration fraction, and k is the deposition rate. For a home with an internal air cleaner that has an air exchange rate of a and a filtration efficiency of f, the infiltration factor equals Pa/(a + k + a f). In the equation above, P, k, and f all may be functions of particle size. On the basis of the physics of aerosol penetration through curved tubes, P would be expected to be a function of particle aerodynamic size and electrostatic charge if air penetrates indoors through narrow, twisting passageways. 22 However, P would be unity if air exchange occurs through open doors or windows or through large, more direct passageways. Studies of indoor and outdoor PM concentrations using mass balance models have estimated P values of slightly less than unity. Using data collected during the summer in State College, PA, Suh et al. 51 estimated P = 0.85 (±0.02) for sulfate. This agrees well with P = 0.84 reported by Koutrakis et al. 41 from a study conducted in New York mainly during the winter. An analysis of data from the Particle Total Exposure Assessment Methodology (PTEAM) study, using a statistical method to estimate both P and k, found that P was not different from unity. 39,40,46 An experimental study in which both k and P were measured also found P = 1, with no dependence on particle size up to 10 µm. 44 The lower values of P for sulfate may have been the result of the homes in those studies being more tightly sealed. However, because these studies measured only sulfate 51 or only, 41 they do not provide any justification for treating P as a function of particle size. The deposition rate, k, also has been studied. 44,52-56 It is known to be a strong function of particle size, reaching a Figure 12. Infiltration ratio (ambient PM indoors/ambient PM outdoors) plotted as a function of air exchange ratio for. minimum between 0.1 and 1.0 µm near the mass mean diameter of the accumulation mode size distribution and increasing for larger and smaller sizes. The deposition rate for coarse particles could be as much as an order of magnitude higher than the deposition rate for fine particles. Deposition rates of 0.39 h -1 for and 1.01 h -1 for have been inferred statistically from the PTEAM database. 39,40,46 However, these values depend on the relative fractions of the fine- and coarse-mode PM making up the total mass of these fractions, so they are specific to the PM mix found in Riverside during the PTEAM study. The infiltration factor, calculated for a range of air exchange rates in non-air-conditioned homes, using P = 1 and ks from PTEAM, is shown in Figure 12. Homes tightly sealed for air-conditioning or heating may have air exchange rates below 0.3. Homes with open windows may have an air exchange rate of greater than ,58 Two studies give a median of 0.5 h -1 for the United States. 59,60 The median for the PTEAM study was 1.02 h -1 during the day and 0.80 h -1 during the night. 61 As can be seen in Figure 12, the infiltration factor for is significantly greater than that for (e.g., at the PTEAM daytime median, 0.68 vs. 0.49; at the nighttime median, 0.66 vs. 0.46). The average personal exposure to ambient particles will be the weighted sum of the outdoor exposure to ambient particles and of the indoor exposure to ambient particles that have infiltrated indoors. The ratio of personal ambient PM exposure to the PM concentration at a community monitor will be lower for coarse particles than for fine particles because of the lower infiltration factor for coarse particles. In addition, most people spend a large fraction of the day indoors; in fact, the U.S. average is only about 2 h a day outside. 62 When a community exposure is calculated by averaging the exposure of people with different times spent outdoors, the lower infiltration factor for coarse particles will lead not only to a lower value for the average personal exposure in the community but also to a broader distribution of the individual values of personal exposure for ambient coarse particles. These relationships will make it more difficult to discern a statistical relationship, if it exists, between health effects and ambient coarse particles. Thus, differences in indoor outdoor relationships, along with differences in variability in time and uniformity in space, argue for separate measurements of fine and coarse particles. From the point of view of exposure assessment for epidemiology studies, the important question is the independence of variations in concentrations of particles generated indoors from particles generated outdoors. If indoor-generated PM concentrations are statistically independent of outdoor-generated PM concentrations, then variations in health endpoints or health effects indicators associated with variations in ambient PM cannot be confounded by indoorgenerated PM. There is no evidence that people alter their 1246 Journal of the Air & Waste Management Association Volume 47 December 1997

