Article pubs.acs.org/est

Fine and Ultrafine Particle Decay Rates in Multiple Homes Lance Wallace* Consultant, 428 Woodley Way, Santa Rosa, California 95409, United States

Warren Kindzierski School of Public Health, University of Alberta, 3-57B South Academic Building, 11405-87 Avenue, Edmonton, Alberta T6G 1C9, Canada

Jill Kearney, Morgan MacNeill, Marie-Ève Héroux, and Amanda J. Wheeler Health Canada, 269 Laurier Avenue West, Ottawa, Ontario K1A 0K9, Canada S Supporting Information *

ABSTRACT: Human exposure to particles depends on particle loss mechanisms such as deposition and filtration. Fine and ultrafine particles (FP and UFP) were measured continuously over seven consecutive days during summer and winter inside 74 homes in Edmonton, Canada. Daily average air exchange rates were also measured. FP were also measured outside each home and both FP and UFP were measured at a central monitoring station. A censoring algorithm was developed to identify indoor-generated concentrations, with the remainder representing particles infiltrating from outdoors. The resulting infiltration factors were employed to determine the continuously changing background of outdoor particles infiltrating the homes. Background-corrected indoor concentrations were then used to determine rates of removal of FP and UFP following peaks due to indoor sources. About 300 FP peaks and 400 UFP peaks had high-quality (median R2 value >98%) exponential decay rates lasting from 30 min to 10 h. Median (interquartile range (IQR)) decay rates for UFP were 1.26 (0.82− 1.83) h−1; for FP 1.08 (0.62−1.75) h−1. These total decay rates included, on average, about a 25% contribution from air exchange, suggesting that deposition and filtration accounted for the major portion of particle loss mechanisms in these homes. Models presented here identify and quantify effects of several factors on total decay rates, such as window opening behavior, home age, use of central furnace fans and kitchen and bathroom exhaust fans, use of air cleaners, use of air conditioners, and indoor−outdoor temperature differences. These findings will help identify ways to reduce exposure and risk.

■

Air Exchange. All homes exchange indoor air with outdoor air at a certain base rate dependent on the construction quality of the home and on indoor−outdoor temperature and pressure differences. These rates are greatly increased by opening windows. Studies in occupied homes have estimated the effect of opening windows on the air exchange rate to range up to as high as 2 h−1 for multiple windows opened wide.6,16 A study in a test house found that a single window opened 7.5 cm increased average air exchange rates by about 56% (0.41 (SD 0.13) from 0.25 (SD 0.10) h−1).17 Another effect on air exchange is the use of exhaust fans. One study found that an attic, kitchen, or bathroom exhaust fan typically increased air exchange rates by about 0.9 h−1.18

INTRODUCTION

Both fine and ultrafine particles (FP and UFP) are associated with morbidity and mortality.1−5 Human exposure to particles in the indoor environment is governed by indoor sources, outdoor particle infiltration, and particle loss mechanisms. To estimate exposures, it is necessary to understand emissions from the various indoor sources (particularly cooking),6,7 infiltration factors as a function of building construction and resident behavior (e.g., use of air cleaners),8−12 and the factors affecting indoor particle concentrations (particularly window and fan usage and deposition on surfaces).13−15

■

PARTICLE LOSS MECHANISMS Indoor particles are lost by exfiltration (air exchange), deposition on indoor surfaces, deposition on duct work when a central furnace or air conditioning fan is running, and filtration by portable or in-duct air cleaners. In addition, for UFP only, coagulation can be important at high concentrations. © 2013 American Chemical Society

Received: Revised: Accepted: Published: 12929

June 11, 2013 October 6, 2013 October 21, 2013 October 21, 2013 dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

old.24,25 However, more recent work has shown that very small particles (modes around 5−6 nm) are produced in copious quantities by both gas and electric stovetop cooking, and improved models of coagulation suggest that the threshold is considerably lower, perhaps around 20 000 cm−3.6,7,17,26 All of the above parameters except coagulation influence the inf iltration factor, a measure of the fraction of outdoor particles entering the residence and remaining airborne under steadystate conditions. The infiltration factor is a crucial measure of humpan exposure to particles of outdoor origin. Several studies have estimated the effect of these parameters (air exchange, deposition, filtration) on the infiltration factor for size-resolved or total FP and UFP.8,10,12,17,26 Based on the steady-state solution to the differential form of the mass balance equation, the infiltration factor Finf may be written

Deposition on Indoor Surfaces (ksurf). Particle deposition has been explored in many laboratory and field studies. Deposition rates are strongly dependent on particle size. Deposition rates are high for the smallest UFP (3−10 nm) due to increased Brownian motion making it more likely that they will strike a surface. The rates decrease toward a minimum near the largest diameter (100 nm) UFP. FP deposition rates increase by size and are affected by gravitational force.13 Deposition rates are affected by the smoothness of adjacent surfaces, with rates increasing as the surface becomes rougher.14 Laboratory studies, mostly on smooth surfaces, have found quite low deposition rates, whereas field studies and laboratory studies on rough surfaces generally find rates 1−2 orders of magnitude higher.19 An early EPA-sponsored study (particle team exposure assessment methodology, or PTEAM) of 178 homes using gravimetric samplers found an average deposition rate of 0.39 (standard error (SE) 0.16) h−1 for PM2.5 and 0.65 (SE 0.28) h−1 for PM10 but could make no estimate of rates for individual homes.20 Other studies have continued to find only statistical averages for groups of homes rather than measured values for individual homes.21,22 Deposition in Ducts (kduct). When a central fan is running, it forces the air in the home to traverse a narrow tunnel several times each hour, thus greatly increasing the probability that some particles will be lost by colliding with the duct surface. Recent studies of deposition rates of UFP between 3 and 100 nm at a test house found deposition rates following a monotonic function of particle diameter on a log−log scale:23

