Impact of mine waste on airborne respirable particulates in northeastern Oklahoma, United States.

Abstract

Atmospheric dispersion of particles from mine waste is potentially an important route of human exposure to metals in communities close to active and abandoned mining areas. This study assessed sources of mass and metal concentrations in two size fractions of respirable particles using positive matrix factorization (U.S. Environmental Protection Agency [EPA] PMF 3.0). Weekly integrated samples of particulate matter (PM) 10 microm in aerodynamic diameter or less (PM10) and fine PM (PM2.5, or PM <2.5 microm in aerodynamic diameter) were collected at three monitoring sites, varying distances (0.5-20 km) from mine waste piles, for 58 consecutive weeks in a former lead (Pb) and zinc (Zn) mining region. Mean mass concentrations varied significantly across sites for coarse PM (PM10-PM2.5) but not PM2.5 particles. Concentrations of Pb and Zn significantly decreased with increasing distance from the mine waste piles in PM10-PM2.5 (P < 0.0001) and PM2.5 (P < 0.0005) fractions. Source apportionment analyses deduced five sources contributing to PM2.5 (mobile source combustion, secondary sulfates, mine waste, crustal/soil, and a source rich in calcium [Ca]) and three sources for the coarse fraction (mine waste, crustal/soil, and a Ca-rich source). In the PM2.5 fraction, mine waste contributed 1-6% of the overall mass, 40% of Pb, and 63% of Zn. Mine waste impacts were more apparent in the PM10-PM2.5 fraction and contributed 4-39% of total mass, 88% of Pb, and 97% of Zn. Percent contribution of mine waste varied significantly across sites (P < 0.0001) for both size fractions, with highest contributions in the site closest to the mine waste piles. Seasonality, wind direction, and concentrations of the Ca-rich source were also associated with levels of ambient aerosols from the mine waste source. Scanning electron microscopy results indicated that the PMF-identified mine waste source is mainly composed of Zn-Pb agglomerates on crustal particles in the PM10-PM2.5 fraction. In conclusion, the differential impacts of mine waste on respirable particles by size fraction and location should be considered in future exposure evaluations.

Full text

Page 1

Impact of Mine Waste on Airborne Respirable Particulates in
Northeastern Oklahoma, United States
Ami R. Zota
Department of Environmental Health, Harvard School of Public Health, Boston, MA
Robert Willis
National Exposure Research Laboratory, U.S. Environmental Protection Agency, Research
Triangle Park, NC
Rebecca Jim
Local Environmental Action Demanded Agency, Vinita, OK
Gary A. Norris
National Exposure Research Laboratory, U.S. Environmental Protection Agency, Research
Triangle Park, NC
James P. Shine
Department of Environmental Health, Harvard School of Public Health, Boston, MA
Rachelle M. Duvall
National Exposure Research Laboratory, U.S. Environmental Protection Agency, Research
Triangle Park, NC
Laurel A. Schaider and John D. Spengler
Department of Environmental Health, Harvard School of Public Health, Boston, MA
ABSTRACT
Atmospheric dispersion of particles from mine waste is
potentially an important route of human exposure to
metals in communities close to active and abandoned
mining areas. This study assessed sources of mass and
metal concentrations in two size fractions of respirable
particles using positive matrix factorization (U.S. Environ-
mental Protection Agency [EPA] PMF 3.0). Weekly inte-
grated samples of particulate matter (PM) 10 ?m in aero-
dynamic diameter or less (PM10) and fine PM (PM2.5, or
PM ?2.5 ?m in aerodynamic diameter) were collected at
three monitoring sites, varying distances (0.5–20 km)
from mine waste piles, for 58 consecutive weeks in a
former lead (Pb) and zinc (Zn) mining region. Mean mass
concentrations varied significantly across sites for coarse
PM (PM10–PM2.5) but not PM2.5particles. Concentrations
of Pb and Zn significantly decreased with increasing dis-
tance from the mine waste piles in PM10–PM2.5(P ?
0.0001) and PM2.5(P ? 0.0005) fractions. Source appor-
tionment analyses deduced five sources contributing to
PM2.5(mobile source combustion, secondary sulfates,
mine waste, crustal/soil, and a source rich in calcium [Ca])
and three sources for the coarse fraction (mine waste,
crustal/soil, and a Ca-rich source). In the PM2.5fraction,
mine waste contributed 1–6% of the overall mass, 40% of
Pb, and 63% of Zn. Mine waste impacts were more appar-
ent in the PM10–PM2.5fraction and contributed 4–39% of
total mass, 88% of Pb, and 97% of Zn. Percent contribu-
tion of mine waste varied significantly across sites (P ?
0.0001) for both size fractions, with highest contributions
in the site closest to the mine waste piles. Seasonality,
wind direction, and concentrations of the Ca-rich source
were also associated with levels of ambient aerosols from
the mine waste source. Scanning electron microscopy re-
sults indicated that the PMF-identified mine waste source
is mainly composed of Zn-Pb agglomerates on crustal
particles in the PM10–PM2.5fraction. In conclusion, the
IMPLICATIONS
This is the first study to use source apportionment model-
ing along with scanning electron microscopy to quantify the
impact of mine waste on respirable particles in residential
areas surrounding an abandoned mining site. Fugitive dust
emissions from mine waste were found predominantly in
the PM10–PM2.5fraction. Impacts were most substantial in
Picher, the source-dominated site, and decreased with in-
creasing distance from the mine waste piles, suggesting
that populations living nearest to these piles may be more
highly exposed. These results will enable more accurate
assessments of human exposure and health effects in com-
munities adjacent to active and abandoned mining areas.
TECHNICAL PAPER
ISSN:1047-3289 J. Air & Waste Manage. Assoc. 59:1347–1357
DOI:10.3155/1047-3289.59.11.1347
Copyright 2009 Air & Waste Management Association
Volume 59 November 2009
Journal of the Air & Waste Management Association 1347

