University of Southern California Dissertations and Theses

Spatial modeling of non-tailpipe emissions and its association with children's lung function

(USC Thesis Other)

Spatial modeling of non-tailpipe emissions and its association with children's lung function

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content SPATIAL MODELING OF NON-TAILPIPE EMISSIONS AND ITS
ASSOCIATION WITH CHILDREN’S LUNG FUCNTION

BY

CHIKA OKONKWO

THESIS

Submitted in partial fulfillment of the requirements
for the degree of Master of Science in Applied Biostatistics and Epidemiology
in the Graduate School of the
University of Southern California,
May 2019

Los Angeles, California

Adviser:

Professor Meredith Franklin

2
Abstract

Although it has been shown that traffic-related tailpipe air pollution adversely affects
children’s lung function especially with traffic noise in the model, few studies have examined
this association using traffic-related non-tailpipe emissions.
Estimates of freeway emission concentrations of oxides of nitrogen (NOx, ppb), reflective
of tailpipe traffic exposure, were spatially assigned to children in Southern California who were
tested for forced vital capacity (FVC, n = 1217) and forced expiratory volume in 1s, (FEV1, n =
1217). Estimates of fine and coarse concentrations of particulate matter mass (PM, µg/m
3
) and
elements (PM metals, ng/m
3
) were obtained on the georeferenced Southern California
participants residential locations. The associations between traffic-related non-tailpipe PM mass
and PM metals- PM2.5 Cu, PM2.5-10 Al, PM2.5 and PM2.5-10 Zn, and the outcomes- FVC and FEV1,
were examined using mixed effects models.
Of the PM mass and metals, both fine and coarse Zn metal showed significant association
with lung function. A 52.6 mL decrease (95% CI – 107.0, 1.82) and a 54.8 decrease (95% CI –
106.0, – 3.6) in FVC was associated with an interquartile range increase in PM2.5 Zn (p ≤ 0.1)
and PM2.5-10 Zn (p ≤ 0.05) respectively, with NOx included in the model.
Our results suggest that non-tailpipe traffic-related exposures were relatively more
important than tailpipe exposures alone in determining lung function association. As a result,
more non-tailpipe PM metals should be included in epidemiological studies of the association
between traffic-related air pollution on lung function.

3
Acknowledgements

This project would not have been possible without the support of many people. I would
like to thank my adviser, Meredith Franklin, who read my numerous revisions and helped make
some sense of the confusion. Also, thanks to my committee members, William Gauderman and
Rima Habre, who also offered extensive guidance and support. Thanks to the University of
Southern California for allowing me the opportunity to complete this project. And finally, thanks
to my parents and numerous friends who endured this process with me by offering support and
love.

4
TABLE OF CONTENTS

Abstract .................................................................................................................................. 2
Acknowledgements ................................................................................................................. 3
1. Introduction ....................................................................................................................... 5
2. Materials and Methods ...................................................................................................... 6
2.1. Study Area ................................................................................................................................ 6
2.2. Environmental Exposures ......................................................................................................... 6
2.2.1. Non-tailpipe air pollution .......................................................................................................................... 6
2.2.2. Tailpipe air pollution ................................................................................................................................. 7
2.3. Health Outcomes ....................................................................................................................... 7
2.4. Statistical Methods .................................................................................................................... 7
2.4.1 Spatial Methods for Exposure Estimation ................................................................................................... 7
2.4.2 Epidemiological Analysis ............................................................................................................................ 9
3. Results ............................................................................................................................... 9
4. Conclusions ..................................................................................................................... 16
References ............................................................................................................................ 18