Finf = Pa /(a + k tot)

where a = air exchange rate (h−1), P = penetration factor (dimensionless), and ktot = ksurf + kduct + kfilt where ksurf = deposition rate on surfaces (h−1), a function of house characteristics, kduct = deposition rate in the ductwork (h−1), kfilt = filtration rate by air cleaners (portable or in-duct) (h−1). Since this is a steady-state solution, it requires that P, a, and ktot be constant for whatever averaging period is desired. In this study, our air exchange rate measurement is a daily average. Since a will often vary diurnally, and kduct may show binary behavior according to a duty cycle when the central fan is operated by a thermostat, and kfilt is nonzero only when the portable air cleaner is on, it is clear that the estimate of Finf is subject to multiple factors resulting in considerable variation and some uncertainty. Indoor sources such as cooking can elevate FP and UFP to very high levels in homes. Elevated exposures following these actions can persist for hours depending on particle loss mechanisms such as air exchange rates, deposition on surfaces, filtration, and coagulation. A number of multiple-home studies have measured indoor particle concentrations, but few have been able to determine the rates of particle removal for individual homes under normal living conditions. Knowledge of the range of typical rates of removal can be useful in estimating human exposure, risk, and mitigation measures. Therefore continuous measurements from this study of FP and UFP in 74 homes in Edmonton Canada were analyzed to determine these rates of removal in each home during week-long measurement periods in summer and winter.

ln(ksurf + kduct) = 2.5(SE0.04) − 0.84(SE0.01)ln D(nm) [central fan always running]

(1)

where D is the electrical mobility diameter. The central fan in this case pulled about 6 house volumes per hour through the duct work. That result can be compared to a similar result for a case of no central fan (friction velocity u* of 3 cm/s) and smooth household surfaces:13 ln ksurf = 1.53(SE0.03) − 1.22(SE0.01)ln D(nm) [central fan off]

(3)

(2)

where the equation has been based on model parameters supplied by the authors 13 (W. W. Nazaroff, personal communication). For a 20 nm particle, these equations indicate that with the central fan on (k+kduct), the loss rate will be about 1 h−1 compared to 0.1 h−1 for the fan off. Filtration (kfilt). The highest deposition rates for particles of any size are obtained by in-duct filtration installations such as electrostatic precipitators (ESP). A multiyear study in an inhabited home found that a duct-mounted ESP could sustain whole-house deposition rates of 3−5 h−1 for UFP, PM2.5, and coarse (PM2.5−10) particles, whereas a good-quality ductmounted membrane filter sank as low as 1 h−1 for particles between 100 and 300 nm, and could only reach a maximum rate of 4 h−1 for certain particle sizes within PM2.5 and UFP.16 Coagulation. At high concentrations, UFP numbers decline rapidly due to coagulation, in which the smaller UFP tend to collide with larger slower-moving particles and thus are removed from the aerosol mixture. Textbook estimates of the concentration above which coagulation is an important loss mechanism have suggested 100 000 cm−3 as a rough thresh-

■

STUDY DESIGN AND METHODS Continuous (1 min) indoor concentrations of FP and UFP were measured for seven consecutive 24 h periods in 50 homes during winter (January−April) and 50 homes during summer (Jun-Aug) 2010, with 26 homes participating in both seasons. The homes were selected randomly based on an a priori hypothesis that age of home would impact indoor air quality. Using the most recent census data (Statistics Canada 2006) for the city of Edmonton, sampling was stratified by age of home, and residences were grouped into five construction year strata (1945 and before, 1946−1960, 1961−1980, 1981−2004, and after 2004) with 10 homes in each stratum. The only exclusion criteria were that participants had to be homeowners and nonsmokers. Samples were collected from homes in 10 randomly selected neighborhoods (2 per stratum). There were nine 7-day sampling periods per season, with approximately six homes being measured concurrently per period. FP 12930

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

emission rates were calculated using measured indoor temperatures and formulas supplied by Brookhaven National Laboratory for the PMCH and PDCH tracers. A baseline questionnaire was administered to collect information about the home (e.g., size, age, upgrades) and participants were asked to complete a daily questionnaire about personal activities that might affect particle concentrations. Cooking, window opening, and fan use were some of the activities for which start and end times were collected. Further details on all methods are provided in the Supporting Information (SI). Modeling Particle Loss Mechanisms. Peaks due to indoor sources are often easy to recognize, since they rise very sharply and tend to decline more slowly. If the main household conditions were not changed during the succeeding hours (that is, new sources were not activated, windows were not opened or closed, exhaust fans were not switched on or off), a smooth exponential decline could often be observed. This decay rate would appear as a straight line on a semilogarithmic plot of either number or mass vs time, provided that the proper indoor background could be estimated and subtracted from the observed values. In previous studies, it has been difficult to estimate this background. However, in this study, a computerized algorithm developed by Kearney et al11 was used to identify and remove FP and UFP peaks due to indoor sources. This allowed an estimate of the infiltration factor for each home and each monitoring day (see SI for more details). The infiltration factors were then multiplied by the current outdoor concentration to provide an estimate of the background to which the indoor concentration was trending at any moment. Regressions were then run of the background-corrected particle number (UFP) or estimated mass (FP) vs time. Declines that were adequately “straight” (R2 > 0.9) were selected as the best estimates of the total decay rate d of the particles. Based on the discussion above, this equation can be written as

