How Does Infiltration Behavior Modify the Composition of Ambient

Seasonal differences in CA and TX data sets were not significant (CA ...... and area sources in the Big Bend regional aerosol visibility and observati...
0 downloads 0 Views 518KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 7315-7321

How Does Infiltration Behavior Modify the Composition of Ambient PM2.5 in Indoor Spaces? An Analysis of RIOPA Data Q I N G Y U M E N G , †,§ B A R B A R A J . T U R P I N , †,|,* J O N G H O O N L E E , †,‡ A N D R E A P O L I D O R I , †,¥ CLIFFORD P. WEISEL,| MARIA MORANDI,⊥ STEVEN COLOME,# JUNFENG ZHANG,| THOMAS STOCK,⊥ AND ARTHUR WINERX Department of Environmental Sciences, Rutgers University, 14 College Farm Road, New Brunswick, New Jersey 08901, Environmental and Occupational Health Sciences Institute, 170 Frelinghuysen Road, Piscataway, New Jersey 08854, School of Public Health, Houston Health Sciences Center, University of Texas, 1200 Hermann Pressler, Houston, Texas 77030, Integrated Environmental Sciences, 5319 University Drive #430, Irvine, California 92612, and Environmental Science and Engineering Program, School of Public Health, University of California, 650 Charles E. Young Drive, 46-081 CHS, Los Angeles, Los Angeles, California 90095

The indoor environment is an important venue for exposure to fine particulate matter (PM2.5) of ambient (outdoor) origin. In this work, paired indoor and outdoor PM2.5 species concentrations from three geographically distinct cities (Houston, TX, Los Angeles County, CA, and Elizabeth, NJ) were analyzed using positive matrix factorization (PMF) and demonstrate that the composition and source contributions of ambient PM2.5 are substantially modified by outdoor-to-indoor transport. Our results suggest that predictions of “indoor PM2.5 of ambient origin” are improved when ambient PM2.5 is treated as a combination of four distinct particle types with differing infiltration behavior (primary combustion, secondary sulfate and organics, secondary nitrate, and mechanically generated PM) rather than as a “single internally mixed entity.” Studywide average infiltration factors (i.e., fraction of ambient PM2.5 found indoors) for Relationship of Indoor, Outdoor, and Personal Air (RIOPA) study homes were 0.51, 0.78, and 0.04 (consistent with P ) 0.6, 0.9, and 0.09; k ) 0.2, 0.1, and 0.6 h-1) for PM2.5 associated with primary combustion, secondary formation (excluding nitrate), and mechanical * Corresponding author phone: 732-932-9800 x6219; fax: 732932-8644; e-mail: : [email protected]. † Rutgers University. | Environmental and Occupational Health Sciences Institute. ⊥ University of Texas. # Integrated Environmental Sciences. X University of California, Los Angeles. § Currently at National Center for Environmental Assessment, U.S. Environmental Protection Agency, B-243-01, Research Triangle Park, NC 27711. ¥ Currently at Department of Civil and Environmental Engineering, University of Southern California, 3620 South Vermont Ave., Los Angeles, CA 90089. ‡ Currently at: South Coast Air Quality Management District, 21865 Copley Dr., Diamond Bar, CA 91765. 10.1021/es070037k CCC: $37.00 Published on Web 09/22/2007

 2007 American Chemical Society

generation, respectively. Modification of the composition, properties, and source contributions of ambient PM2.5 in indoor environments has important implications for exposure mitigation strategies, development of health hypotheses, and evaluation of exposure error in epidemiological studies that use ambient central-site PM2.5 as a surrogate for PM2.5 exposure.

Introduction Epidemiological studies have consistently found a positive association between ambient (outdoor) fine particulate matter (PM2.5) mass concentrations and daily morbidity and mortality (1). This implies an association between adverse health effects and exposure to PM2.5 of ambient origin. On average, U.S. residents spend more than 85% of their time indoors (2), which makes the indoor environment an important venue for exposure to PM of indoor and ambient outdoor origin. Indoor PM2.5 mass concentrations (Ci, µg/ m3) can be described as the sum of indoor-generated PM2.5 (Cig, µg/m3) and ambient outdoor-generated PM2.5 that has penetrated indoors and remains suspended (Cag, µg/m3), where Cag is the product of the ambient outdoor concentration (Ca, µg/m3) and the infiltration factor (FINF, dimensionless) and FINF is a function of the air exchange rate (a, h-1), particle loss rate (k, h-1) and particle penetration coefficient (P, dimensionless):

Ci )

Pa C + Cig ) FINFCa + Cig ) Cag + Cig a+k a

(1)