11 indoor particle-generating activities in response to variations in ambient PM concentrations. However, people might alter the amount of time they spend outdoors, either directly, in response to air pollution warnings, or indirectly, because of variations in weather. Addressing these issues experimentally and statistically will require indoor outdoor personal exposure studies with separate measurements of fine and coarse particles that can differentiate indoor-generated particles from outdoor-generated particles. In addition, surveys will be needed to determine how behavior might be influenced by factors related to ambient air pollution. DISCUSSION Based on 24-h concentrations, was poorly correlated with at most sites. Moreover, the concentration of was frequently higher than the concentration of. was better correlated with than with. Therefore, the daily variation in results primarily from the daily variation in. In the limited database examined, measured daily concentrations of and PM (15-2.5) varied across the city. Therefore, a measurement of at one site would not provide an adequate measure of the average daily concentration of across the community. However, concentrations were relatively evenly distributed across the city. Therefore, a measurement of at one site is more likely to represent the average daily concentration of across the community. The infiltration factor, which gives the fraction of outdoor particles found indoors, and which depends on the penetration indoors and the particle lifetime indoors, is greater for fine particles than for coarse particles, largely because of the lower indoor lifetimes of coarse particles relative to fine particles. Therefore, fine particle concentrations measured at a community monitoring site better reflect the personal exposure to ambient fine particles (indoor plus outdoors), averaged over the community, than is the case for coarse particles. The above observations lead to the hypothesis that the statistical associations observed between daily PM indicators and daily health outcomes may be the result of variations in the fine particle component of the atmospheric aerosol, not of variations in the coarse component. The hypothesis should apply to eastern U.S. cities, where coarse particle concentrations are low and fine particles are regional in nature. However, it may not apply to parts of the dry, western United States or to other areas where coarse particle concentrations are higher and fine particles may be less regional. The hypothesis has been tested only partially, primarily with summertime measurements from Philadelphia. However, this hypothesis leads to three testable predictions or conclusions. (1) Fine particles will show stronger and more statistically significant associations with health outcomes than will coarse or thoracic coarse particles. When this hypothesis was first advanced, 63 some suggestive evidence existed that or other indicators of fine particles might be better predictors of health outcomes than or TSP. Such suggestions came from cross-sectional studies, 64 cohort studies, 65 and time-series studies. 66,67 However, no comparisons of the strength of the statistical associations of fine particles, relative to thoracic coarse particles, had been made. Since then, two time-series studies 68,69 and a further analysis of one cohort study 70 have compared with. All three show a stronger and more statistically significant association for, an indicator of fine particles, than for, an indicator of thoracic coarse particles. The only exception for the seven cities studied was Stubenville, WV, where was high and highly correlated with. (2) In locations where statistical associations have been found between health effects and PM indicators, such as TSP or, a significant correlation will be found between the PM indicator and fine particles. Although neither TSP nor should be used as fine particle indicators, their day-to-day variations may serve as an index to the day-to-day variations in fine particles. This prediction can be tested by examining the correlation between TSP or and or other fine particle indicator. This relationship has been examined in Philadelphia (Figure 6). The average correlation coefficient between and during the summers of 1992 and 1993 was >0.9. It may be possible to find cities where PM indicators are highly correlated with coarse particles. However, a correlation with coarse particles, comparable to or greater than that with fine particles, would not guarantee that coarse particle variations contributed to the statistical association. This occurs because variations in coarse particle concentrations at a community monitoring site are not expected to be as good an indicator of the average community exposure as are fine particles. (3) The failure to find a statistically significant association between indicators of coarse particles and health outcomes does not absolve coarse particles from health effects. This conclusion will be more difficult to evaluate. Existing epidemiologic studies using or TSP probably do not provide much useful information on the acute health effects of coarse particles. The lack of uniformity of concentration across urban areas leads to large errors in the assumption that the daily exposure to coarse particles is the same for everyone in the community. Therefore, even in a community where coarse particle concentrations are higher than fine particle concentrations, a greater population base or a longer time will be required to provide adequate statistical power. To obtain definitive epidemiologic information on the health effects of coarse particles, studies must be conducted in cities where coarse particles are sufficiently uniform across the city to allow the determination of a meaningful community average of coarse particle concentration. To demonstrate this, it first will Volume 47 December 1997 Journal of the Air & Waste Management Association 1247