Page 2

differential impacts of mine waste on respirable particles
by size fraction and location should be considered in
future exposure evaluations.
INTRODUCTION
The U.S. Environmental Protection Agency (EPA) esti-
mates there are over 200,000 inactive and abandoned
hardrock mines in the United States.1Active and aban-
doned mines have the potential to cause damage to
aquatic biota; terrestrial vegetation; and wildlife, air qual-
ity, and cultural resources. Seventy-two mining sites have
been designated to the National Priority List under the
EPA Superfund program. These sites pose an imminent
threat to human or ecological health and warrant federal
intervention.2Coal and hardrock mining continue glo-
bally, regularly creating more abandoned mines.
In metal mining, less than 1% of processed material is
recovered as useful metal.3Mine tailings and other metal-
enriched byproducts of the mining process are often
stored in large piles that can become sources of contam-
ination to surrounding ecosystems and residential areas
through wind-borne dispersal of particles. Suspended air-
borne particles can travel off-site and infiltrate indoors
where they can be directly inhaled, or deposit onto soil or
house dust by settling, impaction, or washout.
Mine waste piles represent a dispersed source of metal
contamination and are particularly abundant at the Tar
Creek Superfund Site, a former lead (Pb) and zinc (Zn)
mining area located in rural Oklahoma. Mine waste, lo-
cally known as “chat,” is largely composed of chert (SiO2);
dolomite (CaMg(CO3)2); and sulfide minerals including
galena (PbS), sphalerite (ZnS), and pyrite (FeS2).4There are
approximately 30 major chat piles in the Tar Creek area,5
which contain elevated concentrations of Zn, Pb, and
cadmium (Cd).6,7Recent research suggests metal concen-
trations in chat particles increase with decreasing particle
size, and ultrafine particles (?1 ?m) contain Zn, Pb, and
Cd at concentrations up to 20 times higher than the bulk
material.7Chat particles may be transported into the
broader environment through various mechanisms such
as wind erosion. Chat may also be deposited on roads,
either through atmospheric settling or when used as a
gravel material, and then become resuspended and dis-
persed with traffic. In addition, local reprocessing of chat
forasphaltandothertransportation
projects6–8may lead to increased aerosol mobilization.
Variable weather conditions including sporadic events
such as windstorms, along with human disturbances, may
lead to spatial and temporal heterogeneity in airborne
contaminants.
A large body of evidence has shown that exposure to
particulate matter (PM) is harmful to human health. Par-
ticle size not only determines the site and efficiency of
pulmonary deposition but also may be an indication of
particle source and composition. Fine PM (PM2.5, or PM
?2.5 ?m in aerodynamic diameter) particles, largely gen-
erated from combustion processes, have been associated
with a range of adverse respiratory and cardiovascular
health effects including mortality.9,10Coarse particles
(PM10–PM2.5), generated primarily from mechanical pro-
cesses, are more commonly associated with respiratory
and cardiovascular morbidity11(e.g., inflammatory lung
construction
injury12). Additionally, experimental and epidemiologic
evidence suggests that metal constituents in PM, such
as vanadium (V), Zn, iron (Fe), and nickel (Ni), play an
important role in inflammatory and cardiovascular
health effects.12–15Inhalation of metals may be partic-
ularly toxic because metals such as manganese (Mn),
Cd, Zn, and Ni can be transported directly to the brain
via olfactory pathways.16–18
Because there is a potential for inhalation exposure to
metal-enriched PM in abandoned mining areas such as
the Tar Creek Superfund Site, it is important to quantify
the impact of mine waste piles on ambient air quality.
Although previous studies have used multivariate statisti-
cal receptor models to identify the contribution of active
mining operations (e.g., smelting) to respirable parti-
cles,19,20few studies have used these techniques to quan-
tify the impact of mine waste, a more indirect and persis-
tent source of respirable particles, in residential areas
surrounding abandoned mining sites. This study is part of
an ongoing effort to understand children’s exposure to
mining-related metal mixtures and subsequent health ef-
fects. The goal of this study was to examine the impact of
chat-related sources on ambient particle concentrations at
the Tar Creek Superfund Site. Specifically, sources of
PM2.5and PM10–PM2.5particles were identified, and their
contributions to mass and metal concentrations were es-
timated using positive matrix factorization (PMF). Predic-
tors of source contributions were identified using regres-
sion analysis. Lastly, the presence of unique, local sources
was qualitatively confirmed using scanning electron mi-
croscopy techniques.
EXPERIMENTAL METHODS
Study Design
Ottawa County, OK, is a predominantly rural area with
high humidity and abundant rainfall between the
months of May and November. The predominant wind
direction is from the south with an average wind speed of
11 km?hr?1.21Three stationary air monitoring sites were
established in the area to capture spatial variability and
represent potentially different human exposure scenarios.
Weekly integrated samples were collected at each site for
58 weeks from July 2005 to September 2006.
Figure 1 shows the locations of the three sites in
relation to the chat piles. Sites were chosen through con-
sultation with our community partners in the area. Site 1,
the source-dominated site, was located in a residential
yard in the town of Picher, surrounded by chat piles on
two sides and within a kilometer of several other piles.
Site 2 was located in the town of Quapaw, approximately
5 km from the bulk of the chat piles, but close to a
well-used county road and several dirt roads that are
thought to be lined with chat material.22Site 3 was lo-
cated in a suburban neighborhood within Miami, the
largest town in Ottawa County, and approximately 18 km
upwind of the Picher site with no chat piles in the nearby
vicinity.
Analytical Methods
Separate filter samples of PM2.5particles and particles
from PM less than 10 ?m in aerodynamic diameter (PM10)
were collected on Teflon filters (2-?m pore size, 37 mm in
Zota et al.
1348 Journal of the Air & Waste Management Association Volume 59 November 2009

Page 3

diameter) using Harvard impactors23attached to Medo air
pumps at 4 L ? min?1. Filters were exposed for 7 days (24
hr a day) and changed once a week. Airflow was measured
and calibrated at the beginning and end of every filter
exposure using calibrated flow meters from Matheson
Tri-gas, Model 603(E500).
Teflon filters were weighed in a temperature- and
humidity-controlled room (18–24 °C, 40 ? 5% relative
humidity). All filters were left to equilibrate for 24 hr
before presampling weighing and 48 hr before postsam-
pling weighing. The elemental content of the ambient
aerosol was quantified by energy-dispersive X-ray fluores-
cence (EDXRF) analysis, a nondestructive and moderately
sensitive analytical method for determining elemental
concentrations of aluminum (Al) through Pb.24,25Analy-
ses were performed at EPA’s National Exposure Research
Laboratory (NERL) in Research Triangle Park, NC, using
an EDXRF spectrometer custom-built for EPA by the Law-
rence Berkeley Laboratory. Black carbon (BC) concentra-
tions were estimated in the PM2.5fraction using reflec-
tance analysis on the particle filters, a method that
provides measurements that are highly correlated with
concentrations measured using thermal/optical meth-
ods.26Sample absorbance values were obtained using an
optical transmissometer data acquisition system (Magee
Scientific). Ultraviolet (UV) and infrared (IR) absorbance
values for each sample were collected at 370 and 880 nm
wavelengths, respectively. A mass absorption coefficient
of 16.6 was used to convert the transmittance to mass
concentration units.27 Computer-controlled scanning
electron microscopy (CCSEM)28coupled with energy-
dispersive X-ray analysis (EDX) (R.J. Lee Instruments,
Ltd., now Aspex Corp.) was conducted on specific filters
to help interpret sources identified from the statistical
receptor models. The metals and reflectance analyses of
particle filters and CCSEM/EDX analyses were conducted
according to standard operating procedures at EPA’s NERL
(Research Triangle Park, NC).
Quality Control and Quality Assurance
Seven percent of the total samples collected were voided
because they failed to meet established flow or sampling
time criteria. Field blanks were transported and handled
like regular samples but the filters were not attached to
the air pumps. Field blanks comprised 10% of the total
samples collected and were used to determine background
contamination. The method limit of detection (LOD) for
each species was calculated as 3 times the average uncer-
tainty of 23 laboratory Teflon blanks divided by the me-
dian volume for all samples (40 m3). If the mean field
blank concentration was greater than the mean plus 3
times the standard deviation (SD) of the laboratory
blanks, then blank correction was applied. Sample con-
centrations of Ca and Fe in one batch of PM2.5samples
were blank-corrected by subtracting the mean field blank
concentration from the sample concentrations in this
batch. Precision of the method was determined by dupli-
cate samples (10% of total samples collected). BC concen-
trations were imputed for two samples using the median
concentration of the sampling location. Samples collected
during the week of June 30, 2006 were eliminated from all
analyses because these samples were impacted by fire-
works and had high levels of potassium (K), copper (Cu),
and strontium (Sr).
Meteorological Data
Meteorological data including daily measures of temper-
ature, average and maximum wind speed, wind direction,
and precipitation were obtained from the National
Figure 1.
Location of mine waste (“chat”) piles and ambient air monitors in Ottawa County, OK.
Zota et al.
Volume 59 November 2009
Journal of the Air & Waste Management Association 1349