Particle size and thermodynamics have a large impact on PM2.5 losses during outdoor-to-indoor transport and in indoor environments (3-7). For example, indoor losses of nitric acid to surfaces can shift the thermodynamic equilibrium of NO3-, substantially depleting nitrate from the particle phase (4). Because particles generated through different mechanisms (i.e., from different source types) have different size distributions, composition, and thermodynamic properties, the outdoor-to-indoor PM2.5 infiltration factor (fraction of outdoor PM2.5 that is found in indoor air; FINF) will vary with source mix. It logically follows that the composition, properties, and source contributions of ambient (outdoor) PM2.5 could be modified in indoor spaces. This is important to note when considering exposure mitigation strategies, developing hypotheses relating sources/components to health endpoints, and evaluating the results of PM epidemiology (i.e., epidemiology based on ambient PM monitoring data). In this work, positive matrix factorization (PMF) was conducted on residential indoor and outdoor PM2.5 species concentrations to describe the indoor and outdoor aerosol as a sum of several factors of covariant species. These factors were then used to investigate the infiltration behavior of major PM2.5 components and to examine how the infiltration process alters the characteristics of ambient PM2.5.

Methods Sampling and Chemical Analysis. This analysis uses 48-h integrated concentrations of indoor and outdoor PM2.5 mass (gravimetric), 24 elements (XRF spectroscopy), and PM2.5 organic carbon and elemental carbon (OC and EC; thermaloptical transmittance with adsorption artifact correction). Nitrate was not measured, although it is a major component of PM2.5 in the Southwest. Species were measured at 279 nonsmoking homes, 1-4 homes at a time, during the VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7315

Relationship of Indoor, Outdoor, and Personal Air (RIOPA) study (1999-2001) in Los Angeles County, CA, Elizabeth, NJ, and Houston, TX. (abbreviated as CA, NJ, and TX, respectively.) Samples are reasonably well distributed across cities (CA: 104; NJ: 78; TX: 81) and seasons (May-Oct: 165; NovApr: 98). Some were particularly close to primary PM sources (within tens of meters) while others were further, impacted by a mix of urban and regional pollutants. Detailed information about the study is provided elsewhere (8-10). PMF Analysis. PMF is well suited for use in exposure studies where indoor, outdoor, and personal concentrations are not necessarily independent, some species are above detection limits in one environment and below in another, and not all source types have been identified/characterized (11). PMF provides factors of covariant species and the contribution of each factor mass to PM mass in each sample. Species can be covariant because they are emitted from a common source, transformed through a common process, or transported together from a common source region. PMF has been applied to ambient PM source apportionment (e.g., (12-14)) and exposure assessment (e.g., (15-18)) studies. In this work, bilinear (PMF2) and trilinear (PMF3) models (11) (see also Supporting Information) were used to identify factors of covariant chemical species. PM2.5 mass concentrations were then regressed on the factor scores, providing the mass concentrations of each factor in each sample (scores) in units of µg/m3 and describing each factor (loadings) as a mass fraction of chemical species in units of ng of measured PM2.5 species/µg of PM2.5 mass. PMF3 analyses of pooled indoor-outdoor data provided estimated factor scores and loadings for each home and a modifier describing the relative factor contribution to the indoor versus outdoor sample at that home. The PMF inputs were the species concentration (xij) and uncertainty (σij) matrices, defined as follows for determined values:

xij ) vij, σij ) (uij2 + dij2)1/2

(2)

for values below detection limits:

xij ) dij/2, σij ) dij

(3)

xij ) vij(geom), σij ) 4vij(geom)

(4)

for missing values:

where vij is the observed concentration; uij is the measurement precision, dij is the detection limit, and vij(geom) is the geometric mean of the observed concentrations. (See Supporting Information for species uncertainties). For each city, the model was run specifying 4 to 12 factors. To better understand the stability of a PMF solution and to ensure the solution was a global minimum, for each number of factors at least five runs were conducted starting from different pseudorandom starting points. During the analysis, the optimal number of factors was determined by two criteria: additional factors provided no further physically meaningful insight of the studied system, and the experimental value of Q was close to the theoretical Q value, where Q is the sum of squares of errors weighted by uncertainties. Also the results were stable, i.e., they varied little between different initial pseudorandom starting points. Describing PM2.5 as a Combination of Indoor and Outdoor “Particle Types”. PMF was used to characterize measured indoor PM2.5 as the sum of factors of covariant PM2.5 species that described indoor-generated and outdoorgenerated PM2.5. For each city, PMF2 was first run using only outdoor samples (N ) 76-103). Then the pooled indoor and 7316

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 21, 2007

outdoor dataset for each city was analyzed by PMF2. Analysis of pooled indoor and outdoor data provides common factor definitions for both types of samples. On the basis of the outdoor-only PMF results, those factors that were deemed to represent indoor-generated PM2.5 were identified and their scores were “pulled down” in the outdoor samples (i.e., with a “G-key matrix”; see Supporting Information), forcing these factors to have concentrations near zero in the outdoor samples. In this way, indoor PM2.5 at each home was described as the sum of several distinct types of PM generated outdoors and generated indoors. Each “particle type” has a consistent definition from home to home, indoors and outdoors within a city, and is allowed to differ between cities. Results provided below focus on insights into ambient PM2.5 infiltration obtained from the pooled indoor-outdoor PMF2 solution. PMF3 analyses are examined as part of the sensitivity analysis. To accomplish the objectives of this paper, PMF factors representing particle types with similar formation mechanisms were combined to describe PM2.5 as the sum of particulate matter formed (1) through combustion (primary), (2) in the atmosphere (secondary), and (3) through mechanical processes (e.g., abrasion). These three “particle types” have distinct size distributions and chemical composition and are likely to have different infiltration behavior, as well as different effects.