5
1. Introduction

Air pollution is known worldwide to cause devastating effects on human health and the
environment. Air quality in the urban environment has risen as an important determinant of
human health. Many studies have examined the association between exposure to traffic-related
air pollution and children’s respiratory health (e.g. Gauderman et al., 2007; McConnell et al.,
2010; Urman et al., 2014; Rice et al., 2016). Long-term exposure to nitrogen oxides (NO, NO2,
NOx) has been shown to lead to a reduction in children’s lung development (Gauderman et al.,
2004, 2007). Traffic-related air pollutants from freeway and non-freeway emissions of NOx as
well as airborne particulate matter with aerodynamic diameters less than 2.5 µm and 10 µm
(PM2.5 and PM10, respectively) have all been associated with many chronic health conditions
(Caiazzo et al., 2013).
Particulate matter (PM) air pollution consists of a complex mixture of microscopic solid
or liquid droplets that vary both in size and chemical composition. Fine particle matter (PM2.5)
has been thought of as posing the greatest risk to human health as it can travel deep into the
lungs and may get into the bloodstream. However, it remains unclear whether it is certain
chemical signatures of PM that pose greater health risk than others (Franklin et al., 2008;
Zanobetti et al., 2009). There are both natural and anthropogenic sources of PM air pollution.
Vehicles largely emit ultrafine particles from exhaust pipe which later becomes PM2.5, while
road abrasion, tire and brake wear largely gives rise to bigger wear particles such as coarse
particulate matter (PM2.5-10). Traffic-related PM in both the fine and coarse fractions also
comprise of metals from non-tailpipe emissions including brake wear, tire wear, and resuspended
road dust (Abu-Allaban et al., 2003; WHO, 2013). To date, traffic-related health studies have
primarily focused on exposure estimates of tailpipe emissions of oxides of nitrogen (NOx) based
on extrapolating concentrations from measurements made at central monitoring sites (Franklin et
al., 2012; Li et al., 2017) or applying line source dispersion models (Franklin and Fruin, 2017).
Few studies have specifically examined traffic-related PM2.5 and health (Requia et al., 2018) and
to date there are no comprehensive exposure-health studies of non-tailpipe emissions, including
brake wear, tire wear and resuspended road dust.
In this study, we developed spatial models for non-tailpipe traffic-related elements in the
fine and coarse PM fraction: Copper (metal from brake wear), Aluminum (metal from soil,
crustal and resuspended road dust) and Zinc (metal from tire wear) to estimate residential
exposures present within seven Southern California communities. We then investigated the
associations between these non-tailpipe and tailpipe traffic-related air pollutants on children’s
respiratory health by taking advantage of data from the Southern California Children’s Health
Study (CHS). Specifically, we tested the associations between children’s lung function
measurements of forced expiratory volume in one second (FEV1, mL), forced vital capacity
(FVC, mL) and NOx – reflective of tailpipe traffic exposure, PM mass and its elements (PM2.5
and PM2.5-10) – reflective of non-tailpipe traffic exposure, with adjustment for important
covariates.

6
2. Materials and Methods

2.1. Study Area

The Southern California Children’s Health Study (CHS) enrolled over 11,000 children in
a series of five cohorts starting in the early 1990s. For this study we focus on the final cohort
(Cohort E), which started in 2003 when children were ages 5-7 and ended in 2012 when children
were ages 14-16. During this period, from November 2008 to December 2009, the CHS-related
Intracommunity Variability (ICV) sampling campaign collected month-long integrated PM
samples in eight communities in the greater Los Angeles, California area: Anaheim, Glendora,
Long Beach, Mira Loma, Riverside, Santa Barbara, San Dimas, and Upland (Figure 1). The ICV
campaign included a spatially dense sampling design with the aim of capturing gradients of
traffic emissions. Within each community, 26 to 29 monitors collecting PM in three size
fractions (PM0.2, PM2.5 and PM10-2.5) were placed at selected CHS residential locations, one
community regional air monitoring site, and neighborhood elementary schools attended by CHS
participants. Some communities (e.g. Anaheim) had many sampling locations that were close to
freeways since a portion of the cohort has high freeway emission exposures (Franklin and Fruin,
2017), while other communities had a mix of sampling locations around freeways and smaller
residential roads. For this study, we excluded Santa Barbara in order to simplify spatial
modeling, focusing on seven southern California communities from the CHS.

Figure 1. Map of study area indicating ICV sampling locations in all 8 CHS communities.

2.2. Environmental Exposures

2.2.1. Non-tailpipe air pollution
The 2008-2009 ICV sampling campaign used Harvard Cascade Impactors (Lee et. al.,
2006) with multiple stages of collection to capture quasi-ultrafine (PM0.2), accumulation mode
7
fine (PM0.2-2.5), and coarse (PM2.5-10) particulate matter measured in two seasons (warm and
cool). Mass from each size fraction was analyzed for chemical speciation by inductively coupled
plasma mass spectrometry (ICP-MS). As we are primarily interested in assessing the health
effects associated with non-tailpipe components of traffic, we focus on developing exposure
models for the components of fine and course mass that are signatures of brake wear (Cu), tire
wear (Zn) and resuspended road dust (Al).

2.2.2. Tailpipe air pollution
Annual and seasonal concentrations of NOx were estimated at each subject’s residence
for the calendar year prior to each child’s lung function test using CALINE4 line source
dispersion model (Benson; Franklin and Fruin, 2017). The CALINE4 dispersion model uses
vehicle traffic volume, roadway geometry and emission rate by roadway link, and meteorological
conditions as inputs to estimate traffic-related NOx concentrations at receptor sites, defined as
CHS residential locations. The estimated freeway NOx exposure is seen as an indicator of
incremental increases in air pollution over background ambient levels due to primary emissions
from local vehicular traffic.

2.3. Health Outcomes

In 2010 when CHS Cohort E study participants were on average 13 years old, pulmonary
function tests were taken by trained respiratory staff. Maximal forced expiratory volume in the
first second (FEV1) and forced vital capacity (FVC) were measured using pressure transducer-
based spirometers (Screenstar Spirometers, Morgan Scientific, Haverhill, Massachusetts, USA).
Other information collected by questionnaire at the time of the pulmonary tests included sex,
age, self-identified race, and ethnic background, occurrences of acute respiratory illness,
exercise, tobacco-smoke exposure (personal smoking or environmental). Written questionnaires
completed at study entry (baseline) also obtained information on demographic characteristics,
parental education, and indoor environmental exposures relating to the child such as housing
characteristics (air conditioning, age of house, presence mildew, pets in the home).

