Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Personal PM2.5 exposure during pregnancy in an environmental health disparities population
(USC Thesis Other)
Personal PM2.5 exposure during pregnancy in an environmental health disparities population
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Personal PM 2.5 Exposure during Pregnancy in an Environmental Health Disparities Population
by
Xu, Yan
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POPULATION, HEALTH and PLACE)
August 2022
Copyright © 2022 Xu, Yan
ii
Dedication
This dissertation is dedicated to my family, especially my father and my brother, for their
unconditional support, as well as my friends for their company along the journey.
iii
Acknowledgements
I am grateful to have Drs. Rima Habre and John Wilson as my co-advisors. Without their
constant advisement and support, I would not have gone this far. I would also like to thank my
committee members, Dr Tracy Bastain, Dr. Shohreh Farzan, and Dr. Manuel Pastor, for their
commitment and support to this dissertation. I am grateful for the support from my colleagues
and friends in the PHP program and MADRES center, especially Dr. Johanna Avelar, Dr.
Douglas Fleming, Dr. Li Yi, Lois Park, Karl O'Sharkey, Dr. Hua Hao, Dr. Yisi Liu, and Dr.
Roxana Khalili. As for data and analysis used in the dissertation, I gratefully acknowledge Jane
Cabison, Marisela Rosales, Lisa Valencia, Wangjing Ke, and the larger MADRES team for their
data collection support, as well as the MADRES participants and their families and the clinic
partners for their time and effort. In addition, I would like to thank John Langstaff from the U.S.
Environmental Protection Agency, and Glen Graham and Chris Holder from ICF for providing
technical support for the APEX model runs.
The dissertation research was supported by NIEHS R01ES027409, NIEHS P30ES007048
pilot funding, the MADRES Center (NIEHS/NIMHD P50ES026086, EPA 83615801, NIMHD
P50MD015705), and first and fifth year fellowships from the USC Spatial Sciences Institute in
the Dornsife College of Letters, Arts and Sciences.
iv
Table of Contents
Dedication ...................................................................................................................................... ii
Acknowledgements ....................................................................................................................... iii
List of Tables ................................................................................................................................ vi
List of Figures .............................................................................................................................. vii
Abbreviations .............................................................................................................................. viii
Abstract .......................................................................................................................................... x
Chapter 1 Introduction ................................................................................................................... 1
1.1. Introduction ........................................................................................................................1
1.2. Dissertation Structure .........................................................................................................5
Chapter 2 The Impact of GPS-derived Activity Spaces on Personal PM2.5 Exposures in the
MADRES Cohort ................................................................................................................... 8
2.1. Introduction ........................................................................................................................8
2.2. Method .............................................................................................................................11
2.2.1. Data Collection .......................................................................................................11
2.2.2. Data Analysis ..........................................................................................................15
2.2.3. Bivariate Analyses ..................................................................................................20
2.2.4. Multivariate Model .................................................................................................21
2.3. Results ..............................................................................................................................22
2.3.1. Descriptive Statistics ...............................................................................................22
2.3.2. Bivariate Analyses ..................................................................................................25
2.3.3. Multivariate Model .................................................................................................28
2.4. Discussion ........................................................................................................................30
2.5. Conclusion .......................................................................................................................36
Chapter 3 Sources of Personal PM 2.5 Exposure in the MADRES Pregnancy Cohort ................. 37
v
3.1. Introduction ......................................................................................................................37
3.2. Method .............................................................................................................................41
3.2.1. Data Collection .......................................................................................................41
3.2.2. Data Analysis ..........................................................................................................45
3.3. Results ..............................................................................................................................49
3.3.1. Descriptive Statistics ...............................................................................................49
3.3.2. Positive Matrix Factorization Analysis ...................................................................51
3.4. Discussion ........................................................................................................................58
3.5. Conclusion .......................................................................................................................62
Chapter 4 Modeling Personal PM2.5 Exposures within Multiple Microenvironments ................ 63
4.1. Introduction ......................................................................................................................63
4.2. Method .............................................................................................................................67
4.2.1. Data Collection .......................................................................................................67
4.2.2. Data Analysis ..........................................................................................................69
4.3. Results ..............................................................................................................................76
4.4. Discussion ........................................................................................................................81
4.5. Conclusion .......................................................................................................................84
Chapter 5 Conclusions ................................................................................................................. 85
References .................................................................................................................................... 91
Appendix A ................................................................................................................................ 115
Appendix B ................................................................................................................................ 125
Appendix C ................................................................................................................................ 128
vi
List of Tables
Table 2.1. Summary of geospatial data sources for residential neighborhood and
activity space-based exposure assessment ........................................................................... 18
Table 2.2. Descriptive statistics of participant demographics (N=213) ....................................... 23
Table 2.3. Home characteristics, indoor sources, and durations of selected activities derived
from questionnaires and exit survey (N=213) ...................................................................... 24
Table 2.4. Bivariate results of personal PM2.5 exposures with GPS-derived time activities
and environmental variables ................................................................................................. 26
Table 2.5. List of all potential variables considered for inclusion in the multivariate model ..... 28
Table 2.6. Results of final generalized linear model of personal PM2.5 mass exposure .............. 29
Table 3.1. Descriptive statistics of participants demographics (N=212) ..................................... 49
Table 3.2. Chemical component concentrations (all in units of ng/m
3
unless
otherwise noted) ................................................................................................................... 50
Table 3.3. Source mass contributions .......................................................................................... 53
Table 3.4. Spearman correlations among PMF-predicted source contributions, colored
from low (blue) to high (red) ................................................................................................ 53
Table 3.5. Spearman correlations between PMF-predicted source contributions and variables
related to personal activities, time-activity patterns, indoor and outdoor environment ....... 54
Table 4.1. Five APEX scenarios modeled in this simulation with associated
conditional variables for the Indoor Residence microenvironment in each ......................... 73
Table 4.2. Demographic characteristics of simulated APEX and actual MADRES
participants ........................................................................................................................... 76
Table 4.3. Spearman correlations among simulated hourly total personal PM2.5 exposures,
microenvironmental exposures, and ambient PM2.5 concentrations in S3 ........................... 80
Table 4.4. PM2.5 comparisons among MADRES personal measurements, APEX estimates,
and US EPA monitoring concentrations (µg/m
3
) ................................................................ 80
vii
List of Figures
Figure 1.1. Dissertation research framework ................................................................................. 6
Figure 2.1. Illustration of multiple activity spaces and residential neighborhood ....................... 19
Figure 2.2. Regression plot between 48-hour integrated personal PM2.5 exposures and
outdoor PM2.5 at the point of residence ................................................................................ 25
Figure 2.3. Measured versus predicted personal PM2.5 concentrations and linear
regression fit ......................................................................................................................... 30
Figure 3.1. Source loading profiles (in % of species) .................................................................. 52
Figure 3.2. Relationship between secondhand smoking mass contributions and home type ...... 55
Figure 3.3. Relationship between aged sea salt and window opening time in the 48-hr
monitoring period ................................................................................................................. 56
Figure 3.4. Relationship between fresh sea salt mass contributions and average wind
direction in the 48-hr monitoring period .............................................................................. 57
Figure 3.5. Relationship between crustal mass contributions and household occupants ............. 58
Figure 4.1. APEX model workflow ............................................................................................. 71
Figure 4.2. APEX scenario S3 simulated results by microenvironment for: (a) stay time
durations (%); (b) PM2.5 concentrations (µg/m
3
); and (c) Personal time-weighted
PM 2.5 exposures (µg/m
3
) ...................................................................................................... 77
Figure 4.3. Contributions of various microenvironments to hourly personal PM2.5
exposures in S3 ..................................................................................................................... 79
viii
Abbreviations
APEX Air Pollutants EXposure (model)
AQS Air Quality System
BIC Bayesian Information Criterion
BMI Body Mass Index
CA California
CALINE California Line Source Dispersion Model
CHAD Consolidated Human Activity Database
CMB Chemical Mass Balance
CO Carbon Monoxide
CSN Chemical Speciation Network
DPA Daily Path Area
GPS Global Positioning System
IRB Institutional Review Board
KDE Kernel Density Estimation
MADRES Maternal And Developmental Risks from Environmental and Social Stressors
MCH Minimum Convex Hull
NDVI Normalized Difference Vegetation Index
NO2 Nitrogen Dioxide
NOx Nitrogen Oxides
O3 Ozone
PEM Personal Environmental Monitor
PM 2.5 Particulate Matter with aerodynamic diameter less than 2.5 µm
ix
PM10 Particulate Matter with aerodynamic diameter less than 10 µm
PMF Positive Matrix Factorization (model)
RMSE Root Mean Square Error
SD Standard Deviation
SSI Spatial Sciences Institute
STN Speciated Trends Network
USC University of Southern California
USEPA U.S. Environmental Protection Agency
XRF X-Ray Fluorescence
x
Abstract
Exposure to particulate matter air pollution with an aerodynamic diameter less than 2.5
μm (PM2.5), particularly during the 3
rd
trimester of pregnancy, has been associated with adverse
impacts on maternal and fetal health. Pregnant women are mobile and locations they spend time
in contribute to their personal PM2.5 exposures, while their total exposures are the mixtures of
multiple sources and affected by multiple factors. Environmental health disparities groups
including racial and ethnic minorities, marginalized, and lower income populations are
disproportionally burdened by elevated PM2.5 exposure and may be more susceptible to its
adverse health effects.
This dissertation used 48-hr integrated, personal PM2.5 measurements and concurrent
GPS records collected from 213 low-income, predominately Hispanic women in their 3
rd
trimester living in Los Angeles, CA, to investigate the impacts of activity spaces on personal
PM 2.5 exposures (Chapter 2), derive the main sources contributing to personal PM2.5 mass
(Chapter 3), and determine the influence of microenvironmental exposures estimated with a
stochastic exposure model and total personal exposures (Chapter 4).
This research found indoor sources dominated personal PM 2.5 exposures, where
combined indoor source contributions (i.e., secondhand smoking, crustal) were more than triple
those of outdoor sources (i.e., traffic, aged and fresh sea salt, and fuel oil). In addition,
environmental exposures encountered within the activity spaces that participants frequented
contributed significantly to personal PM2.5 exposure, with greater exposure to parks and
greenness linked with lower personal exposures. Finally, the simulated personal exposures better
approximated the distribution of personal measurements with the addition of more refined indoor
xi
source terms. However, total predicted PM2.5 exposure was highly correlated with outdoor PM2.5
which is contrary to the patterns observed with measurements.
Overall, the findings of this dissertation shed light on the complexity of sources and
determinants of personal PM2.5 exposures during pregnancy in an environmental health
disparities population, as well as the need for refined exposure assessment methods to capture
the true variability in exposure and aid in the design of relevant interventions to reduce
exposures.
1
Chapter 1 Introduction
This chapter gives an overview of the dissertation research, starting by providing
background to the research studies and knowledge gaps as well as introducing the research goals
and study population. The dissertation structure is also laid out next for readers to follow.
1.1. Introduction
Air pollution is defined as particulate, gaseous, and (semi-)volatile matter “emitted from
an anthropogenic, biogenic, or geogenic source” (Daly & Zannetti, 2007), present in the
microscale, mesoscale, synoptic and global scale of atmospheric motions that can cause short- or
long-term harm to human, animal or plant health, or to the environment (Hickey et al., 2014;
Painter, 1974; Seinfeld & Pandis, 2006). In the twentieth century, several widely publicized
incidents raised public concern about particulate matter (PM) air pollution effects on population
health, such as the historical London Fog episode in 1952 with around 12,000 deaths (Bell &
Davis, 2017). Since then, epidemiological studies have demonstrated that short- and long-term
air pollution exposure is a significant risk factor for various diseases (Benbrahim-Tallaa et al.,
2012; Brandt et al., 2014; Buffler et al., 2005; Chen et al., 2016; Gan et al., 2011; Kim et al.,
2004) and increased mortality (Dockery et al., 1993) as illustrated by the Harvard Six Cities
study (Dockery et al., 1993; Krewski et al., 2003).
Systemic inequities have resulted in persistent environmental health disparities, in which
certain groups are heavily exposed to air pollution, leading to higher health risks (Bae et al.,
2007; Houston et al., 2004; Tian et al., 2013). Studies have shown that low-income Hispanic
populations, especially in California, are disproportionally burdened by elevated air pollution
exposures and worse health outcomes, such as diabetes, lower bone mineral density, and
2
respiratory diseases (Alderete et al., 2017; Chen et al., 2015; Chen et al., 2016; Houston et al.,
2014; Pastor et al., 2004; Pulido et al., 1996). However, there is little known about the major
determinants of PM2.5 (particulate matter with aerodynamic diameter less than 2.5 µm) exposure
in this population (i.e., where and when they experience the highest exposures, and which
sources contribute the most to their personal exposures). Human mobility and the high
spatiotemporal variability in some of the major sources contributing to PM2.5 (such as traffic)
provide an added complexity when trying to accurately estimate personal exposures in
epidemiological studies.
For women during pregnancy, air pollution impacts on both their own as well as their
fetus’ health are major concerns (Dadvand et al. 2014; Ghosh et al. 2014). The health effects
may vary in different time windows, since the fetus develops different organ systems at different
weeks; therefore, exposure in different trimesters may have different health outcomes (Stieb et
al. 2012; Zhu et al. 2015). Public health researchers have conducted studies focused on decreased
birth weight related to in-utero exposure to PM2.5 during pregnancy (Fleischer et al., 2014; Hyder
et al., 2014; Pedersen et al., 2013; Rich et al., 2015; Stieb et al., 2012, 2016; Twum et al., 2016;
Zhu et al., 2015). For example, several studies have demonstrated that PM2.5 exposure during the
3
rd
trimester of pregnancy had the highest impact on the infant’s gestational weight gain and
birthweight (Huang et al., 2015; Rich et al., 2015; Romão et al., 2013; Savitz et al., 2014;
Schembari et al., 2015).
Most of health studies focused on personal exposure to air pollution of outdoor origin
with exposures assigned based on modeled pollutant distribution surfaces. The methods used to
estimate these surfaces have evolved from the nearest monitoring site (i.e., EPA monitoring
sites) (Ebisu et al., 2014; Harris et al., 2014), to spatially interpolated outdoor exposures based
3
on multiple monitoring sites (Clark et al., 2010; Gauderman et al., 2005; Kim et al., 2004) and
sophisticated spatiotemporal models of outdoor concentrations (Brunst et al., 2015; Chen et al.,
2016; Gehring et al., 2010; Hyder et al., 2014). While spatiotemporal modeled air pollution
surfaces provide the advantage of being able to assess the outdoor, residential exposures of large
study populations, they suffer from exposure measurement error which usually leads to
attenuated statistical power in epidemiological analyses.
Since individuals are mobile and spend the majority of their time indoors (Wallace,
1996), their “true” personal exposure is best approximated by the time-weighted average
concentration they experience in and across several microenvironments, most commonly
categorized as indoors, outdoors, and in transit (Gray et al., 2011; Zeger et al., 2000). The
personal exposures for large populations can be estimated using microenvironment models
(USEPA, 2020).
Personal monitoring is the gold standard approach to accurately capture the true personal
exposures in the breathing zone. Accordingly, personal monitoring studies can disentangle the
contributions of indoor and outdoor environments on personal exposures based on when and
where those sampled have interacted with their environments (Adgate et al., 2004a; Rabinovitch
et al., 2016; Steinle et al., 2015). These improvements have resulted in more accurate personal
exposure estimates which greatly reduce measurement error and increase the understanding of
how an individual’s time-activity patterns affect personal exposures.
Furthermore, since PM 2.5 itself is a mixture of various organic and inorganic elements, its
composition, and thus toxicity, may vary based on the sources from which it originated (Berger
et al., 2018; Masiol et al., 2017; Zhai et al., 2017). Several studies have conducted source
apportionment analyses on speciated PM2.5 measurement data collected at designated U.S.
4
Environment Protection Agency (USEPA) ambient monitoring sites that are part of the Speciated
Trends Network (STN). The aim of STN is to resolve the main sources that contribute to the
outdoor PM2.5 mixture and to investigate their impacts on the health of pregnant women (Bell et
al., 2010; Dadvand et al., 2014; Ng et al., 2017; Pereira et al., 2014). However, identifying the
main sources of personal PM2.5 measurements will better serve health studies because toxicity
may vary depending on this personal mixture. The information on major determinants (e.g.,
time-activities, spaces individuals frequented) and main sources of personal PM2.5 exposures will
facilitate the design of interventions that reduce exposures by connecting major sources and
where and when pregnant women are exposed.
The paucity of knowledge of pregnant women’s personal PM2.5 exposures, especially the
lack of information about the major determinants and main sources that contribute to personal
PM 2.5 mass, hinders our ability to assess health effects and provide targeted interventions. This
knowledge gap is often larger for populations burdened by environmental health disparities
because very few studies have focused on their personal PM2.5 exposures. This dissertation aims
to produce a better understanding of personal PM2.5 exposures for an environmental health
disparities population, including the main determinants (e.g., activity spaces, time-activities) and
sources that contribute to personal exposures during pregnancy. To accomplish this, this
dissertation analyzed personal data collected from a sample of 213 women enrolled in the
“Maternal And Developmental Risks from Environmental and Social Stressors (MADRES)” In-
Utero Air Pollution Exposure Study. The sample participants are low-income, predominantly
Hispanic pregnant women living in Los Angeles, CA, with personal PM2.5 measurements and
concurrent GPS tracking data collected over a 48-hr period in their 3
rd
trimester. This study
5
provided a unique opportunity to understand the personal PM2.5 exposures of this at-risk
population.
GPS records can be used to shape individuals’ activity spaces and delineate their time-
activities across microenvironments during the 48-hr sampling period; therefore, environmental
exposures within activity spaces which might affect their personal PM2.5 can be examined.
Furthermore, the personal monitoring filters were speciated and analyzed for chemical
composition information, which in turn was used to identify and resolve the major sources that
contributed to these women’s personal PM2.5 exposures. The integration of individual-level GPS
tracking and personal monitoring data further allowed us to explore how characteristics at the
individual, residential neighborhood and activity space levels interact to affect total and source-
resolved personal PM2.5 exposures during pregnancy.
The MADRES participants in this study provide a convenient sample of women of
childbearing age who are burdened by environmental health disparities in Los Angeles, CA for
whom personal measurements were collected. A stochastic inhalation exposure model would be
used to generate microenvironment level PM2.5 exposures for this population without the
personal measurements. The MADRES personal measurements were used to examine how well
the model outputs approximated personal exposures. The results can help shape the model
parameters and improve model personal exposures from multiple sources in large populations.
1.2. Dissertation Structure
The remainder of this dissertation is organized into three studies, each of which focuses
on one aspect of understanding personal PM2.5 exposures (Figure 1.1). Study 1 (Chapter 2) aims
to investigate the main determinants that affected personal PM 2.5 exposures. The activity spaces
were derived from GPS tracks to delineate the spaces in which individuals interact with their
6
environments (Kwan, 1999; Newsome et al., 1998; Sherman et al., 2005). Generalized linear
models were next applied to examine the impacts of the main factors, i.e., environmental
exposures within activity spaces, time-activities, indoor environments, and outdoor PM2.5, on
variations of personal exposures.
Figure 1.1. Dissertation research framework
Study 2 (Chapter 3) focuses on identifying the main sources that contributed to personal
PM 2.5 exposures. The USEPA-developed Positive Matrix Factorization (PMF) model was used to
resolve the main sources and quantify the mass contributions (Norris et al., 2014). In-depth
personal data analysis was then conducted to confirm the source identities and their origins.
Study 3 (Chapter 4) examines how well personal PM2.5 exposures can be modeled for
environmental health disparities women with childbearing age. The USEPA-developed Air
Pollutants EXposure (APEX) model was used to estimate personal exposures at the
microenvironment level (USEPA, 2020). Multiple scenarios were set to compare the impacts of
7
refining model parameters on personal estimates. MADRES personal measurements were then
used to examine how well the model outputs approximated the actual (i.e., measured) exposures.
Ultimately, this research provided an opportunity to: (1) understand the personal PM2.5
exposures of this environmental health disparities population during pregnancy, including major
determinants, complex PM2.5 sources, and the multiple microenvironments that contributed to
personal exposures; and (2) lay out a foundation to reduce exposure measurement error in health
studies, aid in designing relevant interventions to reduce health disparities. The results may
inform appropriate interventions from urban planning perspectives, such as increasing greenness
and park area city wide to reduce personal PM2.5 exposures, which will potentially benefit both
the mother’s health and the child’s health at birth and beyond.
The final chapter of the dissertation concludes the work by highlighting the major
findings and implications of the three studies as a whole.
8
Chapter 2 The Impact of GPS-derived Activity Spaces on Personal PM2.5
Exposures in the MADRES Cohort
This chapter is focused on investigating how exposures encountered within the activity spaces, as
well as the time activities, home characteristics and residential neighborhood exposures,
contributed to the personal PM2.5 exposures during the 3
rd
trimester of pregnancy among
MADRES participants. The chapter starts by introducing the research background, then followed
by the data and methods used in this study, along with results, discussions and conclusion.
2.1. Introduction
Air pollution is a significant risk factor for various adverse health outcomes including
respiratory infections (Kim et al. 2004), asthma (Brandt et al., 2014), cardiovascular disease
(Gan et al. 2011), diabetes (Chen et al., 2016), and increased mortality (Bell and Davis 2017;
Dockery et al. 1993; Garcia et al. 2016), among others. Studies of health impacts in pregnant
women show air pollution exposure affects the mother’s health (Dadvand et al. 2014; Ghosh et
al. 2014) and may result in adverse birth outcomes (Fleischer et al. 2014; Pereira et al. 2014;
Rich et al. 2015; Ritz et al. 2007). Third trimester exposure to PM2.5 has been associated with
low fetal birthweight (Hyder et al. 2014; Stieb et al. 2016; Twum et al., 2016; Zhu et al. 2015)
and adverse health effects in childhood (Dadvand et al. 2011; Hsu et al. 2015; Rosa et al. 2020).
Environmental health disparities also play a role, with specific racial or ethnic groups and lower
socioeconomic status groups disproportionately exposed to higher concentrations of air pollution
(Bae et al. 2007; Houston et al. 2004; Tian et al., 2013). In turn, these disparities are linked with
increased susceptibility to multiple adverse health effects including obesity (Rossen, 2014),
diabetes (Alderete et al. 2017; Chen et al. 2016), and respiratory outcomes (Grineski et al. 2015).
9
Most epidemiological studies rely on ambient concentrations to represent individuals'
personal exposures to air pollution of outdoor origin (Gauderman et al. 2015; Pun et al. 2017).
These coarse resolution approaches vary from assigning the value of the nearest monitoring site
(Gauderman et al. 2015; Masiol et al. 2017) to spatially interpolated outdoor exposures based on
multiple monitoring sites (Berger et al. 2018; Zhai et al. 2017) or sophisticated spatiotemporal
models of outdoor concentrations (Beckx et al. 2009; Dadvand et al. 2013; Hu et al. 2015;
McGuinn et al. 2016; Weaver et al. 2019). Several studies have used the aforementioned
approaches to investigate the health effects of PM2.5 exposure during pregnancy on maternal and
child health outcomes, e.g. intrauterine inflammation (Nachman et al., 2016), stillbirth (Rammah
et al., 2019), low birth weight (Hyder et al. 2014; Li et al., 2019; Twum et al., 2016), and
childhood over-weight or obesity (Mao et al., 2017). While these models provide a cost-effective
way to assess exposure to outdoor, residential air pollution in large population studies, they
inherently suffer from exposure measurement error since they assume individuals are stationary,
and they do not account for exposures encountered within activity spaces while mobile. They
also ignore pollution sources in the indoor environment and time-activity patterns (i.e., time
spent indoors or in transit).
Activity spaces represent “the local areas within which people move or travel in the
course of their daily activities” (Gesler & Albert, 2000). Environmental exposures within these
“local” areas or activity spaces are thought to be more correlated with personal exposures since
they are more aligned with where and how individuals have contact or interact with their
environments. As such, activity space methods provide promising advances in the field of
environmental exposure science to understand health impacts (Golledge, 1997; Sharp et al.,
2015; Sherman et al., 2005; Tamura et al., 2017; Wang et al., 2018). Several studies have used
10
activity space methods to account for spatiotemporal variations in pollution in relation to an
individual’s whereabouts by recording when and where an individual move and how long they
stay at one place or spend in transit (Gerharz et al., 2009; Goodchild, 2007; Nazelle et al. 2013;
Nyhan et al. 2016; Steinle et al., 2015; Zenk et al. 2011). Several studies have correlated
environmental features in the residential neighborhood (i.e., road network, green space) with
personal exposures (Dadvand et al., 2012b; Kim et al., 2004). However, very few studies to date
have investigated the role of activity space-based exposures on personal PM2.5 exposures.
Moreover, personal monitoring can be used to measure air pollutant concentrations in the
breathing zone and accurately assess total, personal PM2.5 exposure (Dadvand et al. 2012b; Majd
et al. 2018; Shang et al. 2019). Personal monitoring is considered the gold standard external
exposure assessment approach since it captures personal, indoor, and outdoor sources of air
pollution encountered across activity spaces and within microenvironments based on actual time-
activity patterns. As such, it greatly reduces exposure measurement error and can increase
statistical power to observe health associations when they exist (Gray et al., 2011; Zeger et al.
2000). However, despite its advantages, personal monitoring studies have been limited since they
are generally more burdensome and expensive to conduct.
Among the few studies which measured personal PM2.5 exposure in pregnancy, Dadvand
et al. (2012b) monitored 54 participants for 48 hours and found higher residential neighborhood
greenness was associated with lower personal, home-indoor, and home-outdoor PM2.5 levels.
Greenness was also associated with more time spent at home, outdoors. In a study of 17 pregnant
women in the 3
rd
trimester, Zamora et al. (2018) found that personal PM2.5 exposure was
frequently more than double ambient concentrations, and the majority of PM2.5 mass came from
the indoor residential environment. Jedrychowski et al. (2006) found a significant positive
11
association between personal PM2.5 exposures and residential proximity to industrial plants in
407 non-smoking pregnant women in the 2
nd
trimester. Taken together, these studies show that
pregnant women’s personal exposure to PM2.5 can be impacted by a variety of factors including
indoor sources, time-activity patterns, and exposures encountered within residential
neighborhoods and activity spaces.
Therefore, this research project aimed to investigate how exposures encountered within
activity spaces contribute to PM2.5 exposures during the 3
rd
trimester of pregnancy, using highly
resolved personal exposure and geolocation monitoring data. The relationships between personal
PM 2.5 measurements and GPS-extracted activity space-based exposures and time-activity
patterns were first examined; then a model was built to explain the variability in personal PM2.5
exposure based on these as well as individual and residential neighborhood characteristics to
identify key exposure determinants in an environmental health disparities population of primarily
low-income, Hispanic pregnant women living in Los Angeles, CA.
2.2. Method
Personal and environmental data used in this research, along with the main analytical
methods, are laid out in this section.
2.2.1. Data Collection
Given the MADRES personal data used in this study, firstly the study design for
MADRES cohort is briefly introduced here, followed by the description of the personal data used
in this study including 48-hr personal PM2.5 measurements, the concurrent GPS tracks, and
questionnaire answers.
12
2.2.1.1. Study design
This study recruited 213 women who were enrolled in the MADRES cohort study during
their 3
rd
trimester visit for this intensive 48-hour personal PM2.5 exposure monitoring study
between October 2016 and March 2020 (Appendix A, Table S2.1). MADRES is an ongoing
prospective pregnancy cohort with the goal of understanding environmental and social
determinants of maternal and child health outcomes among predominantly low-income, Hispanic
women and their babies. The details of eligibility, enrollment, and follow-up in MADRES are
described elsewhere (Bastain et al., 2019). Here, the aspects related to this personal monitoring
arm of the larger study are briefly outlined. MADRES women were eligible to participate if they
were in the 3
rd
trimester at the time of recruitment, ≥18 yrs old, and could speak either English or
Spanish fluently. Exclusion criteria included: (1) HIV positive status; (2) physical, mental, or
cognitive disabilities that prevent participation; (3) current incarceration; or (4) living in a
smoking household, defined as having at least one smoker living full-time in the same residence
as the pregnant woman. In practice, the non-smoking household criterion was not applied
consistently throughout the study and thus was eliminated. Informed consent was obtained for
each participant. The University of Southern California’s Institutional Review Board (IRB)
approved the study protocol.
2.2.1.2. Personal PM2.5 exposure and geolocation monitoring
Once consented, participants were asked to wear a customized crossbody purse that
contained all the sampling devices for 48 hours as they conducted their usual daily activities. The
purse contained a personal Gilian Plus Datalogging pump (Sensidyne, Inc.) that was
programmed to sample air continuously through an inlet tube at a 1.8 LPM flow rate and a 50%
cycle (starting midnight on day following recruitment till midnight of the second day of
13
sampling, once 48 hours were completed). The tube was connected to a PM2.5 Harvard Personal
Environmental Monitor (PEM) size-selective impactor with a pre-weighed 37 mm Teflon filter
loaded inside (2 µm pore size; Pall, Inc.) to collect a 48-hour integrated (or averaged) sample.
The PEM sampling inlet was mounted on the purse’s shoulder strap to sample air at the
participant's breathing level. Pumps were flow calibrated with the specific PEM sampler prior to
each deployment using a TSI Inc. flow meter. Participants were instructed to carry the purse and
sampling apparatus with them at all times, with a few exceptions to reduce burden and improve
wear compliance. These included when it was unsafe to do so (e.g., driving), while showering or
in high humidity, while sleeping, or while staying in one physical room for a prolonged period.
In these cases, they were instructed to place the sampler near them, elevated above ground level,
away from walls, and unobstructed by any objects.
In addition, an Android smartphone was included in the purse with the madresGPS
geolocation app pre-installed and programmed to log location (GPS and metadata) and motion
sensor data and network connectivity status continuously at 10-sec intervals for the 48-hour
monitoring duration. The madresGPS app logs timestamp, latitude and longitude, location
accuracy (m), and source of location (i.e., smartphone GPS sensor or network (WiFi or cellular)).
The GPS source provided altitude (m), velocity (m/sec), number of satellites in view and in use
when available. Smartphones were connected to a power bank to ensure enough power for a 48-
hr runtime. All sampling devices were demonstrated to participants at recruitment and then
securely sealed in a dedicated section of the purse to prevent loss or damage.
Once the 48-hour monitoring period was completed, the sampling pump shut down and
GPS app stopped logging data automatically. Trained bilingual staff arranged a home pickup
visit usually on the following day to retrieve the sampling devices and complete a short exit
14
survey with participants (described below). PEMs were disassembled and pump and GPS app
data were downloaded (and decrypted in the case of GPS) on the same day in the USC Exposure
Analytics Laboratory. Filters were then analyzed gravimetrically to determine post-sampling
PM 2.5 mass after a minimum of 24 hours equilibration period using a MT5 microbalance (Mettler
Toledo, Columbus, OH, USA) in a dedicated chamber.
2.2.1.3. Questionnaires
At enrollment and during the 1
st
, 2
nd
and 3
rd
trimesters, MADRES participants responded
to interviewer-administered questionnaires during in-person visits or phone interviews by trained
bilingual staff. Questions included demographics (age, race, education, marital status, household
income, country of origin), housing characteristics such as type of dwelling and building age,
indoor sources such as presence and use of gas stoves, heating, and current tobacco smoke
exposure (primary and secondhand). Participants' residential locations were determined based on
reported address at the 3rd trimester timepoint and geocoded for residential neighborhood
exposure assessment.