Results Pooled Indoor-Outdoor PMF2 Results. Good agreement between modeled and measured PM2.5 mass, the stability of factor contributions across PMF runs, and the narrow, normally distributed residuals provide confidence in the “goodness of fit” of the PMF model. Factor concentrations varied by less than 1% across the 5 PMF2 analyses conducted with the same number of factors but a different initial random seed value. The sum of the factor concentrations also compared reasonably well to measured PM2.5 mass (Figure 1), which means that measured mass was satisfactorily explained by the model. There was greater scatter between measured and modeled PM2.5 mass in the CA and TX datasets than in the NJ dataset. This might result from the fact that nitrate is a major component of PM2.5 in southern California and the southwestern U.S. (19, 20), and it was not measured during the RIOPA study. The lower correlations for indoor samples than for outdoor samples (Figure 1) are expected because of larger variations in the strength and composition of indoor emissions. The residuals scaled by their standard deviations were normally distributed and within three standard deviations with a couple of outliers. The exceptions are as follows: residual distributions for Br (all cities) and Cr (NJ) were relatively broad, and several outliers were found for Zn and Cl, implying those elements were affected by extra unidentified factors. For each city the pooled indoor-outdoor PMF2 solution provided 2-3 factors present in indoor samples and not found in the outdoor only PMF2 solution (Figure S2). Note that the sample size (150-200 samples per city) is not sufficient to provide detailed source identification for the ambient outdoor aerosol, and only half of these samples contain information about indoor sources. Also, the variability of indoor sources (both composition and concentration) is expected to be quite large from home to home, which makes the separation and identification of indoor sources even harder. Thus, these 2-3 indoor-generated PM2.5 factors should not be construed to represent pure sources; together, they describe particulate material emitted or formed indoors that is distinctly different from outdoor-generated PM2.5. Better resolution of indoor sources using a multilinear engine (16, 18) is the subject of a subsequent publication. The text below focuses on outdoorgenerated PM2.5 and insights into particle infiltration.

FIGURE 1. Measured and modeled PM2.5 mass concentrations for Los Angeles Co., CA, Elizabeth, NJ, and Houston, TX, indoor and outdoor data sets; results were generated by PMF2 with pooled indoor and outdoor samples. Dashed line is 1:1.

FIGURE 2. Factor profiles for outdoor-generated PM2.5 factors obtained by PMF2 analyses of pooled indoor and outdoor samples for Los Angeles County, CA (a), Elizabeth, NJ (b), and Houston, TX (c) data sets. Four common factors representing outdoor-generated PM2.5 were found across all cities (Figure 2). These are labeled “Secondary Aerosol” (characterized by S and OC), “Mobile/ Road Dust” (characterized by Zn, Fe, OC, EC, and soil elements), “Soil” (characterized by Al, Ca, Fe, Ti, Si), and “Combustion 1 with V” (characterized by V, Ni, EC, and OC). Additional factors labeled “Combustion 2 with K”, “Sea Salt”, and “Sea Salt plus combustion” were found in TX, NJ, and CA analyses, respectively. Each factor was labeled based on the chemical composition profiles, seasonal variability, and knowledge about the composition of sources and atmospheric processing. Each factor was named based on features suggesting the influence of a particular source type. However, more than one source type is likely to contribute to most factors. The first common outdoor factor, labeled “Secondary Aerosol,” is elevated in sulfur and organic carbon but not elemental carbon (Figure 2). The concentrations of this factor were greater during the hot season (May-October) than the cool season (November-April) according to a paired twosided t-test with unequal variance (CA p-value < 0.001; NJ

p-value ) 0.04; TX p-value ) 0.07). This is consistent with expectations for a photochemical aerosol. The second factor that is common to all three geographic locations is labeled “Mobile/Road Dust.” It is distinguished by elevated OC and EC, from fuel combustion, Fe and Zn from motor vehicle brake and diesel additives, and trafficgenerated soil elements (21). The ratios of OC/EC were 1.7, 1.8 and 1.3 for CA, NJ, and TX study homes, respectively. OC/EC ratios of 1-3 are typically observed for diesel and gasoline-powered motor vehicle emissions. A similar factor has been found elsewhere (e.g., (13, 15, 22)). Concentrations of this factor were significantly greater during the cool season than during the hot season for the NJ data set (NJ p-value ) 0.004; two-sided paired t-test, equal variance). Seasonal differences in CA and TX data sets were not significant (CA p-value ) 0.85; TX p-value ) 0.83). Higher concentrations of primary PM are frequently found in wintertime due to lower mixing heights. The factor labeled “Soil” is elevated in typical soil elements: Al, Ca, Fe, K, Si, and Ti. Silicon was elevated in the TX data set, consistent with the presence of high evolution VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7317

