3-EPA
United States
Environmental Protection
Agency
Estimating Contributions of
Outdoor Fine Particles to
Indoor Concentrations and
Personal Exposures:
Effects of Household Characteristics
and Personal Activities
-------
EPA 600/R-06/023
March 2006
Estimating Contributions of Outdoor Fine
Particles to Indoor Concentrations and
Personal Exposures:
Effects of Household Characteristics and
Personal Activities
By
Lance Wallace, Ron Williams, Jack Suggs and Paul Jones
Human Exposure and Atmospheric Sciences Division
National Exposure Research Laboratory
Research Triangle Park, NC, 27711
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC, 27711
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development partially
funded and collaborated in the research described here under contract numbers 68-D2-0134 (QST
Environmental), 68-D2-0187 (SRA Technologies, Inc), 68-D-99-012,68-D5-0040 (Research Triangle Institute),
CR-820076 (University of North Carolina-Chapel Hill), and CR-828186-01-0 (Shaw University). It has been
subjected to the Agency's peer and administrative review, and it has been approved for publication as an EPA
document. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
-------
Abstract
A longitudinal study of personal, indoor, and outdoor exposures to PM: 5 and associated elements was
earned out involving 37 residents of the Research Tnangle Park area in North Carolina. Participant exposures
were monitored for 7 consecutive days in each of four seasons. A main goal of the study was to estimate the
contribution of outdoor PM; 5 to indoor concentrations and personal exposures This contribution depends on
the infiltration factor (the fraction of outdoor PM2 5 remaining airborne after penetrating indoors), which can be
estimated using sulfur as a marker for particles of outdoor origin. The annual average infiltration factors ranged
from 0.26 to 0.89, and depended strongly on air exchange rates. The outdoor contnbutions to personal exposure
were then regressed longitudinally on outdoor concentrations measured at a central momtonng station, with a
range of R: values from 0.19 to 0.88. Vanables significantly affecting indoor air PM: 5 concentrations included
smoking and cooking, the number of persons in the household, burned food, use of a kitchen exhaust fan, and
duration of candle use. These findings might have important implications for epidemiological studies
111
-------
Contents
Notice
Abstract
Figures
s.
v
Tables vii
Acknowledgments ix
Chapter 1: Introduction 1
Chapter 2: Description of Study Methods and Database 3
Chapter 3: Results 4
Calculation of Fmf Using the Indoor/Outdoor Sulfur Ratio 4
Calculation of F^f by Regressing Indoor Sulfur on Outdoor Sulfur 8
Comparison of Methods for Calculating F^ 11
Estimating Indoor and Outdoor Contributions to Indoor PM25 11
Relationship Between Outdoor Particles and Indoor Particles of Outdoor Origin 15
Estimating Contributions of Outdoor Air to Indoor Concentrations Using the RCS Model 15
Estimates of the Contribution of Outdoor Air Particles to Personal Exposure 17
Outdoor Exposure Factor FpeX Estimated Using PM Measurements 19
Estimating the Outdoor Exposure Factor FpeX Using Sulfur Measurements 19
Comparison of Fpex and Finf 19
Comparison of Indoor-Outdoor Sulfur and Personal Sulfur Measurements 20
Use of the Outdoor Exposure Factor to Calculate the Contribution to Personal Exposure Made by
Particles of Outdoor Origin 25
Relationship Between Outdoor Concentrations and the Contribution to Personal Exposure of Particles of
Outdoor Origin 28
Use of Reported Time in Indoor and Outdoor Microenvironments to Predict the Outdoor Exposure
Factor Fpex from the Infiltration Factor Fmf 28
Estimating P andk 32
Calculating Average Values of P and k 32
Calculating Individual Home Values of P andk 33
Estimating F^ from Individual Values of P andk 36
Seasonal Analysis 39
Multivariate Regressions 39
Variables Affecting Air Exchange and the Infiltration Factor 49
Variables Affecting Air Exchange 49
Variables Affecting the Infiltration Factor 53
Chapter 4: Discussion 55
Chapter 5: Conclusions 58
Chapter 6: References 60
Appendix 64
IV
-------
Figures
Number Page
3-1 Valid pairs of indoor and outdoor 24-h average sulfur measurements (ng/m3) 4
3-2 Indoor/outdoor sulfur ratios (/",/) by home averaged across all seasons. Error bars are standard
errors calculated by propagation of error 8
3-3 Indoor/outdoor sulfur ratios by home and by season 8
3-4 Comparison of the results of regressing indoor sulfur on outdoor sulfur with the simple
indoor/outdoor ratio averaged over all visits to a home 11
3-5 Estimates of the fractional contribution of outdoor particles to total indoor PM: 5
concentrations, averaged over all home visits. Error bars are standard errors calculated by
propagation of error 11
3-6 Comparison of average outdoor-generated and indoor-generated particles based on
indoor/outdoor sulfur ratios 11
3-7 Estimates of average outdoor contribution to indoor PM; 5. Error bars are standard errors
calculated by propagation of error techniques applied to the three measurements required to
estimate the outdoor contribution 13
3-8 Estimates of average indoor-generated PM; 5. Error bars are standard errors calculated by
propagation of error techniques applied to the four measurements required to estimate
indoor-generated PM2 5 13
3-9 Indoor-outdoor average contributions to indoor PM25. Summer 2000 13
3-10 Indoor-outdoor average contributions to indoor PM; 5. Fall 2000 13
3-11 Indoor-outdoor average contnbutions to indoor PM;5. Winter 2001 15
3-12 Indoor-outdoor average contributions to indoor PM: 5. Spring 2001 15
3-13 Regression of indoor PM2.5 on residential outdoor PM; 5 data 17
3-14 Estimates of the infiltration factor F.^from sulfur indoor/outdoor ratios compared to
regressions of indoor vs. outdoor fine particles. The regression line shown is for the
18 cases with slopes significantly different from zero 17
3-15 24-h average fine particle personal exposures vs outdoor air concentrations 19
3-16 Personal vs. outdoor PM:5; one outlier removed 19
3-17 Personal vs. outdoor sulfur 19
3-18 Co-located PEMio and HI; 5 sulfur concentrations outdoors. Summer 2000 20
3-19 Co-located PEMK> and HI; 5 sulfur concentrations indoors. Summer 2000 20
3-20 Comparison of infiltration factor Fmf and outdoor exposure factor Fpa by participant 20
3-21 Outdoor contributions to personal exposure. Error bars are standard errors calculated by
propagation of error 25
3-22 Non-outdoor contnbutions to personal exposure. Error bars are standard errors
calculated by propagation of error 25
3-23 Outdoor and non-outdoor contributions to personal PM:5 exposure 25
3-24 Outdoor and non-outdoor contributions to personal PM; 5 exposure. The non-outdoor
contribution is divided into indoor-generated PM: 5 encountered while at home and the
sum of indoor-generated PM;5 while away from home and PM;5 due to the personal cloud 28
3-25 Predicted value of the outdoor exposure factor Fpex using Equation 3-6 compared to
the measured value using the personal/outdoor sulfur ratio 32
3-26 Predicted personal exposure to PM:5 using only F,nf. 32
3-27 Nonlinear least-squares fit to the indoor/outdoor sulfur ratio vs. the air exchange rate.
Bounding curves are + 1 SE 32
3-28 Regression of the outdoor/indoor sulfur ratio vs. residence time 33
3-29 Same regression as in Figure 3-28 without four outliers 33
3-30 Comparison of estimates of P from the linear and nonlinear approaches described in the text.
Only values significantly different from zero are plotted (N = 32 homes) 36
-------
3-31 Comparison of the estimates of k from the linear and nonlinear approaches described in the text.
Only values significantly different from zero are plotted (N = 24 homes) 36
3-32 Comparison of the infiltration factor (F,af) estimates from the simple ratio of indoor sulfur to
outdoor sulfur by home vs. the nonlinear regression of the same ratio using the
measured air exchange rates and the linear regression of the inverse ratio (outdoor/indoor)
against the residence time 36
3-33 Estimates for each home by season of the infiltration factor F^/from regressing indoor
sulfur on outdoor sulfur (Slope) compared to estimates from the simple ratio of indoor
sulfur to outdoor sulfur averaged overall visits in a season 39
3-34 Estimates of Fmfb\ home from the indoor/outdoor sulfur ratio 39
3-35 Estimates of Fm/by home from regressions of indoor on outdoor sulfur 39
3-36 Central-site and residential outdoor concentrations averaged over all visits to a home 40
A-1 Adjusted R2 values from regressing the outdoor contribution to personal exposure on outdoor
PM; 5 measurements just outside the house 64
A-2 Adjusted R2 values from regressing the outdoor contribution to personal exposure on outdoor
PM;.5 Harvard Impactor (HI) measurements at the central site 64
A-3 Adjusted R2 values from regressing the outdoor contribution to personal exposure on outdoor
PM:5 Federal Reference Method (FRM) measurements at the central site 65
VI
-------
Tables
Number Page
3-1 PM: 5 (ug/m3) and Sulfur Concentrations (ng/m3) Observed in Matched Indoor-Outdoor
Samples in the RTF Study 5
3-2 Indoor and Outdoor 24-hour Sulfur Measurements (ng/m3) Averaged Over All Visits
To Each Home 6
3-3 Ratios of the Mean Indoor Sulfur Concentrations to the Mean Outdoor Sulfur Concentrations 7
3-4 Air Exchange Rates (h"1) and Sulfur Indoor/Outdoor Ratios by Season 9
3-5 Results of Regressions of Indoor Sulfur on Outdoor Sulfur by Home, Compared to Average
Values of the Indoor/Outdoor Sulfur Ratio 10
3-6 Estimated Annual Average Contnbutions of Outdoor and Indoor-Generated Particles to
Total Indoor PM25 Concentrations by House (ug/m3) 12
3-7 Estimated Average Contributions of Outdoor Particles and Indoor-Generated Particles to
Total Indoor Concentrations (ug/m3) by House and by Season 14
3-8 Regressions of Indoor PM;5 of Outdoor Origin on Outdoor PM;5 Measured at the Home 16
3-9 Regressions of Indoor on Outdoor PM:5 by House 18
3-10 Sulfur Concentrations (ng/m3) and Ratios in Matched Personal, Indoor, and Outdoor Samples 21
3-11 Personal and Indoor Sulfur Concentrations (ng/m3) by Subject 22
3-12 Ratios of the Mean Personal Sulfur Exposure to the Mean Outdoor Sulfur Concentration
(Fp^) by Subject 23
3-13 Companson of Personal/Outdoor, Indoor/Outdoor, and Personal/Indoor Sulfur Ratios by
House and by Season 24
3-14 Estimated Contribution of Outdoor Particles to Personal Exposure (ug/m3). Standard
Deviations and Standard Errors Calculated by Propagation of Error 26
3-15 Contributions to Personal Exposure from PM25 Particles of Outdoor and Non-Outdoor Origin 27
3-16 Regression of Personal Exposure to Particles of Outdoor Origin on Outdoor Concentrations
Measured Near Residence by Harvard Impactor (HI) 29
3-17 Regression of Personal Exposure to Particles of Outdoor Ongin on Outdoor Concentrations
Measured at Central Site by Federal Reference Method (FRM) 30
3-18 Time (in Minutes) Spent in Various Activities/Locations 31
3-19 Estimates of P and k for Individual Homes Using Nonlinear Fit to the Indoor/Outdoor
Sulfur Ratio 34
3-20 Values for k when P is Bound from Above by 1 35
3-21 Results of Linear Regressions of the Outdoor/Indoor Sulfur Ratio on Residence Time for
36 Homes 37
3-22 Values for k When P is Bound from Above by 1 38
3-23 Multiple Regression of Outdoor Concentrations on Household Characteristics and Personal
Activities 41
3-24 Dependence of Indoor Fine Particle Concentrations on Household Characteristics and Personal
Activities 43
3-25 Dependence of Indoor-Generated and Outdoor-Generated Particles on Household Charactenstics
and Personal Activities 44
3-26 Dependence of Outdoor Sulfur on Outdoor PM: 5 and of Indoor Sulfur on Outdoor
Sulfur and Household Characteristics and Personal Activities 46
3-27 Regressions of Personal Exposures to PM:5 on Household Charactenstics and Personal Activities .. 47
vn
-------
3-28 Regression of the Non-ambient-related Contribution (Perscontrib) to Personal PM2.s Exposure 50
3-29 Regression of the Ambient-Related Contribution to Personal PM25 Exposure 51
3-30 Regressions of Personal Exposure to Sulfur on Indoor and Outdoor Concentrations and
Questionnaire Variables 52
3-31 Variables Affecting Air Exchange Rate 54
3-32 Variables Affecting Indoor/Outdoor Sulfur Ratio 54
3-33 Variables Affecting Indoor/Outdoor Sulfur Ratio: Reduced Model 54
A-l Values of the Average Sulfur Indoor/Outdoor Ratio (F,n/) and the Air Exchange Rates by
House and by Season 66
A-2 Comparison of Seasonal Average Sulfur Indoor/Outdoor Ratios (Su/S^,) with Slopes of
Regressions of Sin on Sout 69
A-3 Comparison of Seasonal Average Sulfur Personal/Outdoor Ratios (Spera/Sout) with
Slopes of Regressions of Sperson Sout 72
A-4 Questionnaire Variables and Definitions 75
Vlll
-------
Acknowledgments
The authors thank Linda Sheldon for many stimulating discussions regarding this report. Robert Kellogg
of Man Tech Environmental performed the X-ray fluorescence analyses that resulted in the excellent sulfur
data. Charles Rodes led the team at Research Triangle Institute International in collecting the field data. The
Environmental Measurements and Analysis Branch of the National Exposure Research Laboratory was
responsible for designing and overseeing the study. In particular, we acknowledge the contributions of Carry
Croghan, Anne Rea, Alan Vette, Carvin Stevens, Ten Conner, and Kelly Leovic. We wish to especially thank
the participants who earned the burden of responsibility for up to 28 days over a year's time.
IX
-------
Chapter 1
Introduction
Many studies worldwide in the last decade have documented an
association between health effects and particle concentrations
measured at central monitoring sites (Schwartz et al., 1996).
Since the health effects are presumably related to personal
exposures, an important research need, identified by the National
Academy of Sciences in 1995, is to determine how personal
exposures correlate with these outdoor concentrations (NRC-
NAS 1998). A number of studies (Abt 2000a,b: Allen et al.,
2003; Clayton et al., 1993; Ebelt et al., 2000; Evans et al., 2000:
Hopke et al., 2003; Howard-Reed et al., 2000; Janssen et al.,
1997, 1998, 1999, 2000; Keeler et al., 2002; Landis et al., 2001;
Liu et al., 2003; Long et al., 2000, 2001, Ozkaynak et al,
1996a,b; Pelhzzan et al., 1992; Rea et al., 2001; Rojas-Bracho et
al., 2000; Samat et al., 2000,2001; Thomas et al., 1993; US EPA
2002,2003; Vette et al., 2001; Wallace et al., 2003a; Williams et
al., 2000a,b, 2003a,b) have measured personal exposure directly
using personal monitors, and the correlations of personal
exposure with outdoor concentrations are straightforward to
determine (Wallace, 2000). However, the correlation that
interests epidemiologists is not that between total personal
exposure and outdoor concentrations, but the correlation between
that component of personal exposure due to outdoor particles and
the outdoor concentrations. This requires the ability to estimate
the contribution to personal exposures from particles originating
outdoors. Only a few studies have reported making this estimate
(Ebelt et al., 2000; Allen et al., 2003, 2004). The goal of this
report is to estimate the contribution of outdoor particles to
personal exposure for a group of 37 persons monitored one week
per season over four seasons in 2000-2001 (US EPA 2002,2003;
Williams et al., 2003a,b). The correlation between this portion
of personal exposure and PM2 5 outdoor concentrations will then
be calculated for each person. Many of the main findings of this
report appear in Wallace and Williams (2005).
Since people spend on average 89% of their time indoors, the
contribution of outdoor particles to indoor concentrations will
also be explored. For many people, the indoor-outdoor
relationship may be the major determinant of the personal-
outdoor relationship.
The contribution of outdoor particles to indoor concentrations is
described by the mass balance equation. The full mass balance
equation includes such phenomena as coagulation, condensation,
and gas-to-particle conversion (Nazaroff and Cass, 1989). We
will consider here a simplified version involving only infiltration,
exfiltration, deposition, and indoor sources. The differential
form of this simplified mass balance equation is
dCm dt = PaCout - (a+k) Cm-SV
(1-1)
where Cm = indoor number or mass concentration (cm"3 or
ug/m3),
Cou, = outdoor number or mass concentration
P = penetration coefficient across building envelope
a = air exchange rate (h"1)
(" = volume of building (cm3 or m3)
5 = source strength (h"1 or ug h"1)
k = deposition rate of particles (h"1)
The equation is assumed to be applicable to all particle sizes,
with all terms except air exchange rate and building volume
considered to be functions of particle size. We assumed that the
entire house is a single well-mixed zone, with instantaneous
mixing of particles throughout the house, and that the measured
air exchange rate in one room applies to the entire house. This
assumption is probably violated by most homes, which are likely
to have different zones on each floor or even on the same floor.
We also assume that the deposition rate is constant over the
period of integration. This assumption will not hold if persons
open windows, turn on fans, run the furnace or the air
conditioner, or otherwise make changes in the household
operating characteristics that will affect particle deposition
during the integration period. Finally, we assume that the
averaging time over which the equation is to be evaluated is
sufficiently long that transient terms due to short-term changes in
the outdoor concentration are negligible compared to the long-
term average concentrations. Since we are dealing with 24-hour
averages, this assumption is probably a good one. Under these
assumptions, the solution to the mass balance equation is
Cm = [Pa'(a-k)] * Cou, - S[V(a-k)]
(1-2)
The coefficient of the outdoor concentration is sometimes called
the infiltration factor Fm/.
The infiltration factor has a major effect on the indoor-outdoor
and the personal-outdoor relationships. This is expected to van-
by household and resident characteristics. For example, a tightly
built house may have a lower penetration coefficient than a
drafty house, although almost no data are available to support
this claim. A house with a large surface area (e.g., many carpets,
rugs, or fibrous wall hangings) may have higher deposition rates
(L ai and Nazaroff 2000). Use of fans or filters may also increase
particle deposition rates (Howard-Reed et al., 2003; Riley et al.,
-------
2002; Thatcher etal., 2002; Thatcher and Layton, 1995; Wallace
et al., 2004a). Open windows will increase F^ by increasing the
air exchange rate (Howard-Reed et al., 2002; Wallace and
Howard-Reed, 2002) and possibly by redirecting infiltrating
particles through the open window (P = 1) rather than through
the rest of the building envelope (P < 1) (Liu and Nazaroff, 2001;
Mosley et al., 2001; Thornburg et al., 2001). Use of air
conditioning has been shown to lower the infiltration factor,
either because of reducing air exchange rates by shutting
windows or increasing deposition rates by recirculating indoor
air through ductwork (Howard-Reed et al., 2003; Sarnat et al.,
2000; Lai et al., 1999; Thornburg, 2004; Wallace et al., 2002).
Despite these clear indications that exposure to outdoor air
particles indoors depends heavily on household and behavioral
characteristics, studies capable of estimating the infiltration
factor reliably for individual homes are rather few. Even more
rare are studies capable of estimating the values of P and k for
individual homes (Allen et al., 2004). In this study we attempt to
estimate both F,nf and its parameters P and k for individual
homes, together with an estimate of the uncertainties involved.
Several investigators have noted that sulfur has few indoor
sources (Ebelt et al., 2000; Samat et al., 2002). If that is the
case, the source term in Equation 1 -2 above may be ignored, and
the equation takes the very simple form
Stn'Sout = Fmf (1-4)
where Sm and S^, are the sulfur concentrations indoors and
outdoors.
Indoor-outdoor comparisons of sulfur concentrations thus
provide a direct way to estimate Fmf for each individual home.
Strictly speaking, the sulfur data provides only information for
particles with similar behavior to that of sulfur with respect to
penetration, deposition, and reactivity. Sulfur particles are
smaller than most other fine particles. Since the particles in the
1-2.5 (am range may have higher deposition velocities than
sulfur, the estimates for the sulfur deposition rate ks would be
slight underestimates for the typical aerosol mixture in the PM2 5
category. To avoid extra notation, we shall use P and k
throughout rather than PS and ks but the reader should remember
that these are values appropriate only for particles in the size
range of sulfur particles. Later we provide evidence that this
'sulfur-related" infiltration factor is actually a very good estimate
of the PM2 5 infiltration factor.
-------
Chapter 2
Description of Study Methods and Database
A full description of the measurement methods used in this study
has been provided in the two preceding reports in this series (US
EPA 2002, 2003) and also in two published articles (Williams et
al, 2003a,b). The following abbreviated description includes
only the methods discussed in this report.
In this study, 37 persons in 36 homes in the Research Triangle
Park, NC area were monitored for up to four seasons, 7 days per
season. (Thirteen of these were monitored for only one, two, or
three seasons; see Appendix A.) Two gravimetric monitor types
were employed for PM? 5 measurements: the Harvard Impactor
(HI), operating at 20 Lpm, and a Personal Exposure Monitor
(PEM), operating at 2 Lpm for a nominal 24-h period. The PEM
was used for personal samples and the HI for indoor-outdoor
samples, the latter measured just outside the home. An HI, a
PEM and a Federal Reference Method (FRM) monitor were also
operated every day of the study at a central site. There were also
fixed PEM 10 monitors co-located with the HI2.5 monitorsfilters
from these instruments were analyzed for a suite of elements
using x-ray fluorescence. Since sulfur particles are expected to
be in the size range <0.5 ^m, the PEM]0 filters should have the
same amount of sulfur as the PEMi 5 filters.)
The HI mass measurements had a precision of about 5%; the
PEM measurements had a precision of about 8% (Williams et al.,
2003a). The precision of the sulfur measurements was calculated
to be about 8% (Kellogg, R, personal communication).
The data analysis concentrated on sulfur as a tracer of outdoor-
generated particles. Indoor-outdoor ratios were used to estimate
the infiltration factor for PMi.,-; personal-outdoor ratios were
used to estimate the portion of personal exposure due to outdoor-
generated particles. The estimates of exposure due to outdoor-
generated particles were then regressed on outdoor
measurements, both at the home and at a central site, to
determine the relationship between central site measurements
and exposure to particles of outdoor origin.
Participants filled out activity logs each day, identifying cooking,
cleaning, and other activities that might affect particle exposures.
A questionnaire on household characteristics was also completed
for each house. The full set of questionnaire variables and their
definitions and units is provided in Appendix A (Table A-4).
These variables were then included in multivariate regressions to
determine their influence on air exchange rates, the infiltration
factor, and the observed PMis and sulfur concentrations. A
detailed description of this analysis is provided in the section on
multivariate regressions.
-------
Chapter 3
Results
A total of 876 person-days had at least one personal, indoor, or
outdoor PM: 5 measurement. About 868 days had at least one
filter analyzed by XRF. Samples were flagged if they failed any
of a number of quality control criteria. For example, flow rates
were required to be within 10% of the target values, filters were
discarded if seen to be torn or pierced. All outdoor PM2s filters
collected between April 11 and April 17 were discarded due to
excessive contamination with pollen. An additional requirement
for the sulfur measurements was that indoor values not be more
than 1.08 times outdoor values (the 1.08 factor was chosen
because the estimated uncertainty of the sulfur measurements
was 8%). Even if an indoor/outdoor ratio exceeding 1.08 were
correct, it would indicate an indoor source of sulfur, and
therefore should not be used in analyses to determine the
infiltration factor. Six samples were flagged for exceeding a
1.08 indoor/outdoor ratio. Another 21 samples were flagged
when regressions suggested an indoor source of sulfur, and it
was discovered that the participant was using a humidifier during
three seasons (7 days per season). Since different combinations
of variables will have different numbers of flags, the number of
valid data points sometimes varies in the following tables.
Calculation of fW Using the Indoor/Outdoor
Sulfur Ratio
Table 3-1 lists the distributional characteristics of all paired
indoor-outdoor samples with validated PM25 and sulfur
measurements. Also included in Table 3-1 are estimates of the
indoor/outdoor sulfur ratio. The average indoor/outdoor sulfur
ratio was 0.59 (0.16 SD).
Assuming all the sulfur is in the form of ammonium sulfate, we
can calculate the mass by multiplying by the ratio of the
molecular weights (3.5). The result is an estimate of 8.0 ug/m3
of ammonium sulfate, or about 41% of the total PM: 5. Several
Eastern cities included in EPA's speciation network ranged from
26-31% in their sulfate/PM:.3 values (AQCD 2003).
All the validated indoor and outdoor sulfur concentrations (N:
775) are displayed in Figure 3-1.
4500
_4000
H
°>3500
| 3000
-------
Table 3-1. PM2.5 (ug/m3) and Sulfur Concentrations (ng/m3) Observed in Matched Indoor-Outdoor Samples in the RTP Study
PM2.5ln
PM2.5 Out
Sulfur In
Sulfur Out
S In/Out
N
774
774
775
775
775
Mean
19.4
19.5
1116
1964
0.59
SD
16
9
653
1123
0.16
Min
2
5
123
404
0.17
10th
7
9
426
765
0.39
25th
10
13
615
1061
0.48
Median
15
19
974
1759
0.59
75th
22
24
1441
2667
0.69
90th
36
32
2034
3610
0.79
Max
119
52
3852
5406
1.06
-------
Table 3-2. Indoor and Outdoor 24-hour Sulfur Measurements (ng/m ) Averaged Over All Visits to Each Home
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/Mean
N
27
26
29
23
27
28
28
6
24
27
8
28
7
27
27
28
26
6
27
27
27
13
13
27
8
24
28
25
34
14
28
23
12
6
20
17
775
S,n
955
1069
1516
882
1199
1293
1476
1270
1087
1049
1661
944
553
701
928
1040
923
1328
1568
1139
1240
1231
1175
909
1827
1829
1335
1059
991
1678
607
522
1746
658
1079
534
1139
SD
372
477
534
359
723
651
901
408
688
558
825
449
224
342
449
554
467
436
769
641
704
607
601
498
679
1009
704
502
446
722
287
322
808
300
327
137
541
SE
72
94
99
75
139
123
170
167
141
107
292
85
85
66
86
105
92
178
148
123
135
168
167
96
240
206
133
100
76
193
54
67
233
123
73
33
126
Sou,
1532
1877
2078
2106
1892
2116
1947
3664
1968
1945
2446
1945
1534
1579
1322
1639
1889
2625
2198
2272
2152
2338
2558
1795
2712
2620
2113
1938
2001
1920
1730
1493
2811
2525
1836
974
2058
SD
612
1030
756
1056
1034
1008
1261
1376
1310
1202
1266
1149
661
774
588
952
1140
996
1218
1229
1322
1175
1508
1153
1292
1457
1276
1147
1269
839
769
978
1139
1130
698
299
1058
SE
118
202
140
220
199
191
238
562
267
231
447
217
250
149
113
180
224
407
234
237
254
326
418
222
457
297
241
229
218
224
145
204
329
461
156
72
252
-------
Table 3-3. Ratios of the Mean Indoor Sulfur Concentrations to the Mean Outdoor Sulfur Concentrations
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Mean
SJS00,
0.62
0.57
0.73
0.42
0.63
0.61
0.76
0.35
0.55
0.54
0.68
0.49
0.36
0.44
0.70
0.63
0.49
0.51
0.71
0.50
0.58
0.53
0.46
0.51
0.67
0.70
0.63
0.55
0.50
0.87
0.35
0.35
0.62
0.26
0.59
0.55
0.56a
SDca,,"
0.35
0.40
0.37
0.27
0.52
0.42
0.67
0.17
0.51
0.44
0.49
0.37
0.21
0.31
0.46
0.50
0.38
0.25
0.53
0.39
0.48
0.37
0.36
0.43
0.41
0.55
0.51
0.41
0.38
0.54
0.23
0.31
0.38
0.17
0.29
0.22
0.39
SEcalc
0.07
0.08
0.07
0.06
0.10
0.08
0.13
0.07
0.10
0.08
0.17
0.07
0.08
0.06
0.09
0.09
0.08
0.10
0.10
0.08
0.09
0.10
0.10
0.08
0.14
0.11
0.10
0.08
0.07
0.14
0.04
0.07
0.11
0.07
0.06
0.05
0.09
3 This value obtained by dividing the mean indoor sulfur concentration for all homes by the mean outdoor sulfur concentration. The arithmetic mean
of the ratios was 0.59.
b Values of the standard deviation and standard error are calculated by propagation of error.