2.4. Statistical Methods

2.4.1 Spatial Methods for Exposure Estimation
To estimate exposures at each CHS study participant’s residential location (excluding
those residing in Santa Barbara), geostatistical methods were applied to first characterize the
spatial patterns in the georeferenced ICV PM mass and speciation concentrations, and then to
interpolate concentrations where measurements were not made. We also assessed differences in
spatial patterns in PM mass by season (cool or warm).
Empirical binned semivariograms were constructed for each of the mass and species of
interest, pooling the concentrations across communities. Separate semivariograms were
constructed by season for fine and coarse mass. The binned semivariance γ(h) is estimated by:

where the h is the spatial lag characterized by Euclidean distance between two measurement
locations si and sj (identified by the projected x,y coordinates of the georeferenced ICV monitor),
8
the number of pairs of observations within a particular distance lag is N(h) and Z(si)-Z(sj) is the
difference in PM mass or species measured at locations si and sj. The empirical semivariogram is
a graph of the distance lag (h) on the x-axis and γ(h) on the y-axis, and is used to explore the
spatial pattern and structure between all pairwise locations (Figure 2).

Figure 2. Illustration of the binned empirical semivariogram (blue +) and fitted theoretical
semivariance function (red dotted line) with location of nugget (!
2
)

sill ($
2
) and range (f) (left);
equations for the Gaussian theoretical semivariance and covariance functions (right).

The spatial parameters that describe the theoretical semivariance functions are the range
(f), sill ($
2
), and nugget (!
2
). The range (f) indicates the distance where the semivariance
reaches an asymptote; the sill ($
2
) refers to the value of the y-axis reached at the range; the
nugget (!
2
) represents the semivariance while the smallest distance (i.e. as hà0). An example of
a theoretical semivariance function and its corresponding covariance is shown in Figure 2.
The weighted least squares (WLS) approach was chosen to fit the theoretical semivariance
function to the binned semivariograms, choosing the best fit from Gaussian, exponential, or
Matern. WLS is a robust method to fit these functions to the data and does not require any
distributional assumptions. With the fitted semivariance function, the shape of the spatial process
for PM mass and the elements was examined, and provided spatial parameter estimates that were
then applied in spatial interpolation.

Kriging is a classical method of spatial interpolation using a Gaussian process model that
provides the best unbiased linear estimation at an unobserved point based on weighting data from
observed points from a pre-prescribed spatial covariance function. The general universal kriging
equation is:

∑ = $
2
H(f) + !
2
I
where H(f) is a valid covariance function with range f and distances h between locations si and
sj. In universal kriging, the trend term is Xβ, and is typically a linear or quadratic function of the
spatial coordinates. Predictions of the spatial process, in this case, PM concentrations, are Z
(
(s0)
where s0 is an unobserved location. In this study universal kriging was performed 1) on a grid
producing ten maps of kriged predictions (four from the seasonal fine and coarse PM and six
Gaussian (theoretical) semivariance
and corresponding Covariance

Semivariance:
g(h)=!
2
+ $
2
(1-exp(-h
2
/f
2
))

Corresponding covariance:
C(h)= $
2
exp(-h
2
/f
2
)
9
from the fine and coarse non-tailpipe PM metals copper, aluminum, zinc), and 2) on the
georeferenced CHS participants residential locations producing exposure estimates for
epidemiological assessments. Kriging standard errors were also generated but are not shown.

2.4.2 Epidemiological Analysis
Mixed effect models incorporating a random effect for community were fit for both
outcome variables, FVC and FEV1, since the study participants are clustered in seven relatively
separate Southern California communities. This model allows for variance in regional pollution
by community and has been used in previous studies involving the CHS data (McConnell et al.,
2010; Franklin and Fruin, 2017; Gauderman et al., 2007). The associations between measured
FVC and FEV1, and the traffic-related pollutants of interest— NOx, PM2.5 and PM2.5-10 mass, and
PM2.5 and PM2.5-10 metals, were examined after adjustment for subject sex, height, body mass
index (BMI), race/ethnicity, parental education level, tobacco smoke, and distance to freeway.
The traffic-related non-tailpipe PM metals of interest examined for the association were PM2.5
copper, PM2.5-10 aluminum and both PM2.5 and PM2.5-10 zinc.

3. Results

The study population of the seven CHS communities consisted of boys (48%) and girls
(52%) of mean age 13.4 years with mean spirometric measurements of FVC (3630 mL, SD = 689
mL) and FEV1 (3460 mL, SD = 547 mL). Details of their physical characteristics, lung function
and exposure levels of NOx are shown in Table 1a. Across communities, the mean concentration
of freeway NOx is 15.3 (SD = 16.6) ppb.