soil types in southern states. The CA soil factor was relatively high in organic carbon (i.e., soil organic matter). This is expected in a soil substantially affected by human activities, either pollution or agriculture. The NJ soil factor was elevated in Al, Si, and OC, consistent with a relatively low degree of soil evolution (23). This factor showed no significant seasonal pattern (CA p-value ) 0.63; NJ p-value ) 0.94; TX p-value ) 0.53; two-sided paired t-test) with low contributions most of the time but occasional high contributions presumably representing wind episodes. The factor labeled “Combustion 1 with V” is distinguished by elevated V and Ni. Electric utilities, large industrial and commercial combustion sources, and other oil combustion activities (e.g., home heating) contribute to this factor. The presence of Pb, Zn, Cl, and K probably reflects other primary industrial combustion sources including incinerators and noncoal electric power generation. A similar factor has been found previously (e.g., (13)). Concentrations of this factor might be expected to be higher in the winter due to lower mixing heights and increased emissions from heating, but increased electric power generation for air conditioning might compensate for this in hot climates. Hot and cool season factor concentrations were not significantly different for NJ and TX (NJ p-value ) 0.67; TX p-value ) 0.52; 2-sided paired t-test). Hot season concentrations were unexplainably higher for the CA data set (CA p-value < 0.001). A “Sea Salt” factor was found in the NJ data set, identified by the presence of Cl. The coexistence of Cl and S could indicate conversion from NaCl to Na2SO4 by the reaction between fresh sea salt and gas-phase sulfuric acid (24). Concentrations of this factor did not exhibit significant seasonal differences, but particularly high concentrations were seen occasionally in the winter. The CA data set also included a “sea salt” factor, but this factor also had substantial contributions from primary combustion tracers (i.e., Fe, Zn, EC, OC) in addition to Cl, suggesting that this factor is a combination of particulate matter formed through mechanical generation (i.e., sea salt) and local primary combustion. The CA factor was labeled “Sea Salt plus Combustion.” The final outdoor factor was labeled “Combustion 2 with K” because it was dominated by OC, EC, and K. This factor was found only in the TX dataset. It is probably associated with local residential wood burning (25, 26) and/or meat cooking (22, 27). This factor had significantly higher concentrations in the cool season than the hot season (TX p-value < 0.001; two-sided paired t-test, unequal variance). With a larger sample size, this type of factor would probably be found in CA and NJ data sets as well. The identified factors covered three types of particle formation mechanisms: primary combustion (Combustion 1, Combustion 2, and Mobile), secondary formation (secondary OC and S), and mechanical generation (soil and sea salt). Method Comparison. PMF3 and PMF2 of pooled datasets provided consistent results (i.e., factor concentrations; Figure S4). For factors representing outdoor-generated PM2.5, the difference between the two methods was typically less than 1 µg/m3. For factors representing indoor-generated PM2.5, the intermethod variability is larger, but differences were still typically within 2 µg/m3. The outdoor-only PMF2 factor concentrations differed from the other two approaches for factors labeled “Combustion 1 with V” in CA and NJ datasets, and “Mobile/road dust” in the NJ dataset. The lower stability in the results for these two primary-combustion factors probably results from the limitations of the small sample size in the outdoor-only analysis. Infiltration Behavior of “Particle Types”. The paired indoor and outdoor factor contributions from the PMF models provide an opportunity to investigate the infiltration behavior of different types of ambient PM2.5. Below, infiltra7318

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 21, 2007

FIGURE 3. Indoor and outdoor contributions for outdoor PM2.5 factors: (a) mechanical generation; (b) primary combustion; (c) secondary formation. The dashed line is the 1:1 line. AER is the air exchange rate. tion factors for the three particle types (primary combustion, secondary formation, and mechanical generation) were estimated from the pooled indoor-outdoor PMF2 factor contributions and compared with those obtained from the PMF3 output, and with estimates obtained using tracer elements (Zn for primary combustion, S for secondary formation, and Si for mechanical formation) (21). Figure 3 shows the indoor and outdoor factor concentrations (segregated by air exchange rate) for the mechanical generation factors (Soil, Sea Salt), for primary combustion factors (Combustion 1, Combustion 2, Mobile) and for the factor labeled “Secondary.” Table 1 provides FINF estimates based on PMF2 (i.e., slopes of the linear regression of Ci on Ca from Figure 3; see also eq 1) and, for comparison, estimates based on PMF3 and for source tracers. For the latter, indoor concentrations of Zn, S, and Si were regressed on outdoor concentrations to provide FINF estimates (slopes) representing primary combustion, secondary formation, and mechanical

