4% United States
Environmental Protection
mm. Agency
Pilot Study on Coarse PM Monitoring
-------
This document contains blank pages to accommodate two-sided printing.
-------
EPA-454/R-15-001
February 2015
Pilot Study on Coarse PM Monitoring
Prepared by:
Jay R. Turner
Department of Energy, Environmental and Chemical Engineering
Washington University at St. Louis
1 Brookings Drive
St. Louis, MO 63130
Steven G. Brown
Hilary A. Minor
Sonoma Technology, Inc.
1455 N. McDowell Blvd., Suite D
Petaluma, CA 94954-6503
Prepared for:
Joann Rice
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Contract No. EP-D-09-097
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC 27711
-------
EPA's Coarse PM Pilot Study
Acknowledgments
Acknowledgments
RTI performed laboratory analyses, overseen by James Flanagan with contributions by
Karin Foarde and Frank Weber, and provided logistical support on instrumentation and
maintenance, led by Jeff Nichol. Washington University in St. Louis students and field operators
conducted the measurements in St. Louis. Maricopa County staff, led by Ben Davis, conducted
the measurements in Phoenix. At Washington University, Varun Yadav led the field operations
and Li Du led the analysis of St. Louis and Phoenix TEOM data. At Sonoma Technology, Inc.,
Adam Pasch and Theresa O'Brien assisted in data handling and quality control, Paul Roberts
was the Principal Investigator, and Mary Jo Teplitz, Ron Teplitz, and Marina Michaels provided
editorial support.
Disclosure: The St. Louis field operations were funded by EPA and were conducted by
students and staff of the Jay Turner group at Washington University in St. Louis. Jay Turner
was also a consultant to Sonoma Technology, Inc., on this project.
-------
EPA's Coarse PM Pilot Study
Table of Contents
Table of Contents
Section Page
Acknowledgments ii
List of Figures v
List of Tables viii
Executive Summary: Key Findings and Recommendations ES-1
1. Introduction 1-1
1.1 Study Objectives 1-2
1.2 Study Design 1-3
1.3 Technical Approach 1-6
1.3.1 Instrumentation 1-7
1.3.2 Sample Analysis 1-7
1.3.3 Nomenclature and Dichot Sampling Concentration Equations 1-7
1.4 Field Operations Overview 1-9
1.4.1 Methods 1-9
1.4.2 Field Operations Summary 1-9
1.5 Guide to This Report 1-11
2. Gravimetric Mass and Composition Overview 2-1
2.1 Gravimetric Mass Comparisons Between Dichot and FRM Samplers 2-1
2.2 Distribution of Species Between Fine and Coarse Mode 2-4
2.3 Collocated Precision of Dichot Measurements 2-5
3. Analysis of Elements 3-1
3.1 Concentrations by Site and Size Fraction 3-1
3.2 Concentrations of Elements: Comparison of Dichot and FRM 3-2
3.3 Attenuation of X-Ray Intensity for Light Elements 3-6
3.4 XRF versus ICP-MS Measurements and Implications 3-10
3.5 Variation in Crustal Composition 3-16
4. Analysis of Carbon 4-1
4.1 Total Carbon Comparisons Between Dichot and FRM Samplers 4-1
4.2 Thermal Fraction Analysis (OC/EC Split) 4-3
4.3 Carbonate Concentrations 4-4
4.4 Carbon Artifacts 4-7
4.4.1 OC Trip Blanks and Field Blanks 4-8
4.4.2 OC Mass Loadings on the Backup Filters 4-8
4.4.3 Comparison of OC Blanks, Front Filters, and Backup Filters 4-10
4.4.4 Summary of Carbon Artifacts 4-13
4.5 Biological Data 4-13
5. Analysis of Ions 5-1
5.1 Approaches to Measuring Ions 5-1
5.2 Nitrate Concentrations on Teflon and Backup Nylon Filter 5-2
5.3 Nitrate Partitioning Between Fine and Coarse Modes 5-2
5.4 Ammonium Balance and Implications for Ammonium Measurements 5-3
5.5 Comparison of Concentrations With and Without Denuders 5-7
5.6 Comparison of Ion and Corresponding Element Concentrations 5-9
-------
EPA's Coarse PM Pilot Study
Table of Contents
Section Page
6. Mass Balance 6-1
6.1 Results for Dichotomous Samplers 6-1
6.2 Mass Balance Implications 6-5
7. Comparison of FDMS TEOM and Filter Coarse PM Measurements 7-1
7.1 Results 7-2
7.2 FDMS TEOM-to-Dichot Summary 7-9
8. References 8-1
Appendix A: Summary Statistics A-1
Appendix B: Summary Ratios of Collocated Dichot Measurements B-1
Appendix C: Summary of Dichot-to-FRM Comparisons C-1
Appendix D: Quartz Fiber Filter Carbon Blanks D-1
Appendix E: Nitrate Correlations with Other Species E-1
iv
-------
EPA's Coarse PM Pilot Study
List of Figures
List of Figures
1-1. Site maps for W 43rd Ave, Phoenix monitoring site 1-4
1-2. Site maps for 13th and Tudor, East St. Louis monitoring site 1-5
1-3. Four-month sampling schedule for collocated dichots 1-6
2-1. Dichot versus FRM mass concentrations for St. Louis and Phoenix for PM2.5, PM10-
2.5, and PM10 2-2
2-2. Stacked bar plots of average composition for (a) PMC at Phoenix; (b) PMf at
Phoenix; (c) PMC at St. Louis; and (d) PMf at St. Louis 2-4
2-3. Average FRM PM10/PM2.5 ratio (log scale) by species for each site 2-5
2-4. Collocated dichot versus dichot plots for (a) gravimetric mass at Phoenix;
(b) gravimetric mass at St. Louis; (c) Si at Phoenix; (d) Si at St. Louis 2-6
2-5. Scatterplot matrices of the observed differences, a* = in(CxB/cxA), in the
collocated concentrations at Phoenix: (a) PMC and (b) PMf 2-10
2-6. Scatterplot vector of the observed differences, a* = in(Cx B/cxA), across the fine
and coarse PM size fractions in the collocated concentrations at Phoenix 2-11
2-7. Average ratio of collocated dichot measurements (log scale) by site, species, and
size, using only data above detection limits for both measurements 2-11
3-1. Percentage of samples above detection limit for dichot PMC by site 3-2
3-2. Scatter plots of PMf and PMC concentrations from dichot and FRM at Phoenix and
St. Louis for S, Zn, Ca, and Fe 3-5
3-3. Cumulative distribution of gravimetric mass concentrations for all dichot sampler
minor flow channel (PMC) samples that were analyzed by XRF and the six samples
with mass loadings reported by RTI both with and without attenuation correction 3-8
3-4. Attenuation factors for dichot minor flow channel samples, PM10 FRM samples,
and dichot major flow channel samples 3-9
3-5. Estimated crustal PMC mass concentrations with and without corrections for self-
attenuation applied to the dichot sampler data 3-10
3-6. PHX dichot PMC aluminum concentrations by XRF (A = 0.51) and ICP-MS 3-11
3-7. PHX dichot PMC aluminum concentrations by ICP-MS with blank correction (mean
laboratory blank of 1.092 |jg/m3 from Table 3-6) and assuming 100% recovery,
and by XRF with (a) XRF attenuation factor of unity; and (b) best-fit attenuation
factor to reconcile the XRF and ICP-MS data 3-15
v
-------
EPA's Coarse PM Pilot Study
List of Figures
Figure Page
*atf
3-8. Simple visualization of major crustal elements: fraction of the sum of Si, Fe, and
Ca by element for primary dichot PMC at (a) Phoenix (N = 51) and (b) St. Louis
(N = 31) 3-17
4-1. Dichot versus FRM total carbon concentrations for St. Louis and Phoenix for PM2.5,
PM10-2.5, and PM10 4-2
4-2. Carbonaceous PM average concentration values at Phoenix (N = 30) and St.
Louis (N = 22) 4-3
4-3. Distribution of organic carbon average concentration values at Phoenix (N = 30)
and St. Louis (N = 22) 4-4
4-4. PMC carbonate from (a) collocated dichots and (b) mean collocated dichot versus
FRM-by-difference 4-5
4-5. Relationship between dichot PMC carbonate and dichot PMC calcium expressed as
molar concentrations in (a) Phoenix and (b) St. Louis 4-6
4-6. OC mass loadings on backup filters: (a) PM2.5 FRM versus PM10 FRM; (b) dichot
major flow versus PM2.5 FRM; and (c) dichot minor flow versus dichot major flow 4-9
4-7. OC mass loading distributions (|jg/filter) for the dichotomous samples 4-11
4-8. Dichotomous sampler OC mass loadings (|jg/filter) for paired front and back filters:
(a) major flow and (b) minor flow 4-12
4-9. OC concentrations for collocated dichot samplers: (a) PM2.5 and (b) PMC 4-13
4-10. Box-whisker plots by site of (a) glucan (ng/m3) and (b) endotoxin (EU/m3) 4-15
5-1. Scatter plots for dichotomous sampler nitrate showing (a) the partitioning of nitrate
between the fine and coarse PM fractions and (b) total fine nitrate and the nitrate
on the front filter 5-3
5-2. Measured ammonium versus calculated ammonium scatter plots at Phoenix and
St. Louis on Teflon and nylon (right four plots) 5-5
5-3. Collocated dichot (dichot with denuder versus dichot without denuder) scatter plots
for (a) NH4, (b) NO3, (c) SO4, and (d) Ca, Fe, and Si at Phoenix 5-8
5-4 Comparison of ion and element concentrations (|jg/m3) at St. Louis for K and K+,
Na and Na+, and S and SO42" via dichot and FRM samplers for PMC and PMf 5-10
5-5. Comparison of ion and element concentrations (|jg/m3) at Phoenix for K and K+,
Na and Na+, and S and SO42" via dichot and FRM samplers for PMC and PMf 5-11
6-1. Mass balance closure between Teflon filter gravimetric mass and the chemical
speciation data for dichot samples 6-2
-------
EPA's Coarse PM Pilot Study
List of Figures
Figure Page
*atf
6-2. Reconstructed mass versus gravimetric mass for the dichotomous sampler Teflon
filters: (a) PMf and (b) PMC for Phoenix and St. Louis 6-3
6-3. RMS residuals between reconstructed mass and gravimetric mass as a function of
the assumed OM/OC ratio: (a) Phoenix and (b) St. Louis 6-4
7-1. East St. Louis and Phoenix 24-hour average dichot TEOM total and nonvolatile
PM mass concentrations versus the 24-hour integrated PM mass concentrations
from filter-based samplers with gravimetric analysis on the Teflon filters: fine PM
and coarse PM 7-3
7-2. Daily-average dichot TEOM time series for fine PM and coarse PM at (a) East St.
Louis, 9/2010-2/2011; and (b) Phoenix, 7/2010-2/2011 7-4
7-3. Daily-average dichot TEOM time series for fine PM and coarse PM at East St.
Louis, 9/2010-2/2011, and Phoenix, 7/2010-2/2011 7-5
7-4. PM2.5 diurnal profiles for East St. Louis, 9/2010 to 2/2011 and Phoenix, 7/2010 to
2/2011: (a) total TEOM mass; (b) nonvolatile TEOM mass; and (c) volatile TEOM
mass 7-6
7-5. PM10-2.5 diurnal profiles for East St. Louis, 9/2010 to 2/2011 and Phoenix, 7/2010 to
2/2011: (a) total TEOM mass and (b) volatile TEOM mass 7-7
7-6. PM10-2.5 weekday and weekend diurnal profiles for East St. Louis, 9/2010 to 2/2011
and Phoenix, 7/2010 to 2/2011 7-8
7-7. Diurnal profiles for the difference in weekday and weekend hourly median PM10-2.5
concentrations for East St. Louis, 9/2010 to 2/2011 and Phoenix, 7/2010 to 2/2011 ....7-8
-------
EPA's Coarse PM Pilot Study
List of Tables
List of Tables
1-1. Definitions of the different types of PM referenced in this document 1-7
1-2. Sample completeness by site and sampler type for operations from June 1, 2010,
through May 31, 2011, excluding sampling days for biological aerosols analysis 1-10
2-1. Gravimetric mass comparisons between the dichot and FRM samplers 2-3
2-2. Measures of agreement between the collocated dichot sampler data for PMC
measured at Phoenix 2-7
2-3. Measures of agreement between the collocated dichot sampler data for PMf
measured at Phoenix 2-7
2-4. Measures of agreement between the collocated dichot sampler data for PMC
measured at St. Louis 2-8
2-5. Measures of agreement between the collocated dichot sampler data for PMf
measured at St. Louis 2-8
3-1. Measures of agreement between the collocated FRM and dichot data for PMC
measured at Phoenix 3-3
3-2. Measures of agreement between the collocated FRM and dichot data for PMf
measured at Phoenix 3-3
3-3. Measures of agreement between the collocated FRM and dichot data for PMC
measured at St. Louis 3-4
3-4. Measures of agreement between the collocated FRM and dichot data for PMf
measured at St. Louis 3-4
3-5. Average attenuation factors (A) back-calculated by RTI for a subset of samples 3-8
3-6. ICP-MS analysis of laboratory blanks with the membrane separated from the
support ring (with adhesive) and of field blanks 3-12
3-7. Recoveries for urban particle matter NIST standard reference material for the
sample digestion protocol and ICP-MS analysis used in this study 3-13
3-8. Attenuation factors for dichot PMC aluminum at PHX 3-15
4-1. Mean mass loadings on the field blanks and dichot backup filters 4-12
4-2. Biomarker field and trip blank data summary 4-14
4-3. Biomarker data summary for blank-corrected data 4-14
VIII
-------
EPA's Coarse PM Pilot Study
List of Tables
Table Page
5-1. Summary of nitrate concentrations via dichot on Teflon, nylon, and total (Teflon +
nylon), plus fraction of total nitrate on Teflon filter 5-2
6-1. Measures of agreement between reconstructed and gravimetric mass for Phoenix
and St. Louis, using OM = 1.6* OC and no correction for OC artifacts 6-3
ix
-------
-------
EPA's Coarse PM Pilot Study
Executive Summary
Executive Nummary; Key Findings and Recv n nendationr*
The U.S. Environmental Protection Agency (EPA) conducted this field study in 2010 and
2011 to evaluate the challenges in sampling and analyzing coarse aerosol, the precision of
coarse PM (PMC) mass species measurements using dichotomous (dichot) samplers, and mass
balance of PMC. The study database is publicly available through the EPA Air Quality System
(AQS) to EPA personnel, atmospheric scientists, and others concerned with the science of PM
air pollution, related health effects, and human exposure to the coarse PM fraction of particulate
matter. Additional samplers—including paired PM10 and PM2.5 Federal Reference Method (FRM)
samplers to calculate PM10-2.5 mass and species concentrations by the difference method, and
semi-continuous monitors—were operated to further characterize coarse PM and aid in the
interpretation of any differences between dichot data and difference method data. The results of
this study may be used to establish routine field operating procedures and laboratory standard
operating procedures (SOPs) for use in PMC speciation monitoring.
ES-1. Primary Objectives
The primary objectives of the coarse PM pilot speciation study were to:
1. develop the target species analyte list for routine speciation monitoring (what species
need to be measured);
2. evaluate and define appropriate analysis methods for routine speciation monitoring and
the necessary SOPs;
3. evaluate the field performance of the dichot samplers for routine speciation monitoring
(e.g., comparing gravimetric mass and speciation to the FRM by difference data and
assessing dichot collocated precision);
4. learn about sampling and operational issues regarding the use of dichots; and
5. evaluate data from the study to inform several issues related to coarse PM speciation
measurements.
ES-2. Study Methods
The coarse PM pilot speciation study included one year of 1 -in-3 day sampling at sites in
Phoenix (Arizona) and East St. Louis (Illinois), from June 2010 through May 2011. At each site,
two Thermo 2025D sequential dichot samplers, one Thermo 2025 sequential PM10 FRM
sampler, one Thermo 2025 sequential PM2.5 FRM sampler, and one Thermo 1405-DF Filter
Dynamics Measurement Systems (FDMS) dichotomous Tapered Element Oscillating
Microbalance (TEOM) monitor were used to make routine measurements. Samples were
collected for laboratory analysis using Teflon®/nylon (T/N) and quartz/quartz (Q/Q) filter
sandwiches.
Analytical methods adopted from the PM2.5 Chemical Speciation Network (CSN) were
used to characterize fine and coarse particle speciation for about half of the sampling events.