Page 4

Oceanographic and Atmospheric Administration (Wash-
ington, DC) weather station for Tulsa International Air-
port, approximately 145 km south of the study area.
Weekly averages for these variables were calculated and
used in regression analyses. Because wind direction data
were originally provided in degrees, a dummy variable
was constructed to correspond to 90° directional incre-
ments (e.g., south, west, etc.). Similarly, a dummy vari-
able was constructed to correspond to the four calendar
seasons.
SUMMARY STATISTICS
After quality assurance/quality control criteria were im-
plemented, a total of 156 PM10and 155 PM2.5samples
were available for data analysis. Coarse concentrations
were not directly measured but were calculated as the
difference between PM10and PM2.5concentrations for
the 150 samples where PM10and PM2.5measurements
were available. To generate summary statistics and com-
pare concentrations across sites, a balanced dataset was
created that included only those samples in which corre-
sponding data were available at all three sites for both size
fractions (n ? 123). For data that fell below the method
LOD, estimated metal concentrations provided by EDXRF
analyses were used in calculation of summary statistics
and statistical models. Correlations were assessed using
Spearman rank correlations. Statistically significant differ-
ences in metal concentrations across sites were deter-
mined using the ANOVA test for differences and Scheffe’s
test for multiple comparisons. Before the ANOVA analy-
sis, the Levene’s test for homogeneity of variance was
implemented. If the assumption of homogenous variance
was not upheld, the Welch’s ANOVA test, which accounts
for unequal group variances, was used instead.
Receptor Modeling: EPA PMF 3.0
A receptor modeling approach using EPA PMF 3.0 was
used to quantify sources in PM2.5and PM10–PM2.5. PMF
uses a constrained, weighted, least squares regression
via the Multilinear Engine (ME-2) to generate source
profiles and source contributions. Further details on the
algorithm can be found in the PMF 3.0 user guide.29
Input data include sample concentration and uncer-
tainty estimates.
Calculation of uncertainty estimates varied according
to chemical species and size fraction. For elemental con-
centrations determined by EDXRF, sample- and element-
specific concentration uncertainties were provided that
equaled 1 SD of error estimates based on analytical preci-
sion. Because sample-specific uncertainty estimates were
not provided for particle mass or BC concentrations, these
uncertainty estimates were defined as 10 and 20% of the
measured concentration, respectively.30Uncertainty esti-
mates for coarse fraction measurements were calculated as
the following:
coarse ?ij??
?PM10?i,j?2? ?PM2.5?i,j?2
2
(1)
where uncertainty ? is the jth species uncertainty esti-
mated for the ith sample.
The model was run in default robust mode to mini-
mize the effects of outliers and included samples from all
three sites because it was assumed that source profiles
would not vary across sites. All models were normalized to
PM mass concentrations. Thirty base runs were executed
for each specific model, and model goodness-of-fit was
evaluated by examining Q (robust) values. The solution
with the lowest Q value was chosen, and 100 bootstrap
simulations were performed to estimate the stability and
uncertainty of that solution, which involved each of the
bootstrapped factors being mapped to exactly one of the
base-case factors. The number of factors in the final solu-
tion was decided using a priori knowledge about local
sources and by maximizing agreement between base run
and bootstrapped results.
Table 1a. Mean (SD) mass (?g/m3) and metal (ng/m3) concentrations in coarse (PM10–PM2.5) particles by location.
PM10–PM2.5
Fraction% <LODPicher QuapawMiami
ANOVA P
Valuea
Sample size
Mass
Al
Si
S
K
Ca
Ti
Mn
Fe
Cu
Zn
As
Se
Br
Pb
41
12 (4.4)
350 (140)
1700 (610)
20 (120)
110 (52)
1200 (560)
30 (14)
7.8 (3.3)
290 (120)
0.95 (0.56)
120 (69)
?LOD
?LOD
0.55 (0.60)
7.9 (4.6)
41
13 (6.1)
360 (130)
1600 (710)
?LOD
110 (44)
1300 (750)
33 (19)
7.7 (3.5)
340 (180)
0.83 (0.54)
39 (31)
?LOD
?LOD
0.48 (0.41)
3.0 (2.4)
41
9.0 (2.9)
300 (130)
1200 (380)
?LOD
96 (33)
970 (400)
22 (8.9)
6.3 (2.1)
220 (78)
0.71 (0.55)
11 (5.6)
?LOD
?LOD
?LOD
1.1 (0.75)
0
1
0
<0.0001
0.12
<0.0001
0.04
0.12
0.02
0.0003
0.02
0.0003
0.13
<0.0001
–
0.55
0.03
<0.0001
59
0
0
0
0
0
36
0
74
85
41
20
Notes:aValues in bold indicate significant (P ? 0.05) differences among sites.
Zota et al.
1350 Journal of the Air & Waste Management Association Volume 59 November 2009