TABLE 1. Infiltration Estimates (slope) with (standard error) for Particle Types source type

CAa

NJ

TX

Overallb

PMF3c

Tracerd

primary combustion secondary formation mechanical generation

0.66 (0.03) 0.82 (0.03) 0.52e (0.02)

0.65 (0.03) 0.63 (0.07) 0.01 (0.02)

0.51 (0.03) 0.48 (0.05) 0.10 (0.01)

0.51 (0.02) 0.78 (0.03) 0.04f,g (0.02)

0.47 (0.02) 0.79 (0.007) 0.04 (0.004)

0.54 (0.03) 0.76 (0.03) 0.07 (0.02)

a Slopes are based on pooled indoor and outdoor PMF2 output for each state. b Slopes are based on pooled indoor and outdoor PMF2 output for all three states. c Slopes are based on pooled indoor and outdoor PMF3 output for all three states. d Slope for each element (Zn for primary combustion; S for secondary formation; Si for mechanical generation (21)) is based on the regression of indoor concentrations on outdoor concentrations for that element in the overall dataset. e The sea salt factor in the CA data set contains co-transport primary combustion elements and therefore is not a good estimate of the infiltration factor of mechanically generated particles. f CA mechanical formation factors were excluded from this calculation. g Slope is statistically significant (p-value ) 0.008).

generation, respectively. It is possible that some PM generated indoors could have a composition so similar to that generated outdoors that it would not separate into its own factor. If this were true, individual values of Ci/Ca could overestimate FINF, but in the regression the indoor-generated PM would contribute to the intercept (Cig in eq 1) and the slope would provide a much more robust (pooled) estimate of FINF (assuming Cag and Cig are independent). FINF estimates provided in Table 1 are quite consistent across methods. Air exchange rates for TX study homes were typically lower than for CA and NJ study homes (9). Thus it makes sense that infiltration factor estimates were generally smaller for the TX dataset. The study-wide FINF estimates were largest for secondary PM2.5 (0.78 ( 0.03; slope ( standard error), smaller for primary combustion PM2.5 (0.51 ( 0.02), and smallest for mechanically generated PM2.5 (0.04 ( 0.02). Primary combustion forms nucleation and condensation mode particles, which experience diffusional losses due to their small size. Secondary formation results in particles in these size modes but also forms particles in the somewhat larger droplet mode, less affected by diffusional losses. Mechanical forces generate coarse mode particles, the tail of which is included in PM2.5. Coarse mode particles have low diffusional losses but large inertial losses, yielding low infiltration factors. Thus FINF results are consistent with expectations. Several controlled experiments have now been conducted to measure P or k using particles of known size (5, 6, 28-36). Using the median value of P (range ) 0.01-0.63; median ) 0.09) and k (range ) 0.3-1.8; median ) 0.6 h-1) reported in the literature for 2 µm particles (a size selected to represent mechanically generated PM2.5) and the measured air exchange rate for each RIOPA home, a mean FINF of 0.05 is obtained. This is similar to FINF ) 0.04, obtained in this research. Literature values of P and k for 0.5 - 0.9 µm particles (a size selected to represent secondary PM2.5) range from P ) 0.6 to 0.99 and k ) 0.08 to 0.35 h-1 (5-7, 28-34). The infiltration factor for secondary aerosol in our study (FINF ) 0.78) is consistent with the infiltration factor calculated from the following four studies: (1) P ) 0.85, k ) 0.08 h-1, mean FINF ) 0.76 ((31); ∼0.5 µm), (2) P ) 0.88, k ) 0.15 h-1, mean FINF ) 0.72 ((30); ∼0.5 µm), (3) P ) 0.9, k ) 0.1 h-1, mean FINF ) 0.78 ((5); ∼0.9 µm), and 4) P ) 0.95, k ) 0.1 h-1, mean FINF ) 0.83 ((31); ∼0.9 µm). Literature values for P of 0.1-0.95 and k of 0.1-3.5 are reported for 0.01-0.3 µm particles, a size range that might reasonably include primary combustion PM2.5. The infiltration factor for primary combustion PM2.5 obtained in our study (FINF ) 0.51) is consistent with that calculated with measured air exchange rate and P and k from the following studies: (1) P ) 0.6, k ) 0.2 h-1, mean FINF ) 0.48 ((5); ∼0.01 µm), (2) P ) 0.87, k ) 0.59 h-1, mean FINF ) 0.50 ((30); ∼0.01 µm), and (3) P ) 0.55, k ) 0.19 h-1, mean FINF ) 0.44 ((30); ∼0.01 µm). Infiltration factor differences for different particle types suggest that the composition of ambient PM2.5 will be modified in indoor spaces. In other words, the composition and size distribution of “PM2.5 of outdoor origin” will differ

FIGURE 4. Contributions of outdoor factors to outdoor PM2.5 and indoor PM2.5 of outdoor origin (the first number is concentration in µg/m3; the second number is the percentage contribution to PM2.5 mass). from outdoor PM2.5. Figure 4 shows the mean contribution of each “particle type” to outdoor PM2.5 and to indoor PM2.5 of outdoor origin. The composition of “PM2.5 of outdoor origin” is significantly different from that of outdoor PM2.5. Secondary aerosol was significantly (on a percentage basis; 95% confidence intervals) enriched in the indoor environment (from 40% outdoors to 55% in the indoor environment), and mechanically generated PM2.5 was significantly (95% confidence intervals) depleted in the indoor environment (from 17% outdoors to 2% in the indoor environment). The relative abundance of primary combustion PM2.5 indoors and outdoors was not significantly altered.