The rest were archived for further study if needed. The analytical methods included gravimetric
mass, elements by x-ray fluorescence (XRF), ions by ion chromatography (IC) from the Teflon
ES-1
-------
EPA's Coarse PM Pilot Study
Executive Summary
filter, and organic carbon (OC) and elemental carbon (EC) by thermal-optical analysis (TOA)
from the quartz filter. Subsets of samples were analyzed for elements by inductively coupled
plasma mass spectrometry (ICP-MS) and were analyzed for carbonate by TOA with sample
acidification. Dichot PMC and PMfine (PMf) concentrations were adjusted to correct for 10% PMf
intrusion into the PMC channel; results in this report incorporate this correction. Coarse PM was
thus measured directly both via the dichot (PMC) and via the difference between FRM PM10 and
PM2.5 measurements (i.e., PM10-2.5).
ES-3. Key Findings
The key study findings were as follows.
Sample completeness. The sample collection completeness objective of 80% was met
for three of the four 2025D sequential dichot samplers. Sample collection completeness
exceeded 90% for the two sequential dichot samplers at Phoenix and was 66% and 87% for the
two sequential dichot samplers in St. Louis. Valid samples were collected from both sequential
samplers on 92% of days at Phoenix, but only 52% of days in St. Louis. A major hardware
failure required one St. Louis dichot sampler to be returned to the manufacturer, leading to low
data completeness at that site. The most common field operations issues were filter exchange
errors and pump failures in the sequential samplers.
Dichot versus FRM by difference. PMC constituents measured on the Teflon filters
(gravimetric mass, elements, and ions) were biased low for the dichot method compared to the
FRM by difference method. In Phoenix, PMC mass from the dichot was, on average, about 20%
lower than the FRM difference method mass (dichot-on-FRM slope = 0.67-0.71, intercept =
2.2-2.7 |a,g/m3 depending on the dichot sampler); in St. Louis, the dichot PMC mass was 10% to
25% lower (dichot-on-FRM slope = 0.83-0.96, intercept statistically indistinguishable from zero
[95% confidence level], depending on the dichot sampler). In contrast, PMC total carbon and
carbonate measured on the quartz filters showed no bias between the two methods, though the
relationship for total carbon exhibited more scatter. The bias for constituents measured on the
Teflon filters is attributed to particle losses from the dichot minor flow channel Teflon filter, which
contains all of the coarse particles and 10% of the fine particles. Losses may occur during the
automated filter exchange in the sequential dichot sampler, during handling, during shipping to
the analytical laboratory, or during any combination of these events. Coarse particles collected
on quartz filters are much less prone to losses because the particles are more deeply
embedded into the filter matrix. Dichot Federal Equivalent Method (FEM) samplers with a
modified shuttle mechanism and firmware to minimize particle loss due to filter exchange are
now available, but were not available from the manufacturer for this study. After the dichots
were modified to be FEM compliant, a follow-up study conducted at Research Triangle Park
(North Carolina) by RTI and EPA resulted in better agreement between the dichot PMC
gravimetric mass and the FRM by difference (PM10-2.5) gravimetric mass, with a dichot-on-FRM
regression slope of 1.05 and an intercept statistically indistinguishable from zero (95%
confidence level). Biases between the dichot method and the FRM by difference method
prevented an evaluation of the potential measurement bias from mixing of PMC and PMf species
components on the PM10 filter.
ES-2
-------
EPA's Coarse PM Pilot Study
Executive Summary
Dichot precision. Dichot collocated precision for PMC gravimetric mass and major soil
constituents (aluminum [Al], calcium [Ca], iron [Fe], silicon [Si], and titanium [Ti]) was in the 8%
to 15% range. In contrast, the collocated precision of organic carbon, which was about 15% of
the mass at both sites, was 19% in Phoenix and 34% in St. Louis. The OC data in St. Louis are
less precise in part from having lower concentrations. Dichot collocated precision for PMf
gravimetric mass was 10% in Phoenix—consistent with the precision for PMf major crustal
species—and 2% in St. Louis.
Acid gas denuders. Ambient nitric acid can adsorb onto filters and cause a positive
artifact for PM nitrate measurements. Sampling conducted in the summertime with collocated
samplers, with and without acid gas denuders, showed insignificant differences in PMf and PMC
nitrate. It appears the sampler inlets can efficiently remove nitric acid and suppress a nitrate
measurement bias.
Organic Carbon. OC mass loadings on the dichot PMC channel backup filters were
statistically indistinguishable from the trip blanks and field blanks OC mass loadings. This is
consistent with very little volatile OC in the PMC size fraction.
Carbonate fraction. Carbonate (CO3) was measured from the dichot PMC quartz filters
on 69 selected sampling events (43 in Phoenix, 26 in St. Louis). Carbonate was also measured
on the dichot PMf quartz filter for 15 of these sampling events. PMf carbonate was below the
3-sigma minimum detection limit (MDL) of 0.52 |jgC/m3 for all samples. However, PMC
carbonate was consistently detected with mean concentrations of approximately 1.2 |jg/m3 and
75th percentile concentrations of approximately 1.6 |jg/m3 in both Phoenix and St. Louis. The
mean carbonate concentrations correspond to 6% and 12% of the PMC mass in Phoenix and St.
Louis, respectively. PMC carbonate was highly correlated with PMC calcium at both sites.
Assuming all carbonate is present as calcium carbonate, about half of the PMC calcium in
Phoenix and two-thirds of the PMC calcium in St. Louis can be explained as being calcium
carbonate.
Biomarker concentrations. Biomarkers (proteins, (1,3)-(3-D-glucans, and endotoxin)
were measured from Teflon filters in the dichot coarse particle channel for 54 sampling events
(28 in Phoenix and 26 in St. Louis). These samples were collected from February through May
2011. In both Phoenix and St. Louis, median PMC glucan concentration was approximately
0.2 ng/m3, and protein concentration was about 0.08 |jg/m3. However, relatively high blank
corrections caused large uncertainties in the proteins data. PMC endotoxin concentrations were
suspect in Phoenix because of dramatic differences in concentrations between analysis
batches, although the batches correspond to adjacent but not overlapping time periods. Median
endotoxin concentration was 0.07 EU/m3 St. Louis.
Mass closure via dichot. Closure between the gravimetric mass and sum-of-species
mass was evaluated for the dichot Teflon filters. The analysis ignored OC artifacts and assumed
that EC and OC loadings on the quartz filters were representative of EC and OC loadings on the
Teflon filters (it is possible that carbonaceous particulate matter is also lost from the dichot PMC
channel filters as reported above for mass, elements, and ions). The analysis also assumed that
the equation commonly used to estimate PMf crustal mass concentration from the major crustal
elements (Al, Ca, Fe, Si, and Ti) is valid for estimating PMC crustal mass concentrations.
ES-3
-------
EPA's Coarse PM Pilot Study
Executive Summary
Assuming an organic-matter-(OM)-to-OC ratio of 1.6, both PMf and PMC mass concentrations
reconstructed from the speciation data at Phoenix were, on average, about 13% higher than the
gravimetric mass. For the St. Louis data, the reconstructed mass was 6% lower for PMf and 1%
higher for PMC compared to the gravimetric mass. Additional analyses were performed using the
OM/OC ratio as an adjustable parameter to obtain best-fit mass balance closure. For PMf, the
best-fit OM/OC ratios were about 1.2 in Phoenix and 1.8 in St. Louis; these ratios are consistent
with estimates reported in the literature, e.g., Simon et al. (2011). For PMC, the best-fit OM/OC
ratio for St. Louis was about 1.5, but subject to large uncertainty; for Phoenix, the ratio was 0.6,
which is physically unrealistic (the ratio cannot be less than unity). This finding that the PMC
reconstructed mass is biased high, especially in Phoenix, suggests systematic errors in the
estimation methodology, such as improper multipliers for estimating crustal PMC from elemental
concentrations, or corrections for X-ray attenuation during XRF analysis of light elements (e.g.,
Al, Si, Ca) that are too large. Two additional confounders are the assumption that EC and OC
are not lost from the Teflon filter (accounting for such losses may improve mass closure, though
it may lead to an overestimate of mass collected on the Teflon filter), and the exclusion of
carbonate from the reconstructed mass calculation (accounting for carbonate would increase
the reconstructed mass concentrations and thus lead to even larger overestimation of the
gravimetric mass).
XRF measurements. Corrections for X-ray attenuation during XRF analysis (self-
attenuation) were evaluated by analyzing Teflon filters from 18 sampling events (10 in Phoenix,
8 in St. Louis) using both XRF and ICP-MS. Dichot PMf and PMC channel filters were analyzed
for all 18 events, and PM10 and PM2.5 FRM filters were analyzed for 10 of the events. For light
elements associated with crustal material (Al, Ca), the coarse particle concentrations by blank-
corrected ICP-MS were greater than the concentrations by XRF. This pattern does suggest that
the corrections for self-attenuation for these constituents are too large. However, quantitative
comparisons were confounded by large ICP-MS blank values for elements such as Al and Ca,
which are present in the membrane filter support ring, the adhesive, and the ink used to stamp
the filter ID number. Smaller corrections for self-attenuation will yield lower PMC concentrations
for these elements and a lower estimate for the crustal PMC mass concentration.
Dichot FDMS TEOM measurements. Hourly PM2.5 and PM10-2.5 concentrations from the
Thermo 1405-DF FDMS TEOM instruments revealed appreciable volatile PM2.5 mass, but
volatile PM10-2.5 mass was too small to be reliably distinguished from measurement error. This is
consistent with expectations that ammonium nitrate and particle-phase semivolatile organic
compounds tend to be in the fine size fraction.
ES-4. Recommendations
The recommendations presented here are based on experiences from the one-year pilot
study with sampling in Phoenix and St. Louis. There are limitations when basing
recommendations on the operations and data for only two sites, and care should be taken to
adapt the recommendations as appropriate for other environmental settings.
Sequential dichot sampling is an attractive approach to particle collection for PMC
measurement. The dichot sampler segregates fine and coarse particles onto separate filters,
ES-4
-------
EPA's Coarse PM Pilot Study
Executive Summary
thus minimizing the potential for measurement artifacts from the mixing of these particles. With
sequential sampling, the instrument can be programmed for multiple sampling events, reducing
the manpower burden for field operations. Care must be taken to maintain the setpoint fine and
coarse channel flowrates to achieve the design PM2.5 cutpoint and to appropriately correct the
coarse channel data for fine particle intrusion.
On the basis of this study, paired dichotomous samplers are recommended, with one
sampler collecting particles onto Teflon filters and the other sampler collecting particles onto
Q/Q filters (see note below). If high concentrations of coarse particle nitrate are expected, e.g.,
at sites where atmospheric processes have converted sea salt into sodium nitrate, a T/N filter
and Q/Q filter combination should also be used during the first year of operations, with analysis
of ions on the T and N filters to assess coarse particle nitrate concentrations. In environments
similar to St. Louis or Phoenix, a denuder does not appear to be necessary; in environments
where there may be significant nitric acid that could absorb onto the Teflon filter and be
quantified as aerosol nitrate, it may be useful to conduct a series of test days to determine
whether a denuder is needed as part of routine sampling. Specific recommendations and
caveats regarding field operations and chemical analyses are discussed below.
Post pilot study note: since the completion of this pilot study, EPA has determined that
backup quartz filters are not necessary for OC artifact correction; therefore, the
recommendation for a paired dichot with Q/Q filters is revised to recommend a paired dichot
with a Q filter only.
Sampling and Field Operations
At both sites, dichot PMC constituents measured on Teflon filters (gravimetric mass, XRF
elements, and ions) were biased low compared to PM10-2.5 data collected under the FRM by
difference method. In contrast, such bias was not observed for PMC constituents measured on
quartz filters in the dichot PMC channel (carbon). It is likely that coarse particles become
dislodged from the dichot PMC channel Teflon filter during the automated filter exchange, but it is
also possible that the particles are dislodged during shipping from the field sites to the analytical
laboratory or during filter handling. Although a shipping protocol recommended by the EPA's
Office of Research and Development (ORD) was used to minimize particle loss, this study did
not conclusively determine which mechanism was responsible (filter exchange or shipping) for
particle loss.
Hardware failures were more frequent than anticipated, especially in St. Louis, with the
most common problem being errors during the automated filter exchanges and pump failures,
together accounting for 8% of dichot sampling events being invalid. Although the Thermo 2025D
sequential dichotomous sampler has since been designated a FEM for PMC, the 2025D
samplers used for this study were not FEM-compliant. The aforementioned issues with particle
losses and field robustness of the sampler may be specific to the non-FEM version of the
2025D, and users should ensure they are using FEM-compliant samplers. Given that a
complete speciation sample requires valid data be collected by two independently operating
samplers, it may be necessary to maintain an inventory of backup hardware to minimize
sampler downtimes.
ES-5
-------
EPA's Coarse PM Pilot Study
Executive Summary
Low-volume (16.7 liter per minute [LPM]) dichot samplers were used in this study.
Detectability and precision were deemed adequate for the constituents of primary interest.
Standard operating procedures (SOPs) for the analytical methods used were included in
the pilot study's Quality Assurance Project Plan (QAPP). These SOPs are appropriate for use in
PMC speciation measurement except for the analysis of elements by XRF (where attenuation
correction factors would need to be revised) and ICP-MS (where a stronger digestion method is
required for the crustal species).
Sample Analyses
Based on experiences in Phoenix and St. Louis, the following baseline measurements
are recommended for PMC speciation.
1. Gravimetric mass concentrations using Teflon filters and following the PM2.5 CSN
method. Analysis must be performed on both the PMf and PMC channel filters. This study
used the filter handling and shipping protocols developed by the EPA's ORD and are
presumed to be adequate for mass and chemical speciation. However, these protocols
should be verified by conducting a specific study to assess potential filter handling and
shipping effects on PMC once a routine network is operational.
2. Elemental mass concentrations by XRF using Teflon filters and following the PM2.5 CSN
method. Analysis must be performed on both the PMf and PMC channel filters. At both
sites, coarse PM mass was dominated by crustal material, so it is important to quantify
the major crustal constituents. However, the corrections for self-attenuation applied to
XRF results for light elements such as Al, Ca, and Si in PMC appear to be too high.
Examination of PMC mass balance closure and comparisons of PMC constituents
measured by XRF and ICP-MS suggest that the corrections for self-attenuation are
necessary, but that the current corrections overestimate the actual concentrations.
Additional work is needed to establish corrections for use with PMC data. The best-fit
corrections determined in this study are subject to confounders that may bias the
estimates. Numerous factors that influence the corrections, such as the particle size
distribution, should be taken into consideration to generate robust corrections. The
comparison should be made using a larger data set, with samples collected at sites that
have high crustal loadings, and ideally including coarse PM from different sources, such
as desert dust and agricultural dust.
3. Elemental and organic carbon concentrations by TOA using quartz filters and following
the PM2.5 CSN method with the Interagency Monitoring of Protected Visual
Environments (IMPROVE) analysis protocol IMPROVE_A. Both the PMf and PMC
channel filters must be analyzed. The IMPROVE_A protocol is recommended because it
would be consistent with the PM2.5 CSN network. Also, the maximum temperature during
analysis by the IMPROVE_A protocol is below the decomposition temperature for
calcium carbonate. However, carbonate might decompose at lower temperatures
because of matrix interactions among particle constituents, and more work should be
done to evaluate whether carbonate, which was observed in PMC at both sites, interferes
with the measurement of EC and OC.
ES-6
-------
EPA's Coarse PM Pilot Study
Executive Summary
4. Carbonate mass concentration by TOA with acidification and using quartz filters.
Carbonate was measured for a subset of the collected samples. PMf carbonate
concentrations were all below the 3-sigma minimum detection limit and did not contribute
to PMf mass. However, PMC carbonate concentrations were, on average, about 6% of
the PMC gravimetric mass at Phoenix and 12% of the PMC gravimetric mass at St. Louis.
PMf carbonate is expected to be low at virtually all sites; therefore, the measurement
should be performed on the PMC channel filter only, because the correction for fine
particle carbonate will be negligible. While carbonate concentrations were similar at
Phoenix and St. Louis despite the dramatically different environments, it is possible that
PMC carbonate might be negligible at some sites and could be dropped from the analysis
plan for such sites after a period of sampling that demonstrates persistently low
carbonate concentrations.
In addition to the above baseline measurements, the following analyses are
recommended depending on site-specific conditions.
1. Anion species mass concentrations by water extraction and IC using Teflon and nylon
filters and following the PM2.5 CSN analysis method. PMC sulfate and nitrate
concentrations were persistently low at both Phoenix and St. Louis. XRF measurement
of total sulfur (S) includes sulfate, and for PMC speciation it is likely unnecessary to
discriminate the sulfate contribution to total sulfur. In contrast to Phoenix and St. Louis,
some locations—particularly sites near coastlines—may have significant concentrations
of PMC nitrate, which should be measured on the Teflon filter. PMC nitrate is expected to
be nonvolatile, and in the absence of fine particle ammonium nitrate, it is possible to
analyze only the Teflon filter. However, in some locations with PMC nitrate, there may be
considerable fine particle ammonium nitrate—in such cases, it will be necessary to also
measure nitrate on a nylon filter placed immediately downstream of the PMf Teflon filter
to properly correct the PMC data for fine particle intrusion in the dichot PMC channel. In
locations where there is abundant nitric acid, a denuder may also be necessary to
ensure that nitric acid is not quantified as aerosol nitrate. As part of a site-specific
assessment of the abundance of coarse particle nitrate, collocated samplers with and
without a denuder should be run for a limited period to assess the need for a denuder as
part of the site's routine operations.