were also measured outside each home and both FP and UFP measurements were made at the National Air Pollution Surveillance NAPS monitoring station in downtown Edmonton. FP and UFP were measured using DustTrak and P-Trak monitors, respectively (TSI Inc., St. Paul, MN), The DustTrak model 8520 devices were equipped with PM2.5 inlets; however, due to differences between the optical measurements of the DustTraks and gravimetric instruments, we refer to the DustTrak measurements as fine particles (FP), rather than PM2.5. The P-Traks (model 8525) have a lower limit of 20 nm with an upper limit of about 1 μm. Since the number of particles is a rapidly decreasing function of their size, the upper limit of the P-Trak does not normally affect the particle number. However, the lower limit of 20 nm means that freshly generated particles from gas or electric stovetop cooking (modes of 5−6 nm) will be missed.6,7 In the case of stovetop cooking (boiling, frying, grilling, sautéing), over 95% of the particles generated are ≈20 000 cm−3. The direct measurement is of d, the actual total decay rate for the particle size of interest. Therefore the magnitude and factors causing this decay are of primary interest. There is a second direct measurement, that of the 24 h average air exchange rate a. However it is also of interest to try to determine the influence of the loss mechanisms other than air exchange by indirect means (mixed models) employing questionnaire responses affecting kduct (e.g., air conditioning 12931

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

used), kfilt (e.g., air cleaner used), and coag (e.g., hightemperature stovetop cooking). These loss mechanisms can be grouped into a factor we call deposition*, the difference between total decay rate and the measured 24 h air exchange rate: deposition* = d − a = k total + ai − a

Table 1. Questionnaire Variables Employed in Mixed Model for Total Decay Rates variable numpeople home volume carpet rooms R2 ESP

(5)

The operative air exchange rate is that during the time of the decay (ai), whereas the only estimate of it is the 24 h average from the PFT measurement (a). If the instantaneous air exchange rate differs from the 24 h average then the deposition* term will include additional uncertainty due to the unmeasured quantity ai. Statistical Methods. A database of the minute-by-minute measurements for the 100 homes was created (991268 records). Daily plots of the P-Trak and DustTrak values were created and manually scanned to detect indoor-generated peaks. These were transformed to background-corrected logarithms and the resulting set of points from the beginning to the end of the peak was regressed against time. The negative of the slope was the estimated decay rate d. Two criteria were used to ensure high-quality decay rate estimates. First, the R2 value was required to exceed 90%. The median R2 value was >98%. Second, the initial peak value was required to be much higher than the indoor background level to minimize errors due to uncertainty in the background estimate. For the UFP, with a typical background level of 2000 cm−3, the minimal peak value was required to be >20 000 cm−3. The median peak value was 63 000 cm−3, far above background concentrations. Similarly, for the FP, the peak value was required to exceed 50 μg/m3, a value close to 10 times typical indoor background concentrations. Statistical software employed included SAS EG (version 4.2), Statistica v9 and v10, and Excel 2003/2007/2010. Given the repeated measures design of this study, a generalized linear mixed model with a variance components covariance structure was employed to estimate the effect on the decay rate of certain variables that might be related to air exchange, deposition, filtration, and coagulation (Table 1). The ranges of these variables are supplied in SI, Table S1. The sources of information on these variables included the baseline questionnaire on household characteristics and a daily questionnaire on household activities. For the window-opening variable, a “window index” was created by multiplying the number of windows open at any moment by the estimated width of the opening. Occupants had been asked to estimate two levels of window opening (6 in.), so the first level was assigned a value of 3 in. (7.6 cm) and the second level a value of 6 in. (15 cm). If no level was indicated, the value assigned was 5 in. (13 cm). The starting and ending times for using bathroom and kitchen fans, air cleaners, and air conditioners were compared to the starting and ending times of each peak and an index indicated the fraction of time during the peak that the appliance was on. Indoor and outdoor temperatures were measured each minute and the mean of the absolute values of the indoor-outdoor temperature differences were determined for each peak. One set of models was run on the measured total decay rate d, a second set on the measured 24 h average air exchange rate a, and a third set of models on the difference d−a (i.e., deposition*). Models were run both on the original and logtransformed measurements. The models with original data are more physically interpretable because of the principle of superposition (that is, the concentrations produced by multiple sources have no interaction and can therefore be added to

peak concentration mean air cleaner by peak mean window index by peak mean bathroom fan by peak mean kitchen fan by peak mean absolute temperature difference by peak mean furnace fan on continuously by peak air conditioner use strata

definition number of persons living in house house volume (m3) number of areas of home with carpeting R2 of regression fit to particle decay presence of electrostatic precipitator on furnace concentration at beginning of peak (P-Trak: cm−3; D-Trak: μg/m3 mean time using air cleaner during peak window index during peak bathroom fan index during peak kitchen fan index during peak mean absolute indoor-outdoor temperature difference during peak furnace fan running continually during peak (winter only) air conditioner running during peak (summer only) strata of home ages 1 - constructed prior to 1946 2 - constructed between 1946 and 1960 3 - constructed between 1961 and 1980 4 - constructed between 1981 and 2004 5 - constructed after 2004