-------
where SD<.atc - the calculated standard deviation of the ratio for
agiven house,
100 ng/m3 and thus may be due to scatter rather than to
indoor sources of sulfur or sulfates. The scatter is due both to
variations in FlnJ with changing seasons and air exchange rates
8
-------
Table 3-4. Air Exchange Rates (h"1) and Sulfur Indoor/Outdoor Ratios by Season
Season
Summer
Fall
Winter
Spring
N
223
187
179
171
airex
0.49
0.61
1.01
0.68
SD
0.57
0.40
0.73
0.49
SJSout
0.50
0.63
0.63
0.62
SDa
0.16
0.14
0.13
0.16
3 Observed standard deviation; SD calculated by propagation of error would be larger.
-------
Table 3-5. Results of Regressions of Indoor Sulfur on Outdoor Sulfur by Home, Compared to Average Values of the Indoor/Outdoor Sulfur Ratio
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
N
27
26
29
23
27
28
28
6
24
27
8
28
7
27
27
28
26
6
27
27
27
13
13
27
8
24
28
25
34
14
28
23
12
6
20
17
Oin/Oout
0.63
0.60
0.74
0.46
0.65
0.61
0.77
0.36
0.58
0.57
0.68
0.54
0.37
0.45
0.69
0.65
0.52
0.52
0.75
0.49
0.61
0.55
0.49
0.55
0.72
0.70
0.66
0.59
0.57
0.89
0.36
0.39
0.62
0.26
0.60
0.56
SE
0.02
0.02
0.02
0.02
0.02
0.02
0.02
0.05
0.02
0.02
0.04
0.02
0.04
0.02
0.02
0.02
0.02
0.05
0.02
0.02
0.02
0.03
0.03
0.02
0.04
0.02
0.02
0.02
0.02
0.03
0.02
0.02
0.03
0.05
0.02
0.03
Slope
0.57
0.45
0.68
0.28
0.60
0.63
0.69
0.27
0.49
0.42
0.63
0.36
0.32
0.41
0.72
0.57
0.36
0.42
0.60
0.50
0.47
0.44
0.39
0.40
0.38
0.67
0.53
0.38
0.30
0.84
0.30
0.27
0.64
0.26
0.41
0.40
SE
0.04
0.02
0.04
0.04
0.07
0.03
0.04
0.06
0.04
0.04
0.07
0.03
0.05
0.03
0.05
0.03
0.04
0.06
0.04
0.03
0.05
0.08
0.03
0.03
0.15
0.04
0.03
0.04
0.03
0.06
0.04
0.04
0.09
0.03
0.05
0.06
Inter.
86
231
103
287
70
-46
132
291
115
227
116
238
60
55
-23
111
238
226
239
-8
232
211
182
190
793
83
215
315
398
74
90
119
-65
10
326
140
SE
72
53
81
97
156
60
83
249
90
88
180
63
78
59
74
50
85
173
91
66
127
217
79
69
438
121
74
99
79
124
83
73
283
94
104
57
P
0.24
0.00
0.21
0.01
0.66
0.45
0.12
0.31
0.21
0.02
0.54
0.00
0.48
0.36
0.76
0.04
0.01
0.26
0.01
0.90
0.08
0.35
0.04
0.01
0.12
0.50
0.01
0.00
0.00
0.56
0.28
0.12
0.82
0.92
0.01
0.03
R2
0.86
0.93
0.93
0.67
0.72
0.96
0.93
0.77
0.88
0.82
0.93
0.86
0.88
0.85
0.88
0.94
0.78
0.90
0.91
0.94
0.77
0.69
0.95
0.85
0.45
0.92
0.92
0.76
0.70
0.94
0.63
0.66
0.81
0.92
0.75
0.76
10
-------
and measurement error. These have the well-known effect of
lower slopes and higher intercepts than the true values.
Therefore the calculated slopes are likely to be underestimates of
the true infiltration factor. Only 4 of the 36 homes had slopes
higher than the mean indoor/outdoor sulfur ratio. The overall
average indoor/outdoor sulfur ratio is 0.56, compared to the
overall average slope of 0.49.
Comparison of Methods for Calculating Finf
The methods for calculating F,,,/ (regression vs. the simple
indoor/outdoor ratio) are compared in Figure 3-4. The
comparison differentiates between the 22 homes with intercepts
not different from zero and the 14 homes with intercepts
significantly different from zero. The latter set of homes cannot
be said with complete confidence to have no indoor sources of
sulfur, although random measurement errors could also cause
nonzero intercepts. The slopes for these homes are generally
lower, as would be expected if measurement errors are affecting
the regressions. Home number 25 is an outlier in this graph
(large red diamond), having the highest intercept of all homes,
but with high uncertainty, due to a small number of samples (N =
8). Without home 25, the 21 remaining homes with intercepts
not different from zero have slopes that are related to the mean
indoor/outdoor ratios with a hish R~ of 0.91.
0.9
0.8
0.7
£ 0.6
I 0.5
5 0.3
1
0-0-2
y = 0.94x - 0.03
R2 = 0.79 N = 22
y= 1.10X-0.22
R2 = 0.76 N = 14
0.4 06
Sulfur Indoor-Outdoor Ratio
0.8
Figure 3-4. Comparison of the results of regressing indoor sulfur on
outdoor sulfur with the simple indoor/outdoor ratio averaged over all
visits to a home.
their indoor PM: 5 concentrations supplied by outdoor air, and 7
homes had less than 40% supplied by outdoor air.
1 1
o
£ 0.8
Q.
o
_c
O 0.4
c
o
« 0.2
It
House ID
Figure 3-5. Estimates of the fractional contribution of outdoor
particles to total indoor PM2 5 concentrations, averaged over all home
visits. Error bars are standard errors calculated by propagation of
error.
The outdoor and indoor-generated contributions to total indoor
PM: 5, averaged overall visits to each house, are shown in Figure
3-6. The relative importance of these two sources varies widely,
with some homes having essentially no indoor contribution and
others having more than 50% indoor contributions.
60
n Indoor-generated
Outdoor contribution
E 40
Estimating Indoor and Outdoor
Contributions to Indoor PM2 5
The indoor/outdoor sulfur ratio may be multiplied by the outdoor
air concentration to estimate the contribution of outdoor air
particles to the total indoor particle level for each home (Table 3-
6). The indoor-generated particles are then the difference
between the total indoor PM: 5 and the outdoor contribution to
indoor PM: 5. The fractional contributions of outdoor air particles
to indoor concentrations, averaged over all visits to each home,
are shown in Figure 3-5. Fourteen homes had more than 80% of
Figure 3-6. Comparison of average outdoor-generated and indoor-
generated particles based on indoor/outdoor sulfur ratios.
The 95% confidence limits are displayed separately for outdoor
and indoor contributions in Figures 3-7 and 3-8. (These values
are taken from Table 3-6.) The indoor contributions cover a
wider range than the outdoor contributions. For example, three
homes had indoor sources producing an average concentration
over the four seasons > 25 ug/nr, compared to no homes with
outdoor contributions that high. On the other hand. 20 homes
had indoor contributions of 5 u.g/mj or lower compared to only
11
-------
Table 3-6. Estimated Annual Average Contributions of Outdoor and Indoor-Generated Particles to Total Indoor PM2 5 Concentrations by House (Mg/m3)
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/ Mean
N Outdoor contribution SD^
to indoor particles
27
26
29
23
27
28
28
6
24
27
8
28
7
27
27
28
26
6
27
27
27
13
13
27
8
24
28
25
34
14
28
23
12
6
20
17
775
11.0
13.8
19.4
9.4
12.2
11.7
18.0
9.6
11.3
10.9
15.5
9.1
4.8
7.4
10.8
10.7
8.7
10.7
17.2
10.8
13.0
10.6
9.9
9.4
21.4
17.6
12.7
10.4
9.7
21.7
5.8
6.4
15.4
6.2
10.4
7.4
11.7
12.9
18.8
16.6
16.8
17.4
15.0
22.6
15.8
20.0
18.9
17.8
16.1
9.6
13.3
11.4
15.1
15.4
11.9
20.4
19.8
20.6
15.6
19.0
16.6
19.1
22.3
18.1
15.8
16.8
20.7
12.3
18.0
16.7
17.4
10.4
8.3
16.5
SEcaic Indoor-generated SDcaic SE^t
particles
2.5
3.7
3.1
3.5
3.4
2.8
4.3
6.5
4.1
3.6
6.3
3.1
3.6
2.6
2.2
2.9
3.0
4.8
3.9
3.8
4.0
4.3
5.3
3.2
6.7
4.5
3.4
3.2
2.9
5.5
2.3
3.7
4.8
7.1
2.3
2.0
3.9
5.0
2.3
1.6
31.0
5.9
5.8
1.6
8.1
4.8
0.2
11.8
3.7
3.8
4.1
7.2
-0.5
4.1
1.4
3.3
-0.3
25.4
14.0
15.1
27.6
32.6
2.2
22.7
1.0
16.3
-0.2
7.3
5.1
1.3
3.0
1.3
0.3
7.8
15.9
21.3
20.0
37.0
20.9
16.5
25.5
16.5
22.0
19.3
22.2
16.9
10.4
14.8
20.3
15.6
18.6
12.2
21.5
20.2
29.6
23.5
33.7
29.1
29.0
24.2
32.1
16.6
22.7
23.2
13.6
20.0
17.3
17.5
10.9
8.8
20.5
3.1
4.2
3.7
7.7
4.0
3.1
4.8
6.7
4.5
3.7
7.8
3.2
3.9
2.8
3.9
3.0
3.6
5.0
4.1
3.9
5.7
6.5
9.4
5.6
10.3
4.9
6.1
3.3
3.9
6.2
2.6
4.2
5.0
7.2
2.4
2.1
4.8
12
-------
one home with an outdoor contribution that low. This reflects
much more variability in indoor particle-generating activities
compared with outdoor particle concentration variability.
Figure 3-7. Estimates of average outdoor contribution to indoor PM25.
Error bars are standard errors calculated by propagation of error
techniques applied to the three measurements required to estimate the
outdoor contribution.
35
30
I"
5 20
CL
1 15
i 10
- 0
-5
-10
House ID
Figure 3-8. Estimates of average indoor-generated PM35. Error bars
are standard errors calculated by propagation of error techniques
applied to the four measurements required to estimate indoor-
generated PM25.
We saw in Figure 3-3 that the indoor/outdoor sulfur ratio varied
by season, and was much lower in summer than the other three
seasons. We next look at the estimates of indoor and outdoor
contributions by season (Table 3-7; Figures 3-9 through 3-12).
Table 3-7 shows that indoor-generated particle concentrations
were 45-46% of the total indoor particle concentrations averaged
across all homes in the summer and fall seasons, but only 31% in
the winter and spring seasons. This was due mostly to a
reduction in indoor sources in the latter two seasons, since the
outdoor contribution did not change greatly over the four
seasons. Because of the small number of measurements in each
house and each season (N = 1-7), the uncertainty, particularly in
the estimates of indoor-generated concentrations, is much larger
than for the equivalent concentrations for each home over all
seasons (as in Table 3-6). The average standard error was 15%
for the estimates of the outdoor contributions, but 66% for the
indoor estimates.
House ID
Figure 3-9. Indoor-outdoor average contributions to indoor PM2 5.
Summer 2000.
70
60
_ so
"E
a Indoor-generated
Outdoor contribution
II I ilil ll
Will
House ID
Figure 3-10. Indoor-outdoor average contributions to indoor PM2 5.
Fall 2000.
13
-------
Table 3-7. Estimated Average Contributions of Outdoor Particles and Indoor-Generated Particles to Total Indoor Concentrations (M9/rn3) by House
and by Season
Summer 2000
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/Mean
N
8
6
7
2
6
7
7
6
6
6
6
7
7
7
6
6
7
6
6
7
7
6
7
7
2
6
6
5
13
7
7
5
7
5
216
Out
14.0
14.2
16.2
7.5
12.0
12.9
17.3
9.6
12.4
10.7
17.0
10.0
4.8
5.0
8.7
13.0
9.4
10.7
16.1
13.9
12.1
9.2
11.4
8.6
14.4
18.6
14.2
10.5
9.8
13.9
3.3
2.6
13.1
6.9
11.3
In
1.6
0.5
-0.1
10.9
16.5
8.5
5.6
8.1
3.5
1.8
15.4
5.7
3.8
4.9
1.5
-0.4
5.6
1.4
8.4
1.0
25.8
15.3
5.1
31.4
51.5
4.1
49.9
0.0
23.0
0.8
6.1
2.7
1.6
2.7
9.5
Fall 2000
N
6
4
6
7
6
5
7
5
7
7
6
7
7
5
7
6
6
7
6
7
6
6
7
7
6
7
7
5
3
5
3
186
Out
8.1
14.9
24.3
13.5
8.3
18.7
26.0
6.2
7.5
6.7
6.1
9.8
7.6
8.9
12.3
5.0
17.7
11.7
8.1
10.7
23.8
23.3
9.8
11.7
7.2
29.5
6.9
4.0
21.1
8.7
4.9
12.4
In
8.1
-1.8
11.8
44.8
8.7
3.5
3.1
11.8
1.1
8.3
2.7
18.5
1.3
0.6
5.3
2.6
28.1
13.0
26.9
23.1
26.2
4.2
27.0
1.8
11.3
-1.2
10.7
4.3
-0.7
3.7
1.9
10.0
Winter 2001
N
6
6
7
7
7
7
6
5
7
7
7
7
7
6
7
7
7
6
5
7
6
7
7
6
7
7
171
Out
10.8
14.3
23.6
8.0
16.6
9.3
15.0
15.6
17.3
10.4
10.4
13.1
11.4
9.7
25.2
16.0
13.9
11.9
21.5
15.2
11.9
12.9
6.7
10.1
12.2
8.0
13.5
In
12.5
-1.9
-6.8
43.9
-1.2
8.2
-0.1
8.8
-1.4
1.5
7.7
1.4
-1.7
1.2
-2.4
-4.9
18.6
27.9
-2.4
12.0
3.3
18.1
6.4
3.0
0.7
0.5
5.9
Spring 2001
N
2
7
7
4
7
5
5
5
2
7
7
3
7
1
7
7
6
7
5
7
6
5
7
4
7
5
142
Out
8.7
12.0
12.9
6.9
7.8
14.7
12.0
7.5
11.2
9.4
7.7
12.6
10.9
2.8
15.2
7.5
7.7
6.4
9.2
11.0
8.3
8.9
6.4
9.1
10.5
7.4
9.4
In
0.7
11.3
2.0
11.7
3.2
-1.8
-1.3
0.2
0.9
-0.7
0.9
3.4
-1.0
2.5
2.8
0.5
19.5
28.1
4.1
3.9
-0.9
9.7
6.1
4.2
0.3
-0.1
4.2
14
-------
30
20
D Indoor-generated
Outdoor contribution
12345678 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2
House ID
Figure 3-11. Indoor-outdoor average contributions to indoor PM2 5.
Winter 2001.
40
35
30 j
, 25
20
15
10
5
0
D Indoor-generated
Outdoor contribution
Figure 3-12.
Spring 2001.
Indoor-outdoor average contributions to indoor PM:;
Relationship Between Outdoor Particles
and Indoor Particles of Outdoor Origin
As described in the Introduction, a quantity of interest to
epidemiologists is the relationship between outdoor particles and
indoor particles of outdoor origin. Since epidemiologists work
with measurements made at a central site, we examined how this
relationship changes as we go from the particles measured just
outside the home to those measured at the central site, and as we
go from particles measured by the same type of equipment to
those measured by different monitors. We first performed
regressions for each home of the estimated contributions to
indoor PM:5 made by particles of outdoor origin vs. the
measured concentrations just outside the home (Table 3-8).
These individual regressions by home have high R: values in
most cases (median = 0.77; range 0.39-0.95). For all 752 valid
daily measurements, the regression has a slope of 0.56 (0.01 SE)
and a non-significant intercept of 0.31 (0.31) (Table 3-8). The
Pearson correlation coefficient was 0.82, with an R" (adjusted)
value of 0.665. (A Spearman rank correlation was also 0.82,
indicating that the Pearson coefficient was not affected by
outliers.) When the regression was run against the HI at the
central site, the R" value was reduced to 0.58 (N = 736). Against
the FRM at the central site, it was reduced to 0.49 (N = 775).
We caution that these R~ values must be considered upper
bounds, because the assumptions of our model, which force the
infiltrated particles of outdoor origin to be proportional to
outdoor concentrations, virtually ensure that R" values will be
high.
Estimating Contributions of Outdoor Air to
Indoor Concentrations Using the RCS Model
Another way of estimating an average infiltration factor, this
time using the PM: 5 mass data, is the Random Component
Superposition (RCS) model developed by Ott and co-workers
(Ott et al., 2000). The RCS model simply assumes that a
regression of all the measured indoor vs. outdoor mass
concentrations will provide an overall average infiltration factor
together with an estimate of the indoor source strength averaged
across all homes (that being the constant term in the regression).
The overall average infiltration factor estimated by this approach
is 0.60 (0.06 SE, N = 774) (Figure 3-13), compared with the
average indoor/outdoor sulfur ratio of 0.589 (0.006 SE, N = 775)
from Table 3-1. The close agreement of these two values for
infiltration factor, one based on particle concentrations alone and
the other based on sulfur concentrations alone, is important
evidence that using the sulfur results to estimate fine particle
parameters is justified, at least for these overall averages. It
should be noted that the RCS model determines only one average
value of Fml for all homes, whereas the methods discussed above
provide estimates for individual homes. The RCS approach as
applied to PM,,, values in Riverside, CA, Philipsburg, NJ, and
Toronto, Canada found similar values of 0.55, 0.60 and 0.61 for
the infiltration factor (Ott et al., 2000) compared to our values
using PM: j sulfur and mass data. Further evidence for the
applicability of the sulfur results to the PM: 5 fraction is the close
agreement between the RCS estimate of 7.7 ug/nv for the
indoor-generated PM: ? compared to 7.8 ng/irr estimated using
the sulfur indoor-outdoor ratios (Table 3-6).
15
-------
Table 3-8. Regressions of Indoor PM2 s of Outdoor Origin on Outdoor PM2 5 Measured at the Home
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/Mean
N
27
26
24
23
27
28
28
6
24
22
8
28
7
27
27
28
26
6
22
22
27
13
13
27
8
21
28
25
34
14
28
23
12
6
20
17
752
Slope
0.53
0.45
0.62
0.35
0.67
0.63
0.81
0.22
0.52
0.37
0.67
0.32
0.31
0.41
0.66
0.60
0.41
0.40
0.57
0.52
0.58
0.38
0.38
0.47
0.78
0.69
0.52
0.37
0.27
0.74
0.31
0.38
0.59
0.26
0.43
0.52
0.56
SE
0.03
0.05
0.04
0.05
0.09
0.03
0.05
0.08
0.06
0.04
0.10
0.04
0.04
0.04
0.07
0.03
0.06
0.10
0.07
0.04
0.07
0.11
0.04
0.06
0.33
0.05
0.04
0.06
0.05
0.05
0.05
0.06
0.15
0.05
0.06
0.05
0.01
Intercept
1.47
3.06
2.34
1.71
-0.56
-0.35
-0.73
3.55
1.05
2.81
0.26
3.26
0.67
0.47
0.55
0.74
1.69
2.32
3.13
-0.30
0.47
2.95
1.88
1.03
-1.59
0.26
2.49
3.58
4.62
2.86
0.72
-0.07
0.53
-0.14
2.86
0.45
0.31
SE
0.65
1.16
0.99
1.18
1.81
0.68
1.25
2.14
1.31
0.86
2.46
0.71
0.64
0.68
1.15
0.52
1.18
2.15
1.40
0.75
1.58
2.20
0.90
1.21
9.98
1.36
0.85
1.24
1.10
1.40
0.92
1.13
3.80
1.23
1.16
0.67
0.31
P
0.03
0.01
0.03
0.16
0.76
0.61
0.56
0.17
0.43
0.00
0.92
0.00
0.34
0.50
0.64
0.17
0.16
0.34
0.04
0.69
0.77
0.21
0.06
0.40
0.88
0.85
0.01
0.01
0.00
0.06
0.44
0.95
0.89
0.91
0.02
0.51
0.32
R2
0.91
0.79
0.93
0.69
0.69
0.93
0.91
0.61
0.76
0.77
0.85
0.75
0.88
0.82
0.77
0.94
0.62
0.75
0.78
0.90
0.74
0.51
0.89
0.68
0.39
0.90
0.86
0.59
0.43
0.95
0.56
0.66
0.58
0.85
0.71
0.89
0.67
16
-------
120
= 060 (006 SE)» *
R2 = 0.10: N = 774
?V&": Vil^r
.V .I----'--.
0 10 20 30 40 SO 60
Outdoor PM?3 (jig/m3)
Figure 3-13. Regression of indoor PM2.5 on residential outdoor
PM2.5data.
Many studios have PM; 5 measurements indoors and outdoors but
do not have sulfur or sulfate measurements to determine Fml.
The RCS model provides an estimate of the average infiltration
factor across all homes in a given study. The question arises: Is
it possible to apply the RCS model estimate of Fln/ for individual
homes from PM measurements alone?
The regression results show that only half (N = 18) of the homes
had slopes that were significantly different from zero (Table 3-
9). Also, three of the slopes were negative and two others were
greater than 1, both unphysical results. By contrast, all 36 sulfur
regressions had slopes that were significantly different from zero
Comparing the slopes of the PM: 5 regressions to the estimates of
Fm/ from the sulfur indoor/outdoor ratios resulted in a low R"
value of 0.11 (Figure 3-14).
1.5
S 0.5
a.
£
-0.5
y = 0.69x + 0.15
Significant Slopes
Nonsignificant slopes
0.2
0.4 0.6 0.8
l Based on Sulfur Indoor/Outdoor Ratios
Figure 3-14. Estimates of the infiltration factor F,nf from sulfur
indoor/outdoor ratios compared to regressions of indoor vs. outdoor fine
particles. The regression line shown is for the 18 cases with slopes
significantly different from zero.
Estimates of the Contribution of Outdoor
Air Particles to Personal Exposure
If total personal exposure E to fine particles is the sum of
exposures while indoors and exposures while outdoors, then the
following equation holds:
=./; c,n +./; c.HII
(3-2)
where /, and _/,' are the fractions of time spent indoors and
outdoors.
Exposure can also be expressed as the sum of outdoor and non-
outdoor sources:
E = E,, +£
(3-3)
where E,, is the exposure due to outdoor sources and Em, is the
exposure due to non-outdoor sources. The exposure due to
outdoor sources may be written as
EH jm Fn,l Cml, + / C,,iu
(3-4)
Now we assume that all of the time spent indoors was spent
indoors at home where a measurement of outdoor concentration
is available, and all of the time spent both outdoors and in transit
is considered to involve exposure to the outdoor concentration.
This outdoor concentration may be that measured at the home or
at the central site if that is available. In this case, the fractions
/ and /, add up to 1. We now define a factor /), such that
E = />v C,ml + £
(3-5)
The factor F/F,.V plays a role with respect to personal exposure that
F,nt played with respect to indoor concentrations. Some
investigators have labeled Fpa as an "attenuation factor,"
marking the decrease in the effect, or attenuation, of outdoor
concentrations on personal exposure. However, this would
imply that attenuation increases as F,,t,, increases, whereas the
reverse is true. We shall call Fpcx the "outdoor exposure factor."
indexing the contribution of outdoor particles to personal
exposure. From Equations 3-3 and 3-4, we have a relationship
for the outdoor exposure factor /),,.,:
* pex Jin*mj Jottl
Rewriting the equation as
ppa- = F,n/-+. /;,,,,
-------
Table 3-9. Regressions of Indoor on Outdoor PM2 5 by House
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
N
27
26
24
23
27
28
28
6
27
22
9
28
7
27
28
28
27
21
22
22
28
13
13
27
8
21
28
25
34
14
28
23
12
6
20
17
Slope
0.37
0.43
0.61
0.67
0.72
0.44
0.99
-0.22
0.35
0.31
-0.25
0.23
0.51
0.33
0.04
0.48
0.80
0.52
0.66
0.44
0.47
0.91
-0.55
0.42
1.66
0.75
0.24
0.36
1.03
0.69
0.22
0.34
0.48
0.11
0.29
0.44
SE
0.21
0.22
0.23
0.76
0.24
0.17
0.16
0.26
0.19
0.05
0.68
0.11
0.25
0.18
0.51
0.05
0.24
0.17
0.12
0.04
0.38
0.69
0.85
0.62
1.18
0.14
0.60
0.12
0.27
0.11
0.18
0.19
0.16
0.10
0.12
0.06
p-level
0.09
0.07
0.01
0.39
0.01
0.02
0.00
0.45
0.08
0.00
0.73
0.04
0.09
0.08
0.94
0.00
0.00
0.01
0.00
0.00
0.23
0.22
0.53
0.51
0.21
0.00
0.69
0.01
0.00
0.00
0.24
0.08
0.01
0.31
0.03
0.00
Intercept
9.4
6.0
6.4
26.0
4.5
9.0
-3.4
23.9
9.4
4.5
34.4
8.6
1.8
6.0
17.5
2.1
-1.1
5.9
6.4
2.3
28.2
6.6
36.6
29.6
4.6
1.8
30.6
4.8
6.8
4.0
9.6
5.8
4.7
6.4
6.6
1.8
SE
4.1
5.6
6.2
17.8
5.0
3.5
3.9
7.5
4.0
0.9
15.6
2.1
3.5
3.3
8.9
0.9
4.4
3.5
2.6
0.8
8.9
14.5
19.6
11.9
35.8
3.7
13.0
2.4
5.5
3.0
3.2
3.6
4.1
2.4
2.3
0.9
p-level
0.03
0.30
0.31
0.16
0.38
0.02
0.39
0.03
0.03
0.00
0.06
0.00
0.64
0.08
0.06
0.03
0.81
0.11
0.02
0.01
0.00
0.66
0.09
0.02
0.90
0.64
0.03
0.06
0.22
0.22
0.01
0.12
0.28
0.06
0.01
0.05
R2
0.08
0.10
0.21
0.00
0.23
0.18
0.59
0.00
0.08
0.67
0.00
0.12
0.35
0.08
0.00
0.76
0.29
0.30
0.58
0.85
0.02
0.06
0.00
0.00
0.12
0.58
0.00
0.24
0.29
0.76
0.02
0.09
0.43
0.07
0.19
0.78
18
-------
Several articles have discussed this idea that personal exposure
to particles of outdoor origin may be greater than indoor
concentrations of such particles (Wilson and Suh, 1997; Wilson,
Mage, and Grant 1998). The central idea is that persons spend
some time outdoors or in vehicles where they are exposed to the
full outdoor concentrations, thus adding to their exposures
indoors where they are "protected" somewhat by the infiltration
factor Finf.