2. Cation species mass concentrations by water extraction and IC using Teflon filters and
following the PM2.5 CSN analysis method. PMC ammonium concentrations were very low
in Phoenix and St. Louis and are expected to be low at most locations. PMC sodium
concentrations were higher than PMf sodium concentrations and were present
predominantly as the monovalent cation (Na+). PMC potassium concentrations were
similar to PMf potassium concentrations, and were present predominantly in forms other
than the monovalent cation (K+). The limited utility from measuring PMC ammonium, ionic
sodium, and ionic potassium does not justify the additional cost for routine operations at
most sites.
ES-7
-------
EPA's Coarse PM Pilot Study
Executive Summary
The following analyses may be considered for special cases, but are not warranted as
routine measurements for PMC speciation.
• Scanning electron microscopy (SEM) can provide substantial insights into particle
sources by classifying particle shape and/or composition. However, Teflon and quartz
filters cannot be used for quantitative analysis by SEM. An additional sampler would be
needed to collect particles onto a suitable substrate, such as polycarbonate membrane
filters. The extra sampling requirement and analytical costs relegate SEM to special
studies rather than routine measurements.
• Biological material—both intact and fragmented—can be a significant contributor to PMC.
Biomarker concentrations for glucans (an indicator for spores) and proteins can provide
insights into spatial and temporal patterns, but to be most valuable to PMC speciation,
multipliers are needed to convert the biomarker concentrations to mass concentrations
of the corresponding biologic material (i.e., mass biologic material per mass of
biomarker). This issue and the extra analytical costs relegate biomarkers to special
studies rather than routine measurements.
• For many elements, ICP-MS provides better sensitivity than XRF. However, this study
demonstrates that detectability and precision using XRF are adequate for the primary
elements of interest. ICP-MS may be attractive for special cases where higher-quality
trace elements data are desired or to confirm that appropriate corrections for self-
attenuation are being used for XRF analysis. EPA's current PM2.5 ICP-MS SOP would
need to be optimized for the specific elements targeted for ICP-MS analysis of PMC
elements. For example, the ICP-MS analyses conducted for this project required the use
of microwave and mixed acid (nitric, hydrochloric, and hydrofluoric) digestion process
because crustal elements were the primary target of the analysis.
ES-8
-------
EPA's Coarse PM Pilot Study
Introduction
1. Introduction
In 1997, the U.S. Environmental Protection Agency (EPA) promulgated revisions to the
National Ambient Air Quality Standards (NAAQS) for particulate matter (PM) and added a
standard for fine particulate matter (PM2.5). In 2006, the EPA issued a final monitoring rule for
thoracic coarse particles. Coarse particles have aerodynamic diameters between 2.5 |jm and
10 |jm; here coarse PM is referred to as PMC, if measured from dichot samplers, or termed
PMio-2.5if from FRM by difference method. The promulgated monitoring requirements specified
the placement of coarse PM speciation samplers at National Core (NCore) monitoring sites. In
2013, the requirement for coarse PM speciation at NCore was revoked because of technical
issues related to the development of appropriate monitoring methods. Sample collection
procedures and analysis methods for coarse PM speciation measurements were explored as
part of the small-scale pilot monitoring study presented here.
In 2009, the Clean Air Scientific Advisory Committee (CASAC) Ambient Air Monitoring
and Methods (AAMM) Subcommittee provided input on sampling and analysis issues for coarse
PM speciation. For coarse PM speciation, the CASAC AAMM strongly recommended the use of
dichotomous samplers (dichots), where coarse particles are directly sampled, rather than
Federal Reference Method (FRM) samplers, where coarse PM is derived from a difference of
PM10 and PM2.5 measurements, i.e., PM10-2.5.
To address concerns of the EPA and CASAC AAMM, a small-scale pilot monitoring
study was deployed, the results of which are presented in this report. This pilot study is
important from several perspectives.
One reason why this study is important is the need to assess whether chemical and
physical characterization of coarse PM differ when the values are determined using the PM10
minus PM2.5 method (termed PM10-2.5 in this report) as compared with characterization of the
PMC fraction derived from the dichotomous sampler. Dichots directly sample the coarse
particles, with 10% of the fine particles drawn through the inlet also present in the sample
stream. There was concern that "mixing" of the PMC fraction with the PM2.5 fraction on a filter
from the PM10 sampler (in the difference method) could lead to changes in aerosol composition
that are different from the changes that occur on the coarse particle filter in the dichot (which
contains only 10% of the PM2.5 mass).
A second important reason for conducting the pilot study was to assess the robustness
of commercial samplers and the training and skills required of the field operator and supporting
laboratory to produce quality data with a high percentage of data capture.
A third reason for the pilot study was to compile a database of coarse PM chemical and
physical information, supplemented by information from measurements not normally made in
the PM2.5 Chemical Speciation Network (CSN) (e.g., protein content, metals determination by
inductively coupled plasma mass spectrometry [ICP-MS], organics speciation by gas
chromatography-mass spectrometry [GC-MS]) and by information derived from collocated
instruments, including a dichotomous tapered element oscillating microbalance (TEOM) monitor
for hourly PM10-2.5 and PM2.5 volatile and nonvolatile mass measurements.
1-1
-------
EPA's Coarse PM Pilot Study
Introduction
1.1 Study Objectives
The primary pilot study objectives were to develop the target species analyte list for
routine speciation monitoring (what species need to be measured); evaluate and define analysis
methods and the necessary SOPs; evaluate the appropriateness of using a dichot sampler;
learn about sampling and operational issues regarding the use of dichots; and evaluate data
from the study to inform other issues (e.g., closure between gravimetric mass and sum-of-
species mass). The coarse PM measurement system includes media preparation, media
shipping, sample handling, routine sampling operations, and laboratory analyses. A list of
species and appropriate measurements needed to reasonably characterize PMC using the low-
volume dichot measurement system is recommended.
The main objectives of the study were as follows:
• Objective 1: Develop the target species analyte list for routine speciation
monitoring. This objective was addressed by starting with the speciate analyte list for
the PM2.5 Chemical Speciation Network and supplementing with additional
measurements such as PMC carbonate.
• Objective 2: Evaluate and define analysis methods for routine speciation
monitoring and the necessary SOPs. Again, the PM2.5 Chemical Speciation Network
was used as a starting point with supplemental measurements added to evaluate the
conventional methods.
• Objective 3: Evaluate the field performance of the dichot samplers for routine
speciation monitoring. To meet project objectives, the PMC dichot and associated
comparison samplers and monitors were operated for one year to provide sufficient
comparison data over a range of atmospheric and seasonal conditions. This information
was needed for the major components of the PMC aerosol, including elements, ions, and
carbon. The needed information was obtained from collocated measurements, trip
blanks, and field blanks. Primary and collocated dichot samplers were used to collect
eight collocated samples of each substrate type per sampling season (with three
sampling seasons per year). In order to accomplish this objective, both dichots were run
with a Teflon/nylon filter pair for eight events per season, and with quartz filters for eight
events per season. Both trip and field blank filters were collected, and four times each
season, field blanks were collected that mimicked a sampling event (but with no air
pulled through the sampler).
• Objective 4: Learn about sampling and operational issues regarding the use of
dichots. Again, the PMC dichot and associated comparison samplers and monitors were
operated for one year to provide information on sampling and operational issues.
• Objective 5: Evaluate data from the study to inform several issues related to
coarse PM speciation measurements. Data analyses were conducted to inform
sampling measurement performance including precision, comparability, and
representativeness. Mass balance closure was examined to identify potential issues in
the speciation measurements.
1-2
-------
EPA's Coarse PM Pilot Study
Introduction
1.2 Study Design
For this pilot study, two monitoring sites were chosen representing different
environmental concentrations and aerosol mixes to operate for nominally one year (May 2010 to
May 2011). The pilot study included sites in Phoenix, Arizona (abbreviated as PHX in tables and
figures), and East St. Louis, Illinois (abbreviated as STL in tables and figures).
At both sites coarse PM was likely to be dominated by crustal elements, but as Phoenix
is in the arid Southwest, concentrations were likely to be higher there. Primarily, two methods
were used to collect coarse PM samples for analysis:
• dichotomous samplers to directly measure PMfine (PMf) and PMC, and
• paired PM10 and PM2.5 FRM samplers to determine PM10-2.5 (difference method).
All samples were collected for 24 hours from midnight to midnight local time. Laboratory
analysis methods consistent with PM2.5 CSN processes (i.e., gravimetric mass, ions by ion
chromatography [IC], elements by XRF, carbon by thermal-optical analysis) were used to
analyze about 50% of the filter samples. The remaining samples were archived for future
analyses.
Sampling commenced in May 2010, with four weeks of nearly daily sampling to refine
the field operations and provide a data set for preliminary evaluation of certain sampling
configurations (e.g., whether the presence of a denuder affected the PMC mass measurements).
Sampling was conducted on a one-in-three day schedule from June 1, 2010, through May 31,
2011, using various sampling configurations to address the technical objectives of the project.
For details, see the PM10-2.5 Speciation Pilot Monitoring Quality Assurance Project Plan
(abbreviated as QAPP and approved in 2010). Additional sampling was conducted periodically
during the study to provide samples for biological content analyses. Under contract EP-D-08-
047, RTI International personnel and subcontractors who regularly serve the EPA/Office of Air
Quality Planning and Standards (OAQPS) PM2.5 CSN provided support for integrated sampler
installation and operation, necessary training, initial equipment audits and flow checks. Filter
preparation and laboratory sample processing and analyses were provided under contract EP-
D-09-010. Under EPA contract EP-D-09-097, Sonoma Technology, Inc. (STI) personnel and Dr.
Jay Turner (Washington University, St. Louis) analyzed the data.
The Phoenix site (Figure 1-1) is at 43rd Avenue and Broadway Road in Phoenix, Arizona
(AQS ID 04-013-4009). The Maricopa County Air Quality Department in Phoenix managed the
day-to-day operations.
The East St. Louis, Illinois, coarse PM speciation pilot site (Figure 1-2) is the PM
Supersite location used previously for PM research (AQS ID 17-163-9010). The St. Louis-
Midwest Supersite is located at 13th Street and Tudor Avenue in East St. Louis, Illinois, which is
about 3 km east of the City of St. Louis, Missouri, central business district. The Air Quality
Laboratory at Washington University in St. Louis managed the day-to-day operations. The
physical footprint managed by Washington University is immediately adjacent to the East St.
Louis compliance monitoring site operated by the Illinois EPA (AQS ID 17-163-0010).
1-3
-------
EPA's Coarse PM Pilot Study
Introduction
Kilometers
Kilometers!
Figure 1-1. Site maps for W 43rd Ave, Phoenix monitoring site. Concentric circles in the
bottom map are 500 m and 1,000 m radii from the site.
1-4
-------
EPA's Coarse PM Pilot Study
Introduction
Kilometers
Kilometers
Figure 1-2. Site maps for 13th arid Tudor, East St. Louis monitoring site. Concentric
circles in the bottom map are 500 m and 1,000 m radii from the site.
1-5
-------
EPA's Coarse PM Pilot Study
Introduction
1.3 Technical Approach
Each dichot had two channels, one for coarse PM and one for fine PM. To collect data
for the calculation of mass balance, at a minimum, a Teflon filter (for ions and elements) and a
quartz fiber filter (for carbon) had to be collected for each size fraction; this means data from two
collocated dichots were needed to achieve mass balance on any given day. To understand
collocated dichot precision, two dichots were run with the same filter media, i.e., either Teflon or
quartz fiber. In addition, the sampling schedule was harmonious with the field work already
occurring on site, which was typically one-in-three day sampling. Filter blanks were also
collected at a regular interval.
To best achieve these goals, a six-day cycle was implemented that alternated between
collocated, mass balance, and field blank collection days for a four-month period, as shown in
Figure 1-3. This cycle was repeated twice more to complete a year of sampling. This sampling
approach resulted in filters for mass balance every sixth day, while the other every-sixth-day
pattern resulted in a series of collocated filters or field blanks. In addition, PM2.5 and PM10 FRM
samples were also collected on Teflon filters in parallel with the dichot measurements. Thus, at
each site there were two dichots plus collocated PM2.5 and PM10 FRM samplers. In addition, a
Thermo 1405-DF Filter Dynamics Measurement Systems (FDMS) dichotomous TEOM - which
is a Federal Equivalent Method (FEM) for PM2.5 but not PM10-2.5 - was operated to obtain hourly
PM2.5 and PM10-2 5 data.
6-day cycle
1
2 3
4
5 6
01
fT/N.Q)01
(T/N,Q)FB01
02
(T/N.Q)02
(T/N.T/N)01
03
(T/N,Q)03
-------
EPA's Coarse PM Pilot Study
Introduction
1.3.1 Instrumentation
The following equipment was installed and operated at each of the two sites:
• Two (2) sequential dichotomous samplers (Thermo 2025D);
• Two (2) sequential Thermo 2025 FRM samplers of the same make and model, one for
PM10 and the other for PM2.5 (for the PM10-2.5 "difference method" measurements);
• One (1) dichotomous semi-continuous mass monitor (Thermo 1405-DF FDMS TEOM);
and
• One (1) eight-stage (0.18, 0.32, 0.56, 1.0, 1.8, 2.5, 5.6, and 10 |jm) MOUDI impactor
(MSP Corporation).
While the Thermo 2025D sequential dichotomous samplers were designated as an FEM
prior to this study, the samplers provided by Thermo were older models that were not FEM-
compliant. Thus, it will be important to obtain additional field operations experience with the
FEM-compliant version of the sampler.
1.3.2 Sample Analysis
Routine sample handling and analysis protocols followed the Standard Operating
Procedures (SOP) used by RTI for the PM2.5 CSN. Details on sample handling and analysis are
provided in the Pilot Study QAPP (U.S. EPA, 2010). The following Whatman® 47 mm filters
(GE Healthcare, Pittsburgh, PA) were used: Teflon membrane (Part Number 7592-304); nylon
membrane (Part Number 7410-004), and quartz fiber (Part Number 1851-047). Teflon filters
were analyzed for mass by gravimetric weighing, for elements by XRF, and for water-soluble
ions by IC. The backup nylon filters were analyzed for ions by IC. Quartz fiber filters were
analyzed for carbon using the IMPROVE_A thermal-optical analysis (TOA) protocol currently
used in the CSN network. Mass and speciation data from these measurements have been
uploaded to the EPA Air Quality System (AQS) database. Additional filters were selected for
analysis of carbonate, biologicals, metals by ICP-MS, microscopy, and speciated organics.
1.3.3 Nomenclature and Dichot Sampling Concentration Equations
Table 1-1 defines the different types of particulate matter referenced in this document.
PM2.5 from the FRM is termed PMfine (PMf) in some cases when the distinction is unambiguous.
Table 1-1. Definitions of the different types of PM referenced in this document.
Abbreviation
Description
PM10-2 5
Coarse PM by difference method
FRM-measured PM10 minus FRM-measured PM2 5
PM10
FRM-measured PM10
PM25
FRM-measured PM25
PMf
Fine PM measured by a dichotomous sampler
PMc
Coarse PM measured by a dichotomous sampler
1-7
-------
EPA's Coarse PM Pilot Study
Introduction
The dichot samplers used in this study include a 16.7 liters-per-minute (LPM) standard
PM10 inlet followed by a virtual impactorthat splits the sample stream into major (15.0 LPM) and
minor (1.67 LPM) flows. The major flow includes only fine particles, whereas the minor flow
includes all of the coarse particles and 10% of the fine particles. PMf mass concentrations are
calculated directly from the major flow, and PMC mass concentrations are calculated by
correcting the minor flow for fine particle intrusion. For a nonvolatile species j collected on the
front filter of the two-filter (routine/backup) sandwiches, the governing equations for species
mass concentration are
PM. ,
3 J
PM.
he
YYl
j, front major
V ¦
major
1
/
^total
V ¦
m minor m
j, front m inor jr j, front m aj or
major
(Eq. 1-1)
jfi j/
j, front minor _ v mjnor
^total ^total
pmjj
where PMj,i< is the mass concentration of species j in size fraction k{f= fine, c = coarse), my,p is
the species j mass loading on the front filter in the specified flow channel p (major or minor); and
Vp is the total air volume sampled by flow channel p (major, minor, or total = major + minor). For
operation with the setpoint flow rates, VminorA/t0tai =0.10; thus, the correction to the coarse
particle concentration for fine particle intrusion is 10% of the fine particle concentration (Dzubay
and Stevens, 1975).