Once the participants completed the 48-hour monitoring period, trained staff conducted
an exit survey during the equipment pick up visit asking about sampling device wear times, time-
activity patterns (e.g., time spent indoors and outdoors, commuting), home operation (e.g.,
ventilation), and presence of any significant indoor sources of PM2.5 such as cooking or smoking
during either day of the 48-hr sampling period. Variables were summarized as the maximum or
largest response across both days for all questions. Exposure to secondhand smoke was
determined based on a response of “a little”, “most” or “all of the time” to the following
question: “How much of the time were you close to cigarette, cigar, hookah or pipe smoke from
people smoking nearby”. Spending time outdoor near traffic was determined based on a response
15
of “sidewalk along the road” or “parking lot” to the following question: “Where were you when
you were outdoors in general”.
2.2.2. Data Analysis
Data analysis is laid out in this sub-section, starting by creating residential neighborhood
exposures, followed by activity space-based exposures, and GPS-derived time-activity patterns.
2.2.2.1. Residential neighborhood environmental exposures
Residential neighborhoods were defined as areas including the residential location and its
surroundings in several ways since the exact spatial extent of influence in not well known in the
literature. These included the residence as a point location, the residential census tract (RN_ct),
and three circular buffers of 100m, 250m, and 500m around the residence (RN_100m,
RN_250m, and RN_500m).
First, daily ambient and traffic-related air pollution and meteorology were estimated at
the residential point location (Bastain et al., 2019). Local traffic-related nitrogen oxides (NOx)
exposure was estimated using the CALINE4 line source dispersion model (Benson, 1992).
Nitrogen dioxide (NO2), PM2.5 and PM 10 (particulate matter with aerodynamic diameter less than
2.5 µm and 10 µm, respectively), and ozone (O3) concentrations were estimated using inverse-
distance squared spatial interpolation of regulatory monitoring data from the USEPA Air Quality
System (AQS). Meteorology (temperature, precipitation, specific humidity, relative humidity,
downward shortwave radiance, and wind speed) was assigned based on a 4 km x 4 km gridded
reanalysis model from Abatzoglou (2013). To correspond to the two sampling days, 48-hr
integrated averages were calculated from these daily measurements.
In addition, several built-environment characteristics were assessed within the census
tract and circular buffer defined residential neighborhoods during the 48-hr monitoring period,
16
including walkability index score, Normalized Difference Vegetation Index (NDVI, the most
commonly-used metric to quantify greenness), access to parks and open spaces, traffic volume
on primary roads, and road lengths (primary roads, secondary roads, and local neighborhood
roads and city streets). These geospatial data sources used for Los Angeles County are shown in
Table 2.1. Road network data were categorized as primary roads (S1100, Interstate highways,
and all other highways with limited access), secondary roads (S1200, main arteries and highways
with at-grade intersections), local neighborhood roads and city streets (S1400, paved non-arterial
street, road, or byway, abbreviated as minor streets)
(https://www2.census.gov/geo/pdfs/reference/mtfccs2018.pdf) (Figure S2.1 shows a map of
these three road classes in Los Angeles, CA).
2.2.2.2. Activity space-based environmental exposures
Activity spaces were constructed using the GPS 10-sec resolution data to examine how
participants’ mobility within 48-hrs affected their personal PM2.5 exposures (Browning and
Soller 2014; Crawford et al. 2014; Perchoux et al. 2016; Sherman et al., 2005; Tamura et al.,
2017). GPS tracks were first processed to remove outliers or erroneous records and retain those
with highest positional accuracy, especially when latitude and longitude were available from
both GPS and network sources. Distance-based outliers were defined based on a maximum
reasonable distance (100 mile/hour) traveled per time elapsed (using a threshold of 45 m/sec
multiplied by time elapsed) and were replaced by the median location (latitude and longitude)
within a moving, centered time window corresponding to approximately one minute (seven
intervals).
Similarly, since the exact spatial extent of influence on personal PM2.5 exposures is not
known, three measures of activity spaces were constructed to examine which might correlate the
17
most as follows: 1) Minimum Convex Hull (MCH) or the smallest area covering all the GPS
points, 2) Daily Path Area (DPA) which focuses on the area along individuals’ routes by
buffering all GPS points to 500 m, and 3) Kernel Density Estimation (KDE) which focuses on
the intensity of GPS points in space. KDE therefore implicitly accounts for duration of time
spent at a certain location, since GPS points will be denser or more intense where participants
spent most time with this equally spaced 10-sec GPS data resolution (Jankowska et al., 2015;
Kwan 1999; Newsome et al., 1998; Sherman et al. 2005; Zenk et al. 2011). KDE was applied
with multiple bins (i.e., 10, 25, 50 m) and neighborhood sizes (i.e., 100, 250, 500 m) to examine
the suitable parameters in terms of its Impact on personal PM2.5 exposures (referred to as
K10/100m; K10/250m; K25/250m; K25/500m; K50/500m). The top 20th percentile area of
intensity in each KDE activity space was also used to calculate exposures as a test of whether
this might be adequately correlated with personal exposure and computationally simpler
compared to using the entire KDE surface (i.e., K10/100m20p; K10/250m20p; K25/250m20p;
K25/500m20p; K50/500m20p).
The same built-environment characteristics (Table 2.1) and ambient PM2.5 and
temperature were also assessed within the activity spaces. Forty-eight-hour average ambient
PM 2.5 concentration and temperature were estimated for 2016-2020 using Empirical Bayesian
Kriging spatial smoothing to complement the inverse distance squared method described earlier.
Figure 2.1 illustrates how different activity space and residential neighborhood methods
are used to calculate exposure along a theoretical GPS trajectory. The blue boundary shows an
example MCH activity space, and the dark green boundary shows the DPA activity space. The
light to dark green (10m bin, 100m neighborhood), blue (25m bin, 250m neighborhood), and
18
\
Table 2.1. Summary of geospatial data sources for residential neighborhood and activity space-based exposure assessment.
19
orange (50m bin, 500m neighborhood) weights correspond to lowest to highest intensity within
multiple KDE areas based on locations an individual spent the most time in. The top 20
th
percentile area of each KDE activity space is illustrated as weight 5 in Figure 2.1. Dark gray
circles (buffers) and polygon (census tract) show the four residential neighborhood definitions
surrounding the residential point location. Residential neighborhood and activity space-based
exposures were derived in ArcGIS Pro 2.5 (Esri, Redland, CA).
Figure 2.1. Illustration of multiple activity spaces and residential neighborhood
2.2.2.3. GPS-derived time-activity patterns
Time spent indoors or in-transit was derived from the processed GPS data using a
previously published method that depends on time and distance thresholds (Cich et al. 2016; Li
et al. 2008; Pérez-Torres et al., 2016; van Dijk 2018; Xiao et al. 2014). It identifies whether a
participant was stationary or moving, as well as the duration of each activity or trip. The
minimum time interval for a stay was defined as 30 mins. If the distance moved within 30 min
was less than 500 m, the participant was identified as staying in one place. Once the stay points
20
were identified, they were classified as home or non-home places by comparing them with the
participants' known residential location with a 75 m buffer threshold to account for potential
noise in geolocation. Twenty-eight participants did not stay overnight at their own residences
during the 48-hr monitoring period, so the place where they stayed from midnight to early
morning was determined to be their "home" place for the sake of deriving time-activity patterns.
Time spent at home was assumed to be indoors and calculated in minutes and in percent (%)
time. Categories of % time indoors were then created based on the data distribution (≤75%, 75 %
to ≤90%, 90% to ≤95%, 95% to ≤98%, >98%) for use in the analysis.
2.2.3. Bivariate Analyses
Bivariate analysis were conducted to screen and select variables for the final regression
model. The Kruskal-Wallis test was used to examine correlations of categorical variables with
personal PM2.5, including time-activity patterns, home characteristics, and indoor air pollution
sources. Some variables were dichotomized or recoded to ensure more balanced and physically
interpretable categories as follows: house vs. apartment, house built before vs. after 1980s, none
or little of time vs. most or all of time for window open, none vs. a little, most, all of time of air
conditioner used at home, less than vs. more than 75% of 48-hr period staying indoors, none or a
little vs. most or all the time spent outdoors, less than vs. more than 30 mins gas stove use on
daily basis, none vs. a little, most, all of time close to smoke from people smoking nearby, none
vs. a little, most, all of time close to smoke from candles or incense burning nearby, and 0-30
mins vs. 30 mins to 1 hr vs. 1-2 hrs vs. more than 2 hrs in terms of commuting time. Sixteen
variables with unbalanced values (≥85% of the records have one value) or too many missing
values (>80%) were dropped from further analysis.
21
Spearman correlations were used to screen continuous variables such as residential
neighborhood and activity space-based exposures for inclusion in the final PM2.5 model and to
evaluate them for collinearity with each other. Variables with absolute correlation > 0.05 or p-
value < 0.25 in the bivariate analyses were retained for subsequent multivariate analysis. In
addition, variables previously reported in the literature as important determinants of personal
PM 2.5 exposure were also retained, including wind speed, relative humidity, year home was
originally built, gas stove usage, secondhand smoking, and park area within activity space.
2.2.4. Multivariate Model
Generalized linear models were fit to explain the variability in personal PM2.5 mass
exposure in relation to multiple variables. These included time-activity patterns (Time-Activity),
demographics (Demographic), home characteristics (Home), indoor sources (Indoorsources),
environmental exposures within residential neighborhoods (EnvExpRN), and environmental
exposures within activity spaces (EnvExpActSp). The model structure was as follows:
𝑌 . = 𝛽
+ 𝛽 ∗ 𝑇𝑖𝑚𝑒 − 𝐴𝑐𝑡𝑖𝑣𝑖𝑡𝑦 + 𝛽 ∗ 𝐷𝑒𝑚𝑜𝑔𝑟𝑎𝑝 ℎ𝑖𝑐 + 𝛽 ∗ 𝐻𝑜𝑚𝑒 + 𝛽 ∗ 𝐼𝑛𝑑𝑜𝑜 𝑟 + 𝛽 ∗ 𝐸 𝑛𝑣𝐸𝑥𝑝 + 𝛽 ∗ 𝐸𝑛𝑣𝐸𝑥𝑝 + 𝜀
Eq. (2.1)
where β0, β (a to f), and ε represent the intercept, coefficients, and error terms, respectively.
Variables selected in bivariate analyses were added to the model one at a time and
retained if they were still significant at p<0.1 level. Variables that were highly correlated (or
collinear) with each other, such as several measures of primary road length in activity spaces,
were treated as alternative factors and substituted into the multivariate model to select the most
22
significant. After the final list of variables was selected using this manual process, forward
stepwise regression with the Sawa Bayesian Information Criterion (BIC) selection criteria was
adopted for building the final model. Parameter estimates for all continuous variables were
scaled to a one SD increase for comparison. BIC criteria was used since it penalizes the addition
of more terms to the model to avoid overfitting. The adjusted R
2
and Root Mean Square Error
(RMSE) were used to examine the fit of the model and p-value (Type III) was used to examine
significance of included variables. All statistical analyses were conducted in SAS 9.4 (SAS
Institute Inc., Raleigh, NC), and plots were generated using JMP Pro 16.1 (SAS Institute Inc.,
Raleigh, NC).
2.3. Results
2.3.1. Descriptive Statistics
Most of the participants (>98%) resided in central and east Los Angeles, CA. The
majority were Hispanic (79%), employed during the 3
rd
trimester (41%), and with up to grade 12
education (54%). The mean age was 28 years at consent (range 18-45 years), and mean birth
order of index child at the time of pregnancy (i.e., parity) was 2 (range 1-6). The majority of
participants reported annual household incomes less than $30,000 (67%, N=135) (Table 2.2). In
terms of personal monitoring device wearing compliance during 48-hr sampling period, 192
participants (91%) reported wearing it most of the time while awake, 183 (86%) reported putting
it nearby as instructed while sleeping, and 200 (94%) reported putting it nearby as instructed
when not wearing it during the day time (Table S2.2).
Table 2.3 presents distributions of home characteristics, indoor PM sources, and
durations of selected activities for the participants. Based on the exit survey referring to the 48-hr
sampling period, 60% of participants opened windows more than half of the time, 63% spent
23
Table 2.2. Descriptive statistics of participant demographics (N=213).
Variable
Mean (SD) or
n (%)
Variable
Mean (SD) or
n (%)
Maternal Age (years) 28.3 (6.0) Employment
Birth order of index child at time of
pregnancy
2 (1.2 ) Homemaker 58 (27.2%)
Race
Student 21 (9.9%)
White, non-Hispanic 12 (5.6%) Employed 87 (40.8%)
Asian, non-Hispanic 2 (0.9%) Temporary Medical Leave 9 (4.2%)
African American, non-Hispanic 24 (11.3%) Unemployed 35 (16.4%)
Hispanic 169 (79.3%) Missing 3 (1.4%)
Other 4 (1.9%) Household income in the last year
Missing 2 (0.9%) Less than $15,000 44 (20.7%)
Education
$15,000 to $29,999 47 (22.1%)
< 12th grade 51 (23.9%) $30,000 to $49,999 29 (13.6%)
Completed high school 65 (30.5%) $50,000 to $99,999 7 (3.3%)
Some college 63 (29.6%) $100,000 or more 8 (3.8%)
Completed college 25 (11.7%) Don't know 76 (35.7%)
Some Graduate school 7 (3.3%) Missing 2 (0.9%)
Missing 2 (0.9%)
little or no time outdoors, 61% spent some time near traffic, and 34% spent more than 2 hrs
commuting. In terms of indoor PM sources, 83 (39%) were exposed to smokers, and 52 (24%)
were close to burning candles or incense. Based on the 3
rd
trimester questionnaire, 57% of
participants lived in an apartment, 45% had a household size > 3 persons, and 43% lived in a
home built after the 1980s. In addition, 139 (65%) used stoves > 30 mins/day at home during the
3
rd
trimester.
Summary statistics of 48-hr integrated personal PM2.5 exposure and modeled outdoor
PM 2.5 at residential location and within some activity spaces are shown in Table S2.3. Overall,
48-hr personal PM2.5 exposures (mean = 23.3 μg/m
3
, SD = 18.9) were much higher and more
variable than corresponding outdoor residential levels (mean = 11.8 μg/m
3
, SD = 5.5).
Approximately 25% had personal exposures two to four times higher than outdoor residential
PM 2.5. Outdoor PM 2.5 within multiple activity spaces was very similar to residential PM 2.5, which
24
was also much lower compared to personal PM2.5. Figure 2.2 shows the relationship between
personal and outdoor PM2.5 at residential location.
Table 2.3. Home characteristics, indoor sources, and durations of selected activities
derived from questionnaires and exit survey (N=213).
Home Characteristics n (%) Indoor Air Pollution Source n (%)
*Which best describes the home in which
you currently live most of the time?
** How much of the time were you close to smoke from
candles or incense burning nearby?
House 75 (35.2%) None of the time 160 (75.1%)
Apartment 122 (57.3%) A little, most, or all of the time 52 (24.4%)
Missing 16 (7.5%) Missing 1 (0.5%)
*How many people counting yourself live in
your household?
*About how long is the gas stove, range or oven used on
an average day while you are at home?
1 and 2 people 26 (12.2%) Less than 30 minutes 40 (18.8%)
3 people 30 (14.1%) More than 30 minutes 139 (65.2%)
4 people 40 (18.8%) Missing 34 (16.0%)
5 people 20 (9.4%)
**How much of the time were you close to cigarette,
cigar, hookah or pipe smoke from people smoking nearby?
More than 5 people 35 (16.4%) None of the time 128 (60.1%)
Missing 62 (29.1%) A little, most, or all of the time 83 (39.0%)
*About when was this home building
originally built?
Missing 2 (0.9%)
Built after 1980s 91 (42.7%) Time-Activities
Built before 1980s 69 (32.4%)
**How much of the time did you spend outdoors (not
commuting in a car, bus or train)?
Missing 53 (24.9%) None or a little of the time 135 (63.4%)
*Is there carpeting in your home? Most or all of the time 77 (36.2%)
No 106 (49.8%) Missing 1 (0.5%)
Yes 92 (43.2%) **When outdoor, whether were you near traffic?
Missing 15 (7.0%) No 82 (38.5%)
Home Ventilation Yes 130 (61.0%)
** How long the window open in your
home?
Missing 1 (0.5%)
None or little of the time 85 (39.9%) **How many hours did you spend on commute?
Most or all of the time 127 (59.6%) 0 to 30 min 56 (26.3%)
Missing 1 (0.5%) 30 min to 1 hr 47 (22.1%)
**How much of the time was the air
conditioner used in your home, when you were
there with the sampler?
1 to 2 hrs 39 (18.3%)
None of the time 157 (73.7%) > 2 hrs 41 (19.2%)
A little, most, or all of the time 55 (25.8%) Missing 30 (14.1%)
Missing 1 (0.5%)
* From the 3
rd
trimester questionnaire; ** Reported or derived from exit survey referring to 48-hour
monitoring period.
25
Figure 2.2. Regression plot between 48-hour integrated personal PM 2.5 exposures
and outdoor PM2.5 at the point of residence
2.3.2. Bivariate Analyses
Starting with questionnaire/exit variables, the bivariate results with personal PM2.5 are
shown in Table S2.4. The top five variables significantly associated with personal PM2.5
exposure included the following (sorted by p-value): (a) home type; (b) home carpeting; (c) time
spent close to smoke from candles or incense burning; (d) education level; and (e) number of
people living in a household, parity and time spent outdoors. Living in an apartment (compared
to a house), proximity to smoke from burning candles or incense, and spending more time
outdoors was associated with higher personal exposures. Participants with 5+ people or more
children in their household had higher personal exposures compared to less occupants or less
children at home.
The descriptive statistics and correlation coefficient for environmental exposures within
residential neighborhoods and activity spaces and time-activity patterns with personal PM2.5
exposure are shown in Table 2.4. The top 3 variables positively associated with personal PM2.5
26
Table 2.4: Bivariate results of personal PM2.5 exposures with GPS-derived time activities and
environmental variables.
Variables N Mean (SD)
Spearman Correlation
( p-value)
Time-Activity
***Time spent indoors 199 2,569.5 (523.4) -0.18 (0.009)
Residential Neighborhood Exposure
Air pollutants
PM2.5 (µg/m
3
) 209 11.8 (5.5) 0.09 (0.206)
O3 (ppb) 209 24.8 (8.3) -0.14 (0.037)
NO2 (ppb) 209 17.3 (8.6) 0.15 (0.031)
Freeway and highway traffic-related NOx (ppb) 204 1.8 (1.8) 0.04 (0.530)
Meteorology
Downward shortwave radiance (w/m
2
) 209 224.6 (83.0) -0.17 (0.014)
Relative humidity (%) 209 60.2 (12.4) -0.10 (0.167)
Wind speed (m/s) 209 2.4 (0.7) -0.05 (0.439)
Greenness (NDVI) and Parks
Mean NDVI within RN_100 m 213 -0.02 (0.04) -0.05 (0.429)
Total park area within RN_250m 213 2957.6 (8166.1) 0.10 (0.147)
Road lengths and traffic volume
Primary roads within RN_250 m 213 39.4 (115.1) 0.10 (0.140)
Secondary road within RN_ct 213 75.3 (281.9) -0.06 (0.396)
Minor streets within RN_500 m 213 339.0 (102.7) -0.19 (0.006)
Mean traffic volume within RN_250 m 213 9535.2 (45898.4) 0.04 (0.589)
Built environment exposures
Mean WIS within RN_250 m 213 14.4 (2.0) -0.12 (0.087)
Activity Space Exposure
Air pollutants
Mean outdoor PM2.5 within KDE area (K10/100m) 199 11.4 (5.5) 0.04 (0.530)
Meteorology
Mean daily temperature within KDE area (K50/500m20p) 199 18.3 (4.3) -0.15 (0.030)
Greenness (NDVI) and Parks
Mean NDVI within KDE area (K25/250m) 199 -0.1 (0.1) -0.15 (0.037)
Mean park area within DPA 199 28,390.4 (59,052.1) -0.06 (0.388)
Road lengths and traffic volume
Primary road within KDE area (K50/500m) 199 425.7 (865.9) 0.12 (0.094)
Secondary road within DPA 199 453.6 (613.5) 0.04 (0.571)
Minor streets within KDE area (K10/100m) 199 126.7 (38.3) 0.10 (0.161)
Mean traffic volume within DPA 199
181,777.6
(108,620.1)
0.14 (0.044)
Built environment exposures
Mean WIS within KDE area (K10m/100n) 199 15.3 (2.1) -0.16 (0.027)
Nitrogen dioxide (NO2), ozone (O3), particulate matter with aerodynamic diameter less than 2.5 µm (PM2.5),
daily path area (DPA), kernel density estimation (KDE), residential neighborhood (RN).
*** From 48-hour GPS tracks. Values presented in bold font shows significant p-values at p<0.05 level.
27
(sorted by descending r value) were: (a) NO2 at residential location, (b) mean traffic volume
within DPA, and (c) primary road lengths within KDE area. The top 3 most negatively correlated
variables with personal PM2.5 were: (a) minor street lengths within RN_500m; (b) time spent
indoors, and (c) downward shortwave radiance. PM2.5 at residential neighborhood was more
strongly associated with personal PM2.5 compared to PM2.5 within KDE area.
To illustrate how different residential neighborhood versus activity space methods could
result in variable correlations with each other and with personal PM2.5 exposure, primary roads
were used as an example. Table S2.5 presents primary road lengths encountered by participants
in their activity spaces or residential neighborhoods, along with the bivariate relationships with
personal PM2.5. Primary road lengths within KDE (K10/250m, K25/250m, K25/500m,
K50/500m) activity spaces and within residential circular buffers (RN_250m, RN_500m) were
most significantly associated with personal PM2.5 (Spearman r 0.05 to 0.13). Table S2.6 shows
the Spearman correlations between various activity space measures of primary road lengths
ranging from low (blue) to high (red). The primary roads exposure variables which were most
significantly associated with personal PM2.5 were also highly correlated with each other (r > .5),
so only one was selected to include in the final model (based on lowest p-value as explained
earlier). Tables S2.7 and S2.8 (NDVI), and S2.9 (park area) show similar results for the
remaining activity space and residential measures. The final list of variables selected for
multivariate modeling is shown in Table 2.5.
28
Table 2.5. List of all potential variables considered for inclusion in the multivariate model.
Time-activity patterns Environmental exposures at residential neighborhoods
**Time spent outdoors O 3 (ppb)
**Time outdoor and near traffic NO 2 (ppb)
***Time spent indoors PM 2.5 (µg/m
3
)
**Average commuting time Relative humidity (%)
Demographics Downward shortwave radiance (w/m
2
)
*Education level Wind speed (m/s)
*Birth order of index child at time of pregnancy Mean length of minor streets within RN_500 m
Home characteristics Average WIS within RN_250 m
*Home type Environmental exposures within activity spaces
**Window open time Average NDVI value within KDE area (K25/250m)
*Household crowding Mean length of major road within DPA
*Home built year Sum length of freeway within KDE area (K50/500m)
*Home carpeting Mean traffic volume within DPA
**Air conditioner used at home Mean park area within DPA
Indoor sources Average daily PM 2.5 within KDE area (K10/100m)
**Time close to smoke from candles burning Average daily temperature within KDE area (K50/500m 20p)
*Average stove use time at home
**Having someone smoking nearby
Nitrogen dioxide (NO2), ozone (O3), particulate matter with aerodynamic diameter less than 2.5 µm (PM2.5), daily path area
(DPA), kernel density estimation (KDE), residential neighborhood (RN).
* From the 3
rd
trimester questionnaire; ** From exit survey referring to 48-hour monitoring period;
*** From 48-hour GPS tracks.
2.3.3. Multivariate Model
Table 2.6 summarizes the results of the final personal PM 2.5 model obtained with
stepwise linear regression. Variables referring to parity, home ventilation, environmental
exposures within selected activity spaces and residential neighborhoods, indoor sources, outdoor
environment, and time-activities were included in the final model. Among them, longer window
opening time, more greenness (higher NDVI) exposure within KDE area, longer duration of
staying indoors, greater park area experienced within DPA, and higher exposure to minor streets
within RN_500m were associated with lower personal PM2.5 exposures. Whereas, parity, primary
road exposure within the KDE area, outdoor PM2.5 at residence, secondary road exposure within
DPA, and candles or incense burning indoors increased personal PM2.5 exposures. Commuting
29
time was also included in the final model but seemed to have a non-linear relationship with
personal PM2.5. The final model (adjusted R
2
= 0.34 and intercept = 25.57) suggests that less than
half of the variability in personal PM2.5 mass was explained by all these factors. Figure 2.3 shows
the plot of measured versus predicted personal PM2.5 exposure based on the final model.
Table 2.6. Results of final generalized linear model of personal PM2.5 mass exposure.
Incremental model
performance once
variable added
Variable Parameter Estimate
*
Pr > |t| Pr > F BIC Adj. R
2
Intercept 25.57
1.000 1120.07 0.00
Parity 5.81
<.0001 1101.19 0.10
Window open time (on average in 3
rd
trimester)
0.002 1092.79 0.14
Less than half of the time ref
Half to all the time -5.48 0.027
Length of primary road within KDE area
(K50/500m) 2.82
0.005 1086.71 0.17
Average NDVI value within KDE area
(K25/250m) -3.09
0.018 1082.89 0.19
Average time of commute (in 48 hours)
0.013 1077.58 0.23
None ref
<= 1 hr -0.65 0.893
1-2 hrs 7.29 0.139
2-3 hrs -3.24 0.526
More than 3 hrs -7.62 0.154
Missing -1.36 0.803
Duration of staying indoors (in 48 hours)
0.021 1074.14 0.27
≤75%
ref
75% to ≤90%
-15.03 0.002
90% to ≤95%
-20.09 <.0001
95% to ≤98%
-9.99 0.029
> 98% -10.45 0.023
Outdoor PM2.5 concentration at residence 2.05
0.016 1070.72 0.29
Mean length of secondary road within DPA 5.57
0.043 1069.00 0.30
Mean park area within DPA -3.62
0.023 1066.42 0.32
Candles or incense burning (in 48 hours)
0.047 1065.05 0.33
No ref
Yes 5.69 0.036
Mean length of minor streets within RN_500m -2.53
0.040 1063.55 0.34
Particulate matter with aerodynamic diameter less than 2.5 µm (PM2.5), daily path area (DPA), kernel density estimation (KDE),
residential neighborhood (RN).
*Parameter estimates of all continuous variables are scaled to one SD increase as summarized in Table S2.10.
30
Figure 2.3. Measured versus predicted personal PM2.5 concentrations and linear regression fit
2.4. Discussion
In this study, 48-hr integrated personal PM2.5 measurements and concurrently recorded
continuous GPS data were leveraged to assess environmental exposures in activity spaces, derive
time-activity patterns, and investigate determinants of personal exposure among 213 pregnant
women in the 3
rd
trimester in Los Angeles, CA. Given the higher burden of collecting personal
monitoring data especially during pregnancy, this study provided a unique opportunity to
understand the multiple complex factors that contribute to personal PM2.5 exposure in this
environmental health disparities population. The novel approach revealed that the exposures
encountered within activity spaces, particularly greenness (NDVI), park area, and road lengths,
were the significant contributors to PM2.5 exposures. In addition, indoor environment, time-
activities, and outdoor PM2.5 at residential locations also affected the variation of exposures.
This study found that experiencing more park area and more greenness within
individuals’ activity spaces was associated with significant reductions in personal PM2.5
exposure. To our knowledge, this is probably the first study to document this direct relationship
31
between these natural and built environment features and personal PM2.5 exposure during
pregnancy. Most previous studies examining the impact of parks, green space or greenness on
exposure and health used relatively coarse approaches or data (Chen et al., 2019; Crouse et al.,
2019; Riggs et al., 2021; Son et al., 2021; Yitshak-Sade et al., 2019). For example, they used
PM 2.5 values from monitoring sites or modeling estimates to approximate personal PM2.5
exposures, calculated park area or average NDVI value within a fixed distance of residences,
then established connections at the population level. Among the few personal monitoring studies,
Dadvand et al. (2012b) examined the associations between personal PM2.5 samples of pregnant
women and surrounding greenness, represented by average NDVI within 100m/250m/500m
residential buffers. They also found higher residential greenness was associated with lower
personal PM2.5 exposures, with strongest relationship seen for 100m buffer. In this study, NDVI
exposure within multiple GPS-derived activity spaces were directly assessed that correspond to
where participants actually went and spent time. The results revealed that KDE activity space
greenness measures (with larger neighborhood sizes) were more strongly associated with lower
personal PM2.5 exposures compared to other activity spaces or residential neighborhood
measures. This is probably because KDE measures are most representative of exposures
experienced in space and time (of all the ones investigated). Others have also reported positive
associations between greener residential neighborhood and birth weight (Dadvand et al., 2012a;
Donovan et al., 2011). Taken together, the findings might suggest that increasing greenness in
places where pregnant women visit and stay could result in beneficial reductions to personal
exposure, which in turn might also improve physical health and well-being of both mother and
her baby.
32
Furthermore, the results also revealed a significant effect of spending time near roads on
personal PM2.5 exposures, where primary and secondary roads within activity spaces were
selected into the final model capturing potentially different aspects of road impacts on personal
exposures. Primary road lengths within KDE (K50/500m) – which is a space- and time-
integrated measure of being on or close to roads – was significantly associated with higher
personal PM2.5 exposure. In addition, mean length of secondary roads within DPA – a measure
which strongly correlates with any encounter of secondary roads (which have a very specific
geographic distribution in Los Angeles, CA, as shown in Figure S2.1) during participants’
movement or mobility – was also significantly associated with higher personal PM2.5 exposure.
These results show how different activity space measures with potentially variable spatial extents
could capture different aspects or contributions of the built environment to exposure.
This study found that the indoor environment has a large impact on personal exposure,
and probably one of the largest in relative magnitude, both in terms of indoor sources such as
combustion (e.g., candles or incense burning) and the number of children in the home
(approximated by birth order or parity). These results are in line with other exposure studies
showing indoor PM2.5 had significant contribution to personal exposures (Brown et al., 2009;
Kim et al., 2005; Koistinen et al., 2004; Lai et al., 2004; Zamora et al., 2018), or there was strong
correlation between personal and indoor PM2.5 (Adgate et al., 2002; Crist et al., 2008). The
results showed that indoor combustion source contributed twice as much as outdoor PM2.5
estimates to personal PM2.5 (on a per SD change basis in outdoor PM2.5). Given MADRES
participants spent around 94% of their time indoors (Table S2.11), the indoor environment
presumably dominated their personal exposures. Therefore, reducing indoor PM 2.5 sources could
greatly reduce personal PM2.5. Comparing to other studies which investigated indoor combustion
33
sources mainly from smoking or cooking (Brown et al., 2009; Kim et al., 2005; Meng et al.
2009; Wheeler et al., 2011; Zamora et al., 2018), the results only identified candles or incense
burning contribution to personal exposure. This could be because the survey measures did not
fully capture the presence of secondhand smoking (no primary smoking in these participants) or
cooking, or because the 48-hr sampling period did not always capture these if they occurred.
Future planned chemical analysis of these personal sampler filters will help us resolve PM2.5
sources and improve our understanding.