As an example, suppose fin = 0.89, as found by the NHAPS
study, and/, = {fraction of time outdoors (6%) plus fraction of
time in vehicles (5%)} =0.11. Setting Finf= 0.60 as found by the
RCS model, we find that Fpex would be 0.64, an increase of 7%.
Outdoor Exposure Factor Fpex Estimated
Using PM Measurements
A regression of personal PM2.5 on outdoor PM2.5 concentrations
provides an estimate of Fpex averaged across all participants: 0.64
+ 0.01 SE (Figure 3-15). However, the fit is poor (R2 = 0.09).
The intercept in this case (11.6 + 1.4 ug/m3) is an estimate of £,
the average exposure due to non-outdoor sources. The
difference between the indoor source contributions (estimated at
8.0 ug/m3 in Table 3-1) and the non-outdoor source contribution
is sometimes attributed to the "personal cloud," which in this
case would equal 3.6 + 2.0 ug/m3.
250
200
3 150
:
I
| 100
B.
50
R* = 0.09: N = 776
.". :" '' t
, «. J _ .-\ *
10 20 30 40
Outdoor PMj , uig/m3)
50
60
Figure 3-15. 24-h average fine particle personal exposures vs. outdoor
air concentrations.
One outlier in Figure 3-15 (personal PM2.5 > 200 ug/m3) has a
strong influence on the slope of the line. If the outlier is
removed, the slope decreases from 0.64 to 0.59 (Figure 3-16),
while the intercept increases from 11.6 to 12.3. This would
increase the personal cloud from 3.6 to 4.3 ug/m3. The similarity
of the slopes for the personal vs. outdoor and indoor vs. outdoor
regressions (0.59-0.64 compared to 0.61) suggests that the time
spent indoors drives the relationship between personal exposure
and outdoor concentrations. That is, the infiltration factor F-mf,
which governs the reduction of particle concentrations as they
enter a house, is very similar to the outdoor exposure factor Fpac,
which governs the reduction in outdoor concentrations
contributing to personal exposure.
10 20 30 40 50 60
Figure 3-16. Personal vs. outdoor PM2.s; one outlier removed.
Estimating the Outdoor Exposure Factor
Fpex Using Sulfur Measurements
Another estimate of the outdoor exposure factor is given by the
regression of personal sulfur exposures vs. outdoor sulfur
concentrations (Figure 3-17). The slope is 0.49, identical to the
slope of the indoor/outdoor sulfur regression (Figure 3-1), while
the intercept of 90 ng/m3 is less than the indoor/outdoor intercept
of 1 50 ng/m3. The estimate using sulfur (R2=0.78) is preferred to
the estimate using PM2.5 (R2=0.09).
3500
3000
2500 -
=. 2000
ta
| 1500
1000
500
y = 0.49X + 89.5
Rz = 0.78; N = 750
.*"
1000 2000 3000 4000 5000
Outdoor Sulfur Concentration (ng/m3)
6000
Figure 3-17. Personal vs. outdoor sulfur.
Comparison of Fpex and Finf
Both the PM2.5 and sulfur regressions fail to support the
argument based on Equation 3-7 that F^ is larger than F-mf. One
possibility is that time spent in office buildings, which usually
recirculate a portion of the air through filters, will involve lower
exposures to sulfur than time spent in homes without such
recirculation and refiltration. Some evidence for this can be
found in the PTEAM study, where persons who worked away
19
-------
from home had lower particle exposures than those who stayed
home (Ozkaynak et al., 1996). In that case, the proper equation
for total exposure would have to take account of the different
concentrations in homes and offices:
£ Jinhome *~ inhume ~*~ Join *~ oul+J inoffice ^inoffice (j"°)
where the relationship Cinoffli.e < CMmmt would hold. This will
reduce the magnitude of the outdoor factor.
Comparison of Indoor-Outdoor Sulfur and
Personal Sulfur Measurements
Since the personal samples were collected using the 2 Lpm PEM,
but the indoor and outdoor samples were collected using the 20
Lpm HI, \ve need to address the question of how the two
samplers agree. During the summer season, 166 co-located
indoor and outdoor samples were collected. As showii in Figures
3-18 and 3-19, the agreement was excellent (R: = 0.97 for both).
with the slopes near 1 for both. This agreement between
different monitor types is justification for comparing the
indoor/outdoor sulfur ratio (the infiltration factor F,,,,) with the
personal/outdoor sulfur ratio (the outdoor exposure factor Fp^).
6000
5000
4000
£ 3000
i
2000
1000
y=1.03(0.03SE)x+41 (28)
R1 = 0.97. N = 166
2000 3000 4000
HI (ng/mj)
Figure 3-18. Co-located PEM2sand HI25 sulfur concentrations
outdoors. Summer 2000
4000 -
I
3500 -
i
3000
- 2500
y = 0.98 (0.03 SE) X +47 (33)
R2 = 0.97. N = 166
500 1000 1500 2000 2500
HI (ng/m1)
3000 3500 4000
Figure 3-19. Co-located PEM25and HI25 sulfur concentrations
indoors Summer 2000
The overall comparison shows that personal sulfur exposures
were somewhat smaller than indoor concentrations at home
(Table 3-10). When compared on a home-by-home basis, only
33% of the participants had sulfur exposures higher than indoor
concentrations (Table 3-11; Figure 3-20). The ratio of the mean
personal sulfur concentration averaged over all visits to a home
to the mean outdoor sulfur concentration (Fpfx) is 0.54 (Table 3-
12).
0.9 .
0.8
0.7 -
0.6 -
0.5
0.4.
0.3
0 2
a,
o .
it ; ) u 14 A 11 .^ i& s /; i u i. 17 /i f t » it w n ft ! it n » u " " ' i n w u '
Participant
Figure 3-20. Comparison of infiltration factor F,n, and outdoor exposure
factor Fpe, by participant.
Finally, only 32% of 117 cases compared by home and season
had higher personal sulfur exposures than indoor sulfur levels
(Table 3-13). Only in the summer season was the average
exposure higher than the average indoor concentration, but only
by 1%. 18 of 35 homes had higher exposures than indoor
concentrations in the summer, compared to only 8,4, and 8 cases
in the other seasons. In most cases, /}, was less than F,,,;,
contrary to the assumption that personal exposures to particles ot
outdoor origin would be larger than indoor concentrations due to
the time spent outdoors. This supports the idea that indoor
concentrations in buildings, where our participants spent some ot
20
-------
Table 3-10. Sulfur Concentrations (ng/m3) and Ratios in Matched Personal, Indoor, and Outdoor Samples
Concentrations/ratios
Personal
Indoor
Outdoor
Indoor/outdoor
Personal/outdoor
Personal/indoor
N
780
780
750
745
750
780
Mean
1062
1116
1944
0.59
0.55
0.97
SD
632
655
1140
0.17
0.14
0.23
Min
137
123
323
0.17
0.16
0.28
10th
393
423
755
0.38
0.38
0.75
25th
556
621
1037
0.48
0.46
0.83
Median
915
957
1671
0.58
0.54
0.92
75th
1419
1442
2592
0.69
0.63
1.03
90th
1975
2039
3647
0.80
0.73
1.21
Max
3254
3852
5406
1.08
1.08
2.64
21
-------
Table 3-11. Personal and Indoor Sulfur Concentrations (ng/m3) by Subject
Subject
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Mean
N
27
24
26
25
26
27
27
7
28
27
10
27
7
26
27
27
26
7
28
28
28
14
13
25
8
24
28
27
28
14
28
25
12
6
6
20
17
780
Personal
854
910
1338
934
1254
1082
1210
1256
1124
940
1671
892
643
707
841
951
878
1398
1464
1104
1369
1329
1209
1012
1686
1562
1101
896
996
1469
811
485
1692
1056
1143
940
511
1101
SE
71
86
108
105
141
117
155
165
141
105
264
85
112
69
80
108
93
189
139
128
149
189
213
119
246
175
106
88
109
161
65
54
224
170
249
82
41
132
Indoor
932
1058
1508
853
1196
1281
1488
1234
1130
1029
1542
959
553
720
950
1028
976
1421
1597
1120
1211
1285
1175
880
1827
1833
1335
1079
989
1678
607
524
1746
997
658
1079
534
1135
SE
74
86
108
72
145
127
176
145
128
106
251
87 '
85
66
87
108
103
177
146
120
134
165
167
98
240
206
133
97
88
193
54
65
233
154
123
73
33
126
Pers/ln
0.91
0.84
0.88
1.08
1.07
0.85
0.79
1.01
0.97
0.90
1.08
0.93
1.16
0.97
0.88
0.91
0.91
0.97
0.91
0.97
1.12
1.01
0.99
1.15
0.91
0.85
0.84
0.82
0.99
0.88
1.43
0.97
0.97
1.05
1.77
0.86
0.96
0.99
SE
0.01
0.02
0.02
0.07
0.04
0.02
0.02
0.04
0.04
0.02
0.06
0.02
0.10
0.02
0.02
0.01
0.04
0.04
0.02
0.02
0.04
0.04
0.04
0.06
0.03
0.02
0.02
0.02
0.06
0.02
0.08
0.04
0.02
0.03
0.17
0.04
0.04
0.04
22
-------
Table 3-12. Ratios of the Mean Personal Sulfur Exposure to the Mean Outdoor Sulfur Concentration (F^) by Subject
Subject
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Sum/Mean
N
25
23
21
23
26
27
27
6
25
21
8
27
7
26
27
27
25
6
22
22
28
13
14
23
14
21
28
25
27
14
28
22
13
27
6
21
20
765
Spere/Sout
0.57
0.49
0.64
0.46
0.67
0.51
0.62
0.35
0.57
0.46
0.73
0.45
0.42
0.44
0.63
0.59
0.46
0.50
0.65
0.49
0.66
0.54
0.47
0.55
0.59
0.60
0.52
0.45
0.53
0.77
0.47
0.33
0.63
0.60
0.45
0.52
0.53
0.54a
SDb
0.34
0.33
0.37
0.37
0.54
0.38
0.58
0.18
0.54
0.44
0.53
0.34
0.26
0.30
0.42
0.49
0.38
0.27
0.55
0.45
0.57
0.40
0.44
0.47
0.36
0.50
0.41
0.36
0.47
0.46
0.29
0.29
0.37
0.42
0.32
0.29
0.33
0.40
SEb
0.07
0.07
0.08
0.08
0.11
0.07
0.11
0.08
0.11
0.10
0.19
0.06
0.10
0.06
0.08
0.09
0.08
0.11
0.12
0.10
0.11
0.11
0.12
0.10
0.10
0.11
0.08
0.07
0.09
0.12
0.05
0.06
0.10
0.08
0.13
0.06
0.07
0.09
a Ratio of the mean personal sulfur concentration to the mean outdoor sulfur concentration. The mean of the ratios is also 0.54.
"Standard Deviations (SO) and Standard Errors (SE) Calculated by Error Propagation.
23
-------
Table 3-13. Comparison of Personal/Outdoor, Indoor/Outdoor, and Personal/Indoor Sulfur Ratios by House and by Season
Summer
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/Mean
N
6
6
5
2
6
6
6
6
6
5
6
7
7
7
6
6
6
6
6
7
7
6
7
7
7
6
7
5
14
7
7
4
7
6
215
P/Oa
0.50
0.46
0.72
0.41
0.57
0.43
0.54
0.35
0.47
0.45
0.74
0.39
0.42
0.39
0.60
0.57
0.45
0.50
0.59
0.54
0.55
0.46
0.47
0.54
0.51
0.52
0.45
0.39
0.42
0.91
0.49
0.25
0.52
0.45
0.50
I/O"
0.57
0.57
0.83
0.28
0.50
0.57
0.67
0.35
0.48
0.45
0.68
0.40
0.36
0.39
0.58
0.57
0.42
0.51
0.62
0.53
0.43
0.43
0.43
0.41
0.56
0.62
0.54
0.40
0.39
0.96
0.25
0.20
0.54
0.26
0.49
P/lc
0.88
0.81
0.86
1.45
1.13
0.75
0.81
1.02
0.97
1.00
1.09
0.98
1.16
1.02
1.04
0.99
1.08
0.99
0.96
1.01
1.27
1.08
1.10
1.32
0.92
0.84
0.83
0.97
1.07
0.95
1.95
1.27
0.97
1.74
1.01
N
6
6
7
7
6
7
7
7
7
7
6
7
7
6
7
6
7
7
7
6
7
6
7
7
6
7
7
6
6
7
6
205
Fall
P/O
0.58
0.48
0.64
0.46
0.66
0.57
0.70
0.58
0.44
0.50
0.44
0.63
0.65
0.38
0.71
0.44
0.74
0.63
0.48
0.52
0.71
0.69
0.61
0.51
0.62
0.67
0.51
0.39
0.75
0.52
0.54
0.58
I/O
0.63
0.51
0.70
0.47
0.66
0.63
0.84
0.69
0.57
0.56
0.46
0.74
0.72
0.48
0.79
0.43
0.70
0.64
0.58
0.51
0.80
0.78
0.77
0.66
0.59
0.82
0.40
0.38
0.84
0.65
0.54
0.64
P/l
0.93
0.93
0.91
0.98
1.01
0.89
0.82
0.84
0.77
0.89
0.96
0.85
0.91
0.80
0.90
1.01
1.05
0.98
0.83
1.02
0.89
0.88
0.80
0.77
1.06
0.82
1.29
1.01
0.89
0.80
1.00
0.91
N
7
6
7
7
7
7
7
7
7
2
7
6
7
7
7
7
7
7
6
7
7
7
7
7
5
7
7
179
Winter
P/O
0.60
0.48
0.59
0.48
0.66
0.50
0.58
0.57
0.50
0.60
0.56
0.45
0.64
0.55
0.47
0.71
0.41
0.71
0.60
0.56
0.58
0.49
0.56
0.56
0.47
0.54
0.56
0.55
I/O
0.67
0.63
0.67
0.48
0.72
0.59
0.76
0.60
0.59
0.67
0.62
0.47
0.77
0.67
0.56
0.82
0.47
0.68
0.57
0.68
0.66
0.61
0.68
0.48
0.50
0.67
0.63
0.62
P/l
0.89
0.76
0.88
1.00
0.92
0.85
0.77
0.94
0.86
0.90
0.91
0.95
0.82
0.82
0.85
0.86
0.86
1.05
1.06
0.81
0.88
0.80
0.82
1.18
0.95
0.81
0.90
0.89
N
6
5
2
7
7
7
7
5
2
6
7
7
7
6
2
2
7
4
2
7
6
7
7
7
7
7
146
Spring
P/O
0.63
0.59
0.59
0.48
0.80
0.53
0.71
0.70
0.47
0.43
0.46
0.66
0.57
0.57
0.66
0.52
0.75
0.64
0.74
0.54
0.47
0.62
0.37
0.29
0.51
0.51
0.56
I/O
0.64
0.66
0.74
0.39
0.74
0.61
0.85
0.63
0.50
0.46
0.46
0.78
0.62
0.58
0.68
0.51
0.67
0.62
0.78
0.67
0.60
0.69
0.30
0.37
0.53
0.50
0.59
P/l
0.99
0.90
0.79
1.25
1.09
0.87
0.84
1.12
0.93
0.93
0.99
0.85
0.92
0.98
0.97
1.02
1.13
1.04
0.95
0.81
0.78
0.90
1.22
0.80
0.96
1.03
0.95
a Personal/outdoor sulfur concentration ratio (outdoor exposure factor)
" Indoor/outdoor sulfur concentration ratio (infiltration factor)
c Personal/indoor sulfur concentration ratio
24
-------
their time, may be lower than in homes, due to recirculation and
filtration of outdoor air. The fact that the summer season gave
similar estimates for personal exposure and indoor
concentrations of sulfur is additional evidence for this
hypothesis, since air conditioning was in general use in summer
and would provide extra filtration for the indoor sulfur particles.
Use of the Outdoor Exposure Factor Fpex to
Calculate the Contribution to Personal
Exposure Made by Particles of Outdoor
Origin
We next used our estimated outdoor exposure factors Fnex for
each subject (based on the sulfur personal/outdoor ratio) to
calculate the contribution of particles of outdoor origin to
personal exposure (Table 3-14; Figures 3-21 through 3-23). On
average, particles of outdoor origin contributed 10.9 ug/nv
(47%) of the total personal exposure compared with 12.5 ug/rrr
(53%) for the particles of non-outdoor origin. The range of the
non-outdoor-generated particle contributions (5-33 ug/mj) was
much greater than the range of the outdoor-generated
contributions (6-19 ug/m').
The contribution of panicles of non-outdoor origin to personal
exposure can be subdivided into particles generated indoors
while the person is at home and particles from all other sources.
» ***************
Participant
Figure 3-21. Outdoor contributions to personal exposure. Error bars
are standard errors calculated by propagation of error.
Participant
Figure 3-22. Non-outdoor contributions to personal exposure. Error
bars are standard errors calculated by propagation of error.
70
01
d>
~ 40
B.
3 30
c
o
at
| 20
10
[3 Non-outdoor contribution
Outdoor contribution
Participant
Figure 3-23. Outdoor and non-outdoor contributions to personal PM2s
exposure.
The particles from other sources include particles generated
while the person is indoors at other locations or in a vehicle, as
well as particles from the "personal cloud" that may be generated
throughout the day at all locations (Yakovlcva ct al, 1999). The
indoor-generated particle concentrations estimated from the Flnl
values can be multiplied by the fraction of time spent indoors at
home (from the activity logs) to calculate the contribution of
indoor-generated particles to personal exposure. The outdoor
contribution is calculated by multiplying the outdoor
concentration times the fraction of time spent outdoors. The
contribution of indoor-generated particles at all other indoor
locations as well as the particles from the personal cloud is then
found by subtracting the sum of the particles of outdoor origin
and the indoor-generated particles while at home from the total
personal exposures (Table 3-15; Figure 3-24).
25
-------
Table 3-14. Estimated Contribution of Outdoor Particles to Personal Exposure (ng/m3). Standard Deviations and Standard Errors Calculated by
Propagation of Error
Subject
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Sum/Mean
^ Outdoor contribution to -p. _ Nonoutdoor contribution to _. _
personal exposure personal exposure
25
23
21
23
26
27
27
6
25
21
8
27
7
26
27
27
25
6
22
22
28
13
14
23
14
21
28
25
27
14
28
22
13
28
6
21
20
766
9.9
11.6
15.5
9.8
12.8
9.8
14.5
9.7
11.2
7.9
16.4
8.4
5.6
7.3
10.3
9.7
8.2
10.5
13.3
9.0
14.2
10.6
9.9
9.8
17.8
14.8
10.5
8.5
9.6
18.8
7.8
5.9
15.5
9.6
10.6
9.0
7.1
10.9
13.3 2.7
17.9 3.7
17.4 3.8
19.4 4.0
17.2 3.4
15.9 3.1
23.8 4.6
16.5 6.7
20.5 4.1
18.9 4.1
17.9 6.3
16.3 3.1
10.0 3.8
13.3 2.6
12.6 2.4
15.8 3.0
16.1 3.2
12.4 5.1
18.7 4.0
19.1 4.1
20.8 3.9
16.3 4.5
21.2 5.7
16.8 3.5
20.1 5.4
22.6 4.9
17.9 3.4
16.3 3.3
17.6 3.4
20.5 5.5
11.8 2.2
17.7 3.8
16.0 4.4
13.7 2.6
18.6 7.6
11.2 2.5
10.1 2.3
16.9 4.0
10.9
7.4
19.0
29.6
16.6
11.7
10.4
13.4
9.0
5.1
13.5
5.5
7.9
5.2
10.0
7.8
12.8
5.9
6.6
6.7
31.9
18.4
15.7
27.1
33.0
19.5
13.7
9.1
9.0
5.9
8.7
9.9
5.4
13.9
7.3
10.3
6.6
12.5
16.2 3.2
20.3 4.2
33.1 7.2
36.1 7.5
29.8 5.8
25.6 4.9
29.7 5.7
17.8 7.3
22.8 4.6
19.7 4.3
21.9 7.8
16.9 3.2
13.7 5.2
14.6 2.9
16.7 3.2
17.8 3.4
23.5 4.7
13.2 5.4
20.1 4.3
20.4 4.3
24.9 4.7
23.5 6.5
26.9 7.2
26.2 5.5
52.7 14.1
27.1 5.9
22.0 4.1
18.7 3.7
19.0 3.7
24.5 6.5
14.3 2.7
21.4 4.6
17.9 5.0
10.4 3.4
19.4 7.9
16.5 3.6
11.6 2.6
21.8 5.2
26
-------
Table 3-15. Contributions to Personal Exposure from PM2 5 Particles of Outdoor and Non-Outdoor Origin
Subject Fraction of time at
home
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Mean
SD
0.90
0.77
0.89
0.83
0.78
0.88
0.84
0.75
0.73
0.90
0.85
0.84
0.89
0.86
0.92
0.79
0.88
0.84
0.87
0.71
0.61
0.88
0.70
0.81
0.69
0.83
0.87
0.84
0.91
0.87
0.67
0.95
0.83
0.97
0.74
0.87
0.68
0.82
0.08
Outdoor contribution to
personal exposure3
9.9
11.6
15.5
10.0
12.8
9.8
14.5
9.7
11.1
7.9
16.4
8.4
5.6
7.3
9.8
9.7
8.2
10.5
13.3
9.0
14.6
10.6
10.0
9.8
19.3
14.6
10.5
8.5
9.6
18.8
7.8
5.9
14.9
9.5
10.6
9.1
7.1
10.9
3.3
Home contribution" Other0 PM2 5 foutd
Personal exposure
5.0
2.1
3.7
26.4
4.7
5.2
1.3
6.1
3.5
0.6
10.0
3.3
3.4
3.6
6.8
-0.4
3.7
1.2
4.3
0.8
15.5
12.4
10.7
21.5
22.6
2.2
19.7
0.9
16.1
-0.2
4.9
5.0
1.0
11.6
2.2
1.1
0.2
6.6
7.0
5.9
5.3
15.2
3.2
12.0
6.5
9.1
7.3
5.5
4.5
3.6
2.3
4.5
1.7
3.2
8.2
9.1
4.7
2.4
5.9
16.5
6.0
5.0
5.6
10.5
17.3
-6.0
8.3
-7.1
6.0
3.8
4.9
4.4
2.3
5.1
9.2
6.4
5.9
4.8
20.8
19.0
34.5
39.6
29.5
21.5
24.9
23.1
20.2
13.0
29.9
14.0
13.5
12.6
19.8
17.5
21.0
16.4
19.9
15.6
46.5
29.0
25.7
36.9
52.4
34.1
24.2
17.6
18.6
24.6
16.5
15.8
20.3
23.4
17.9
19.4
13.6
23.3
9.3
0.48
0.61
0.45
0.25
0.43
0.46
0.58
0.42
0.55
0.61
0.55
0.60
0.41
0.58
0.49
0.55
0.39
0.64
0.67
0.58
0.31
0.37
0.39
0.27
0.37
0.43
0.43
0.48
0.52
0.76
0.47
0.37
0.73
0.41
0.59
0.47
0.52
0.49
0.12
c
0.24
0.11
0.11
0.67
0.16
0.24
0.05
0.26
0.17
0.05
0.33
0.24
0.25
0.29
0.34
-0.02
0.18
0.07
0.22
0.05
0.33
0.43
0.42
0.58
0.43
0.06
0.81
0.05
0.87
-0.01
0.30
0.32
0.05
0.50
0.12
0.06
0.01
0.25
0.22
» '
lolher
0.28
0.28
0.44
0.08
0.41
0.30
0.37
0.32
0.27
0.35
0.12
0.16
0.33
0.13
0.16
0.47
0.43
0.29
0.12
0.38
0.35
0.21
0.19
0.15
0.20
0.51
-0.25
0.47
-0.38
0.24
0.23
0.31
0.22
0.10
0.28
0.47
0.47
0.26
0.18
' product of Fpex and the mean residential outdoor concentration
product of the indoor-generated concentration and the fraction of time at home
0 sum of particle concentrations indoors at other locations times the fraction of time spent there plus contributions from the personal cloud; obtained
by subtraction of outdoor and home contributions from total personal exposure
" Proportion of personal exposure contributed by outdoor-generated particles
° Proportion of personal exposure contributed by indoor-generated particles while at home
' Proportion of personal exposure contributed by indoor-generated particles while indoors other than at home plus personal activities throughout the
day
27
-------
so
so
40
30
I
? 20
10
0
-10
D Personal cloud plus particles generated indoors away from home
Particles generated indoors at home
3 Outdoor-generated contribution
Participant
Figure 3-24. Outdoor and non-outdoor contributions to personal PM25
exposure The non-outdoor contribution is divided into indoor-generated
PM; 5 encountered while at home and the sum of indoor-generated PM; 5
while away from home and PM25 due to the personal cloud.
From Table 3-15, the contribution of at-home indoor-generated
particles to personal exposure varied widely, from about 0 to 26
ug/nr (mean 6.6 ug/mj). The contribution of particles generated
while the person was away from home (either from exposures
indoors at other locations or from personal activities throughout
the day) ranged from -7 to 17 u.g/nr (mean 5.9 u,g/m'). Two of
the 37 participants had negative values for this variable due to
uncertainties in estimating the outdoor and home contributions.
On average, the home contribution plus the contribution of the
personal cloud and exposures in locations other than the home
added up to a bit more than the contribution from outdoor-
generated particles ( 12.5 ug/m3 vs. 10.9 ug/mj). On an individual
basis, however, some participants have virtually all of their
exposures contributed from outdoor particles, while others have
most of their exposure from indoor-generated or personal cloud
particles (Figure 3-24).