produce the total concentration). The models with logtransformed data are generally closer to normality assumptions. The deposition* (d−a) model was expected to be influenced mostly by variables that would have an effect either on the deposition rate ksurf, the duct deposition rate kduct, the filtration rate from air cleaner use kfilt, or the coagulation rate coag, the latter applied for UFP only. However, this model has an additional uncertainty caused by the difference between the daily air exchange rate and the air exchange rate during the peak. Because the 24 h air exchange rate estimate was used in the calculation of d−a, we also included variables that might contribute to a higher air exchange rate during the peak than was measured over the 24 h period (e.g., window opening, fan use). Specifically, we created variables that were the difference of the value of each variable during the peak compared with a 24 h average value of each variable. For instance if the kitchen fan was used 10% of the 24 h period, and 100% of the time during the peak, then the difference was 90% (0.90).

■

RESULTS AND DISCUSSION Quality Assurance. A laboratory intercomparison of 12 PTraks showed biases (relative to the median value) ranging from −11% to +14%. The median bias-corrected precision was 4% (range 2−11%). Field comparisons of colocated and concurrent P-Traks (N = 14 577 min) resulted in a Spearman correlation coefficient of 0.99 and a median precision of 3.5% (IQR 1.7−6.6%). The two DustTraks at the central site had a median precision of 7% in summer and 10% in winter. Fifty homes were monitored each season, including 26 that were monitored in both seasons (74 separate homes). Valid data were obtained for 659 of 700 possible days (94% completeness) amounting to 990 000 min of FP measurements and 386 000 min of UFP measurements. An example of concurrent indoor-outdoor UFP concentrations is provided in Figure 1. Examples of daily UFP variation, indoor-outdoor FP 12932

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

about 1−3 h for the period of elevated exposures following an indoor source. A main indoor source was cooking, reported on 591 of the 694 person-days of the study. Frequencies of using the stovetop, oven, toaster oven, and toaster are provided in the SI (Table S2). On a per-home basis, 58 of the 74 homes showed at least one measurable FP decay rate and 60 had at least one UFP decay rate (SI Table S3, Figures S4−S7). Thirteen homes had from 7 to 32 FP peaks and 20 homes had from 7 to 25 UFP peaks (i.e., from 1 to about 4 peaks per day from indoor sources). Sometimes a single source produced concurrent UFP and FP peaks. In such cases the decay rates for the two size categories can be directly compared. There were 129 such peaks, and the comparison showed that the median UFP deposition* rates were 12−16% greater than the FP rates (SI, Table S4) Predictive Models. The models for FP and UFP using the log-transformed data are provided in Table 3, and those using the raw data are shown in SI Table S5. For the log-transformed data, in summer, both window opening and use of the kitchen exhaust fan were strongly associated (p < 0.005) with larger FP decay rates. In winter, the stratification of homes by construction year appeared as a significant effect for FP, with homes constructed after 1960 (strata 3-5) showing much lower total decay rates, probably due to better construction resulting in the observed lower air exchange rates. The R2 effect may be due to the fact that longer duration decays can achieve higher R2 values due to the larger number of time pointsbut also longer durations tend to have lower decay rates. For both FP and UFP in the summer, the indoor-outdoor temperature difference was negatively associated with the decay rate, contrary to theory, possibly due to persons being more likely to open windows when the temperatures are moderate. In the winter, this relationship was reversed for the UFP (p < 0.001). For UFP in winter, use of an air cleaner and also the bathroom exhaust fan resulted in strong increases in the decay rate. The effect of the peak concentration in increasing total decay rate could be due to the increased coagulation rate with higher concentrations. If so, this would be one of the first indications of a possible effect of coagulation observed in field studies to date. Also the effect of the furnace fan being on continuously and thus forcing significantly larger volumes of the air in the house through the duct work attained significance at the p < 0.01 level, an effect that has not to our knowledge been found before in field studies. For the models of the raw data, the strongest predictor in the FP total decay rate model for summer (SI Table S5) was the use of the kitchen exhaust fan (p < 0.0001), which lowered decay rates by 3.5 times the mean value of 1.66 h−1. This effect was only partially due to the increase in average daily air exchange rate produced by the fan; some of the effect would have been due to removal of many particles before they had a chance to be transported to the rest of the house. The window opening index appeared as an important predictor variable in both seasons. For UFP in summer, air conditioning was associated with a very large reduction of nearly 0.9 h−1 in the total decay rate, possibly due to the lower air exchange rate due to closed windows. For the homes with central air conditioning (about half of the total homes reporting use of air conditioning), there would be additional loss due to deposition in ducts, but this was apparently outweighed by the reduced air exchange rate. For UFP in winter, use of an air cleaner was associated with an even larger increase in the decay rate of

concentrations, and concurrent outdoor FP concentrations for six homes are found in the SI (Figures S1−S3).

Figure 1. Log-transformed UFP concentrations in a home (blue) compared to outdoor concentrations at the central NAPS site (red).