The results revealed that parity was more significantly associated with personal PM2.5
than household size although these two variables are significantly correlated (Spearman r=0.24).
The impact of multiple occupants in the home on personal exposure is less reported in the
literature. These findings could reflect the fact that children (compared to adults) tend to be more
active and stay in closer interaction with their mothers, or that mothers with more children
cooked or cleaned more frequently for example. In addition, as reported in the literature an effect
of window opening on reducing personal exposures was also found (Brown et al. 2009; Sarnat et
al., 2006). Window opening increases ventilation in the home and could dilute PM
concentrations emitted from indoor sources. It is also possible that window opening introduces
PM of outdoor origin indoors when outdoor air quality is poor; however, the personal
measurements were well spread over the sampling period which increases confidence in the
representativeness of this finding across seasons (Table S2.1).
Individual’s time-activities such as commuting and spending time near roads and traffic
(regardless of activity) also affected personal PM2.5 exposure. Previous studies also reported
commuting impact on personal PM2.5, with magnitude of influence highly dependent on
commute modes and ventilation settings (Ham et al., 2017; Huang et al., 2012; Kaur et al., 2007;
34
Qiu and Cao, 2020). The non-linear or non-monotonous relationship between commuting time
and personal PM2.5 exposures in this study could be due to the fact of insufficient data on in-
transit ventilation, commuting mode, or other factors known to modify exposure to PM2.5 in
transit. This study also found significant outdoor PM2.5 contributions to personal exposure, and
this is to be expected and highlights the rationale behind many studies of outdoor air pollution
health effects that are using outdoor residential estimates as proxies of personal exposure to
PM 2.5 of outdoor origin.
Finally, despite the sophisticated data collected in the research, the model did not explain
a large portion of the variability in personal PM2.5 exposure (Adj.R
2
= 0.34). This could be due to
several reasons. One important reason could be that organic carbon (OC) contributes a large
fraction of indoor and personal PM2.5 mass, and there are major sources of OC indoors (Habre et
al., 2014a, 2014b; Turpin et al., 2017). Turpin et al. (2007) found organic particulate matter was
the major constituent of PM2.5 generated indoors, which contributed 48% of PM2.5 mass inside
individual homes in Los Angeles. Habre et al. (2014a) attributed 46% of indoor PM2.5 mass to
indoor sources related to OC in New York. Other studies also confirmed large contributions of
OC or organic matter to indoor or personal PM2.5 (Hasheminassab et al., 2014; Habre et al.,
2014b; Schachter et al., 2016; Shang et al., 2019). This study was not able to measure OC in this
study using Teflon filters; therefore, a large portion of the PM mass could be missed this way
(and especially OC1, the most volatile thermal fraction of OC). Other reasons could relate to the
complexity of personal exposure and the multiple factors that contribute to it, where despite the
sophisticated data collection and modeling, other important determinants of exposure might not
have been captured. For example, there was not information on cleaning, vacuuming or dusting
which resuspend particles and dust and could have contributed to personal exposures (Habre et
35
al., 2014a; Hasheminassab et al., 2014; He et al., 2004; Koistinen et al., 2004; Long et al., 2000;
Molnár et al., 2014). Some home characteristics, e.g., type of residence, carpeting or AC usage,
were associated with personal PM2.5 in bivariate models (p < .15) but ended up dropping out in
multivariate model. Factors such as secondhand smoking and cooking in this study, which are
well-recognized as important contributors to indoor and person exposures (He et al., 2004; Hu et
al., 2012; Long et al., 2000), did not meet the bivariate screening criteria for multivariate
analysis, and this could depend on the form of questions used or other biases. Data on these
factors were collected in the questionnaires; nonetheless, if the question did not have enough
resolution or the data did not have enough variability to capture the real complexity of these
factors, the ability to model their full contribution to personal PM2.5 might be limited.
The strengths of this research include a study population selected from a highly
characterized prospective pregnancy cohort in a health disparities population, the 48-hr
integrated personal PM 2.5 monitoring and concurrent GPS data, and the sophisticated activity
space modeling to incorporate mobility and capture environmental impacts on personal
exposures. This rich dataset provided the ability to examine complex factors to understand
personal exposure. Some limitations include small sample size which is characteristic of personal
exposure studies that are more burdensome to conduct, no organic carbon measurements, and
perhaps reduced generalizability of the findings to other areas that do not resemble Los Angeles,
CA. However, the results may generalize to other environmental health disparities contexts and
studies. The 48-hour monitoring period in the 3
rd
trimester might also not be representative of the
full pregnancy or entire 3
rd
trimester exposure; however, the samples were somewhat evenly
spread out across seasons and years of the study.
36
2.5. Conclusion
The findings show that environmental exposures encountered within activity spaces,
along with indoor environment, time-activities, and outdoor PM2.5, significantly contribute to
personal PM2.5 exposure during pregnancy. Characterizing the impact of environmental
exposures and sources encountered in activity spaces and across microenvironments can shed
light on solutions and interventions to reduce personal exposures. Especially the finding of a
direct association between greater greenness exposure in the activity space and lower personal
exposure in the 3
rd
trimester of pregnancy need to be noted which could have direct relevance to
built-environment design and planning to promote health and well-being.
37
Chapter 3 Sources of Personal PM2.5 Exposure in the MADRES Pregnancy
Cohort
In this chapter, the main sources are first identified and their mass contributions to personal
PM 2.5 exposure of MADRES participants in their 3
rd
trimester of pregnancy are quantified. The
factors such as time-activity patterns, environmental exposures encountered within activity
spaces, home characteristics, and outdoor environment at the residence that were correlated with
these sources were examined next to further confirm their identities and understand their origin
(i.e., personal activity related, indoor origin, outdoor origin). The chapter starts by introducing
the research background, followed by the data and method used in this study, along with results,
discussion and conclusion.
3.1. Introduction
Epidemiological studies have shown that prenatal exposure to PM2.5 is associated with
adverse maternal and fetal health outcomes (Dadvand et al., 2013; Hu et al., 2015; Jedrychowski
et al., 2012). Exposure in the 3
rd
trimester of pregnancy specifically has been associated with low
birth weight and other impaired growth outcomes given this is the time when most fetal weight
gain occurs (Guo et al., 2018; Percy et al., 2019; Sun et al., 2016). The toxicity of PM2.5 and its
subsequent impact on health is driven by its chemical composition and main sources contributing
to it (Berger et al., 2018; Hasheminassab et al., 2014a; Masiol et al., 2017; Rohr & Wyzga, 2012;
Saffari et al., 2013; Stanek et al., 2011; Stieb et al., 2012; Sun et al., 2016; Zhai et al., 2017;
Zhang et al., 2008). Personal exposure to PM2.5 is impacted by indoor, outdoor, and personal
activity related sources in the various microenvironments individuals typically encounter. For
example, behaviors, time-activity patterns, and household, neighborhood or activity space
characteristics can impact the types and quantities of sources individuals are exposed to (Chen et
38
al., 2020; Larson et al., 2004; Shang et al., 2019). As such, identifying and quantifying the main
sources of personal PM2.5 can shed light on particular mixtures that might pose a greater risk and
would otherwise be missed by investigating exposure to total PM2.5 mass concentration as a
whole. This is particularly important in environmental health disparities contexts and for specific
vulnerable populations such as pregnant women for whom meaningful recommendations to
reduce exposures and health risks are needed (Brown et al., 2007; Han et al., 2017;
Hasheminassab et al., 2014a).
Earlier studies have resolved and quantified main sources of PM exposure using source-
and receptor-oriented modeling approaches. Source-oriented approaches start at the source and
model the emissions, transport, dilution, and transformation of pollutants and estimate
concentrations at receptor sites for one or more specific sources (Kim et al., 2005; Lippmann,
2009; Reff et al., 2009). Based on the fundamental mass balance principle (Watson et al., 2008),
the receptor-oriented approach utilizes speciated measurements at receptor sites or points of
interest to identify the major sources (or source groups) impacting that receptor and quantify
their respective contributions to the observed concentrations (Hasheminassab et al., 2014a,
2014b; Hopke, 2003). Two of the most commonly used receptor-oriented models are the
Chemical Mass Balance (CMB) model which assumes that the major sources impacting a
receptor site are known along with their profiles or chemical signatures (Fujita et al., 2003;
Harley et al., 1992; Hasheminassab et al., 2013; Schauer et al., 2002; Zhai et al., 2017) and the
PMF model which solves for and does not explicitly assume known sources and profiles (Berger
et al., 2018; Brown et al., 2007; Hadley, 2017; Han et al., 2017; Hasheminassab et al., 2014a,
2014b; Heo et al., 2009; Hopke, 2016; Masiol et al., 2017; Paatero & Tapper, 1994; Pekney et
al., 2006; Rohr et al., 2014; Song et al., 2001; Wang & Hopke, 2013).
39
And while several studies have derived outdoor air pollution sources (Arhami et al.,
2009; Cheung et al., 2011a, 2011b; Daher et al., 2013; Hasheminassab et al., 2013, 2014c; Heo et
al., 2013; Schauer et al., 1996; Sowlat et al., 2016) and investigated their health impacts (Bell et
al., 2010; Dadvand et al., 2014; Ng et al., 2017; Pereira et al., 2014; Rohr et al., 2014; Schachter
et al., 2016), very few studies have been able to accomplish this for personal exposure. For
example, Hasheminassab et al. (2013) and Hasheminassab et al. (2014a) resolved several sources
of outdoor PM (particle size range 0.25-10 µm) including vehicular emissions, wood smoke,
natural gas combustion, ship emissions, secondary aerosols, fresh and aged sea salt, and soil/road
dust. Through conducting concurrent indoor and outdoor PM sampling at three retirement
homes, Hasheminassab et al. (2014c) found that mobile sources were the major contributor to
both indoor (39±21%) and outdoor (46±17%) PM2.5 mass in Los Angeles, CA. However, sources
that contribute to personal exposures can be more complex or difficult to discern since
individuals get exposed to PM2.5 in multiple microenvironments and locations, while being
mobile or stationary, sometimes in close proximity to indoor or personal activity related sources
and while being impacted by outdoor or infiltrated air pollution (Jenkins et al., 1992; MacIntosh
et al., 2000; Ott et al., 2006; Wallace, 1996). A few studies have examined sources of indoor and
outdoor PM air pollution in residential settings, for example, in homes of children with asthma
(Habre et al., 2014a, 2014b). They found that risk of asthma symptoms in children varied by
PM 2.5 source (Habre et al., 2014b).
However, even fewer studies conducted source apportionment analyses on personal
monitoring samples, and most have ranged from 12 to 48 hours in duration (Brinkman et al.,
2009; Chen et al., 2020; Kim et al., 2005; Koistinen et al., 2004; Larson et al., 2004; Molnár et
al., 2014; Ryan et al., 2015; Shang et al., 2019). Personal monitoring is considered to be the gold
40
standard external exposure assessment method to accurately understand what individuals are
exposed to in their breathing zones (MacIntosh et al., 2000; Ott et al., 2006). Nevertheless, due to
the high cost and burden of collecting high quality personal exposure data, very few studies have
been able to conduct this type of monitoring especially in pregnant women (Choi et al., 2006,
2012; Jedrychowski et al., 2004, 2009; Rundle et al., 2012; Tonne et al., 2004), and even fewer
conducted source apportionment analyses on personal PM2.5 samples (Minguillón et al., 2012).
Özkaynak et al. (1996) found that personal exposure to PM10 in 178 nonsmoking residents in
Riverside, CA, was much higher than outdoor and indoor concentrations, and that these only
explained 16% and 50% of the variation in personal exposures, respectively. In addition, they
reported that cooking and smoking were important sources of personal exposure and that indoor
and outdoor measurements alone were not sufficient to fully capture variation in personal
exposure. Minguillón et al. (2012) found cosmetics and train/subway sources among others
contributed to personal PM2.5 exposures of 54 pregnant women with wide variation in
contributions across participants. They report that questionnaire data helped identify the
train/subway source, but limitations (e.g., recall error, accuracy of time and location of travel and
activities) could introduce noise when resolving the sources.
To the best of our knowledge, no studies to date have investigated sources of personal
PM 2.5 exposure in an environmental health disparities population during pregnancy. This study
aimed to understand the main sources and determinants of exposure for this specific vulnerable
population using data from a personal monitoring sub-study of the MADRES cohort in Los
Angeles, CA. MADRES aims to address critical gaps in understanding the impacts of air
pollution, environmental exposures, and social stressors on the maternal and child health in a
low-income, predominantly Hispanic women in urban Los Angeles (Bastain et al., 2019). To
41
accomplish this goal, the personal PM2.5 samples were analyzed first for chemical composition.
Source apportionment analysis was next conducted using the USEPA PMF model (Norris et al.,
2014) and the relationships between predicted source contributions and a suite of questionnaire-
collected and GPS-derived activity space and residential characteristics, personal behaviors, and
time-activity patterns were investigated to confirm source identities and understand what
contributes to their variation.
3.2. Method
In this section, the personal and environmental data used in this research are described
along with the USEPA-developed PMF model and the statistical analysis used to achieve the
research goals.
3.2.1. Data Collection
The study design for MADRES is briefly described first. Then the personal exposure data
of MADRES participants including personal PM2.5 measurements, concurrent GPS tracks,
questionnaires, and environmental exposures at residential locations and within GPS-derived
activity spaces, along with EPA speciated data, are described.
3.2.1.1. Study design
A total of 212 women in their 3
rd
trimester who were enrolled in the larger MADRES
cohort study were recruited into this personal monitoring sub-study between October 2016 and
March 2020. MADRES is an ongoing prospective pregnancy cohort focused on predominantly
low-income, Hispanic women and their babies residing in Los Angeles, CA. The details of
eligibility, enrollment, and follow-up of MADRES participants are described elsewhere (Bastain
et al., 2019). Briefly, eligible participants for this study were in the 3
rd
trimester at the time of
42
recruitment, ≥18 years of age, and could speak either English or Spanish fluently. In the initial
design, people living in a smoking household were excluded to reduce the impact from smoking
on personal PM2.5 exposures. However, the non-smoking household criterion was not applied
consistently throughout the study and was eliminated. Informed consent was obtained for each
participant. The University of Southern California’s Institutional Review Board (IRB) approved
the study protocol.
3.2.1.2. Personal PM2.5 measurements
Personal, 48-hr integrated PM2.5 measurements were collected using a Gilian Plus
Datalogging Pump (Sensidyne, Inc.) operating on a 50% cycle at 1.8 lpm flow rate and
connected to a PM2.5 Harvard PEM size-selective impactor with a 37 mm Teflon filter (2 µm
pore size; Pall, Inc.). Participants were asked to wear the sampling device for the entire data
collection period with a few exceptions. These included when it is unsafe to do so (e.g., driving),
showering, or sleeping, in which case they were instructed to place the device near them in an
unobstructed location.
Filters were analyzed gravimetrically to determine post-sampling PM2.5 mass using a
MT5 microbalance (Mettler Toledo, Columbus, OH, USA) in a dedicated chamber at the USC
Exposure Analytics Laboratory. Filters were then sent to Research Triangle Institute
International (RTI Inc., Research Triangle Park, NC) to determine elemental composition of 33
species using X-Ray Fluorescence (XRF). The chemical components included barium (Ba),
calcium (Ca), chlorine (Cl), copper (Cu), iron (Fe), potassium (K), magnesium (Mg), manganese
(Mn), sodium (Na), nickel (Ni), sulfur (S), silicon (Si), titanium (Ti), zinc (Zn), aluminum (Al),
bromine (Br), cobalt (Co), phosphorus (P), lead (Pb), selenium (Se), strontium (Sr), vanadium
(V), cesium (Cs), zirconium (Zr), chromium (Cr), rubidium (Rb), arsenic (As), indium (In),
43
silver (Ag), antimony (Sb), tin (Sn), cerium (Ce), and cadmium (Cd). Filters were also analyzed
for concentrations of black carbon (BC), brown carbon (BrC), and environmental tobacco smoke
(ETS) using a four-wavelength optical reflectance method (Lawless et al., 2004; Yan et al.,
2011).
3.2.1.3. Questionnaires
MADRES participants filled out interviewer-administered questionnaires in trimester-
specific visits and an exit survey after completing the 48-hr monitoring period. Data that might
directly or indirectly relate to PM 2.5 sources and personal exposures were collected, including
demographics (e.g., age, race, education, employment, income), pre-pregnancy body mass index
(BMI), housing characteristics (e.g., type of dwelling, building age), time-activity patterns (e.g.,
time spent indoors and outdoors, commuting), home ventilation (e.g., window open, air
conditioner use), current tobacco smoke exposure (primary and secondhand), and presence of
any significant indoor sources of PM2.5 such as cooking or candle burning (Bastain et al., 2019).
Participants' residential locations at the 3
rd
trimester study timepoint were geocoded for
residential exposure assessment.
3.2.1.4. Residential Environmental Exposure Assessment
Daily ambient concentrations of NO2, PM2.5, PM10, and O3 obtained from the USEPA
AQS were interpolated at the residence using inverse distance squared weighted interpolation
(Bastain et al., 2019). Daily local traffic-related NOx concentrations at the residence were
estimated using the CALINE4 line source dispersion model by roadway class (Benson, 1992).
Daily meteorology (temperature, precipitation, specific humidity, relative humidity, downward
shortwave radiance, wind direction and wind speed) was assigned at the residence based on a 4
km x 4 km gridded model developed by Abatzoglou (2013). Forty-eight-hour integrated averages
44
were calculated from all daily measurements to correspond to the personal monitoring period.
Specifically, wind direction was the average direction of degree in 48-hr period, where a
direction of 0 degrees is due North on a compass and a wind coming from the south has a wind
direction of 180 degrees. For analytical purposes, we categorized wind direction into four
categories as follows: 0-90 degrees as wind blowing from NE, 91-180 degrees as SE, 181-270
degrees as SW, and 271-360 degrees as NW.
3.2.1.5. GPS-Derived Time-Activity Patterns and Environmental Exposures within Activity
Spaces
Participants’ 48-hr GPS records were collected using smartphones with the madresGPS
app pre-installed and programmed to log geolocation (GPS and metadata) and motion sensor
data continuously at 10-sec intervals. Time-activity patterns were derived from analyzing GPS
records. Using the method described in Cich et al. (2016), Li et al. (2008), Pérez-Torres et al.
(2016), van Dijk (2018), and Xiao et al. (2014), durations of staying at home or other places
were extracted, as well as time on the road. All stays were assumed indoors and time spent
indoors in the 48-hr period were calculated in minutes then converted it to a percentage out of
the total 48 hours for use in the analysis.
KDE activity spaces were also constructed for each participant based on GPS trajectories
to examine how exposures encountered within correlated with sources, where KDE implicitly
integrates time and space to account for durations of time spent at certain locations (Jankowska
et al., 2015, 2017; Kwan, 1999; Newsome et al, 1998; Sherman et al., 2005; Zenk et al., 2011).
KDE was applied with pre-defined bin (i.e., 25 m) and neighborhood sizes (i.e., 250 m) to
examine the impact on personal PM2.5 exposures (i.e., K25/500m).
Built-environment characteristics including NDVI (greenness), parks and open spaces,
traffic volume on primary roads, walkability index scores, road lengths by categories (i.e.,
45
primary and secondary roads, and minor streets), ambient daily PM2.5 and temperature were
assigned for the KDE activity spaces (data sources described in more detail in Table 2.1).
Geospatial analyses for creating activity spaces, residential neighborhoods, and environmental
exposure data were conducted in ArcGIS Pro 2.5 (Esri, Redlands, CA).
3.2.1.6. EPA PM2.5 Speciation Data for Los Angeles, CA
Ambient PM2.5 speciated data was downloaded from the USEPA monitoring site located
in downtown Los Angeles. The concentration of these PM2.5 components are 24-hr averaged
values, which are collected every three days from the Chemical Speciation Network (CSN)
(Solomon et al., 2014). The data includes the measurement of the major chemical components of
PM 2.5 using the Met One SASS/SuperSASS Teflon - Energy Dispersive XRF method, including
carbonaceous material, and a series of trace elements.
3.2.2. Data Analysis
The analytical methods are laid out in this sub-section, starting with descriptive statistics
and followed by performing of the PMF analysis to identify main sources, as well as bivariate
analysis to further understand factors that influence the distribution of each source and help
confirm its identity or origin.
3.2.2.1. Descriptive Statistics
The descriptive statistics were calculated in SAS 9.4 (SAS Institute Inc 2013) to check
the distributions of population demographics, housing characteristics, home ventilation, indoor
sources of PM2.5, time-activities, personal PM2.5 mass and the measured chemical components
and optical carbon fractions.
46
3.2.2.2. Positive Matrix Factorization Analysis
The USEPA PMF 5.0 model was used to resolve and identify major sources of PM2.5 and
quantify their mass contributions using the measured chemical species concentrations and
sample-specific uncertainties as inputs. Briefly, the PMF model uses factor analysis to identify
source contributions and profiles for a given number of sources through solving the following
equation (Norris et al., 2014; Paatero & Tapper, 1994; Paatero, 1997):
𝑋 = 𝑔 𝑓 + 𝑒 Eq. (3.1)
where Xij represents the concentration of chemical species j in sample i, gik represents the mass
contribution of each factor k in sample i, fkj represents the loading of chemical species j on factor
k, and eij is the residual error for sample i and species j.
The PMF model solves Eq. (3.1) by minimizing the sum of squares object function Q for
a given number of factors k (Brown et al., 2015; Paatero & Tapper, 1994; Paatero, 1997):
𝑄 = [
𝑒 𝑢 ]
Eq. (3.2)
where uij is the uncertainty of species j in sample i. The model decomposes the concentrations
matrix into a contributions g matrix and profiles f matrix and constrains results to be positive (or
not significantly negative) (Brown et al., 2007; Paatero & Tapper, 1994). Each observation is
individually weighted by its uncertainty in Eq. (3.2); therefore, samples with higher analytical
uncertainties will have less influence on the solution.
Based on the PMF-calculated signal-to-noise ratio (S/N), which indicates the degree of
noise in each species’ measurements (Norris et al., 2014), we categorized species with S/N ≤0.2
as “Bad”, species with 0.2 < S/N <1 as “Weak”, and species with S/N > 1 as “Strong”. “Bad”
species were excluded from the subsequent analysis. “Weak” species were retained and used in
47
the analysis; however, their uncertainty values were increased by a factor of 3 to reduce their
impact on the solution. Although Pb and V had S/N < 0.2, they were included in the analysis as
potentially important tracers of traffic and fuel oil, respectively, and set to “Weak”.
Of the 36 species measured, the following 16 were finally included in the PMF analysis
as “Strong”: BC, BrC, Ba, Ca, Cl, Cu, Fe, K, Mg, Mn, Na, Ni, S, Si, Ti, and Zn. We also
included 9 “Weak” species as follows: Al, Br, Co, ETS, P, Pb, Se, Sr, and V. PM2.5 mass was
designated as the total variable which automatically defaults to “Weak” to reduce its impact on
the solution. An extra 10% modeling uncertainty was added in the model to account for sampling
or modeling errors not captured in the sample-specific analytical uncertainties (Norris et al.,
2014). In order to maintain sample size, missing values were replaced by the species’ median
value. Out of all available samples, 2.3% (5 out of 217) were excluded as outliers from the
analysis based on species’ concentrations.
The solutions with five to seven factors and 20 model runs were scanned first to decide
upon a reasonable factor number. The Q values for no undue influence from outliers and no local
minimum solution were checked next. Based on loading chemicals in profiles and prior
knowledge, the optimal sources from PMF that provided the most physically interpretable
solution were identified (Brown et al., 2007). Once the optimal factor number was decided, 100
model runs were executed and the convergent solution with the lowest Qrobust value, where Qrobust
is the calculated goodness-of-fit parameter excluding points with uncertainty-scaled residuals
greater than 4, was selected (Norris et al., 2014). Residuals were checked for normality, along
with R
2
values in terms of whether species were well modeled.
Diagnostics analysis of Displacement (DISP), Bootstrap (BS) (100 bootstraps, 0.6
minimum correlation), and Bootstrap-Displacement (BS-DISP) were performed to estimate the
48
variability in the PMF solution under different scenarios. DISP focuses on effects of rotational
ambiguity in the profiles or loadings; BS identifies whether there are a small set of observations
that can disproportionately influence the solution; and BS-DISP include effects of random errors
and rotational ambiguity (Norris et al., 2014). Fpeak rotations, where positive F peak values
sharpen the F matrix and negative values sharpen the G matrix were performed next. The
optimal Fpeak value for solution rotation was chosen based on the smallest change in Q (or dQ)
(Norris et al., 2014).
3.2.2.3. Bivariate analysis
To further confirm the identities and expected trends in the PMF-predicted source
contributions, the relationships with several variables described earlier including demographics,
time-activity patterns, home characteristics, indoor air pollutant sources, residential ambient air
pollutant concentrations and meteorological conditions, and environmental exposures within
activity spaces, were examined.
The descriptive statistics were calculated first and used to check the distribution of the
final, PMF-predicted and rotated source contributions for normality and outliers. The Spearman
correlations between the predicted source contributions (in mass concentration units) and
between the sources and variables hypothesized to relate to personal PM2.5 exposure from that
source were calculated next. The Kruskal-Wallis test was then used to test whether source
contributions were significantly different (rank test) across levels of categorical independent
variables. Categorical variables with unbalanced values (≥85% of the records have one value) or
with too many missing values (≥80% of the records have missing values) were excluded from
the bivariate evaluation and dropped from further analysis.
49
3.3. Results
3.3.1. Descriptive Statistics
Most of the participants (>98%) resided in central and east Los Angeles, CA. The
majority were Hispanic (78%), working (48%) during the 3
rd
trimester, and with up to grade 12
education (55%). The mean age was 28 yr at consent (range 18-45 yr), and mean parity was 2
(range 1-6). The majority of participants reported annual household incomes less than $30,000
(67%, N=135) and in terms of pre-pregnancy BMI, 63 participants (30%) were overweight and
82 (39%) were obese (Table 3.1).
Table 3.1. Descriptive statistics of participants demographics (N=212).
Variable Mean (SD) or n (%) Variable Mean (SD) or n (%)
Maternal Age (years) 28.3 (6.0) Maternal Ethnicity and Origin
Parity 2 (1.2 ) Non‐Hispanic 41 (19.3%)
Race
US‐Born Hispanic 75 (35.4%)
White, non-Hispanic 12 (5.7%) Foreign‐Born Hispanic 87 (41.0%)
Asian, non-Hispanic 2 (0.9%) Missing 9 (4.2%)
African American, non-Hispanic 23 (10.8%) Employment
Hispanic 166 (78.3%) Homemaker 57 (26.9%)
Other 4 (1.9%) Student 21 (9.9%)
Missing 5 (2.4%) Employed 84 (39.6%)
Education
Temporary Medical
Leave
9 (4.2%)
< 12th grade 50 (23.6%) Unemployed 35 (16.5%)
Completed high school 66 (31.1%) Missing 6 (2.8%)
Some college 59 (27.8%) Working Status
Completed college 25 (11.8%) No 106 (50.0%)
Some graduate training after college 7 (3.3%) Yes 101 (47.6%)
Missing 5 (2.4%) Missing 5 (2.4%)
Pre-Pregnancy Obesity Categories based on Body Mass
Index
Household income in the last year
Underweight 6 (2.8%) Less than $15,000 44 (20.7%)
Normal 57 (26.9%) $15,000 to $29,999 47 (22.1%)
Overweight 63 (29.7%) $30,000 to $49,999 29 (13.6%)
Class 1 Obese 51 (24.1%) $50,000 to $99,999 7 (3.3%)
Class 2 Obese 18 (8.5%) $100,000 or more 8 (3.8%)
Class 3 Obese 13 (6.1%) Don't know 76 (35.7%)
Missing 4 (1.9%) Missing 2 (0.9%)
50
Chemical component concentrations are provided in Table 3.2. The mean and SD
personal PM2.5 mass concentrations during the 48-hr sampling period were 22.3 and 16.6 μg/m
3
,
respectively. The optical carbon fractions BC, BrC, and ETS combined constituted on average
17% (3.7 μg/m
3
) of the total PM 2.5 mass. Among the elemental components measured, S, Na, Si
were presented at the highest concentrations.
Table 3.2. Chemical component concentrations (all in units of ng/m
3
unless otherwise noted).
N Mean SD
PM 2.5 mass (μg/m
3
) 209 22.33 16.61
Optical Carbon Fractions
BC (μg/m
3
) 209 1.05 1.71
BrC (μg/m
3
) 206 1.08 0.82
ETS (μg/m
3
) 210 1.58 6.11
Elements
Al 212 1.76 6.68
Ba 212 2.01 1.92
Br 212 0.42 0.44
Ca 212 12.13 20.00
Cl 212 17.91 35.86
Co 212 0.07 0.11
Cu 212 2.65 1.73
Fe 212 17.31 15.69
K 212 14.96 20.11
Mg 212 5.53 8.86
Mn 212 0.36 0.41
Na 212 43.34 42.69
Ni 212 0.33 0.39
P 212 0.77 2.52
Pb 212 0.20 0.37
S 212 56.88 41.54
Se 212 0.22 0.26
Si 212 23.47 28.79
Sr 212 0.25 0.95
Ti 212 1.44 1.80
V 212 0.09 0.16
Zn 212 1.86 2.47
The distributions of home characteristics, indoor PM sources, and selected time-activities
as reported in questionnaires or derived from GPS data are presented in Table S3.1 (Appendix
51
B). Based on the exit survey, 60% spent some time near traffic when outdoors, and 34% spent
more than 2 hrs per day commuting during the monitoring period. When with the sampler, 60%
of participants opened windows more than half of the time, 26% used air conditioning and 37%
used fans at home. In terms of indoor PM sources, 80 (38%) were close to cooking smoke and 51
(24%) close to burning candles or incense, and 83 (39%) were exposed to smokers. Based on the
3
rd
trimester questionnaire, 56% of participants lived in an apartment, 44% were part of a
household with > 3 persons, and 43% lived in a home built after the 1980s. In addition,
participants’ GPS-estimated duration of staying indoors at home was 78.5% (SD=19.6) and
staying in non-home locations was 15.2% (15.7).
3.3.2. Positive Matrix Factorization Analysis
We replaced missing values of 14 observations with species’ median values including
PM 2.5 mass (3 observations), BC (3), BrC (6), and ETS (2). A five-factor solution combined the
two sources later identified as fuel oil and secondhand smoking, while seven factors resulted in a
non-interpretable factor with a single high loading of Zn, resulting in a six-factor solution as the
optimal, physically interpretable solution (Qrobust =5845.3 and Qtrue=6143.1). An Fpeak rotation
of -0.1 was then applied with 100 bootstraps which resulted in no unmapped factors (compared
to one factor with two unmapped bootstrap runs in the base model (Table S3.2)). These six
factors together explained 48% of the variability in PM2.5. The species BC, Cl, K, S, Ca, and Zn
had non-normal residuals (Table S3.3). The PMF results are presented below for each predicted
source along with any bivariate analyses that supported its identification or explained some of
the variation in its mass contributions.
Traffic. The first source identified was traffic with high loadings of BC, Zn, and Ba
(Figure 3.1). It contributed on average 2.4% of the personal PM2.5 mass (Table 3.3). Traffic was
52
Figure 3.1. Source loading profiles (in % of species)
53
moderately positively correlated with crustal (described on page 58) and inversely correlated
with fresh sea salt and fuel oil sources (Table 3.4). This source was positively correlated with
outdoor NO2 and PM 2.5 and negatively correlated with O3 in the residential environment. It was
also positively correlated with total traffic-related NOx concentrations from local roadways
around the residence, as modeled by the CALINE4 dispersion model. In addition, length of
primary roads within KDE area was positively correlated with this source (Table 3.5).