Relationship Between Outdoor
Concentrations and the Contribution to
Personal Exposure of Particles of Outdoor
Origin
As described in the Introduction, our ultimate aim is to determine
the relationship between personal exposure to particles of
outdoor origin and ambient measurements. We regressed our
estimate of the outdoor contribution to personal exposure on the
concentrations just outside each home (Table 3-16). Adjusted R~
values ranged from 0.42-0.93 (median of 0.81). The overall R:
was 0.71 (N = 750). When the regression was run against the
HI at the central site, the overall R~ was reduced to 0.64 (N =
735). The regressions of exposure to particles of outdoor origin
against central site FRM concentrations is the one that interests
epidemiologists, who use values at the central site to estimate
personal exposure. For this set of regressions, the mean R: was
0.60(N = 743), with a range from 0.19 to 0.88 (median = 0.73).
Again, we caution that our assumptions force high R' values, and
these should be considered upper bounds for the true R: values.
Tables 3-16 and 3-17 provide clear evidence of some
degradation in the relationship between personal exposure to
particles of outdoor origin and the outdoor concentrations as we
consider measurements made further from the home. 30 of the
37 participants showed higher correlations with outdoor
concentrations measured just outside the home than with
concentrations measured at the central site by the FRM. The
range in R2 differences was -0.24 to +0.52, with an average
differenceofO.il.
Would these regressions show improved results if we considered
each season separately? There would be an advantage perhaps in
that the infiltration factor would probably be less variable within
a season than across seasons. However, a drawback would be
the limitation to no more than 7 observations per household per
season. After testing this on the seasonal values, we found that
the drawback of fewer observations outweighed the advantage of
reducing cross-season variability. That is, adjusted R" values by
season included many that were well below the minimum of 0.19
observed for the full set of year-round observations. The
adjusted R: results for seasonal regressions with the residential
outdoor and central-site outdoor monitors are shown in Figures
A-l to A-3 in the Appendix.
Use of Reported Time in Indoor and
Outdoor Microenvironments to Predict the
Outdoor Exposure Factor Fpex from the
Infiltration Factor Finf
In some studies, personal exposure is not measured; therefore, a
general theory has been developed using time-activity data and
measurements of indoor and outdoor concentrations as a way of
estimating personal exposure. Equation 3-6 provides a way to
predict the outdoor exposure factor Fpl,v from the fractions ot
time indoors and outdoors and the measured infiltration factor
F,m-. Since we also have a measured value of Fpiv, using the
sulfur exposure/outdoor ratio, we can compare values predicted
by Equation 3-6 to measured values.
From the daily time-activity questionnaires, we can find the time
spent in various activities (Table 3-18). We define the fraction
of time outdoors/',,,, as the time either outdoors or in travel, with
a mean value of 10%. The fraction of time indoors/,, (mean
90°'o) includes indoors at home and all other locations. For each
participant and each day we can calculate Fp^ from Equation 3-
6. Table 3-18 and Figure 3-25 show that the predicted value of
/), exceeds the measured value at nearly all points of the
distribution. Although the R: is 0.57, we can almost match it
28
-------
Table 3-16. Regression of Personal Exposure to Particles of Outdoor Origin on Outdoor Concentrations Measured Near Residence by Harvard
Impactor (HI)
Person
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Sum/Mean
N
25
24
21
23
27
27
27
6
25
21
9
27
7
27
27
27
25
6
22
22
28
13
14
23
14
21
28
25
27
14
28
22
13
27
6
21
21
750
Slope
0.48
0.44
0.50
0.44
0.68
0.56
0.65
0.31
0.53
0.35
0.32
0.42
0.72
0.44
0.67
0.55
0.40
0.45
0.54
0.55
0.55
0.48
0.48
0.52
0.69
0.60
0.40
0.36
0.40
0.58
0.30
0.27
0.55
0.40
0.46
0.43
0.50
0.53
SE
0.04
0.05
0.05
0.05
0.06
0.05
0.07
0.04
0.08
0.03
0.03
0.10
0.12
0.04
0.05
0.03
0.07
0.09
0.04
0.04
0.04
0.10
0.06
0.04
0.11
0.06
0.03
0.04
0.06
0.06
0.07
0.04
0.17
0.06
0.12
0.07
0.05
0.01
Inter
1.47
1.36
3.12
0.46
-0.08
-0.93
-0.81
1.10
0.76
1.83
2.45
0.00
-0.11
-0.12
-0.90
0.59
1.29
1.02
2.43
-1.10
-1.10
1.14
0.02
0.68
-2.20
0.20
2.48
1.90
2.38
4.00
2.77
1.39
1.77
2.56
-0.46
1.61
0.46
0.26
SE
0.83
1.20
1.29
1.21
1.28
0.98
1.66
1.24
1.76
0.66
0.69
1.39
3.11
0.70
0.97
0.55
1.25
2.03
0.95
0.73
0.73
2.19
1.36
0.83
3.21
1.57
0.61
0.75
1.19
1.69
1.17
0.78
4.34
1.22
3.01
1.28
0.76
0.26
P
0.09
0.27
0.03
0.71
0.95
0.35
0.63
0.42
0.67
0.01
0.00
1.00
0.97
0.87
0.36
0.30
0.31
0.64
0.02
0.15
0.15
0.61
0.99
0.42
0.51
0.90
0.00
0.02
0.06
0.04
0.03
0.09
0.69
0.05
0.89
0.23
0.55
0.40
R2
0.84
0.77
0.84
0.76
0.82
0.84
0.79
0.91
0.63
0.84
0.80
0.77
0.74
0.83
0.85
0.93
0.59
0.81
0.88
0.92
0.85
0.63
0.83
0.87
0.76
0.83
0.88
0.78
0.62
0.88
0.42
0.67
0.44
0.60
0.74
0.64
0.81
0.71
29
-------
Table 3-17. Regression of Personal Exposure to Particles of Outdoor Origin on Outdoor Concentrations Measured at Central Site by Federal
Reference Method (FRM^
Subject
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
35
36
37
38
Sum/
Mean
N
25
24
21
23
27
27
27
6
25
21
9
27
7
27
27
27
25
6
22
22
28
13
14
23
14
21
28
25
27
14
28
22
13
27
6
21
21
750
Slope
0.58
0.45
0.48
0.40
0.76
0.50
0.57
0.36
0.55
0.46
0.62
0.35
0.54
0.48
0.71
0.58
0.35
0.52
0.47
0.58
0.51
0.43
0.45
0.41
0.68
0.60
0.41
0.34
0.37
0.60
0.31
0.33
0.57
0.23
0.56
0.46
0.53
0.52
SE
0.06
0.06
0.06
0.05
0.10
0.04
0.07
0.06
0.09
0.06
0.18
0.04
0.15
0.05
0.05
0.07
0.10
0.22
0.06
0.07
0.14
0.12
0.06
0.09
0.10
0.07
0.04
0.04
0.06
0.09
0.06
0.06
0.11
0.08
0.09
0.06
0.08
0.02
p (slope)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.00
0.01
0.00
0.00
0.00
0.00
0.08
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.00
0.00
0.00
0.00
Intercept
0.95
2.71
5.81
1.49
0.24
0.39
2.07
0.49
1.11
1.05
3.08
2.58
-0.92
-0.11
0.25
0.85
2.09
0.08
5.18
-1.47
4.08
1.64
1.39
2.48
-1.56
1.56
3.01
2.23
3.61
5.77
2.66
0.85
0.93
6.06
-2.39
1.99
0.38
1.29
SE
1.08
1.16
1.35
1.19
1.75
0.87
1.77
1.54
1.82
0.97
4.32
0.67
1.88
0.77
0.83
1.11
1.77
4.54
1.22
1.32
2.93
2.64
1.20
1.68
3.12
1.70
0.73
0.83
1.14
2.30
1.09
1.00
2.95
1.40
2.20
1.02
1.10
0.31
P (Int)
0.39
0.03
0.00
0.22
0.89
0.65
0.25
0.77
0.55
0.29
0.50
0.00
0.64
0.88
0.76
0.45
0.25
0.99
0.00
0.28
0.17
0.55
0.27
0.15
0.63
0.37
0.00
0.01
0.00
0.03
0.02
0.41
0.76
0.00
0.34
0.07
0.73
0.00
R2
0.77
0.74
0.76
0.73
0.69
0.84
0.69
0.88
0.59
0.74
0.57
0.77
0.67
0.8
0.87
0.74
0.34
0.47
0.71
0.77
0.33
0.5
0.83
0.49
0.76
0.78
0.82
0.73
0.56
0.76
0.47
0.59
0.68
0.19
0.88
0.73
0.67
0.60
30
-------
Table 3-18. Time (in Minutes) Spent in Various Activities/Locations
Activity/Location
Indoors at home
Cooking
Cleaning
Grooming
Other indoor locations
Travel
Outdoors
Unknown
Exposed to smoke
Outdoors + travel
Indoors
fou, (fraction of time outdoors or in travel)
fm (fraction of time indoors)
Infiltration factor Fmf (indoor/outdoor S)
Measured outdoor factor Fpex (personal/outdoor S)
Predicted outdoor factor F^x from Equation 3-6
N
727
727
727
727
727
727
727
727
727
727
727
727
727
727
727
727
Mean
1179
98
39
57
90
78
67
11
14
145
1295
0.10
0.90
0.59
0.55
0.63
SD
181
72
45
47
82
61
63
22
44
103
103
0.07
0.07
0.16
0.14
0.15
Min
495
0
0
0
0
0
0
0
0
0
840
0.00
0.58
0.17
0.16
0.19
10th
915
15
0
15
0
0
0
0
0
30
1155
0.02
0.80
0.38
0.38
0.44
25th
1065
45
0
30
30
30
15
0
0
60
1230
0.04
0.85
0.47
0.45
0.53
Median
1215
75
30
45
75
60
45
0
0
120
1320
0.08
0.92
0.58
0.54
0.63
75th
1320
135
60
75
120
105
105
15
0
210
1380
0.15
0.96
0.68
0.63
0.72
90th
1395
195
105
120
200
165
150
30
45
285
1410
0.20
0.98
0.79
0.73
0.82
Max
1440
480
210
480
540
345
390
135
375
600
1440
0.42
1.00
1.06
1.08
1.05
31
-------
by using only Fmf (not F,,,.v) to estimate personal exposure
(Figure 3-26). The latter regression has a higher slope (0.85 vs.
0.80), a lower intercept (0.12 vs. 0.19) and an R2 of 0.49,
indicating that the extra information from the time-activity diary
was not sufficient or was not precise enough to produce a better
result than simply using the indoor/outdoor sulfur ratio alone.
1.2
y = 0.80 (0.03 SE) x t 0.19 (0.01)
5 £ R; = 0.57: N = 727
0 02 0.4 0.6 0.8 1
Measured ratio of personal vs outdoor sulfur
Figure 3-25. Predicted value of the outdoor exposure factor Fpex using
Equation 3-6 compared to the measured value using the personal/outdoor
sulfur ratio.
y = 0.667 (0.025)x + 4.0 (0.7)
0.49 N = 753
40 60 80 100 120 140 160
Observed Personal PM2 5 Exposure (jig/m3)
Figure 3-26. Predicted personal exposure to PM25 using only Fml.
Equation 3-2 relates personal exposure to the time-weighted
averages of the indoor and outdoor microenvironments. We
have measurements of tine particle concentrations both indoors
and outdoors, but they are only 24-h averages and are not
necessarily the actual concentrations experienced by the persons
at the time they were in those microenvironments. Also, fine
particle concentrations were not measured in some locations such
as the car and workplace. If we assign the measured residential
outdoor value to all the outdoor and transport
microenvironments, and the measured indoor concentrations in
the home to all the other indoor locations, we can test how well
Equation 3-2 does in predicting personal exposure by comparing
with the measured personal exposure. Results (Figure 3-26)
show a moderate ability of Equation 3-2 to predict exposure,
with an adjusted R2 of 0.49 (N=753). However, the average
personal exposure predicted from Equation 3-2 was only 19.4
ug/nv compared to the observed average exposure of 23.1 ng/m3.
The difference of 3.7 ug/mj provides an independent estimate of
the magnitude of the personal cloud, and this value agrees well
with the earlier estimates above. The inability of Equation 3-2 to
estimate personal exposure has been shown previously
(Ozkaynak et al., 1996; Pellizzari et al., 1999).
Estimating P and k
Calculating Average Values ofP and k
Equation 1-3 for the infiltration factor is nonlinear in the air
exchange rate a. Therefore if we plot our measured 24-h average
indoor/outdoor sulfur ratios versus the measured air exchange
rate, we can solve for the overall average P and k by minimizing
some appropriate function. We tried minimizing the squares of
the differences between the measured and modeled
concentrations (ordinary Gaussian least squares approach) and
also tried minimizing the absolute differences (a procedure
giving less weight to outliers). Both approaches gave almost
identical results, so we report only the results from the ordinary
least squares approach. Figure 3-27 gives the results for all the
individual 24-h measurements (N = 720). The estimate for the
penetration coefficient P averaged across all measurements is
0.85 (0.02 SE). The estimate for the average deposition rate k is
0.22 h"' (0.01 SE). The R2 value was moderately high at 0.45.
1.2 -]
0.2
P = 0.85 (0.016 SE): * = 0.22 (0.014) h'1
R- (adj) = 0.446: N = 720
Air Exchange Rate (h ')
Figure 3-27. Nonlinear least-squares fit to the indoor/outdoor sulfur ratio
vs. the air exchange rate. Bounding curves are ± 1 SE.
A second (linear) approach to calculating average values of P
and k follows from the simple time-averaged mass balance
equation for sulfur with no indoor sources:
»AIH, = Pa/(a+k)
(3-9)
This equation is not linear for one of the unknowns (k), and it
also mixes the values off and k together in one factor. We can
partially separate P from k by inverting the equation (Long et al.,
2001):
32
-------
Suu/Sin=k/(Pa)+l/P
(3-10)
y = 0.278ic » 1.22
RJ = 0.356 N * 716
Note that writing the equation this way isolates P in a single
term, the intercept of the regression. This value can then be used
to calculate k from the other term. Since the inverse of the air
exchange rate a is the residence time T, we can put this equation
into the form of a linear regression on r:
Soll/S,n = (k/P)r
(3-11)
The intercept of the regression will be our best estimate of P and
the slope will lead to an estimate of k (based on our value for P).
One problem with this approach is that the intercept is always an
extension of a line determined by points that may be far from the
intercept. For example, participants 4 and 6 in the following
graphs have no values of r lower than 2 h or 1 h, respectively, so
that the extrapolation of the line to a value of 0 goes well beyond
the domain where all the points lie. This will lead to more
variation and a greater uncertainty in our estimates of P.
As an initial check on whether the sulfate data provides useful
data for determining P and k we can run the regression on the
full data set (Figure 3-28).
3
o>
5 3
o
I2
y = 0.228 (0.012 SE) « * 1.328 (0.035)
R2 = 0.335: N =720
r/P
P = 0.75 (0.04): k = 0.17 (0.05) h''
10
Residence time T (h)
15
Figure 3-28. Regression of the outdoor/indoor sulfur ratio vs. residence
time.
However, there are four very influential outliers at unusually
high residence times (air exchange rates < 0.1 h"'). These are
unlikely values given other measurements in the same house. If
we rerun the regression without these outliers we obtain
somewhat larger values for P and k, as well as an improved R2
(Figure 3-29). The overall average deposition rate k= 0.23 (0.01
SE) h"' while the overall average penetration coefficient P =
Residence time i (h)
Figure 3-29. Same regression as in Figure 3-28 without four outliers.
0.81 (0.03 SE). Since measurement error generally leads to
lower slopes and higher intercepts, it may be that these estimates
are lower bounds for the actual average k and P. These linear
estimates agree well with the nonlinear estimates of 0.22 h"1 for k
and 0.85 for P.
Calculating Individual Home Values ofP
and k
In general the different ways of analyzing the data do not
disagree violently, and therefore we can go ahead and try to
calculate k and P for the individual homes. We first assume that
P and k do not change greatly across seasons for any home. This
is unlikely to be the case if certain household characteristics such
as window opening, use of fans and filters, changes by season.
Nonetheless, by making this assumption, we can make use of all
measurements for each house in a single regression. We can
employ a nonlinear fit to the observed values of the
indoor/outdoor sulfur ratio (see Equations 1-3 and 1-4),
calculating the best value of P and k for each home averaged
across all seasons (Table 3-19). The results are mixed, with R2
values ranging from zero to 0.94. 32 of 37 estimates for P were
significantly different from zero, but only 16 estimates of A: were
significantly different from zero (significant results shown in
boldface). Of the 32 significant estimates for P, three were
considerably greater than 1, a physical impossibility. The range
of significant P estimates was from 0.52 to 1.40, but the
interquartile range was more tightly clustered between 0.66 and
0.98.
There are 5 values for P (3 of them significant) that are well
above 1, ranging from 1.16 to 2.02. We can rerun the
regressions for these 5 cases bounding P from above at 1 (Table
3-20). All five of the new estimates of k are significantly
different from zero, compared to three when P was unbounded.
The R2 values are decreased slightly compared to the unbounded
case, but remain quite high, from 0.31 to 0.81.
33
-------
Table 3-19. Estimates of P and k for Individual Homes Using Nonlinear Fit to the Indoor/Outdoor Sulfur Ratio
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
32
33
34
37
38
31
36
35
N
22
23
28
21
20
26
25
6
21
25
8
28
7
27
24
28
19
6
27
27
26
13
13
27
8
22
27
24
31
28
20
10
19
15
14
5
0
P
0.85
0.61
0.83
0.61
1.16
0.67
0.88
0.44
0.80
0.71
0.67
0.91
0.30
0.45
1.03
0.67
0.60
1.17
1.00
0.52
0.88
0.97
0.82
0.77
1.37
0.83
0.77
0.75
1.46
0.60
0.55
2.02
0.79
0.68
SE
0.10
0.04
0.04
0.15
0.15
0.07
0.04
0.23
0.09
0.08
0.05
0.12
0.08
0.05
0.18
0.03
0.07
0.86
0.07
0.04
0.08
0.23
0.10
0.10
0.15
0.07
0.04
0.08
0.38
0.20
0.09
2.05
0.06
0.06
P(P)
0.000
0.000
0.000
0.001
0000
0.000
0.000
0.134
0.000
0.000
0.000
0.000
0.013
0.000
0.000
0.000
0.000
0.246
0.000
0.000
0.000
0.001
0.000
0.000
0.000
0.000
0.000
0.000
0.001
0.006
0.000
0.352
0.000
0.000
Did not converge
Did not converge
Same as house 29
k
0.26
-0.01
0.12
0.09
0.41
0.06
0.14
0.06
0.17
0.14
-0.01
0.29
-0.04
0.00
0.23
0.02
0.03
0.37
0.35
0.02
0.21
0.28
0.50
0.18
0.48
0.13
0.15
0.13
0.69
0.18
0.22
0.93
0.12
0.07
SE
0.13
0.04
0.04
0.09
0.12
0.06
0.05
0.16
0.08
0.07
0.03
0.10
0.04
0.04
0.13
0.03
0.02
0.50
0.10
0.03
0.07
0.16
0.15
0.08
0.12
0.07
0.06
0.06
0.31
0.15
0.12
1.36
0.04
0.03
PW
0.049
0.861
0.012
0.326
0.003
0.377
0.013
0.746
0.036
0.065
0.725
0.006
0.358
0.994
0.097
0.557
0.192
0.498
0.002
0.505
0.004
0.106
0.007
0.033
0.007
0.071
0.021
0.046
0.033
0.246
0.086
0.511
0.004
0.048
R2
0.29
0.00
0.27
0.17
0.53
0.04
0.32
0.05
0.35
0.23
0.02
0.51
0.11
0.00
0.25
0.02
0.15
0.40
0.51
0.02
0.53
0.48
0.74
0.28
0.94
0.22
0.27
029
0.53
0.15
0.37
0.39
0.55
0.37
34
-------
Table 3-20. Values for k when P is Bound from Above by 1
Subject N R2 k SE
25 8 0.81 0.21 0.03
29 31 0.48 0.33 0.03
34 10 0.31 0.26 0.03
18 6 0.40 0.28 0.02
5 20 0.50 0.28 0.03
35
-------
A similar approach to the above nonlinear use of the
indoor/outdoor sulfur ratio is to use the linear equation involving
the inverse of that ratio, as well as the inverse of the air exchange
rate (Table 3-21). 24 homes had both slope and intercept
significantly different from zero (shown in boldface).
Five values of P were well above 1, so the regressions were
rerun bounding P from above by 1 (Table 3-22). All 5 new
estimates of A" are significantly different from zero, compared to
none when P was unbounded. The R" values arc decreased
slightly compared to the unbounded case, but remain quite high,
from 0.32 to 0.82.
The estimates of P and k from the nonlinear approach (Table 3-
19) are compared to the estimates from the linear approach
(Table 3-21) in Figures 3-30 and 3-31. Only values significantly
different from zero are included. Figure 3-30 suggests that the
nonlinear approach gave consistently smaller estimates of P at
the high end of the range. In fact, of the five estimates exceeding
unity due to the linear approach, all five were lower, and two
were less than one, in the nonlinear approach.
The median value for P was 0.81 (interquartile range 0.66 to
0.90). The median for k was 0.24 h"1 (interquartile range 0.12 to
0.35 h'1).
Having calculated P and k for each home from the linear
regression of the outdoor/indoor sulfur ratio vs. the residence
time, the estimated average infiltration factor for each home can
be calculated from the equation
ml = Pa/(a+k)
(3-12)
where a in this case is the arithmetic, geometric, or harmonic
average of the air exchange rates for each home.
2
1.8
16
1.4
1.2
0. 1
08
0.6
0.4
0.2
P(lln)
P(nlin)
29 25 15 22 21 19 31
Figure 3-30. Comparison of estimates of P from the linear and nonlinear
approaches described in the text. Only values significantly different from
zero are plotted (N = 32 homes).
1.2
1
0.3
£ 0.6
0.1
0.2
k(lm|
K(nfm)
I
23 26 a 22 H 21 19 16 4 12 1 28 24 5 27 26 7 37 10 9 3 U X 17
House ID
Figure 3-31. Comparison of the estimates of k from the linear and
nonlinear approaches described in the text. Only values significantly
different from zero are plotted (N = 24 homes).
Estimating Finf from Individual Values ofP
and k
These estimated values of P and k from the two approaches
(linear and nonlinear) can be used to estimate F,n/-for each home.
These estimates are compared to the observed indoor/outdoor
sulfur ratio in Figure 3-32. Although the excellent agreement
(R: = 0.96 for the nonlinear estimate and 0.99 for the linear
estimate) of the F,,,/ estimates may be an artifact due to the
related nature of the three calculations, at least this agreement
suggests that the P and k values used to estimate Fml from both
the linear and nonlinear regressions (which both involve the air
exchange rate or its inverse) are at least consistent with the
measured indoor/outdoor sulfur ratios (which do not involve the
air exchange rate). However, the many cases in which neither
the linear nor the nonlinear approach gave values for k
significantly different from zero, and the several cases in both
approaches in which values of P were greater than unity,
suggests that the assumption of constant values for k and P
across seasons for each home is violated.
1
0.9
0.8
0.7
0.6
2
y = 0.97x * 0.003
R! = 0.96. N = 37
y = 0.98x » 0.004
R' = 0.99, N = 37
0.2
0 0.2 0.4 0.6 0.8 1
P - from Sulfur Indoor-Outdoor Ratio
Figure 3-32. Comparison of the infiltration factor (F,nf) estimates from the
simple ratio of indoor sulfur to outdoor sulfur by home vs. the nonlinear
regression of the same ratio using the measured air exchange rates and
the linear regression of the inverse ratio (outdoor/indoor) against the
residence time.
36
-------
Table 3-21. Results of Linear Regressions of the Outdoor/Indoor Sulfur Ratio on Residence Time for 36 Homes
House
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
31
32
33
34
36
37
38
Sum/Mean
N
22
23
28
21
20
26
25
6
21
25
8
28
7
27
24
28
19
6
27
27
26
13
13
27
8
22
27
24
31
14
28
20
10
5
19
15
720
1/P
1.19
1.66
1.24
1.08
1.12
1.47
1.14
2.25
1.45
1.51
1.49
1.13
3.03
2.20
0.85
1.51
1.66
0.99
1.00
1.93
0.96
0.88
1.19
1.23
0.65
1.17
1.28
1.15
0.52
1.05
1.26
2.59
0.58
3.46
1.28
1.56
1.33
SE
0.13
0.11
0.06
0.45
0.21
0.15
0.05
1.36
0.13
0.15
0.10
0.16
0.79
0.24
0.18
0.07
0.20
0.65
0.06
0.14
0.12
0.29
0.17
0.21
0.07
0.10
0.08
0.17
0.25
0.08
0.62
0.35
0.54
1.78
0.10
0.11
0.03
P
0.84
0.60
0.81
0.93
0.90
0.68
0.88
0.44
0.69
0.66
0.67
0.89
0.33
0.45
1.18
0.66
0.60
1.01
1.00
0.52
1.05
1.14
0.84
0.81
1.54
0.85
0.78
0.87
1.91
0.96
0.79
0.39
1.73
0.29
0.78
0.64
0.75
SE
0.09
0.04
0.04
0.39
0.17
0.07
0.04
0.27
0.06
0.06
0.05
0.13
0.09
0.05
0.26
0.03
0.07
0.67
0.06
0.04
0.13
0.37
0.12
0.14
0.16
0.07
0.05
0.13
0.91
0.07
0.39
0.05
1.64
0.15
0.06
0.04
0.00
k/P
0.31
0.00
0.11
0.32
0.25
0.10
0.16
0.14
0.14
0.17
-0.01
0.32
-0.05
0.04
0.30
0.03
0.07
0.29
0.36
0.04
0.35
0.36
0.64
0.29
0.40
0.19
0.24
0.27
0.58
0.16
0.47
0.19
0.44
0.14
0.15
0.09
0.23
SE
0.09
0.07
0.04
0.11
0.09
0.08
0.05
0.33
0.04
0.06
0.04
0.06
0.17
0.07
0.09
0.05
0.03
0.19
0.06
0.04
0.05
0.10
0.12
0.08
0.03
0.07
0.07
0.07
0.09
0.11
0.16
0.07
0.21
0.44
0.03
0.03
0.01
k
0.26
0.00
0.09
0.29
0.22
0.07
0.14
0.06
0.10
0.11
-0.01
0.28
-0.02
0.02
0.35
0.02
0.04
0.29
0.35
0.02
0.37
0.41
0.54
0.23
0.61
0.16
0.18
0.24
1.11
0.15
0.37
0.08
0.77
0.04
0.12
0.05
0.17
R2
0.34
0.00
0.20
0.28
0.25
0.02
0.33
0.00
0.32
0.20
0.00
0.52
0.00
0.00
0.30
0.00
0.22
0.19
0.60
0.00
0.66
0.52
0.72
0.29
0.96
0.25
0.31
0.40
0.55
0.07
0.21
0.27
0.28
0.00
0.53
0.39
0.34
37
-------
Table 3-22. Values for k when P is Bound from Above by 1
Subject
15
22
25
29
34
N
24
13
8
31
10
k
0.22
0.32
0.26
0.28
0.44
SE
0.02
0.03
0.03
0.03
0.05
R2
0.32
0.55
0.82
0.55
0.34
38
-------
Seasonal Analysis
Up to this point, the analyses have been carried out using values
averaged across all visits during the year. However, some
seasonal variation was noted in the values for the infiltration
factor F-mj and the personal exposure factor F/>ex. In this section,
we test whether we can obtain useful results on an individual
home or person on a seasonal basis by regressing personal and
indoor sulfur on outdoor sulfur concentrations by season. The
advantage to this is that the seasonal variations, if any, will be
observed; the disadvantage is that limiting the regression to a
maximum of 7 values is likely to lead to wider variance and
more uncertainty in estimating the two quantities of interest.