After correction for the indoor background concentration, the slightly curved decays in Figure 1 should appear straight on a semilog plot if the decay is exponential. The result of correcting the background for both UFP and FP using our censoring algorithm shows the nearly straight-line declines indicating a clean exponential decay across multiple hours with correct choice of background (Figure 2).

Figure 2. Natural logarithms of background-corrected indoor UFP (cm−3) and FP (μg/m3) concentrations, with best-fit regressions. Both regressions had R2 values >99%. Outdoor concentrations also shown.

A total of 410 P-Trak peaks and 290 DustTrak peaks had measurable decay rates, with R2 values >90% for all but 8 of the 700 slopes. Median total decay rates were 1.08 h−1 for FP and 1.26 h−1 for UFP (Table 2). After subtracting the measured air exchange rates, median particle loss rates due to deposition and filtration were 0.70 h−1 for FP and 0.92 h−1 for UFP. (The UFP decays may also have included losses due to coagulation for the higher peaks.) Deposition, filtration, and coagulation thus accounted for most of the total decay rate for both FP (72%) and UFP (77%). The interquartile range of peak durations was 12933

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

Table 2. Basic Statistics for Decay, Depositiona, and Air Exchange Rates for 700 FP and UFP Peaks N

mean

SD

min

10%

25%

median

75%

90%

max

Fineparticles (DustTrak) all data total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes) summer total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes) winter total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes) all data total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes) summer total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes) winter total decay rate (h−1) depositiona (h−1) air exchange rate (h−1) duration (minutes)

290 288 288 290

1.38 0.99 0.39 134

1.10 0.97 0.32 103

0.19 −0.30 0.05 28

0.41 0.20 0.14 53.5

0.62 0.37 0.17 69

1.08 0.70 0.26 100

1.75 1.30 0.53 165

2.59 2.18 0.83 278.5

7.43 7.33 2.60 624

109 107 107 109

1.66 1.21 0.47 114

1.38 1.26 0.41 90

0.19 −0.30 0.08 28

0.41 0.19 0.12 42

0.74 0.41 0.18 62

1.22 0.87 0.34 90

2.06 1.54 0.60 130

3.88 3.15 0.99 206

7.43 7.33 2.60 510

181 181 181 181

1.21 0.86 0.34 146

0.84 0.73 0.24 109

0.28 −0.01 0.05 28 Ultrafine Particles

0.40 0.22 0.14 58 (P-Trak)

0.57 0.34 0.17 77

1.00 0.60 0.23 105

1.56 1.19 0.48 180

2.36 1.86 0.72 308

4.86 4.43 1.00 624

414 410 410 414

1.51 1.17 0.35 163

1.04 0.94 0.30 124

0.33 −0.63 0.04 29

0.53 0.35 0.11 55

0.82 0.56 0.17 80

1.26 0.92 0.23 126

1.83 1.50 0.48 194

2.85 2.27 0.70 315

9.07 8.58 2.35 734

205 201 201 205

1.59 1.21 0.38 157

1.20 1.11 0.35 105

0.34 −0.63 0.05 29

0.52 0.34 0.11 55

0.80 0.55 0.16 81

1.28 0.90 0.29 125

1.90 1.46 0.51 193

3.04 2.33 0.73 307

9.07 8.58 2.35 552

209 209 209 209

1.44 1.13 0.31 169

0.85 0.75 0.23 139

0.33 0.02 0.04 32

0.54 0.37 0.12 55

0.86 0.56 0.17 80

1.25 0.97 0.21 126

1.81 1.50 0.40 194

2.74 2.27 0.70 338

4.43 3.53 1.00 734

Deposition on surfaces + deposition in ductwork + filtration + (P-Trak only) coagulation + difference between instantaneous and 24 h average air exchange rate. a

nearly 1.6 h−1. The indoor-outdoor temperature difference was positively associated with decay rates, and was also very strong (p < 0.001). Carpeted rooms were associated with an increase in the decay rate for ultrafine particles, but were only marginally significant (p < 0.06). Larger homes had a strong (p < 0.0001) negative effect on total decay rate for both FP and UFP. Seasonal models were also created for the air exchange rate a, with the strongest effects shown by the window-opening index (SI Table S6). Home volume and construction year, together with air conditioning, were also significant variables. Results of the FP and UFP models for deposition* (d−a) are provided in SI Tables S7 and S8. Comparisons with Other Studies of Deposition Rates in Residences. Early studies including estimation of deposition rates for individual homes and specific particle size categories were those of Abt et al., 2000 (3 homes for one week),29 Long et al., 2001 (10 homes for one week),21 and Wallace et al., 2002 (1 home for 4.5 years).18 All three studies employed a scanning mobility particle sizer (SMPS) capable of providing deposition information for about 100 sizes of UFP ranging from 10 to 1000 nm, and an aerodynamic particle sizer (APS) capable of providing deposition information for about 50 sizes of FP ranging from 0.5 to 20 μm. All three studies also employed semicontinuous measurement of air exchange rates using repeated injection of a tracer gas. These studies tended to

support the general understanding from theoretical and chamber studies that deposition follows a “V”-shaped curve with the highest values at the smallest and largest particle sizes, but with rates 1−2 orders of magnitude higher than those predicted for spherical particles on smooth surfaces.13 Given these specific values, it is also possible to convolute the observed particle size distributions with the deposition rates to calculate a total deposition rate for the aerosol mixture. Different lower size cutoffs could be employed to predict what might be observed for, say, a P-Trak (lower cutoff 20 nm) and DustTrak (upper cutoff 2.5 μm). Large multihome studies are forced to use instrumentation (e.g., P-Traks and DustTraks) that is less capable of differentiating among particle sizes, and also less capable of providing continuous air exchange rates (e.g., the PFT method providing only 24 h or longer average vaplues). This leads to some limitations that are discussed below. Among these larger studies was another Canadian study in which total decay and deposition rates were estimated for UFP in a study of multiple homes in Windsor, Ontario.11 Two summer seasons with 38 homes in each season produced a median deposition rate estimate of 0.76 and 0.79 h−1, compared to 0.90 h−1 in this study. The winter season in Windsor produced a median of 0.61 h−1 compared to 0.97 h−1 in this study. Both studies found 12934