12 be necessary to show that existing coarse particle monitors have sufficient precision or to develop new monitors that have adequate precision. Also, new epidemiologic studies should be conducted in cities where fine and coarse particles are not highly correlated and where coarse particles dominate. Conducting epidemiologic studies with PM resolved into fine particles and thoracic coarse particles is only one step in the search for the key components of PM that are responsible for human health effects. Clinical and laboratory animal exposure studies will be needed to identify biologically active components. Once biologically active components have been suggested or identified by such studies, those specific components should be included in exposure measurements for time-series epidemiology. To the extent that specific components, or specific sources, are not highly correlated with each other or with PM mass, epidemiologic studies using time series of components or sources might suggest biologically active components or sources. CONCLUSIONS In addition to size, fine and coarse particles differ in sources, formation mechanisms, composition, atmospheric lifetimes, spatial distribution, temporal variability, and, probably, in biological effects. Fine and coarse particle concentrations, at most sites where measurements are available, are poorly correlated. The combination of fine and coarse PM into reduces the day-to-day variability of and, therefore, its effectiveness as a particle indicator for time-series epidemiology. Measurements in several cities suggest that fine particle concentrations are more spatially uniform than coarse particle concentrations. Therefore, the day-to-day variability of the concentration of ambient fine particles at one site may be expected to give a reasonable indication of the variability in the concentration of ambient fine particles across the community. However, either because of less uniformity or lower measurement precision, the concentrations of coarse particles at one site will give a much poorer indication of the variability of the concentration of coarse particles across the community. Most epidemiologic studies have used TSP or as the PM indicator. The personal exposure to ambient particles (indoor plus outdoor) and, therefore, the average community exposure, will be a smaller fraction of the average ambient concentration for coarse particles than for fine particles. Based on the above observations, the hypothesis is advanced that the reported statistical associations of variations in health outcomes with variations in PM indicators, such as TSP and, may reflect an underlying association of health outcomes with variations in fine particles. Because of their differences in sources, composition, and properties, fine and coarse particles should be considered as separate classes of pollutants and measured separately in research and epidemiologic studies. The geographic distribution of fine and coarse particles needs to be determined before measurements at a central site may be used with confidence to represent concentrations of either fine or coarse particles across a large urban area. ACKNOWLEDGMENTS The authors thank Eric Smith, Michelle Wayland, and Terence Fitz-Simons for making available R 2 s for the paired relationships between, (data from AIRS). Thanks also are due to David Mage, Allan Marcus, Dennis Kotchmar, and Lawrence Cox for helpful discussions. The information in this document has been funded by the U.S. Environmental Protection Agency, in part through cooperative agreement CR with the Harvard School of Public Health. It has been subjected to EPA s peer and administrative review and approved for publication. An early version of this paper was presented at the Seminar on Trends in Aerosol Research at Gerhard Mercator University in Duisburg, Germany, in January Mention of trade names or commercial products does not constitute endorsement or recommendation for use. REFERENCES 1. Air Quality Criteria for Particulate Matter; U.S. Environmental Protection Agency, National Center for Environmental Assessment: Research Triangle Park, NC, 1996; EPA/600/P-95/001bF; Chapter 13. Also, NTIS, Springfield, VA; PB Bachmann, J.D.; Damberg, R.J.; Caldwell, J.C.; Richmond, H.M.; Polkowsky, B.; Smith, E.G.; Woodruff, T.J.; Martin, K.M.; Edwards, C.; Koman, P.D. Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of Scientific and Technical Information. OAQPS staff paper; U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards: Research Triangle Park, NC, 1996; EPA/452/R-96/013. Also, NTIS, Springfield, VA; PB REB. 3. Dockery, D.W.; Pope, C.A., III. Ann. Rev. Public Health 1994, 15, Air Quality Criteria for Particulate Matter; U.S. Environmental Protection Agency, National Center for Environmental Assessment: Research Triangle Park, NC, 1996; EPA/600/P-95/001bF; Chapter 12. Also, NTIS, Springfield, VA; PB Vedal, S. J. & Air Waste Manage. Assoc. 1997, 47, Samet, J.M.; Zeger, S.L.; Berhane, K. The association of mortality and particulate air pollution, In Particulate Air Pollution and Daily Mortality: Replication and Validation of Selected Studies, the Phase I Report of the Particle Epidemiology Evaluation Project. Health Effects Institute: Cambridge, MA, 1995; pp Beck, B.D.; Brain, J.D. Prediction of the pulmonary toxicity of respirable combustion products from residential wood and coal stoves, Proceedings of the 1982 Residential Wood and Coal Combustion Specialty Conference; Air Pollution Control Association: Pittsburgh, PA, 1982; pp Dreher, K.; Jaskot, R.; Richards, J.; Lehmann, J.; Winsett, D.; Hoffman, A.; Costa, D. Am. J. Respir. Crit. Care Med. 1996, 153, A Whitby, K.T. Atmos. Environ. 1978, 12, Air Quality Criteria for Particulate Matter; U.S. Environmental Protection Agency, National Center for Environmental Assessment: Research Triangle Park, NC, 1996; EPA/600/P-95/001bF; Chapter 3. Also, NTIS, Springfield, VA; PB Lundgren, D.A.; Burton, R.M. Inhalation Toxicol. 1995, 7, Fed. Regist. 1987, 52, CFR Chow, J.C. J. & Air Waste Manage. Assoc. 1995, 46, Whitby, K.T.; Charlson, R.E.; Wilson, W.E.; Stevens, R.K. Science (Washington, DC) 1974, 183, Miller, F.J.; Gardner, D.E.; Graham, J.A.; Lee, R.E., Jr.; Wilson, W.E.; Bachmann, J.D. J. Air Pollut. Control Assoc. 1979, 29, Rodes, C.E.; Evans, E.G. Atmos. Environ. 1985, 19, Journal of the Air & Waste Management Association Volume 47 December 1997