For species such as nitrate that are collected on both the front (Teflon) filter and back
(nylon) filter, the governing equations are
PM
jfl
j, front major j, back major
jj
K
PM'
K
total
K
major
+ Tfl
front minor j, back minor
\ V ¦ /
' V v J'
+ Tfl
front major j, back major
(Eq. 1-2)
PMhc+-
m,
'j, back minor ( minor
vm
V,
total
K
total
fm ^
j, back major
V ¦
major J
where the prime species mass concentration includes mass on both the front and back filters.
1-8
-------
EPA's Coarse PM Pilot Study
Introduction
1.4 Field Operations Overview
1.4.1 Methods
Field operations methods are detailed in the Pilot Study QAPP (U.S. EPA, 2010). All
work performed and data collected at the Pilot Monitoring Program site locations were based on
the following quality objectives:
• Multiple samplers were installed at each pilot site with inlets as close to the same height
as possible and within a 1- to 4-meter separation from each other.
• All routine field sampling information (start time, end time, average flow rate,
temperature and pressure data, meteorological conditions, etc.) and verification quality
assurance (QA) checks were recorded on hard-copy field data forms prepared for use
with each sampler.
• All integrated sampler field data were verified and placed in the RTI database and
periodically provided to EPA/STI. These data were also posted to the EPA AQS
database. The TEOM mass monitor data was submitted to STI via the AirNow-Tech
website. RTI was the central repository for all integrated sampler data (including raw
data) and related field information.
• All sampler parameters (flow rate, ambient and filter temperature, and barometric
pressure) were verified against NIST-traceable standards prior to beginning and at the
completion of the study, or after any sampler maintenance.
• For this project, the target completeness objective (completeness being the percentage
of valid data compared to the total expected data) for all species and measurements was
80% of all scheduled measurements. In addition to individual measurement
completeness, the program completeness (sampling events with all attempted
measurements having valid data) was tracked, because program completeness dictates
the robustness of the data set across the entire measurement strategy.
The PM10-2.5 Pilot Study QAPP details audit procedures and routine operations, as well
as sample handling and laboratory procedures.
1.4.2 Field Operations Summary
Table 1-2 summarizes the study sample collection completeness for June 1, 2010,
through May 31, 2011, excluding the extra sampling days programmed to collect samples for
biological content analysis. At each site, the two dichotomous samplers were distinguished by
the designations "A" and "B," or primary (P) and collocated (C). Sampler hardware failures were
the most common reasons for invalid samples. Sampling completeness for the Phoenix
operations was above 90% for all samplers, and each sampler met the 80% completeness
target. For St. Louis, the sampling completeness was much lower because of issues with the
2025D (dichotomous) samplers, and one sampler failed to meet the 80% completeness target.
The low sample collection completeness in St. Louis was particularly problematic for this
methods evaluation study, which places high value on the number of days with valid sample
collection for all four samplers (only 44% of all sampling days for St. Louis).
1-9
-------
EPA's Coarse PM Pilot Study
Introduction
Table 1-2. Sample completeness by site and sampler type for operations from June 1,
2010, through May 31, 2011, excluding sampling days for biological aerosols analysis.
About half of valid samples were archived by RTI rather than analyzed. The total number
of possible sample days is 122.
Sample Type
Phoenix
St. Louis
Valid
Sample
Days
% of Total
Possible
Sample Days
Valid
Sample
Days
% of Total
Possible
Sample Days
PM2.5 FRM (2025)
115
94%
101
83%
PM10 FRM (2025)
114
93%
112
92%
Dichot A (2025D)
113
93%
80
66%
Dichot B (2025D)
120
98%
106
87%
Both Dichots
112
92%
71
58%
All Samplers3
105
86%
54
44%
a Sampling events with all four samplers (PM2.5 and PM10 FRMs, Dichots A and B) having
valid sample collection.
At both sites, there were sampler issues in the beginning of the study. These issues
were most often due to issues with the filter exchange mechanism. Overall, there were fewer
valid dichot measurements in St. Louis than in Phoenix.
At Phoenix, one sample from Dichot B and four samples from Dichot A were lost
because of issues with the filter exchange mechanism. In addition, Dichot A sampling events
were lost because of a short run time (one), a flow rate problem (one), and operator error (two);
one sample from each dichot was lost because of non-operational events (e.g., filter
mishandling).
At St. Louis, for Dichot A, 28 samples were not collected because an electronic board
failure required the sampler to be returned to the manufacturer for repairs, and there was a
delay in receiving a functioning replacement. Other samples were lost or not collected because
of filter exchange errors (five samples), operator error (four samples, two due to a denuder
installed for quartz/quartz [Q/Q] sampling), pump failure (four), and laboratory technician error
(one). Once the original Dichot A was replaced in mid-January, data completeness for the
remaining 45 sampling events was 96%, indicating that the majority of lost samples were due to
the malfunctioning unit. Dichot B samples were lost because of a pump failure that was initially
misdiagnosed as a filter exchange mechanism error (six), a second/replacement pump failure
(six), filter exchange errors (three), and operator error (one). One problem with the pump
failures is that they often initially manifest as a filter exchange error when the actuator line
pressure drops too low to properly advance the filter shuttle mechanism. PM2.5 FRM sampling
issues were predominantly limited to filter exchange errors.
While the Thermo 2025D sequential dichotomous air sampler was designated as a FEM
prior to this study, FEM-compliant samplers were not available and therefore, not provided by
Thermo. Given the sampler delivery lead time and time constraints, only the two dichotomous
and two FRM samplers sent to the Phoenix site were verified operational at RTI prior to
1-10
-------
EPA's Coarse PM Pilot Study
Introduction
deployment. These four samplers were evaluated for flow rate, leaks, temperature, barometric
pressure, and filter exchange. Double quartz (Q/Q) filter cassettes required modification in
thickness to reduce filter cassette jamming. Samplers deployed in St. Louis were also newly
purchased but were not verified operational at RTI; minimal sampler testing was performed at
St. Louis due to time constraints, likely contributing to equipment problems during the study.
1.5 Guide to This Report
Section 1 provides an overview of the pilot study and field operations.
Section 2 compares dichot and FRM samplers for gravimetric mass in order to gauge
precision and bias. In this section, to put analyses presented in later sections in context, the
overall composition of the fine and coarse aerosol at each site is discussed. The order in which
the results of the study are presented here builds up to the question of mass balance closure.
As crustal elements are the largest contributor to coarse aerosol mass, Section 3
provides details on the collocated precision of these measurements, a discussion of potential
biases regarding the correction in for X-ray attenuation in XRF measurements, and a description
of how the XRF measurements compared to ICP-MS measurements.
Section 4 presents details on carbonaceous aerosol, OC/EC splits, and the influence of
biological material on OC.
Section 5 examines ions, including an analysis of nitrate loss on Teflon filters and an
assessment of the usefulness of ion measurements as part of a long-term monitoring network.
Building on the results in Sections 2 through 5, Section 6 then presents mass balance
results for dichot measurements, i.e., an examination of how well the measured species
reconstruct the measured gravimetric mass, including exploration of "best fit" OM/OC ratios to
achieve mass closure.
Section 7 provides additional collocated measurement comparisons using hourly FDMS
TEOM data, and examines volatile versus non-volatile coarse PM.
The appendices provide supporting information as follows.
• Appendix A provides tables of summary statistics (concentrations and minimum
detection limits [MDLs] by species and size fraction) for the samplers in St. Louis and
Phoenix.
• Appendix B is a table of the summary ratios of collocated dichot measurements.
• Appendix C is a table summarizing the dichot-to-FRM comparisons.
• Appendix D provides information on quartz fiber filter carbon blanks.
• Appendix E is a table showing nitrate correlations with other species.
1-11
-------
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
2 ic Usss *V,r^v>mon Overview
PMC constituents measured on the Teflon filters (gravimetric mass, elements, and ions)
were biased low for the dichot compared to the FRM by difference method (Section 2.1). In
Phoenix, PMC mass from the dichot was on average about 20% lower than the FRM difference
method mass, and in St. Louis, the dichot PMC mass was 10% to 25% lower, depending on the
dichot sampler. These findings are consistent with those found for pre-FEM 2025D sequential
dichots evaluated during a multi-site evaluation of candidate methodologies for PM10-2.5 (U.S.
EPA, 2011). In contrast, PMC total carbon measured on the quartz filters showed no bias
between the two methods, though the relationship exhibited more scatter. The bias for
constituents measured on the Teflon filters is attributed to particle losses from the dichot minor
flow channel Teflon filter, which contains all of the coarse particles and only 10% of the fine
particles. On average, PMC was predominantly composed of crustal oxides at both sites, with
15% of the mass attributed to OC and less than 5% of the mass from other species, such as
sulfate and nitrate (Section 2.2). As shown in Section 2.3, dichot collocated precision for PMC
constituents was typically in the 8% to 15% range for species with high rates of detectability. An
exception was organic carbon, which had a collocated precision of 19% in Phoenix and 34% in
St. Louis. The relatively less precise OC results in St. Louis arise in part from lower OC
concentrations than in Phoenix. Organic carbon is further discussed in Section 4.
2.1 Gravimetric Mass Comparisons Between Dichot and FRM
Samplers
Figure 2-1 shows scatter plots of the gravimetric mass measured by the dichotomous
samplers at Phoenix and St. Louis compared to the FRM measurements for PM2.5, PM10-2.5, and
PM10. Using the FRM samplers, the PM10-2.5 concentration is not directly measured, but instead
is calculated from the difference between PM10 and PM2.5 measurements.
Similarly, dichot PMC is not directly measured, but instead must be corrected for fine PM
particle intrusion into the dichot minor flow channel; dichot PM10 is not directly measured, but is
rather the sum of the PMf and PMC measurements. At each site, there were two dichotomous
samplers, labeled "primary" and "collocated." Summary statistics for gravimetric mass
comparisons are presented in Table 2-1. For PMf gravimetric mass, excellent agreement was
observed at St. Louis for both the primary and collocated dichot samplers compared to the PM2.5
FRM (Figure 2-1 a). The mean value of the daily dichot-to-FRM mass ratio was 1.01 for both
dichots, with mean absolute relative differences of 4% to 5%. In contrast, Figure 2-1 d shows
that at Phoenix, the dichot PMf gravimetric mass was biased low with respect to the FRM, with
mean ratios of 0.91 for both dichots. The mean absolute relative differences of 11% to 13%
were worse than observed in St. Louis. Samples at high concentrations and outside the ±20%
envelope (i.e., below the lower dashed line in Figure 2-1 d) all correspond to samples collected
in February and early March 2011. Dichot PMC agreed reasonably well with FRMio-2.son these
days, so it is the dichot PMf measurement that is suspect. The reason for the discrepancies on
these sample days is not known.
2-1
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
PM
2.5
PM
10-2.5
PM
10
o STL - collocate
a STL - primary
o SI L - collocate
a STL-primary
O STL - collocate
a STL - primary
PM10_25 FRM, jjg/m
PMin FRM, ug/m
5 10
PM25 FRM, ^g/m
O PHX- collocate
A PHX-primary
(e)
/
/ /
/ /
/ /
/ /
/ /
/ /
/ / x
/ / /
/ / /
/ / /
/
/ .
* 0- 2
A
/ jtaSB*
O
,
O PHX-collocate
a PHX-primary
5 10 15 20
PM25 FRM, pg/m3
20 40 (
PMi0-25 FRM, jjg/m3
o PHX-collocate
a PHX-primary
20 40 60
PM10 FRM, jjg/m3
Figure 2-1. Dichot versus FRM mass concentrations for St. Louis (top row) and Phoenix
(bottom row) for PM2 5 (left), PM10-2.5 (center), and PM10 (right). Triangles are data from
the primary dichot sampler and circles are data from the collocated dichot sampler.
Diagonal lines are 1:1 (solid) and +20% of 1:1 (dashed).
Dichot PMC gravimetric mass was consistently biased low compared to FRM10-2.5 for both
dichots at both sites (Figures 2-1 b, e). At St. Louis, the dichot-to-FRM mean ratio for the primary
dichot was 0.90 while for the collocated dichot it was only 0.76. At Phoenix, the two dichots
show quite similar performances, with dichot-to-FRM mean ratios of approximately 0.82.
Overall, the mean absolute relative differences of 12-26% across the sites and four samplers is
driven more by systematic bias than random measurement error, as demonstrated by
regression slopes significantly different from unity.
Dichot PM10 gravimetric mass comparison metrics1 (Table 2-1) fall between the PMf and
PMC results. For St. Louis, the PM10 values are near the midpoint of PMf and PMC values, which
1 These metrics include a reduced major axis (RMA) regression, which is a type of orthogonal regression. Ordinary
least squares (OLS) regression minimizes the sum of square differences between the reported and predicted
y-values (i.e. the vertical distance between the best-fit line and measured values) and assumes the x-values are
exact. In contrast, RMA minimizes the sum of square differences between the reported and the predicted values
based on the distance perpendicular to the best-fit line and the measured values. Thus, RMA takes into consideration
uncertainty in both the x- and y-values.
2-2
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
reflects the similar contributions from fine and coarse PM mass to PM10 mass. For Phoenix, the
PM10 values are close to the PMC values, since PM10 in Phoenix is generally dominated by
coarse PM. The dichot-to-FRM bias observed for PM10 demonstrates that the PMf and PMC
biases do not arise from poor cutpoint performance of the dichot virtual impactor. There must be
other sources of significant bias, such as functional differences in the standard PM10 inlet
performance (e.g., from differences in the cleaning protocol and frequency), particle losses
during filter exchanges within the dichot sequential samplers, or particle losses from the filters
during shipping and handling.
Table 2-1. Gravimetric mass comparisons between the dichot and FRM samplers.
P = primary dichot and C = collocated dichot. Mean and median absolute relative
differences were calculated using the FRM as the reference sampler. Mean and median
ratios have been calculated for dichot-to-FRM comparisons. Regression statistics are
from reduced major axis (RMA) regressions of dichot mass on FRM mass.
Total Samples^
PHX-P
80
PHX-C
55
STL-P
45
STL-C
40
PMf (PM2.5)
Mean Absolute Relative Difference
(Median Absolute Relative Diff.)
11% (9%)
13% (7%)
4% (2%)
5% (4%)
Mean Ratio (Median Ratio)
0.91 (0.92)
0.91 (0.94)
1.01 (1.01)
1.01 (1.00)
Regression Slope, |jg/m3
0.91 +0.07
0.65 + 0.08
1.00 + 0.03
0.97 + 0.05
Regression Intercept
-0.1 +0.7
2.0 + 0.8
0.1 +0.4
0.3 + 0.5
r2
0.893
0.798
0.987
0.978
PMC (PM10-2.5)
Mean Absolute Relative Difference
(Median Absolute Relative Diff.)
19% (19%)
20% (19%)
12% (10%)
26% (27%)
Mean Ratio (Median Ratio)
0.83 (0.83)
0.81 (0.81)
0.90 (0.92)
0.76 (0.74)
Regression Slope, |jg/m3
0.71 +0.07
0.67 + 0.10
0.96 + 0.05
0.83 + 0.08
Regression Intercept
2.2 + 1.9
2.7 + 2.5
-0.6 + 0.6
-0.6 + 1.1
r2
0.787
0.719
0.971
0.904
PM10
Mean Absolute Relative Difference
(Median Absolute Relative Diff.)
17% (16%)
18% (17%)
5% (4%)
13% (12%)
Mean Ratio (Median Ratio)
0.84 (0.84)
0.82 (0.83)
0.96 (0.96)
0.87 (0.88)
Regression Slope, |jg/m3
0.79 + 0.07
0.71 +0.08
0.97 + 0.04
0.91 +0.06
Regression Intercept
1.1 +2.5
3.2 + 2.7
-0.3 + 0.9
-0.7 + 1.3
r2
0.845
0.839
0.983
0.960
2-3
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
2.2 Distribution of Species Between Fine and Coarse Mode
Next, the average composition was examined at Phoenix and St. Louis for PMC and PMf,
shown in Figure 2-2. At this stage, the objective was not to determine mass balance closure but
rather to examine which classes of species dominate each size fraction. The PMf composition
was similar to the composition observed with longer-term measurements at both sites, where
Phoenix PMf is predominantly crustal oxides and carbonaceous aerosol, while St. Louis is
mostly ammonium sulfate and carbonaceous aerosol. For PMC, both sites were dominated by
crustal oxides; OC was about 15% of the mass, and other species, including sulfate and nitrate,
made up less than 5% of the average mass. Appendix A shows summaries of average
concentration and MDL, the fraction of samples above MDL and 3*MDL, and the average
uncertainty greater than MDL by species, site, and size (fine or coarse) for species analyzed
from the Teflon and quartz filters.