Page 5

The source apportionment model for PM2.5in-
cluded concentration and uncertainty data from 155
samples. Fourteen possible elements and two carbon
fractions were considered for inclusion in the PMF
model. Species were evaluated based on their detection
frequency, signal-to-noise ratio, and usefulness as
source tracers. The following analytes were included in
the PM2.5source apportionment analysis: Al, silicon
(Si), sulfur (S), K, Ca, titanium (Ti), Mn, Fe, Zn, sele-
nium (Se), bromine (Br), Pb, and BC.
Because PM10–PM2.5fraction metal concentrations
were not directly measured, a more conservative ap-
proach was used when conducting the PM10–PM2.5frac-
tion PMF analysis. To be included in the PMF analysis,
elements were assessed on the following two criteria: (1)
well detected (?70%) in the PM10–PM2.5fraction and (2)
had reasonably higher concentrations in PM10as com-
pared with PM2.5. On the basis of these criteria, 9 ele-
ments and 150 samples were included in the PM10–PM2.5
PMF model. This list was similar to that of the PM2.5
model but omitted S, Se, Br, and BC.
Regression Analyses
Source-specific mass concentration estimates were ex-
tracted for each sample from the PMF output. Univari-
ate and bivariate summary statistics and distributional
plots were examined for all variables. Mine waste esti-
mates were positively skewed and log-transformed be-
fore regression analysis. Multiple linear regression mod-
els were used to describe the relationships between
PMF-estimated mine waste concentrations and predic-
tor variables of interest, which included season, wind
and speed, precipitation, site location, and estimated
mass concentration data of the other PMF sources. To
obtain the final model, backward elimination was used
with a threshold P value of less than 0.05 for retaining
the variable in the model. Regression analyses were
conducted in SAS version 9.1.
RESULTS
Summary Statistics
Mean (?SD) PM10mass concentrations in Picher (23 ?
6.4 ?g/m3) and Quapaw (24 ? 7.6 ?g/m3) were signifi-
cantly higher (P ? 0.01) than concentrations in Miami
(20 ? 5.1 ?g/m3). Similar to PM10, PM10–PM2.5particle
mass concentrations in Picher (12 ? 4.4 ?g/m3) and Qua-
paw (13 ? 6.1 ?g/m3) were higher than Miami (9 ? 2.9
?g/m3), suggesting a difference in local sources between
these areas (Table 1, a and b). PM2.5concentrations did
not differ by site and approximated 11 ?g/m3at all three
sites. PM10was more strongly correlated with PM10–PM2.5
than with PM2.5in Picher (r ? 0.81) and Quapaw (r ?
0.85). By contrast, in the more commercially developed
town of Miami, PM10concentrations were more strongly
correlated with PM2.5(r ? 0.82). Seasonal variability was
present in PM2.5and PM10, with highest concentrations
in the summer and lowest concentrations in the winter
(data not shown).
Mass and selected metal concentrations for PM10–
PM2.5and PM2.5particulates by location are presented in
Table 1, a and b. There was a significant spatial trend in Pb
and Zn concentrations, with both elemental concentra-
tions decreasing with increasing distance from the chat
piles. In the more locally influenced PM10–PM2.5fraction,
mean Pb concentrations were significantly different at all
three sites ([Picher: 7.9 ? 4.6] vs. [Quapaw: 3 ? 2.4] vs.
[Miami: 1.1 ? 0.75] ng/m3, P ? 0.0001). Similarly, Zn
concentrations in PM10–PM2.5particles varied signifi-
cantly at all three sites, with a 10-fold difference between
Picher (120 ? 69 ng/m3) and Miami (11 ? 5.6 ng/m3).
Elements associated with crustal sources such as Al, Si, Ca,
and Fe were generally higher in Quapaw and Picher com-
pared with Miami.
As expected, there were fewer differences in metal
concentrations across sites in the PM2.5fraction, likely
because of the impact of regional pollution sources and
the slower deposition rate of small particles. Pb and Zn
concentrations showed a similar but less pronounced
Table 1b. Mean (SD) mass (?g/m3) and metal (ng/m3) concentrations in PM2.5particles by location.
PM2.5
Fraction% <LODa
PicherQuapawMiami
ANOVA
P Valuea
Sample size
Mass
Al
Si
S
K
Ca
Ti
Mn
Fe
Cu
Zn
As
Se
Br
Pb
41
11 (4.0)
62 (75)
220 (150)
870 (480)
57 (21)
97 (44)
4.2 (3.7)
1.4 (0.64)
58 (38)
1.1 (1.2)
22 (10)
0.64 (0.48)
0.58 (0.31)
2.6 (1.1)
3.5 (2.6)
41
11 (3.9)
67 (77)
230 (160)
860 (490)
57 (19)
91 (45)
4.5 (3.5)
1.8 (0.84)
70 (41)
0.90 (0.52)
11 (5.0)
0.62 (0.32)
0.57 (0.27)
2.5 (0.84)
2.2 (1.4)
41
11 (3.8)
57 (84)
190 (170)
890 (480)
61 (20)
77 (33)
4.5 (4.1)
1.5 (0.71)
54 (43)
0.80 (0.61)
7.9 (3.0)
0.56 (0.33)
0.58 (0.30)
2.4 (0.88)
1.9 (1.4)
00.98
0.83
0.44
0.96
0.52
0.09
0.95
0.04
0.20
0.20
<0.0001
0.58
0.99
0.73
0.0003
49
0
0
0
0
6
1
0
37
0
41
23
0
10
Notes:aValues in bold indicate significant (P ? 0.05) differences among sites.
Zota et al.
Volume 59 November 2009
Journal of the Air & Waste Management Association 1351

Page 6

spatial gradient compared with the PM10–PM2.5frac-
tion. Pb concentrations at Picher (3.5 ? 2.6 ng/m3)
were elevated relative to Miami (1.9 ? 1.4 ng/m3) by a
factor of 2. A similar trend was also observed for Zn in
the PM2.5fraction.
Characterization of Sources
When EPA PMF 3.0 was applied to PM2.5data, the model
converged, yielding a five-factor solution. There was good
agreement between the predicted and measured PM2.5
mass (R2? 0.75). The R2values for the elements ranged
from 0.50 to 0.99. Over 87 bootstraps out of 100 were
mapped to the original base factor.
A three-factor solution was extracted from the PM10–
PM2.5fraction data. The PMF solution for the PM10–PM2.5
data exhibited a better goodness-of-fit compared with
the solution for the PM2.5fraction data. The correlation
between predicted and measured PM mass yielded an R2
? 0.89, and the R2for the elements ranged from 0.85–
0.99. All 100 bootstraps were mapped to the original
base factor.
The source profiles deduced from the PMF models for
the PM2.5and PM10–PM2.5fractions are presented in Fig-
ure 2. Consistent with the composition of mine waste,
there was a factor present in both size fractions, the source
profiles of which were dominated by Zn and Pb with
modest contributions of crustal elements (Al, Si, and Ca;
Figure 2). The average Zn-to-Pb ratios in this source profile
were 9 and 16 in the PM2.5and PM10–PM2.5fractions,
respectively, which is in agreement with the ratio found
in “parent” chat particles less than 37 ?m collected in the
Tar Creek area (Zn-to-Pb ratio ? 15; unpublished data).
The percent mass of Zn and Pb in the PMF-deduced source
profile in PM2.5and PM10–PM2.5fractions were also sim-
ilar to that of the “parent” chat. Zn was found in the
1–10% range and Pb was found in less than 1% of the
total mass.7
The source profile of soil and crustal materials was
characterized by the component loaded on Al, Si, Ti, Mn,
Fe, and K. These elements have commonly been used to
identify crustal sources.31,32In PM10–PM2.5and PM2.5
fractions, there was a source rich in Ca. This factor also
showed contributions from Si, Al, and BC (PM2.5fraction
only) and likely incorporates the impact of unpaved and
paved road dust. In both size fractions, the Al-to-Ca ratio
was approximately 0.1 in the Ca-rich factor, which is
similar to unpaved road dust as characterized by Chow et
al.31Also, consistent with the unpaved road dust source
profile,31the percent mass of Ca in the PMF-deduced
Ca-rich profile was 15% in the PM10–PM2.5fraction. The
last two factors were only found in the PM2.5particles.
These were mobile source combustion and secondary sul-
fates from coal combustion and were readily identified by
comparison with previously reported profiles.31–33
Mass and Elemental Apportionments
PM2.5Fraction. Average contributions of each source to
PM2.5are summarized in Table 2. Most PM2.5mass was
apportioned to secondary sulfates and mobile sources
with less than 5% attributed to the mine waste source.
Elemental apportionment found that approximately 40%
of Pb and 63% of Zn were apportioned to mine waste.
Variations in mass apportionments were observed across
sites. Average mine waste contributions ranged from 1%
in Miami to 6% in Picher (Figure 3a). Contributions of the
Ca-rich source were highest in Quapaw and mobile source
contributions were highest in Miami. Secondary sulfate
estimates were similar at all sites. Variation was not only
0.001
0.001
0.01
0.01
0.1
0.1
1
1
0.001
0.001
0.01
0.01
0.1
0.1
1
1
0.001
0.001
0.01
0.01
0.1
0.1
1
1
0.001
0.001
0.01
0.01
0.1
0.1
1
1
0.001
0.001
0.01
0.01
0.1
0.1
1
1
Al
Al
Si
Si
SKCaTi
Ti
Mn
Mn Fe
FeZn
Zn Se
SeBr
Br Pb Bc
PbBC
Concentration (µg/g)
Coarse (PM10 - PM2.5)
Fine (PM2.5)
Mine Waste
Crustal
Ca-rich
Secondary
Mobile Source
S K Ca
Figure 2.
derived from EPA PMF 3.0 for (a) mine waste, (b) crustal/soil, (c)
Ca-rich, (d) secondary sulfates, and (e) mobile source material.
Source profiles for PM2.5and PM10–PM2.5fraction
Table 2. Mean source contributions (in ?g/m3) of coarse (PM10–PM2.5)
and fine (PM2.5) particulate from PMF.
PM10–PM2.5
(n ? 150)
PM2.5
(n ? 155)
Mine waste
Crustal
Ca-rich
Secondary
Mobile sources
2.2 (0.20)
4.6 (0.25)
4.1 (0.24)
–
–
0.31 (0.025)
1.2 (0.15)
1.7 (0.098)
2.9 (0.22)
4.4 (0.20)
Notes: Values in parentheses represent standard error (SE).
Zota et al.
1352 Journal of the Air & Waste Management AssociationVolume 59 November 2009