Table 3.3. Source mass contributions.
Sources
Average mass contribution
(μg/m
3
) (SD)
Percent contribution to total
PM 2.5 mass (%)
Traffic 0.4 (0.5) 2.4
Secondhand Smoking 11.7 (9.3) 64.2
Aged Sea Salt 0.9 (0.9) 4.8
Fresh Sea Salt 0.8 (2.0) 4.5
Fuel Oil 2.1 (1.6) 11.4
Crustal 2.3 (4.1) 12.6
Table 3.4. Spearman correlations among PMF-predicted source contributions,
colored from low (blue) to high (red).
Traffic
Secondhand
Smoking
Aged
Sea Salt
Fresh
Sea Salt
Fuel Oil Crustal
Traffic
Secondhand Smoking -0.09
Aged Sea Salt -0.01 -0.29
Fresh Sea Salt -0.20 -0.24 0.07
Fuel Oil -0.20 -0.03 -0.07 -0.01
Crustal 0.32 -0.03 -0.08 -0.08 0.14
Values in bold font represent significant p-values at p<0.05 level.
Secondhand Smoking. The second source we identified had a high loading of BrC and
ETS (Figure 3.1). With an average mass contribution of 11.7 μg/m
3
, it contributed the majority
of personal PM2.5 mass (64.2% on average) (Table 3.3). Participants living in apartments seemed
to have slightly higher exposure to this source compared to those living in house (12.8 vs. 10.2
54
Table 3.5. Spearman correlations between PMF-predicted source contributions and variables
related to personal activities, time-activity patterns, indoor and outdoor environment.
Source Predictor Correlation
Traffic Outdoor (48-hour) air pollution at residence
O 3 -0.35
NO 2 0.61
PM 2.5 0.43
Total NO x from local traffic on Citilab road classes 1-5 0.14
Length of primary roads within KDE activity space 0.15
Secondhand Smoking
Ambient air pollutant concentrations (overlapping 24 hours)
at Downtown Los Angeles central site
Potassium Ion 0.12
Potassium -0.05
Element Carbon 0.03
Organic Carbon 0.09
Aged Sea Salt
Outdoor (48-hour) air pollution and meteorology at
residence
O 3 0.53
Wind speed -0.22
Temperature 0.55
Fresh Sea Salt Outdoor (48-hour) meteorology at residence
Wind speed 0.27
Relative humidity 0.16
Ambient air pollutant concentrations (overlapping 24 hours)
at Downtown Los Angeles central site
Ambient Chloride Ion 0.25
Ambient Chlorine 0.20
Fuel Oil Outdoor (48-hour) air pollution at residence
O 3 -0.17
NO 2 0.16
Crustal
Outdoor (48-hour) air pollution and meteorology at
residence
PM 10 0.24
Relative humidity -0.47
Precipitation -0.16
Values in bold font represent significant p-values at p<0.05 level.
μg/m
3
, respectively, not significant, Figure 3.2). The secondhand smoking source was also
negatively correlated with greater window opening time (11.7 vs.12.3 μg/m
3
, not significant).
Regarding the question of “if greater than none, how many people were smoking nearby”
included in the exit survey, 70 participants (out of 212) provided positive answers and those
55
experienced with more than one people smoking nearby had higher contributions from this
source than those with only one person smoking nearby (13.1 vs. 10.9 μg/m
3
, not significant). To
eliminate the possibility that this could be an outdoor biomass burning signal, we checked its
correlation with outdoor K, K+, elemental and organic carbon measures at the downtown
speciation site (n=148), all of which showed insignificant weak correlations (Table 3.5).
Figure 3.2. Relationship between secondhand smoking mass contributions and home type
Aged Sea Salt. The third source we identified had high loadings of Na, Mg, and S (Figure
3.1). It contributed on average 4.8% of the personal PM2.5 mass (Table 3.3). Aged sea salt was
negatively correlated with the secondhand smoking source (Table 3.4). It was strongly positively
correlated with outdoor O3 concentration and temperature and negatively correlated with wind
speed (Table 3.5). Aged sea salt was also significantly positively correlated with window
opening time, with an increasing trend in its average mass contributions from windows open
none of the time (0.3 μg/m
3
) to a little of the time (0.6 μg/m
3
), most of the time (1 μg/m
3
), and all
of the time (1.2 μg/m
3
) (Figure 3.3).
56
Figure 3.3. Relationship between aged sea salt and window opening time in the 48-hr monitoring
period
Fresh Sea Salt. The fourth source we identified had high loadings of Cl, Na, and Mg
(Figure 3.1). It contributed on average 4.5% of the personal PM2.5 mass (Table 3.3). Fresh sea
salt was negatively correlated with traffic and secondhand smoking sources (Table 3.4). The
mass contributions of this source were highest on days when average wind direction originated
from the west (NW followed by SW, significant, Figure 3.4). Fresh sea salt was also positively
correlated with wind speed and relative humidity at residence. To eliminate the possibility of this
being an indoor source correlated with aerosolized minerals from domestic water use or salt used
in cooking (Özkaynak et al., 1996; Schachter et al., 2020; Wallace 1996), we checked its
relationships with humidifier usage and time close to smoke from cooking, respectively. Even
though the sample size was unbalanced (30 out of 212 reported using a humidifier), average
mass contributions were lower (not significant) when people used a humidifier compared to not
(0.9 vs. 0.5 μg/m
3
, respectively). Similarly, mass contributions were lower when participants
reported spending more time close to smoke from cooking in the 48 hours (and not significant).
57
In addition, fresh sea salt was moderately positively correlated with ambient Cl and Cl- as
measured at the Downtown Los Angeles central site (Table 3.5).
Figure 3.4. Relationship between fresh sea salt mass contributions and average wind direction in
the 48-hr monitoring period
Fuel Oil. The fifth identified source had high loadings of Cu, Ni, and V (Figure 3.1).
With an average mass contribution of 2.1 μg/m
3
, it contributed 11.4% of personal PM2.5 mass on
average (Table 3.3). Fuel oil was positively correlated with crustal and negatively correlated with
the traffic source (Table 3.4). The participants living in homes originally built before 1980 had
higher exposures to this source than those living in newer homes (2.4 vs. 1.9 μg/m
3
, not
significant). In addition, it was positively correlated with outdoor NO2 and negatively correlated
with O3 (Table 3.5).
Crustal. The last source we identified had high loadings of Ca, Si, Ti, and Al (Figure
3.1). It contributed the second largest share of personal PM2.5 mass (12.6% on average), with an
average mass contribution of 2.3 μg/m
3
(Table 3.3). Crustal was moderately positively correlated
with traffic and fuel oil sources (Table 3.4). Households with more than three occupants were
58
associated with greater contributions of this source than households with three or fewer
occupants (1.3 vs. 2.8 μg/m
3
, significant, Figure 3.5). It was also positively correlated with
outdoor NO2 and PM 10 at the residence and negatively correlated with outdoor relative humidity
and precipitation.
Figure 3.5. Relationship between crustal mass contributions and household occupants
3.4. Discussion
In this study, six main sources were identified along with their contributions to personal
PM 2.5 mass concentrations collected from 212 low-income, predominantly Hispanic pregnant
women living in Los Angeles, CA, during the third trimester of pregnancy. Of the six sources
identified, secondhand smoking and crustal appeared to be of indoor origin, while traffic, aged
and fresh sea salt, and fuel oil were of outdoor origin. Secondhand smoke was the single largest
contributor to total personal PM2.5 mass concentrations. The combined indoor source
contributions (77%) were more than triple those of outdoor sources (23%), highlighting the
importance of the indoor environment in contributing to personal exposures.
59
In order to avoid overloading the samplers with particles from primary tobacco smoke
which would also overshadow any chemical fingerprints from other sources if present, by design,
we excluded participants who reported smoking themselves (this did not occur in this
population) or those with an active smoker permanently residing in their household (despite this
latter criterion not being consistently applied throughout the study). Despite this, secondhand
smoking was still identified as the source with the largest contribution to personal PM2.5
exposures. The mass contributions of this source did not show any clear trends over time as the
study progressed, suggesting that recruitment decisions did not significantly influence the
findings. This source had high loadings of BrC and ETS, and some loadings of Br and K which
were related to tobacco smoke in previous studies (Benner et al., 1989; Lawless et al., 2004;
Müller et al., 2011). Secondhand smoke is a well-known contributor to indoor air pollution
(Mueller et al., 2011; Nazaroff & Singer, 2004). The results showed that participants living in
apartments tended to have marginally higher exposure to secondhand smoking than those living
in detached houses. This could suggest greater potential of secondhand smoke infiltration from
adjacent units in an apartment building or from visitors smoking (Fabian et al., 2016; Price et al.,
2006; Wilson et al., 2011). Nevertheless, as both BrC and K are also strongly related to biomass
burning (e.g., Hasheminassab et al., 2014a, 2014b; Meng et al., 2007; Palm et al., 2020; Runa et
al., 2021), the correlations between secondhand smoking source and outdoor potassium were
checked to eliminate the possible source of biomass burning.
The results showed both fresh sea salt and aged sea salt as outdoor sources, with high
loading of Cl, Mg, Na, and Mg, Na, S, respectively. Previous work identified sea salt sources
with similar loading profiles (e.g., Cheung et al., 2011a; Corral et al., 2020; Habre et al., 2021;
Hasheminassab et al., 2014a, 2014b). Despite only having average wind direction over the 48-hr
60
monitoring period (not most frequent wind direction), fresh sea salt mass contributions were
higher with westerly winds and higher wind speeds, which provided greater potential for
aerosolization and airborne transport of sea salt particles from the Pacific Ocean. Habre et al.
(2021) found sea salt mass contributions to PM2.5 mass in southern California to be highest in
coastal communities. As fresh sea salt ages and undergoes photochemical reactions that also lead
to secondary O3 formation with warmer temperatures and more stagnant wind conditions (lower
wind speed), chlorine is lost and sulfates are formed (Gard et al., 1998; Habre et al., 2021). Thus,
aged sea salt resembles fresh sea salt in its loading profiles, except with S instead of Cl. Lower
wind speed provides more chemical reaction time between the sea salt particles and contributes
to the loss of chlorine and an increase in the formation of O3 (Crawford et al., 2019; Knipping &
Dabdub, 2003).
The high loadings of Al, Ca, Si is expected in natural crustal materials, and the lack of or
less abundant loadings of Ba, Zn and Cu indicated that this was not resuspended road dust which
could have tire and brake wear impacts (Cheung et al., 2011a; Lough et al., 2005). Crustal
elements originate outdoors and can enter the indoor environment as windblown dust or as dust
tracked indoors on residents’ shoes. Once indoors, crustal materials will typically settle and get
resuspended as indoor sources (or emissions of indoor origin) when disturbed by human
movement or other activities (i.e. vacuuming). Therefore, the presence of more occupants in a
household provides greater opportunities for re-suspension of crustal dust which mirrors the
results reported here (Habre et al., 2014a). As such, crustal was labelled as an indoor origin
source despite the possibility of our participants getting exposed to crustal dust outside of their
homes as well.
61
The results indicate that fuel oil and traffic sources contributed to personal PM2.5
exposures as well. Similar to previous studies, the fuel oil source had high loadings of Ni and V
which are known tracers of heavy residual fuel oil combustion in large industrial applications
and in marine engine emissions (Corbin et al., 2018; Corral et al., 2020; Larson et al., 2004;
Maykut et al., 2003; Meng et al., 2007; Minguillón et al., 2012). BC serves as a marker for the
traffic related source (Habre et al., 2014a; Hasheminassab et al., 2014b), while species such as
Zn, Ba, and Fe come from motor vehicle exhaust emission, brake and diesel additives (Ålander
et al., 2005; Corral et al., 2020; Meng et al., 2007; Onat et al., 2013). This source was correlated
with residential estimates of CALINE NOx and outdoor pollutants related to traffic, which can
be related to the finding that participants spent the majority of their time at home. The correlation
between their traffic mass contributions and activity space based primary road exposures also
revealed that these women visited many places, which were aligned with the time-activities
derived from their GPS tracks.
The strengths of this study include the 48-hr personal PM2.5 measurements and detailed
chemical composition analysis that allowed us to apportion the major sources that contributed to
personal exposures. By integrating concurrently collected questionnaire data and geospatially
modeled environmental exposures in activity spaces (from GPS) and in the residential
neighborhood, the results further corroborate these sources, their origin (primarily indoor vs
outdoor), and exposure effects. With approximately three-fourths of personal exposures
contributed by indoor sources, our findings highlight the importance of the indoor environment
contributions to total PM2.5 exposures during pregnancy and the potentially incomplete
understanding of this population’s exposures by solely relying on outdoor air pollution measures.
The PMF model also only explained a portion of the variability in personal PM2.5 mass
62
concentrations (R
2
= 0.48). One possible reason could be that we did not measure organic carbon
(OC) species in this study which are known to contribute a large fraction of indoor PM2.5 mass
concentrations (Habre et al., 2014a, 2014b; Turpin et al., 2017), and the possible volatilization of
lightweight organic carbon fractions from the Teflon filters used in the sampling design. The
sample size of the study, while considered large in personal monitoring settings, and the short
monitoring period may limit the generalizability and representativeness of personal PM2.5
exposures beyond this study area and across the full pregnancy and postpartum periods.
However, this is one of the few studies to conduct a thorough characterization of sources
impacting personal PM2.5 exposures of predominantly Hispanic and low-income women during
pregnancy in an environmental health disparities context.
3.5. Conclusion
PM 2.5 is a mixture of organic and inorganic elements, and its composition and thus
toxicity can vary based on its sources. Given the complexity of PM 2.5 itself and multiple factors
affecting personal exposures, it is critical to disentangle and understand the relative importance
of different sources contributing to personal PM2.5 exposures. Our findings also provide new
insights of how multiple sources from indoor and outdoor environments contributed to the
personal PM2.5 exposures of low-income, predominantly Hispanic/Latina pregnant women in Los
Angeles. The results may facilitate investigating the health effects related to each source, as well
as recommending source-specific interventions to an environmental health disparities population
during pregnancy.
63
Chapter 4 Modeling Personal PM2.5 Exposures within Multiple
Microenvironments
This chapter examined whether or not the APEX model developed by the USEPA could estimate
the range of personal PM2.5 exposures for MADRES participants, as well as how APEX
parameters could be adjusted to capture more of the complexity in personal PM2.5 exposures
contributed by indoor sources and the interaction of the indoor and outdoor environments. The
chapter starts by introducing the research background, followed by the data and methods used in
this study, and then finishes up with results, discussion, and conclusions.
4.1. Introduction
Exposure to PM2.5 is associated with several adverse health outcomes including
respiratory and cardiovascular morbidity (Brandt et al., 2014; Gan et al., 2011; Kim et al., 2004).
PM 2.5 exposure during pregnancy has also been shown to affect maternal (Dadvand et al., 2014;
Ghosh et al., 2014) and fetal health (Dadvand et al., 2011; Fleischer et al., 2014; Hsu et al., 2015;
Pereira et al., 2014; Rich et al., 2015; Ritz et al., 2007; Rosa et al., 2020). To accurately estimate
its health risks, it is crucial to have accurate measures/estimates of total personal PM2.5 exposure
in health studies since this will reduce exposure measurement error and increase statistical power
to observe associations (Baxter et al., 2013; Hu et al., 2017). In addition, given PM 2.5 itself is a
mixture with several complex factors contributing to total personal exposure (i.e. time-activity
patterns, indoor sources, behaviors, etc.), there is a need to understand where and when highest
exposures occur, which sources contribute the most across various microenvironments people
spend time in, and how to intervene to reduce risk. This is especially important for
environmental health disparities research since disadvantaged populations often experience
64
disproportionately higher exposures to certain sources and can be more susceptible to their
adverse health effects (Bae et al., 2007; Houston et al., 2004; Tian et al., 2013).
Personal monitoring is considered the “gold standard” for assessing external exposure, in
which participants carry or wear portable devices to sample air pollutants in their breathing zones
as they go about their daily activities (Choi et al., 2006, 2008, 2012; Jedrychowski et al., 2004,
2009; Minguillón et al., 2012; Rundle et al., 2012). Nevertheless, due to the high cost and burden
for participants and researchers, it is difficult to conduct high quality personal monitoring in
large populations and over long periods of time. Therefore, models that can accurately predict
total personal exposure for large populations and account for the various sources and factors that
contribute to it would be highly desirable.
Since individuals are mobile, their personal PM2.5 exposure is driven by their daily time-
activities, by outdoor PM2.5, and by PM2.5 concentrations in microenvironments they spend time
in (Duan 1982; Wallace, 1996; Wallace and Williams 2005). Accordingly, microenvironmental
models have been developed to estimate personal exposure by integrating information on time
spent within key microenvironments and PM2.5 concentrations within them, assuming well-
mixed conditions (Lai et al., 2004; Liu et al., 2003; Rabinovitch et al., 2016; Steinle et al., 2015).
Several microenvironmental models have been developed to support population level
applications (Berrocal et al., 2011; Breen et al., 2014; Hänninen et al., 2003; Hsu et al., 2020;
Lim et al., 2012). Among them, the APEX inhalation exposure model developed by the USEPA
has been widely used in air pollution exposure and risk assessment, as well as health studies
(Dionisio et al., 2017; Johnson et al., 2018; Rosenbaum et al., 2008; Sarnat S. et al., 2013;
USEPA, 2019a, 2020).
65
For example, Sarnat S.E. et al. (2013) found that APEX estimated personal exposures to
carbon monoxide (CO) and NOx from outdoor origin produced better risk estimates of
emergency department visits for asthma and wheeze than ambient concentrations in Atlanta, GA.
However, Johnson et al. (2018) compared APEX-simulated microenvironmental PM2.5 with
corresponding measurements in three study areas within central Los Angeles, CA, and identified
various sources of uncertainties in APEX inputs and predictions, namely lack of spatial
resolution for ambient PM2.5 and the non-representativeness of some of the APEX parameter
(e.g., air exchange rate, decay rate) distributions.
APEX uses a stochastic, microenvironmental approach to estimate personal exposures to
several air pollutants such as PM2.5 for individuals randomly drawn based on age, race, and
gender distributions within census tracts in specified geographic areas (USEPA, 2020). Activity
patterns of simulated individuals are simulated by random draws from the USEPA’s
Consolidated Human Activity Database (CHAD) diaries, and their daily trajectories are assigned
to user-selected microenvironments (McCurdy et al., 2000; USEPA, 2020).
Microenvironments and how they are operated can be customized for various settings or
populations. For example, studies have shown that incorporating information on use of windows
for ventilation and indoor source emissions may improve estimates of indoor concentrations
(Johnson et al., 2018; Sarnat S.E. et al., 2013; Weisel et al., 2005). However, customization can
also make it challenging to compare exposure estimates across models and studies. Previous
studies have used fixed-site measured pollutant concentrations or exposures estimated from other
methods to check the accuracy of APEX outputs (Johnson et al. 2018; Sarnat S.E. et al. 2013).
Compared to the fixed-site monitoring data, Johnson et al. (2018) found APEX underestimated
PM2.5 concentrations in all of the microenvironments identified in this study.
66
In this study, personal PM2.5 exposure measurements collected in the 3
rd
trimester of
pregnancy in a sub-study of the MADRES pregnancy cohort were leveraged. MADRES aims to
understand the effects of air pollution, environmental exposures, and social stressors on maternal
and child health in a predominantly Hispanic, low-income population in Los Angeles, CA. Data
from this personal monitoring sub-study provides a unique opportunity to compare to the
distribution of APEX-predicted personal PM2.5 exposures in a synthetic population simulated to
resemble the larger environmental health disparities community the MADRES population (and
eventually this sub-study) draws from. By learning from questionnaire information on key
parameters (i.e., home ventilation, indoor sources, time-activity patterns, etc.) in MADRES, this
study could also evaluate the extent to which the inputs need to be refined or resolved to get
closer to reproducing the range of the personal measurement data for this particular
environmental health disparities population.
Nevertheless, this comparison will not be perfect because: (a) APEX simulates
hypothetical people and cannot be used to predict exposure for the same individuals and time
periods in MADRES (USEPA, 2019a, 2019b, 2020); and (b) several assumptions are embedded
in this comparison. For example, simulated individuals in APEX comprise a random sample
drawn from the defined population universe in Los Angeles, while MADRES participants in the
larger cohort from which the personal monitoring subset was selected were recruited from
several prenatal care providers mainly serving medically underserved populations (Bastain et al.,
2019). As such, the MADRES cohort was not designed to be a representative sample of
environmental health disparities populations in Los Angeles, CA, but closely reflects a specific
population’s characteristics within the larger and more diverse Los Angeles, CA populations
noted here. Therefore, for the purpose of this analysis, we assume that participants with personal
67
monitoring data constitute an imperfect sample of MADRES, which is also a convenience
sample of women of childbearing age living in Los Angeles neighborhoods experiencing
environmental health disparities.
Therefore, the overall aim in this work was to examine whether APEX can estimate and
explain personal PM2.5 exposures seen in MADRES at scale, and if not, how much refinement or
resolution of APEX inputs is needed to adequately reproduce the distribution and range of
personal measurements. As such, the analysis spanned four stages: (1) running APEX with as
close to default settings as possible to estimate personal PM2.5 exposures for a simulated
population with similar demographic characteristics as the MADRES participants; (2)
incrementally adding and customizing parameters in APEX to capture more refined ventilation
impacts and indoor source emissions in four scenarios; (3) comparing APEX estimates with
personal measurements to select an optimal scenario; and (4) describing predicted exposures
patterns and trends in the larger health disparities simulated population from the optimal,
selected scenario.
4.2. Method
In this section, the personal exposure measurement data of the MADRES participants, the
APEX model inputs, and the main methods applied in this research are described.
4.2.1. Data Collection
Given MADRES personal exposure measurement data being used in this study, the
MADRES study and the sampling sub-study are introduced and the input data for APEX are
described.
68
4.2.1.1. MADRES cohort and personal monitoring sub-study
MADRES is an ongoing prospective pregnancy cohort with the goal of understanding
environmental and social stressors that might affect childhood and pregnancy-related obesity
among predominantly low-income, Hispanic women and their babies in Los Angeles, CA
(Bastain et al., 2019). Women at less than 30 weeks gestation, ≥18 years of age, and able to
speak either English or Spanish fluently were recruited into MADRES from four prenatal care
providers in Los Angeles (Bastain et al., 2019). Informed consent was obtained from each
participant, and the USC’s IRB approved the study protocol.
The personal monitoring sub-study recruited 213 women in their 3
rd
trimester of
pregnancy from MADRES between October 2016 and March 2020. Their personal PM2.5
exposures and geolocation were monitored using 48-hr integrated personal sampling and
continuous GPS tracking at 10-sec intervals, respectively. In addition, an exit survey was
conducted at the end of the 48-hr monitoring period to ask about home operation (e.g.,
ventilation) and presence of any significant indoor sources of PM 2.5 such as cooking or smoking
during the 48-hr sampling period. Trimester specific questionnaires on demographics (e.g., age,
race), indoor sources such as presence and use of gas stoves, home operation (e.g., windows
open or not, AC usage), and current tobacco smoke exposure (primary and secondhand) were
also used to define the study population, adjust the microenvironment settings, and add indoor
emission sources to the APEX model, as described below.
4.2.1.2. APEX model input data
The APEX model provides flexibility in terms of setting microenvironment parameters
and adding multiple emission sources to predict personal exposures to air pollutants for large
populations. Therefore, five scenarios were set up by varying the parameters and emission
69
sources, aiming to find the optimal settings for predicting personal PM2.5 exposures for a large
health disparities population. The model (Version 5.2, October 2019) was downloaded from the
USEPA website (https://www.epa.gov/fera/human-exposure-modeling-air-pollutants-exposure-
model). The 2010 census tract-based population counts (by gender, race and age), along with the
activity diaries (questionnaires, events, and statistics) from the Consolidated Human Activities
Database (CHAD) (McCurday et al. 2000) were downloaded from the same website. As the
CHAD dataset covers the whole nation, the updated CHAD-California dataset was acquired from
the USEPA support team.
In addition, the ambient regulatory PM 2.5 monitoring data (concentrations and district
boundaries) for the modeling period of 2016-2020, as well as hourly temperature measurements
and meteorology zones for the EPA monitoring sites located in our study domain were
downloaded (https://aqs.epa.gov/aqsweb/airdata/download_files.html#Raw).
4.2.2. Data Analysis
This section describes the descriptive statistics gathered from MADRES personal
measurements and by extracting time-activities from the GPS data, and then lays out the
operational steps for the APEX model. Next, the model estimates and MADRES measurements
are compared, and the APEX outputs are examined at the microenvironment level. The section
closes with a description of the sensitivity analyses that were performed.
4.2.2.1. MADRES personal PM2.5 exposure and geolocation monitoring data
Descriptive statistics were calculated on the personal PM2.5 measurements to check their
overall distributions. The personal measurements were also stratified by ethnicity to examine
exposure variations between Hispanic and non-Hispanic women. Their 48-hr GPS tracks were
used to extract the time spent indoors for comparison as well, using a previously published
70
method (Cich et al., 2016; Li et al., 2008; Pérez-Torres et al., 2016; van Dijk, 2018; Xiao et al.,
2014) with time (30 min) and distance (e.g., 500 m) thresholds to estimate the time spent
indoors. The stay locations to the 3
rd
trimester residence were examined to confirm whether they
stayed at home or not. There was not sufficient information to differentiate the
microenvironments such as Outdoor, Near-Road and Vehicle and these were assigned to the “on-
road” category.
4.2.2.2. APEX model runs
Microenvironment setting
The five microenvironments pre-defined by APEX were adopted, along with the methods
(i.e., MASSBAL and FACTORS) for calculating PM2.5 concentrations in each microenvironment
(USEPA, 2019a, 2019b). The mass balance method (MASSBAL) was used to calculate
concentrations for the Indoor-Residence and Indoor-Other microenvironments. MASSBAL
assumes that an enclosed microenvironment (e.g., residence) is a single, well-mixed volume with
the air concentration approximately spatially uniform, and the amount of outside air flowing into
the microenvironment equals that flowing out of the microenvironment (USEPA, 2019b).
Therefore, the PM 2.5 concentrations in microenvironments such as Indoor-Residence are affected
by the inflow of air, outflow of air, removal of PM2.5 due to deposition, filtration, and chemical
degradation, and emissions from PM2.5 sources inside Indoor-Residence. FACTORS was used to
calculate PM2.5 concentrations for the Outdoor, Near-Road and Vehicle microenvironments, in
which it applies linear functions to relate microenvironment PM2.5 concentrations to the current
ambient concentration.
71
Defining the modeling domain
The framework for using APEX to estimate personal exposures is summarized in Figure
4.1. A circular study area with a 30 km radius that covered most of the MADRES participants’
activity spaces was defined at the outset (Appendix C, Figure S4.1). The modeling period was set
as October 1, 2016 to March 11, 2020, matching the data collection period of personal samples
for the MADRES participants.
Figure 4.1. APEX model workflow
Defining the study population
The census tracts with centroids within the study area were used to establish the study
domain for the simulated population. A total of 500 women aged 18 to 46 years living in Los
Angeles County were randomly selected and each was assigned to a home sector and work
sector, if employed. Their age and racial characteristics mirrored those in the 2010 census tract-
level population count within the study area.
72
Generating activity diaries
APEX used the matching demographic characteristics (e.g., sex, race, employment status)
for a simulated person and daily temperatures (e.g., MaxTemp, AvgTemp) for a simulated day to
select an activity profile from the CHAD-California database. Then the model used the records
matched on these characteristics to generate activity diaries and link them to microenvironments.
Calculating microenvironment PM2.5 concentrations in five scenarios
APEX used the outdoor PM2.5 concentrations to calculate hourly PM2.5 concentrations in
every microenvironment by linking the microenvironments people are in and the parameters such
as air exchange rates (AER) and decay rate governing concentrations in those
microenvironments. Given that multiple factors (e.g., keeping windows open or closed, emission
sources) can affect microenvironment parameters such as AER (Abt et al., 2000; Cao & Frey,
2011; Habre et al., 2014a, 2014b; Howard-Reed et al., 2002; Jiao et al., 2012; Wallace et al.,
2002; Yamamoto et al., 2010) and further affect microenvironment PM2.5 concentrations, five
scenarios were defined (labelled as S1, S2, etc., summarized in Table 4.1) for the model runs.
The goal was to understand how the parameter setting changes for the Indoor-Residence
microenvironment (e.g., probabilities of window opening depending on temperature, or emission
rates of various indoor sources) might affect personal PM2.5 estimates.
In S1, we used generic APEX-provided settings to calculate AER for Indoor-Residence
(e.g., average temperature, the probability of using an air conditioner at home) and ambient
PM 2.5 concentrations collected from EPA monitoring sites as the pollutant input. The parameters
describing the distributions for microenvironmental concentrations estimates were taken from
Johnson et al. (2018). Given the positive connections between windows open, air exchange rates
73
Table 4.1. Five APEX scenarios modeled in this simulation with associated conditional variables
for the Indoor Residence microenvironment in each.
Scenarios Conditional Variables PM 2.5 Source Being Modeled
S1
Temperature ranges (categories) in Fahrenheit, Home AC
probabilities (Yes/No)
Ambient
S2
Temperature ranges (categories) in Fahrenheit, Home AC
probabilities (Yes/No), Home windows open (Yes/No)
Ambient
S3
Temperature ranges (categories) in Fahrenheit, Home AC
probabilities (Yes/No), Home windows open (Yes/No),
Home gas stove probability (Yes/No)
Ambient and indoor (gas stove
use for cooking)
S4
Temperature ranges (categories) in Fahrenheit, Home AC
probabilities (Yes/No), Home windows open (Yes/No),
Home candle burning probability (Yes/No)
Ambient and indoor (candle
burning)
S5
Temperature ranges (categories) in Fahrenheit, Home AC
probabilities (Yes/No), Home windows open (Yes/No),
Home gas stove probability (Yes/No), Home candle burning
probability (Yes/No)
Ambient and indoor (gas stove
use for cooking and candle
burning)
and indoor pollutant concentrations shown in the literature (He et al., 2004; Howard-Reed et al.,
2002; Sarnat J.A. et al., 2013; Schembari et al., 2013; Wallace et al., 2002; Yamamoto et al.,
2010), window openings were added to the parameters specified in S1 for calculating Indoor-
Residence AER in S2. Based on previous studies, we assumed the window openings doubled
AER (Howard-Reed et al., 2002; Wallace et al., 2002). Both S1 and S2 model the contribution of
PM2.5 of outdoor origin to total exposures. In S3, PM2.5 emissions from gas stove use for cooking
were added as an indoor source of PM2.5, in addition to the parameters specified in S2. The
probability of using gas stoves (Yes=0.92) was extracted from the MADRES questionnaire, and
the use levels were taken from Hu et al. (2012). In S4 indoor PM2.5 emissions from candle or
incense burning were added to the parameters specified in S2. The probability was again
extracted from the MADRES exit survey (Yes=0.25), and the associated parameter distributions
were also taken from Hu et al. (2012). Finally, in S5 the concentrations from both indoor sources
74
used in S3 and S4 were combined (cooking by gas stoves and candles/incense burning). The
complete list of microenvironment parameters used for the five model runs along with their
distributions is shown in Table S4.1 (Appendix C). To facilitate comparisons, we used the same
seed number for all five scenarios.