Table 1a. Characteristics of the study population.
Number of subjects % or mean (SD)
Subjects
Boys 615 48%
Girls 671 52%
Age (years) 1286 13.4 (0.6)
Height (cm) 1286 159.0 (8.2)
BMI (kg/m
2
) 1286 21.5 (4.7)

Race
Asian 74 6%
African American 26 2%
Caucasian 568 44%
Mixed 179 14%
Other 291 23%
Unknown or missing 148 11%

Ethnicity
Hispanic 663 52%
Non-Hispanic 557 43%
Unknown or missing 66 5%

10
Outcomes
Forced Vital Capacity (mL) 1217 3630 (689)
Forced Expiratory Volume (mL) 1217 3460 (547)

Exposures
Freeway NO x (ppb) 1286 15.3 (16.6)
Distance to freeway (km) 1286 1.4 (1.1)
Exposure to smoke 1274 9%

Communities
Anaheim 140 11%
Glendora 266 21%
Long Beach 96 7%
Mira Loma 209 16%
Riverside 153 12%
San Dimas 211 16%
Upland 211 16%

The community specific distributions of freeway NOx are shown in Figure 3. Of these
communities, Anaheim had the highest average and most variable freeway NOx concentration
while Glendora had the lowest average freeway NOx concentration. Across communities, the
mean or median concentrations of sampled PM and non-tailpipe PM metals are shown in Table
1b. The community specific distributions of PM2.5 copper, PM2.5-10 aluminum, PM2.5 and PM2.5-10
zinc, and PM2.5 and PM2.5-10 mass are also depicted in Figure 3. Of the communities, Long Beach
had the highest average PM2.5 copper and both PM2.5 and PM2.5-10 zinc concentrations; while
Mira Loma had the highest average PM2.5-10 aluminum concentration and both PM2.5 and PM2.5-10
concentrations.

Table 1b. Summary statistics of fine and coarse PM (in µg/m
3
) and PM metals (in ng/m
3
) across
7 CHS communities
Number of
observations
Mean (SD) Median (IQR)
PM 2.5 396 15.2 (4.0) 15.0 (3.3)
PM 2.5 (Cool) 198 14.5 (5.1) 14.2 (4.3)
PM 2.5 (Warm) 198 15.8 (2.2) 15.6 (3.0)
PM 2.5-10 396 13.3 (4.9) 13.0 (4.2)
PM 2.5-10 (Cool) 198 11.8 (5.0) 11.8 (3.9)
PM 2.5-10 (Warm) 198 14.8 (4.4) 14.0 (3.1)
PM 2.5 Copper 396 4.5 (2.6) 3.5 (3.5)
PM 2.5-10 Copper 396 11.9 (6.4) 9.4 (7.7)
PM 2.5 Aluminum 396 56.6 (32.0) 46.1 (30.2)
PM 2.5-10 Aluminum 396 436.0 (340.6) 343.8 (229.1)
PM 2.5 Zinc 396 7.8 (8.1) 4.6 (8.1)
PM 2.5-10 Zinc 396 10.6 (8.3) 7.8 (8.3)

11

Figure 3. Distributions of tailpipe freeway NOx (top left), and key non-tailpipe PM2.5 and PM2.5-
10 mass and metals measured through the ICV sampling campaign in the 7 CHS communities.

Empirical semivariograms enabled proper visualization of the shape of the spatial process
of the PM concentrations across the sampled seven southern California communities and guided
fitting the theoretical function to the observed data. For all PM and PM metals, the Gaussian
semivariogram function resulted in the smallest sum of square error (SSE) compared to
exponential and Matern functions. Spatial parameter estimates and SSE for each WLS fitted
semivariance function are shown in Table 2. We observed spatial patterns for all fine and coarse
12
PM metals as evidenced by the semivariogram plots (Figure 4). The estimated spatial ranges (the
distance at which spatial correlation becomes negligible) of the fitted semivariogram functions
are between 9.9 km and 15.9 km (Table 2) indicating that the spatial variability in these
components is quite localized. Similarly, we also observed spatial autocorrelation at small
distances of seasonal fine and coarse PM (Figure 5), with significantly smaller scale spatial
variability in the warm season than the cool season. While the concentration scales are quite
different between the PM metals, we observe the WLS fit for fine copper was best, having the
lowest sum of square error (SSE = 187.8) Coarse aluminum and zinc had the worst fits (SSE =
1128.7 and 1829.5, respectively) but their relative concentrations were higher, and there is
unusually high semivariance of zinc at small distances (Figure 4). For PM by season, we
observed relatively large sum of square error for all except PM2.5-10 warm (422.7), most likely
resulting from having smaller concentrations in the winter and a good model fit (Figure 5).

Figure 4. Semivariogram of metals measured Figure 5. Semivariogram of PM measured
across 7 CHS communities by season. across 7 CHS communities by season.