Since air exchange rates are the only measured variable
contributing to the infiltration factor (the others being the
unmeasured P and k parameters), we compare the (harmonic)
average air exchange by house and by season to the estimate of
/"^obtained by taking the average indoor/outdoor sulfur ratio by
house and by season (Appendix Table A-l). Then we compare
the estimate of Fin/ from the indoor/outdoor sulfur ratio to the
corresponding estimate obtained by regressing indoor on outdoor
sulfur and taking the slope of the regression as an estimate of Fml
for each house and season (Appendix Table A-2). The estimates
of Fmf from the ratio are compared to the estimates from the
slope in Figure 3-33. The R" value is 0.62, and the estimates
from the regression slopes vary more widely (from 0 to >1) than
the estimates from the indoor/outdoor ratios (0.2 to <1).
1.2
0.2 4
0.2
0.3
0.4
0.5
0.6
Ratio
0.7
0.8
Figure 3-33. Estimates for each home by season of the infiltration factor
F,n,from regressing indoor sulfur on outdoor sulfur (Slope) compared to
estimates from the simple ratio of indoor sulfur to outdoor sulfur averaged
over all visits in a season.
Another way of comparing the estimates using the regressions to
the estimates using the ratios is provided by Figures 3-34 and 3-
35. Figure 3-34 presents the homes ordered by season and by the
value of Finj as determined from the ratio; Figure 3-35 presents
the comparable results from the regressions. Comparing the two
figures shows many cases in which the regression approach
predicts low values for the infiltration factor (< 0.4), as well as a
smaller number of cases in which the approach predicts very
high values, including one impossible result (>1). This analysis
confirms that the ratio provides the more stable estimate of the
infiltration factor.
1
0.9
-- 0.8
J 0.7
!o.6
2
6 nil - -
%
V
^
.
IU-J
« 0.2
0.1
n
Summer;
- Fall \
Winter '
Spring
Figure 3-34. Estimates of fm, by home from the indoor/outdoor sulfur
ratio.
1.2
a.
* 0.8
o °-6
|
1 0.4
Summer
Fall
Winter
Spring
Figure 3-35. Estimates of Fm, by home from regressions of indoor on
outdoor sulfur.
Multivariate Regressions
All participants filled out a questionnaire on their household
characteristics and daily activities. A total of 54 questionnaire
variables were included in the dataset (see Table A-4 in the
Appendix for the variable names and definitions). We carried out
a series of multiple regressions on the personal, indoor, and
outdoor fine particle and sulfur concentrations vs. these
questionnaire responses to try to identify sources of increased
exposure.
Our first priority was to investigate the 54 independent variables
for possible collinearity. This procedure is explained fully in the
book Regression Diagnostics by Belsley, Kuh, and Welsch
(1980). A factor matrix is prepared of the 54 variables and a
"condition number" is calculated for each of the eigenvalues of
the matrix. Belsley, Kuh. and Welsch (1980) recommend that if
39
-------
a condition number exceeds 30, then variables that have a heavy
weight on that eigenvalue should be inspected and either
combined or a way found to drop one of the variables. Four
eigenvalues did in fact exceed a condition number of 30. The
pairs of variables with the heaviest loadings on each eigenvalue
were ROOMS/AREA, DRYER/DRYER VENT,
DSTFACTORavg/C FUEL, and TEMPC/TEMPDELTA. Since
AREA seemed the more precise factor, we dropped ROOMS.
Since DRYER seemed the more fundamental variable, we
dropped DRYER_VENT. And since TEMPDELTA was a
calculated variable including a number of imputed values, we
dropped it. The choice between DSTFACTORavg and C_FUEL
was more difficult. The dust factor as estimated by the
technician has previously (in the PTEAM Study, for example)
been one of the strongest predictors of airborne particles.
However, the cooking fuel (electric or gas) is the more
fundamental variable, so we reluctantly dropped
DSTFACTORavg from the list of variables. The final list
contained 50 variables dealing with household characteristics,
personal activities, and two measured quantities (air exchange
and outdoor temperature).
To include 50 variables in the regression and achieve an overall
significance of p<0.05, the Bonferroni criterion was applied by
dividing the chosen significance level by the number of
variables. This gives us a p-value of 0.001 as the value required
to achieve an overall significance of p<0.05 for the final model.
In all of the following tables dealing with multiple regression
results, the variables that meet the Bonferroni criterion for
significance are listed in boldface type.
The first step in the multiple regression approach was to carry
out a stepwise regression (combined forward selection and
backward elimination) on all 50 household characteristics and
personal activities as well as one or two measured particulate
matter or sulfur variables using Statistica 6.1 software. A second
regression was run including only those variables at the p < 0.05
level. It is important to run the second regression on the smaller
set of variables since this will include more cases (typically, for
our data set, 20-70 additional records that were dropped because
of the casewise elimination employed for all 50 independent
variables.) To check our results, all of the initial 50-variable
regressions were rerun on SAS statistical software using a
standard backward elimination stepwise regression. Normally,
SAS and Statistica identified the same significant variables, with
minor differences in the slopes, intercepts, and R: values (the
latter were always within 0-3% of each other). When a variable
differed, both final versions were run one more time in Statistica,
and the run with the higher R" value was selected.
Although major collinearities were avoided, less strong ones
remain. This could cause one or two of a collinear pair of
variables to fail to register as significant in the first run involving
50 variables; they will then have no chance to be considered in
the final run picking out only the significant variables from the
first run. In some situations, we inserted a variable that we
suspected might be relevant after the "final" run and discovered
that it too was significant, but never succeeded in changing the
final R: value by more than 1% when adding individual
variables, indicating that the most important variables in each
regression were identified.
In carrying out analyses of how household characteristics and
personal activities affect indoor concentrations of particles, we
need to carefully examine how the study design may have
created unequal conditions among the homes. First, not all
homes were visited for an equal number of seasons. Only 24
homes were visited in all four seasons. Second, even within a
season, only 3-6 homes were visited each week. Therefore
different homes within any given season encountered different
outdoor conditions. The homes were scrambled each season so
that the same homes were not visited together in a given week
over two seasons, so this should have evened out the outdoor
conditions encountered to some extent, at least for the 24 homes
visited all four seasons. However, for the entire group of 36
homes, there was extensive variation in the outdoor PM:5
concentrations, whether at the central site or outside the homes,
ranging over a factor of two from the highest to the lowest
(Figure 3-36).
35
30
25
20 -
I
v .
ambHI2S
FRM2S
PM25ui.ii
Figure 3-36. Central-site and residential outdoor concentrations
averaged over all visits to a home.
Clearly, if homes at the high end of the outdoor concentrations
happen by chance to have some characteristics substantially
different from those at the low end, a simple regression of
outdoor concentrations at the central site on these household
characteristics may show significant results, even though the
relationship cannot be causal (household characteristics cannot
affect outdoor concentrations). In fact, some of these
"impossible" relationships are observed. For example, a
regression of the two central site monitors and the outdoor
residential monitor on various household characteristics showed
a number of artifactual relationships (Table 3-23). The building
age and air exchange rates were among the strongest variables in
40
-------
Table 3-23. Multiple Regression of Outdoor Concentrations on Household Characteristics and Personal Activities
Variable
PM25out
Intercept
AGE
Windopen
airex
outdoor
DRYER
ambHI25
Intercept
airex
AGE
C_FUEL
Windopen
FAN
MILDEWavg
spacehtr
VAC
FRM
Intercept
airex
AGE
FAN
C_FUEL
Windopen
MILDEWavg
spacehtr
VAC
Slope
13.78
0.13
3.94
-3.18
0.01
1.84
10.58
-2.82
0.11
2.95
2.45
-3.66
-2.77
-4.63
2.00
10.60
-2.56
0.10
-3.73
2.46
1.97
-2.74
-4.27
1.95
Std.Err.
1.13
0.02
0.65
0.57
0.01
0.87
1.82
0.52
0.02
0.72
0.66
1.00
1.03
1.96
1.02
1.68
0.49
0.02
0.93
0.65
0.60
0.96
1.87
0.95
p-level N R2(adj.)
720 0.12
0.000000
0.000000
0.000000
0.000000
0.003668
0.034377
743 0.10
0.000000
0.000000
0.000004
0.000042
0.000202
0.000264
0.007241
0.018484
0.049494
772 0.10
0.000000
0.000000
0.000002
0.000067
0.000168
0.001016
0.004431
0.022320
0.039863
41
-------
all three cases, along with open windows in two of the three
cases, but it is likely that these variables simply represent
differential outdoor air conditions encountered at the different
times that the homes were monitored. The variables "explain"
only 10-12% of the variance of the outdoor monitors, but they
represent a caution nonetheless in our interpretation of
"significant" variables in all other of the multiple regressions we
will consider. In particular, those regressions that include both
an outdoor measurement and one or more of the variables AGE
and airex (and windopen) will be putting the variables into
"double jeopardy", examining their effect both explicitly and
implicitly as part of the outdoor particle measurement. Since the
signs of airex and windopen are opposite, one might think that
because they are correlated they end up with opposite signs, a
common occurrence in regressions with correlated variables.
Although it might seem that air exchange rates should be fairly
well correlated with open windows, the Spearman correlation ot
airex and windopen was only 0.24. Nonetheless, the first
regression in Table 3-23 was rerun twice, dropping airex from
the first run and then restoring it and dropping windopen from
the second run, to test whether some kind of cross-tenn
relationship was occurring. However, in both cases the variable
left in continued to be significant with the sign in the same
direction as before, and the adjusted R" dropped in both cases,
once to 0.10 and once to 0.08. Therefore, the "best" model for
the residential outdoor PM: 5 measurements continues to be the
one listed first in Table 3-23. Again, for the definitions of the
variables appearing in the next five tables, see Table A-4 in the
Appendix.
The indoor model is
PM25in =a+ fi,,n*PM25oitt + fi*x,
(3-13)
where the x, are the 50 appropriate continuous/categorical
variables.
The model above can be repeated using FRM25 as the
independent variable:
PM25m = a + ft,,ul*FRM25 + [i*x,
(3-14)
The point of using the FRM25 as the outdoor variable is that
epidemiological studies are often limited to the central-site
monitor.
The results of these model runs are prov ided in Table 3-24. The
first regression is a simple regression of indoor PM-> 5 on outdoor
PM: 3. As can be seen, the relationship is particularly poor, with
an R" of only 0.09. Since epidemiologists are often restricted to
use of a fixed-site urban monitor, the regression is rerun using
the central-site FRM monitor. The R: remains low at 0.11.
These results are consistent with other studies showing low
cross-sectional relationships between indoor and outdoor PM
levels.
The R: value is increased to above 0.40 for multiple regressions
including the questionnaire responses and either the residential
outdoor monitor or the central-site monitor. In each case, the
outdoor PM concentration has the greatest influence on the
indoor PM (as judged by the p-value). Estimates forthe outdoor
contribution to indoor PM ranged from 48% (using residential
outdoor PM) to 57% (using central site FRM). The difference in
these estimates can be attributed to the consistently lower values
returned by the FRM. Investigators have noted that the FRM
may underestimate PM levels due to loss of volatile species such
as nitrates.
Burned food added about 12 ug/nr to the indoor concentration,
while a nearby dirt road added 8.4 to 8.9 ug/nr, and use of an
exhaust fan added about 5 ug/nr. It is likely that the use of the
exhaust fan because of cooking or burned food added to the
indoor concentration.
Homes with persons who reported being near cigarette smoke
also had higher indoor fine particle concentrations. The number
of smokers increased indoor PM concentrations by 4-6 ug/nr,
while time spent near smokers increased concentrations by 0.04
ug/nr per minute exposed.
Cooking increased PM: 5 concentrations by about 0.03 ug/nv per
minute. Homes with electric stoves had PM: 5 concentrations
6.6-6.8 ug/m"1 higher than homes with gas stoves, but it should be
remembered that the homes with electric stoves also had higher
outdoor PM: 5 levels (see Table 3-23). Each additional person
living in the household added 1.3 ug/'nr to the daily average
PM: 5 concentrations. Vacuuming appeared capable of
increasing daily average PM: 5 concentrations by about 4 ug nr.
although the variable was only marginally significant (using the
Bonferroni criterion) in one of the two regressions. Finally,
homes with a clothes dryer were associated with a decrease of
4.7 ug/m' in their daily average concentrations. The presence of
a clothes dryer (nearly all of which were vented outdoors)
increases air exchange since almost the same volume of air must
enter the house to replace the heated air vented outdoors.
The next series of two multiple regressions takes advantage of
our ability to split indoor concentrations into indoor-generated
and outdoor-generated fine particles (Table 3-25).
The model for indoor-generated tine particles is the following:
Inconthh = a + ft, *x,
(3-15)
"Incontrib" is the contribution of indoor sources after subtracting
the outdoor contnbution determined by the indoor outdoor sulfur
ratio; therefore, this model does not include the outdoor PM^5
variable. The model was initially run with this variable included
42
-------
Table 3-24. Dependence of Indoor Fine Particle Concentrations on Household Characteristics and Personal Activities
Variable
PM25in
Intercept
PM25out
PM25in
Intercept
FRM25
PM25in
Intercept
PM25out
DIRT_RD
Burning
C_FUEL
Numpeopl
Exhstfan
cooking
numsmok
DRYER
vacuum
smoke
Otherjndoor
Unknown
outdoor
PM25in
Intercept
FRM25
Burning
DIRT_RD
Exhstfan
C_FUEL
Numpeopl
cooking
vacuum
DRYER
Otherjndoor
Unknown
smoke
numsmok
outdoor
Slope
8.47
0.54
7.21
0.67
-10.50
0.48
8.89
11.80
6.77
1.34
5.04
0.03
6.33
-4.66
3.89
0.04
0.09
0.06
-0.02
-11.15
0.57
11.94
8.42
5.34
6.57
1.30
0.03
4.04
-4.65
0.10
0.06
0.04
4.55
-0.02
Std.Err.
1.37
0.06
1.36
0.07
2.91
0.05
1.42
1.93
1.36
0.29
1.11
0.01
1.71
1.40
1.20
0.01
0.03
0.02
0.01
2.91
0.06
1.91
1.40
1.11
1.36
0.28
0.01
1.19
1.40
0.03
0.02
0.01
1.76
0.01
p-level N
775
0.000000
0.000000
760
0.000000
0.000000
762
0.000325
0.000000
0.000000
0.000000
0.000001
0.000003
0.000007
0.000099
0.000225
0.000912
0.001220
0.001506
0.002624
0.012012
0.016669
747
0.000138
0.000000
0.000000
0.000000
0.000002
0.000002
0.000006
0.000045
0.000750
0.000909
0.001069
0.004758
0.006304
0.009757
0.021123
R2(adj.)
0.09
0.11
0.42
0.42
43
-------
Table 3-25. Dependence of Indoor-Generated and Outdoor-Generated Particles on Household Characteristics and Personal Activities
Variable
Incontrib
Intercept
Burning
C_FUEL
DIRT_RD
airex
Numpeopl
Exhstfan
S_WIN
cooking
smoke
Otherjndoor
outdoor
Cleaning_1
DRYER
Unknown
Outcontin
Intercept
PM25out
Outcontin
Intercept
PM25out
TempC
AGE
Windopen
cigsmokd
A_C
windowall
Numpeopl
FLRCOVav
vacuum
airex
fry
Slope
-12.03
12.06
8.70
8.01
-4.42
1.43
4.89
-4.56
0.03
0.05
0.10
-0.03
2.60
-3.64
0.05
0.04
0.58
-0.64
0.59
-0.15
0.06
1.68
035
0.39
0.0022
-018
-0.02
0.64
055
-0.47
Std.Err.
2.88
1.97
1.45
1.45
0.83
0.30
1.15
1.12
0.01
0.01
0.03
0.01
1.00
1.45
0.02
0.31
0.01
0.54
0.01
0.02
0.01
0.23
0.06
0.07
0.0006
0.06
0.01
0.25
0.23
0.21
p-level N R2(adj.)
709 0.37
0.000033
0.000000
0.000000
0.000000
0.000000
0.000003
0.000024
0.000055
0.000064
0.000068
0.001016
0.002000
0.009668
0.012452
0.024292
775 0.69
0.889273
0.000000
686 0.82
0.239662
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000069
0.002316
0.002557
0.011674
0.014491
0.025024
44
-------
as an independent variable to confirm that it had been fully
accounted for. Results showed that it was not significant and
could be omitted from the final model.
The indoor-generated contribution (Incontrib) depends on many
of the variables appearing in the Equation 3-13 and 3-14
regressions for indoor PM:5 with about the same coefficient
values, but does not depend on the outdoor concentration. The
value of R: is 0.37 and again, burned food, cooking, use of
electric stoves, nearness to a dirt road, use of the exhaust fan, the
number of people living in the home, and exposure to ETS turn
up as significant variables affecting indoor-generated particles.
A new variable, the air exchange rate (airex), reduces indoor-
generated concentrations as airex increases.
The model for the outdoor contribution to indoor concentrations
is the following:
Outcontin = a + p~uul*PM25out + /3*x,
(3-16)
"Outcontin'' is the contribution of outdoor sources to indoor
PM2 5 levels determined by the indoor/outdoor sulfur ratio.
Results (Table 3-25) show that air exchange (airex, windopen,
windowall) again significantly influences the results, this time
increasing the outdoor contribution as their values increase. The
additional variables increase R2 to 0.82, compared with an R2 of
0.69 for outdoor PM alone. The fact that air exchange variables
appear as highly significant in these two regressions is important
confirmation of our assumptions in using sulfur to identify
indoor-generated and outdoor-generated PM. That is, if the
assumption of no indoor sulfur source were violated in some
homes, we would not be able to separate indoor-generated PM
from outdoor-generated PM in those homes, and the relationship
with air exchange would be weakened by the amount of the
misclassification. Air exchange variables are not selected as
significant in regressions of total indoor PM, which lumps both
indoor-generated and outdoor-generated PM, because the
influence of air exchange can work in two directions depending
on whether indoor air PM is greater than or less than outdoor
PM. But increased air exchange must increase outdoor-generated
PM and it must decrease indoor-generated PM; therefore our
finding of a strong effect in the expected directions when we
analyze the indoor- and outdoor-generated concentrations
separately must reflect a successful separation of the two
sources.
A model for indoor sulfur concentration is the following:
SH, = a+pou*Sottl+p*xl (3-17)
Once again we first examine whether the outdoor concentration
shows artifactual dependence on household variables by looking
at outdoor sulfur (S,M) alone (Table 3-26). Several artifacts
appearedair exchange, storm windows, time spent outdoors.
However, these variables increased R2 from 0.59 using only
PM2.5 outdoors as the independent variable, to 0.65 using all
significant variables, a relatively small increase.
The results for indoor sulfur (Sm) suggest that there may be more
sources of indoor sulfur than we have assumed above. For
example, the number of pilot lights (pilot) was a strongly
significant variable, adding 133 ng/m3 to the indoor sulfur
concentration per each additional pilot light. Natural gas
sometimes contains some sulfur, although if the level exceeds a
certain amount it must be scrubbed. Smoking was also
significant (cigsmokcf), possibly due to the use of matches. The
effect of open windows was also clear and expected. Homes
with at least one open window (windopen) had increased indoor
sulfur concentrations of 120 ng/m3. A quantitative measure
(windowalf) was an increase of 0.32 ng/m3 for every inch-hour a
window was open. Thus a window opened 6 inches wide for a
day would result in an increase of 48 ng/m3 S. Building age
(AGE) was shown to be artifactually associated with outdoor
PM; therefore, its appearance in this multiple regression may
also be artifactual. Adding these variables to the simple
regression of indoor on outdoor sulfur increased the R: from 0.70
to 0.83.
We can estimate personal exposure from (1) outdoor
concentrations alone; (2) indoor concentrations alone; (3)
outdoor and indoor concentrations together; (4) outdoor
concentrations and indoor concentrations together with
household characteristics; and (5) outdoor concentrations alone
together with household characteristics. These choices result in
the following models of personal exposure:
PM25pers =a+ ftoul*PM25out
PM25pers = a+ pm,*FRM25
PM25pers = a +/?,* PM25'm
(3-18)
(3-19)
(3-20)
PM25pers = a + ftmi*PM25ont +p,n*PM25in (3-21)
PM25pers = a + P
-------
Table 3-26. Dependence of Outdoor Sulfur on Outdoor PM2 5 and of Indoor Sulfur on Outdoor Sulfur and Household Characteristics and Personal
Activities
Variable
Soul
Intercept
PM25out
Souf
Intercept
PM25out
airex
S_WIN
outdoor
numsmok
A_C
Sln
Intercept
S0uf
Sm
Intercept
Soul
AGE
PILOT
windowall
TempC
Windopen
clgsmokd
FAN
A_C
FLRCOVav
Numpeopl
fry
Unknown
DIRT_RD
vacuum
Slope
47.09
97.14
594.47
96.98
-354.67
-216.48
-1.50
-209.72
-40.72
151.44
0.49
-19.76
0.51
7.91
133.01
0.32
-9.74
120.11
28.21
-140.52
33.49
-1.98
-20 14
-60.97
1.21
81.66
61.71
Std.Err.
66.44
3.01
85.83
2.85
43.81
54.80
0.41
82.71
18.05
27.27
0.01
48.56
0.01
0.73
13.43
0.05
1.59
23.85
5.66
30.52
7.29
0.51
6.13
20.94
0.47
32.60
25.90
p-level
0.478737
0.000000
0.000000
0.000000
0.000000
0.000086
0.000240
0.011441
0.024394
0.000000
0.000000
0.684261
0.000000
0.000000
0.000000
0.000000
0.000000
0.000001
0.000001
0.000005
0.000005
0.000104
0.001074
0.003705
0.009623
0.012471
0.017422
N R2(adj.)
720 0.59
720 0.65
720 0.70
741 0.83
46
-------
Table 3-27. Regressions of Personal Exposures to PM2 5 on Household Characteristics and Personal Activities
Variable
PM2Spers
Intercept
PM25out
PM25pers
Intercept
FRM25
PM25pers
Intercept
PM25in
PM25pers
Intercept
PM25in
PM25out
PM25pers
Intercept
PM25in
PM25out
grill
cooking
PM25pers
Intercept
numsmok
PM25out
CANDLDUR
Burning
C_FUEL
DIRT_RD
Other_indoor
A_C
Exhstfan
Cleaning_1
cooking
outdoor
PM25pers
Intercept
FRM25
numsmok
Slope
10.78
0.66
10.44
0.73
9.98
0.67
5.66
0.63
0.27
4.32
0.60
0.26
11.54
0.02
-11.15
13.76
0.55
0.05
11.19
7.42
6.05
0.11
1.32
4.29
3.46
0.02
-0.02
-11.81
0.60
13.31
Std.Err.
1.56
0.07
1.53
0.08
0.69
0.03
1.12
0.03
0.05
1.21
0.03
0.05
3.22
0.01
3.63
1.54
0.06
0.01
2.23
1.56
1.69
0.03
0.40
1.36
1.11
0.01
0.01
3.69
0.07
1.61
p-level N R2(adj.)
750 0.10
0.000000
0.000000
734 0.11
0.000000
0.000000
0.000000 727 0.45
0.000000
727 0.47
0.000001
0.000000
0.000001
727 0.49
0.000370
0.000000
0.000001
0.000359
0.003309
737 0.37
0.002239
0.000000
0.000000
0.000000
0.000001
0.000003
0.000363
0.000800
0.001039
0.001652
0.001968
0.009918
0.013482
721 0.39
0.001449
0.000000
0.000000
47
-------
Table 3-27. Continued
Variable
Slope
Std.Err.
p-level
R(adj.;
CANDLDUR
Burning
C_FUEL
A_C
Other_indoor
Exhstfan
DIRT_RD
Cleaning_1
cooking
outdoor
0.05
11.71
7.44
1.42
0.11
4.49
5.45
3.52
0.02
-0.02
0.01
2.25
1.59
0.41
0.03
1.37
1.70
1.13
0.01
0.01
0.000000
0.000000
0.000003
0.000492
0.000553
0.001106
0.001407
0.001927
0.010664
0.015074
48
-------
When only outdoor PM: 5 and the household characteristics were
included in the regression, the most significant variable other
than outdoor PM2.s was the number of smokers, adding between
13.3 and 13.8 ug/m3 to the daily average personal exposure. The
use of candles was highly significant, adding 0.05 ug/m per
minute burned to daily average personal exposure. Burned food
added between 11.2 and 11.7 ug/m3 to personal exposure.
Electric stoves increased particle concentrations by 7.4 ug/m3.
Although these household characteristics and personal activities
were useful in improving the R2 for personal exposure,
collectively they were not able to improve it as much as simply
adding in the measured indoor PM25 concentration (R" = 0.37-
0.39 for the outdoor + household characteristics variables
compared to 0.47 for the outdoor + indoor concentrations).
By assuming that no sulfur is created indoors, we can estimate
the contribution of personal activities to personal exposure
(Perscontrib) by multiplying the outdoor PM2 5 by the ratio of
the personal sulfur concentration to the outdoor sulfur
concentration, and subtracting that product from the observed
personal exposure. As with the similarly defined indoor-
generated particle contribution to total indoor concentrations, we
can regress Perscontrib on the household characteristics and
personal activities variables without including the outdoor
concentration as an input, since we used it to determine the
values of Perscontrib. Again, we checked to confirm that, when
included, the outdoor concentration was not significant. The
model is
Perscontrib = a + ft, *x,
(3-25)
The focus on the non-outdoor part of personal exposure again
showed that air exchange was an important variable, tending to
decrease the personal contribution by 4.1 ug/m3 per unit change
in the air exchange variable (Table 3-28). Smoking, burned
food, and cooking fuel variables significantly influence personal
exposure concentration. Candle burning increased personal
exposure by 0.05 ug/m3 per minute burned.
The factors affecting the outdoor contribution to PM2.5 personal
exposure were investigated using multiple regression (Table 3-
29). Here the model is
Outcontribpers = a + /3oul*pm25out + fi,*x,
(3-26)
The most influential factor, as expected, was the outdoor PM2.5
concentration. This is so much stronger than all the rest of the
factors put together (t-value of 45 compared to values <7 for the
other variables) that it can be considered the single variable
driving personal exposure to particles of outdoor origin. The
next strongest factor (AGE) may be an artifact due to the
unbalanced design of the study, in which not all homes were
monitored on the same day. The next three factors have to do
with air exchange, which depends on window opening behavior
(windopen, windowall) and on indoor-outdoor temperature
differences (roughly approximated by TempC). The increased air
exchange rates due to increases in these variables brings more
outdoor particles indoors, and thus affects personal exposure to
the extent the person is at home. Note that the R2 of 0.82 for this
model of the influence of outdoor particles on personal exposure
precisely matches that for the influence of outdoor particles on
indoor particle concentrations (last part of Table 3-25).