Page 7

observed across locations, but also within locations. At
Picher, the source-dominated site, there was a 10-fold
difference between minimum and maximum estimated
concentrations of mine waste (0.11–1.4 ?g/m3).
PM10–PM2.5
source to PM10–PM2.5mass concentrations are also sum-
marized in Table 2. Mean contributions of mine waste,
crustal, and Ca-rich across sites were 20, 42, and 38%,
respectively. Elemental apportionments found that 88%
of Pb and 97% of Zn were associated with mine waste. Not
only were mass and percent contributions of mine waste
greater in the PM10–PM2.5fraction relative to the PM2.5
fraction, but spatial differences were also more substantial
with 39, 11, and 4% percent contributions in Picher,
Quapaw, and Miami, respectively (Figure 3b). Mine waste
was the largest contributor to PM10–PM2.5mass in Picher
(Figure 3b), and in Picher and Quapaw, mine waste con-
centrations ranged 2 orders of magnitude—from 0.1 to 10
?g/m3. The percent contributions of crustal and Ca-rich
sources were similar between Quapaw and Miami; how-
ever, the concentrations apportioned to these sources
were greater in Quapaw.
Fraction. Averagecontributionsofeach
Predictors of Mine Waste Source Contribution
Predictors of mine waste attributable mass were examined
using the estimated concentrations (?g/m3) from PMF
source apportionment models in correlation and linear
regression analysis. Mine waste estimates in PM2.5and
PM10–PM2.5fractions were strongly correlated (r ? 0.76,
P ? 0.0001), but PM2.5fraction mine waste was not sig-
nificantly correlated with any of the other PM2.5sources.
Predictors of PM2.5fraction mine waste in multivariate
models included sampling site, season, wind direction,
and PM10–PM2.5fraction mine waste concentrations (Ta-
ble 3). These factors collectively accounted for 64% of the
variability in PM2.5fraction mine waste estimates. Mine
waste concentrations were highest in Picher and lowest in
Miami. Concentrations were highest in the fall and lowest
in the winter. Elevated mine waste concentrations were
also associated with easterly winds. Wind speed, precipi-
tation, and temperature were not associated with PM2.5
mine waste.
PM10–PM2.5fraction mine waste was positively cor-
related with PM10–PM2.5Ca-rich mass (r ? 0.26, P ?
0.001) and negatively correlated with PM10–PM2.5crustal
mass (r ? ?0.16, P ? 0.05). In univariate analysis, precip-
itation was inversely associated with PM10–PM2.5mine
waste concentrations, and sampling site was a significant
predictor. The final multivariate model included sam-
pling site, season, and PM10–PM2.5Ca-rich concentra-
tions (Table 3). PM10–PM2.5mine waste concentrations
were highest in Picher and lowest in Miami. PM10–PM2.5
mine waste levels increased with increasing temperature,
with highest levels in the summer and lowest levels in the
winter. PM10–PM2.5Ca-rich concentrations remained a
significant predictor after accounting for the other vari-
ables, whereas precipitation was no longer significant in
the multivariate model. In addition, a statistically signif-
icant interaction (P ? 0.0001) was observed between sam-
pling site and Ca-rich concentrations, suggesting a differ-
ential relationship between Ca-rich and mine waste by
location. Figure 4 shows the graphical interpretation of
this statistical interaction term. PM10–PM2.5 fraction
mine waste increased linearly with PM10–PM2.5fraction
Ca-rich estimates in both Quapaw and Miami. Con-
versely, no relationship was observed between the mine
waste and Ca-rich source in Picher. Collectively, these
factors accounted for 79% of the variability in PMF-
estimated PM10–PM2.5mine waste concentrations. PM10–
PM2.5mine waste was not associated with wind speed or
direction.
Scanning Electron Microscopy
CCSEM was used to characterize Zn- and Pb-bearing par-
ticles in a PM10sample from the Picher site that had a
high factor score for the mine waste source in the PM10–
PM2.5size fraction. Several hundred individual particles
larger than 0.4 ?m with detectable concentrations of Zn
or Pb were sized and analyzed for elemental composition
by EDX. Consistent with source apportionment modeling
results, CCSEM found that these particles were predomi-
nantly in the PM10–PM2.5fraction (mass median aerody-
namic diameter ? 6.8 ?m). Although EDX analysis does
Figure 3.
PM10–PM2.5mass by location.
Mass contribution (%) of sources to (a) PM2.5and (b)
Zota et al.
Volume 59 November 2009
Journal of the Air & Waste Management Association 1353