(a) PHX PMc
(b) PHX PMf
12
ID
4
FRM PHX-Dichot A PHX-Dichot B
ill
FRM PHX-Dichot A PHX-Dichot B
Other Metals
l Soil
l S04
I N03
NH4
I EC
IOC
24
20
16
12
8
4
0
(c) STL PMc
(d) STL PMf
111
FRM STL-Dichot A STL-Dichot B
FRM STL-Dichot A STL-Dichot B
Figure 2-2. Stacked bar plots of average composition for (a) PMC at Phoenix; (b) PMf at
Phoenix; (c) PMC at St. Louis; and (d) PMf at St. Louis. All concentrations are in |ag/m3.
2-4
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
As seen in Figure 2-3 and detailed in Appendix A, the PM10/PM2.5 ratio from FRM
measurements is typically much greater than unity, with values greater than two representing an
enhancement in coarse PM species abundance relative to fine PM. Ammonium, sulfate (both by
IC and by XRF sulfur), bromine, and rubidium are almost exclusively in the fine fraction.
Because the PM10-2.5 concentration will be the difference between two very similar values,
measurement of these species in PM10-2.5 using the FRM-by-difference method can result in
large relative uncertainties. However, from a mass balance perspective, these species
contribute little to PM10-2.5 mass. EC, OC, soluble potassium (by IC), lead, and zinc are, on
average, nearly evenly distributed between the fine and coarse fractions. The remaining
elements tend to be more abundant in the coarse fraction, with the species commonly
associated with crustal material (e.g., aluminum [Al], calcium [Ca], magnesium [Mg], and silicon
[Si]) enhanced in PM10-2.5 by more than a factor of five (i.e., a PM10/PM2.5 ratio greater than ten).
Based on these coarse PM concentration and relative abundance patterns, the analysis
of measurement precision focuses primarily on crustal species and secondarily on OC.
o STL + PHX
CO
01
CL
or
O)
O)
TO
ci3
>
<
2-
10-
8-
6-
4-
2-
1-
o
+
o
?o
o
+
o
o
-T-0--+--
+
+
o +
o
o
+
+
o
o
o
9 ^ t +
o
o
?
o o
+ +
+
+ +
o
o
o
+ +
O O (.)
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
a:cc:u-u_u_u_u-u_U-u_u-u_0U-U-U-0U-0U-0U-U-U-U-0U-U-U-U- 5
ooK-K-K-K-K-K-K-K-K-K-— iCtcta:— ,cz — ^|
Figure 2-3. Average FRM PM10/PM2 5 ratio (log scale) by species for each site.
2.3 Collocated Precision of Dichot Measurements
A major component of the sampling plan was to periodically run the collocated dichots
using the same filter type, making it possible to evaluate dichot measurement precision.
Figure 2-4 shows collocated dichot sampler comparisons for gravimetric mass and silicon
including all data. As demonstrated later in this section, silicon collocated performance is
representative of most crustal species. For these two parameters, the agreement appears good
for both sites and for both particle sizes. Gravimetric mass data for PMC at Phoenix does exhibit
relatively more scatter at high concentrations. Measures of agreement for PMC and PMf at both
sites are reported in Tables 2-2 through 2-5 for gravimetric mass, elements used in the soil
2-5
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
equation, and OC. (In those tables, OC-TOR is organic carbon measured by thermal-optical
reflectance.) The gravimetric mass data sets were conditioned to include only those samples
with XRF elements also reported. Linear regressions of Dichot B concentrations on Dichot A
concentrations were performed using reduced major axis regression, which assumes a constant
measurement uncertainty for the two samplers. For most slopes, the 95% confidence intervals
include unity, and for most intercepts, the 95% confidence intervals include zero.
gravimetric mass
(b)
o
gravimetric mass
(a)
30
2 25
O PHX PMc
A PHX PMf
O STL PMc
A STL PMf
dichot A concentration, ng/m
dichot A concentration, ng/m
silicon
(c)
silicon
(d)
cn 1.5
$ 1,0
O PHX PMc
A PHX PMf
O STL PMc
A STL PMf
123456789
dichot A concentration, ng/m
dichot A concentration, ng/m
Figure 2-4. Collocated dichot versus dichot plots for (a) gravimetric mass at Phoenix;
(b) gravimetric mass at St. Louis; (c) Si at Phoenix; (d) Si at St. Louis. The solid lines are
1:1 lines and the dashed lines in panels (a) and (b) are the reduced major axis
regressions for the PMC data; the regression coefficients are reported in Tables 2-2
through 2-5.
2-6
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
Table 2-2. Measures of agreement between the collocated dichot sampler data for PMC
measured at Phoenix. Slopes and intercepts are reported for reduced major axis
regressions of Dichot B on Dichot A.
PHXPMc
(N=22)
Mean
(pg/m3)
Min, Max
(Mg/m3)
Slope
+/- 95%
C.I.
Intercept
+/- 95%
C.I.
(pg/m3)
1
B/A
Ratio:
Mean
(Median)
Collocate
d
Precision
(pg/m3)
Grav.
mass
18.87
8.09,
37.88
0.91 ±0.13
2.2 ±2.6
0.91
0.97 (0.95)
1.86 (10%)
Al
1.02
0.41, 2.57
1.19 ± 0.17
-0.22 ±0.19
0.90
0.95 (0.94)
0.12 (12%)
Ca
1.19
0.44, 2.81
1.14 ± 0.17
-0.19 ±0.22
0.90
0.96 (0.95)
0.15 (12%)
Fe
0.81
0.30, 1.78
1.04 ±0.18
-0.06 ±0.16
0.87
0.96 (0.97)
0.11 (13%)
Si
3.15
1.45, 8.09
1.22 ±0.18
-0.77 ±0.61
0.90
0.96 (0.95)
0.37 (12%)
Ti
0.061
0.024,
0.14
1.21 ±0.21
-0.01 ±0.01
0.86
0.97 (0.94)
0.01 (15%)
OC-TOR
(ISM 4)
2.29
0.91, 4.01
1.04 ±0.38
0.30 ± 0.84
0.67
1.23 (1.26)
0.43 (19%)
Table 2-3. Measures of agreement between the collocated dichot sampler data for PMf
measured at Phoenix. Slopes and intercepts are reported for reduced major axis
regressions of Dichot B on Dichot A.
PHX PMf
(N=22)
Mean
(pg/m3)
Min, Max
(pg/m3)
Slope
+/- 95% C.I.
Intercept
+/- 95% C.I.
(pg/m3)
r2
B/A
Ratio:
Mean
(Median)
Collocated
Precision
(pg/m3)
Grav. mass
8.10
3.24, 18.61
1.01 ±0.21
-0.56 ± 1.94
0.8
1.09
(1.02)
1.35 (17%)
Al
0.14
0.03, 0.43
1.12 ± 0.11
-0.0005 ±
0.02
0.96
1.13
(1.07)
0.02 (15%)
Ca
0.17
0.04, 0.46
1.17 ± 0.15
-0.003 ± 0.03
0.92
1.14
(1.11)
0.03 (17%)
Fe
0.19
0.04, 0.53
1.01 ±0.11
0.01 ±0.02
0.95
1.08
(1.07)
0.02 (11%)
Si
0.43
0.10, 1.26
1.15 ± 0.12
-0.004 ± 0.06
0.95
1.13
(1.12)
0.07 (16%)
Ti
0.008
0.002,
0.025
1.33 ±0.27
-0.0007 ±
0.002
0.81
1.26
(1.20)
0.002 (29%)
OC-TOR
(ISM 4)
2.11
0.58, 3.75
0.91 ±0.21
0.18 ±0.49
0.87
1.05
(1.01)
0.25 (12%)
2-7
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
Table 2-4. Measures of agreement between the collocated dichot sampler data for PMC
measured at St. Louis. Slopes and intercepts are reported for reduced major axis
regressions of Dichot B on Dichot A.
STL PMC
(N=11)
Mean
(pg/m3)
Min, Max
(Mg/m3)
Slope
+/- 95%
C.I.
Intercept
+/- 95%
C.I.
(pg/m3)
1
B/A
Ratio:
Mean
(Median)
Collocate
d
Precision
(pg/m3)
Grav.
mass
9.77
1.33,
21.31
1.07 ±0.09
0.67 ± 0.95
0.99
0.84 (0.89)
1.08 (11%)
Al
0.27
0.01, 0.73
1.04 ±0.11
-0.05 ±0.04
0.98
0.81 (0.83)
0.03 (11%)
Ca
1.14
0.13, 3.64
0.98 ±0.09
-0.08 ±0.14
0.99
0.90 (0.92)
0.12 (10%)
Fe
0.29
0.08, 0.83
1.00 ±0.02
-0.03 ±0.03
0.99
0.88 (0.91)
0.03 (9%)
Si
0.83
0.10, 1.79
1.03 ± 0.10
-0.09 ±0.10
0.98
0.90 (0.91)
0.07 (8%)
Ti
0.012
0.000,
0.031
1.05 ± 0.17
-0.002 ±
0.003
0.95
0.86 (0.87)
0.002 (13%)
OC-TOR
(ISM 4)
1.71
0.49, 4.53
0.74 ±0.38
0.25 ±0.80
0.40
0.93 (1.00)
0.59 (34%)
Table 2-5. Measures of agreement between the collocated dichot sampler data for PMf
measured at St. Louis. Slopes and intercepts are reported for reduced major axis
regressions of Dichot B on Dichot A.
STL PMf
(N=11)
Mean
(pg/m3)
Min, Max
(pg/m3)
Slope
+/- 95%
C.I.
Intercept
+/- 95%
C.I.
(pg/m3)
r2
B/A
Ratio:
Mean
(Median)
Collocate
d
Precision
(pg/m3)
Grav.
mass
11.21
6.33, 17.64
0.99 ±0.09
0.12 ± 1.04
0.99
1.00
(1.02)
0.27 (2%)
Al
0.05
0.01, 0.27
1.04 ±0.09
-0.002 ±
0.008
0.99
1.04
(1.03)
0.01 (13%)
Ca
0.09
0.01, 0.20
1.17 ± 0.35
0.001 ±
0.03
0.85
1.23
(1.15)
0.02 (20%)
Fe
0.08
0.03, 0.19
1.10 ±0.23
-0.0001 ±
0.02
0.93
1.11
(1.04)
0.01 (15%)
Si
0.11
0.01, 0.53
0.99 ±0.10
0.01 ±0.02
0.98
1.18
(1.15)
0.01 (12%)
Ti
0.003
0.000,
0.019
1.10 ±0.36
0.001 ±
0.002
0.81
1.72
(1.15)
0.002
(55%)
OC-TOR
(ISM 4)
2.59
0.82, 5.30
1.07 ±0.20
-0.51 ±0.60
0.92
0.87
(0.90)
0.33 (13%)
2-8
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
For PMC at Phoenix (Table 2-2) the mean ratio of 0.97 for gravimetric mass
demonstrates that there is little mass bias between the samplers. Mean ratios for the crustal
species are consistently in the range 0.95-0.97, which shows that, on average, Dichot B
concentrations are a few percent lower than Dichot A concentrations for these species. In
contrast, the mean ratio is 1.23 for OC, which shows that, on average, Dichot B concentrations
are 23% higher than Dichot A concentrations. Collocated precision, defined as the root mean
square (RMS) of the Dichot A-to-Dichot B concentration differences divided by the square root
of 2, is in the range 12% to 19% for the PMC components, with the highest value observed for
OC. For PMf at Phoenix (Table 2-3) the mean ratio of 1.09 for gravimetric mass demonstrates
some bias, with Dichot B reading higher than Dichot A. This pattern is observed for all of the
major crustal species with mean ratios ranging from 1.08 (Fe) to 1.26 (Ti). Similar to PMC, the
PMf collocated precision estimates are in the range of 10% to 20% for each of the reported
species.
Collocated precision for the St. Louis data was, with a few exceptions, similar to the
Phoenix data. Precision was poorer in St. Louis for PMC OC-TOR and PMf titanium, and better in
St. Louis for PMf gravimetric mass.
Measurement error for the collocated dichot data collected at Phoenix was further
examined by calculating the concentration differences between samplers (expressed as the
natural logarithm of the concentration ratio) for all species. Figure 2-5 shows scatterplots for the
concentration differences between dichot samplers for the elements used in the soil equation,
XRF sulfur, and gravimetric mass. For PMC (Figure 2-5a), all of these species combinations
show strong correlation. Interpreting the concentration difference between dichot samplers to be
measurement error, these strong correlations mean that these species share the same
dominant source of measurement error (Hyslop and White, 2011), and this measurement error
leads to the small but consistent deviation in the mean ratios from unity for these species shown
in Table 2-2. One possible source of such measurement error could be functional differences
between the dichots in performance of the PM10 size-selective inlets.
2-9
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
(a) PHX PMC (b)PHXPMf
ACa
ACa
AFe
/
AFe
/r
!
-
3
ASi
r
•
r
t
ASi
'•*'» i"
O" '>:•
.. i
ATI
&
.r
»
< r
*
:f '
s s
ATI
* - , ¦=
:t
*
AS
13*
.i*
¦>%
%
s
* %i:
* »*<.
$
<• > «
AS
.>>•
or,''-
«a
.. ^
. ^
Amass
'.I,
,
'A
i-;£
Amass
s*
'*
.Ml ACa AFe .ASi Alt AS Ml ACa AFe .Si ATi AS
Figure 2-5. Scatterplot matrices of the observed differences, a* = In (CxB / CxA), in the
collocated concentrations at Phoenix: (a) PMC and (b) PMf.
For comparison, Figure 2-5b shows scatterplots for the observed concentration
differences for PMf from the collocated dichots at Phoenix. Ca, Fe, Si, and to some extent sulfur
(S) show strong correlation. Correlations for Al and Ti with the other species are weaker; this
weaker correlation might arise from the relatively greater contribution of analytical errors to the
overall measurement errors for these species, which are at relatively low concentrations in the
PMf size range. Another potential source of measurement error is differences in the cutpoint
curves between virtual impactors in the two dichots. This error could explain the Dichot
B-to-Dichot A ratios for gravimetric mass and crustal species being less than one for PMC and
greater than one for PMf (Tables 2-2 and 2-3). In this case, for a given species, the
concentration differences between dichots should be anti-correlated for the PMC and PMf size
ranges.
Figure 2-6 shows the observed differences for the major soil species. In each case, the
correlation is very weak, and thus, if there are differences in virtual impactor performance
between the two dichot samplers, the effect on concentration is masked by other measurement
errors.
2-10
-------
EPA's Coarse PM Pilot Study
Gravimetric Mass and Composition Overview
AAI
ACa
AFe
ASi
ATi
AS
PMf
%
o
°o °o
o o
. o
^ o o0 0 0
0
°cP 0
0
° 0 4
0 0
0
0
A 0O 0 O
O 0
Q
°o°°®
0
0
0
0 0 0 0
« 0% 0
S^o
O
0
O (5> O
O
o° 0 0
0 00
0 0
0 ?
O 0
0
0 0
f° 0
0 0
0 0
O <5b
0 0
0 0 0 0
0
PMC
Figure 2-6. Scatterplot vector of the observed differences, a* = In (CxB / Cx 4), across
the fine and coarse PM size fractions in the collocated concentrations at Phoenix.
The average and median concentration ratios between collocated dichots by species,
site, and size (fine or coarse) are summarized in Appendix B. Figure 2-7 shows the average
Dichot B-to-Dichot A ratio by species for both sites and size fraction, using only data above the
MDL. For most species, the average ratio is near unity. One noteworthy exception is
ammonium, which has low abundance in PMC and has high uncertainty from the correction for
fine particle intrusion into the dichot minor flow channel.
CH 01 \-
O O o o
I— I- H i-
ll
o o o o
LU LU O O
O o) —
I— < <
o
2 "I
o o
w (/) I T CO
=3NN
o
I
-s-
c O i_
ro
CC
o
Q
5a
o
.c
o
Q
0
O)
2
0
H
9'
X
+
A Phoenix, PMc
Phoenix, PMf
• STL, PMc
X STL, PMf
* w
• *
~ ~
A
+
•
"I i
X,
¥*
¥
•
A
+ * +$ + + + + *
* jJ.i
+ X # • A *
+
x*
X
i
X
+
X
« *
A*X
+
A
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
CCHCCHC D)^= <5 roir TOT3 0 o O >7 GO =3 0 D)CQ II Q 7 Q Q. il i (» i 0 (J. C O , n iZ > C
nnnnn<<-
-------
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
• viii,»/ >is of Elements
Species that were typically above the MDL in the PM2.5 CSN were generally above the
MDL in PMC (Section 3.1), including the crustal species, OC and EC, sulfate, nitrate, and metals
such as barium (Ba), chlorine (CI), copper (Cu), potassium (K), manganese (Mn), sodium (Na),
and zinc (Zn).