Finally, personal exposure to sulfur was regressed against
outdoor sulfur and the household characteristics, using the
following set of models:
Spers = a + fi,n*Sin
Spers = a+ fioll,*Sout
Spers = a + /?, *Soitt
(3-27)
(3-28)
(3-29)
Spers = a + ftuu,*Sout + pn*Sin + b*x, (3-30)
Spers = a+pol,,*Sout + P*x, (3-31)
These correspond to having greater or less information about the
outdoor and indoor sulfur concentrations and the questionnaire
variables.
In general, very high values of R2 are achieved (Table 3-30).
The worst case is having only the outdoor sulfur concentration,
and that alone explains 0.78 of the variance in personal exposure
to sulfur. If the indoor sulfur were known alone, the R" could be
improved to 0.88. Using both measured indoor and outdoor
concentrations to estimate personal exposure increases R" to
0.91. Other influential variables included time spent near a
smoker and the number of pilot lights in the house, both possible
sources of sulfur as mentioned above.
Variables Affecting Air Exchange and the
Infiltration Factor
Our primary focus has been on documenting the influence of
outdoor fine particles on indoor concentrations and personal
exposures. Two of the most influential factors are air exchange
and the infiltration factor.
Variables Affecting Air Exchange
Previous work (Howard-Reed et al., 2002; Wallace et al., 2002)
identified window opening and the absolute indoor-outdoor
temperature difference as two main variables affecting air
exchange. To create a variable approximating the absolute
indoor-outdoor temperature difference, we took the absolute
value of the difference between the outdoor temperature
(TempC) and a typical indoor temperature of 72 F (22.2 C). The
new variable was called AbsTempDif.
49
-------
Table 3-28. Regression of the Non-ambient-related Contribution (Perscontrib) to Personal PM2 5 Exposure
Variable
Perscontrib
Intercept
numsmok
CANDLDUR
C_FUEL
Burning
airex
Cleaning_1
DIRT_RD
Exhstfan
outdoor
A_C
Other indoor
cooking
Slope
-9.68
11.45
0.05
8.44
10.83
-4.12
4.26
5.85
4.61
-0.03
1.31
0.09
0.02
Std.Err.
3.55
1.57
0.01
1.62
2.25
0.97
1.14
1.69
1.38
0.01
0.41
0.03
0.01
p-level N R2(adj.)
677 0.30
0.006489
0.000000
0.000000
0.000000
0.000002
0.000024
0.000194
0.000569
0.000850
0.001324
0.001585
0.005808
0.022302
50
-------
Table 3-29. Regression of the Ambient-Related Contribution to Personal PM2 5 Exposure
Intercept
PM25out
AGE
windowall
Windopen
TempC
cigsmokd
MILDEWavg
smoke
cooking
A_C
DIRT_RD
Travel
Pets
dust
FLRCOVav
Grooming
AREA
B
-0.952
0.521
0.044
0.003
1.078
-0.062
0.221
1.113
0.009
0.005
0.234
0.959
0.004
0.754
-0.688
-0.012
-0.006
-0.001
SE
0.588
0.012
0.007
0.000
0.211
0.012
0.056
0.296
0.002
0.001
0.068
0.295
0.002
0.298
0.271
0.004
0.002
0.000
t(691)
-1.6
45.2
6.7
6.4
5.1
-5.0
3.9
3.8
3.6
3.6
3.5
3.3
2.9
2.5
-2.5
-2.8
-3.0
-3.2
p-level N R2(adj.)
0.105780 710 0.82
0.000000
0.000000
0.000000
0.000000
0.000001
0.000101
0.000187
0.000288
0.000311
0.000593
0.001196
0.004144
0.011484
0.011302
0.005730
0.002672
0.001409
51
-------
Table 3-30. Regressions of Personal Exposure to Sulfur on Indoor and Outdoor Concentrations and Questionnaire Variables
Variable
Spers
Intercept
Sin
Spers
Intercept
Sin
Sout
Spers
Intercept
Sout
Spers
Intercept
Sin
Sout
DIRT_RD
smoke
windowall
MILDEWavg
C_FUEL
Grooming
outdoor
Otherjoc
TempC
Spers
Intercept
Sout
PILOT
windowall
AGE
AREA
smoke
DIRT_RD
FLRCOVav
Windopen
TempC
MILDEWavg
Travel
BUSY_RD
FAN
cooking
cigsmokd
sweep
Slope
42.50
0.91
-12.71
0.66
0.17
89.52
0.49
-1.59
0.65
0.17
117.68
0.87
0 14
81 04
-55.33
-0.45
0.34
0.22
2.46
28.25
0.50
123.65
036
3.63
-0.10
1.03
111 72
-1.62
72.43
-459
92.35
0.47
70.43
-76.58
0.33
1398
^5.29
Std.Err.
15.67
0.01
14.32
0.02
0.01
21.54
0.01
31.71
0.02
0.01
18.99
0.16
0.03
20.08
15.33
0.14
0.11
0.08
1.07
58.70
0.01
13.27
0.05
074
0.02
0.26
31.23
0.48
22.40
1 51
30.90
0.16
24.83
29.86
0.14
601
21.61
p-level N R2(adj.)
727 0.88
0.006860
0.000000
727 0.91
0.374879
0.000000
0.000000
750 0.78
0.000036
0.000000
722' 0.93
0.959998
0.000000
0.000000
0 000000
0.000000
0.000006
0 000060
0.000327
0.001283
0.002197
0.010413
0.022455
715 0.85
0.630470
0 000000
0.000000
0.000000
0.000001
0.000040
0.000105
0.000372
0.000862
0.001280
0.002480
0.002903
0 003808
0 004686
0.010533
0.017800
0.020369
0.036424
-------
Because we saw increased air exchange rates in summer, we
considered using Season as a separate independent variable. The
reasoning here is that perhaps people would tend to keep the
windows closed and the air conditioner running in Summer even
on a relatively cool day, whereas on a day with the identical
temperature in another season they might not bother turning on
the air conditioner. Thus, both temperature and season could
have separate effects. First, Season was coded by average
temperature, using a 3-point scale (Winter = 1, Fall and Spring =
2, Summer= 3); however, a regression including either season or
TempC showed that TempC produced slightly higher R". It was
then thought that a 3-point scale might be too restrictive, so a 5-
point scale was created by month (e.g., Jan-Feb = 1, July-August
=5). There were only 10 months, with March and December
missing. The new variable was called MonthbyTemp. This time
when the regression was run putting in either TempC or
MonthbvTemp, the R~ using the MonthbyTemp variable was
slightly higher than that using TempC (Table 3-31).
The two strongest variables are in fact those that we already
know from previous studies: the window opening width and the
absolute indoor-outdoor temperature difference. The latter has a
coefficient of 0.04 ach/°C, which agrees well with the estimate of
0.02 ach/°C in Wallace et al. (2002). The next strongest variable
is the number of persons in the home. This variable, as well as
Ihepets variable, was included on the questionnaire because of
results from past studies (e.g., Thatcher and Layton, 1995)
indicating that people and pets can increase particle levels due to
resuspension and going outdoors more often, which will bring in
outdoor particles when the door is open. The pets variable also
appears, although just missing significance. The next strongest
variable is age of the homeprevious studies have noted that
older homes are constructed more loosely. The use of the
exhaust fan tends to increase air exchange, as noted in a previous
study (Wallace et al., 2002). Another variable is the
MonthbyTemp variable, reflecting the increased use of air
conditioners (and therefore closed windows) in the summer. The
presence of a clothes dryer (nearly all of which were vented
outdoors) increases air exchange since almost the same volume
of air must enter the house to replace the heated air vented
outdoors. We have no explanation for the appearance here of the
VAC and FLRCOVav variables, although their effect on the total
R" value of the model is very small.
Variables Affecting the Infiltration Factor
The multiple regression on the indoor/outdoor sulfur ratio was
run on all variables, including airex (Table 3-32). However, the
five strongest variables contributing to this ratio can all be seen
to be variables already contributing to the air exchange rate.
Therefore these variables are, in a sense, being double-counted.
They are obscuring and weakening the actual relationship with
the air exchange rate. Therefore the regression was run again
including all variables except those already found to contribute
to the air exchange rate (Table 3-33). The final model shows
that the air exchange rate (together with the windows open
variable) is the strongest variable affecting the sulfur
indoor/outdoor ratio. The air exchange rate is the only one of our
variables that explicitly appears in the equation for the
infiltration factor: (Pa/(a+k)). The three remaining variables
provide very little contribution (about 3%) to the total R2 of the
model. One of these, A/C, appears to have the wrong sign, since
presence of an air conditioner would be expected to reduce the
infiltration factor. However, this variable ranges from 1 to 5 air
conditioners and therefore is measuring the falloff in efficiency
as one goes from central air conditioning to multiple window air
conditioning.
53
-------
Table 3-31. Variables Affecting Air Exchange Rate
Variable
a/rex
Intercept
windowall
ABSTempDIF
Numpeopl
AGE
VAC
FLRCOVav
DRYER
MILDEWavg
Monthbytemp
Mealsckd
Exhstfan
Unknown
Pets
Slope Std.Err.
0.5156 0.1355
0.0010 0.0001
0.0434 0.0040
0.0693 0.0094
0.0079 0.0013
-0.3079 0.0496
-0.0043 0.0008
0.2071 0.0458
-0.2406 0.0544
-0.0673 0.0153
-0.1488 0.0390
0.1375 0.0410
0.0025 0.0008
0.1586 0.0523
t(711) p-level N R2(adj.)
744 0.51
3.8 0.00015
13.2 0.000000000000
10.8 0.000000000000
7.3 0.000000000001
6.3 0.00000000062
-6.2 0.00000000093
-5.1 0.00000044
4.5 0.0000073
-4.4 0.000011
-4.4 0.000012
-3.8 0.00014
3.4 0.00085
3.3 0.0012
3.0 0.0025
TableS- 32. Variables Affecting Indoor/Outdoor Sulfur Ratio
Variable
Sinout
Intercept
Monthbytemp
AGE
airex
Windopen
windowall
C_FUEL
DIRT_RD
VAC
DUSTY_RD
Slope Std.Err.
0.68439 0.02916
-0.03656 0.00343
0.00299 0.00037
0.06455 0.00925
0.06168 0.01027
0.00014 0.00002
-0.06117 0.01047
0.06170 0.01362
-0.05484 0.01323
-0.04057 0.01203
Table 3-33. Variables Affecting Indoor/Outdoor Sulfur Ratio:
Variable
Sinout
Intercept
airex
Windopen
A_C
numsmok
BUSY_RD
Slope Std.Err. t(699)
0.409 0.010 40.3
0.121 0.008 15.1
0.090 0.010 9.2
0.015 0.003 4.3
0.055 0.015 3.7
0.036 0.011 3.2
t(638) p-level N R2 (adj.)
659 0.46
23.5 0.00000000000000
-10.7 0.00000000000000
8.1 0.00000000000000
7.0 0.000000000008
6.0 0.0000000031
5.9 0.0000000062
-5.8 0.0000000083
4.5 0.0000070
-4.1 0.000039
-3.4 0.00079
Reduced Model
p-level N R2 (adj.)
704 0.43
0.000000000000
0.000000000000
0.000000000000
0.000018
0.00021
0.0013
54
-------
Chapter 4
Discussion
To reach our primary goal of estimating the contribution of
outdoor PMi.s to personal exposures, we set an intermediate goal
of estimating the contribution of outdoor particle concentrations
to indoor particle concentrations. Since personal exposures
depend heavily on indoor concentrations, this may be a good first
approximation. We calculated the contribution of outdoor
particles by multiplying the outdoor concentration by the ratio of
the indoor to outdoor sulfur concentrations:
C =
*- mu '
1 C
olil *- Mill
(4-1)
where Cmo is the concentration indoors of particles infiltrating
from outdoors, and where the S,n/So:i, term is an estimate of Fmj.
We also calculated Fmt using a combination of the
indoor/outdoor sulfur ratios and the air exchange measurements
and found excellent agreement (R2 = 0.96-0.99), indicating that
the estimates of Fml for individual homes were quite stable. This
is one of the first studies to calculate infiltration factors and
outdoor exposure factors for individual homes and persons,
respectively, and is also one of the first to carry out the
regressions of that portion of personal exposure due to particles
of outdoor origin on outdoor concentrations.
Limitations of the Regressions and
Influence of Assumptions
When we run a regression of , on C,,,,,, we are in fact
regressing a term on a portion of itself:
p ,r = A C +
1 inj *- out ** *- out
(4-2)
The infiltration factor does not have a high degree of day-to-day
variability: Fm, = 0.59 (0.16 SD) for all homes, and an even
smaller standard deviation within homes. Therefore our
regression is not very different from regressing a term on a
constant fraction of itself, which will result in an R2 of 1. This is
a result that is forced by our basic assumptions no indoor
sources of sulfur, no coagulation, condensation, particle-gas
conversion, or indoor chemistry, negligible transient terms,
instantaneous perfect mixing, penetration and deposition
characteristics of PMi.s identical to those of sulfur - leading to a
simple linear relation between the outdoor concentration and the
fraction of outdoor air particles infiltrating the house. The
regression on outdoor air measured at all the homes resulted in
an R2 of 0.71, a value that is certainly higher than reality due to
our assumptions. The individual regressions on 36 homes gave
even higher values of R" (median 0.77). When the regressions
were run on the outdoor PMi.j concentrations measured by the
FRM at the central site, the overall R2 value was lowered to 0.60
(median 0.73), but with the same caveat that our assumptions
force a high correlation and therefore these estimates are not be
considered best estimates, but rather upper bounds.
A regression of personal PM2 5 on outdoor PM: 5 concentrations
provides an estimate of F^ of 0.59 + 0.01 SE, with one
influential outlier removed. With the outlier included, the
estimate was 0.64 + 0.02 SE. The intercept with the outlier
excluded (12.3 + 1.4 ug/m3) is an estimate of £, the exposure
due to non-outdoor sources. The difference between the indoor
source contributions estimated at 7.7 ug/m3 in Table 3-5 and the
non-outdoor source contribution is sometimes attributed to the
"personal cloud," which in this case equals 4.6 + 2.0 ug/m3.
The similarity of the slopes for the personal vs. outdoor and
indoor vs. outdoor PMi, regressions (0.59 compared to 0.60)
suggests that the time spent indoors drives the relationship
between personal exposure and outdoor concentrations. That is,
the infiltration factor F,,,/, which governs the reduction of particle
concentrations as they enter a house, is very similar to the
outdoor exposure factor Fpcx, which governs the reduction in
outdoor concentrations contributing to personal exposure.
A major assumption has been that fine particles in general will
behave like sulfur in terms of penetration and deposition. How
good is this assumption? Sulfur is found primarily in the fine
fraction, and within that fraction it typically has diameters <0.5
55
-------
(.an, smaller than most other elements. From theoretical and
experimental studies, deposition velocities appear to reach a
minimum around diameters of 0.1-0.2 um (Lai and Nazaroft,
2000). Deposition rates for sulfur and other elements were
estimated in the PTEAM study (Ozkaynak et al, 1996). The rate
for sulfur was 0.16 h"' (0.02 SE). Rates for other elements were
generally higher (e.g., iron, 0.70 h"1 (0.30 SE)). Therefore sulfur
will stay elevated indoors longer than most other elements, and
possibly longer than most of the remainder of the fine particle
mass. This leads to overestimates of the contribution of outdoor
particles to indoor PM: 5. Evidence for the amount of the
overestimate is provided by the results for iron concentrations.
The indoor/outdoor ratio for iron was 0.38 (0.18 SD; N = 766).
This was the lowest ratio of all the elements. For Si it was 0.58
(N = 766); for Mn it was 0.52 (N = 409). These were elements
that appeared to have few indoor sources, judging from their
high indoor-outdoor product-moment correlations (0.81 for Fe,
0.74 for Si, N = 762 samples for each; 0.58 for Mn, N = 460
samples; these correlations were almost as high as the value of
0.85 for sulfur). For the other elements, the indoor/outdoor
correlations were lower and the indoor/outdoor ratios higher due
to indoor sources, and cannot be used to help estimate the
amount of the overestimate due to using the sulfur ratios. Iron
itself makes up only a small fraction of PM: 5 mass, but it could
be a marker for the behavior of a larger fraction.
Sulfur accounted for 20% of indoor PM: 5, assuming it is in the
form of ammonium sulfate. If the remaining 80° o of the indoor
particle mass behaved more like iron, outdoor air particles would
account for only 42% of indoor air PM; 5, rather than 58% as was
found using sulfur alone. However, it is unlikely that the
remainder of the indoor air particles behave more like iron,
because the RCS model regression of indoor PM: 5 on outdoor
PM: j (Figure 3-13) resulted in a slope very close to the slope
predicted by the sulfur indoor-outdoor ratio. These
considerations indicate that the overestimate of the outdoor
contribution due to relying on the sulfur ratios is not a major one
and does not affect the relative order of the homes in terms of the
fraction of indoor PM: 5 produced by outdoor particles.
Based on previous studies, we attempted to use the particle
measurements to estimate an infiltration factor for each home;
however, we found poor agreement with the estimates in this
study using the sulfur ratio. Most other studies of personal
exposure and/or indoor concentrations of fine particles have
fewer measurements per home than this study. Therefore our
results suggest that particle mass measurements alone cannot
provide reliable estimates of /",,.
We estimated the outdoor exposure factor FfH., using two
methods: personal, outdoor sulfur ratios and a model using time
indoors and outdoors. The model consistently overestimated the
outdoor factor. We speculate that this is because people spend
time in unmonitored environments, and our assumption that
exposures are the same in those environments as in the
monitored home environment may be incorrect. We found that
estimating personal exposure using only the indoor/outdoor
sulfur ratio (F,n/) gave better results than estimates using the
modeled outdoor exposure factor F/KY based on time spent
indoors and outdoors.
Although several different methods were employed in making
these estimates, the methods ultimately depended on the sulfur
measurements only. Because of the way P and k are related in
the equation for F,,,h an infinite number of solutions are available
for any given infiltration factor. The solution surface for P and k
is very flat (meaning a very wide error associated with all point
estimates); therefore a slight error in any measurement can lead
to a very large error in the estimation of P and k. For example,
each of the slightly different methods produced four or five
physically impossible estimates for P Although bounds can be
established to prevent P from exceeding unity, the existence of
these nonphysical solutions suggests that measurement errors or
violations of our assumptions may have caused large errors in
estimating these parameters. Since we have no independent
methods of estimating these parameters, we are unable to
validate our estimates of P and k for individual homes. The
estimates of P and k that were significantly different from zero
included about'24 of the 36 homes. The median value for the
penetration coefficient P was 0.81 (interquartile range 0.66 to
0.90). The median for the deposition rate k was 0.24 h'1
(interquartile range 0.12 to 0.35 h"'). These estimates arc similar
to those obtained in some other studies (Liu et al., 2003; Wallace
et al., 2002; Thatcher et al., 2003).
Because not all homes were done in the same time periods, there
were certain unavoidable differences in the outdoor particle
concentrations encountered at the time they were monitored. We
attempted to identify possible artifacts (significant associations
with no possible causal explanations) by regressing central-site
and residential outdoor concentrations vs. the questionnaire
variables. We did find some artifacts, and therefore caution that
some apparently significant relationships may be due to these
temporal variations rather than a true causal relationship. This
may be particularly true for certain unvarying household
characteristics such as building age, location near a road, and
type of cooking fuel (gas vs. electric) because these unvarying
characteristics enter the regressions as a block of entries repeated
up to 28 times per house. Therefore their number of degrees of
freedom is greatly reduced and any variation due to temporal
heterogeneity will be multiplied by this repeated appearance.
This caution extends to the variables AGE, DIRT ROAD.
C_FUEL (cooking fuel). S_WIN (storm windows). VAC. and
AREA among others.
Nonetheless, regressions of indoor concentrations and personal
exposures showed greatly improved R: results when
questionnaire variables were added to the outdoor
concentrations. The improvement was from R: about 0.1 to R:
about 0.4, for both indoor and personal estimates. Among the
56
-------
significant contributors to indoor and personal exposure were
smokers in the household, cooking, and number of people in a
household, all variables that have previously been found to
contribute to indoor PM concentrations (Ozkaynak et al, 1996b;
Wallace et al., 2004b). Some variables not previously identified
as contributing to indoor or personal exposures were burning
food, duration of candle use, number of pilot lights, use of an
exhaust fan, proximity to a dirt road, and presence of electric
stoves (compared to gas stoves). The latter two variables are to
be treated with caution since they are unvarying household
characteristics and therefore may have been subject to temporal
unevenness of sampling. Vacuuming was another variable that
achieved significance at times. The presence of a clothes dryer
was significantly associated with reduced indoor concentrations
and personal exposure.
An important result was the appearance of the air exchange
variables (including windows open or closed) in the regressions
on estimated concentrations of indoor-generated and outdoor-
generated particles. The fact that these variables appeared in the
expected directions whereas they did not appear when regressing
on the mixture of both indoor-generated and outdoor-generated
particles is confirmation that our division into the two
components using the sulfur ratios was successful.
We found several variables that were important in affecting air
exchange rates. Opening windows was the single most important
variable, as has been shown in previous studies (Howard-Reed et
al., 2002). The indoor-outdoor temperature difference (which
affects the indoor-outdoor pressure difference and therefore the
driving force for air exchange) was the next most powerful
variable. The coefficient for this variable (0.04 ach/°C) was of
similar magnitude to that found by Wallace et al. (2002). The
number of people in the home and the presence of pets, both
significant variables in this study but not often found in other
studies, contribute to increased air exchange through the
increased number of times going in and out of the house. The use
of a kitchen exhaust fan was associated with an increase of 0.14
ach. This also compares well with the finding of a 0.8 ach
increase associated with use of an attic exhaust fan (Wallace et
al., 2002), given the much longer periods that an attic exhaust
fan may run compared to a kitchen exhaust fan.
The single strongest variable affecting the infiltration factor F,nf
is the air exchange rate, as can be seen from the equation for Fm]
and the observed relationship shown in Figure 3-27. All other
variables entering our analysis are either highly correlated with
air exchange rate or appear to be artifacts arising from the
unequal monitoring of homes on different days.
57
-------
Chapter 5
Conclusions
The sulfur measurements showed excellent agreement between
the personal samplers and the fixed indoor-outdoor Harvard
Impactors. More than 160 co-located measurements both
indoors and outdoors resulted in slopes insignificantly different
from 1 and intercepts insignificantly different from zero, with R"
values of 0.97 for both indoor and outdoor locations.
Uncertainties of individual sulfur measurements were estimated
at 8%.
The data appear to be consistent with the hypothesis that there
are few indoor sources of sulfur. Only about 1% of
measurements showed higher levels indoors than out. Although
nearly all regressions of indoor sulfur vs. outdoor sulfur gave
positive intercepts, these intercepts were often relatively small
and could be due to measurement error. Therefore we accepted
the indoor/outdoor sulfur ratio as our best estimate of Fml for
PM:.S.
Estimates of F,nl averaged over all seasons varied over a large
range by household (0.26-0.87), but half of the homes were
within 20% of the median value of 0.59. The infiltration factor
was significantly lower for many homes in the summer season,
presumably due to the closed windows and increased
recirculation and filtration of air associated with the use of air
conditioners. Evidence from measurements of iron and silicon
suggested that the estimates of the infiltration factor using the
sulfur indoor/outdoor ratio are likely to overestimate the
influence of outdoor air due to the low deposition velocity of
sulfur compared to other likely constituents of PM: 5. However,
other considerations lead us to think that the overestimate is not
large; also, it would not affect the relative ranking of the homes
with respect to the outdoor contributions to indoor PM: 5.
In general, the outdoor exposure factor FpL,x was very similar to
F!nf. This is expected, since persons spend most of their time
indoors, but the very close agreement is also an indication that
the sulfur measurements were reproducible and agreed well even
though the personal monitor collected 10 times less material than
the indoor/outdoor monitors.
Fpn was usually smaller by a few percent than Fm(, although a
simplistic application of time-activity budgets indicates that it
should be larger by about 6 or 7% for persons spending a typical
amount of time outdoors and in vehicles. This may be due to
time spent in unmonitored indoor locations (e.g., office
buildings, department stores) that have mechanical ventilation,
recirculation, and filtration, thus lowering exposure to sulfur
while in those locations. The fact that /), was slightly larger
than F,,,/ during the summer season, when many homes were
closed and using recirculated and filtered air, supports this
hypothesis. We conclude that the model advocated in earlier
publications for determining /*,, by using time-activity budgets
is not useful for studies that do not measure indoor
concentrations in schools, workplaces, shopping malls, and other
locations where persons spend substantial amounts of time. The
preferred way to estimate Fpcx is by using personal sulfur
measurements, but lacking those, it may be just as useful to use
Fm/ as the best estimate of Fpcx.
Regressions of indoor air concentrations due to particles of
outdoor origin vs. outdoor concentrations had generally high R:
values, as did regressions of personal exposure to particles of
outdoor origin vs. outdoor concentrations. However, this result
is partially due to the many assumptions in our approach leading
to a simple linear relation between indoor concentrations of
particles of outdoor origin and outdoor concentrations.
Therefore these estimates should not be treated as best estimates,
but rather as upper bounds to the actual amount of variance in
personal exposure to PM:5 of outdoor origin that can be
explained by outdoor measurements at the central site.
We investigated whether particle measurements alone could be
used to estimate the infiltration factor. However, comparisons
with the sulfur ratio method indicated that particle measurements
58
-------
alone cannot be used to give reliable estimates of Finf.
The two unmeasured parameters contributing to Fln/ (P and k)
were estimated using both linear and nonlinear approaches. The
median value for the penetration coefficient P was 0.81
(interquartile range 0.66 to 0.90). The median for the deposition
rate k was 0.24 h'1 (interquartile range 0.12 to 0.35 h"1). Recall
that these values of k and P are for particles in the same size
range as sulfur particles. However, we conclude that the
unphysical values obtained for P in some cases casts doubt on all
the estimates of P and k for individual homes; we are unable to
validate these estimates.
Air exchange rates were found to depend primarily on opening
windows and on the absolute indoor-outdoor temperature
difference. Other contributors included use of a kitchen exhaust
fan, presence of a vented clothes dryer, and number of persons in
a household. The infiltration factor was primarily dependent on
air exchange, as expected from the basic equilibrium equation.
Regressions of indoor concentrations and personal exposures
showed considerably improved R2 estimates by consideration of
questionnaire variables. In particular, smoking, cooking, number
of persons in a household, and burned food were important
contributors. Air exchange rates were also very important
variables, but only after the total indoor particle concentration
had been split into indoor-generated and outdoor-generated
portions. The strong effect of air exchange (increasing outdoor-
generated particle concentrations and decreasing indoor-
generated particle concentrations) was an important confirmation
of the success of our efforts to resolve these two contributors to
total indoor particle levels.