Table 2. Spatial parameter estimates using gaussian function of PM by season and PM metals
across 7 CHS communities.
Nugget* Partial
Sill
Range SSE
PM 2.5 Copper 0.9 2.3 9.9 187.8
PM2.5-10 Copper 5.3 25.9 15.9 219.8
PM 2.5 Aluminum 87.3 742.5 11.0 415.8
PM2.5-10 Aluminum 6.6 x10
3
1.0 x10
5
12.9 1128.7
PM 2.5 Zinc 3.6 10.3 10.2 629.3
PM2.5-10 Zinc 3.0 37.2 12.3 1829.5
By season
PM 2.5 (Cool) 1.1 60.9 13.5 1067.9
PM 2.5 (Warm) 1.3 5.9 6.4 1537.2
PM2.5-10 (Cool) 1.2 52.6 12.6 1469.8
PM2.5-10 (Warm) 3.4 11.6 13.6 422.7
* Units of the nugget and partial sill are in squared concentrations (i.e. (ng/m
3
)
2
for metals and

(µg/m
3
)
2
for mass),
and range is in kilometers.
13
From the kriging maps of seasonal PM (Figure 6), we observed higher values of kriging
estimates in cool season for both PM2.5 (max. estimate = 25 µg/m
3
) and PM2.5-10 (max. estimate =
25 µg/m
3
) compared to the warm season, PM2.5 (max. estimate = 20 µg/m
3
) and PM2.5-10 (max.
estimate = 22 µg/m
3
). This result aligns with existing knowledge about air pollution from
primary sources. We observe high levels of PM2.5 during the cool season in Mira Loma,
Riverside, San Dimas and Anaheim; while during warm season, Mira Loma, Upland, and
Anaheim communities have the highest levels of PM2.5. Highest levels of PM2.5-10 during cool
season are also seen in Mira Loma, but also in Long Beach, while during the warm season, Mira
Loma and Riverside have highest concentrations. We also observed that the predicted estimates
at unobserved locations for fine and coarse particulate matter during cool season fall mostly
under higher values (ranging from light orange to dark orange), while the predicted estimates at
unobserved locations for fine and coarse PM during warm season are fall mostly under low
values (yellow to light orange). Please note the small number of negative estimated PM2.5-10
concentrations in the cool season. These are likely an artifact of the kriging model but are not in
a community where there were any CHS subjects, so we did not remove them from the maps.

PM2.5 Cool PM2.5-10 Cool

PM2.5 Warm PM2.5-10 Warm

Figure 6. Kriged surfaces of seasonal PM2.5 and PM2.5-10 over the CHS communities.

Maps of the gridded kriged estimates over the study region for both fine and coarse non-
tailpipe traffic-related particulate matter of copper, aluminum and zinc metals are shown in
Figure 7. We observe that the spatial patterns by PM metal are similar for both PM2.5 and PM2.5-
10. The communities with fairly high levels of copper (from brake wear) are Long Beach,
Anaheim and Mira Loma. The communities with fairly high levels of aluminum (from soil,
14
crustal and resuspended road dust) are Mira Loma and Riverside. Finally, Anaheim, Long Beach
and Mira Loma have the highest observed concentrations of zinc (from tire wear).

PM2.5 Copper PM2.5-10 Copper

PM2.5 Aluminum PM2.5-10 Aluminum

PM2.5 Zinc PM2.5-10 Zinc

Figure 7. Kriged surfaces of PM2.5 and PM2.5-10 metals over the CHS communities.

To narrow down our investigation of the key non-tailpipe exposures of interest, we focus
on a handful of predicted concentrations in our epidemiological analysis: fine copper, coarse
aluminum, and fine and coarse zinc. After assigning each CHS study participant kriged PM mass
and metals and CALINE NOx, we examined the Pearson correlations between these non-tailpipe
and tailpipe exposures (Table 3). Between freeway NOx and distance to freeway, the correlation r
= – 0.54, indicating there is a strong association that as distance from the freeway increases
15
freeway NOx concentrations decrease, as expected. The correlations between freeway NOx and
PM metals are far smaller but indicate some interesting patterns. Coarse aluminum is negatively
correlated with freeway NOx and positively correlated with distance to freeway, indicating that it
may not be driven by traffic as much as it is driven by other sources such as windblown dust.
The correlations of brake and tire wear metals with NOx are positive and all in the range of 0.33
to 0.36, and with distance to freeway they are negative but very small -0.07 to -0.1. This suggests
that the metals are explaining other components of traffic, or that tailpipe and non-tailpipe
emissions are not as highly correlated as commonly expected.
Results from the mixed effects models that included adjustment for age, height, height
squared, BMI, BMI squared, sex, race, ethnicity, parental education, exposure to tobacco smoke,
and a random intercept for community are shown in Table 4. We did not include distance to
freeway in the models due to its moderately high correlation with freeway NOx. Models
including tailpipe (NOx) and non-tailpipe concentrations together as well as separately were
examined. Overall, effect estimates were stronger for FVC than FEV1. Interestingly, NOx was
not significantly associated with either outcome when modeled on its own or jointly with the
non-tailpipe exposures. The non-tailpipe metals also all indicated a decrease in lung function
with an IQR increase in exposure. Furthermore, when NOx was included in the joint exposure
models, the magnitudes of the non-tailpipe effect estimates were stronger. Specifically, we
observed a marginally significant 16.5 mL decrease (95% CI – 33.6, 0.55) and a 52.6 mL
decrease (95% CI – 107.0, 1.82) in FVC associated with an IQR increase in freeway PM2.5
aluminum and PM2.5 zinc (p ≤ 0.1) when NOx was included in the model. For an IQR increase in
freeway PM2.5-10 zinc, we observed a statistically significant 54.8 decrease (95% CI – 106.0, –
3.6) in FVC (p ≤ 0.05). Similar results were found with the PM metals alone without adjustment
for freeway NOx.