Page 8

not definitively identify specific minerals, the analysis
showed that Zn- and Pb-rich particles were observed in
two major varieties. The first major type consisted of
heterogeneous particles in which Zn- and Pb-rich particles
were co-present with S, consistent with metal sulfide min-
erals which have been identified in mine waste.4,6,7Iso-
lated particles of ZnS or PbS were rarely observed. Rather,
these particles were frequently attached to PM10–PM2.5
silicate or aluminosilicate particles. Figure 5 shows SEM
micrographs of a typical particle of this type. The second-
ary electron image (Figure 5a) highlights surface morphol-
ogy, whereas the backscattered electron (BSE) image (Fig-
ure 5b) reveals chemical heterogeneity within the
particle. Bright areas in the BSE image are features associ-
ated with high average atomic number and were consis-
tent with the mineral ZnS, as seen in the upper EDX
spectrum (Figure 5c). These ZnS features appear to be
adsorbed onto or aggregated with an aluminosilicate ma-
trix (Figure 5d), the EDX spectrum of which is largely
consistent with the PMF-deduced crustal/soil factor. The
second type of frequently occurring particle were Zn- or
Pb-rich silicate particles, which may be indicative of
hemimorphite, a primary and secondary mineral that
forms in oxic, Zn-rich conditions and has been identified
in chat particles less than 37 ?m.7
DISCUSSION
To the authors’ knowledge, this study is the first to use
source apportionment techniques to quantify the impact
of mine waste on airborne respirable particulates in resi-
dential areas close to an abandoned mining site. PMF
results for PM2.5and PM10–PM2.5particulates isolated a
unique factor whose chemical profile was similar to that
of particles from chat piles. With a composition similar to
that of chat, the PMF-deduced mine waste factor was
dominated by Pb and Zn and included traces of crustal
elements (e.g., Al, Si, and Ca). The percent mass contri-
bution of Zn and Pb, and the Zn-to-Pb ratio, were also in
Figure 4. Scatterplots of PMF estimated PM10–PM2.5mine waste
concentrations vs. PMF-estimated PM10–PM2.5Ca-rich concentrations
by site location. Regression lines represent the slopes of the associa-
tion for the three sites and are statistically significant for Quapaw (P ?
0.0001) and Miami (P ? 0.0001) but not Picher (P ? 0.12).
Table 3. Multivariate regression models of PM2.5and PM10-PM2.5mine waste attributable mass (?g/m3).
Mass Attributed to Mine Waste Source
PM2.5Fractiona
PM10–PM2.5Fractiona
?(SE)P value
?(SE)P value
Coarse fraction mine waste mass (?g/m3)a
Coarse fraction Ca-rich mass (?g/m3)a
Siteb
Miami
Quapaw
Picher
Seasonb
Winter
Spring
Summer
Fall
Wind directionb
West
South
East
Coarse Ca-rich mass site interactionc
Coarse Ca-rich Miamic
Coarse Ca-rich Quapawc
Coarse Ca-rich Picherc
Model R2
0.42 (0.07)
–
?0.0001
–
?0.0001
––
1.01 (0.14)
?0.0001
?0.0001
00
0.07 (0.17)
0.83 (0.22)
1.30 (0.24)
3.33 (0.22)
?0.0001 0.0034
00
0.56 (0.18)
0.13 (0.17)
0.83 (0.17)
0.34 (0.16)
0.77 (0.15)
0.17 (0.15)
–
–
–
–
0.006–
–
–
–
0
0.70 (0.21)
0.90 (0.42)
–
–
–
–
0.64
–
–
–
–
?0.0001
0
?0.14
?0.75
0.79
(0.16)
(0.16)
Notes:aLog-transformed;bReference group in parentheses,cP ? 0.05.
Zota et al.
1354 Journal of the Air & Waste Management Association Volume 59 November 2009

Page 9

good agreement with measurements made on “parent”
chat.7Additionally, scanning electron microscopy analy-
sis of a filter with a high PMF-estimated mine waste
concentration qualitatively confirmed that the PMF-
identified mine waste source corresponded to particles of
mine waste origin.
Source apportionment analysis revealed that the
mine waste source predominantly impacted particles in
the PM10–PM2.5fraction. Pb and Zn concentrations as
well as estimated mine waste concentrations were all
higher in PM10–PM2.5particles in comparison to PM2.5.
Mine waste source contributions were most substantial in
Picher, where mine waste was the largest source of PM10–
PM2.5mass, contributing an average of 4.5 ?g/m3(?40%
of total mass). PM10–PM2.5fraction impacts were also
observed at sites further away from the chat piles partic-
ularly in Quapaw, where average contributions exceeded
10%. Particles of crustal origin and a source rich in Ca
were major contributors of PM10–PM2.5mass in Miami
and Quapaw.
Although a mine waste source was identified in the
PMF solution for PM2.5, its contributions to PM2.5mass
were modest even in the source-dominated site of Picher.
Our results are consistent with a previous study in South
Africa, which found small and geographically limited im-
pacts of mine tailings on atmospheric lead.34Pb concen-
trations in our PM2.5samples were generally low and
below the National Ambient Air Quality Standard (0.15
?g/m3). Zn concentrations were higher than those typi-
cally found in rural areas and similar to urban areas35
where respiratory effects (e.g., asthma) have been associ-
ated with Zn PM2.5levels.36Concentrations of Cd, also
elevated in mine waste, were all below the LOD in our
samples. In contrast, a recent study found increasing
metal concentrations with decreasing particle size down
to 1 ?m in “parent” chat from the Tar Creek site.7Our
results suggest that although PM2.5chat particles may be
highly enriched in metals, they are low in abundance or
less likely to be windborne because of the presence of
larger particles on the surface of the chat piles.
Similar to other source apportionment studies, sec-
ondary sulfates from coal combustion and mobile sources
were the largest contributors to PM2.5mass.33Model di-
agnostics suggest that the factors resolved in the PM2.5
data were less stable and more uncertain than those char-
acterized in the PM10–PM2.5data. PM2.5source apportion-
ment results may have overestimated contributions for
mobile source combustion and omitted other minor
sources such as vegetative burning. The analysis of addi-
tional carbon fractions and metals ions could have as-
sisted in the further identification of sources and im-
proved the model fit.
Although mine waste impacts were isolated using
EPA PMF source apportionment, it was difficult to deter-
mine the relative importance of various sources and trans-
port mechanisms because of the dispersed, heterogeneous
nature of contemporary mine waste. Wind erosion of dust
particles from chat piles, mechanically generated chat-
laden dust from paved and unpaved roads, and the ongo-
ing removal and processing of chat for construction
Figure 5. Results of CCSEM analysis from a PM10particle filter identified as having a high factor score for the mine waste source in PMF
models: (a) secondary electron image, (b) BSE image, (c) EDX spectrum of ZnS inclusion in top of particle, and (d) EDX spectrum of
aluminosilicate matrix at bottom of particle.
Zota et al.
Volume 59 November 2009
Journal of the Air & Waste Management Association 1355