59
-------
Chapter 6
References
Abt, E.. Suh, H.H., Allen, G. and Koutrakis, P. (2000a).
Characterization of indoor particle sources: A study conducted
in the metropolitan Boston area. Environmental Health
Perspectives 108( 1 ):35-44.
Abt, E., Suh, H.H., Catalano, P.J. and Koutrakis, P. (2000b).
Relative contribution of outdoor and indoor particle sources to
indoor concentrations. Environmental Science ami Technology
34:3579-3587.
Allen R., Larson, T., Sheppard, L., Wallace, L. and Liu, L-J S.
(2003). Use of Real-time Light Scattering Data to Estimate
the Contribution of Infiltrated and Indoor-Generated Particles
to Indoor Air. Environmental Science and Technology
37:34X4-3492.
Allen R., Larson, T., Sheppard, L.. Wallace, L. and Liu, L-J S.
(2004). Estimated hourly personal exposures to ambient and
non-ambient paniculate matter among sensitive populations in
Seattle, Washington. Journal of the Air and Waste
Management Association 54:1197-1211.
AQCD (2003). Fourth External Review Draft of Air Quality
Criteria for Paniculate Matter (June, 2003). National Center
for Environmental Assessment, U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Belsley, D.A., Kuh. E., and Welsch, R.E., Regression
Diagnostics, John Wilev & Sons, 1980
Clayton. C.A., Perritt. R.L., Pellizzari, E.D., Thomas, K.W.,
Whitmorc, R.W., Wallace. L.A., Ozkaynak, H.. and Spengler,
J.D. (1993). Panicle Total Exposure Assessment
Methodology (PTEAM) study: distributions of aerosol and
elemental concentrations in personal, indoor, and outdoor air
samples in a southern California community. Journal of
Exposure Analysis and Environmental Epidemiology 3:227-
250.
Ebelt, ST., Petkau, A.J., Vedal, S., Fisher, T.V., and Brauer, M.
(2000). Exposure of chronic obstructive pulmonary disease
patients to paniculate matter: relationship between personal
and ambient air concentrations. Journal of the Air and Waste
Management Association 50:1081 -1094.
Evans, G.F., Highsmith, R.V., Sheldon, L.S., Suggs, J.C.,
Williams, R.W.. Zweidmger, R.B., Creason, J.P., Walsh, D..
Rodes, C.E. and Lawless^.A. (2000). The 1999 Fresno
paniculate matter exposure studies: comparison of community,
outdoor, and residential PM mass measurements. Journal of
the Air and Waste Management Association 50( 11): 1887-
1896.
Highsmith. V.R., Rodes, C.E. and Hardy. R.J. (1988). Indoor
panicle concentrations associated with use of tap water in
portable humidifiers. Environ Sci Tech 22:1109-1112.
Hopke PK, Ramadan Z, Paatero P, Norris GA, Landis ML,
Williams RW, and Lewis CW (2003). Receptor modeling of
ambient and personal exposure samples: 1998 Baltimore
Paniculate Matter Epidemiology-Exposure Study.
Atmospheric Environment 37(23):3289-3302.
60
-------
Howard-Reed, C. Rea, A. W., Zufall, M. J., Burke, J. M,
Williams, R. W., Suggs, J. C., Sheldon, L. S., Walsh, D., and
Kwock, R. (2000). Use of a continuous nephelometer to
measure personal exposure to particles during the U.S.
Environmental Protection Agency Baltimore and Fresno panel
studies. Journal of the Air and Waste Management Association
50:1125-1132.
Howard-Reed C, Wallace LA, and Emmerich SJ (2003). Effect
of ventilation systems and air filters on decay rates of particles
produced by indoor sources in an occupied townhouse.
Atmospheric Environment 37(38):5295-5306.
Howard-Reed, C.H., Wallace, L.A., and Ott, W.R. (2002). The
effect of opening windows on air change rates in two homes.
Journal of the Air and Waste Management Association
52:147-159.2002.
Janssen, N. A. H., Hoek, G., Harssema, H., and Brunekreef, B.
(1997). Childhood exposure to PM]0: relation between
personal, classroom, and outdoor concentrations. Occupational
and Environmental Medicine 54:888-894.
Janssen, N.A.H., Hoek, G., Brunekreef, B., Harssema, H.,
Mensink, I., and Zuldhof, A. (1998). Personal sampling in
adults: relation among personal, indoor, and outdoor air
concentrations. American Journal of Epidemiology 147:537-
547.
Janssen, N. A. H., Hoek, G., Harssema, H., and Brunekreef, B.
(1999). Personal exposure to fine particles in children
correlates closely with ambient fine particles. Archives of
EnvironmentalHealth 54:95-101.
Janssen, N. A. H., De Hartog, J. J., Hoek, G., Brunekreef, B.,
Lanki, T., Timonen, K. L., and Pekkanen, J. (2000). Personal
exposure to fine particulate matter in elderly subjects: relation
between personal, indoor, and outdoor concentrations. Journal
of the Air and Waste Management Association 50:1133-1143.
Keeler, G. J., Dvonch, J. T., Yip, F. Y., Parker, E. A., Israel, B.
A., Marsik, F. J., Morishita, M., Barres, J. A., Robins, T. G.,
Brakefield-Caldwell, W., and Sam, M. (2002). Assessment of
personal and community-level exposures to particulate matter
among children with asthma in Detroit, Michigan, as part of
Community Action Against Asthma (CAAA). Environ. Health
Perspect. 110(suppl. 2):173-181.
Lai, A.C.K., Burne, M.A., and Goddard, A.J.H. (1999).
Measured deposition of aerosol particles on a two-dimensional
ribbed surface in a turbulent duct flow. Journal of Aerosol
Science 30(9): 1201-1214.
Lai, A.C.K. and Nazaroff, W.W. (2000). Modeling indoor
particle deposition from turbulent flow onto smooth surfaces.
Journal of Aerosol Science 31:463-476.
LaRosa, L.E., Buckley, T.J., and Wallace, L.A. (2002). Real-
time indoor and outdoor measurements of black carbon in an
occupied house: An Examination of Sources. Journal of the
Air and Waste Management Association 52:174-185.
Landis, M.S., Morris, GA., Williams, R.W., and Weinstein, J.P
(2001). Personal exposures to PM2 5 mass and trace elements
in Baltimore, Maryland. Atmospheric Environment 35:6511-
6524.
Liu, L.-J. S., Box, M., Kalman, D., Kaufman, J., Koenig, J.,
Larson, T., Sheppard, L., Slaughter, C., Lewtas, J., and
Wallace, L.A. (2003). Exposure assessment of particulate
matter for susceptible populations in Seattle, WA.
EnvironmentalHealth Perspectives 111(7):909-918.
Liu, D.L. and Nazaroff, W.W. (2001). Modeling pollutant
penetration across building envelopes. Atmospheric
Environment 35:4451-4462.
Long, C.M., Suh, H.H., Catalano, P.J. and Koutrakis, P. (2001).
Using time- and size-resolved particulate data to quantify
indoor penetration and deposition behavior. Environmental
Science & Technology 35(10):2089-2099.
Long, C.M., Suh, H.H. and Koutrakis, P. (2000).
Characterization of indoor particle sources using continuous
mass and size monitors. Journal of the Air and Waste
Management Association 50(7): 1236-1250.
Mosley, R.B., Greenwcll, D.J., Sparks, L.E., Guo, Z., Tucker,
W.G., Fortmann, R. and Whitfield, C. (2001). Penetration of
Ambient Fine Particles into the Indoor Environment. Aerosol
Science and Technology 34:( 1) 127-136.
National Research Council-National Academy of Science
(1998). Research priorities for airborne particulate matter I:
immediate priorities and a long-range research portfolio.
National Academy Press, Washington, DC.
Nazaroff, W.W. and Cass, G.R. (1989). Mathematical modeling
of indoor aerosol dynamics. Environmental Science &
Technology 23:157'-166.
Ott, W., Wallace, L. and Mage D. (2000). Predicting Particulate
(PMIO) Personal exposure distributions using a random
component superposition statistical model. Journal oj the Air
and Waste Management Association 50:1390-1406.
61
-------
Ozkaynak, H., Xue, J., Weker, R., Butler, D., Koutrakis, P., and
Spengler, J. (1996a). The Particle TEAM (PTEAM) study-
analysis of the data. Volume 111. Final Report. EPA Contract
#68-02-4544. Research Triangle Park, NC: US
Environmental Protection Agency.
Ozkaynak. H., Xue, J., Spengler, J.D., Wallace, L.A., Pellizzari,
E.D., and Jenkins, P. (1996b). Personal exposure to airborne
particles and metals: results from the Particle TEAM Study in
Riverside, CA. Journal of Exposure Analysis and
Environmental Epidemiology: 6:57-78.
Pellizzari, E.D., Thomas, K.W., Clayton, C.A., Whitmore, R.W.,
Shores, R.C., Zelon, H.S., and Perritt, R.L. (1992). Particle
Total Exposure Assessment Methodology (PTEAM): Riverside.
California pilot study. Volume I. Final Report. EPA Contract
# 68-02-4544, Research Triangle Park, NC: US
Environmental Protection Agency.
Rea, A., Zufall, M., Williams, R., Reed, C, and Sheldon, L.
(2001). The influence of human activity patterns on personal
PM exposure: a comparative analysis of filter-based and
continuous particle measurements. Journal of the Air and
Waste Management Association 51:1271-1279.
Riley, W.J., McKone, T.E., Lai, A.C. and Nazaroff, W.W.
(2002). Indoor particulate matter of outdoor origin:
importance of size- dependent removal mechanisms.
Environmental Science & Technology 36(2):200-207.
Rojas-Bracho, L., Suh, H.H., and Koutrakis, P. (2000).
Relationships among personal, indoor, and outdoor fine and
coarse particle concentrations for individuals with COPD.
Journal of Exposure Analysis and Environmental
Epidemiology: 10:294-306.
Sarnat, J.A., Koutrakis, P. and Suh, H.H. (2000). Assessing the
relationship between personal particulate and gaseous
exposures of senior citizens living in Baltimore, MD. Journal
of the Air and Waste Management Association 50(7): 1184-
1198.
Sarnat, J. A., Schwartz, J., Catalano, P. L, and Suh, H. H.
(2001). Gaseous pollutants in particulate matter
epidemiology: confounders or surrogates? Environmental
Health Perspectives 109:1053-1061.
Sarnat, J. A., Long, C. M., Koutrakis, P., Coull, B. A., Schwartz,
J., and Suh, H. H. (2002). Using sulfur as a tracer of outdoor
fine particulate matter. Environmental Science and Technology
36:5305-5314.
Schwartz, J., Dockery. D. W., and Neas, L. M. (1996). Is daily
mortality associated specifically with fine particles? Journal of
the Air and Waste Management Association 46:927-939.
Thatcher, T. and Layton, D. (1995). Deposition, resuspension,
and penetration of particles within a residence. Atmospheric
Environment 29( 13): 1487-1497.
Thatcher, T. L.. Lai, A. C. K., Moreno-Jackson, R., Sextro, R.
G., and Nazaroff, W. W. (2002). Effects of room furnishings
and air speed on particle deposition rates indoors. Atmospheric
Environment 36:\8\\-\8\9.
Thomas, K.W., Pellizzari, E.D., Clayton, C.A., Whitaker, D.A.,
Shores, R.C., Spengler, J.D., Ozkaynak, H., and Wallace, L.A.
(1993). Particle Total Exposure Assessment Methodology
(PTEAM) 1990 study: method performance and data quality
for personal, indoor, and outdoor monitoring. Journal of
Exposure Analysis and Environmental Epidemiology: 3:203-
226.
Thornburg, J., Ensor, D.S., Rodes, C.E., Lawless, P.A., Sparks,
L.E. and Mosley, R.B. (2001). Penetration of particles into
buildings and associated physical factors. Part 1: model
development and computer simulations. Aerosol Science and
Technology 34:284-296.
Thornburg J.W., Rodes C.E., Lawless P.A., Stevens C.D., and
Williams R.W. (2004). A pilot study of the influence of
residential HAC duty cycle on indoor air quality. Atmospheric
Environment 38:1567-1577.
US EPA (2002). Preliminary:particulate matter mass
concentrations associated with longitudinal panel studies:
assessing human exposures of high-risk subpopulations to
particulate matter. National Exposure Research Laboratory,
Office of Research and Development, EPA/600/R-01/086.
Washington, D.C.: U.S. Environmental Protection Agency.
US EPA (2003). Exposure of High-Risk Subpopulations to
Particles. Final Report (APM-21). National Exposure
Research Laboratory, Office of Research and Development,
EPA/600/R-03/145. Washington, D.C.: U.S. Environmental
Protection Agency.
Vette, A.F., Rea, A.W., Lawless, P.A., Rodes, C.E., Evans, G.,
Highsmith, V.R. and Sheldon, L. (2001). Characterization of
indoor-outdoor aerosol concentration relationships during the
Fresno PM exposure studies. Aerosol Science and Technology
34:118-126.
Wallace L.A. and Williams, R. (2005). Use of personal-indoor-
outdoor sulfur concentrations to estimate the infiltration factor,
outdoor exposure factor, penetration coefficient, and
deposition rate for individual homes. Environ Sci Tech
39:1707-1714.
62
-------
Wallace L.A., Emmerich S.J., and Howard-Reed C. (2004b).
Source strengths of ultrafine and fine particles due to cooking
with a gas stove. Environmental Science & Technology
38(8):2304-2311.
Wallace L.A., Emmerich S.J., and Howard-Reed C. (2004a).
Effect of central fans and in-duct filters on deposition rates of
ultrafine and fine particles in an occupied townhouse.
Atmospheric Environment 38(4):405-413.
Wallace L.A., Mitchell H., O'Connor G.T., Liu L.-J., Neas L.,
LippmannM., Kattan M., Koenig J., Stout J.W., Vaughn B.J.,
Wallace D., Walter M., and Adams K. (2003). Particle
concentrations in inner-city homes of children with asthma:
the effect of smoking, cooking, and outdoor pollution.
Environmental Health Perspectives 111:1265-1272.
Wallace, L.A. and Howard-Reed, C.H. (2002). Continuous
monitoring of ultrafine, fine, and coarse particles in a
residence for 18 months in 1999-2000. Journal of the Air and
Waste Management Association 52(7):828-844.
Wallace, L.A., Howard-Reed, C.H. and Emmerich, S.J. (2002).
Continuous measurements of air change rates in an occupied
house for one year: the effect of temperature, wind, fans, and
windows. Journal of Exposure Analysis and Environmental
Epidemiology 12:296-306.
Wallace, L.A. (2000a) Correlations of Personal Exposure to
Particles with Outdoor Air Measurements: A Review of
Recent Studies. Aerosol Science and Technology 32:15-25.
Wallace, L.A. (2000b). Real-Time Monitoring of Particles,
PAH, and CO in an Occupied Townhouse, Applied
Occupational and Environmental Hygiene 15(1): 39-47.
Williams, R., Creason, J., Zweidinger, R., Watts, R., Sheldon, L.
and Shy, C. (2000a). Indoor, outdoor and personal exposure
monitoring of paniculate air pollution: the Baltimore elderly
epidemiology-exposure pilot study. Atmospheric Environment
34:4193-4204.
Williams, R., Suggs, J., Zweidinger, R., Evans, G., Creason, J.,
Kwok, R., Rodes C., Lawless, P., and Sheldon, L. (2000b).
Comparison of PM2 5 and PM,0 monitors. Journal of Exposure
Analysis and Environmental Epidemiology 10:497-505.
Williams R.W., Suggs, J, Rea A., Leovic K., Vette A., Croghan
C., Sheldon L., Rodes C., Thornburg J., Ejire A. et al. (2003a).
The Research Triangle Park paniculate matter panel study:
PM mass concentration relationships, Atmospheric
Environment 37(38):5349-5363.
Williams R, Suggs J, Rea A, Sheldon L, Rodes C, and
Thornburg J (2003b). The Research Triangle Park paniculate
matter panel study: modeling ambient source contribution to
personal and residential PM mass concentrations, Atmospheric
Environment 37(38):5365-5378.
Wilson, W. E., Mage, D. T., and Grant, L. D. (2000). Estimating
separately personal exposure to ambient and nonambient
paniculate matter for epidemiology and risk assessment: why
and how. Journal of the Air and Waste Management
Association 50:1167-1183.
Wilson, W. and Suh, H. (1997). Fine and coarse particles:
concentration relationships relevant to epidemiological studies.
Journal o] the Air and Waste Management Association
47:1238-1249.
Yakovleva, E., Hopke, P. and Wallace, L. (1999). Positive
Matrix Factorization in Determining Sources of Particles
Measured in EPA's Particle TEAM Study, Environmental
Science & Technology 33:3645-3652.
63
-------
Appendix
In this Appendix we document the seasonal variation of the
infiltration factor Fmland the ambient exposure factor Fpl,x. Table
A-l presents the seasonal average of Fml and the seasonal
harmonic average of the air exchange rate for each home.
Table A-2 compares the seasonal average indoor/outdoor ratio
for sulfur to the slopes obtained when regressing indoor on
outdoor sulfur. The slope is usually smaller than the ratio,
averaging 91% of its value, and also has a wider range, with a
standard deviation almost twice that of the ratio (0.26 compared
to 0.14). This is to be expected, since measurement error causes
lower slopes in regressions, and the regressions are based on a
maximum of 7 values (with a few exceptions), and usually only
5, 6, or 7; this leads to more variability than is desirable.
Table A-3 compares the personal/outdoor ratios (/",,) averaged
over one season with the slopes determined from regressions of
personal on outdoor sulfur. As with the /",/ values in Table A-2,
the slopes are lower than the ratios and once again the standard
deviations are nearly twice as high (0.19 compared to 0.11).
The results of regressing the outdoor contribution to personal
exposures (by season) on the outdoor monitors arc provided in
Figures A-l to A-3, which provide boxplots of the adjusted R"
values. Note that the definition of the adjusted R~ parameter
allows negative values. The range of these seasonally calculated
R" values is quite a bit larger than the range of the year-round R:
and suggests that the increased variability due to the smaller
number of observations in each regression outweighs whatever
advantage was gained in looking at the smaller time period.
1.2
1.0
0.8
0.6
H 04
(N
±
0.2
00
-02
-04
1 Summer 3 Winter
2 Fall 4 Spring
Season
D2.
c Median
I I 25%-75%
~T 5%-95%
c Outliers
Figure A-1. Adjusted R" values from regressing the outdoor contribution
to personal exposure on outdoor PM25 measurements just outside the
house.
1.2
1.0
0.6
0.0
1 Summer 3 Winter
2 Fall 4 Spring
Season
c Median
I I 25%-75%
~T 5%-95%
c Outliers
Figure A-2. Adjusted R2 values from regressing the outdoor contribution
to personal exposure on outdoor PM25 Harvard Impactor (HI)
measurements at the central site.
64
-------
1.2
1.0
0.8
0.6
tsi
5 0.4
0.2
0.0
-0.2
-0.4
1 Summer 3 Winter
2 Fall 4 Spring
Season
T~ 5%-95%
<- Outliers
* Extremes
Figure A-3. Adjusted R2 values from regressing the outdoor contribution
to personal exposure on outdoor PM2 5 Federal Reference Method (FRM)
measurements at the central site.
65
-------
Table A-1. Values of the Average Sulfur Indoor/Outdoor Ratio (Fml) and the Air Exchange Rates by House and by Season
House
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
6
6
6
6
7
7
7
7
8
9
9
9
9
10
10
10
10
11
11
12
12
12
12
13
14
Season
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Winter
Summer
Fall
Winter
Spring
Summer
Summer
N
8
6
2
6
6
4
7
6
8
6
7
7
3
7
4
7
7
6
7
7
5
7
7
7
7
5
6
6
6
5
5
5
6
7
5
7
6
2
7
7
7
7
7
7
S In/Outa
0.58
0.64
0.76
0.65
0.59
0.53
0.64
0.66
0.78
0.70
0.69
0.76
0.32
0.48
0.51
0.45
0.51
0.65
0.74
0.60
0.65
0.58
0.61
0.67
0.83
0.75
0.86
0.36
0.48
0.55
0.62
0.69
0.46
0.57
0.56
0.64
0.69
0.67
0.40
0.57
0.62
0.54
0.37
0.41
SE
0.03
0.03
0.06
0.03
0.03
0.04
0.03
0.03
0.03
0.03
0.03
0.03
0.05
0.03
0.04
0.03
0.03
0.03
0.03
0.03
0.04
0.03
0.03
0.03
0.03
0.04
0.03
0.03
0.03
0.04
0.04
0.04
0.03
0.03
0.04
0.03
0.03
0.06
0.03
0.03
0.03
0.03
0.03
0.03
Airexb
0.63
0.79
1.29
0.82
0.46
0.81
1.35
0.71
1.05
0.50
0.87
1.07
0.21
0.26
0.34
0.22
0.43
0.36
0.60
0.45
0.57
0.60
0.58
0.56
1.06
1.50
1.94
0.24
0.23
0.40
0.84
0.66
0.30
0.54
0.67
0.66
0.39
4.49
0.27
0.45
0.64
0.33
0.22
0.23
SE
0.10
0.11
0.19
0.11
0.11
0.14
0.10
0.11
0.10
0.11
0.10
0.10
0.16
0.10
0.14
0.10
0.10
0.11
0.10
0.10
0.12
0.10
0.10
0.10
0.10
0.12
0.11
0.11
0.11
0.12
0.12
0.12
0.11
0.10
0.12
0.10
0.11
0.19
0.10
0.10
0.10
0.10
0.10
0.10
66
-------
Table A-1. Continued
House
14
14
14
15
15
15
15
16
16
16
16
17
17
17
17
18
19
19
19
19
20
20
20
20
21
21
21
21
22
22
23
23
24
24
24
24
25
25
26
26
26
26
27
27
-
Season
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Summer
Fall
Summer
Fall
Winter
Spring
Summer
Fall
Summer
Fall
Winter
Spring
Summer
Fall
N
6
7
7
7
7
3
7
7
7
7
7
7
5
1
6
6
6
7
7
7
7
6
7
7
7
6
6
7
6
7
7
6
7
7
7
6
2
6
6
6
5
5
6
7
S In/Out*
0.46
0.45
0.47
0.61
0.73
0.76
0.77
0.59
0.71
0.68
0.61
0.42
0.53
0.43
0.59
0.52
0.62
0.79
0.84
0.73
0.56
0.43
0.47
0.51
0.43
0.68
0.67
0.67
0.44
0.65
0.43
0.55
0.41
0.55
0.60
0.64
0.46
0.81
0.61
0.78
0.70
0.73
0.54
0.77
SE
0.03
0.03
0.03
0.03
0.03
0.05
0.03
0.03
0.03
0.03
0.03
0.03
0.04
0.08
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.06
0.03
0.03
0.03
0.04
0.04
0.03
0.03
Airex"
0.33
0.51
0.35
0.42
0.49
0.72
0.56
0.41
0.65
1.28
0.95
0.11
0.14
0.51
0.27
0.30
0.61
1.33
1.54
0.93
0.24
0.23
0.52
0.35
0.28
0.53
0.63
0.58
0.27
0.48
0.57
1.02
0.33
0.35
0.51
0.69
0.27
0.67
0.50
0.79
1.15
0.79
0.61
1.25
SE
b~7i
0.10
0.10
0.10
0.10
0.16
0.10
0.10
0.10
0.10
0.10
0.10
0.12
0.27
0.11
0.11
0.11
0.10
0.10
0.10
0.10
0.11
0.10
0.10
0.10
0.11
0.11
0.10
0.11
0.10
0.10
0.11
0.10
0.10
0.10
0.11
0.19
0.11
0.11
0.11
0.12
0.12
0.11
0.10
67
-------
Table A-1. Continued
House
27
27
28
28
28
28
29
29
29
29
31
31
32
32
32
32
33
33
33
33
34
34
36
37
37
37
38
38
38
Season
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Summer
Fall
Winter
Spring
Fall
Winter
Spring
Mean
SD
N
7
7
5
7
6
6
13
6
5
7
7
7
7
7
7
7
5
5
4
6
7
3
5
5
7
7
3
5
7
720
S In/Out3
0.66
0.68
0.41
0.65
0.61
0.62
0.40
0.59
0.75
0.70
0.97
0.80
0.25
0.40
0.48
0.31
0.23
0.35
0.49
0.40
0.54
0.74
0.25
0.58
0.66
0.55
0.52
0.64
0.50
0.59
0.14
SE
0.03
0.03
0.04
0.03
0.03
0.03
0.02
0.03
0.04
0.03
0.03
0.03
0.03
0.03
0.03
0.03
0.04
0.04
0.04
0.03
0.03
0.05
0.04
0.04
0.03
0.03
0.05
0.04
0.03
Airex"
2.61
0.67
0.28
0.47
0.86
0.37
0.31
0.42
0.61
0.53
2.77
1.11
0.23
0.27
0.33
0.25
0.15
0.25
1.12
0.64
0.36
0.45
0.25
0.27
0.69
0.30
0.18
0.54
0.26
0.64
0.56
SE
0.10
0.10
0.12
0.10
0.11
0.11
0.08
0.11
0.12
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.12
0.12
0.14
0.11
0.10
0.16
0.12
0.12
0.10
0.10
0.16
0.12
0.10
WWHXKiBnHUWeUUAKHaK
a Average ratio of indoor/outdoor sulfur concentrations
b Harmonic average of air exchange rate (h"1)
68
-------
Table A-2. Comparison of Seasonal Average
Sulfur indoor/Outdoor Ratios (Sin/Sout) with Slopes of Regressions of Sin on Sout
House
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
9
9
9
9
10
10
10
10
11
11
12
12
12
12
13
14
Season
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
S
F
W
V
S
F
W
V
S
W
S
F
W
V
S
S
N
8
6
7
6
6
7
7
6
8
7
7
7
3
7
6
7
7
6
7
7
7
7
7
7
7
7
7
7
6
6
6
7
5
6
7
7
7
6
2
7
7
7
7
7
7
Sin/Sout
(Fir,f)
0.58
0.64
0.66
0.65
0.59
0.51
0.64
0.66
0.78
0.71
0.69
0.76
0.32
0.48
0.50
0.45
0.51
0.65
0.71
0.74
0.60
0.63
0.58
0.61
0.67
0.83
0.76
0.84
0.36
0.48
0.57
0.61
0.69
0.46
0.57
0.60
0.64
0.69
0.67
0.40
0.57
0.62
0.54
0.37
0.41
SE
0.01
0.03
0.06
0.03
0.04
0.02
0.02
0.01
0.03
0.02
0.04
0.02
0.06
0.02
0.02
0.06
0.01
0.04
0.05
0.08
0.03
0.02
0.01
0.01
0.02
0.02
0.02
0.03
0.02
0.02
0.05
0.01
0.10
0.01
0.03
0.04
0.06
0.03
0.01
0.02
0.02
0.03
0.06
0.03
0.04
Intercept
99
96
-64
106
552
18
39
22
537
60
190
365
880
104
45
166
198
-250
-68
-148
164
-100
-63
52
37
-288
-3
-257
291
-14
284
23
263
76
45
192
1002
382
253
69
23
132
60
155
SE
111
101
210
135
222
153
113
77
193
173
88
119
108
157
116
87
105
222
185
403
270
187
47
66
238
54
53
119
247
227
223
26
477
164
94
119
266
368
184
90
57
100
78
113
P (Int)
0.41
0.40
0.77
0.48
0.07
0.91
0.75
0.79
0.03
0.74
0.08
0.03
0.08
0.54
0.72
0.12
0.12
0.32
0.73
0.73
0.57
0.62
0.24
0.47
0.88
0.00
0.96
0.08
0.31
0.95
0.27
0.40
0.62
0.67
0.66
0.17
0.01
0.36
0.23
0.47
0.70
0.24
0.48
0.23
Slope
0.52
0.57
0.72
0.57
0.30
0.51
0.60
0.64
0.48
0.68
0.54
0.62
0.00
0.42
0.47
0.31
0.44
0.83
0.77
0.81
0.54
0.67
0.63
0.57
0.66
0.97
0.76
1.07
0.27
0.49
0.34
0.58
0.53
0.43
0.54
0.37
0.22
0.55
0.31
0.51
0.60
0.41
0.32
0.28
SE
0.06
0.06
0.15
0.09
0.10
0.05
0.09
0.06
0.10
0.07
0.06
0.04
0.03
0.07
0.07
0.04
0.04
0.15
0.13
0.17
0.10
0.06
0.03
0.04
0.07
0.02
0.05
0.10
0.06
0.07
0.17
0.03
0.16
0.06
0.06
0.13
0.09
0.12
0.06
0.05
0.04
0.04
0.05
0.07
p (Slope)
0.0001
0.0008
0.0045
0.0027
0.0447
0.0001
0.0010
0.0005
0.0027
0.0002
0.0002
0.0001
0.9711
0.0014
0.0025
0.0004
0.0001
0.0055
0.0018
0.0050
0.0030
0.0001
0.0001
0.0001
0.0002
0.0001
0.0001
0.0001
0.0141
0.0017
0.1111
0.0001
0.0484
0.0014
0.0003
0.0389
0.0518
0.0098
0.0033
0.0002
0.0001
0.0002
0.0010
0.0111
(adj.)