Table 3. Pearson correlation coefficients between key exposure
variables.
Freeway
NO x
(ppb)
Distance to
Freeway
(km)
Freeway NO x (ppb) 1 - 0.54
Distance to Freeway (km) - 0.54 1
PM 2.5 Cu (ng/m
3
) 0.33 - 0.10
PM 2.5-10 Al ng/m
3
) - 0.13 0.20
PM 2.5 Zn (ng/m
3
) 0.36 - 0.08
PM 2.5-10 Zn (ng/m
3
) 0.34 - 0.07
PM 2.5 (µg/m
3
) - 0.0069 0.078
PM 2.5-10 (µg/m
3
) 0.025 0.085

16
Table 4. Freeway NOx and PM metals exposures effect estimates and outcome (in mL per
IQR)*.
Model * Outcome Effect Estimate (95%
CI) **
Effect Estimate
(95% CI)
PM 2.5 Cu FVC – 40.9 (– 90.7, 8.8) – 35.8 (– 85.0, 13.5)
PM 2.5-10 Al FVC – 16.5 (– 33.6, 0.55)
a
– 15.0 (– 32.1, 2.1)
a

PM 2.5 Zn FVC – 52.6 (– 107.0, 1.82)
a
– 47.5 (– 100.4, 5.5)
a

PM 2.5-10 Zn FVC – 54.8 (– 106.0, – 3.6)
b
– 49.2 (– 99.1, 0.6)
b

NO x FVC 1.03 (– 25.5, 27.6 -
PM 2.5 FVC – 19.6 (– 40.3, 1.1) – 18.3 (– 39.6, 3.0)
PM 2.5-10 FVC – 28.4 (– 65.0, 8.3) – 25.3 (– 62.6, 12.0)
PM 2.5 Cu FEV 1 – 23.3 (– 69.9, 23.2) – 16.2 (– 64.6, 32.1)
PM 2.5-10 Al FEV 1 – 5.9 (– 23.6, 11.7) – 6.7 (– 25.4, 12.0)
PM 2.5 Zn FEV 1 – 25.8 (–76.7, 25.1) – 17.3 (– 69.5, 34.9)
PM 2.5-10 Zn FEV 1 – 16.2 (– 65.3, 32.9) – 10.1 (– 60.4, 40.3)
NO x FEV 1 9.86 (– 14.6, 34.3) -
PM 2.5 FEV 1 – 4.9 (– 26.1, 16.3) – 5.4 (– 28.2, 17.4)
PM 2.5-10 FEV 1 – 3.4 (– 39.6, 32.8) – 2.6 (– 39.9, 34.7)
a
p ≤ 0.1.
b
p ≤ 0.05.
* All models include covariate adjustment for age at time of lung function test, sex, height, BMI, race, ethnicity,
parental education, tobacco smoke, and a random intercept for community.
** Effect estimates with freeway NOx included in the model.

4. Conclusions

Our analysis of PM mass and key non-tailpipe tracer metals for brake wear (Cu), tire
wear (Zn), and resuspended road dust (Al) show several differences from tailpipe emissions. On
a community level, tailpipe emissions were clearly highest in Anaheim, where many of the CHS
study subjects lived closer to freeways and major arterials compared to other communities
(Franklin and Fruin, 2017). Non-tailpipe metals were in general higher in Mira Loma and Long
Beach where there is a lot of truck traffic, particularly in the ports of Long Beach and Los
Angeles. The spatial patterns in the PM mass and metals indicate small scale spatial variability,
with maximum spatial correlations (indicated by the range) between 6 km and 16 km. This
indicates that spatial variability at within-community scales is important, particularly when
characterizing exposures for the CHS subjects.
Seasonally, the fine mass concentrations did not differ drastically. Coarse mass
concentrations were higher in the warm season, likely due to contributions of windblown dust
under the hot, dry and windier conditions typical of inland southern California (Franklin et al.,
2018).
Overall, the epidemiological results indicate that NOx did not have as strong or
significant an impact on lung function as did the PM mass or metals. Marginal associations with
NOx were not statistically significant, and they were in the wrong direction. When included in
the joint tailpipe and non-tailpipe models they were also not statistically significant or in the
correct direction, whereas the effect estimates for all PM mass and metals were negative
indicating what is expected: that lung function decreases as air pollution concentrations increase.
17
Of the PM mass and metals, zinc showed the strongest association with lung function.
The association with FVC was statistically significant at the alpha=0.05 level, indicating that
with an IQR increase in freeway PM2.5-10 zinc, there is a 54.8 decrease (95% CI – 106.0, – 3.6) in
FVC. Similar associations with PM zinc were found in previous studies (Bloemsma, Hoek and
Smit, 2016) including a study of susceptible Italian adults (Lagorio et al., 2006), and a panel
study of university students in China (Wu et al., 2013). We also found that zinc and copper in the
fine fraction showed suggestive but large decreases in lung function.
Decreases in lung function of 40-50 mL per IQR increase in these PM metal components
are clinically relevant and indicate the importance of examining different aspects of the traffic
mixture. Not only do these findings suggest that non-tailpipe exposures are relatively more
important than tailpipe exposures, it also suggests that mass alone is not necessarily a sufficient
metric for characterizing PM exposures. Over the past decade, tailpipe exposures have decreased
significantly in California due to strict vehicle emissions standards (Gauderman et al., 2015).
Nevertheless, vehicle miles traveled continues to increase, particularly in Southern California,
indicating the relative importance of examining non-tailpipe components of traffic.