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

Table 3. Decay Rate Model for FP and UFP (Log-Transformed Data)a season

effect

summer

intercept mean window index by peak mean kitchen fan by peak mean abs temp diff by peak intercept R2 mean window index by peak strata 5. homes constructed after 2004 4. homes constructed between 1981 and 2004 3. homes constructed between 1961 and 1980 2. homes constructed between 1946 and 1960 1. homes constructed prior to 1946

winter

summer

winter

intercept peak concentration mean window index by peak mean abs temp diff by peak intercept peak concentration mean bathroom fan by peak mean abs temp diff by peak mean air cleaner use by peak home volume mean furnace fan on continuously by peak

coeff

stderr Fine Particles 0.387 0.164 0.015 0.005 2.210 0.739 −0.049 0.016 3.733 1.689 −3.692 1.722 0.047 0.014 −0.377 0.232 −0.437 0.243 −0.561 0.211 0.058 0.229 0 Ultrafine Particles 0.378 0.127 0.013 0.006 0.015 0.004 −0.043 0.011 0.380 0.228 0.009 0.004 1.334 0.636 0.020 0.006 0.691 0.200 −0.002 0.000 0.279 0.101

ecoeff

DF

t

p

1.47 1.02 9.12 0.95 41.80 0.02 1.05

89.1 74.1 109.8 108.5 166.5 164.8 159.5

2.36 3.09 2.99 −3.09 2.21 −2.14 3.42

0.69 0.65 0.57 1.06 1.00

38.4 41.8 33.3 38.7

−1.62 −1.80 −2.66 0.26

0.0206 0.0028 0.0034 0.0026 0.0284 0.0335 0.0008 0.0263 0.1126 0.0796 0.0121 0.8001

1.46 1.01 1.02 0.96 1.46 1.01 3.80 1.02 2.00 1.00 1.32

139.3 201.5 202.7 199.5 61.7 191.9 202.4 127.0 20.6 35.7 34.0

2.97 2.25 3.58 −3.84 1.67 2.28 2.1 3.46 3.46 −5.73 2.77

0.0035 0.0258 0.0004 0.0002 0.1008 0.0237 0.0372 0.0007 0.0024 0.0000 0.0090

a

The model is fully described in SI (eq 2). coeff = Coefficient of the associated independent variable. stderr = standard error of the coefficient. ecoeff = factor by which the variable increases or decreases the total decay rate. DF = degrees of freedom. t = Student’s t value. p = probability result is due to chance.

kitchen. In that study, when sources were operating, the fraction of particles from 10 to 20 nm was 26%. Thus the PTrak would have missed at least 26% of the particles observed in that study (more than 26% if there were particles below 10 nm in size produced by the sources). A further limitation of the present study is that air exchange rates were only available as 24 h averages. This caused unavoidable uncertainty in the ability to separate the total decay rate into its constituent parts including the instantaneous air exchange rate and the remaining loss rate due to deposition, filtration, and coagulation. An attempt to estimate the size of this effect is provided in the SI. Briefly, we find that because the deposition rate is normally several times higher than the air exchange rate, errors in the latter will be reduced in the former. The very low median air exchange rates in Edmonton in both winter (0.21 h−1) and summer (0.28 h−1) are a further limitation on the size of the error expected due to the diurnal variation of air exchange rates. Since under normal closedwindow configurations in winter, the main effect on air exchange rates is expected to be the indoor-outdoor absolute temperature difference, we expect air exchange rates were higher at night than in the daytime when most peaks occurred, and therefore our estimation of the deposition* rate (d−a) would have been biased low. In summer, on the other hand, open windows in the daytime would have caused instantaneous air exchange rates during most of the peaks to be higher than the daily average used in our calculations, and therefore our estimates of deposition* would have been biased high. Estimates of the resulting error are provided in the SI (Figure S8 and