Estimating personal PM2.5 exposures
Based on the microenvironment concentrations, APEX then calculated the PM2.5
exposure for each simulated person within every microenvironment. Time-averaged daily
personal PM2.5 exposures were also estimated for the simulated individuals.
4.2.2.3. Comparison of APEX estimates with MADRES measurements
The distribution of APEX predicted personal PM2.5 exposures in S1 were initially
compared to the personal measurements using descriptive statistics, on a yearly basis and for the
whole modeling period. Minimum and maximum values were used to check whether the range of
APEX estimates were within personal measurements. Mean values were used to compare the
overall performance for each scenario, while standard deviation values were used to check inter-
personal variations in APEX predictions. The normality of personal PM2.5 measurements and
APEX estimates was checked and the Wilcoxon Sign Rank test was used to examine whether
predicted vs. measured median PM2.5 exposures were statistically significantly different. The
same evaluations were conducted for S2 through S5 to check the impact of parameters (e.g.,
window conditions, indoor PM2.5 sources) on estimated personal exposures and how well they
reproduced the range of personal measurements.
The sum of minutes in each microenvironment were converted to the percentage duration
from the overall modeling period or 48-hr sampling period for the aforementioned comparisons.
We examined whether there were significant differences in terms of duration of time spent in In-
75
Residence and In-Other microenvironments as estimated by APEX or with GPS data from
MADRES using Wilcoxon Sign Rank non-parametric tests. Given the personal exposures for
MADRES participants were integrated values the estimated and measured PM2.5 concentrations
at the microenvironment level could not be compared. Therefore, the distributions of
microenvironment exposures among APEX results estimated by the five scenarios were
compared to highlight the impact of different parameters on microenvironment exposures. In
addition, the hourly ambient PM2.5 concentrations with predicted total personal and
microenvironmental PM2.5 exposures were compared.
Once the optimal scenario based on the closest reproducibility of the range of personal
PM 2.5 measurements was selected, the predictions could be described in more detail. The
durations (in percent) and exposures (µg/m
3
) by microenvironment level are presented for the
whole modeling period to gain a better understanding of time-activity patterns and associated
exposures for APEX individuals. The microenvironment exposures were next compared on an
hourly basis along with the personal exposures and ambient PM2.5 and Spearman correlations
among personal exposures, microenvironment exposures and ambient PM2.5 were also conducted
to check their relationships.
4.2.2.4. Sensitivity analyses
Since MADRES participants are low-income predominantly Hispanic pregnant women,
and the APEX study area covers high income neighborhoods such as Beverly Hills, one of the
concerns is that the simulated population might have very different socioeconomic
characteristics and PM2.5 exposures from the larger environmental health disparities population
represented by MADRES participants. Therefore, sensitivity analyses were conducted to test
whether including only MADRES census tracts in the simulation resulted in predicted exposures
76
that more closely resembled the personal PM2.5 measurements gathered in MADRES. To
conduct this analysis, the overall modeling outputs into two groups – one group only including
the estimated exposures for simulated individuals living in MADRES census tracts, and the other
group only including the estimated exposures for those living in non-MADRES census tracts.
The between-group differences was then calculated along with comparisons with personal
measurements. Given MADRES participants are predominantly Hispanic women, tests were also
conducted to see whether the limiting the simulation results to only Hispanic women resulted in
more similar PM2.5 estimates as well.
4.3. Results
Table 4.2 summarizes the demographic characteristics of the 500 individuals simulated
with APEX and the MADRES participants. Since we used the same seed to initialize the model
but different parameters and added indoor PM2.5 sources in several scenarios, we observed slight
differences in terms of population composition. MADRES participants were four years younger
on average; and 79% of them were Hispanic, compared to just 52% in the model simulations.
Table 4.2. Demographic characteristics of simulated APEX and actual MADRES participants.
MADRES S1 S2 S3 S4 S5
(N=213) (N=500) (N=500) (N=500) (N=500) (N=500)
Age (years) - Mean (SD) 28.3 (6.00) 32.3 (8.26) 32.3 (8.26) 32.4 (8.25) 32.4 (8.25) 32 (8.42)
Race - n (%)
White, non-Hispanic 12 (5.6%) 89 (17.8%) 89 (17.8%) 115 (23.0%) 115 (23.0%) 127 (25.4%)
Asian, non-Hispanic 2 (0.9%) 87 (17.4%) 87 (17.4%) 62 (12.4%) 62 (12.4%) 73 (14.6%)
African American, non-
Hispanic
24 (11.3%) 51 (10.2%) 51 (10.2%) 45 (9.0%) 45 (9.0%) 39 (7.8%)
Hispanic 169 (79.3%) 257 (51.4%) 257 (51.4%) 266 (53.2%) 266 (53.2%) 247 (49.4%)
Other 6 (2.8%) 16 (3.2%) 16 (3.2%) 12 (2.4%) 12 (2.4%) 14 (2.8%)
Estimated times spent in various microenvironments were similar across APEX scenarios
(Table S4.2). Figure 4.2a shows that individuals spent the majority of their time indoors in S3.
77
Figure 4.2. APEX scenario S3 simulated results by microenvironment for: (a) stay time
durations (%); (b) PM2.5 concentrations (µg/m
3
); and (c) Personal time-weighted PM 2.5
exposures (µg/m
3
) (*Numbers indicate means and error bars indicate standard deviations)
78
APEX simulated individuals spent 7% (SD=8) less time at home compared to MADRES
participants (79%) (SD=20) and more time at other indoor locations (18% vs. 15%). Actual time-
activity patterns were also more variable than APEX simulated ones.
APEX estimated PM2.5 concentrations in Near Road, Outdoor, and Vehicle
microenvironments were much higher than in the two indoor microenvironments across all five
APEX scenarios (Table S4.3). In-Residence PM2.5 concentrations increased between S1
(mean=8.8 µg/m
3
, SD=1.6), S2 (9.8 µg/m
3
, 1.5), and S3 (10.5 µg/m
3
, 1.7) due to the impact of
window opening and the combined impact of window openings and indoor cooking,
respectively. S4 (9.9 µg/m
3
, 1.6) integrated indoor candle or incense burning, while S5 (10.4
µg/m
3
, 1.7) combined both indoor sources of cooking and indoor candle or incense burning.
Through comparing how well these scenarios reproduced personal measurements, S3 was the
optimal one given it had the best approximation to personal measurements (with the highest
mean value among scenarios), followed by S5 with similar estimates. Using S3 results as an
example, the time spent in both indoor microenvironments was higher than the others (Figure
4.2a); both indoor microenvironmental PM2.5 concentrations were lower than the others (Figure
4.2b), while time-weighted exposures in both indoor microenvironments were higher than
exposures from Outdoor, Near-Road, and Vehicle microenvironments (Figure 4.2c). In-
Residence microenvironment contributed most of the personal exposures (67-71% in different
scenarios), followed by the exposures in In-Other (16-19%), Vehicle (7-8%), Outdoor (5-6%)
and Near-Road (0.3-0.4%) on an hourly basis (Table S4.4).
Figure 4.3 shows hourly personal PM2.5 exposures contributed by the various
microenvironments throughout the day. In-residence exposures dominated the evening hours and
contributed substantially during the daytime hours as well. In-Other microenvironmental
79
exposures also had sizeable contributions between 10 a.m. and 5 p.m., during which simulated
individuals were probably in work locations or other indoor environments. Vehicle and outdoor
exposures had observable shares between 9 a.m. and 8 p.m., while the contributions from Near-
Road were negligible all of the time.
Figure 4.3. Contributions of various microenvironments
to hourly personal PM2.5 exposures in S3 (*power scale of 1.5 used for Y-axis values)
Table 4.3 presents Spearman correlations among simulated, hourly microenvironment
exposures, ambient PM2.5 concentrations, and personal exposures from S3. Hourly personal
exposure estimates were strongly correlated with ambient PM2.5 concentrations (r=0.83).
Personal exposures had the strongest correlation with In-Residence exposures (r=0.54). In-
Residence exposures were negatively correlated with Outdoor exposures (r=-0.24), but positively
correlated with ambient PM2.5 (r=0.46). Tables S4.5 and S4.6 show similar correlations in S1 and
S2.
80
Table 4.3. Spearman correlations among simulated hourly total personal PM2.5 exposures,
microenvironmental PM2.5 exposures, and ambient PM2.5 concentrations in S3.
Total Personal
PM2.5 Exposure
In-Residence
PM2.5
In-Other
PM2.5
Outdoor
PM2.5
Near-Road
PM2.5
Vehicle
PM2.5
Ambient
PM2.5
APEX Estimates
In-Residence PM2.5
0.54
In-Other PM2.5
0.02
-0.65
Outdoor PM2.5
0.09
-0.24 0.10
Near-Road PM2.5
0.02
-0.07 -0.002 0.04
Vehicle PM2.5
0.11
-0.29 0.28 0.32 0.07
Ambient PM2.5
0.83
0.46 0.05 0.004 -0.003 0.02
Values in bold font represent significant p-values at p<0.05 level.
Overall mean MADRES personal PM2.5 measurements were almost twice as high as
mean ambient concentrations at monitoring stations and two or more times higher than the
APEX personal exposure estimates (Table 4.4). In addition, 48-hr integrated MADRES
measured exposures were more variable than daily APEX estimates.
Table 4.4. PM2.5 comparisons among MADRES personal measurements,
APEX estimates, and USEPA monitoring station concentrations (µg/m
3
).
PM 2.5 Concentrations (µg/m
3
)
Personal
measurements
APEX
estimates
Ambient
concentrations
S1 S2 S3 S4 S5
Minimum 1.8 0.1 0.1 0.2 0.1 0.3 0
Maximum 140.2 111.1 114.1 112.8 123.5 117.5 121.0
Mean (SD) 23.3 (19) 9.5 (5) 10.2 (6) 10.7 (6) 10.2 (6) 10.6 (5) 11.7 (7)
Sensitivity analyses were also conducted to test the impact of including non-MADRES
residential tracts on estimated personal PM2.5 exposures. Compared to the simulated individuals
living in the non-MADRES census tracts, the individuals within the MADRES census tracts had
slightly higher exposures (0.3-2.4% higher in different scenarios). Most of simulated Hispanic
women (i.e., 247-266 out of 500 simulated individuals) had significantly lower exposures
compared to MADRES Hispanic participants (i.e.,169 out of 213 individuals included in this
comparison).
81
4.4. Discussion
In this study, personal PM2.5 exposures were simulated for 500 individuals randomly
selected to represent the larger population of women of child-bearing age living in Los Angeles,
CA, from October 1, 2016 to March 11, 2020. The simulated exposures were compared with the
personal measurements in a sub-study of 213 women enrolled in the MADRES study. Although
MADRES participants represent an imperfect sample of the simulated population, this study
provided a unique opportunity to examine whether the APEX model could estimate the range of
personal exposures for a larger environmental health disparities population from which the
MADRES cohort is drawn. Furthermore, by comparing model estimates within
microenvironments to personal measurements, the evaluation can be made to see whether more
nuanced inputs can generate estimated exposures closer to the distribution of the real exposures
and capture the complexity in total personal exposure.
The results show that the estimated personal PM2.5 exposures were significantly lower
than MADRES personal measurements, indicating that the model underestimated personal
exposures. Personal exposures are modelled as time-weighted averages of microenvironmental
PM 2.5 concentrations which integrate both time-activities of individuals and pollutant
concentrations in each microenvironment (Duan 1982; Johnson et al., 2018; Sarnat S.E. et al.,
2013). This study gathered and used the microenvironmental parameters from Johnson et al.
(2018) to calculate microenvironmental PM 2.5 concentrations and encountered the same issues
documented by Johnson et al. (2018). The lack of knowledge about one or more critical
parameters (e.g., AER, decay rates, emission rates) determining indoor concentrations may result
in the underestimation of microenvironment concentrations.
82
The results also show strong correlations between APEX personal exposure estimates and
ambient PM2.5 (Spearman r=0.83), while the parallel correlation between personal measurements
and outdoor PM2.5 concentrations at residence (r=0.09) was fairly low among MADRES
participants. This large difference indicates that APEX estimates are perhaps mostly driven by
PM 2.5 sources of outdoor origin. APEX estimates might not be fully capturing the complexity of
personal exposures including indoor PM2.5 sources, the role of personal behaviors, individual
activities, and home characteristics. Nonetheless, this also points to ways to refine the model
inputs and improve personal exposure estimates. Given that indoor cooking and window opening
behaviors were common among MADRES participants, the parameters that reflected these
behaviors were added to the scenarios, and this improved the ability of the model outputs
approximate to personal measurements. This result suggested ways to fine-tune the APEX
parameters so the model can better describe the exposures of populations of special interest like
in this dissertation.
Compared to the small subset of MADRES participants, APEX underestimated durations
of staying at home for simulated individuals. Previous studies have shown that the durations of
staying in microenvironments, particularly indoor at residence, is an important factor affecting
total personal exposure occurring indoors (Adgate et al., 2004b; Jenkins et al., 1992; Kim et al.,
2005; Turpin et al., 2007). In this study, the CHAD California data was used to generate activity
diaries for simulated individuals. The majority of the CHAD data specifically describing
activities for Californians was collected between 1987 and 1992 (McCurdy et al., 2000). Even
though several activity studies were incorporated into the original CHAD in November 2016, the
most recent study was conducted from 2003 through 2011 (Graham et al., 2019). In addition,
among 23 studies incorporated in CHAD, only two studies were conducted in the Los Angeles
83
area with real-time diaries collected from students aged 10 to 17 years (Graham et al., 2019).
Given the CHAD data might be outdated, the activity diaries used by APEX might not represent
current day time-activities for individuals. The differences of spatial range for study areas, age
range of respondents, and the type of survey designs used in the CHAD and MADRES studies,
along with the specific time-activity patterns for pregnant women, might also contribute to the
time duration spent in microenvironments differences between APEX individuals and MADRES
participants. Therefore, the inclusion of more recent and representative diary data for the
specified study area, age and socioeconomic range, even for a special population group such as
pregnant women in CHAD datasets may produce estimates that better approximate personal
measurements in a similar future study.
While APEX provides some flexibility in terms of capturing the characteristics of the
population at hand, there are still some areas that would benefit from further customization. For
example, the default microenvironment setting does not provide the option to set up different
home characteristics (e.g., living in an apartment or a house) for In-Residence among simulated
individuals; however, around 60% of MADRES participants live in an apartment while 40% live
in a house. Some studies have shown that air exchange rates (AER) can be twice as high for
apartments compared to single-family homes in certain contexts (Price et al., 2006). This means
that there may be important differences in In-Residence exposures for apartments and for houses,
and that we might define multiple In-Residence settings accordingly. In terms of multiple PM2.5
sources that could influence personal exposures, such as secondhand smoking which happens
across several microenvironments (Fabian et al., 2016; Zamora et al., 2018), the current model
setting does not seem to allow specifying such conditions very well.
84
The strength of this research included applying APEX to model personal exposures for
the population with pre-defined microenvironments and multiple emission sources. The outputs
improve our understanding of personal exposures at the microenvironment level. Furthermore,
the results reveal the need for the refinement of the model inputs to reproduce the distribution of
personal measurements. One limitation is that the MADRES sub-study participants with personal
monitoring are an imperfect subset of the larger MADRES cohort, which in itself is an imperfect
(in statistical terms) of the sample drawn from the larger environmental health disparities
population we aimed to simulate with APEX. In addition, the lack of sufficient knowledge
regarding the distributions of model parameters, as well as indoor emission sources within
microenvironments in our own data, affected our ability to model personal PM2.5 exposures.
4.5. Conclusion
The research findings show that the APEX model does a great job at modeling personal
exposure to PM2.5 of outdoor origin. It demonstrates a much greater improvement compared to
just relying on outdoor data, since it incorporates ventilation conditions and allows changes in
ventilation (e.g., open windows and air conditioner usage) based on actual temperature.
However, it seems more involved or complex to try to recreate all the different sources that
contribute to total personal exposure than is currently possible using APEX. The results may lead
to a better understanding of how the APEX model can be used to estimate personal PM2.5
exposures, along with potential improvements in input specifications to better approximate
personal exposures.
85
Chapter 5 Conclusions
Exposure to air pollution and PM2.5 more specifically is an important environmental risk
factor that has been associated with various adverse health outcomes. Pregnancy in particular is
considered a sensitive exposure window with potential for long term impacts on maternal and
child health. Systemic inequities over time lead to persistent environmental health disparities,
which result in disadvantaged groups such as the low-income Hispanic population in Los
Angeles being disproportionally exposed to air pollution and more susceptible to its health risks.
Personal exposure to PM2.5 is complex, as PM2.5 itself is a mixture of various sources with
varying physicochemical properties and toxicity that could contribute to varying health
outcomes. Human mobility, time-activity patterns, and behaviors may also contribute to
variations of personal exposure.
Most health studies rely on outdoor PM2.5 estimates to investigate the associated health
outcomes, assuming they are the best surrogate of personal PM2.5 exposure of outdoor origin.
However, ignoring the impacts of factors such as activity spaces and time-activities on personal
exposure might result in exposure measurement error. In addition to the health risks of outdoor
PM 2.5, understanding the effects of total personal PM 2.5 exposures and indoor exposure
specifically is increasingly important given their high contribution to personal exposures and
their potential impacts on health. To date, knowledge around understanding personal PM2.5
exposures of low-income Hispanic pregnant women has been limited due to the complexity of
the sources contributing to their personal exposures, the multitude of co-occurring risk factors in
this sensitive window of time, and the greater systemic disadvantages they experience.
In this dissertation, personal PM 2.5 measurements and concurrent geolocation records for
a population of low-income, predominantly Hispanic pregnant women provided a unique
86
opportunity to fill this gap. Three aspects of their personal exposures were examined in three
separate studies. Study 1 was focused on investigating the main determinants, e.g.,
environmental exposures within GPS-derived activity spaces, time-activity patterns, indoor
sources, etc. and their impacts on personal exposures. Study 2 was focused on investigating the
main sources and their contributions to personal PM2.5 mass distinguished based on their
chemical fingerprints. Study 3 examined the contribution of microenvironments to personal
exposures and whether total personal exposure could be estimated for larger populations using a
well-known and population stochastic inhalation exposure model.
The results from Study 1 revealed a direct association between greater green cover and
parks and open space exposure in activity spaces and lower personal PM2.5 exposure which has
not been reported in previous studies. In addition, compared to the impact of outdoor residential
PM 2.5, indoor PM2.5 sources and indoor activities had a greater contribution to personal exposure
(on a standard deviation scaled basis). Study 2 identified six main sources based on their
chemical fingerprints that contributed to total personal PM 2.5 mass concentration, with combined
indoor source contributions greater than three times those of outdoor sources. The APEX
inhalation model results from Study 3 captured the contribution of outdoor PM2.5 to personal
exposure, since predicted total personal exposure was highly correlated with outdoor PM2.5,
contrary to the weak correlation observed when using personal measurements. However, the
Indoor-Residence microenvironment contributed the majority of estimated personal exposures.
Overall, the model seemed to underestimate total personal exposures when compared to personal
measurement data despite the addition of different combinations of indoor source emission terms
selected based on the most commonly reported or observed sources in earlier studies. Refinement
of inputs such as more accurate indoor source terms and current time-activity budgets that
87
represent environmental health disparities populations would likely yield improved personal
exposure estimates.
Taken together, the findings of the three studies characterize personal PM2.5 exposures of
the low-income, predominantly Hispanic pregnant women. The results point to the significant
impact of GPS-derived activity spaces on the variation of personal exposures. Compared to
residential neighborhoods, environmental exposures within activity spaces, particularly KDE
area, are more correlated with where and how individuals interacted with their environments.
Therefore, using activity spaces may detect the associations between built-environment and
personal PM2.5 exposures in more accurate ways (e.g., greenness within KDE). This also reveals
the possible exposure measurement error when outdoor PM2.5 estimates at the residence are used
to approximate personal exposures of outdoor origin in health studies. This quantification of
environmental impacts could, in turn, facilitate the design of potential interventions, e.g.,
promoting “greener” urban spaces from the policy and practice perspective.
Similarly, the results for indoor candle or incense burning, duration of staying indoors,
indoor activities and home ventilation (Study 1), sources of secondhand smoking and crustal
(Study 2), and major contribution of Indoor-Residence exposure (Study 3), show the significant
contributions to total personal exposures across all three investigations. The identified sources
and factors reveal the importance of indoor environment when assessing personal PM2.5
exposures, which also improve our understanding of the disproportional exposures that this low-
income Hispanic population burdened. In addition to regulating outdoor PM2.5 concentrations,
interventions or standards that target the indoor environment, e.g., reducing indoor PM2.5
emissions or requiring building designers and operators to increase removal of indoor PM 2.5,
need to be developed in a scientific, evidence-based manner to provide adequate health
88
protection. This finding also raises the awareness to examine the health effects of the source-
specific PM 2.5 exposures, given the varied species and toxicity related to each source and the fact
that individuals are exposed to these mixtures and not a single pollutant or chemical at a time.
Lastly, the results demonstrate the possibility of using modeling approaches to estimate
personal exposures, particularly the personal exposures of outdoor origin. However, with the
significant contributions of indoor sources on personal measurements, more work needs to be
accomplished to model indoor microenvironment exposures from non-outdoor sources (e.g.,
emission from indoor PM2.5 sources or human activities), which may improve personal exposure
predictions accordingly. To facilitate modeling PM2.5 concentrations of the indoor
microenvironments, a database of indoor source emission distributions, as well as a library of
home ventilation effects on AER distributions, under wide ranging conditions that represent a
diverse population would make a significant contribution to this field. In addition, more recent
and representative travel and activity diary data covering a wide range of geographies, age and
socioeconomic status (SES) are recommended to be included in the CHAD or similar national
time-activity and travel behavior datasets, which may result in the improved estimates of time
spent in each microenvironment. Integrating SES information into population data for modeling
will make it possible to refine simulations, which may further improve exposure predictions for
individuals that are part of environmental health disparities populations. These tangible
recommendations, combined with modeling exposure of outdoor origin, may provide an
actionable way for improved personal exposure prediction in large populations and over longer
periods of time, which might greatly reduce the cost and burden of understanding personal
exposures. Collecting personal exposure measurements in tailored and targeted assessments;
89
however, can provide important validation data to continuously improve models and achieve a
greater understanding of personal exposures of different populations at scale.
My research demonstrates the complexity of how this pregnant environmental health
disparities population get exposure to PM2.5, and my findings provide foundation to refine
source-specific estimates of personal exposure to PM 2.5 of outdoor origin and total personal
PM 2.5. By doing so may reduce exposure measurement error, which these more accurate
exposure estimates can help epidemiological health studies. The greenness finding reveal the
direct link between urban design, city planning, and greener activity spaces on reducing personal
exposures, which may improve public health. In addition, my results also reveal the need for
smarter, more contextually aware interventions targeting main sources and determinants,
particularly indoor environment due to its significant contribution on total exposures. The
research also demonstrate the importance of interdisciplinary approach, or collaboration of
multiple disciplines, e.g., geography, exposure sciences, urban planning, public health, and
demography, to understand the complexity of personal PM2.5 exposure for this particular
population.
As the strength of my research, to my knowledge, this is one of the very few studies that
conducted a thorough investigation on personal PM2.5 exposures of predominantly Hispanic and
low-income women during pregnancy in an environmental health disparities context. The
personal PM2.5 monitoring and concurrent GPS data constitute a rich dataset which enabled this
investigation. In terms of the generalizability of my research, some of my findings, e.g.,
greenness and traffic impact, indoor sources, and outdoor PM can be generalizable to other urban
areas and other population; and the ways of approaching it may be transferable to other
environmental health disparities contexts and studies. As for the limitations, the sample size of
90
200 might be considered low, however for personal monitoring studies that are quite expensive
and burdensome to conduct to provide the highest quality data, this is considered fairly decent.
Of course, if it weren’t for the pandemic we would have expected a slightly larger sample size.
Overall, the dissertation findings help to dissect the complexity of personal PM2.5
exposures of this susceptible low-income predominantly Hispanic population during the critical
window of pregnancy. These findings can be further applied to advance environmental health
research and recommend appropriate interventions with the aim to control or minimize personal
PM2.5 exposures, which may further reduce health disparities.
91
References
Abatzoglou, J. T. 2013. “Development of Gridded Surface Meteorological Data for Ecological
Applications and Modelling.” International Journal of Climatology 33(1), 121–131.
https://doi.org/10.1002/joc.3413
Abt, E., Suh, H. H., Allen, G., & Koutrakis, P. 2000. “Characterization of Indoor Particle
Sources: A Study Conducted in the Metropolitan Boston Area.” Environmental Health
Perspectives 108(1), 35–44. https://doi.org/10.2307/3454293
Adgate, J. L., Church, T. R., Ryan, A. D., Ramachandran, G., Fredrickson, A. L., Stock, T. H.,
… Sexton, K. 2004a. “Outdoor, Indoor, and Personal Exposure to VOCs in Children.”
Environmental Health Perspectives 112(14), 1386–1392.
https://doi.org/10.1289/ehp.7107
Adgate, J. L., Eberly, L. E., Stroebel, C., Pellizzari, E. D., & Sexton, K. 2004b. “Personal,
Indoor, and Outdoor VOC exposures in a Probability Sample of Children.” Journal of
Exposure Analysis & Environmental Epidemiology 14(1), S4-S13.
https://doi.org/10.1038/sj.jea.7500353
Adgate, J. L., Ramachandran, G., Pratt, G. C., Waller, L. A., & Sexton, K. 2002. “Spatial and
Temporal Variability in Outdoor, Indoor, and Personal PM2.5 Exposure.” Atmospheric
Environment 36(20), 3255–3265. https://doi.org/10.1016/S1352-2310(02)00978-0
Ålander, T., Antikainen, E., Raunemaa, T., Elonen, E., Rautiola, A., & Torkkell, K. 2005.
“Particle Emissions from a Small Two-Stroke Engine: Effects of Fuel, Lubricating Oil,
and Exhaust Aftertreatment on Particle Characteristics.” Aerosol Science and Technology
39(2), 151–161. https://doi.org/10.1080/027868290910224
Alderete, T. L., Habre, R., Toledo-Corral, C. M., Berhane, K., Chen, Z., Lurmann, F. W., …
Gilliland, F. D. 2017. “Longitudinal Associations between Ambient Air Pollution with
Insulin Sensitivity, β-cell Function, and Adiposity in Los Angeles Latino Children.”
Diabetes 66(7), 1789-1796. https://doi.org/10.2135/cropsci2013.11.0764
Arhami, M., Sillanpää, M., Hu, S., Olson, M. R., Schauer, J. J., & Sioutas, C. 2009. “Size-
Segregated Inorganic and Organic Components of PM in the Communities of the Los
Angeles Harbor.” Aerosol Science and Technology 43(2), 145–160.
https://doi.org/10.1080/02786820802534757
Bae, C. H. C., Sandlin, G., Bassok, A., & Kim, S. 2007. “The Exposure of Disadvantaged
Populations in Freeway Air-Pollution Sheds: A Case Study of the Seattle and Portland
Regions.” Environment and Planning B: Planning and Design 34(1), 154–170.
https://doi.org/10.1068/b32124
Bastain, T. M., Chavez, T., Habre, R., Girguis, M. S., Grubbs, B., Toledo-Corral, C., … Breton,
C. 2019. “Study Design, Protocol and Profile of the Maternal and Developmental Risks
from Environmental and Social Stressors (MADRES) Pregnancy Cohort: A Prospective
92
Cohort Study in Predominantly Low-Income Hispanic Women in Urban Los Angeles.”
BMC Pregnancy and Childbirth 19(1), 1–16. https://doi.org/10.1186/s12884-019-2330-7
Baxter, L. K., Burke, J., Lunden, M., Turpin, B. J., Rich, D. Q., Thevenet-Morrison, K., …
Özkaynak, H. 2013. “Influence of Human Activity Patterns, Particle Composition, and
Residential Air Exchange Rates on Modeled Distributions of PM2.5 Exposure Compared
with Central-Site Monitoring Data.” Journal of Exposure Science & Environmental
Epidemiology 23(3), 241–247. https://doi.org/10.1038/jes.2012.118
Beckx, C., Int Panis, L., Arentze, T., Janssens, D., Torfs, R., Broekx, S., & Wets, G. 2009. “A
Dynamic Activity-Based Population Modelling Approach to Evaluate Exposure to Air
Pollution: Methods and Application to a Dutch Urban Area.” Environmental Impact
Assessment Review 29(3), 179–185. https://doi.org/10.1016/j.eiar.2008.10.001
Bell, M. L., Belanger, K., Ebisu, K., Gent, J. F., Lee, H. J., Koutrakis, P., & Leaderer, B. P.
2010. “Parental Exposure to Fine Particulate Matter and Birth Weight: Variations by
Particulate Constituents and Sources.” Epidemiology 21(6), 884–891.
https://doi.org/10.1097/EDE.0b013e3181f2f405.Prenatal
Bell, M. L., & Davis, D. L. 2017. “Reassessment of the Lethal London Fog of 1952: Novel
Indicators of Acute and Chronic Consequences of Acute Exposure to Air Pollution.”
Environmental Health Perspectives 109(suppl 3), 389–394.
https://doi.org/10.1289/ehp.01109s3389
Benbrahim-Tallaa, L., Baan, R. A., Grosse, Y., Lauby-Secretan, B., El Ghissassi, F., Bouvard,
V., … Straif, K. 2012. “Carcinogenicity of Diesel-Engine and Gasoline-Engine Exhausts
and Some Nitroarenes.” The Lancet Oncology 13(7), 663–664.
Benner, C. L., Bayona, J. M., Caka, F. M., Tang, H., Lewis, L., Crawford, J., … Eatough, D. J.
1989. “Chemical Composition of Environmental Tobacco Smoke. 2. Particulate-Phase
Compounds.” Environmental Science & Technology 23(6), 688–699.
https://doi.org/10.1021/es00064a007
Benson, P. E. 1992. “A Review of the Development and Application of the CALINE3 and 4
Models.” Atmospheric Environment Part B, Urban Atmosphere 26(3), 379–390.
https://doi.org/10.1016/0957-1272(92)90013-I
Berger, K., Malig, B. J., Hasheminassab, S., Pearson, D. L., Sioutas, C., Ostro, B., & Basu, R.
2018. “Associations of Source-Apportioned Fine Particles with Cause-Specific Mortality
in California.” Epidemiology 29(5), 639–648.
https://doi.org/10.1097/EDE.0000000000000873
Berrocal, V. J., Gelfand, A. E., Holland, D. M., Burke, J., & Miranda, M. L. 2011. “On the Use
of a PM 2.5 Exposure Simulator to Explain Birthweight.” Environmetrics 22(4), 553–571.
https://doi.org/10.1111/j.1743-6109.2008.01122.x.Endothelial
Brandt, S., Perez, L., Künzli, N., Lurmann, F., Wilson, J., Pastor, M., & McConnell, R. 2014.
“Cost of Near-Roadway and Regional Air Pollution-Attributable Childhood Asthma in
93
Los Angeles County.” Journal of Allergy and Clinical Immunology 134(5), 1028–1035.
https://doi.org/10.1016/j.jaci.2014.09.029
Breen, M. S., Long, T. C., Schultz, B. D., Crooks, J., Breen, M., Langstaff, J. E., … Buckley, T.