18
References

Abu-Allaban, M., Gillies, J.A., Gertler, A.W., Clayton, R. and Proffitt, D. (2003) Tailpipe,
resuspended road dust, and brake-wear emission factors from on-road vehicles.
Atmospheric Environment, 37, 5283–5293.
Benson, P.E. CALINE4 -- A Dispersion Model for Predicting Air Pollutant Concentrations near
Roadways, Report No. FHWA/CA/TL-84/15. Office of Transportation Laboratory,
California Department of Transportation, Sacramento, CA;
1989. http://www.dot.ca.gov/research/researchreports/1981-1988/84-15.pdf.
Bloemsma, L.D., Hoek, G. and Smit, L.A.M. (2016) Panel studies of air pollution in patients
with COPD: Systematic review and meta-analysis. Environmental Research, 151, 458–
468.
Caiazzo, F., Ashok, A., Waitz, I.A., Yim, S.H.L. and Barrett, S.R.H. (2013) Air pollution and
early deaths in the United States. Part I: Quantifying the impact of major sectors in 2005.
Atmospheric Environment, 79, 198–208.
Franklin, M. and Fruin, S. (2017) The role of traffic noise on the association between air
pollution and children’s lung function. Environmental Research, 157, 153–159.
Franklin, M., Kalashnikova, O., Garay, M. and Fruin, S. (2018) Characterization of Subgrid-
Scale Variability in Particulate Matter with Respect to Satellite Aerosol Observations.
Remote Sensing, 10, 623.
Franklin, M., Koutrakis, P. and Schwartz, J. (2008) The Role of Particle Composition on the
Association Between PM2.5 and Mortality: Epidemiology, 19, 680–689.
Franklin, M., Vora, H., Avol, E., McConnell, R., Lurmann, F., Liu, F., et al. (2012) Predictors of
intra-community variation in air quality. Journal of Exposure Science & Environmental
Epidemiology, 22, 135–147.
Gauderman, W.J., Edward, A., Frank, G., Hita, V., Duncan, T., Kiros, B., et al. (2004) The
Effect of Air Pollution on Lung Development from 10 to 18 Years of Age. The New
England Journal of Medicine, 11.
Gauderman, W.J., Urman, R., Avol, E., Berhane, K., McConnell, R., Rappaport, E., et al. (2015)
Association of Improved Air Quality with Lung Development in Children. New England
Journal of Medicine, 372, 905–913.
Gauderman, W.J., Vora, H., McConnell, R., Berhane, K., Gilliland, F., Thomas, D., et al. (2007)
Effect of exposure to traffic on lung development from 10 to 18 years of age: a cohort
study. , 369, 7.
Lagorio, S., Forastiere, F., Pistelli, R., Iavarone, I., Michelozzi, P., Fano, V., et al. (2006) Air
pollution and lung function among susceptible adult subjects: a panel study.
Environmental Health, 12.
19
Li, L., Lurmann, F., Habre, R., Urman, R., Rappaport, E., Ritz, B., et al. (2017) Constrained
Mixed-Effect Models with Ensemble Learning for Prediction of Nitrogen Oxides
Concentrations at High Spatiotemporal Resolution. Environmental Science &
Technology, 51, 9920–9929.
McConnell, R., Islam, T., Shankardass, K., Jerrett, M., Lurmann, F., Gilliland, F., et al. (2010)
Childhood Incident Asthma and Traffic-Related Air Pollution at Home and School.
Environmental Health Perspectives, 118, 1021–1026.
Requia, W.J., Higgins, C.D., Adams, M.D., Mohamed, M. and Koutrakis, P. (2018) The health
impacts of weekday traffic: A health risk assessment of PM2.5 emissions during
congested periods. Environment International, 111, 164–176.
Rice, M.B., Rifas-Shiman, S.L., Litonjua, A.A., Oken, E., Gillman, M.W., Kloog, I., et al. (2016)
Lifetime Exposure to Ambient Pollution and Lung Function in Children. American
Journal of Respiratory and Critical Care Medicine, 193, 881–888.
Urman, R., McConnell, R., Islam, T., Avol, E.L., Lurmann, F.W., Vora, H., et al. (2014)
Associations of children’s lung function with ambient air pollution: joint effects of
regional and near-roadway pollutants. Thorax, 69, 540–547.
Wu, S., Deng, F., Wang, X., Wei, H., Shima, M., Huang, J., et al. (2013) Association of lung
function in a panel of young healthy adults with various chemical components of ambient
fine particulate air pollution in Beijing, China. Atmospheric Environment, 77, 873–884.
Zanobetti, A., Franklin, M., Koutrakis, P. and Schwartz, J. (2009) Fine particulate air pollution
and its components in association with cause-specific emergency admissions.
Environmental Health, 8.