that air exchange accounted for only about 25% of the total decay rates. In a study of seven northern California homes, out of 38 peaks providing estimated decay rates in the homes, 29 were associated with cooking.30 Individual deposition rates were not reported, since the investigators did not measure air exchange rates during the monitoring period . (They did measure air exchange rates at other times when the house was unoccupied, finding a geometric mean rate of 0.59 h−1.) They did report geometric mean decay rates for 36 of the 38 peaks (omitting two cases where only a single value was available) ranging from 1.1 to 1.8 h−1, in good agreement with our interquartile range estimates of 0.8−1.8 h−1. (The interquartile range is chosen here since we are comparing 420 peaks to 38). If their geometric mean air exchange rate estimate of 0.59 h−1 is subtracted from their decay rates, the range for deposition rates becomes about 0.5 to 1.2 h−1, compared to our IQR of 0.5 to 1.5 h−1. A recent study31 investigated UFP concentrations in 18 homes and found a median deposition rate of 1.01 h−1, very close to our findings of 0.90 and 0.97 h−1 in summer and winter, respectively. Limitations. A main limitation of the study is that the lower limit of 20 nm for the P-Trak probably caused an underestimation of the peak UFP values due to indoor sources producing freshly generated particles 63 000 cm−3, generally much higher than the median outdoor levels of 7500 cm−3 (summer) and 9700 cm−3 (winter). Also, infiltration factors for UFP have been found to be well below those for FP, on the order of 0.25 rather than 0.5, and therefore even high outdoor concentrations are strongly attenuated indoors.11,35 Finally, the high UFP concentrations near roadways are dominated by the smallest UFP. Zhu31 states that 50% of the total particle number within 50 m of the expressway were in the 6−25 nm size range, most of which is not even accessible to the P-Trak used in this study. Therefore we consider this limitation relatively minor. Finally, all data were collected from a single city, which may limit the wider applicability of this model. Future studies could be improved by extending the lower limit of the UFP monitors to at least 10 nm and perhaps even lower to about 4 nm to catch the 5−6 nm peak of stovetop cooking. Better time resolution of both air exchange measurements and personal activities would also be desirable. A main contribution of this study has been the development of a method to estimate a background concentration of indoor particles. Calculation of decay rates cannot be done without knowledge of the background concentrations to which indoor concentrations are trending. For the rare cases when background concentrations are constant at the same concentration both before and after the full peak, the measured background levels can be used. However, this almost never happens in a real situation, since the background is always changing, cannot be measured during the peak, and the peak itself often is not allowed to complete its decay since new peaks intervene. The instantaneous background concentration is the product of the outdoor concentration and the infiltration factor, which is often not known or calculated in most studies. In this study, however, the daily average infiltration factor for each home was calculated using a censoring algorithm for the UFP and FP11 and therefore the instantaneous background concentrations could be estimated for each minute for each home. This study provides measured total decay rates and total deposition rates for a substantial number of homes (58−60) and monitoring days (>300). Progress was made in identifying variables affecting the different components of decay rates. For example, carpeted rooms were identified as a possible effect regarding deposition on surfaces (ksurface), furnace fan on continuously might be related to deposition on ductwork and

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*(L.W.) Phone: 707-843-5258; e-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The study depended on the cooperation of 74 homeowners in making their home available and keeping a record of their activities for 7 or sometimes 14 days. We acknowledge Keith Van Ryswyk, Hongyu You, Ryan Kulka, and Tim Shin for fieldwork and database management. We thank an anonymous reviewer for an extraordinarily careful and thorough review that has substantially improved the manuscript.

■

REFERENCES

(1) Dockery, D. W.; Pope, C. A.; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G.; Speizer, F. E. An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med. 1993, 329, 1753−1759. (2) Stolzel, M.; Breitner, S.; Cyrys, J.; Pitz, M.; Wolke, G.; Kreyling, W.; et al. Daily mortality and particulate matter in different size classes in Erfurt, Germany. J. Exposure Sci. Environ. Epidemiol. 2007, 17 (5), 458−467. (3) Weichenthal, S.; Dufresne, A.; Infante-Rivard, C. Indoor ultrafine particles and childhood asthma: Exploring a potential public health concern. Indoor Air 2007, 17, 81−91. (4) HEI. Understanding the Health Effects of Ambient Ultrafine Particles.Report by the HEI Review Panel on Ultrafine Particles; Health Effects Institute. Boston, MA, 2013; (5) Oberdorster, G.; Oberdorster, E.; Oberdorster, J. Nanotoxicology: An emerging discipline evolving from studies of ultrafine particles. Environ. Health Perspect. 2005, 113, 823−839. (6) Wallace, L. A.; Emmerich, S. J.; Howard-Reed, C. Source strengths of ultrafine and fine particles due to cooking with a gas stove. Environ. Sci. Technol. 2004, 38 (8), 2304−2311. (7) Wallace, L. A.; Wang, F.; Howard-Reed, C.; Persily, A. Contribution of gas and electric stoves to residential ultrafine particle concentrations between 2 and 64 nm: Size distributions and emission and coagulation rates. Environ. Sci. Technol. 2008, 42, 8641−8647. (8) Allen, R.; Larson, T.; Sheppard, L.; Wallace, L.; Liu, L-J S. Use of real-time light scattering data to estimate the contribution of infiltrated and indoor-generated particles to indoor air. Environ. Sci. Technol. 2003, 37, 3484−3492. (9) Howard-Reed, C. H.; Wallace, L. A.; Ott, W. R. The effect of opening windows on air change rates in two homes. J. Air Waste Manage. Assoc. 2002, 52, 147−159. 12936