J. 2014. “GPS-Based Microenvironment Tracker (MicroTrac) Model to Estimate Time-
Location of Individuals for Air Pollution Exposure Assessments: Model Evaluation in
Central North Carolina.” Journal of Exposure Science & Environmental Epidemiology
24(4), 412–420. https://doi.org/10.1038/jes.2014.13
Brinkman, G. L., Milford, J. B., Schauer, J. J., Shafer, M. M., & Hannigan, M. P. 2009. “Source
Identification of Personal Exposure to Fine Particulate Matter Using Organic Tracers.”
Atmospheric Environment 43(12), 1972–1981.
https://doi.org/10.1016/j.atmosenv.2009.01.023
Brown, K. W., Sarnat, J. A., Suh, H. H., Coull, B. A., & Koutrakis, P. 2009. “Factors Influencing
Relationships between Personal and Ambient Concentrations of Gaseous and Particulate
Pollutants.” Science of the Total Environment 407(12), 3754–3765.
https://doi.org/10.1016/j.scitotenv.2009.02.016
Brown, S. G., Frankel, A., & Hafner, H. R. 2007. “Source Apportionment of VOCs in the Los
Angeles Area Using Positive Matrix Factorization.” Atmospheric Environment 41(2),
227–237. https://doi.org/10.1016/j.atmosenv.2006.08.021
Browning, C. R., & Soller, B. 2014. “Moving Beyond Neighborhood: Activity Spaces and
Ecological Networks As Contexts for Youth Development.” Cityscape 16(1), 165–196.
Brunst, K. J., Ryan, P. H., Brokamp, C., Bernstein, D., Reponen, T., Lockey, J., … LeMasters,
G. 2015. “Timing and Duration of Traffic-Related Air Pollution Exposure and the Risk
for Childhood Wheeze and Asthma.” American Journal of Respiratory and Critical Care
Medicine 192(4), 421–427. https://doi.org/10.1164/rccm.201407-1314OC
Buffler, P. A., Kwan, M. L., Reynolds, P., & Urayama, K. Y. 2005. “Environmental and Genetic
Risk Factors for Childhood Leukemia: Appraising the Evidence.” Cancer Investigation
23(1), 60–75. https://doi.org/10.1081/CNV-46402
Cao, Y., & Frey, H. C. 2011. “Geographic Differences in Inter-Individual Variability of Human
Exposure to Fine Particulate Matter.” Atmospheric Environment 45(32), 5684–5691.
https://doi.org/10.1016/j.atmosenv.2011.07.034
Chen, M., Dai, F., Yang, B., & Zhu, S. 2019. “Effects of Neighborhood Green Space on PM2.5
Mitigation: Evidence from Five Megacities in China.” Building and Environment 156,
33–45. https://doi.org/10.1016/j.buildenv.2019.03.007
Chen, X. C., Jahn, H. J., Ward, T. J., Ho, H. C., Luo, M., Engling, G., & Kraemer, A. 2020.
“Characteristics and Determinants of Personal Exposure to PM2.5 Mass and Components
in Adult Subjects in the Megacity of Guangzhou, China.” Atmospheric Environment 224,
117295. https://doi.org/10.1016/j.atmosenv.2020.117295
94
Chen, Z., Salam, M. T., Karim, R., Toledo-Corral, C. M., Watanabe, R. M., Xiang, A. H., …
Gilliland, F. D. 2015. “Living Near a Freeway is Associated with Lower Bone Mineral
Density among Mexican Americans.” Osteoporosis International 26(6), 1713–1721.
https://doi.org/10.1007/s00198-015-3051-z
Chen, Z., Salam, M. T., Toledo-Corral, C., Watanabe, R. M., Xiang, A. H., Buchanan, T. A., …
Gilliland, F. D. 2016. “Ambient Air Pollutants Have Adverse Effects on Insulin and
Glucose Homeostasis in Mexican Americans.” Diabetes Care 39(4), 547–554.
https://doi.org/10.2337/dc15-1795
Cheung, K., Daher, N., Kam, W., Shafer, M. M., Ning, Z., Schauer, J. J., & Sioutas, C. 2011a.
“Spatial and Temporal Variation of Chemical Composition and Mass Closure of Ambient
Coarse Particulate Matter (PM10-2.5) in the Los Angeles Area.” Atmospheric Environment
45(16), 2651–2662. https://doi.org/10.1016/j.atmosenv.2011.02.066
Cheung, K., Daher, N., Shafer, M. M., Ning, Z., Schauer, J. J., & Sioutas, C. 2011b. “Diurnal
Trends in Coarse Particulate Matter Composition in the Los Angeles Basin.” Journal of
Environmental Monitoring 13(11), 3277–3287. https://doi.org/10.1039/c1em10296f
Choi, H., Jedrychowski, W., Spengler, J., Camann, D. E., Whyatt, R. M., Rauh, V., … Perera, F.
P. 2006. “International Studies of Prenatal Exposure to Polycyclic Aromatic
Hydrocarbons and Fetal Growth.” Environmental Health Perspectives 114(11), 1744–
1750. https://doi.org/10.1289/ehp.8982
Choi, H., Rauh, V., Garfinkel, R., Tu, Y., & Perera, F. P. 2008. “Prenatal Exposure to Airborne
Polycyclic Aromatic Hydrocarbons and Risk of Intrauterine Growth Restriction.”
Environmental Health Perspectives 116(5), 658–665. https://doi.org/10.1289/ehp.10958
Choi, H., Wang, L., Lin, X., Spengler, J. D., & Perera, F. P. 2012. “Fetal Window of
Vulnerability to Airborne Polycyclic Aromatic Hydrocarbons on Proportional
Intrauterine Growth Restriction.” PLoS ONE 7(4), e35464.
https://doi.org/10.1371/journal.pone.0035464
Cich, G., Knapen, L., Bellemans, T., Janssens, D., & Wets, G. 2016. “Threshold Settings for
TRIP/STOP Detection in GPS Traces.” Journal of Ambient Intelligence and Humanized
Computing 7(3), 395–413. https://doi.org/10.1007/s12652-016-0360-9
Clark, N. A., Demers, P. A., Karr, C. J., Koehoorn, M., Lencar, C., Tamburic, L., & Brauer, M.
2010. “Effect of Early Life Exposure to Air Pollution on Development of Childhood
Asthma.” Environmental Health Perspectives 118(2), 284–290.
https://doi.org/10.1289/ehp.0900916
Corbin, J. C., Mensah, A. A., Pieber, S. M., Orasche, J., Michalke, B., Zanatta, M., … Gysel, M.
2018. “Trace Metals in Soot and PM2.5 from Heavy-Fuel-Oil Combustion in a Marine
Engine.” Environmental Science & Technology 52(11), 6714–6722.
https://doi.org/10.1021/acs.est.8b01764
95
Corral, A. F., Dadashazar, H., Stahl, C., Edwards, E., Zuidema, P., & Sorooshian, A. 2020.
“Source Apportionment of Aerosol at a Coastal Site and Relationships with Precipitation
Chemistry: A Case Study over the Southeast United States.” Atmosphere 11(11), 1212.
https://doi.org/10.3390/atmos11111212
Crawford, J., Cohen, D. D., Chambers, S. D., Williams, A. G., & Atanacio, A. 2019. “Impact of
Aerosols of Sea Salt Origin in a Coastal Basin: Sydney, Australia.” Atmospheric
Environment 207, 52–62.
Crawford, T. W., Jilcott Pitts, S. B., McGuirt, J. T., Keyserling, T. C., & Ammerman, A. S.
2014. “Conceptualizing and Comparing Neighborhood and Activity Space Measures for
Food Environment Research.” Health & Place 30, 215–225.
https://doi.org/10.1016/j.healthplace.2014.09.007
Crist, K. C., Liu, B., Kim, M., Deshpande, S. R., & John, K. 2008. “Characterization of Fine
Particulate Matter in Ohio: Indoor, Outdoor, and Personal Exposures.” Environmental
Research 106(1), 62–71. https://doi.org/10.1016/j.envres.2007.06.008
Crouse, D. L., Pinault, L., Balram, A., Brauer, M., Burnett, R. T., Martin, R. V., … Weichenthal,
S. 2019. “Complex Relationships between Greenness, Air Pollution, and Mortality in a
Population-Based Canadian Cohort.” Environment International 128, 292–300.
https://doi.org/10.1016/j.envint.2019.04.047
Dadvand, P., de Nazelle, A., Figueras, F., Basagaña, X., Su, J., Amoly, E., … Nieuwenhuijsen,
M. J. 2012a. “Green Space, Health Inequality and Pregnancy.” Environment
International 40, 110–115. https://doi.org/10.1016/j.envint.2011.07.004
Dadvand, P., de Nazelle, A., Triguero-Mas, M., Schembari, A., Cirach, M., Amoly, E., …
Nieuwenhuijsen, M. 2012b. “Surrounding Greenness and Exposure to Air Pollution
during Pregnancy: An Analysis of Personal Monitoring Data.” Environmental Health
Perspectives 120(9), 1286–1290. https://doi.org/10.1289/ehp.1104609
Dadvand, P., Figueras, F., Basagana, X., Beelen, R., Martinez, D., Cirach, M., …
Nieuwenhuijsen, M. J. 2013. “Ambient Air Pollution and Preeclampsia: A
Spatiotemporal Analysis.” Environmental Health Perspectives 121(11), 1365–1371.
Dadvand, P., Ostro, B., Amato, F., Figueras, F., Minguillón, M. C., Martinez, D., …
Nieuwenhuijsen, M. 2014. “Particulate Air Pollution and Preeclampsia: A Source-Based
Analysis.” Occupational and Environmental Medicine 71(8), 570–577.
https://doi.org/10.1136/oemed-2013-101693
Dadvand, P., Rankin, J., Rushton, S., & Pless-Mulloli, T. 2011. “Association between Maternal
Exposure to Ambient Air Pollution and Congenital Heart Disease: A Register-Based
Spatiotemporal Analysis.” American Journal of Epidemiology 173(2), 171–182.
https://doi.org/10.1093/aje/kwq342
Daher, N., Hasheminassab, S., Shafer, M. M., Schauer, J. J., & Sioutas, C. 2013. “Seasonal and
Spatial Variability in Chemical Composition and Mass Closure of Ambient Ultrafine
96
Particles in the Megacity of Los Angeles.” Environmental Sciences: Processes and
Impacts 15(1), 283–295. https://doi.org/10.1039/c2em30615h
Daly, A., & Zannetti, P. 2007. “An Introduction to Air Pollution – Definitions, Classifications,
and History.” In Ambient Air Pollution, edited by P. Zannetti, D. Al-Ajmi and S. Al-
Rashied, 1-14. The Arab School for Science and Technology and The EnviroComp
Institute.
Delfino, R. J., Quintana, P. J. E., Floro, J., Gastañaga, V. M., Samimi, B. S., Kleinman, M. T., …
McLaren, C. E. 2004. “Association of FEV1 in Asthmatic Children with Personal and
Microenvironmental Exposure to Airborne Particulate Matter.” Environmental Health
Perspectives 112(8), 932–941. https://doi.org/10.1289/ehp.6815
Dionisio, K. L., Nolte, C. G., Spero, T. L., Graham, S., Caraway, N., Foley, K. M., & Isaacs, K.
K. 2017. “Characterizing the Impact of Projected Changes in Climate and Air Quality on
Human Exposures to Ozone.” Journal of Exposure Science & Environmental
Epidemiology 27(3), 260–270. https://doi.org/10.1038/jes.2016.81
Dockery, D. W., Pope III, C. A., Xu, X., Spengler, J. D., Ware, J. H., Fay, M. E., … Speizer, F.
E. 1993. “An Association between Air Pollution and Mortality in Six US Cities.” The
New England Journal of Medicine 329(24), 1753–1759.
Donovan, G. H., Michael, Y. L., Butry, D. T., Sullivan, A. D., & Chase, J. M. 2011. “Urban
Trees and the Risk of Poor Birth Outcomes.” Health & Place 17(1), 390–393.
https://doi.org/10.1016/j.healthplace.2010.11.004
Duan, N. 1982. “Models for Human Exposure to Air Pollution.” Environment International 8(1-
6), 305-309. https://doi.org/10.1016/0160-4120(82)90041-1
Ebisu, K., Belanger, K., & Bell, M. L. 2014. “Association between Airborne PM2.5 Chemical
Constituents and Birth Weight – Implication of Buffer Exposure Assignment.”
Environmental Research Letters 9(8), 084007. https://doi.org/10.1088/1748-
9326/9/8/084007
Fabian, M. P., Lee, S. K., Underhill, L. J., Vermeer, K., Adamkiewicz, G., & Levy, J. I. 2016.
“Modeling Environmental Tobacco Smoke (ETS) Infiltration in Low-Income
Multifamily Housing before and after Building Energy Retrofits.” International Journal
of Environmental Research and Public Health 13(3), 1–15.
https://doi.org/10.3390/ijerph13030327
Fleisch, A. F., Luttmann ‐Gibson, H., Perng, W., Rifas ‐Shiman, S. L., Coull, B. A., Kloog, I., …
Oken, E. 2017. “Prenatal and Early Life Exposure to Traffic Pollution and
Cardiometabolic Health in Childhood.” Pediatric Obesity 12(1), 48–57.
Fleischer, N. L., Merialdi, M., van Donkelaar, A., Vadillo-Ortega, F., Martin, R. V., Betran, A.
P., … O’Neill, M. S. 2014. “Outdoor Air Pollution, Preterm Birth, and Low Birth
Weight: Analysis of the World Health Organization Global Survey on Maternal and
97
Perinatal Health.” Environmental Health Perspectives 122(4), 425–430.
https://doi.org/10.1289/ehp.1306837
Fujita, E. M., Campbell, D. E., Zielinska, B. B., Sagebiel, J. C., Bowen, J. L., Goliff, W. S., …
Lawson, D. R. 2003. “Diurnal and Weekday Variations in the Source Contributions of
Ozone Precursors in California’s South Coast Air Basin.” Journal of the Air and Waste
Management Association 53(7), 844–863.
https://doi.org/10.1080/10473289.2003.10466226
Gan, W. Q., Koehoorn, M., Davies, H. W., Demers, P. A., Tamburic, L., & Brauer, M. 2011.
“Long-Term Exposure to Traffic-Related Air Pollution and the Risk of Coronary Heart
Disease Hospitalization and Mortality.” Environmental Health Perspectives 119(4), 501–
507. https://doi.org/10.1289/ehp.1002511
Garcia, C. A., Yap, P. S., Park, H. Y., & Weller, B. L. 2016. “Association of Long-Term PM2.5
Exposure with Mortality Using Different Air Pollution Exposure Models: Impacts in
Rural and Urban California.” International Journal of Environmental Health Research
26(2), 145–157. https://doi.org/10.1080/09603123.2015.1061113
Gard, E. E., Kleeman, M. J., Gross, D. S., Hughes, L. S., Allen, J. O., Morrical, B. D., …
Prather, K. A. 1998. “Direct Observation of Heterogeneous Chemistry in the
Atmosphere.” Science 279(5354), 1184–1187.
https://doi.org/10.1126/science.279.5354.1184
Gauderman, W. J., Avol, E., Lurmann, F., Kuenzli, N., Gilliland, F., Peters, J., & McConnell, R.
2005. “Childhood Asthma and Exposure to Traffic and Nitrogen Dioxide.” Epidemiology
16(6), 737–743. https://doi.org/10.1097/01.ede.0000181308.51440.75
Gauderman, W. J., Urman, R., Avol, E., Berhane, K., McConnell, R., Rappaport, E., …
Gilliland, F. 2015. “Association of Improved Air Quality with Lung Development in
Children.” New England Journal of Medicine 372(10), 905–913.
https://doi.org/10.1056/NEJMoa1414123
Gehring, U., Wijga, A. H., Brauer, M., Fischer, P., de Jongste, J. C., Kerkhof, M., … Brunekreef,
B. 2010. “Traffic-Related Air Pollution and the Development of Asthma and Allergies
during the First 8 Years of Life.” American Journal of Respiratory and Critical Care
Medicine 181(6), 596–603. https://doi.org/10.1164/rccm.200906-0858OC
Gerharz, L. E., Krüger, A., & Klemm, O. 2009. “Applying Indoor and Outdoor Modeling
Techniques to Estimate Individual Exposure to PM2.5 from Personal GPS Profiles and
Diaries: A Pilot Study.” Science of the Total Environment 407(18), 5184–5193.
https://doi.org/10.1016/j.scitotenv.2009.06.006
Gesler, W. M., & Albert, D. P. 2000. “How Spatial Analysis Can Be Used in Medical
Geography.” In Spatial Analysis, GIS and Remote Sensing, edited by D. P. Albert, W. M.
Gesler, and B. Levergood, 19–46. Boca Raton: CRC Press.
98
Ghosh, G., Grewal, J., Mannisto, T., Mendola, P., Chen, Z., Xie, Y., & Laughon, S. K. 2014.
“Racial/Ethnic Differences in Pregnancy-Related Hypertensive Disease in Nulliparous
Women.” Ethnicity and Disease 24(3), 283–289.
Golledge, R. G. 1997. Spatial Behavior: A Geographic Perspective. New York: Guilford Press.
Goodchild, M. F. 2007. “Citizens as Voluntary Sensors: Spatial Data Infrastructure in the World
of Web 2.0.” International Journal of Spatial Data Infrastructures Research 2(2), 24–32.
Graham, S., Isaacs, K., McCurdy, T., Langstaff, J., Hartley, S., Cohen, J., & Holder, C. 2019.
“The Consolidated Human Activity Database (CHAD) Documentation and User’s
Guide.” US Environmental Protection Agency EPA-452/B-19-001.
https://www.epa.gov/sites/default/files/2019-11/documents/chadreport_october2019.pdf
Gray, S. C., Gelfand, A. E., & Miranda, M. L. 2011. “Hierarchical Spatial Modeling of
Uncertainty in Air Pollution and Birth Weight Study.” Statistics in Medicine 30(17),
2187–2198. https://doi.org/10.1002/sim.4234
Grineski, S. E., Herrera, J. M., Bulathsinhala, P., & Staniswalis, J. G. 2015. “Is There a Hispanic
Health Paradox in Sensitivity to Air Pollution? Hospital Admissions for Asthma, Chronic
Obstructive Pulmonary Disease and Congestive Heart Failure Associated with NO2 and
PM 2.5 in El Paso, TX, 2005-2010.” Atmospheric Environment 119, 314–321.
https://doi.org/10.1016/j.atmosenv.2015.08.027
Guo, T., Wang, Y., Zhang, H., Zhang, Y., Zhao, J., Wang, Q., … Ma, X. 2018. “The Association
between Ambient PM2.5 Exposure and the Risk of Preterm Birth in China: A
Retrospective Cohort Study.” Science of the Total Environment 633(12), 1453–1459.
https://doi.org/10.1016/j.scitotenv.2018.03.328
Habre, R., Coull, B., Moshier, E., Godbold, J., Grunin, A., Nath, A., … Koutrakis, P. 2014a.
“Sources of Indoor Air Pollution in New York City Residences of Asthmatic Children.”
Journal of Exposure Science & Environmental Epidemiology 24(3), 269–278.
https://doi.org/10.1038/jes.2013.74
Habre, R., Girguis, M., Urman, R., Fruin, S., Lurmann, F., Shafer, M., … Gilliland, F. 2021.
“Contribution of Tailpipe and Non-Tailpipe Traffic Sources to Quasi-Ultrafine, Fine and
Coarse Particulate Matter in Southern California.” Journal of the Air and Waste
Management Association 71(2), 209–230.
https://doi.org/10.1080/10962247.2020.1826366
Habre, R., Moshier, E., Castro, W., Nath, A., Grunin, A., Rohr, A., … Koutrakis, P. 2014b. “The
Effects of PM2.5 and Its Components from Indoor and Outdoor Sources on Cough and
Wheeze Symptoms in Asthmatic Children.” Journal of Exposure Science &
Environmental Epidemiology 24(4), 380–387. https://doi.org/10.1038/jes.2014.21
Hadley, O. L. 2017. “Background PM2.5 Source Apportionment in the Remote Northwestern
United States.” Atmospheric Environment 167, 298–308.
https://doi.org/10.1016/j.atmosenv.2017.08.030
99
Ham, W., Vijayan, A., Schulte, N., & Herner, J. D. 2017. “Commuter Exposure to PM2.5, BC,
and UFP in Six Common Transport Microenvironments in Sacramento, California.”
Atmospheric Environment 167, 335–345. https://doi.org/10.1016/j.atmosenv.2017.08.024
Han, F., Kota, S. H., Wang, Y., & Zhang, H. 2017. “Source Apportionment of PM2.5 in Baton
Rouge, Louisiana during 2009-2014.” Science of the Total Environment 586, 115–126.
https://doi.org/10.1016/j.scitotenv.2017.01.189
Hänninen, O., Kruize, H., Lebret, E., & Jantunen, M. 2003. “EXPOLIS Simulation Model: PM2.5
Application and Comparison with Measurements in Helsinki.” Journal of Exposure
Analysis & Environmental Epidemiology 13(1), 74–85.
https://doi.org/10.1038/sj.jea.7500260
Harley, R. A., Hannlgan, M. P., & Cass, G. R. 1992. “Respeciation of Organic Gas Emissions
and the Detection of Excess Unburned Gasoline in the Atmosphere.” Environmental
Science & Technology 26(12), 2395–2408. https://doi.org/10.1021/es00036a010
Harris, G., Thompson, W. D., Fitzgerald, E., & Wartenberg, D. 2014. “The Association of PM2.5
with Full Term Low Birth Weight at Different Spatial Scales.” Environmental Research
134, 427–434. https://doi.org/10.1016/j.envres.2014.05.034
Hasheminassab, S., Daher, N., Ostro, B. D., & Sioutas, C. 2014a. “Long-Term Source
Apportionment of Ambient Fine Particulate Matter (PM2.5) in the Los Angeles Basin: A
Focus on Emissions Reduction from Vehicular Sources.” Environmental Pollution 193,
54–64. https://doi.org/10.1016/j.envpol.2014.06.012
Hasheminassab, S., Daher, N., Saffari, A., Wang, D., Ostro, B. D., & Sioutas, C. 2014b. “Spatial
and Temporal Variability of Sources of Ambient Fine Particulate Matter (PM2.5) in
California.” Atmospheric Chemistry and Physics 14(22), 12085–12097.
https://doi.org/10.5194/acp-14-12085-2014, 2014
Hasheminassab, S., Daher, N., Schauer, J. J., & Sioutas, C. 2013. “Source Apportionment and
Organic Compound Characterization of Ambient Ultrafine Particulate Matter (PM) in the
Los Angeles Basin.” Atmospheric Environment 79, 529–539.
https://doi.org/10.1016/j.atmosenv.2013.07.040
Hasheminassab, S., Daher, N., Shafer, M. M., Schauer, J. J., Delfino, R. J., & Sioutas, C. 2014c.
“Chemical Characterization and Source Apportionment of Indoor and Outdoor Fine
Particulate Matter (PM2.5) in Retirement Communities of the Los Angeles Basin.”
Science of the Total Environment 490, 528–537.
https://doi.org/10.1016/j.scitotenv.2014.05.044
He, C., Morawska, L., Hitchins, J., & Gilbert, D. 2004. “Contribution from Indoor Sources to
Particle Number and Mass Concentrations in Residential Houses.” Atmospheric
Environment 38(21), 3405–3415. https://doi.org/10.1016/j.atmosenv.2004.03.027
100
Heo, J. B., Hopke, P. K., & Yi, S. M. 2009. “Source Apportionment of PM2.5 in Seoul, Korea.”
Atmospheric Chemistry and Physics 9(14), 4957–4971. https://doi.org/10.5194/acp-9-
4957-2009
Heo, J., Dulger, M., Olson, M. R., McGinnis, J. E., Shelton, B. R., Matsunaga, A., … Schauer, J.
J. 2013. “Source Apportionments of PM2.5 Organic Carbon Using Molecular Marker
Positive Matrix Factorization and Comparison of Results from Different Receptor
Models.” Atmospheric Environment 73, 51–61.
https://doi.org/10.1016/j.atmosenv.2013.03.004
Hickey, N., Boscarato, I., & Kaspar, J. 2014. “Air Pollution from Mobile Sources: Formation
and Effects and Abatement Strategies.” In Current Environmental Issues and Challenges,
edited by G. Cao and R, Orrù, 15–43. Dordrecht: Springer. https://doi.org/10.1007/978-
94-017-8777-2_2
Hopke, P. K. 2003. “Recent Developments in Receptor Modeling.” Journal of Chemometrics
17(5), 255–265. https://doi.org/10.1002/cem.796
Hopke, P. K. 2016. “Review of Receptor Modeling Methods for Source Apportionment.”
Journal of the Air and Waste Management Association 66(3), 237–259.
https://doi.org/10.1080/10962247.2016.1140693
Houston, D., Li, W., & Wu, J. 2014. “Disparities in Exposure to Automobile and Truck Traffic
and Vehicle Emissions Near the Los Angeles-Long Beach Port Complex.” American
Journal of Public Health 104(1), 156–164. https://doi.org/10.2105/AJPH.2012.301120
Houston, D., Wu, J., Ong, P., & Winer, A. 2004. “Structural Disparities of Urban Traffic in
Southern California: Implications for Vehicle-Related Air Pollution Exposure in Minority
and High-Poverty Neighborhoods.” Journal of Urban Affairs 26(5), 565–592.
https://doi.org/10.1111/j.0735-2166.2004.00215.x
Howard-Reed, C., Wallace, L. A., & 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(2), 147–159. https://doi.org/10.1080/10473289.2002.10470775
Hsu, H. H. L., Chiu, Y. H. M., Coull, B. A., Kloog, I., Schwartz, J., Lee, A., … Wright, R. J.
2015. “Prenatal Particulate Air Pollution and Asthma Onset in Urban Children:
Identifying Sensitive Windows and Sex Differences.” American Journal of Respiratory
and Critical Care Medicine 192(9), 1052–1059. https://doi.org/10.1164/rccm.201504-
0658OC
Hsu, W. T., Chen, J. L., Candice Lung, S. C., & Chen, Y. C. 2020. “PM2.5 Exposure of Various
Microenvironments in a Community: Characteristics and Applications.” Environmental
Pollution 263, 114522. https://doi.org/10.1016/j.envpol.2020.114522
Hu, H., Ha, S., Henderson, B. H., Warner, T. D., Roth, J., Kan, H., & Xu, X. 2015. “Association
of Atmospheric Particulate Matter and Ozone with Gestational Diabetes Mellitus.”
101
Environmental Health Perspectives 123(9), 853–859.
https://doi.org/10.1289/ehp.1408456
Hu, T., Singer, B. C., & Logue, J. M. 2012. “Compilation of Published PM2.5 Emission Rates for
Cooking, Candles and Incense for Use in Modeling of Exposures in Residences.”
Lawrence Berkeley National Lab LBNL-5890E. https://doi.org/10.2172/1172959
Hu, X., Belle, J. H., Meng, X., Wildani, A., Waller, L. A., Strickland, M. J., & Liu, Y. 2017.
“Estimating PM2.5 Concentrations in the Conterminous United States Using the Random
Forest Approach.” Environmental Science & Technology 51(12), 6936–6944.
https://doi.org/10.1021/acs.est.7b01210
Huang, C., Nichols, C., Liu, Y., Zhang, Y., Liu, X., Gao, S., … Ren, A. 2015. “Ambient Air
Pollution and Adverse Birth Outcomes: A Natural Experiment Study.” Population Health
Metrics 13(1), 1–7.
Huang, J., Deng, F., Wu, S., & Guo, X. 2012. “Comparisons of Personal Exposure to PM2.5 and
CO by Different Commuting Modes in Beijing, China.” Science of the Total Environment
425, 52–59. https://doi.org/10.1016/j.scitotenv.2012.03.007
Hyder, A., Lee, H. J., Ebisu, K., Koutrakis, P., Belanger, K., & Bell, M. L. 2014. “PM2.5
Exposure and Birth Outcomes: Use of Satellite- and Monitor-Based Data.” Epidemiology
25(1), 58–67. https://doi.org/10.1097/EDE.0000000000000027.PM
Jankowska, M. M., Natarajan, L., Godbole, S., Meseck, K., Sears, D. D., Patterson, R. E., &
Kerr, J. 2017. “Kernel Density Estimation as a Measure of Environmental Exposure
Related to Insulin Resistance in Breast Cancer Survivors.” Cancer Epidemiology
Biomarkers & Prevention 26(7), 1078–1084. https://doi.org/10.1158/1055-9965.EPI-16-
0927
Jankowska, M. M., Schipperijn, J., & Kerr, J. 2015. “A Framework for Using GPS Data in
Physical Activity and Sedentary Behavior Studies.” Exercise and Sport Sciences Reviews
43(1), 48-56. https://doi.org/10.1249/JES.0000000000000035
Jedrychowski, W., Bendkowska, I., Flak, E., Penar, A., Jacek, R., Kaim, I., … Perera, F. P. 2004.
“Estimated Risk for Altered Fetal Growth Resulting from Exposure to Fine Particles
during Pregnancy: An Epidemiologic Prospective Cohort Study in Poland.”
Environmental Health Perspectives 112(14), 1398–1402.
https://doi.org/10.1289/ehp.7065
Jedrychowski, W. A., Perera, F. P., Maugeri, U., Spengler, J., Mroz, E., Flak, E., … Jacek, R.
2012. “Prohypertensive Effect of Gestational Personal Exposure to Fine Particulate
Matter. Prospective Cohort Study in Non-Smoking and Non-Obese Pregnant Women.”
Cardiovascular Toxicology 12(3), 216–225. https://doi.org/10.1007/s12012-012-9157-z
Jedrychowski, W., Perera, F., Mrozek-Budzyn, D., Mroz, E., Flak, E., Spengler, J. D., …
Skolicki, Z. 2009. “Gender Differences in Fetal Growth of Newborns Exposed Prenatally
102
to Airborne Fine Particulate Matter.” Environmental Research 109(4), 447–456.
https://doi.org/10.1016/j.envres.2009.01.009
Jedrychowski, W. A., Perera, F. P., Pac, A., Jacek, R., Whyatt, R. M., Spengler, J. D., …
Sochacka-Tatara, E. 2006. “Variability of Total Exposure to PM2.5 Related to Indoor and
Outdoor Pollution Sources. Krakow Study in Pregnant Women.” Science of the Total
Environment 366(1), 47–54. https://doi.org/10.1016/j.scitotenv.2005.08.001
Jenkins, P. L., Phillips, T. J., Mulberg, E. J., & Hui, S. P. 1992. “Activity Patterns of
Californians: Use of and Proximity to Indoor Pollutant Sources.” Atmospheric
Environment Part A, General Topics 26(12), 2141–2148. https://doi.org/10.1016/0960-
1686(92)90402-7
Jiao, W., Frey, H. C., & Cao, Y. 2012. “Assessment of Inter-Individual, Geographic, and
Seasonal Variability in Estimated Human Exposure to Fine Particles.” Environmental
Science & Technology 46(22), 12519–12526. https://doi.org/10.1111/j.1743-
6109.2008.01122.x.Endothelial
Johnson, T. R., Langstaff, J. E., Graham, S., Fujita, E. M., & Campbell, D. E. 2018. “A
Multipollutant Evaluation of APEX Using Microenvironmental Ozone, Carbon
Monoxide, and Particulate Matter (PM2.5) Concentrations Measured in Los Angeles by
the Exposure Classification Project.” Cogent Environmental Science 4(1), 1–26.
https://doi.org/10.1080/23311843.2018.1453022
Kaur, S., Nieuwenhuijsen, M. J., & Colvile, R. N. 2007. “Fine Particulate Matter and Carbon
Monoxide Exposure Concentrations in Urban Street Transport Microenvironments.”