Abstract (if available)

Abstract Although it has been shown that traffic-related tailpipe air pollution adversely affects children’s lung function especially with traffic noise in the model, few studies have examined this association using traffic-related non-tailpipe emissions. ❧ Estimates of freeway emission concentrations of oxides of nitrogen (NOₓ, ppb), reflective of tailpipe traffic exposure, were spatially assigned to children in Southern California who were tested for forced vital capacity (FVC, n = 1217) and forced expiratory volume in 1s, (FEV₁, n = 1217). Estimates of fine and coarse concentrations of particulate matter mass (PM, μg/m³) and elements (PM metals, ng/m³) were obtained on the georeferenced Southern California participants residential locations. The associations between traffic-related non-tailpipe PM mass and PM metals—PM₂.₅ Cu, PM₂.₅₋₁₀ Al, PM₂.₅ and PM₂.₅₋₁₀ Zn, and the outcomes—FVC and FEV₁, were examined using mixed effects models. ❧ Of the PM mass and metals, both fine and coarse Zn metal showed significant association with lung function. A 52.6 mL decrease (95% CI – 107.0, 1.82) and a 54.8 decrease (95% CI – 106.0, – 3.6) in FVC was associated with an interquartile range increase in PM₂.₅ Zn (p ≤ 0.1) and PM₂.₅₋₁₀ Zn (p ≤ 0.05) respectively, with NOₓ included in the model. ❧ Our results suggest that non-tailpipe traffic-related exposures were relatively more important than tailpipe exposures alone in determining lung function association. As a result, more non-tailpipe PM metals should be included in epidemiological studies of the association between traffic-related air pollution on lung function.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Assessment of the mortality burden associated with ambient air pollution in rural and urban areas of India

PDF

Disparities in exposure to traffic-related pollution sources by self-identified and ancestral Hispanic descent in participants of the USC Children’s Health Study

PDF

Long term effects of air pollution on lung function in the European community respiratory health survey

PDF

Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data

PDF

Spatial analysis of PM₂.₅ air pollution in association with hospital admissions in California

PDF

Ambient air pollution and lung function in children

PDF

Effectiveness of individual- and household-level protective actions in reducing symptoms associated with hydrogen sulfide chronic low-level exposure

PDF

Uncertainty quantification in extreme gradient boosting with application to environmental epidemiology

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Predicting ototoxicity evaluated by SIOP in children receiving cisplatin

PDF

Assessing the impact of air pollution on adverse birth outcomes in a low resource setting

PDF

Mixed-method analysis for school-aged children with access and functional needs during the Covid-19 pandemic

PDF

Association of traffic-related air pollution and lens opacities in the Los Angeles Latino Eye Study

PDF

Investigating the role of urban emission sources on redox-active PM compounds and the chemical analysis of the standardized diesel exhaust particles

PDF

Associations of ambient air pollution exposures with perceived stress in the MADRES cohort

PDF

Sensor-based mobile health approaches for personal air pollution and pediatric asthma studies

PDF

Cross-sectional association of blood pressure, antihypertensive medications, MRI volumetric measures and cognitive function scores in an aging population

PDF

Associations between ambient air pollution and hypertensive disorders of pregnancy

PDF

Airway inflammation and respiratory health in the Southern California children's health study

PDF

Validation of the Children’s International Mucositis Evaluation Scale (ChIMES) in pediatric cancer and SCT

Asset Metadata

Creator Okonkwo, Chika Augusta (author)

Core Title Spatial modeling of non-tailpipe emissions and its association with children's lung function

School Keck School of Medicine

Degree Master of Science

Degree Program Applied Biostatistics and Epidemiology

Publication Date 04/28/2019

Defense Date 04/25/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Air pollution,children's respiratory health,lung function,non-tailpipe emissions,OAI-PMH Harvest,tailpipe emissions,Traffic

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Franklin, Meredith (committee chair), Gauderman, William (committee member), Habre, Rima (committee member)

Creator Email chikaoko@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-150478

Unique identifier UC11660639

Identifier etd-OkonkwoChi-7317.pdf (filename),usctheses-c89-150478 (legacy record id)

Legacy Identifier etd-OkonkwoChi-7317.pdf

Dmrecord 150478

Document Type Thesis

Format application/pdf (imt)

Rights Okonkwo, Chika Augusta

Type texts

Source 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 a...

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