Unlike PM2.5, ammonium was above the MDL only 60% of the time in St. Louis and 50%
of the time in Phoenix, reflective of the low ammonium concentrations in PMC. Agreement of
individual species' PMC concentrations between dichot and FRM measurements (Section 3.2)
were generally similar to those seen for gravimetric mass (Section 2.1).
At St. Louis, dichot and FRM measurements compared fairly well for both fine and
coarse fractions. At Phoenix, there was a consistent bias for PMC, with FRM concentrations
higher than dichot concentrations for most species, especially the crustal elements. Given that
crustal oxides make up the majority of the PMC mass, and that FRM PM10-2.5 measurements of
these species were consistently higher than dichot PMC measurements, the sensitivity of these
species' measurements to the assumptions in the XRF method was examined (Section 3.3).
Attenuation factors are negligible for PM2.5 measurements; however, attenuation factors
must be applied for PMC measurements. The attenuation factors used in this study for PMC were
further investigated for a subset of samples. Results from ICP-MS analysis were compared to
XRF results for a subset of samples (Section 3.4), which further indicated that the coarse
particle concentrations measured by blank-corrected ICP-MS were greater than those
measured by XRF. This pattern does suggest that the XRF attenuation corrections for these
constituents are too large, although more is needed to verify these preliminary findings.
Lastly, Section 3.5 shows that there was little daily variation in the crustal composition,
though the composition was somewhat different at the two monitoring sites.
3.1 Concentrations by Site and Size Fraction
Similar to CSN measurements, many species that were analyzed were often below
detection. Twenty species were above the MDL more than 80% of the time at both sites.
Appendix A summarizes average concentration and MDL, fraction of samples above MDL and
3*MDL, and the average uncertainty by species, site, and size (fine or coarse) for species
analyzed from the Teflon and quartz filters. Elements were analyzed by XRF, and ions were
analyzed by IC. Potassium and sodium were analyzed by both methods. Figure 3-1 shows the
fraction of samples above the MDL for dichot PMC. Species were detected at similar rates at
both sites, with a few exceptions: the detection rates for cobalt (Co), chromium (Cr), rubidium
(Rb), and strontium (Sr) were higher at Phoenix. This is consistent with the higher PMC crustal
concentrations in Phoenix compared to St. Louis, but might also reflect differences in species
abundance in the crustal material at these sites.
3-1
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
• St Louis
A Phoenix
o
ct:
o
op
A
(/)
C O 1_ Ci-O^
<<NNliJO
LJLLJLJ
1111111 i1111111111411 1111 'I' 111111111 1 lj
-A-
o •
• 00
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
<5raif:raT3!=>i|-|-
<
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Table 3-1. Measures of agreement between the collocated FRM and dichot data for PMC
measured at Phoenix. Slopes and intercepts are reported for reduced major axis
regressions of dichot on FRM.
PHXPMc
(N=35)
FRM
Mean
(pg/m3)
FRM
Min, Max
(Mg/m3)
Slope
+/- 95%
C.I.
Intercept
+/- 95%
C.I.
(pg/m3)
1
D/F
Ratio:
Mean
(Median)
Collocate
d
Precision
(pg/m3)
Grav. mass
22.7
9.6, 54.9
0.64 ±0.10
3.6 ±2.5
0.8
0
0.83
(0.85)
5.1 (25%)
S
0.12
0.04, 0.28
1.11 ±0.27
0.00 ±0.04
0.5
3
1.17
(1.08)
0.03 (24%)
Zn
0.06
0.01, 0.57
0.46 ± 0.04
0.01 ±0.00
0.9
4
0.83
(0.77)
0.04 (82%)
Ca
1.43
0.53, 4.06
0.65 ±0.13
0.22 ± 0.22
0.6
8
0.86
(0.87)
0.40 (31%)
Fe
0.99
0.35, 2.34
0.71 ±0.12
0.10 ± 0.13
0.7
9
0.85
(0.85)
0.23 (26%)
Si
4.23
1.59, 11.26
0.65 ±0.14
0.65 ± 0.65
0.6
4
0.85
(0.87)
1.10 (29%)
Table 3-2. Measures of agreement between the collocated FRM and dichot data for PMf
measured at Phoenix. Slopes and intercepts are reported for reduced major axis
regressions of dichot on FRM.
PHX PMf
(N=35)
FRM
Mean
(pg/m3)
FRM
Min, Max
(pg/m3)
Slope
+/- 95% C.I.
Intercept +/-
95% C.I.
(pg/m3)
a
D/F Ratio:
Mean
(Median)
Collocated
Precision
(pg/m3)
Grav. mass
8.4
3.5, 19.6
0.98 ± 0.08
-0.5 ±0.77
0.94
0.92 (0.93)
0.8 (10%)
S
0.31
0.09, 1.02
1.03 ±0.04
-0.01 ±0.01
0.99
0.99 (1.01)
0.01 (3%)
Zn
0.015
0.003, 0.047
1.00 + 0.10
0.000 + 0.002
0.91
1.00 (0.95)
0.002 (14%)
Ca
0.18
0.04, 0.63
0.85 ±0.10
0.00 ±0.02
0.88
0.88 (0.87)
0.04 (24%)
Fe
0.24
0.04, 0.67
0.92 ±0.12
-0.01 ±0.03
0.86
0.89 (0.90)
0.05 (22%)
Si
0.42
0.10, 1.74
0.91 ±0.10
-0.01 ±0.05
0.90
0.91 (0.89)
0.08 (20%)
3-3
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Table 3-3. Measures of agreement between the collocated FRM and dichot data for PMC
measured at St. Louis. Slopes and intercepts are reported for reduced major axis
regressions of dichot on FRM.
STL PMC
(N=18)
FRM
Mean
(pg/m3)
FRM
Min, Max
(Mg/m3)
Slope
+/- 95% C.I.
Intercept +/-
95% C.I.
(pg/m3)
r2
D/F Ratio:
Mean
(Median)
Collocated
Precision
(pg/m3)
Grav. mass
12.3
3.1, 24
0.99 ±0.08
-0.9 ± 1.1
0.98
0.90 (0.93)
1.0 (8%)
S
0.10
-0.13, 0.24
0.42 ±0.19
0.07 ±0.03
0.28
-0.75 (0.84)
0.06 (57%)
Zn
0.025
0.004, 0.136
0.96 ±0.08
-0.001 ±0.003
0.98
1.01 (0.99)
0.004 (17%)
Ca
1.51
0.15, 3.92
0.95 ±0.08
0.02 ±0.15
0.97
0.97 (0.98)
0.13 (9%)
Fe
0.39
0.1, 0.98
0.84 ±0.08
0.03 ±0.04
0.97
0.94 (0.9)
0.05 (13%)
Si
1.14
0.18, 2.75
0.88 ±0.08
0.07 ±0.12
0.97
0.96 (0.96)
0.12 (11%)
Table 3-4. Measures of agreement between the collocated FRM and dichot data for PMf
measured at St. Louis. Slopes and intercepts are reported for reduced major axis
regressions of dichot on FRM.
STL PMf
(N=18)
FRM
Mean
(pg/m3)
FRM
Min, Max
(pg/m3)
Slope
+/- 95%
C.I.
Intercept +/-
95% C.I.
(pg/m3)
D
D/F Ratio:
Mean
(Median)
Collocated
Precision
(pg/m3)
Grav. mass
12.0
6.3, 18.1
1.01 ±0.09
0.0 ±1.1
0.97
1.01 (1)
0.4 (3%)
S
0.88
0.44, 1.36
0.99 + 0.11
0.00 + 0.10
0.95
1.00 (1.01)
0.03 (3%)
Zn
0.019
0.003, 0.067
0.98 + 0.13
0.000 + 0.003
0.930
1.00 (0.95)
0.003 (16%)
Ca
0.09
0.02, 0.20
1.13 + 0.21
-0.01 +0.02
0.87
1.08 (1.00)
0.02 (22%)
Fe
0.09
0.02, 0.24
1.07 + 0.13
0.00 + 0.01
0.95
1.05 (1.00)
0.01 (11%)
Si
0.11
0.02, 0.54
0.96 + 0.08
0.00 + 0.01
0.98
1.04 (0.98)
0.01 (9%)
3-4
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
S, PHX
S, STL
1.2
£ 1.0
"So
c 0.8
o
'+¦>
flj
£ 0.6
0J
u
u 0.4
0.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
FRM Concentration (ng/m3)
OPMc
APMf
0.2 0.4 0.6 0.8 1.0 1.2
FRM Concentration (ng/m3)
Zn, PHX
Zn, STL
E 0.50
OPMc
A PMf
0.00 0.10 0.20 0.30 0.40 0.50 0.60
FRM Concentration (ng/m3)
0.15
/ O
CO
"I0-12
CUD
3
c
•B 0.09
ru
i-
c
01
i 0 06
/
/
/ '
'
/ /
/oV°
U
4-*
o
« 0.03
Q
°PMc
APMf
0.00
i
0.00 0.03 0.06 0.09 0.12 0.15
FRM Concentration (ng/m3)
Figure 3-2. Scatter plots of PMf and PMC concentrations from dichot (y-axis) and FRM (x-
axis) at Phoenix (left column) and St. Louis (right column) for S, Zn, Ca, and Fe. Solid
lines are the 1:1 lines and dashed lines indicate the +1-20% difference between
measurements. (Figure continued on next page.)
3-5
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Ca, PHX
OPMc
A PMf
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
FRM Concentration (ng/m3)
Fe, PHX
O PMc
A PMf
Ca, STL
0.0 0.5 1.0 1.5 2.0 2.5 3.0
FRM Concentration (ng/m3)
O PMc
A PMf
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
FRM Concentration (ng/m3)
Fe, STL
euo
2.
O PMc
A PMf
0.0 0.5 1.0 1.5 2.0 2.5 3.0
FRM Concentration (ng/m3)
Figure 3-2 (continued). Scatter plots of PMf and PMC concentrations from dichot (y-axis)
and FRM (x-axis) at Phoenix (left column) and St. Louis (right column) for S, Zn, Ca, and
Fe. Solid lines are the 1:1 lines and dashed lines indicate the ± 20% difference between
measurements.
3.3 Attenuation of X-Ray Intensity for Light Elements
Measurement error can arise from both sampling errors and analytical errors. A
potentially significant source of analytical error for PMC elements is X-ray attenuation during
XRF analysis (self-attenuation). This error is not captured in the collocated dichot comparisons
and is only indirectly captured in the dichot-to-FRM comparisons through the dependence of the
attenuation effect on the size distribution of collected particles.
3-6
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
RTI applied attenuation factors (A), calculated using proprietary software, to estimate
attenuation-corrected mass loadings (ml) from the XRF instrument-reported mass loadings
(IT) no corrj) ¦
m
nocorr.i #«— #> -¦ \
m = — (Eq. 3-1)
Attenuation factors are in the range A< 1, and they vary with both element and PM size
range (PMf, PMC, PM10, and PM2.5). For A = 1, there is no attenuation correction, while for A < 1,
the attenuation-corrected mass loading is higher than the instrument-reported mass loading.
These adjustments are made to the "raw" mass loadings on the filter reported by the XRF. RTI
provided A for a subset of PM10 FRM samples and dichot major flow channel (PMf filter) and
minor flow channel (PMC filter) samples. PM2.5 FRM attenuation factors are identical to the
dichot PMf attenuation factors. For PM2.5 FRM, PM10 FRM, and dichot PMf samples, the
attenuation factor linearly propagates through the calculation of ambient concentrations. For
dichot PMC, however, the dichot minor flow attenuation factors do not linearly propagate through
the calculation of ambient calculations because the dichot minor flow channel mass loadings are
corrected for fine particle intrusion.
Attenuation factors are specific to each sample (Gutknecht et al., 2010) and are
automatically generated and applied by the signal processing software. RTI provided the
uncorrected mass loadings for a subset of samples and back-calculated the sample-specific
attenuation factors. The results are summarized in Table 3-5. Figure 3-3 demonstrates that
samples selected for the dichot minor flow channel were well distributed over the observed
range of gravimetric mass loadings. Attenuation factors were unity for all elements with atomic
numbers (Z) greater than 20. For each element and sample type, the standard deviation of the
sample-specific attenuation factors was relatively small, so the arithmetic mean attenuation
factor is deemed representative of all samples. Figure 3-4 shows the mean attenuation factors
stratified by element and sample type.
Because of the physics of self-attenuation, coarse particles are affected more than finer
particles. Since PM10 includes both PMC and PMf, the attenuation factors for PM10 should be
intermediate between those of PMC and PM2.5. Thus, for any element, attenuation factors occur
in the order APMC < APM10 < (APM2.5 = APMf). When comparing attenuation results for PMC with
the dichot vs. the FRM (i.e., PM10-PM2.5 vs. PMC [dichot]), the intrusion of PMf into the PMC
deposit must be taken into account.
3-7
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Table 3-5. Average attenuation factors (A) back-calculated by RTI for a subset of
samples.
Element
Z
Dichot Major Flow
Channel (PMf), AM 3
Dichot Minor Flow
Channel (PMC), N=G
PM10FRM,
N=9
N>
MDL
A
N>
MDL
A
N>
MDL
A
Na
11
12
0.96 ± 0.02
6
0.44 ± 0.00
8
0.51 ±0.00
Mg
12
11
0.97 ±0.01
6
0.46 ± 0.00
7
0.54 ± 0.00
Al
13
12
0.98 ±0.01
6
0.51 ±0.00
7
0.58 ± 0.00
Si
14
12
0.99 ±0.01
6
0.57 ±0.00
8
0.61 ±0.00
P(phosphorus)
15
3
0.99 ± 0.00
5
0.70 ±0.00
7
0.84 ± 0.00
S
16
12
0.99 ± 0.00
6
0.85 ±0.00
8
0.96 ±0.01
CI
17
12
0.99 ± 0.00
6
0.79 ±0.00
7
0.85 ± 0.00
K
19
13
1.00 ±0.00
6
0.87 ±0.00
8
0.90 ± 0.00
Ca
20
12
1.00 ±0.00
6
0.86 ±0.00
8
0.88 ± 0.00
--
>20
12
1.00 ±0.00
6
1.00 ±0.00
7
1.00 ±0.00
2
1
O reported with attenuation correction only
• reported with and without attenuation correction
0 100 200 300 400 500 600 700 800
dichot minor flow channel gravimetric mass, j_ig/filter
Figure 3-3. Cumulative distribution of gravimetric mass concentrations for all dichot
sampler minor flow channel (PMC) samples that were analyzed by XRF (open circles) and
the six samples with mass loadings reported by RTI both with and without attenuation
correction (closed circles). The sample shown at the 99th percentile was the sample with
the highest mass loading.
3-8
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
o
03
c
o
V-'
03
3
a
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
CO
> /in
o
10
20
30
40
crustal PMc without attenuation correction, |_ig/m3
Figure 3-5. Estimated crustal PMC mass concentrations with (y-axis) and without (x-axis)
corrections for self-attenuation applied to the dichot sampler data.
3.4 XRF versus ICP-MS Measurements and Implications
To assess the adequacy of the attenuation factors being applied to the RTI XRF results,
filters from 18 sampling events (10 from Phoenix, 8 from St. Louis) were selected for elemental
analysis at RTI using both XRF and ICP-MS. ICP-MS was chosen as an independent "referee"
method because of its sensitivity and immunity to the attenuation effect that is seen in XRF. The
10 sets of filters from Phoenix exhibited a broader range of crustal species concentrations than
those from St. Louis. Samples were analyzed by XRF and then digested for analysis by ICP-MS
using a microwave-assisted, mixed acid extraction method2 to get the more difficult metals into
solution (e.g., Al, Si, Mg, Fe, and Cr).
Figure 3-6 shows the dichot PMC data for Al with A = 0.51 for XRF and no adjustments
to the ICP-MS data. The oblique solid line represents 1:1 agreement between x and y. The
sample within the square was an outlier for many elements and has been excluded from the
remainder of the analysis. Al concentrations by ICP-MS are biased high, and a reduced major
axis (RMA) regression3 of ICP-MS concentration on XRF concentration has intercept
2 Microwave-assisted digestion was performed on the filters using 2.5 mL nitric acid, 2.5 mL hydrochloric acid, and
0.5 mL hydrofluoric acid. (This digestion matrix is more aggressive than the 8% HCI/3% HNO3 matrix recommended
for analysis of PM2.5 samples on Teflon filters (RTI, 2010).) A two-step digestion protocol was conducted using a
CEM MARS5 unit, with the first step a 15-minute ramp to 100°C at 400W, followed by a 20-minute ramp to 200°C at
1600W. The pressure limit was 800 psi for both steps. Deionized water was used to bring the contents to a final
volume of 50 mL. Analysis was conducted using a Thermo X-Series II ICP-MS with a collision cell and cell gas 10%
hydrogen and 90% helium.