Atmospheric Environment 41(23), 4781–4810.
https://doi.org/10.1016/j.atmosenv.2007.02.002
Kim, D., Sass-Kortsak, A., Purdham, J. T., Dales, R. E., & Brook, J. R. 2005. “Sources of
Personal Exposure to Fine Particles in Toronto, Ontario, Canada.” Journal of the Air and
Waste Management Association 55(8), 1134–1146.
https://doi.org/10.1080/10473289.2005.10464710
Kim, J. J., Smorodinsky, S., Lipsett, M., Singer, B. C., Hodgson, A. T., & Ostro, B. 2004.
“Traffic-Related Air Pollution Near Busy Roads: The East Bay Children’s Respiratory
Health Study.” American Journal of Respiratory and Critical Care Medicine 170(5),
520–526. https://doi.org/10.1164/rccm.200403-281OC
Knipping, E. M., & Dabdub, D. 2003. “Impact of Chlorine Emissions from Sea-Salt Aerosol on
Coastal Urban Ozone.” Environmental Science & Technology 37(2), 275–284.
Koistinen, K. J., Edwards, R. D., Mathys, P., Ruuskanen, J., Künzli, N., & Jantunen, M. J. 2004.
“Sources of Fine Particulate Matter in Personal Exposures and Residential Indoor,
Residential Outdoor and Workplace Microenvironments in the Helsinki Phase of the
EXPOLIS Study.” Scandinavian Journal of Work, Environment & Health 30(suppl 2),
36-46.
103
Krewski, D., Burnett, R. T., Goldberg, M. S., Hoover, B. K., Siemiatycki, J., Jerrett, M., …
White, W. H. 2003. “Overview of the Reanalysis of the Harvard Six Cities Study and
American Cancer Society Study of Particulate Air Pollution and Mortality.” Journal of
Toxicology and Environmental Health Part A 66(16-19), 1507-1552.
https://doi.org/10.1080/15287390306424
Kwan, M. P. 1999. “Gender and Individual Access to Urban Opportunities: A Study Using
Space–Time Measures.” The Professional Geographer 51(2), 210–227.
https://doi.org/10.1111/0033-0124.00158
Lai, H. K., Kendall, M., Ferrier, H., Lindup, I., Alm, S., Hänninen, O., … Nieuwenhuijsen, M. J.
2004. “Personal Exposures and Microenvironment Concentrations of PM2.5, VOC, NO2
and CO in Oxford, UK.” Atmospheric Environment 38(37), 6399–6410.
https://doi.org/10.1016/j.atmosenv.2004.07.013
Larson, T., Gould, T., Simpson, C., Liu, L. J. S., Claiborn, C., & Lewtas, J. 2004. “Source
Apportionment of Indoor, Outdoor, and Personal PM2.5 in Seattle, Washington, Using
Positive Matrix Factorization.” Journal of the Air and Waste Management Association
54(9), 1175–1187. https://doi.org/10.1080/10473289.2004.10470976
Lawless, P. A., Rodes, C. E., & Ensor, D. S. 2004. “Multiwavelength Absorbance of Filter
Deposits for Determination of Environmental Tobacco Smoke and Black Carbon.”
Atmospheric Environment 38(21), 3373–3383.
https://doi.org/10.1016/j.atmosenv.2004.03.038
Li, Q., Zheng, Y., Xie, X., Chen, Y., Liu, W., & Ma, W. Y. 2008. “Mining User Similarity Based
on Location History.” In Proceedings of the 16th ACM SIGSPATIAL international
conference on Advances in geographic information systems, 1-10.
https://doi.org/10.1145/1463434.1463477
Li, R., Hopke, P. K., Dozier, A., Thurston, S. W., Thevenet-Morrison, K., Croft, D., … Rich, D.
Q. 2019. “Term Birth Weight and Ambient Air Pollutant Concentrations during
Pregnancy, among Women Living in Monroe County, New York.” Journal of Exposure
Science & Environmental Epidemiology 29(4), 500–509. https://doi.org/10.1038/s41370-
019-0131-8
Lim, S., Kim, J., Kim, T., Lee, K., Yang, W., Jun, S., & Yu, S. 2012. “Personal Exposures to
PM 2.5 and Their Relationships with Microenvironmental Concentrations.” Atmospheric
Environment 47, 407–412. https://doi.org/10.1016/j.atmosenv.2011.10.043
Lippmann, M. 2009. “Semi-Continuous Speciation Analyses for Ambient Air Particulate Matter:
An Urgent Need for Health Effects Studies.” Journal of Exposure Science &
Environmental Epidemiology 19(3), 235–247. https://doi.org/10.1038/jes.2008.65
Liu, L. J. S., Box, M., Kalman, D., Kaufman, J., Koenig, J., Larson, T., … Wallace, L. 2003.
“Exposure Assessment of Particulate Matter for Susceptible Populations in Seattle.”
Environmental Health Perspectives 111(7), 909–918. https://doi.org/10.1289/ehp.6011
104
Long, C. M., Suh, H. H., & 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. https://doi.org/10.1080/10473289.2000.10464154
Lough, G. C., Schauer, J. J., Park, J. S., Shafer, M. M., Deminter, J. T., & Weinstein, J. P. 2005.
“Emissions of Metals Associated with Motor Vehicle Roadways.” Environmental Science
& Technology 39(3), 826–836. https://doi.org/10.1021/es048715f
MacIntosh, D. L., & Spengler, J. D. 2000. “Human Exposure Assessment.” World Health
Organization & International Programme on Chemical Safety.
https://apps.who.int/iris/handle/10665/42181
Majd, E., McCormack, M., Hansel, N., & Koehler, K. 2018. “Characterizing Personal Fine and
Ultrafine Particle Exposures among Inner-City Children with Asthma.” In ISEE
Conference Abstracts, 2018(1).
Mao, G., Nachman, R. M., Sun, Q., Zhang, X., Koehler, K., Chen, Z., … Wang, X. 2017.
“Individual and Joint Effects of Early-Life Ambient PM2.5 Exposure and Maternal Pre-
Pregnancy Obesity on Childhood Overweight or Obesity.” Environmental Health
Perspectives 125(6), 1–10. https://doi.org/10.1289/EHP261
Masiol, M., Hopke, P. K., Felton, H. D., Frank, B. P., Rattigan, O. V., Wurth, M. J., & LaDuke,
G. H. 2017. “Source Apportionment of PM2.5 Chemically Speciated Mass and Particle
Number Concentrations in New York City.” Atmospheric Environment 148, 215–229.
https://doi.org/10.1016/j.atmosenv.2016.10.044
Maykut, N. N., Lewtas, J., Kim, E., & Larson, T. V. 2003. “Source Apportionment of PM2.5 at an
Urban IMPROVE Site in Seattle, Washington.” Environmental Science & Technology
37(22), 5135–5142. https://doi.org/10.1021/es030370y
McCurdy, T., Glen, G., Smith, L., & Lakkadi, Y. 2000. “The National Exposure Research
Laboratory’s Consolidated Human Activity Database.” Journal of Exposure Analysis &
Environmental Epidemiology 10(6), 566–578. https://doi.org/10.1038/sj.jea.7500114
McGuinn, L. A., Ward-Caviness, C. K., Neas, L. M., Schneider, A., Diaz-Sanchez, D., Cascio,
W. E., … Devlin, R. B. 2016. “Association between Satellite-Based Estimates of Long-
Term PM2.5 Exposure and Coronary Artery Disease.” Environmental Research 145, 9 –
17. https://doi.org/10.1016/j.envres.2015.10.026
Meng, Q. Y., Spector, D., Colome, S., & Turpin, B. 2009. “Determinants of Indoor and Personal
Exposure to PM2.5 of Indoor and Outdoor Origin during the RIOPA Study.” Atmospheric
Environment 43(36), 5750–5758. https://doi.org/10.1016/j.atmosenv.2009.07.066
Meng, Q. Y., Turpin, B. J., Lee, J.-H., Polidori, A., Weisel, C. P., Morandi, M., … Winer, A.
2007. “How Does Infiltration Behavior Modify the Composition of Ambient PM2.5 in
Indoor Spaces? An Analysis of RIOPA Data.” Environmental Science & Technology
41(21), 7315–7321. https://doi.org/10.1021/es070037k
105
Minguillón, M. C., Schembari, A., Triguero-Mas, M., de Nazelle, A., Dadvand, P., Figueras, F.,
… Querol, X. 2012. “Source Apportionment of Indoor, Outdoor and Personal PM2.5
Exposure of Pregnant Women in Barcelona, Spain.” Atmospheric Environment 59, 426–
436. https://doi.org/10.1016/j.atmosenv.2012.04.052
Molnár, P., Johannesson, S., & Quass, U. 2014. “Source Apportionment of PM2.5 Using Positive
Matrix Factorization (PMF) and PMF with Factor Selection.” Aerosol and Air Quality
Research 14(3), 725–733. https://doi.org/10.4209/aaqr.2013.11.0335
Mueller, D., Uibel, S., Braun, M., Klingelhoefer, D., Takemura, M., & Groneberg, D. A. 2011.
“Tobacco Smoke Particles and Indoor Air Quality (ToPIQ) - The Protocol of a New
Study.” Journal of Occupational Medicine and Toxicology 6(1), 1–5.
https://doi.org/10.1186/1745-6673-6-35
Müller, A. L. H., Bizzi, C. A., Pereira, J. S. F., Mesko, M. F., Moraes, D. P., Floresa, E. M. M.,
& Muller, E. I. 2011. “Bromine and Chlorine Determination in Cigarette Tobacco Using
Microwave-Induced Combustion and Inductively Coupled Plasma Optical Emission
Spectrometry.” Journal of the Brazilian Chemical Society 22(9), 1649–1655.
https://doi.org/10.1590/s0103-50532011000900005
Nachman, R. M., Mao, G., Zhang, X., Hong, X., Chen, Z., Soria, C. S., … Wang, X. 2016.
“Intrauterine Inflammation and Maternal Exposure to Ambient PM2.5 during
Preconception and Specific Periods of Pregnancy: The Boston Birth Cohort.”
Environmental Health Perspectives 124(10), 1608–1615. https://doi.org/10.1289/EHP243
Nazaroff, W. W., & Singer, B. C. 2004. “Inhalation of Hazardous Air Pollutants from
Environmental Tobacco Smoke in US Residences.” Journal of Exposure Analysis &
Environmental Epidemiology 14(1), S71-S77. https://doi.org/10.1038/sj.jea.7500361
Nazelle, A. De, Seto, E., Donaire-Gonzalez, D., Mendez, M., Matamala, J., Nieuwenhuijsen, M.
J., & Jerrett, M. 2013. “Improving Estimates of Air Pollution Exposure through
Ubiquitous Sensing Technologies.” Environmental Pollution 176, 92–99.
https://doi.org/10.1016/j.envpol.2012.12.032
Newsome, T. H., Walcott, W. A., & Smith, P. D. 1998. “Urban Activity Spaces: Illustrations and
Application of a Conceptual Model for Integrating the Time and Space Dimensions.”
Transportation 25(4), 357–377. https://doi.org/10.1023/a:1005082827030
Ng, C., Malig, B., Hasheminassab, S., Sioutas, C., Basu, R., & Ebisu, K. 2017. “Source
Apportionment of Fine Particulate Matter and Risk of Term Low Birth Weight in
California: Exploring Modification by Region and Maternal Characteristics.” Science of
The Total Environment 605, 647–654. https://doi.org/10.1016/j.scitotenv.2017.06.053
Norris, G., Duvall, R., Brown, S., & Bai, S. 2014. “EPA Positive Matrix Factorization (PMF) 5.0
Fundamentals and User Guide.” US Environmental Protection Agency EPA/600/R-
14/108. https://www.epa.gov/sites/default/files/2015-
02/documents/pmf_5.0_user_guide.pdf
106
Nyhan, M., Grauwin, S., Britter, R., Misstear, B., McNabola, A., Laden, F., … Ratti, C. 2016.
““Exposure Track” - The Impact of Mobile-Device-Based Mobility Patterns on
Quantifying Population Exposure to Air Pollution.” Environmental Science &
Technology 50(17), 9671–9681. https://doi.org/10.1021/acs.est.6b02385
Onat, B., Sahin, U. A., & Akyuz, T. 2013. “Elemental Characterization of PM2.5 and PM1 in
Dense Traffic Area in Istanbul, Turkey.” Atmospheric Pollution Research, 4(1), 101–105.
https://doi.org/10.5094/APR.2013.010
Ott, W. R., Steinemann, A. C., & Wallace, L. A. 2006. Exposure Analysis. Boca Raton: CRC
Press.
Özkaynak, H., Xue, J., Spengler, J., Wallace, L., Pellizzari, E., & Jenkins, P. 1996. “Personal
Exposure to Airborne Particles and Metals: Results from the Particle Team Study in
Riverside, California.” Journal of Exposure Analysis & Environmental Epidemiology
6(1), 57–78.
Paatero, P. 1997. “Least Squares Formulation of Robust Non-Negative Factor Analysis.”
Chemometrics and Intelligent Laboratory Systems 37(1), 23–35.
https://doi.org/10.1016/S0169-7439(96)00044-5
Paatero, P., & Tapper, U. 1994. “Positive Matrix Factorization: A Non ‐Negative Factor Model
with Optimal Utilization of Error Estimates of Data Values.” Environmetrics 5(2), 111 –
126.
Painter, D. E. 1974. Air Pollution Technology. Reston: Reston Publishing Company Inc.
Palm, B. B., Peng, Q., Fredrickson, C. D., Lee, B. H., Garofalo, L. A., Pothier, M. A., …
Thornton, J. A. 2020. “Quantification of Organic Aerosol and Brown Carbon Evolution
in Fresh Wildfire Plumes.” Proceedings of the National Academy of Sciences of the
United States of America 117(47), 29469–29477.
https://doi.org/10.1073/pnas.2012218117
Pastor, M., Sadd, J. L., & Morello-Frosch, R. 2004. “Waiting to Inhale: The Demographics of
Toxic Air Release Facilities in 21st-Century California.” Social Science Quarterly 85(2),
420–440. https://doi.org/10.1111/j.0038-4941.2004.08502010.x
Pedersen, M., Giorgis-Allemand, L., Bernard, C., Aguilera, I., Andersen, A.-M. N., Ballester, F.,
… Danileviciute, A. 2013. “Ambient Air Pollution and Low Birthweight: A European
Cohort Study (ESCAPE).” The Lancet Respiratory Medicine 1(9), 695–704.
Pekney, N., Davidson, C., Zhou, L., & Hopke, P. 2006. “Application of PSCF and CPF to PMF-
Modeled Sources of PM2.5 in Pittsburgh.” Aerosol Science and Technology 40(10), 952–
961. https://doi.org/10.1080/02786820500543324
Perchoux, C., Chaix, B., Brondeel, R., & Kestens, Y. 2016. “Residential Buffer, Perceived
Neighborhood, and Individual Activity Space: New Refinements in the Definition of
107
Exposure Areas – The RECORD Cohort Study.” Health & Place 40, 116–122.
https://doi.org/10.1016/j.healthplace.2016.05.004
Percy, Z., DeFranco, E., Xu, F., Hall, E. S., Haynes, E. N., Jones, D., … Chen, A. 2019.
“Trimester Specific PM2.5 Exposure and Fetal Growth in Ohio, 2007–2010.”
Environmental Research 171, 111–118. https://doi.org/10.1016/j.envres.2019.01.031
Pereira, G., Bell, M. L., Lee, H. J., Koutrakis, P., & Belanger, K. 2014. “Sources of Fine
Particulate Matter and Risk of Preterm Birth in Connecticut, 2000-2006: A Longitudinal
Study.” Environmental Health Perspectives 122(10), 1117–1122.
https://doi.org/10.1289/ehp.1307741
Pérez-Torres, R., Torres-Huitzil, C., & Galeana-Zapién, H. 2016. “Full on-Device Stay Points
Detection in Smartphones for Location-Based Mobile Applications.” Sensors 16(10).
https://doi.org/10.3390/s16101693
Price, P. N., Shehabi, A., & Chan, R. 2006. “Indoor-Outdoor Air Exchange Rates of California
Apartments and Commercial Buildings.” Lawrence Berkeley National Lab LBNL-60682.
https://doi.org/10.2172/1010623
Pulido, L., Sidawi, S., & Vos, R. O. 1996. “An Archaeology of Environmental Racism in Los
Angeles.” Urban Geography 17(5), 419–439.
Pun, V. C., Kazemiparkouhi, F., Manjourides, J., & Suh, H. H. 2017. “Long-Term PM2.5
Exposure and Respiratory, Cancer, and Cardiovascular Mortality in Older US Adults.”
American Journal of Epidemiology 186(8), 961–969. https://doi.org/10.1093/aje/kwx166
Qiu, Z., & Cao, H. 2020. “Commuter Exposure to Particulate Matter in Urban Public
Transportation of Xi’an, China.” Journal of Environmental Health Science and
Engineering 18(2), 451–462. https://doi.org/10.1007/s40201-020-00473-0
Rabinovitch, N., Adams, C. D., Strand, M., Koehler, K., & Volckens, J. 2016. “Within-
Microenvironment Exposure to Particulate Matter and Health Effects in Children with
Asthma: A Pilot Study Utilizing Real-Time Personal Monitoring with GPS Interface.”
Environmental Health 15(1), 1–10. https://doi.org/10.1186/s12940-016-0181-5
Rammah, A., Whitworth, K. W., Han, I., Chan, W., & Symanski, E. 2019. “PM2.5 Metal
Constituent Exposure and Stillbirth Risk in Harris County, Texas.” Environmental
Research 176, 108516. https://doi.org/10.1016/j.envres.2019.05.047
Reff, A., Bhave, P. V., Simon, H., Pace, T. G., Pouliot, G. A., Mobley, J. D., & Houyoux, M.
2009. “Emissions Inventory of PM2.5 Trace Elements across the United States.”
Environmental Science & Technology 43(15), 5790–5796.
https://doi.org/10.1021/es802930x
Rich, D. Q., Liu, K., Zhang, J., Thurston, S. W., Stevens, T. P., Pan, Y., … Zhang, J. 2015.
“Differences in Birth Weight Associated with the 2008 Beijing Olympics Air Pollution
108
Reduction: Results from a Natural Experiment.” Environmental Health Perspectives
123(9), 880–887. https://doi.org/10.1289/ehp.1408795
Riggs, D. W., Yeager, R., Conklin, D. J., DeJarnett, N., Keith, R. J., DeFilippis, A. P., …
Bhatnagar, A. 2021. “Residential Proximity to Greenness Mitigates the Hemodynamic
Effects of Ambient Air Pollution.” American Journal of Physiology-Heart and
Circulatory Physiology 320(3), H1102–H1111.
https://doi.org/10.1152/ajpheart.00689.2020
Ritz, B., Wilhelm, M., Hoggatt, K. J., & Ghosh, J. K. C. 2007. “Ambient Air Pollution and
Preterm Birth in the Environment and Pregnancy Outcomes Study at the University of
California, Los Angeles.” American Journal of Epidemiology 166(9), 1045–1052.
https://doi.org/10.1093/aje/kwm181
Rohr, A. C., Habre, R., Godbold, J., Moshier, E., Schachter, N., Kattan, M., … Koutrakis, P.
2014. “Asthma Exacerbation is Associated with Particulate Matter Source Factors in
Children in New York City.” Air Quality, Atmosphere & Health 7(2), 239–250.
https://doi.org/10.1007/s11869-013-0230-y
Rohr, A. C., & Wyzga, R. E. 2012. “Attributing Health Effects to Individual Particulate Matter
Constituents.” Atmospheric Environment 62, 130–152.
https://doi.org/10.1016/j.atmosenv.2012.07.036
Romão, R., Pereira, L. A. A., Saldiva, P. H. N., Pinheiro, P. M., Braga, A. L. F., & Martins, L. C.
2013. “The Relationship between Low Birth Weight and Exposure to Inhalable
Particulate Matter.” Cadernos de Saude Publica 29, 1101–1108.
Rosa, M. J., Hair, G. M., Just, A. C., Kloog, I., Svensson, K., Pizano-Zárate, M. L., … Sanders,
A. P. 2020. “Identifying Critical Windows of Prenatal Particulate Matter (PM 2.5)
Exposure and Early Childhood Blood Pressure.” Environmental Research 182, 109073.
https://doi.org/10.1016/j.envres.2019.109073
Rosenbaum, A., Hartley, S., Holder, C., Turley, A., & Graham, S. 2008. “Nitrogen Dioxide
(NO2) Exposure Assessment in Support of U. S. EPA’s NAAQS Review: Application of
AERMOD and APEX to Philadelphia County.” Epidemiology 19(6), S344–S345.
https://doi.org/10.1097/01.ede.0000340563.36534.ff
Rossen, L. M. 2014. “Neighbourhood Economic Deprivation Explains Racial/Ethnic Disparities
in Overweight and Obesity among Children and Adolescents in the USA.” Journal of
Epidemiology and Community Health 68(2), 123–129. https://doi.org/10.1136/jech-2012-
202245
Runa, F., Islam, M. S., Jeba, F., & Salam, A. 2021. “Light Absorption Properties of Brown
Carbon from Biomass Burning Emissions.” Environmental Science and Pollution
Research 29, 21012-21022. https://doi.org/10.1007/s11356-021-17220-z
Rundle, A., Hoepner, L., Hassoun, A., Oberfield, S., Freyer, G., Holmes, D., … Whyatt, R. 2012.
“Association of Childhood Obesity with Maternal Exposure to Ambient Air Polycyclic
109
Aromatic Hydrocarbons during Pregnancy.” American Journal of Epidemiology 175(11),
1163–1172. https://doi.org/10.1093/aje/kwr455
Ryan, P. H., Brokamp, C., Fan, Z.-H., & Rao, M. B. 2015. “Analysis of Personal and Home
Characteristics Associated with the Elemental Composition of PM2.5 in Indoor, Outdoor,
and Personal Air in the RIOPA Study.” Health Effects Institute 185, 3–40.
http://europepmc.org/abstract/MED/26934775
Saffari, A., Daher, N., Shafer, M. M., Schauer, J. J., & Sioutas, C. 2013. “Seasonal and Spatial
Variation in Reactive Oxygen Species Activity of Quasi-Ultrafine Particles (PM 0.25) in
the Los Angeles Metropolitan Area and Its Association with Chemical Composition.”
Atmospheric Environment 79, 566–575. https://doi.org/10.1016/j.atmosenv.2013.07.058
Sarnat, J. A., Sarnat, S. E., Flanders, W. D., Chang, H. H., Mulholland, J., Baxter, L., …
Özkaynak, H. 2013. “Spatiotemporally Resolved Air Exchange Rate as a Modifier of
Acute Air Pollution-Related Morbidity in Atlanta.” Journal of Exposure Science &
Environmental Epidemiology 23(6), 606–615. https://doi.org/10.1038/jes.2013.32
Sarnat, S. E., Coull, B. A., Schwartz, J., Gold, D. R., & Suh, H. H. 2006. “Factors Affecting the
Association between Ambient Concentrations and Personal Exposures to Particles and
Gases.” Environmental Health Perspectives 114(5), 649–654.
https://doi.org/10.1289/ehp.8422
Sarnat, S. E., Sarnat, J. A., Mulholland, J., Isakov, V., Özkaynak, H., Chang, H. H., … Tolbert,
P. E. 2013. “Application of Alternative Spatiotemporal Metrics of Ambient Air Pollution
Exposure in a Time-Series Epidemiological Study in Atlanta.” Journal of Exposure
Science & Environmental Epidemiology 23(6), 593–605.
https://doi.org/10.1038/jes.2013.41
Savitz, D. A., Bobb, J. F., Carr, J. L., Clougherty, J. E., Dominici, F., Elston, B., … Matte, T. D.
2014. “Ambient Fine Particulate Matter, Nitrogen Dioxide, and Term Birth Weight in
New York, New York.” American Journal of Epidemiology 179(4), 457–466.
https://doi.org/10.1093/aje/kwt268
Schachter, E. N., Moshier, E., Habre, R., Rohr, A., Godbold, J., Nath, A., … Kattan, M. 2016.
“Outdoor Air Pollution and Health Effects in Urban Children with Moderate to Severe
Asthma.” Air Quality, Atmosphere & Health 9(3), 251–263.
https://doi.org/10.1007/s11869-015-0335-6
Schachter, E. N., Rohr, A., Habre, R., Koutrakis, P., Moshier, E., Nath, A., … Kattan, M. 2020.
“Indoor Air Pollution and Respiratory Health Effects in Inner City Children with
Moderate to Severe Asthma.” Air Quality, Atmosphere & Health 13, 247–257.
https://doi.org/10.1007/s11869-019-00789-3
Schauer, J. J., Fraser, M. P., Cass, G. R., & Simoneit, B. R. T. 2002. “Source Reconciliation of
Atmospheric Gas-Phase and Particle-Phase Pollutants during a Severe Photochemical
Smog Episode.” Environmental Science & Technology 36(17), 3806–3814.
https://doi.org/10.1021/es011458j
110
Schauer, J. J., Rogge, W. F., Hildemann, L. M., Mazurek, M. A., Cass, G. R., & Simoneit, B. R.
T. 1996. “Source Apportionment of Airborne Particulate Matter Using Organic
Compounds as Tracers.” Atmospheric Environment 30(22), 3837–3855.
https://doi.org/10.1016/1352-2310(96)00085-4
Schembari, A., De Hoogh, K., Pedersen, M., Dadvand, P., Martinez, D., Hoek, G., …
Nieuwenhuijsen, M. J. 2015. “Ambient Air Pollution and Newborn Size and Adiposity at
Birth: Differences by Maternal Ethnicity (the Born in Bradford Study Cohort).”
Environmental Health Perspectives 123(11), 1208–1215.
https://doi.org/10.1289/ehp.1408675
Schembari, A., Triguero-Mas, M., de Nazelle, A., Dadvand, P., Vrijheid, M., Cirach, M., …
Nieuwenhuijsen, M. J. 2013. “Personal, Indoor and Outdoor Air Pollution Levels among
Pregnant Women.” Atmospheric Environment 64(2), 287–295.
https://doi.org/10.1016/j.atmosenv.2012.09.053
Seinfeld, J. H., & Pandis, S. N. 2006. Atmospheric Chemistry and Physics. New York: John
Wiley & Sons Inc.
Shang, J., Khuzestani, R. B., Tian, J., Schauer, J. J., Hua, J., Zhang, Y., … Zhang, Y. 2019.
“Chemical Characterization and Source Apportionment of PM 2.5 Personal Exposure of
Two Cohorts Living in Urban and Suburban Beijing.” Environmental Pollution 246,
225–236. https://doi.org/10.1016/j.envpol.2018.11.076
Sharp, G., Denney, J. T., & Kimbro, R. T. 2015. “Multiple Contexts of Exposure: Activity
Spaces, Residential Neighborhoods, and Self-Rated Health.” Social Science & Medicine
146, 204–213. https://doi.org/10.1016/j.socscimed.2015.10.040
Sherman, J. E., Spencer, J., Preisser, J. S., Gesler, W. M., & Arcury, T. A. 2005. “A Suite of
Methods for Representing Activity Space in a Healthcare Accessibility Study.”
International Journal of Health Geographics 4, 1–21. https://doi.org/10.1186/1476-
072X-4-24
Solomon, P. A., Crumpler, D., Flanagan, J. B., Jayanty, R. K. M., Rickman, E. E., & McDade, C.
E. 2014. “U.S. National PM2.5 Chemical Speciation Monitoring Networks – CSN and
IMPROVE: Description of Networks.” Journal of the Air and Waste Management
Association 64(12), 1410–1438. https://doi.org/10.1080/10962247.2014.956904
Son, J. Y., Sabath, M. B., Lane, K. J., Miranda, M. L., Dominici, F., Di, Q., … Bell, M. L. 2021.
“Long-Term Exposure to PM2.5 and Mortality for the Older Population: Effect
Modification by Residential Greenness.” Epidemiology 32(4), 477–486.
https://doi.org/10.1097/EDE.0000000000001348
Song, X. H., Polissar, A. V., & Hopke, P. K. 2001. “Sources of Fine Particle Composition in the
Northeastern US.” Atmospheric Environment 35(31), 5277–5286.
https://doi.org/10.1016/S1352-2310(01)00338-7
111
Sowlat, M. H., Hasheminassab, S., & Sioutas, C. 2016. “Source Apportionment of Ambient
Particle Number Concentrations in Central Los Angeles Using Positive Matrix
Factorization (PMF).” Atmospheric Chemistry and Physics 16(8), 4849–4866.
https://doi.org/10.5194/acp-16-4849-2016
Stanek, L. W., Sacks, J. D., Dutton, S. J., & Dubois, J. J. B. 2011. “Attributing Health Effects to
Apportioned Components and Sources of Particulate Matter: An Evaluation of Collective
Results.” Atmospheric Environment 45(32), 5655–5663.
https://doi.org/10.1016/j.atmosenv.2011.07.023
Steinle, S., Reis, S., Sabel, C. E., Semple, S., Twigg, M. M., Braban, C. F., … Wu, H. 2015.
“Personal Exposure Monitoring of PM2.5 in Indoor and Outdoor Microenvironments.”
Science of the Total Environment 508, 383–394.
https://doi.org/10.1016/j.scitotenv.2014.12.003
Stieb, D. M., Chen, L., Beckerman, B. S., Jerrett, M., Crouse, D. L., Omariba, D. W. R., …
Dugandzic, R. M. 2016. “Associations of Pregnancy Outcomes and PM2.5 in A National
Canadian Study.” Environmental Health Perspectives 124(2), 243–249.
https://doi.org/10.1289/ehp.1408995
Stieb, D. M., Chen, L., Eshoul, M., & Judek, S. 2012. “Ambient Air Pollution, Birth Weight and
Preterm Birth: A Systematic Review and Meta-Analysis.” Environmental Research 117,
100–111. https://doi.org/10.1016/j.envres.2012.05.007
Sun, X., Luo, X., Zhao, C., Zhang, B., Tao, J., Yang, Z., … Liu, T. 2016. “The Associations
between Birth Weight and Exposure to Fine Particulate Matter (PM2.5) and Its Chemical
Constituents during Pregnancy: A Meta-Analysis.” Environmental Pollution 211, 38–47.
https://doi.org/10.1016/j.envpol.2015.12.022
Tamura, K., Elbel, B., Chaix, B., Regan, S. D., Al-Ajlouni, Y. A., Athens, J. K., … Duncan, D.