Discussion This work demonstrates that that the composition and source contributions of ambient outdoor PM2.5 can be substantially altered in indoor spaces, due to differential infiltration behavior. Infiltration factors were estimated for three distinct particle types with differing infiltration behavior: primary combustion (FINF ) 0.51 ( 0.02), secondary formation (i.e., secondary sulfate and organics; FINF ) 0.78 ( 0.03), and mechanically generated PM (FINF ) 0.04 ( 0.02). A fourth distinctly different particle type of potential significance that was not measured in this study, is secondary nitrate. The scrubbing of gas-phase nitric acid by indoor surfaces can lead to substantial losses of particle phase ammonium nitrate in indoor environments (4). These results suggest that the source contributions, chemical composition, mass concentration, size distribution, and behavior of “indoor PM2.5 of VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7319

ambient origin” will be better predicted when ambient PM2.5 is treated as a combination of four distinct particle types with differing infiltration behavior (primary combustion, secondary sulfate and organics, secondary nitrate, and mechanically generated PM2.5) rather than as a “single internally mixed entity.” The infiltration factor, the fraction of ambient (outdoor) PM that is found indoors, depends on the air exchange rate, particle size distribution, thermodynamic properties of the PM species, and house characteristics, which vary from hometo-home and/or by city and season (8, 37). The treatment of ambient PM2.5 as a combination of four particle types (rather than one) accommodates, to a certain degree, variations in particle properties through variations in the contribution of each particle type to the PM2.5 mix. Variations in air exchange rate can be accommodated when air exchange rate, P, and k are known. The infiltration factors obtained in this study are consistent with the following P and k values: P ) 0.6, k ) 0.2 h-1 for primary combustion ((5, 30); P ) 0.9, k ) 0.1 h-1 for secondary sulfate and organics (5, 30, 31), and P ) 0.09, k ) 0.6 h-1 for mechanical generation (31). Lunden (4) successfully predicted indoor nitrate from outdoor nitrate concentrations in an unoccupied home under a wide range of air exchange rates (0.2-6 h-1) with a model that included physical losses (P ) 0.8; k ) 0.12 h-1 for particle deposition) and nitrate evaporation (k ) 0.3-18 h-1 for evaporation); fitted deposition velocities were reported. One limitation of this work is that secondary sulfates and organics are presented as one particle type despite some important differences in their behavior. Most notably, many organic PM compounds are semivolatile and their gas-particle partitioning is altered with outdoor-to-indoor transport when they encounter a change in temperature or a change in the quantity of carbonaceous PM into which they can sorb (e.g., the addition of indoor-generated OC) (10). Because PMF was conducted on the indoor and outdoor data together (i.e., the “secondary formation” factor was constrained to be identically defined in indoor and outdoor samples) and because secondary sulfate and OC were represented by a single factor, this effect is probably dampened to some degree in the analysis above. The change in organic gas-particle partitioning with infiltration warrants further investigation. The modification of ambient PM2.5 in indoor environments, where people spend most of their time, has important implications for exposure mitigation strategies and the development of health hypotheses. The implications to epidemiological studies in which central-site outdoor PM2.5 is used as a surrogate for human exposure to ambient outdoor-generated PM2.5, should be investigated.

Acknowledgments This research was supported by the Health Effects Institute (#98-23-2), the Mickey Leland National Urban Air Toxics Center, the NIEHS Center of Excellence (ES05022), the NJ Agricultural Experiment Station, an EPA/NCEA-ORISE research fellowship (for Q. Meng), and an EPA STAR Graduate Fellowship (for D. Shendell). We acknowledge valuable discussions with Drs. Pentti Paatero and Melissa Lunden. Research was conducted, in part, under contract to the Health Effects Institute (HEI), an organization jointly funded by the United States Environmental Protection Agency (R828112) and automotive manufacturers. The contents of this article do not necessarily reflect the views of HEI nor the views and policies of EPA or of motor vehicle and engine manufacturers. The PMF model was used under a licensing agreement with Dr. Pentti Paatero of the University of Helsinki, Finland.