0.92
0.94
0.79
0.89
0.60
0.95
0.88
0.95
0.77
0.94
0.94
0.97
-1.00
0.87
0.90
0.92
0.96
0.85
0.85
0.78
0.82
0.95
0.99
0.97
0.95
1.00
0.97
0.95
0.77
0.92
0.39
0.99
0.70
0.92
0.93
0.53
0.48
0.80
0.82
0.94
0.97
0.94
0.88
0.71
Ratio3
0.90
0.88
1.09
0.87
0.51
0.99
0.94
0.97
0.62
0.96
0.78
0.81
0.00
0.88
0.94
0.68
0.85
1.29
1.09
1.10
0.89
1.07
1.08
0.93
0.98
1.17
1.00
1.28
0.75
1.01
0.59
0.95
0.77
0.94
0.94
0.61
0.35
0.80
0.78
0.89
0.96
0.76
0.87
0.69
69
-------
Table A-2. Continued
House
14
14
14
15
15
15
15
16
16
16
16
17
17
17
17
18
19
19
19
19
20
20
20
20
21
21
21
21
22
22
23
23
24
24
24
24
25
25
26
26
26
26
27
27
27
Season
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
S
F
W
V
S
F
W
V
S
F
W
V
S
F
S
F
S
F
W
V
S
F
S
F
W
V
S
F
W
N
6
7
7
7
7
6
7
7
7
7
7
7
6
7
6
6
6
7
7
7
7
6
7
7
7
7
6
7
6
7
7
6
7
7
7
6
2
6
6
6
7
5
7
7
7
Sin/Sout
(F.nf)
0.46
045
0.47
0.61
0.73
0.66
0.77
0.59
0.71
0.68
0.61
0.42
0.52
0.55
0.59
0.52
0.62
0.79
0.84
0.73
0.56
0.43
0.47
0.51
0.43
0.70
0.67
0.67
0.44
0.65
043
0.55
0.41
0.55
0.60
0.64
0.46
0.81
0.61
0.78
0.69
0.73
0.54
0.77
0.66
SE
0.04
0.04
0.02
0.04
002
0.05
0.03
0.02
0.03
0.01
0.02
0.03
0.05
0.06
0.03
0.03
0.03
0.01
0.05
0.02
0.04
0.02
0.02
0.01
0.02
0.03
0.03
0.02
0.03
0.03
0.02
0.02
0.02
0.03
0.04
0.04
0.01
0.02
0.01
0.02
0.02
0.04
0.01
0.03
0.01
Intercept
54
-71
316
85
-15
-212
-116
246
-49
34
-217
4
285
-31
102
226
-118
-19
135
-102
755
-25
-33
-152
-151
-73
173
-55
586
89
88
15
248
125
67
-129
-95
-66
-63
100
-680
-81
-2
-3
SE
111
139
164
116
60
145
147
126
106
15
77
340
198
222
87
173
510
31
115
504
197
108
59
192
167
312
66
91
177
128
168
63
174
127
66
218
173
212
349
87
1169
73
126
50
P (Int)
0.65
063
0.11
0.50
0.82
0.22
034
0.11
0.67
0.08
0.04
0.99
0.22
0.89
0.31
0.26
0.83
0.57
0.29
0.85
0.01
0.82
0.60
0.46
0.41
0.82
0.06
0.58
0.03
0.52
0.62
0.82
0.21
0.37
0.35
0.59
0.62
0.77
0.87
0.30
0.60
0.32
0.99
0.96
Slope
0.42
0.53
0.33
0.55
0.75
0.91
0.88
0.50
0.45
0.64
0.81
042
0.35
0.59
0.51
0.42
0.65
080
0.69
0.77
032
0.45
0.50
0.57
0.48
0.73
0.50
0.72
0.22
0.60
0.40
0.54
0.32
0.48
0.51
0.76
0.86
0.63
0.80
0.57
0.99
0.57
0.77
0.66
SE
0.07
0.12
0.07
0.07
0.04
0.15
0.09
0.04
0.07
0.01
0.06
0.12
0.07
0.19
0.05
0.06
0.14
0.02
0.10
0.18
0.05
0.06
0.05
0.07
0.05
0.11
0.05
0.08
0.06
0.06
004
0.04
0.06
0.05
0.05
0.19
0.08
0.05
0.11
0.10
0.43
0.02
0.08
0.04
p (Slope)
0.0040
0.0062
0.0050
0.0007
0.0001
0.0033
0.0002
0.0001
0.0001
0.0001
0.0001
0.0190
0.0090
0.0250
0.0006
0.0025
0.0086
0.0001
0.0009
0.0084
0.0016
0.0017
0.0001
0.0006
0.0001
0.0012
0.0006
0.0003
0.0247
0.0001
0.0003
0.0002
0.0035
0.0002
0.0002
00150
0.0003
0.0002
0.0016
0.0020
0.1038
0.0001
0.0002
0.0001
R2
(adj.)
0.87
0.77
078
0.90
0.98
0.88
093
0.96
0.96
1.00
0.96
0.64
0.81
0.60
095
0.90
0.82
1.00
0.89
0.74
0.86
0.92
0.95
0.91
0.95
0.88
0.95
093
0.69
0.95
0.93
0.97
0.81
0.94
0.94
0.76
096
0.97
092
0.85
0.52
0.99
0.94
0.98
Ratio3
0.91
1.17
0.69
0.89
1.02
1.38
1.15
0.83
0.63
0.94
1.31
1.00
0.67
1.07
0.88
0.80
1.06
1.02
0.82
1.05
0.58
1.04
1.07
1.12
1.11
1.04
0.75
1.08
0.49
0.92
0.93
0.97
0.77
0.86
0.86
1.20
1.06
1.03
1.03
083
1.34
1.05
1.00
1.00
70
-------
Table A-2. Continued
House
27
28
28
28
28
29
29
29
29
31
31
32
32
32
32
33
33
33
33
34
34
36
37
37
37
38
38
38
Mean
SD
Season
V
S
F
W
V
S
F
W
V
S
F
S
F
W
V
S
F
W
V
S
F
S
F
W
V
F
W
V
N
7
5
7
7
6
13
7
7
7
7
7
7
7
7
7
5
6
5
7
7
5
6
6
7
7
3
7
7
775
Sin/Sout
(Fmf)
0.68
0.41
0.65
0.62
0.62
0.40
0.61
0.71
0.70
0.97
0.80
0.25
0.40
0.48
0.31
0.23
0.40
0.49
0.42
0.54
0.72
0.26
0.59
0.66
0.55
0.52
0.64
0.50
0.59
0.14
SE
0.04
0.02
0.01
0.02
0.05
0.02
0.05
0.06
0.03
0.02
0.03
0.02
0.02
0.03
0.02
0.04
0.07
0.05
0.04
0.03
0.03
0.02
0.04
0.03
0.03
0.03
0.03
0.01
0.03
0.02
Intercept
257
126
-20
173
980
89
182
259
135
59
-319
2
35
-6
335
95
454
-89
165
-41
-21
10
-152
-94
565
-23
136
96
89
241
SE
273
241
25
68
319
210
311
142
167
106
160
71
91
112
129
31
373
146
83
174
331
93
326
129
113
87
112
59
167
137
P (Int)
0.39
0.64
0.46
0.05
0.04
0.68
0.58
0.13
0.46
0.60
0.10
0.97
0.72
0.96
0.05
0.06
0.29
0.58
0.10
0.83
0.95
0.92
0.66
0.50
0.00
0.84
0.28
0.17
0.47
0.30
Slope
0.56
0.37
0.67
0.42
0.14
0.36
0.47
0.39
0.60
0.92
0.95
0.25
0.38
0.48
0.16
0.12
0.05
0.60
0.29
0.56
0.74
0.26
0.69
0.73
0.30
054
046
0.41
0.54
0.20
SE
0.11
0.07
0.01
0.07
0.15
0.06
0.22
0.15
0.10
0.07
0.06
0.05
0.05
0.07
0.05
0.02
0.27
0.17
0.03
0.06
0.11
0.03
0.21
0.08
0.04
0.07
013
005
0.08
0.06
P (Slope)
0.0040
0.0210
0.0001
0.0024
0.4100
0.0001
0.0840
0.0508
0.0019
0.0001
0.0001
0.0032
0.0005
0.0011
0.0306
0.0078
0.8700
0.0350
0.0003
0.0002
0.0060
0.0017
0.0290
0.0003
0.0011
0.0869
0.0172
0.0006
0.03
0.13
R2
(^j.)
0.80
0.88
1.00
0.84
-0.03
0.73
0.38
0.48
085
0.97
0.97
0.82
0.91
0.88
0.57
0.91
-0.24
0.76
0.93
0.94
0.92
0.92
0.67
0.93
0.88
0.96
0.65
0.90
0.83
0.25
Ratio3
0.82
0.89
1.03
0.67
0.22
0.90
0.77
0.55
0.86
0.95
1.19
0.98
0.94
1.01
052
0.55
0.12
1.23
0.69
1.03
1.02
0.99
1.17
1.10
0.54
1.05
0.73
0.81
0.91
0.26
"Ratio of slope to Sin/Sout.
71
-------
Table A-3. Comparison of Seasonal Average Sulfur Personal/Outdoor Ratios (Spers/Sout) with Slopes of Regressions of Spers on Sout
Subject
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
9
9
9
9
10
10
10
10
11
11
12
12
12
12
13
Season
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
S
F
W
V
S
F
W
V
S
W
S
F
W
V
S
N
6
6
7
6
6
6
6
6
5
7
7
2
2
7
7
7
6
7
7
7
6
7
7
7
6
7
7
7
6
6
7
7
5
5
7
7
2
6
3
7
7
7
6
7
Spers/Sout
(Fpex)
050
058
059
065
047
048
047
059
0.72
064
060
0.59
041
0.48
050
047
059
0.65
066
082
044
055
0.50
054
054
066
056
070
036
047
058
057
070
044
046
051
062
076
062
040
051
056
044
043
SE
002
003
005
0.05
004
0.03
004
002
004
0.04
0.02
001
0.01
0.04
005
006
004
0.05
003
004
0.04
003
002
002
002
005
004
0.04
001
002
007
0,07
005
002
003
003
025
004
005
002
002
0.06
003
004
Slope
052
0.57
066
0.54
0.29
046
0.56
074
045
068
0.50
033
035
0.55
046
0.66
062
079
042
068
055
045
050
099
078
1 00
033
043
056
046
072
048
0.36
048
049
077
0.28
045
060
042
039
SE
009
005
0 14
0 11
012
009
016
009
0.44
0 12
0.03
009
013
0 12
0.10
023
0.07
006
0.13
008
0.04
005
0 11
006
0 12
0 11
004
0.06
030
020
0 14
005
0.05
008
012
013
004
006
009
005
0 10
P
00049
0.0003
0.0051
0.0078
00683
00073
0.0240
0.0012
03799
0.0026
0.0001
00150
0.0440
0.0051
00085
0.0336
00004
0.0001
00327
0.0004
00001
0.0003
00111
00001
00013
00003
00007
00016
0 1230
00660
00128
00031
00008
00023
00148
0 1075
00015
00009
00012
00008
00094
Intercept
-53
17
-76
141
345
55
-103
-173
439
-95
122
282
182
-131
307
0
61
12
21
-341
-76
109
131
-680
-180
-334
77
120
31
100
-54
-91
104
17
746
-118
337
72
-56
10
37
SE
184
80
197
169
254
269
216
110
736
304
53
218
206
282
279
315
105
152
369
254
63
79
429
150
119
137
138
198
400
192
393
161
76
76
370
127
135
105
129
121
159
P
079
084
0.71
045
0.25
0.85
0.66
0.19
059
077
0.07
0.25
042
0.66
0.33
1 00
059
094
0.96
0.24
0.28
023
0.77
001
019
006
0.61
0.58
0.94
0.62
090
061
023
083
011
0.52
0.05
0.52
068
094
082
R2 (adj)
0.86
0.96
0.78
082
051
0.83
070
0.93
0.01
0.83
0.97
0.67
0.51
0.78
0.82
055
0.92
0.96
065
092
0.97
0.93
079
098
087
093
094
0.92
0.29
0.43
0.87
095
0.90
084
076
0.94
087
089
088
094
0.72
Ratio3
1.06
0.98
1.11
083
062
0.96
1 18
1.26
0.63
1 06
084
0.69
070
1.17
078
1 00
0.93
0.97
096
1.24
1 12
0.85
093
1 51
1 38
1 44
093
092
096
080
1 03
1 08
079
0,95
065
1 23
069
087
1 07
095
0.92
72
-------
Table A-3. Continued
« «*
Subject
14
14
14
14
15
15
15
15
16
16
16
16
17
17
17
17
18
19
19
19
19
20
20
20
20
21
21
21
21
22
22
23
23
24
24
24
24
25
25
26
26
26
26
27
Season
S
F
W
V
S
F
W
V
S
F
W
V
S
F
W
V
S
S
F
W
V
S
F
W
V
S
F
W
V
S
F
S
F
S
F
W
V
S
F
S
F
W
V
S
N
7
7
6
7
6
7
7
7
6
7
7
7
6
6
7
6
6
6
7
7
2
7
6
7
2
7
7
7
7
6
7
7
7
7
6
6
4
7
7
6
6
7
2
7
Spers/Sout
(F >
\rr>ext
0.40
0.43
0.43
0.48
0.59
0.61
0.61
0.64
0.57
0.66
0.57
0.56
0.44
0.43
0.46
0.57
0.51
0.59
0.72
0.72
0.66
0.55
0.45
0.39
0.52
0.55
0.74
0.75
0.73
0.48
0.60
0.47
0.50
0.53
0.53
0.63
0.64
0.52
0.70
0.52
0.70
0.56
0.74
0.46
SE
0.02
0.02
0.04
0.03
0.03
0.02
0.05
0.03
0.01
0.03
0.03
0.03
0.03
0.06
0.06
0.03
0.02
0.03
0.01
0.03
0.06
0.03
0.03
0.02
0.03
0.03
0.01
0.05
0.05
0.04
0.03
0.03
0.02
0.02
0.04
0.10
0.03
0.02
0.02
0.02
0.01
0.02
0.02
0.01
Slope
0.35
0.51
0.65
0.23
0.69
0.67
0.86
0.81
0.53
0.66
0.46
0.90
0.51
0.27
0.47
0.55
0.47
0.61
0.68
0.66
0.42
0.39
0.50
0.54
0.71
0.52
1.21
0.22
0.68
0.49
0.46
0.64
0.48
0.49
0.67
0.49
0.86
0.47
0.65
0.42
0.43
SE
0.05
0.05
0.14
0.05
0.08
0.03
0.14
0.05
0.06
0.06
0.04
0.05
0.09
0.08
0.19
0.08
0.07
0.11
0.02
0.07
0.07
0.09
0.04
0.07
0.03
0.08
0.15
0.11
0.05
0.13
0.04
0.04
0.04
0.16
0.16
0.07
0.06
0.05
0.05
0.09
0.03
P
0.0011
0.0001
0.0097
0.0084
0.0011
0.0001
0.0019
0.0001
0.0008
0.0001
0.0001
0.0001
0.0053
0.0293
0.0574
0.0023
0.0028
0.0055
0.0001
0.0002
0.0018
0.0103
0.0001
0.0007
0.0001
0.0015
0.0004
0.1198
0.0001
0.0120
0.0001
0.0001
0.0002
0.0377
0.0547
0.0012
0.0001
0.0006
0.0002
0.0051
0.0001
Intercept
_.
-99
-247
526
-129
-56
-219
-224
107
-16
107
-389
-176
250
6
34
76
-49
45
49
412
69
-102
33
81
197
-505
660
-92
-66
39
-275
71
125
-32
68
-317
187
136
120
90
SE
83
68
178
132
137
45
146
82
177
100
47
59
271
216
228
134
201
417
41
82
265
156
44
265
97
98
167
323
111
473
62
114
99
215
193
278
141
199
157
79
105
P
0.49
0.21
0.24
0.01
0.40
0.27
0.19
0.04
0.58
0.88
0.07
0.00
0.55
0.31
0.98
0.81
0.73
0.91
0.32
0.57
0.18
0.68
0.07
0.91
0.44
0.10
0.03
0.11
0.45
0.89
0.56
0.06
0.52
0.59
0.88
0.82
0.07
0.40
0.43
0.19
0.43
R (adj)
__
0.96
0.80
0.74
0.93
0.99
0.85
0.97
0.94
0.95
0.96
0.98
0.85
0.67
0.46
0.90
0.89
0.85
0.99
0.94
0.85
0.80
0.97
0.90
0.99
0.87
0.92
0.37
0.97
0.70
0.95
0.98
0.97
0.63
0.84
0.88
0.97
0.95
0.97
0.78
0.97
Ratio"
0.87
1.21
1.51
0.49
1 16
1.09
1.41
1.26
0.93
1.01
0.80
1.62
1.17
0.61
1.01
0.95
0.93
1.03
0.95
0.92
0.77
0.88
1.26
0.98
0.95
0.70
1.66
0.47
1.12
1.04
0.92
1.20
0.90
0.79
1.05
0.95
1.23
0.89
0.93
0.75
0.93
73
-------
Table A-3. Continued
Subject
27
27
27
28
28
28
28
29
29
29
29
31
31
32
32
32
32
33
33
33
33
34
34
35
36
37
37
37
38
38
38
Mean
SD
Season
F
W
V
s
F
W
V
s
F
W
V
s
F
S
F
W
V
S
F
W
V
s
F
S
S
F
W
V
F
W
V
N
7
7
7
5
7
7
6
7
6
7
7
7
7
7
7
7
7
4
6
5
7
7
6
7
6
7
7
7
7
7
7
750
_,.,,,,,,,:,
Spers/Sout
(Fp.«)
0.62
0.58
0.54
0.39
0.50
0.50
0.48
0.43
0.61
0.59
0.62
0.91
0.66
0.47
0.53
0.60
0.38
0.30
0.40
0.48
0.32
0.52
0.74
0.41
0.45
0.55
0.54
0.53
0.54
0.57
0.52
0.55
0.11
SE
0.01
0.02
0.03
0.02
0.02
0.02
0.04
0.03
0.06
0.05
0.06
0.02
0.03
0.03
0.04
0.05
0.04
0.08
0.07
0.04
0.03
0.02
0.04
0.04
0.03
0.09
0.02
0.03
0.04
0.04
0.04
0.04
0.03
Slope
0.60
0.54
0.50
036
0.54
0.33
0.10
0.49
0.97
0.28
0.72
0.94
0.76
0.56
0.47
0.38
0.15
0.16
0.15
0.51
0.26
0.54
0.72
0.28
0.51
0.26
0.60
0.32
0.57
0.47
0.36
0.53
0.19
SE
0.04
0.06
0.10
0.06
0.06
0.08
0.13
0.12
0.37
0.11
0.20
0.06
0.07
0.06
0.09
0.09
0.19
0.06
0.30
0.12
0.01
0.04
0.14
0.07
0.09
0.33
0.10
0.07
0.06
0.19
0.18
0.10
0.07
P
0.0001
0.0004
0.0034
0.0089
0.0002
0.0085
04929
0.0090
0.0571
0.0522
0.0152
0.0001
0.0001
0.0003
0.0033
00099
0.4704
0.1091
0.6425
0.0233
0.0001
0.0001
0.0064
0.0122
0.0057
0.4687
0.0016
0.0058
0.0002
0.0566
0.0934
0.03
0.10
Intercept
20
49
98
109
-61
144
785
-156
-503
252
-150
-49
-207
-110
74
267
516
117
318
-35
81
-38
85
313
-133
381
-83
473
-39
79
157
41
231
SE
65
75
235
211
108
72
282
428
541
100
325
87
183
98
170
145
466
86
416
105
37
130
417
219
256
484
151
174
83
160
193
190
125
P
0.77
0.54
0.69
0.64
0.60
0.10
0.05
0.73
0.40
0.05
0.66
0.60
0.31
0.31
0.68
0.13
0.32
0.31
0.49
0.76
0.08
0.78
0.85
0.21
0.63
0.47
0.61
0.04
0.66
0.64
0.45
0.48
0.29
R2(adj)
0.97
0.92
0.81
0.90
0.94
0.73
-0.09
0.73
0.55
0.47
0.67
0.98
0.95
0.93
0.82
0.72
-0.07
0.69
-0.18
0.81
0.98
0.96
0.84
069
0.85
-0.07
0.86
0.77
0.94
0.46
0.35
0.79
0.24
Ratio3
0.97
0.92
0.92
0.92
1.10
0.66
0.20
1.12
1.58
0.47
1.17
1.04
1.15
1.20
0.88
063
039
0.53
0.38
1.08
0.80
1.03
0.97
0.67
1.13
0.46
1.12
0.60
1.06
0.82
0.70
0.96
0.22
' Ratio of slope to Spers/Sout.
74
-------
Table A-4. Questionnaire Variables and Definitions
# Variable Name Definition
Units
ug/m
ng/m3
ng/m3
ng/m3
1 Persdate Julian date'100+Subject ID
2 DatelD Date (YYYYMMDD)*100 + SUBJECT id
3 EPA ID epa id
4 Subject SUBJECT ID
5 Cohort Cohort (Cardiovascular or hypertensive)
6 House House ID
7 Season Season
8 Date Julian date
9 Month Month
10 Day Day
11 Year Year
12 ambHI25 Central-site Harvard Impactor PM25
13 FRM25 Central-site Federal Reference Method PM2 5 ug/m3
14 Flaghi25 Validity code: 2 = valid; used when central-site HI25 used in regressions
15 FlagFRM Validity code. 2 = valid; used when FRM is a variable in regressions
16 Flagairex Validity code: 2 = valid; used when airex is a variable in regressions
17 persS personal sulfur
18 Sin Indoor sulfur
19 Sout Outdoor sulfur
20 Sinout Indoor/outdoor sulfur ratio
21 FlagSinout Validity code: 2 = valid; used for all indoor/outdoor regressions
22 SpersSout personal/outdoor sulfur ratio
23 Flagspersout Validity code: 2 = valid; used for all personal/outdoor regressions
24 PM25in Indoor PM25
25 PM25out Outdoor PM2 5
26 PM25pers Personal PM25
27 Outcontin Outdoor contribution to indoor PM25 concentrations
28 Outcontpers Outdoor contribution to personal PM25 concentrations
29 Incontrib Indoor-generated contribution to indoor PM25 concentrations
30 Perscontrib Non-outdoor contribution to personal PM25 concentrations
31 Perscloud Remainder after accounting for time-weighted indoor-outdoor exposures ug/m3
32 airex air exchange rate h"1
33 Cleaning Time spent cleaning minutes
34 Grooming Time spent grooming minutes
35 Otherjndoor Time in other indoor locations minutes
36 Otherjoc Time in other locations minutes
37 Travel Time spent in travel minutes
38 Unknown Time in Unknown location minutes
39 cooking Time spent cooking minutes
40 outdoor Time spent outdoors minutes
41 smoke Time spent near smokers minutes
_42 TempC Outdoor Temperature (°C) °C
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
ug/m3
75
-------
Table A-4. Continued
# Variable Name Definition
Units
43 Tempdelta Absolute Outdoor-Indoor Temp Difference using thermostat setting °F
44 Numpeopl Number of persons living in house
45 numsmok Number of smokers in house
46 cigsmokd Number of cigarettes smoked in house
47 Mealsckd Number of meals cooked
48 Burning Was food burned today?
49 Exhstfan Was exhaust fan used today?
50 Candles Were candles used today?
51 CANDLDUR Duration of candle use minutes
52 Incense Was incense used today?
53 INCENDUR Duration of incense use minutes
54 Windopen Were windows open today?
55 windowall Sum of products of open windows X width opened X duration open inch-hours
56 spacehtr Was a space heater used today?
57 Cleaning_1 Did cleaning occur today?
58 Pets Any dog or cat pets?
59 broil broiled food today
60 fry fried food today
61 grill grilled food today
62 sautee sauteed food today
63 sweep swept floors today
64 vacuum vacuumed today
65 dust dusted today
66 TYPE Type of building (1 = detached, 2 = duplex, 4 = apartment, 6 = trailer)
67 AGE age of building (years) years
68 BUSY_RD high-traffic road nearby
69 DIRT_RD dir road nearby
70 DUSTY_RD Dust from nearby construction etc
71 GAR_USE Park cars in attached garage?
72 A_C air conditioning unit(s) in home
73 FAN whole-house or attic fan
74 S_WIN Storm windows (0 = none, 0.5 = partial, 1 = all)
75 C_FUEL cooking fuel (1 = gas, 2 = electricity)
76 C_FAN range hood?
77 PILOT Number of pilot lights (0-3)
78 DRYER clothes dryer?
79 DRY_VENT clothes dryer vented outside?
80 VAC vacuum bag type (0 = none, 1 = standard, 2 = HEPA)
81 AREA area of house (square footage) feet2
82 Rooms Number of rooms
83 FLRCOVav Percent of floor covered by carpet %
84 MILDEWavg mildew noticed by technician
::^^
76
------- |