3 Reduced major axis regression is an orthogonal regression method. Smith (2009) discusses the use of ordinary
least squares and reduced major axis regressions.
3-10
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
0.88 ± 0.27 |jg/m3, which is statistically greater than zero at the 95% confidence level. This
intercept value implies background contamination for the ICP-MS measurement (as described
below). The slope of 0.75 ± 0.16 is less than unity and implies that the attenuation factor applied
to the XRF data is too small (i.e., the data are overcorrected).
3
O)
ZS.
CO
CL
o
>> 2
_Q
C
o
TO
C
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Table 3-6. ICP-MS analysis of laboratory blanks with the membrane separated from the
support ring (with adhesive) and of field blanks. For each cell, the top row is the median
and the bottom row is the mean and standard deviation. Values are concentrations
reported in |jg/m3foran effective air volume sampled of 24 m3 (16.7 LPM for 24 hours).
Element
Laboratory Blanks (N = 7)
Field Blanks (N = 4)
Membrane
Support Ring
Na
0.003
0.003 ±0.001
0.294
0.271 ±0.041
0.418
0.493 ±0.159
Mg
0.003
0.003 ±0.001
0.428
0.419 ±0.026
0.378
0.379 ±0.116
Al
0.004
0.004 ±0.004
1.152
1.088 ±0.111
1.163
1.406 ±0.525
Si
-0.018
0.004 ±0.081
1.271
1.414 ±0.590
1.660
2.647 ±2.176
P
-0.002
-0.004 ± 0.006
0.251
0.242 ±0.018
0.246
0.253 ±0.018
S
-0.011
-0.011 ±0.005
0.030
0.032 ± 0.004
0.091
0.140 ±0.134
K
0.004
0.006 ±0.005
0.009
0.009 ± 0.002
0.028
0.127 ±0.214
Ca
-0.021
-0.017 ±0.009
0.229
0.210 ±0.040
0.334
0.711 ±0.812
Fe
0.003
0.009 ±0.016
0.007
0.011 ±0.008
0.049
0.246 ± 0.422
Ti
-0.001
0.000 ±0.001
0.013
0.016 ±0.008
0.012
0.012 ±0.002
Zn
-0.001
-0.001 ±0.001
0.027
0.034 ± 0.024
0.025
0.026 ± 0.022
Analysis of a Standard Reference Material (SRM) is commonly used to characterize
recoveries. RTI analyzed five samples (each nominally 10 mg) of NIST SRM 1648a, which is an
urban particulate matter reference material. One sample had anomalous recoveries for several
elements and was excluded from the data analysis. Table 3-7 summarizes the results.
Recoveries were about 90% for several elements of interest to this study, such as Al and Ca.
3-12
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
Table 3-7. Recoveries for urban particle matter NIST standard reference material for the
sample digestion protocol and ICP-MS analysis used in this study. S.E. is the standard
error of the mean.
SRM 1648a Recovery (%)
Element
(N =
4)a
Median
Mean ± S.E.
Na
90
89 ± 1
Mg
91
90 ±2
Al
94
91 ±4
S
88
89 ± 1
K
91
90 ±2
Ca
89
89 ±2
Ti
69
71 ±5
V (vanadium)
79
96 ±27
Cr
73
72 ±5
Mn
81
82 ±3
Fe
91
92 ± 1
Co
81
81 ±2
Ni (nickel)
103
100 ±4
Cu
83
83 ±2
Zn
119
113 ±8
As (arsenic)
103
102 ±2
Se (selenium)
123
116 ±8
Sr
96
95 ±2
Ag (silver)
110
108 ±9
Cd (cadmium)
111
107 ±5
Sb (antimony)
95
91 ±7
Pb (lead)
86
86 ±2
a Excludes one run that exhibited low recoveries for Al, Mg, K, Zn,
and Se, and high recoveries for V and As.
In light of these results, the following approach was taken to evaluate the XRF
attenuation factors. For a given element analyzed by both XRF and ICP-MS:
1. For the dichot minor flow channel filters (i.e., PMC filters), use the XRF attenuation
factors A, reported in Table 3-5 and Equation 3-1 to back-calculate the XRF mass
loadings without correction for self-attenuation, ITIno corr,i-
2. Estimate a revised attenuation factor A! and use Equation 3-1 to calculate revised mass
loadings mi. Calculate XRF PMC elemental concentrations, PMC', using the revised mass
loadings for the PMC filter data. Attenuation factors for the dichot minor flow channel
filters (i.e., PMf filters) are close to unity (Table 3-5) and were not revised. Thus, no
3-13
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
adjustments were made to the PMf concentrations that are used to correct the PMC
concentrations for fine particle intrusion into the dichot PMC filter channel.
3. Blank-correct the ICP-MS mass loadings data using the laboratory blanks results (i.e.,
the mass loadings data underlying Table 3-6). Mean loadings from the support ring and
membrane were summed and subtracted from the dichot PMC and PMf filter mass
loadings.
4. Calculate ICP-MS PMC elemental concentrations using the blank-corrected dichot PMC
and PMf filter mass loadings.
5. Calculate the square difference between the PMC concentration values for blank-
corrected ICP-MS data and XRF data with revised attenuation factors (PMC').
6. Calculate the sum of squared differences (SOSD) over the nine samples and iterate on
the revised attenuation factor A! to determine the best-fit A! that minimizes the SOSD.
7. Repeat the above steps, adding to Step 5 an adjustment for the ICP-MS recoveries
using the mean recoveries data reported in Table 3-7.
Figure 3-7 shows the PMC aluminum concentrations after blank-correcting the ICP-MS
data using the mean laboratory blank value in Table 3-6 (support ring plus membrane) and
using no adjustment for XRF attenuation (A = 1, left panel) and the best-fit XRF attenuation
factor (A = 0.73, right panel). For Al, the best-fit attenuation factor was 0.73, assuming 100%
recovery for the digestion and ICP-MS analysis. This attenuation factor is 43% higher than the
value of 0.51 used by RTI (Table 3-5) and would decrease the RTI-reported PMC aluminum
concentrations by about 30%. If the Al recovery of 91% (Table 3-7) is included in the analysis,
the best-fit attenuation factor is 0.63 and decreases the RTI-reported PMC aluminum
concentrations by about 20%.
Table 3-8 shows the best-fit attenuation factors for the light elements. Best-fit
attenuation factors (assuming 100% recovery for the ICP-MS analysis) are in good agreement
with the RTI-reported values for Na and S, about 5% higher for Ca, and 20% to 45% higher for
(in increasing order) K, P, Si, and Al. The best-fit attenuation factor for K is greater than unity;
this suggests that another source of error is present; such as poor digestion recovery or
calibration bias for XRF or ICP-MS. Silicon has a very low coefficient of linear correlation
(r2 = 0.07) and high RMS error (RMSE; 78%), which makes the results for silicon suspect. The
best-fit attenuation factor for Mg is about 20% lower than the RTI-reported value. When
recoveries for ICP-MS analysis are included, the best-fit attenuation factors are as much as
20% lower than the factors that are estimated assuming 100% recovery, but are still higher than
the RTI-reported factors for Al, K, and Ca (recovery data are not available for Si and P).
3-14
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
4
Al - dichot PMc - PHX
XRF attenuation factor = 0.73
blank-corrected I CP-MS
S
CL
o
>, 2
-Q
c
o
-E
, 2
-Q
c
o
¦E
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
The response surface for the minimization (SOSD versus A) was examined by
calculating the attenuation factors for 110% of the minimum SOSD. The attenuation factor
ranges, shown in Table 3-8 in parenthesis after the best-fit A assuming 100% recovery, have
relatively narrow bounds except for P and Si. Attenuation factors were also calculated by
excluding each data point one at a time ("leave one out" analysis) and minimizing the SOSD. In
this case, the range of attenuation factors for Al was 0.70 to 0.75.
These preliminary results suggest that the RTI-reported dichot PMC filter attenuation
factors for some crustal elements (e.g., Al, Ca, and possibly Si) are perhaps too low and thus
the attenuation-corrected concentrations overestimate the true concentrations. The software
used by RTI to apply attenuation factors should be examined to tabulate the assumptions being
made about particle size distribution and particle composition. These assumptions could be
adjusted within reasonable ranges to assess whether a better fit between XRF and ICP-MS can
be achieved. The comparison between XRF and ICP-MS should be repeated for a larger data
set, with samples collected at sites that have high crustal loadings, and ideally including coarse
PM from different sources, such as desert dust and agricultural dust. XRF should be performed
on each filter, and then the membrane should be separated from the support ring prior to
digesting the particle-laden membrane for ICP-MS analysis. This approach will dramatically
decrease the blank levels and thus reduce the uncertainty introduced by the blank correction.
Samples suitable for particle size analysis, e.g., by electron microscopy, should be collected in
parallel so the assumptions used to develop the attenuation factors can be evaluated.
3.5 Variation in Crustal Composition
The defining feature of the PMC elemental data is the high concentrations of species
associated with crustal material. Since crustal species are the dominant component of PMC
mass, the variation of crustal composition between sites and over the course of the study was
examined. Seasonal variation of crustal composition could be expected if there was influence of
dust from long-range transport (e.g., Asian dust event) rather than locally generated dust, and if
the mix of regional and local dust varied across seasons. Figure 3-8 shows a simple
visualization of the crustal composition, comparing the fraction of Si, Fe, and Ca to the sum of
these elements. At both sites, there is little temporal variability in the ratio of these species.
However, Si accounts for nearly twice the fraction of crustal mass measured at Phoenix
compared to St. Louis, the latter having more mass from Ca. This finding is consistent with likely
different soil composition at the two sites, with St. Louis having karst topography, which could
lead to higher abundance of calcium in the crustal PM profile.
3-16
-------
EPA's Coarse PM Pilot Study
Analysis of Elements
(a) Phoenix
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Ti
I Si
Fe
I Ca
lAI
o o o
r>J J
o o o
CM (N (N
CTi
tH
r>J
i-n ro
CO
r>J
O O
-------
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
4. Analysis of Carbon
After crustal species, carbon was the second largest contributor to PMC mass, on
average about 15% (see Section 2). Unlike crustal species, there is only marginal agreement
between dichot PMC and FRM PM10-2.5 carbon measurements; most points were beyond a ±20%
difference (Section 4.1), though without the bias towards the FRM that was seen with crustal
species and gravimetric mass.
Analysis of OC and EC by thermal fraction indicate that the composition of OC and EC is
different between Phoenix and St. Louis, though there is little difference at each site between
the fine and coarse OC and EC composition by thermal fraction (Section 4.2).
Carbonate was measured for a subset of samples (Section 4.3). PMf carbonate was
below the detection limit for all samples analyzed. In contrast, PMC carbonate was on average
6% of PMC gravimetric mass at Phoenix and 12% of PMC gravimetric mass at St. Louis.
Carbonate carbon was on average 9% of PMC total carbon (TC) at Phoenix and 14% of PMC TC
at St. Louis. Thus, at these two sites, carbonate significantly contributes to gravimetric mass
and total carbon.
Unlike fine PM, OC mass loadings on backup filters for the dichot minor flow (used to
calculate PMC concentrations) are statistically indistinguishable from the trip and field blanks
(Section 4.4). This finding, together with other analyses presented in this section, lead to the
conclusion that there is very little volatile OC in the coarse PM size range, and thus a low
capacity for negative artifacts. Since coarse OC may comprise biological material, 54 samples
for biological analysis were collected between February and May 2011 (see Section 4.5).
Concentrations of biomarkers (proteins, (1,3)-(3-D-glucans, and endotoxin) were relatively low
compared to the OC, with a median PMC glucan concentration of about 0.2 ng/m3, protein
concentrations of about 0.08 |jg/m3 in both Phoenix and St. Louis, and endotoxin concentrations
of 0.017 EU/m3 at St. Louis; endotoxin values at Phoenix were well outside typical variability and
were suspect.
4.1 Total Carbon Comparisons Between Dichot and FRM Samplers
Figure 4-1 shows a comparison of total carbon (TC; sum of OC and EC) as measured in
each size fraction by dichot and FRM. Most of the measurements for PM2.5 at both sites are
within 20% (as indicated by dashed lines on the plot), and are well correlated (r2 varying from
0.91 to 0.96). PM10 measurements similarly have a high correlation (greater than 0.90), and
most measurements are within 20%. However, the dichot PMC and FRM PM10-2.5 measurements
have lower correlations than either PM2.5 or PM10 measurements, with more points beyond
±20% difference. At each site, each of the dichots yields different correlation statistics when
compared with the FRM measurements (i.e., correlation of dichot to FRM varies between 0.64
and 0.84 at St. Louis and between 0.67 and 0.84 at Phoenix, depending on which dichot is
compared to the FRM). However, the striking feature is that PMC total carbon does not exhibit
the dichot-to-FRM bias that is prevalent for PMC gravimetric mass and the crustal species
(Figure 2-1 and Tables 3-1 and 3-3).
4-1
-------
o STL -Collocate
a STL - Primary
123456789 10
PM2,5 FRM, ng/m3
12345678
PM10.2,s FRM, (ig/m3
o STL - Collocate
a STL - Primary
o STL - Collocate
a STL - Primary
4*
I
ro
o PHX- Collocate
a PHX - Primary
123456789 10
PM2.5 FRM, ng/m3
12345678
PM1W, FRM, |ig/m3
o PHX - Collocate
a PHX - Primary
o PHX-Collocate
a PHX - Primary
2 4 6 8 10 12 14 16
PM10 FRM, ng/m3
Figure 4-1. Dichot versus FRM total carbon concentrations for St. Louis (top row) and Phoenix (bottom row) for PM2 5 (left),
PM10-2.5 (center), and PM10 (right). Triangles are data from the primary dichot sampler and circles are data from the collocated
dichot sampler. Diagonal lines are 1:1 (solid) and +20% of 1:1 (dashed).
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
4.2 Thermal Fraction Analysis (OC/EC Split)
In Section 3.6, the variation in PMC crustal composition between sites was examined. In
this section, the variation in carbon fractions across sites and size fractions is examined.
Figure 4-2 shows the carbonaceous PM average concentration values by site, size fraction, and
carbon subtraction from the dichotomous samples. These data have not been adjusted for
carbon artifacts. The average coarse PM total carbon concentration (height of the stacked bar)
is nearly equal to the fine concentration in Phoenix and is less than the fine concentration in
St. Louis. For both fine and coarse PM, the EC and OC concentrations are higher in St. Louis
than in Phoenix.
Figure 4-3 shows the relative distribution of the organic carbon subtractions. For each
size range, the distributions at Phoenix and St. Louis are quite similar. The fine PM has
relatively higher OC1 and OC2 concentrations, whereas the coarse PM has relatively higher
OC3 and OC4 concentrations. This suggests differences in the composition of organic carbon
between the fine and coarse fractions at both sites.
PM Carbon (no artifact correction)
CD
C
o
03
&_
c
<1)
O
c
o
o
<1)
CD
05
&_
<1)
>
05
OC1
OC2
I OC3
I OC4
I EC2
Figure 4-2. Carbonaceous PM average concentration values at Phoenix (N = 30) and St.
Louis (N = 22). No artifact correction has been applied to these data. Concentration
values are stratified by the IMPROVE protocol carbon subfractions, with PCR = pyrolytic
carbon by the reflectance method and EC1* = EC1 - PCR. The average EC3
concentration was zero for all cases and is not shown.
4-3
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
Figure 4-3. Distribution of organic carbon average concentration values at Phoenix
(N = 30) and St. Louis (N = 22). No artifact correction has been applied to these data.
Concentration values are stratified by the IMPROVE protocol carbon subfractions, with
PCR = pyrolytic carbon by the reflectance method.
4.3 Carbonate Concentrations
Carbon can be present in ambient particulate matter as carbonate (CO32). Assuming that
carbonate decomposes at temperatures greater than 800°C, carbonate should not confound
measurement of EC and OC used during TOA temperature protocols with maximum
temperatures below 800°C (Chow and Watson, 2002). The maximum temperature for the
IMPROVE_A protocol is 800°C. Carbonate in geological samples is most commonly present as
calcium carbonate; its presence will bias low the estimate of crustal-derived PM using
conventional soil estimation equations, which assume that calcium is present as calcium oxide
(CaO).