Supporting Information Available PMF model structure; species data quality measures; PMF2 of outdoor samples alone; additional pooled indoor and 7320

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 21, 2007

outdoor PMF2 factor results; PMF3 results; comparison of three PMF models. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) U.S. EPA. Air quality criteria for particulate matter (October 2004); U.S. Environmental Protection Agency: Washington, DC, 2004. (2) Klepeis, N. E.; Nelson, W. C.; Ott, W. R.; Robinson, J. P.; Tsang, A. M.; Switzer, P.; Behar, J. V.; Hern, S. C.; Engelmann, W. H. The national human activity pattern survey (NHAPS): A resource for assessing exposure to environmental pollutants. J. Exposure Anal. Environ. Epidemiol. 2001, 11, 231-252. (3) Thatcher, T. L.; Layton, D. W. Deposition, resuspension, and penetration of particles within a residence. Atmos. Environ. 1995, 29, 1487-1497. (4) Lunden, M. M.; Revzan, K. L.; Fischer, M. L.; Thatcher, T. L.; Littlejohn, D.; Hering, S. V.; Brown, N. J. The transformation of outdoor ammonium nitrate aerosols in the indoor environment. Atmos. Environ. 2003, 37, 5633-5644. (5) Thatcher, T. L.; Lunden, M. M.; Revzan, K. L.; Sextro, R. G.; Brown, N. J. A concentration rebound method for measuring particle penetration and deposition in the indoor environment. Aerosol Sci. Technol. 2003, 37, 847-864. (6) Liu, D. L.; Nazaroff, W. W. Particle penetration through building cracks. Aerosol Sci. Technol. 2003, 37, 565-573. (7) Hussein, T.; Korhonen, H.; Herrmann, E.; Hameri, K. H.; Lehtinen, K. E. J.; Kulmala, M. Emission rates due to indoor activities: Indoor aerosol model development, evaluation, and applications. Aerosol Sci. Technol. 2005, 39, 1111-1127. (8) Weisel, C. P.; Zhang, J. F.; Turpin, B. J.; Morandi, M. T.; Colome, S.; Stock, T. H.; Spektor, D. M.; Korn, L.; Winer, A.; Alimokhtari, S. et al. Relationship of indoor, outdoor and personal air (RIOPA) study: Study design, methods and quality assurance/control results. J. Exposure Anal. Environ. Epidemiol. 2005, 15, 123137. (9) Meng, Q. Y.; Turpin, B. J.; Korn, L.; Weisel, C. P.; Morandi, M.; Colome, S.; Zhang, J. F. J.; Stock, T.; Spektor, D.; Winer, A. et al. Influence of ambient (outdoor) sources on residential indoor and personal PM2.5 concentrations: Analyses of RIOPA data. J. Exposure Anal. Environ. Epidemiol. 2005, 15, 17-28. (10) Polidori, A.; Turpin, B.; Meng, Q. Y.; Lee, J. H.; Weisel, C.; Morandi, M.; Colome, S.; Stock, T.; Winer, A.; Zhang, J. et al. Fine organic particulate matter dominates indoor-generated PM2.5 in RIOPA homes. J. Exposure Sci. Environ. Epidemiol. 2006, 16, 321-331. (11) Paatero, P. Least squares formulation of robust non-negative factor analysis. Chemom. Intell. Lab. Syst. 1997, 37, 23-35. (12) Polissar, A. V.; Hopke, P. K.; Paatero, P. Atmospheric aerosol over Alaska - 2. Elemental composition and sources. J. Geophys. Res.-Atmos. 1998, 103, 19045-19057. (13) Song, X. H.; Polissar, A. V.; Hopke, P. K. Sources of fine particle composition in the northeastern U.S. Atmos. Environ. 2001, 35, 5277-5286. (14) Pekney, N. J.; Davidson, C. I.; Robinson, A.; Zhou, L. M.; Hopke, P.; Eatough, D.; Rogge, W. F. Major source categories for PM2.5 in Pittsburgh using PMF and UNMIX. Aerosol Sci. Technol. 2006, 40, 910-924. (15) Yakovleva, E.; Hopke, P. K.; Wallace, L. Receptor modeling assessment of particle total exposure assessment methodology data. Environ. Sci. Technol. 1999, 33, 3645-3652. (16) Hopke, P. K.; Ramadan, Z.; Paatero, P.; Norris, G. A.; Landis, M. S.; Williams, R. W.; Lewis, C. W. Receptor modeling of ambient and personal exposure samples: 1998 Baltimore particulate matter epidemiology-exposure study. Atmos. Environ. 2003, 37, 3289-3302. (17) Larson, T.; Gould, T.; Simpson, C.; Liu, L. J. S.; Claiborn, C.; Lewtas, J. Source apportionment of indoor, outdoor, and personal PM2.5 in Seattle, Washington, using positive matrix factorization. J. Air Waste Manage. Assoc. 2004, 54, 1175-1187. (18) Zhao, W. X.; Hopke, P. K.; Norris, G.; Williams, R.; Paatero, P. Source apportionment and analysis on ambient and personal exposure samples with a combined receptor model and an adaptive blank estimation strategy. Atmos. Environ. 2006, 40, 3788-3801. (19) Malm, W. C.; Sisler, J. F. Spatial patterns of major aerosol species and selected heavy metals in the United States. Fuel Process. Technol. 2000, 65, 473-501. (20) Tolocka, M. P.; Solomon, P. A.; Mitchell, W.; Norris, G. A.; Gemmill, D. B.; Wiener, R. W.; Vanderpool, R. W.; Homolya, J. B.; Rice, J. East versus west in the U.S.: Chemical characteristics of PM2.5 during the winter of 1999. Aerosol Sci. Technol. 2001, 34, 88-96.