Page 10

projects may all contribute to the mine waste particles
observed in these air samples. For example, CCSEM/EDX
analyses documented two different types of particles that
appeared to be of mine waste origin.
Additionally, linear regression results, which in-
cluded a statistically significant interaction term between
sampling site and Ca-rich concentrations in the PM10–
PM2.5fraction, may suggest differential transport mecha-
nisms by location. Mine waste estimates increased lin-
early with Ca-rich source estimates in Quapaw and Miami
but not in Picher. One possible interpretation of this
interaction is that fugitive dust from the chat piles may be
the predominant source of airborne mine waste in Picher.
Conversely, resuspension of chat, previously deposited on
paved and unpaved roads, may be a more important
transport mechanism at the Quapaw and Miami sites,
which are farther away from the chat piles. Another pos-
sible explanation for this observed interaction is that the
Ca-rich source concentrations are a surrogate for some
other unmeasured environmental factor or anthropo-
genic activity that varies between Picher and the other
sites and is positively associated with mine waste concen-
trations. Future studies should compare the geochemical
properties and particle size distribution profiles of wind-
blown mine waste versus mechanically resuspended mine
waste at relevant mining sites.
This study has several important strengths. The
unique study design and collaboration allowed us to eval-
uate chronic airborne exposures to mining-related metals
in an underserved community. Although abandoned
hardrock mines and associated solid mine waste are in-
creasing in a global context, limited data are available on
the ambient air quality at these sites. Various methods
were used, including receptor modeling and scanning
electron microscopy, to examine mine waste impacts on
ambient PM. Another major strength of the study is the
unique collaboration, which included university, com-
munity, and government partners. Most notably, the Lo-
cal Environmental Action Demanded (L.E.A.D.) agency, a
community partner in this study, successfully conducted
field sampling from using strict quality assurance/quality
control guidelines for over 1 yr. This approach should be
a model for other community-based assessments of air
quality.
There were also some weaknesses to this study.
Although seasonal variations in mine waste concentra-
tions were observed using regression analysis, weekly
integrated samples obscured the ability to assess acute
exposures, reconstruct backward trajectories, and ana-
lyze relationships with wind speed. However, the
longer sampling time allowed for increased particle
mass deposition and thus improved elemental detection
limits. Coarse fraction concentrations were calculated in-
directly by subtracting PM10and PM2.5measurements,
which may increase measurement error. However, the
measurement error in this approach did not overwhelm
the data given the strong model diagnostics for the PMF
PM10–PM2.5fraction results—including excellent replica-
tion of the base factors in the bootstrapping simulations.
Lastly, the scanning electron microscopy methods could
only be used in a qualitative manner. Future air pollution
studies should expand the quantitative use of geochemi-
cal techniques in source apportionment studies and com-
pare source apportionment estimates for mine waste in
respirable particles to those from fugitive dust models.
CONCLUSIONS
This study estimated the contribution of mine waste, an
important local pollution source, to airborne respirable
particles using source apportionment techniques. Mass
contribution of mine waste particulates was more appar-
ent in the PM10–PM2.5fraction, with average contribu-
tions 3-fold greater in the PM10–PM2.5particles compared
with the PM2.5fraction (20% vs. 6%). There were also
large differences observed in mine waste contribution by
location. Impacts were largest at the source-dominated
site and decreased with increasing distance from the chat
piles, suggesting that populations living nearest to the
chat piles may be more highly exposed. In conclusion,
this study characterized temporal and spatial variability of
metal concentrations and potential sources in two size
fractions of respirable, ambient particles. Future studies
from this center will examine the impact of mining-
related sources in the indoor environment and relation-
ships to biological measures in children.
ACKNOWLEDGMENTS
The authors acknowledge the active involvement and
continued cooperation of our community partners—the
L.E.A.D. agency and Integris Baptist Medical Center of
Miami, OK. The authors thank the residents of Ottawa
County, who were instrumental in the collection of this
valuable information, for welcoming us onto their land
and allowing us to assemble and maintain our air sam-
pling stations for over a year. The authors thank Rebecca
Lincoln for her assistance with sampling and manuscript
feedback, Meredith Franklin for her programming assis-
tance, and Beatriz Vinas for her laboratory assistance.
This study was supported by the U.S. National Institute
of Environmental Health Sciences (NIEHS), National
Institutes of Health (NIH) grant P42-ES05947, Superfund
Basic Research Program, NIEHS Center grant 2 P30-ES
00002, NIEHS Children’s Center grant 1P01ES012874, and
EPA STAR research assistance agreement no. RD-83172501.
The contents of this paper are solely the responsibility of
the authors and do not necessarily represent the official
views of the NIEHS, NIH, or EPA. EPA, through its Office
of Research and Development, collaborated in the re-
search described here, which has been subjected to EPA
review and approved for publication.
REFERENCES
1. EPA’s National Hardrock Mining Framework; U.S. Environmental Protec-
tion Agency; Office of Solid Waste and Emergency Response; Office of
Enforcement and Compliance Assurance; Office of Water: Washing-
ton, DC, 1997.
2. Hardrock Mining on Federal Lands; National Research Council; National
Academy Press: Washington, DC, 1999.
3. Fields, S. The Earth’s open wounds: Abandoned and orphaned mines;
Environ. Health Perspect. 2003, 111, A154-A161.
4. McKnight, E.T.; Fischer, R.P. Geology and Ore Deposits of the Picher Field,
Oklahoma and Kansas; U.S. Government Printing Office: Washington,
DC, 1970.
5. Luza, K.V. Stability Problems Associated with Abandoned Underground
Mines in the Picher Field Northeastern Oklahoma; Oklahoma Geological
Survey: Norman, OK, 1986.
Zota et al.
1356 Journal of the Air & Waste Management AssociationVolume 59 November 2009