T. 2017. “Residential and GPS-Defined Activity Space Neighborhood Noise Complaints,
Body Mass Index and Blood Pressure Among Low-Income Housing Residents in New
York City.” Journal of Community Health 42(5), 974–982.
https://doi.org/10.1007/s10900-017-0344-5
Tian, N., Xue, J., & Barzyk, T. M. 2013. “Evaluating Socioeconomic and Racial Differences in
Traffic-Related Metrics in the United States Using a GIS Approach.” Journal of
Exposure Science & Environmental Epidemiology 23(2), 215–222.
https://doi.org/10.1038/jes.2012.83
Tonne, C. C., Whyatt, R. M., Camann, D. E., Perera, F. P., & Kinney, P. L. 2004. “Predictors of
Personal Polycyclic Aromatic Hydrocarbon Exposures among Pregnant Minority Women
in New York City.” Environmental Health Perspectives 112(6), 754–759.
https://doi.org/10.1289/ehp.5955
Turpin, B. J., Weisel, C. P., Morandi, M., Colome, S., Stock, T., Eisenreich, S., & Buckley, B.
2007. “Relationships of Indoor, Outdoor, and Personal Air (RIOPA): Part II. Analyses of
Concentrations of Particulate Matter Species.” Health Effects Institute 130 (Pt 2), 1–77.
112
Twum, C., Zhu, J., & Wei, Y. 2016. “Maternal Exposure to Ambient PM2.5 and Term Low
Birthweight in the State of Georgia.” International Journal of Environmental Health
Research 26(1), 92–100. https://doi.org/10.1080/09603123.2015.1061110
USEPA. 2019a. “Air Pollutants Exposure Model Documentation (APEX, Version 5.2) Volume I:
User’s Guide.” US Environmental Protection Agency EPA-452/R-19-005a.
https://www.epa.gov/fera/apex-user-guides
USEPA. 2019b. “Air Pollutants Exposure Model Documentation (APEX, Version 5.2) Volume
II: Technical Support Document.” US Environmental Protection Agency EPA-452/R-19-
005b. https://www.epa.gov/fera/apex-user-guides
USEPA. 2020. “An Introduction to APEX.” US Environmental Protection Agency.
https://www.epa.gov/sites/default/files/2019-
11/documents/apex5_introduction_document.pdf
van Dijk, J. 2018. “Identifying Activity-Travel Points from GPS-Data with Multiple Moving
Windows.” Computers, Environment and Urban Systems 70, 84–101.
https://doi.org/10.1016/j.compenvurbsys.2018.02.004
Wallace, L. 1996. “Indoor Particles: A Review.” Journal of the Air and Waste Management
Association 46(2), 98–126. https://doi.org/10.1080/10473289.1996.10467451
Wallace, L. A., Emmerich, S. J., & Howard-Reed, C. 2002. “Continuous Measurements of Air
Change Rates in An Occupied House for 1 Year: The Effect of Temperature, Wind, Fans,
and Windows.” Journal of Exposure Analysis & Environmental Epidemiology 12(4),
296–306. https://doi.org/10.1038/sj.jea.7500229
Wallace, L., & Williams, R. 2005. “Use of Personal-Indoor-Outdoor Sulfur Concentrations to
Estimate the Infiltration Factor and Outdoor Exposure Factor for Individual Homes and
Persons.” Environmental Science & Technology 39(6), 1707–1714.
https://doi.org/10.1021/es049547u
Wang, J., Kwan, M. P., & Chai, Y. 2018. “An Innovative Context-Based Crystal-Growth
Activity Space Method for Environmental Exposure Assessment: A Study Using GIS and
GPS Trajectory Data Collected in Chicago.” International Journal of Environmental
Research and Public Health 15(4), 1–24. https://doi.org/10.3390/ijerph15040703
Wang, Y., & Hopke, P. K. 2013. “A Ten-Year Source Apportionment Study of Ambient Fine
Particulate Matter in San Jose, California.” Atmospheric Pollution Research 4(4), 398–
404. https://doi.org/10.5094/APR.2013.045
Watson, J. G., Chen, L. W. A., Chow, J. C., Doraiswamy, P., & Lowenthal, D. H. 2008. “Source
Apportionment: Findings from the U.S. Supersites Program.” Journal of the Air and
Waste Management Association 58(2), 265–288. https://doi.org/10.3155/1047-
3289.58.2.265
113
Weaver, A. M., McGuinn, L., Neas, L., Mirowsky, J., Devlin, R. B., Dhingra, R., … Diaz-
Sanchez, D. 2019. “Neighborhood Sociodemographic Effects on the Associations
between Long-Term PM2.5 Exposure and Cardiovascular Outcomes and Diabetes
Mellitus.” Environmental Epidemiology 3(1), e038.
https://doi.org/10.1097/ee9.0000000000000038
Weisel, C. P., Zhang, J., Turpin, B. J., Morandi, M. T., Colome, S., Stock, T. H., … Shendell, D.
2005. “Relationship of Indoor, Outdoor and Personal Air (RIOPA) Study: Study Design,
Methods and Quality Assurance/Control Results.” Journal of Exposure Analysis &
Environmental Epidemiology 15(2), 123–137. https://doi.org/10.1038/sj.jea.7500379
Wheeler, A. J., Wallace, L. A., Kearney, J., Van Ryswyk, K., You, H., Kulka, R., … Xu, X.
2011. “Personal, Indoor, and Outdoor Concentrations of Fine and Ultrafine Particles
Using Continuous Monitors in Multiple Residences.” Aerosol Science and Technology
45(9), 1078–1089. https://doi.org/10.1080/02786826.2011.580798
Wilson, K. M., Klein, J. D., Blumkin, A. K., Gottlieb, M., Winickoff, P., Wilson, A. K. M., &
Jonathan, D. 2011. “Tobacco-Smoke Exposure in Children Who Live in Multiunit
Housing.” Pediatrics 127(1), 85–92. https://doi.org/10.1542/peds.2010-2046
Xiao, X., Zheng, Y., Luo, Q., & Xie, X. 2014. “Inferring Social Ties between Users with Human
Location History.” Journal of Ambient Intelligence and Humanized Computing 5(1), 3–
19. https://doi.org/10.1007/s12652-012-0117-z
Yamamoto, N., Shendell, D. G., Winer, A. M., & Zhang, J. 2010. “Residential Air Exchange
Rates in Three Major US Metropolitan Areas: Results from the Relationship among
Indoor, Outdoor, and Personal Air Study 1999-2001.” Indoor Air 20(1), 85–90.
https://doi.org/10.1111/j.1600-0668.2009.00622.x
Yan, B., Kennedy, D., Miller, R. L., Cowin, J. P., Jung, K., Perzanowski, M., … Chillrud, S. N.
2011. “Validating A Nondestructive Optical Method for Apportioning Colored
Particulate Matter into Black Carbon and Additional Components.” Atmospheric
Environment 45(39), 7478–7486. https://doi.org/10.1016/j.atmosenv.2011.01.044
Yitshak-Sade, M., James, P., Kloog, I., Hart, J. E., Schwartz, J. D., Laden, F., … Zanobetti, A.
2019. “Neighborhood Greenness Attenuates the Adverse Effect of PM2.5 on
Cardiovascular Mortality in Neighborhoods of Lower Socioeconomic Status.”
International Journal of Environmental Research and Public Health 16(5), 1–10.
https://doi.org/10.3390/ijerph16050814
Zamora, M. L., Pulczinski, J. C., Johnson, N., Garcia-Hernandez, R., Rule, A., Carrillo, G., …
Koehler, K. 2018. “Maternal Exposure to PM2.5 in South Texas, A Pilot Study.” Science
of the Total Environment 628, 1497–1507.
https://doi.org/10.1016/j.scitotenv.2018.02.138
Zeger, S. L., Thomas, D., Dominici, F., Samet, J. M., Schwartz, J., Dockery, D., & Cohen, A.
2000. “Exposure Measurement Error in Time-Series Studies of Air Pollution: Concepts
114
and Consequences.” Environmental Health Perspectives 108(5), 419–426.
https://doi.org/10.1289/ehp.00108419
Zenk, S. N., Schulz, A. J., Matthews, S. A., Odoms-Young, A., Wilbur, J. E., Wegrzyn, L., …
Stokes, C. 2011. “Activity Space Environment and Dietary and Physical Activity
Behaviors: A Pilot Study.” Health & Place 17(5), 1150–1161.
https://doi.org/10.1016/j.healthplace.2011.05.001
Zhai, X., Mulholland, J. A., Russell, A. G., & Holmes, H. A. 2017. “Spatial and Temporal
Source Apportionment of PM 2.5 in Georgia, 2002 to 2013.” Atmospheric Environment
161, 112–121. https://doi.org/10.1016/j.atmosenv.2017.04.039
Zhang, Y., Schauer, J. J., Shafer, M. M., Hannigan, M. P., & Dutton, S. J. 2008. “Source
Apportionment of in Vitro Reactive Oxygen Species Bioassay Activity from
Atmospheric Particulate Matter.” Environmental Science & Technology 42(19), 7502–
7509. https://doi.org/10.1021/es800126y
Zhu, X., Liu, Y., Chen, Y., Yao, C., Che, Z., & Cao, J. 2015. “Maternal Exposure to Fine
Particulate Matter (PM2.5) and Pregnancy Outcomes: A Meta-Analysis.” Environmental
Science and Pollution Research 22(5), 3383–3396. https://doi.org/10.1007/s11356-014-
3458-7
115
Appendix A
Table S2.1. Summary of personal monitoring data collection time for MADRES participants.
116
Figure S2.1. Los Angeles, CA, primary, secondary, and local neighborhood roads and city streets
used in the analysis
117
Table S2.2. Personal sampler wearing compliance (N=213).
N (%)
Wear compliance while awake during the daytime
Missing 1 (0.5%)
none 10 (4.7%)
medium 10 (4.7%)
high 192 (90.1%)
Wear compliance while sleeping during the nighttime
Missing 1 (0.5%)
none 16 (7.5%)
medium 13 (6.1%)
high 183 (85.9%)
Place nearby when not worn it during the daytime
Missing 1 (0.5%)
none 7 (3.3%)
medium 5 (2.3%)
high 200 (93.9%)
Table 2.3. The distribution of personal and outdoor (residential and
selected activity spaces) PM2.5 mass concentrations (µg/m
3
).
N Mean (SD) Min Median Max
Personal PM 2.5 213 23.3 (18.9) 1.8 18.4 140.2
Outdoor PM 2.5
At residential location 209 11.8 (5.5) 2.9 10.9 35.1
Within MCH area 199 11.3 (5.5) 0.5 10.7 33.8
Within DPA 199 11.3 (5.5) 0.5 10.7 33.6
Within KDE area (K10/100m) 199 11.4 (5.5) 0.5 10.6 33.7
Within (K25/250m) 199 11.4 (5.5) 0.5 10.7 33.7
Within KDE area (K50/500m) 199 11.4 (5.5) 0.5 10.7 33.6
*MCH: minimum convex hull, DPA: daily path area, KDE: kernel density estimation
118
Table S2.4. Bivariate relationships between personal PM2.5 exposures and questionnaire variables.
119
120
Table S2.5. Bivariate association between primary road lengths within activity spaces and
residential neighborhoods with personal PM2.5 exposure (N=213).
Length of Primary Roads (m) by Method Mean (SD)
Spearman
Correlation
p-value
Activity Space
Daily Path Area (DPA) 80,163.8 (121,844.5) 0.02 0.817
Minimum Convex Hull (MCH) 108,645.0 (239,961.7) -0.04 0.605
Kernel Density Estimation (KDE) measures (bin size, neighborhood size)
10 m, 100 m 6.2 (17.5) 0.02 0.802
10 m, 100 m, top 20
th
percentile 12,736.9 (29,165.6) -0.01 0.885
10 m, 250 m 12.7 (36.0) 0.12 0.085
10 m, 250 m, 20
th
percentile 18,026.6 (40,688.9) 0.01 0.836
25 m, 250 m 79.7 (226.5) 0.13 0.073
25 m, 250 m, 20
th
percentile 17,993.2 (40,590.0) 0.02 0.775
25 m, 500 m 106.4 (216.7) 0.11 0.112
25 m, 500 m, 20
th
percentile 20,884.9 (44,589.5) 0.08 0.291
50 m, 500 m 425.7 (865.9) 0.12 0.094
50 m, 500 m, 20
th
percentile 20,847.9 (44,508.5) 0.07 0.331
Residential Neighborhood
Residential census tract (RN_ct) 1,309.6 (2,508.8) -0.04 0.516
100 m buffer around residence (RN_100 m) 4.5 (47.9) 0.05 0.477
250 m buffer around residence (RN_250 m) 208.3 (644.2) 0.10 0.140
500 m buffer around residence (RN_500 m) 1,266.3 (2,168.4) 0.07 0.340
121
Table S2.6. Spearman correlations of total primary road lengths within activity spaces and
residential neighborhoods, colored from low (blue) to high (red) (N=213).
122
Table S2.7. Associations between personal PM2.5 and NDVI within activity spaces and
residential neighborhood.
Variables Spearman Correlation p-value
Residential Neighborhood Residence, 100 m buffer -0.05 0.429
Residence, 250 m buffer 0.01 0.911
Residence, 500 m buffer 0.04 0.605
Residence, census tract 0.01 0.914
Activity Space -DPA Daily Path Area -0.1 0.172
Activity Space -MCH Minimum Convex Hull -0.03 0.710
Activity Space - KDE (20p) KDE, 10m, 100m, 20p -0.12 0.084
KDE, 10m, 250m, 20p -0.12 0.104
KDE, 25m, 250m, 20p -0.11 0.112
KDE, 25m, 500m, 20p -0.08 0.250
KDE, 50m, 500m, 20p -0.05 0.464
Activity Space - KDE KDE, 10m, 100m -0.03 0.644
KDE, 10m, 250m -0.01 0.924
KDE, 25m, 250m -0.15 0.037
KDE, 25m, 500m -0.04 0.550
KDE, 50m, 500m -0.02 0.802
Values presented in bold font show significant p-values at p<0.05 level
123
Table S2.8. Spearman correlations of average NDVI within activity spaces and residential neighborhoods,
colored from low (blue) to high (red).
124
Table S2.9. Associations between personal PM2.5 and park area (mean and sum) within activity
spaces and residential neighborhoods.
Variables (mean area)
Spearman
Correlation
p-value Variables (sum area)
Spearman
Correlation
p-value
Residence, 100 m buffer 0.03 0.71 Residence, 100 m buffer 0.03 0.71
Residence, 250 m buffer 0.1 0.16 Residence, 250 m buffer 0.1 0.15
Residence, 500 m buffer 0.08 0.22 Residence, 500 m buffer 0.08 0.26
Residence, census tract 0.01 0.93 Residence, census tract 0.01 0.94
Daily Path Area -0.06 0.39 Daily Path Area -0.06 0.42
Minimum Convex Hull -0.06 0.39 Minimum Convex Hull -0.05 0.47
KDE, 10m, 100m, 20p 0.1 0.17 KDE, 10m, 100m, 20p 0.03 0.70
KDE, 10m, 250m, 20p 0.07 0.36 KDE, 10m, 250m, 20p 0.004 0.95
KDE, 25m, 250m, 20p 0.07 0.36 KDE, 25m, 250m, 20p 0.002 0.98
KDE, 25m, 500m, 20p 0.0003 1.00 KDE, 25m, 500m, 20p -0.03 0.66
KDE, 50m, 500m, 20p 0.002 0.98 KDE, 50m, 500m, 20p -0.03 0.66
KDE, 10m, 100m 0.08 0.28 KDE, 10m, 100m 0.05 0.50
KDE, 10m, 250m 0.1 0.18 KDE, 10m, 250m 0.08 0.25
KDE, 25m, 250m 0.09 0.19 KDE, 25m, 250m 0.08 0.25
KDE, 25m, 500m 0.06 0.41 KDE, 25m, 500m 0.08 0.28
KDE, 50m, 500m 0.06 0.41 KDE, 50m, 500m 0.08 0.28
Table S2.10. Summary of scaled parameter estimates for continuous variables in the final model.
Effect
Scaled
Estimate
p-value
Model
Estimate
Std Scale
Birth order of index child at time of pregnancy
5.81 <.0001
4.689
1.24
Length of primary road within KDE area (K50/500m)
2.82 0.018
0.003
865.86
Average NDVI value within KDE area (K25/250m)
-3.09 0.01
-0.239
12.92
Outdoor PM2.5 concentration at residence 2.05 0.092
0.372
5.50
Mean length of secondary road within DPA 5.57 0.001
0.009
613.54
Mean park area within DPA -3.62 0.009
-0.0001
59,052.13
Mean length of minor streets within RN_500 m -2.53 0.04
-0.025
102.65
Table S2.11. Duration of time spent in each microenvironment (%) (N=199).
Min Max Median Mean Std
Indoor 0 100 96.3 94.2 8.5
At home 0 100 81.0 78.9 18.8
Not at home 0 77.2 11.2 15.2 15.6
Outdoor 0 28.4 3.7 5.3 5.1
125
Appendix B
Table S3.1. Home characteristics, indoor source, and time-activities derived
from questionnaires and exit survey (N=212).
Variables n (%) Variables n (%)
Home Characteristics **How open were your windows or doors in general?
*Which best describes the home in which you
currently live most of the time?
A little to half way 86 (40.6%)
House 75 (35.4%) Most to all the way 92 (43.4%)
Apartment 118 (55.7%) Missing 34 (16.0%)
Missing 19 (9.0%)
**How much of the time was a portable or ceiling fan
used in your home, when you were there with the sampler?
*How many people counting yourself live in your
household?
None of the time 129 (60.8%)
1 and 2 people 26 (12.3%) A little, most, or all of the time 78 (36.8%)
3 people 29 (13.7%) Missing 5 (2.4%)
4 people 40 (18.9%) Indoor Air Pollution Source
5 people 20 (9.4%)
**How much of the time were you close to smoke from
candles or incense burning nearby?
More than 5 people 34 (15.9%) None of the time 158 (74.5%)
Missing 63 (29.7%) A little, most, or all of the time 51 (24.1%)
*About when was this home building originally
built?
Missing 3 (1.4%)
Built after 1980s 90 (42.5%)
**How much of the time were you close to smoke or
fume from cooking?
Built before 1980s 68 (32.1%) None of the time 129 (60.8%)
Missing 54 (25.5%) A little, most, or all of the time 80 (37.7%)
*Is there carpeting in your home?
Missing 3 (1.4%)
No 103 (48.6%)
**How much of the time were you close to cigarette,
cigar, hookah or pipe smoke from people smoking nearby?
Yes 91 (42.9%) None of the time 125 (59.0%)
Missing 18 (8.5%) A little, most, or all of the time 83 (39.2%)
*Do you have pets at home?
Missing 4 (1.9%)
No 134 (63.2%) Time-Activities
Yes 74 (34.9%)
**How much of the time did you spend outdoors (not
commuting in a car, bus or train)?
Missing 4 (1.9%) None or a little of the time 133 (62.7%)
*Does your home have heating?
Most or all of the time 76 (35.8%)
No 73 (34.4%) Missing 3 (1.4%)
Yes 120 (56.6%) **When outdoor, whether were you near traffic?
Missing 19 (9.0%) No 81 (38.2%)
Home Ventilation
Yes 128 (60.4%)
** How long the window open in your home, when
you were there with sampler?
Missing 3 (1.4%)
None or little of the time 82 (38.7%) **How many hours did you spend on commute?
Most or all of the time 127 (59.9%) 0 to 30 min 17 (8.0%)
Missing 3 (1.4%) 30 min to 1 hr 44 (20.8%)
126
**How much of the time was the air conditioner
used in your home, when you were there with the
sampler?
1 to 2 hrs 47 (22.2%)
None of the time 154 (72.6%) > 2 hrs 72 (34.0%)
A little, most, or all of the time 55 (25.9%) Missing 32 (15.1%)
Missing 3 (1.4%)
* question from 3
rd
trimester questionnaire; ** question from exit survey
Table S3.2. Bootstrapping Results for base solution, final rotated Fpeak solution and model
variability/error diagnostics.
Legend
Factor 1 Traffic
Factor 2 Secondhand smoking
Factor 3 Aged sea salt
Factor 4 Fresh sea salt
Factor 5 Fuel oil
Factor 6 Crustal
Mapping of bootstrap factors to base factors (BS mapping, 100 bootstraps, 0.6 minimum correlation)
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Unmapped
98 0 2 0 0 0 0
7 69 15 6 0 1 2
0 0 98 2 0 0 0
0 0 0 100 0 0 0
1 0 1 0 98 0 0
0 0 0 0 0 100 0
Mapping of Fpeak (rotated) bootstrap factors to base factors
Base
Factor 1
Base
Factor 2
Base
Factor 3
Base
Factor 4
Base
Factor 5
Base
Factor 6 Unmapped
Boot Factor 1
100 0 0 0 0 0 0
Boot Factor 2
3 94 3 0 0 0 0
Boot Factor 3
0 0 100 0 0 0 0
Boot Factor 4
0 0 0 100 0 0 0
Boot Factor 5
0 0 0 0 100 0 0
Boot Factor 6
0 0 0 0 0 100 0
DISP Diagnostics
Error Code: 0
Largest Decrease in Q: 0
%dQ: 0
Swaps by Factor: 0 0 0 0 0 0 0
BS-DISP Diagnostics
BS-DISP Displaced Species: BrC
# of Cases Accepted: 98
% of Cases Accepted: 98%
127
Largest Decrease in Q: -9.13
%dQ: -0.16
# of Decreases in Q: 0
# of Swaps in Best Fit: 1
# of Swaps in DISP: 1
Swaps by Factor: 1 0 0 1 0 0
Table S3.3: PMF model results showing R
2
and normality of residuals for each species.
Species R
2
Normal residuals?
PM mass 0.48 Yes
Carbon Species
BC 0.16 No
BrC 0.53 Yes
ETS 0.12 No
Elements
Al 0.5 No
Ba 0.41 Yes
Br 0.24 Yes
Ca 0.53 No
Cl 0.85 No
Co 0.33 No
Cu 0.77 Yes
Fe 0.78 Yes
K 0.13 No
Mg 0.84 Yes
Mn 0.54 Yes
Na 0.86 Yes
Ni 0.35 Yes
P 0.0001 No
Pb 0.13 No
S 0.83 No
Se 0.09 Yes
Si 0.62 Yes
Sr 0.04 No
Ti 0.7 Yes
V 0.04 No
Zn 0.3 No
128
Appendix C
Figure S4.1. MADRES study area used for APEX model runs
129
Table S4.1. APEX microenvironment parameters.
Microenvironment Parameters Conditions Distributions
Indoor-Residence AER Temp < 68; AC; room, window-open LogN (1.344, 1.863, 0.1, 10)
Temp 68-76; AC; room, window-open LogN (3.348, 2.223, 0.1, 10)
Temp > 76; AC; room, window-open LogN (1.898, 1.644, 0.1, 10)
Temp < 50; AC; none, window-open LogN (1.086, 3.087, 0.1, 10)
Temp 50-67; AC; none, window-open LogN (1.494, 2.085, 0.1, 10)
Temp 68-76; AC; none, window-open LogN (2.744, 2.283, 0.1, 10)
Temp > 76; AC; none, window-open LogN (1.976, 1.967, 0.1, 10)
Temp < 68; AC; room, window-close LogN (0.672, 1.863, 0.1, 10)
Temp 68-76; AC; room, window-close LogN (1.674, 2.223, 0.1, 10)
Temp > 76; AC; room, window-close LogN (0.949, 1.644, 0.1, 10)
Temp < 50; AC; none, window-close LogN (0.543, 3.087, 0.1, 10)
Temp 50-67; AC; none, window-close LogN (0.747, 2.085, 0.1, 10)
Temp 68-76; AC; none, window-close LogN (1.372, 2.283, 0.1, 10)
Temp > 76; AC; none, window-close LogN (0.988, 1.967, 0.1, 10)
Indoor-Other AER All LogN (1.109, 3.015, 0.07, 13.8)
Indoor-Residence ES gas stove LogN (1700, 10, 100, 2000)
Indoor-Residence ES gas stove (duration) Uniform (0.5, 1)
Indoor-Residence Vol
Normal (120, 30, 50, 300)
Indoor-Residence ES candle burning Normal (110, 60, 10, 200)
Indoor-Residence ES candle burning (duration) Uniform (0.6, 1)
Indoors-All Decay rate All Uniform (0.1, 1.1)
All MEs Penetration
1
All MEs Proximity All Normal (1.0, 0.07, 0.9, 1.1)
Table S4.2. Duration of time spent in each microenvironment (as a percentage per day).
Microenvironment Mean (SD)
MADRES (N=213) S1 (N=500) S2 (N=500) S3 (N=500) S4 (N=500) S5 (N=500)
In-Residence 78.46 (19.59) 72.06 (7.90) 72.06 (7.90) 72.50 (8.33) 72.50 (8.33) 71.79 (8.21)
In-Other 15.22 (15.66) 18.24 (7.39) 18.24 (7.39) 17.97 (7.58) 17.97 (7.58) 18.73 (7.51)
Outdoor 3.86 (3.63) 3.86 (3.63) 3.68 (3.38) 3.68 (3.38) 3.61 (3.68)
Near-Road
0.24 (0.49) 0.24 (0.49) 0.25 (0.57) 0.25 (0.57) 0.25 (0.52)
Vehicle 5.56 (1.03) 5.56 (1.03) 5.57 (1.09) 5.57 (1.09) 5.58 (1.04)
130
Table S4.3. Estimated mean microenvironment PM2.5 concentrations (µg/m
3
).
Microenvironments PM2.5 Concentrations (µg/m
3
), Mean (SD)
S1 S2 S3 S4 S5
In-Residence 8.8 (1.6) 9.8 (1.5) 10.5 (1.7) 9.9 (1.6) 10.4 (1.7)
In-Other 9.7 (1.7) 9.7 (1.7) 9.5 (1.6) 9.4 (1.5) 9.5 (1.7)
Outdoor 14.2 (0.6) 14.2 (0.6) 14.2 (0.7) 14.2 (0.7) 14.3 (0.7)
Near-Road 14.1 (2.2) 14.1 (2.2) 14.1 (2.4) 14.1 (2.4) 14.1 (2.3)
Vehicle 14.2 (0.4) 14.2 (0.4) 14.2 (0.4) 14.2 (0.4) 14.2 (0.4)
Table S4.4. Estimated total personal and microenvironment PM2.5 exposures
extracted from hourly outputs (µg/m
3
).
Microenvironments PM2.5 Exposures, Mean (SD)
S1
S2 S3 S4 S5
Estimated Personal Exposure
9.5 (7.26)
10.2 (7.63) 10.65 (7.73) 10.19 (7.69) 10.6 (7.69)
In-Residence
6.34 (6.73) 7.05 (7.39) 7.58 (7.63)
7.14 (7.46) 7.47 (7.59)
In-Other
1.77 (4.7) 1.77 (4.7) 1.72 (4.61)
1.69 (4.55) 1.79 (4.72)
Outdoor
0.55 (2.86) 0.55 (2.86) 0.52 (2.86)
0.52 (2.85) 0.51 (2.78)
Near-Road
0.03 (0.75) 0.03 (0.75) 0.03 (0.72) 0.03 (0.72) 0.04 (0.74)
Vehicle 0.79 (2.9) 0.79 (2.9) 0.79 (2.96) 0.79 (2.96) 0.79 (2.97)
Table S4.5. Spearman correlations among estimated hourly personal PM2.5 exposures,
microenvironment exposures, and ambient PM2.5 concentrations in S1.
In-Residence
In-Other -0.65
Outdoor -0.24 0.09
Near-Road -0.07 -0.0001 0.03
Vehicle -0.28 0.28 0.31 0.07
Ambient PM 2.5 0.44 0.05 0.00 -0.001 0.02
Personal
Exposures
0.44 0.12 0.13 0.03 0.15 0.81
In-Residence In-Other Outdoor Near-Road Vehicle Ambient PM 2.5
Values in bold font represent significant p-values at p<0.05 level.
131
Table S4.6. Spearman correlations among estimated hourly personal PM2.5 exposures,
microenvironment exposures, and ambient PM2.5 concentrations in S2.
In-Residence
In-Other -0.65
Outdoor -0.24 0.09
Near-Road -0.07 -0.0001 0.03
Vehicle -0.29 0.28 0.31 0.07
Ambient PM 2.5 0.46 0.05 0.003 -0.001 0.02
Personal
Exposures
0.49 0.06 0.10 0.03 0.12 0.84
In-Residence In-Other Outdoor Near-Road Vehicle Ambient PM 2.5
Values in bold font represent significant p-values at p<0.05 level.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Personal exposure to particulate matter PM2.5 sources during pregnancy and birthweight
PDF
Prenatal environmental exposures and fetal growth in the MADRES cohort
PDF
Sensor-based mobile health approaches for personal air pollution and pediatric asthma studies
PDF
Urban air pollution and environmental justice: three essays
PDF
Investigating daily effects of activity space-based built environment exposures on physical activity behaviors in Hispanic women during pregnancy and early postpartum
PDF
Indoor air quality for human health in residential buildings
PDF
Associations of cumulative pollution burden and environmental health vulnerabilities with gestational weight gain in a cohort of predominantly low-income Hispanic women
PDF
Effectiveness of individual- and household-level protective actions in reducing symptoms associated with hydrogen sulfide chronic low-level exposure
PDF
Susceptibility to autism likelihood associated with prenatal ambient air pollution exposures: identifying vulnerable populations
PDF
Associations of ambient air pollution exposures with perceived stress in the MADRES cohort
PDF
Prenatal exposure to organophosphate esters and early neurodevelopment
PDF
Air pollution, mitochondrial function, and growth in children
PDF
Investigating the temporal trends, sources, and toxicity of ambient particulate matter (PM) in different metropolitan environments, and development of a novel aerosol generation setup for inhalat...
PDF
Incomplete data & insufficient methods: transgender population health research in the US
PDF
Impact of urban source emissions on ambient particulate matter (PM) composition and associated toxicity in local and regional scales using source appointment models
PDF
Size-resolved particulate matter (PM) in urban areas: toxico-chemical characteristics, sources, trends and health implications
PDF
Characterizing self-reported spatial and provider barriers to maternal health care utilization in Malawi
PDF
Alzheimer’s disease: dysregulated genes, ethno-racial disparities, and environmental pollution
PDF
Biomarkers of age-related health changes: associations with health outcomes and disparities
PDF
Ultrafine particle concentration inside vehicles: models for exposure assessment
Asset Metadata
Creator
Xu, Yan
(author)
Core Title
Personal PM2.5 exposure during pregnancy in an environmental health disparities population
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Health and Place,Population
Degree Conferral Date
2022-08
Publication Date
05/27/2022
Defense Date
05/09/2022
Publisher
University of Southern California. Libraries
(digital)
Tag
activity spaces,air pollutants exposure (APEX) model,Air pollution,environmental disparities population,GPS,OAI-PMH Harvest,personal exposure,PM2.5,positive matrix factorization (PMF) model,Pregnancy,source apportionment analysis,stochastic microenvironment approach
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Habre, Rima (
committee chair
), Wilson, John Peter (
committee chair
), Bastain, Theresa (
committee member
), Farzan, Shohreh (
committee member
), Pastor, Manuel (
committee member
)
Creator Email
xuyan1904@yahoo.com,yxu@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111336172
Unique identifier
UC111336172
Identifier
etd-XuYan-10732.pdf (filename)
Legacy Identifier
etd-XuYan-10732
Document Type
Dissertation
Rights
Xu, Yan
Internet Media Type
application/pdf
Type
texts
Source
20220527-usctheses-batch-944
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
activity spaces
air pollutants exposure (APEX) model
environmental disparities population
GPS
personal exposure
PM2.5
positive matrix factorization (PMF) model
source apportionment analysis
stochastic microenvironment approach