(21) Hopke, P. K. Receptor modeling in environmental chemistry; John Wiley & Sons, Inc.: New York, 1985. (22) Chow, J. C.; Watson, J. G.; Kuhns, H.; Etyemezian, V.; Lowenthal, D. H.; Crow, D.; Kohl, S. D.; Engelbrecht, J. P.; Green, M. C. Source profiles for industrial, mobile, and area sources in the Big Bend regional aerosol visibility and observational study. Chemosphere 2004, 54, 185-208. (23) Miller, R. W.; Gardiner, P. T. Soils in our environment, 9th ed.; Prentice Hall, Inc.: Upper Saddle River, NJ, 2001. (24) Seinfeld, J. H.; Pandis, S. N. Atmospheric chemistry and physics; John Wiley & Son, Inc.: New York, 1998. (25) Fine, P. M.; Cass, G. R.; Simoneit, B. R. T. Chemical characterization of fine particle emissions from fireplace combustion of woods grown in the northeastern United States. Environ. Sci. Technol. 2001, 35, 2665-2675. (26) Fine, P. M.; Cass, G. R.; Simoneit, B. R. T. Chemical characterization of fine particle emissions from the fireplace combustion of woods grown in the southern United States. Environ. Sci. Technol. 2002, 36, 1442-1451. (27) Watson, J. G.; Chow, J. C. Source characterization of major emission sources in the Imperial and Mexicali valleys along the U.S./Mexico border. Sci. Total Environ. 2001, 276, 33-47. (28) Abt, E.; Suh, H. H.; Catalano, P.; Koutrakis, P. Relative contribution of outdoor and indoor particle sources to indoor concentrations. Environ. Sci. Technol. 2000, 34, 3579-3587. (29) Vette, A. F.; Rea, A. W.; Lawless, P. A.; Rodes, C. E.; Evans, G.; Highsmith, V. R.; Sheldon, L. Characterization of indoor-outdoor aerosol concentration relationships during the Fresno PM exposure studies. Aerosol Sci. Technol. 2001, 34, 118-126. (30) Long, C. M.; Suh, H. H.; Catalano, P. J.; Koutrakis, P. Using timeand size-resolved particulate data to quantify indoor penetration and deposition behavior. Environ. Sci. Technol. 2001, 35, 2089-2099.

(31) Mosley, R. B.; Greenwell, D. J.; Sparks, L. E.; Guo, Z.; Tucker, W. G.; Fortmann, R.; Whitfield, C. Penetration of ambient fine particles into the indoor environment. Aerosol Sci. Technol. 2001, 34, 127-136. (32) Thatcher, T. L.; Lai, A. C. K.; Moreno-Jackson, R.; Sextro, R. G.; Nazaroff, W. W. Effects of room furnishings and air speed on particle deposition rates indoors. Atmos. Environ. 2002, 36, 1811-1819. (33) Jeng, C. J.; Kindzierski, W. B.; Smith, D. W. Modeling entry of micron-sized and submicron-sized particles into the indoor environment. Aerosol Sci. Technol. 2003, 37, 753-769. (34) Wallace, L. A.; Emmerich, S. J.; Howard-Reed, C. Effect of central fans and in-duct filters on deposition rates of ultrafine and fine particles in an occupied townhouse. Atmos. Environ. 2004, 38, 405-413. (35) Hussein, T.; Hameri, K.; Heikkinen, M. S. A.; Kulmala, M. Indoor and outdoor particle size characterization at a family house in Espoo-Finland. Atmos. Environ. 2005, 39, 3697-3709. (36) Zhu, Y. F.; Hinds, W. C.; Krudysz, M.; Kuhn, T.; Froines, J.; Sioutas, C. Penetration of freeway ultrafine particles into indoor environments. J. Aerosol. Sci. 2005, 36, 303-322. (37) Meng, Q. Y.; Turpin, B. J.; Polidori, A.; Lee, J. H.; Weisel, C.; Morandi, M.; Colome, S.; Stock, T.; Winer, A.; Zhang, J. F. PM2.5 of ambient origin: Estimates and exposure errors relevant to PM epidemiology. Environ. Sci. Technol. 2005, 39, 51055112.

Received for review January 6, 2007. Revised manuscript received July 23, 2007. Accepted August 9, 2007. ES070037K

VOL. 41, NO. 21, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7321