Page 11

6. Datin, D.L.; Cates, D.A. Sampling and Metal Analysis of Chat Piles in the
Tar Creek Superfund Site; Prepared by the Tar Creek Superfund Site,
Ottawa County, OK, for the Oklahoma Department of Environmental
Quality: Oklahoma City, OK, 2002.
7. Schaider, L.A.; Senn, D.B.; Brabander, D.J.; McCarthy, K.D.; Shine, J.P.
Characterization of Zinc, Lead, and Cadmium in Mine Waste: Impli-
cations for Transport, Exposure, and Bioavailability; Environ. Sci. Tech-
nol. 2007, 41, 4164-4171.
8. Five Year Review Tar Creek Superfund Site, Ottawa County, Oklahoma; U.S.
Environmental Protection Agency, Region 6: Washington, DC, 2000.
9. Dockery, D.W.; Pope, C.A., III; Xu, X.; Spengler, J.D.; Ware, J.H.; Fay,
M.E.; Ferris, B.G., Jr.; Speizer, F.E. An Association between Air Pollu-
tion and Mortality in Six U.S. Cities; N. Engl. J. Med. 1993, 329,
1753-1759.
10. Krewski, D.; Burnett, R.; Jerrett, M.; Pope, C.A.; Rainham, D.; Calle, E.;
Thurston, G.; Thun, M. Mortality and Long-Term Exposure to Ambi-
ent Air Pollution: Ongoing Analyses Based on the American Cancer
Society Cohort; J. Toxicol. Environ. Health A 2005, 68, 1093-1109.
11. Brunekreef, B.; Forsberg, B. Epidemiological Evidence of Effects of
Coarse Airborne Particles on Health; Eur. Respir. J. 2005, 26, 309-318.
12. Schwarze, P.E.; Ovrevik, J.; Lag, M.; Refsnes, M.; Nafstad, P.; Hetland,
R.B.; Dybing, E. Particulate Matter Properties and Health Effects: Con-
sistency of Epidemiological and Toxicological Studies; Hum. Exp. Toxi-
col. 2006, 25, 559-579.
13. Ghio, A.J. Biological Effects of Utah Valley Ambient Air Particles in
Humans: a Review; J. Aerosol Med. 2004, 17, 157-164.
14. Kodavanti, U.P.; Schladweiler, M.C.; Gilmour, P.S.; Wallenborn, J.G.;
Mandavilli, B.S.; Ledbetter, A.D.; Christiani, D.C.; Runge, M.S.; Karoly,
E.D.; Costa, D.L.; Peddada, S.; Jaskot, R.; Richards, J.H.; Thomas, R.;
Madamanchi, N.R.; Nyska, A. The Role of Particulate Matter-Associ-
ated Zinc in Cardiac Injury in Rats; Environ. Health Perspect. 2008, 116,
13-20.
15. Lippmann, M.; Ito, K.; Hwang, J.S.; Maciejczyk, P.; Chen, L.C. Cardio-
vascular Effects of Nickel in Ambient Air; Environ. Health Perspect.
2006, 114, 1662-1669.
16. Sunderman, F.W., Jr. Nasal Toxicity, Carcinogenicity, and Olfactory
Uptake of Metals; Ann. Clin. Lab. Sci. 2001, 31, 3-24.
17. Tjalve, H.; Henriksson, J. Uptake of Metals in the Brain via Olfactory
Pathways; Neurotoxicology 1999, 20, 181-195.
18. Thompson, K.; Molina, R.M.; Donaghey, T.; Schwob, J.E.; Brain, J.D.;
Wessling-Resnick, M. Olfactory Uptake of Manganese Requires DMT1
and is Enhanced by Anemia; FASEB J. 2007, 21, 223-230.
19. McDonald, J.D.; Zielinska, B.; Sagebiel, J.C.; McDaniel, M.R.; Mousset-
Jones, P. Source Apportionment of Airborne Fine Particulate Matter in
an Underground Mine; J. Air & Waste Manage. Assoc. 2003, 53, 386-
395.
20. Williamson, B.J.; Udachin, V.; Purvis, O.W.; Spiro, B.; Cressey, G.;
Jones, G.C. Characterisation of Airborne Particulate Pollution in the
Cu Smelter and Former Mining Town of Karabash, South Ural Moun-
tains of Russia; Environ. Monitor. Assess. 2004, 98, 235-259.
21. The Climate of Ottawa County; University of Oklahoma; Oklahoma
Climatological Survey: Norman, OK, 2007.
22. Jim, R. Local Environmental Action Demanded, Vinita, OK, Personal
communications, 2005.
23. Marple, V.A.; Rubow, K.L.; Turner, W.; Spengler, J.D. Low Flow-Rate
Sharp Cut Impactors for Indoor Air Sampling—Design and Calibra-
tion; JAPCA 1987, 37, 1303-1307.
24. Dzubay, T.G.; Rickel, D.G. In Electron Microscopy and X-Ray Applications
to Environmental and Occupational Health Analysis; Russell, P.A., Hutch-
ings, A.E., Eds.; Ann Arbor Science: Ann Arbor, MI, 1978.
25. Compendium Method IO-3.3: Determination of Metals in Ambient Partic-
ulate Matter Using X-Ray Fluorescence (XRF) Spectroscopy; U.S. Environ-
mental Protection Agency; Center for Environmental Research Infor-
mation; Office of Research and Development: Cincinnati, OH, 1999.
26. Kinney, P.L.; Aggarwal, M.; Northridge, M.E.; Janssen, N.A.; Shepard,
P. Airborne Concentrations of PM(2.5) and Diesel Exhaust Particles on
Harlem Sidewalks: a Community-Based Pilot Study; Environ. Health
Perspect. 2000, 108, 213-218.
27. Morphy, T. Magee Scientific, Berkeley, CA, Personal communication,
2008.
28. Mamane, Y.; Willis, R.; Conner, T. Evaluation of Computer-Controlled
Scanning Electron Microscopy Applied to an Ambient Urban Aerosol
Sample; Aerosol Sci. Technol. 2001, 34, 97-107.
29. EPA Positive Matrix Factorization (PMF) 3.0 Fundamentals and User Guide;
U.S. Environmental Protection Agency; National Exposure Research
Laboratory; Office of Research and Development: Research Triangle
Park, NC, 2008.
30. Reff, A.; Eberly, S.I.; Bhave, P.V. Receptor Modeling of Ambient Par-
ticulate Matter Data Using Positive Matrix Factorization: Review of
Existing Methods; J. Air & Waste Manage. Assoc. 2007, 57, 146-154.
31. Chow, J.C.; Watson, J.G.; Kuhns, H.; Etyemezian, V.; Lowenthal, D.H.;
Crow, D.; Kohl, S.D.; Engelbrecht, J.P.; Green, M.C. Source Profiles for
Industrial, Mobile, and Area Sources in the Big Bend Regional Aerosol
Visibility and Observational Study; Chemosphere 2004, 54, 185-208.
32. Thurston, G.D.; Spengler, J.D. A Quantitative Assessment of Source
Contributions to Inhalable Particulate Matter Pollution in Metropol-
itan Boston; Atmos. Environ. 1985, 19, 9-25.
33. Hopke, P.K.; Ito, K.; Mar, T.; Christensen, W.F.; Eatough, D.J.; Henry,
R.C.; Kim, E.; Laden, F.; Lall, R.; Larson, T.V.; Liu, H.; Neas, L.; Pinto,
J.; Stolzel, M.; Suh, H.; Paatero, P.; Thurston, G.D. PM Source Appor-
tionment and Health Effects: 1. Intercomparison of Source Apportion-
ment Results; J. Expo. Sci. Environ. Epidemiol. 2006, 16, 275-286.
34. Monna, F.; Poujol, M.; Losno, R.; Dominik, J.; Annegarn, H.; Coetzee,
H. Origin of Atmospheric Lead in Johannesburg, South Africa; Atmos.
Environ. 2006, 40, 6554-6566.
35. Toxicological Profile for Zinc; Agency for Toxic Substances and Disease
Registry: Atlanta, GA, 2005.
36. Hirshon, J.M.; Shardell, M.; Alles, S.; Powell, J.L.; Squibb, K.; Ondov, J.;
Blaisdell, C.J. Elevated Ambient Air Zinc Increases Pediatric Asthma
Morbidity. Environ. Health Perspect. 2008, 116, 826-831.
About the Authors
During this study, Dr. Ami R. Zota was a postdoctoral
research fellow in the Department of Environmental Health
at the Harvard School of Public Health in Boston, MA, and
the Silent Spring Institute in Newton, MA. Dr. Robert Willis
is a research physicist in the Environmental Characteriza-
tion and Apportionment Branch of EPA NERL in Research
Triangle Park, NC. Dr. Gary A. Norris is a research physical
scientist in the Environmental Characterization and Appor-
tionment Branch of EPA NERL. Dr. James P. Shines is an
associate professor in the Department of Environmental
Health at the Harvard School of Public Health. Ms. Rebecca
Jim heads L.E.A.D., a community-based nonprofit organi-
zation in Miami, OK. Dr. Rachelle M. Duvall is a research
physical scientist in the Environmental Characterization and
Apportionment Branch of EPA NERL. Dr. Laurel A. Schaider
is a research associate in the Department of Environmental
Health at the Harvard School of Public Health. Dr. John D.
Spengler is the Akira Yamaguchi Professor of Environmen-
tal Health in the Department of Environmental Health at the
Harvard School of Public Health. Please address corre-
spondence to: Ami R. Zota, Program on Reproductive
Health and the Environment, Department of Obstetrics,
Gynecology, and Reproductive Sciences, University of Cal-
ifornia–San Francisco, 1330 Broadway Street, Suite 1100,
Oakland, CA 94612; phone: ?1-510-986-8928; e-mail:
zotaar@obgyn.ucsf.edu.
Zota et al.
Volume 59 November 2009
Journal of the Air & Waste Management Association 1357