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937

Environmental Science & Technology

Article

(10) Howard-Reed, C.; Wallace, L. A.; Emmerich, S. J. Effect of ventilation systems and air filters on decay rates of particles produced by indoor sources in an occupied townhouse. Atmos. Environ. 2003, 37 (38), 5295−5306. (11) Kearney, J.; Wallace, L.; MacNeill, M.; Xu, X.; VanRyswyk, K.; You, H.; Kulka, R.; Wheeler, A. J. Residential indoor and outdoor ultrafine particles in Windsor, ON. Atmos. Environ. 2011, 45, 7583− 7593, DOI: 10.1016/j.atmosenv.2010.11.002. (12) Wallace, L. A.; Williams, R.; Rea, A.; Croghan, C. Continuous weeklong measurements of personal exposures and indoor concentrations of fine particles for 37 health-impaired North Carolina residents for up to four seasons. Atmos. Environ. 2006, 40, 399−414. (13) Lai, A. C. K.; Nazaroff, W. W. Modeling indoor particle deposition from turbulent flow onto smooth surfaces. J. Aerosol Sci. 2000, 31, 463−476. (14) Lai, A. C. K.; Nazaroff, W. W. Supermicron particle deposition from turbulent chamber flow onto smooth and rough vertical surfaces. Atmos. Environ. 2005, 39 (27), 4893−4900. (15) 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 (4), 405−413. (16) Wallace, L. A.; Howard-Reed, C. H.; Emmerich, S. J. Continuous measurements of air change rates in an occupied house for one year: The effect of temperature, wind, fans, and windows. J. Expos. Anal. Environ. Epidemiol. 2002, 12, 296−306. (17) Rim, D.; Wallace, L.; Persily, A. Infiltration of outdoor ultrafine particles into a test house. Environ. Sci. Technol. 2010, 44, 5908−5913. (18) Wallace, L. A.; Howard-Reed, C. H. Continuous monitoring of ultrafine, fine, and coarse particles in a residence for 18 months in 1999−2000. J Air Waste Manage. Assoc. 2002, 52 (7), 828−844. (19) Thatcher, T. L.; Lai, A. C.; 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. (20) Ö zkaynak, H.; Xue, J.; Spengler, J. D.; Wallace, L. A.; Pellizzari, E. D.; Jenkins, P. Personal exposure to airborne particles and metals: Results from the Particle TEAM Study in Riverside, CA. J. Exposure Sci. Environ. Epidemiol. 1996, 6, 57−78. (21) Long, C. M.; Suh, H. H.; Catalano, P. J.; Koutrakis, P. Using time- and size-resolved particulate data to quantify indoor penetration and deposition behavior. Environ. Sci. Technol. 2001, 35, 2089−2099. (22) Bennett, D. H.; Koutrakis, P. Determining the infiltration of outdoor particles in the indoor environment using a dynamic model. J. Aerosol Sci. 2006, 37 (6), 766−785. (23) Rim, D.; Green, M.; Wallace, L.; Persily, A.; Choi, J.-I. Evolution of ultrafine particle size distributions following indoor episodic releases: Relative importance of coagulation, deposition and ventilation. Aerosol Sci. Technol. 2012, 46 (5), 494−503. (24) Hinds, W. C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles; Wiley: New York, 1999. (25) Seinfeld, J. H.; Pandis, S. N. Atmospheric Chemistry and Physics, 2nd ed.; . Wiley: Hoboken, NJ, 2006. (26) Rim, D.; Wallace, L. A.; Persily, A. K. Indoor ultrafine particles of outdoor origin: Importance of window opening area and fan operation condition. Environ. Sci. Technol. 2013, 47, 1922−1929. (27) Wallace, L. A.; Wheeler, A. J.; Kearney, J.; Van Ryswyk, K.; You, H.; Kulka, R.; Rasmussen, P.; Brook, J.; Xu, X. Validation of continuous particle monitors for personal, indoor, and outdoor exposures. J. Exposure Sci. Environ. Epidemiol. 2010, 21, 49−64. (28) Dietz, R. N.; Goodrich, R. W.; Cote, E. A.; Wieser, R. F. Detailed Description and Performance of Passive Perfluorocarbon Tracer System for Building Ventilation and Air Exchange Measurements, STM STP 904; American Society for Testing and Materials: Philadelphia, PA, 1986. (29) 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. (30) Bhangar, S.; Mullen, N. A.; Hering, S. V.; Kreisberg, N. M.; Nazaroff, W. W. Ultrafine particle concentrations and exposures in seven residences in northern California. Indoor Air 2011, 21, 132−144.

(31) Stephens, B.; Siegel, J. A. Penetration of ambient submicron particles into single-family residences and associations with building characteristics. Indoor Air 2012, 22, 501−513. (32) Zhu, Y.; Hinds, W.; et al. Concentration and size distribution of ultrafine particles near a major highway. J Air Waste Management Assoc. 2002, 52, 1032−1042. (33) Hoek.; et al. Indoor-outdoor relationships of particle number and mass in four European cities. Atmos. Environ. 2008, 42, 156−169. (34) Wheeler, A. J.; Wallace, L. A.; Kearney, J.; Van Ryswyk, K.; You, H.; Kulka, R.; Brook, J. R.; Xu, X. Personal, indoor, and outdoor concentrations of fine and ultrafine particles using continuous monitors in multiple residences. Aerosol Sci. Technol. 2011, 45, 1078−1089. (35) MacNeill, M.; Wallace, L.; Kearney, J.; Allen, R. W.; Van Ryswyk, K.; Judek, S.; Xu, X.; Wheeler, A. Factors Influencing Variability in the Infiltration of PM2.5 Mass and its Components. Atmos. Environ. 2012, 61, 518−532.

12937

dx.doi.org/10.1021/es402580t | Environ. Sci. Technol. 2013, 47, 12929−12937