The Desert Research Institute (DRI) used a TOA protocol with sample acidification to
analyze selected quartz fiber filter samples for carbonate. Initially, dichot filters from 15 sampling
events (six from Phoenix and nine from St. Louis) were selected, which reflected a range of
Si/Ca ratios, based on the hypothesis that differences in Si/Ca ratio may be indicative of
differences in PMC composition, and thus the amount of calcium carbonate (CaCC>3) in a given
sample. For PMf, only one sample exceeded the reported carbonate uncertainty value of
0.17 |jgC/m3—a Phoenix sample with PMf carbonate concentration of 0.18 |jg C/m3.4 5 In
4 Carbonate concentrations are reported as carbonate carbon, i.e., |jgC/m3.
5 The 3ct MDL reported by DRI, based on a standard set of laboratory blanks, is 0.93 |jgC/m3, which corresponds to
0.51 |jgC/m3 for PMf and 0.46 |jgC/m3 for PMC. The error structure has the form Unc, = -j(CV ^C,)2 +(MDLI3)2, where
Una is the uncertainty for sample CVis the coefficient of variance from replicate analyses, and C, is the sample
concentration. Thus, at low concentrations, the uncertainty is MDL/3.
4-4
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
contrast, for this relatively small set of samples, the dichot PMC carbonate was on average
0.43 |jg C/m3 for Phoenix and 0.36 |jg C/m3 for St. Louis. Dichot PMC carbonate corrections for
fine particle intrusion into the minor flow channel were at most 5%, so no additional dichot PMf
filter samples were analyzed.6
Subsequently, 12 collocated quartz filters sampling events were selected (six from each
site) with carbonate analysis on both dichot PMC filters; PM10 and PM2.5 FRM filters were
analyzed for six of these samples. Dichot PMC filters from an additional 42 mass balance
protocol sampling events were analyzed (31 from Phoenix and 11 from St. Louis).
Figure 4-4a shows the PMC carbonate data for collocated dichot samples. The
collocated precision was statistically indistinguishable between the sites, with a pooled
collocated precision of 0.053 |jg C/m3 (22% relative precision). This precision is much better
than the laboratory-re ported uncertainty of 0.15 |jg C/m3 for low carbonate concentrations, and
suggests that the reported MDL of 0.46 |jg C/m3 is conservative. Figure 4-4b compares the PMC
carbonate from the dichots (mean of the collocated values) to FRM by difference (i.e., PM10-2.5)
for the six samples for which collocated dichot and FRM filters were analyzed for carbonate.
The precision of 0.072 |jg C/m3 (23% relative precision) is only slightly degraded compared to
the collocated dichots. The ratio of means is 1.00, which demonstrates no bias between the
dichot and FRM carbonate measurements, similar to results for total carbon; this result is in
sharp contrast to gravimetric mass and species measurements from the Teflon filters, which
showed higher loadings for FRM PM10-2.5 than for dichot PMC.
0.7
0.7
• PHX
O STL
• PHX
O STL
0.6
™E 0.6
O
O)
^ 0.5
0.5
£
(C
_a 0.4
0.4
(O
O
<
0.3
_ 0.3
0)
Q_
£
co n 0
CO 0.2
o
Q_
0.2
_£=
O
b
0.0
0.0
0.0
0.2 0.3 0.4
FRM PM10_25 Carbonate, ng C/m3
0.5
0.7
0.0
0.1
0.2 0.3 0.4 0.5 0.6
Dichot Sampler B Carbonate, ng C/m3
0.7
(a) (b)
Figure 4-4. PMC carbonate from (a) collocated dichots and (b) mean collocated dichot
versus FRM-by-difference.
6 For the remainder of this section, dichot PMC carbonate data do not include a correction for fine particle carbonate.
4-5
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
PMC carbonate from the mass balance protocol sampling events was compared to PMC
mass, carbon (TC, EC-TOR, and OC-TOR), major crustal species (Al, Ca, Fe, Si) and S.
Correlation coefficients (r) were statistically significant at the 95% confidence level (CL) for all
comparisons except EC in Phoenix and OC and S in St. Louis. However, at both sites, the
carbonate was most highly correlated with PMC Ca (r= 0.85 for Phoenix, r= 0.97 for St. Louis);
in Phoenix, none of the other comparisons exceeded r= 0.7.
Figure 4-5 shows the relationship between PMC carbonate and PMC calcium on a molar
basis; data shown in Figure 4-5 include 37 samples from Phoenix and 20 samples from St.
Louis. Carbonate can on average explain two-thirds of the PMC calcium, and there are virtually
no samples with a carbonate-to-calcium molar ratio greater than one. For the subset of these
sampling events with FRM PM10-2.5 data available, PMC carbonate can explain 67% of the
PM10-2.5 Ca in St. Louis (N = 12), but only 52% of the PM10-2.5 calcium in Phoenix (N = 24). With
the majority of calcium in CaCC>3 rather than CaO form, using the typical equation for calculating
crustal mass (see Section 6) likely underestimates the contribution from crustal material.
However, this result should be qualified by the differences in particle losses between the dichot
Teflon and quartz filters used to measure calcium and carbonate, respectively, since a larger
correction of fine particle intrusion is used for calcium than for carbonate.
The grand mean PMC carbonate concentrations were 0.23 |jg C/m3 for Phoenix (N = 43)
and 0.27 |jg C/m3 for St. Louis (N = 26). The molecular weight of carbonate is five times as
great as the molecular weight of carbon. For those sampling events with PMC mass available,
the carbonate mass was, on average, 6% of the PMC gravimetric mass in Phoenix (range 0% to
20%, N = 37) and 12% of the PMC gravimetric mass in St. Louis (range 0% to 21%, N = 20).
100
a)
IS
o PHX
e so
£
rc 40
S«b
100
PMc calcium (nmoles/m3)
(a)
PMc calcium (nmoles/m3)
(b)
Figure 4-5. Relationship between dichot PMC carbonate and dichot PMC calcium
expressed as molar concentrations in (a) Phoenix and (b) St. Louis.
4-6
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
Dichot PMC carbonate mass was on average 5% of the FRM PM10-2.5 mass in Phoenix (N = 24)
and 11% of the FRM PM10-2.5 mass in St. Louis (N = 12).
These results demonstrate that carbonate can be a major contributor to PMC mass. For
these data, PMf carbonate was negligible, and measurement of carbonate on the dichot minor
flow channel filter is adequate to characterize PMC carbonate; i.e., no correction for fine particle
intrusion appears necessary. While pure calcium carbonate decomposes at temperatures above
800°C, analysis conditions and matrix effects with other PM constituents might cause the
carbonate carbon to decompose at lower temperatures (Chow and Watson, 2002). Thus, given
the relatively large ratio of PMC carbon present as carbonate to PMC carbon present as EC and
OC (averages of 9% in Phoenix and 14% in St. Louis), more work is needed to evaluate
whether carbonate evolves during TOA by the IMPROVE_A protocol and thereby biases high
the EC, OC, or both.
4.4 Carbon Artifacts
Filter-based sampling to quantify organic carbon in ambient particulate matter is prone to
measurement artifacts. Positive artifacts can arise from the adsorption of vapor onto the sample
(including the quartz filter), while negative artifacts can arise from the volatilization of collected
particulate matter. The two major PM2.5 monitoring networks deployed in the United States take
different approaches to addressing these artifacts.
In the past, IMPROVE network OC data were adjusted on the assumption that the OC
measured on backup quartz filters at a subset of sites is representative of the positive artifact at
all sites in the network. Adjustments were derived and applied on a monthly basis. Although
field blanks, trip blanks, and backup quartz filter blanks are collected by the CSN, the OC data
currently reported for the CSN network are not adjusted for artifacts; the trip blanks, field blanks,
and backup filter data are available to the user through EPA's AQS data base. A workgroup of
EPA and IMPROVE technical experts was convened to explore OC artifact corrections in both
the CSN and IMPROVE monitoring networks. The initial recommendation based on a series of
exploratory analyses is to use monthly network-wide quartz filter field blanks for adjustment.
Field blank filters are collected at all CSN sites and housed in the sampler for the entire duration
of the sampling event. The field blanks are less variable over time and space, decrease the
additive artifact, and do not over-correct by including the multiplicative factor. At the finalization
of this report, the IMPROVE network has begun implementing these recommendations and the
CSN plans to begin January 2015.
This section summarizes an analysis of carbon artifacts from this study. All samples
collected for carbon analysis used a Q/Q filter sandwich. A subset of samples was analyzed for
chemical speciation with carbon analysis performed at the DRI using the IMPROVE_A thermal-
optical analysis protocol. This analysis focuses on organic carbon measured by optical
reflectance (OC-TOR).
4-7
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
4.4.1 OC Trip Blanks and Field Blanks
OC loadings on trip blanks and field blanks provide a context for characterizing and
interpreting OC on the front and backup filters. The blanks data were examined and the key
results are summarized in this section, with additional details provided in Appendix D. Seven
sets of trip blanks were analyzed by OC-TOR. Each set included two to four filters per site. The
May 2010 data were excluded from the analysis because of data quality issues (discussed in
Appendix D); after this date, there was evidence that the trip blanks data could be pooled across
all samples and sites. The pooled trip blanks data have mean and median OC mass loadings of
5.4 ± 5.7 and 4.2 |jg/filter, respectively (N = 44). These mass loadings are the same as the
estimated IMPROVE_A analysis MDL for OC of 5.3 |jg/filter.7
The same statistical analyses were performed for the seven sets of field blanks that
were collected at the same times as the trip blanks. Each set included four to twelve filters per
site, with the maximum case being front and back filters in the PM2.5 FRM, PM10 FRM, and both
channels of both dichot samplers. While in principle the sampler type and filter position (front
versus back) could be treated as additional factors, their effects were deemed inconsequential,
and samples pooled across these factors were treated as pseudo-replicates. As described in
Appendix D, the site had a statistically insignificant effect, whereas the date had a statistically
significant effect on OC mass loadings, even after removing the May 2010 data. The pooled
field blanks data have mean and median OC mass loadings of 6.6 ± 5.5 and 5.2 |jg/filter,
respectively (N = 130).
Our interpretation of the blanks data concludes that (1) in May, which was a one-month
intensive to shake down the field operations prior to the start of routine sampling in June, there
were anomalously high trip blank and field blank mass loadings; and (2) for subsequent months,
the trips blanks mass loadings were statistically indistinguishable, while the field blanks mass
loadings decreased as the study progressed. The temporal behavior confounds the use of the
95th percentile OC mass loading as a robust estimate of the lower quantifiable limit (LQL) but it
can be used as a conservative estimate. Excluding the May 2010 samples, the 95th percentile
trip blank and field blanks mass loadings are each 19 |jg/filter. The May 2010 measurements
are not included in the results presented in the remainder of this section.
4.4.2 OC Mass Loadings on the Backup Filters
Backup filter OC mass loadings can arise from contamination and vapor adsorption
during sampling. Adsorption can be from "native" vapor (semivolatile OC entering the sampler in
the vapor phase) or semivolatile OC volatilized (desorbed) from particles on the front filter as a
result of the filter face velocity and pressure drop across the filter during sampling. PM2.5 and
PM10 backup filters are exposed to the same flow rate (16.7 LPM) with incrementally more
particulate matter OC on the PM10 front filter than on the PM2.5 front filter. Figure 4-6 shows the
OC mass loading on the back filter for these two samplers. For both Phoenix and St. Louis, the
data are scattered about the 1:1 line. The median PM2.5/PM10 OC ratios for the backup filters are
7 Carbon analysis MDL values for this study were estimated by assuming the same MDL as the CSN network when
expressed as |ig/cm2 filter area. Reported CSN network MDL values are 1.5 |ig/filter for OC and 0.42 |ig/filter for EC
and were scaled using filter diameters of 25 mm for the CSN network and 47 mm for this study.
4-8
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
nearly equal to unity (1.02 and 1.10 for Phoenix and St. Louis, respectively). Nearly all OC mass
loadings are greater than the 95th percentile trip blank, and several OC mass loadings are
greater than the 95th percentile field blank. Based on these lines of evidence, there is no
compelling evidence for coarse particle volatilization. In particular, the PM2.5 and PM10 FRM
backup filters have nearly identical OC mass loading, which means that coarse particle
volatilization from the PM10 front filter is likely negligible.
PM25 FRM backup filter OC, ug/filter
PM10 FRM backup filter OC, ug/filter
40
30
20
10
PHX
STL
0
0
10
20
30
40
dichot major flow backup filter OC, ug/filter
Figure 4-6. OC mass loadings on backup filters: (a) PM2 5 FRM versus PM10 FRM;
(b) dichot major flow versus PM2 5 FRM; and (c) dichot minor flow versus dichot major
flow. Data from May 2010 are excluded.
Filters in the dichot major flow channel are exposed to 90% of the flow rate and 90% of
the fine particles sampled by the PM2.5 FRM. Figure 4-6b shows the OC mass loading on the
backup filters for these two samplers. Data are widely scattered, with median dichot/PIVh.s OC
ratios of 0.81 for Phoenix and 0.99 for St. Louis. There appears to be, on average, modestly
4-9
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
less OC mass on the dichot major flow channel backup filters compared to the PM2.5 FRM
backup filters. This pattern could arise from less adsorption of native vapor at the lower flow rate
of the dichot major flow channel (the flow rate is 10% lower than the PM2.5 FRM flow rate), or
less volatilization of fine PM from the dichot front filter compared to the PM2.5 FRM front filter
(the dichot major flow channel collects only 90% of the fine PM entering the sampler), or both.
More data are needed to determine whether the suggested patterns are real.
Figure 4-6c shows the OC mass loadings on backup filters for the dichot minor flow
(used to measure coarse PM) against the dichot major flow (used to measure fine PM). There is
virtually always more OC on the backup filter for the major flow compared to the minor flow. The
excess at St. Louis is greater than the excess at Phoenix, with median major/minor OC ratios of
1.8 for Phoenix and 2.6 for St. Louis. Again, it is not clear whether the trend occurs from higher
adsorption of native vapor at the nine-fold higher flow rate for the major flow, or from particle
volatilization from the fine particles which are distributed 90% to the major flow and 10% to the
minor flow, or both.
4.4.3 Comparison of OC Blanks, Front Filters, and Backup Filters
Enhanced OC on the major flow backup filter compared to the minor flow backup filter is
consistent with measurements conducted at each site using a dichotomous TEOM with FDMS
modules (Section 7). For mid-to-late summer through mid-winter, the TEOM data exhibited
FDMS volatile component concentrations for the major flow that were 3.3 times (Phoenix) to
4.3 times (St. Louis) higher than measured for the minor flows. For both sites, the major-to-
minor flow TEOM-derived volatile OC enhancement is approximately 1.7 times greater than
major-to-minor flow OC enhancement on the dichot sampler backup filters. The FDMS volatility
component is an upper bound on the semivolatile losses from the integrated sampler filter, and
this analysis suggests that fine PM on the integrated sampler front filter could be volatilizing with
at least a portion of the evolved OC adsorbing onto the backup filter. For the data collected over
the mid-to-late summer through mid-winter period, there was relatively little nitrate, so the
semivolatile fine PM is assumed to be dominated by OC.
Figure 4-7 shows box-whisker plots for OC mass loadings on trip and field blanks, the
dichot major and minor flow backup filters, and dichot major and minor flow front filters. The
minor flow backup filter mass loadings are statistically indistinguishable from the trip and field
blanks loadings (95% confidence level), and the backup filter loading is more than 7% of the
corresponding front filter loading for 25% of the samples. Major flow backup filter mass loadings
are statistically higher than the trip and field blanks loadings (95% confidence level), and the
backup filter loading is more than 17% of the corresponding front filter loading for 25% of the
samples. These trends reaffirm that, at least for these two sites, backup filter measurements
have more OC in fine PM, but might not bring added value to characterizing OC in coarse PM.
Figure 4-8 shows the OC mass loadings relationships for the paired front and back
filters of the major and minor flows. Solid and dashed horizontal lines are the 95th percentile trip
blanks and field blanks OC mass loadings, respectively. As previously discussed, the field
blanks mass loadings decreased as the study progressed, and the 95th percentile value was not
a good representation of the LQL. Thus, the trip blanks 95th percentile mass loading was used
as an LQL proxy. Figure 4-8a shows that most of the major flow backup filter OC mass loadings
4-10
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
are above the trip blanks-based LQL. Front and backup filter mass loadings are weakly
correlated (0.29). In contrast, a recent analysis of the URG-3000N sampler data for Missouri
PM2.5 CSN sites shows a stronger positive correlation between the front and backup filter mass
loadings (not shown). OC mass loadings on the front and backup filter might positively correlate
if semivolatile OC vapor concentrations increase with increasing front filter OC particulate matter
loading, or if more particulate matter OC volatilization occurs at increasing front filter OC
particulate matter loading, or both. Front and back filter mass loadings for the minor flow are
uncorrelated (-0.07), consistent with the backup filter mass loadings being indistinguishable
from the trip and field blanks.
Figure 4-7. OC mass loading distributions (jjg/filter) for the dichotomous samples. The
interior solid blank line is the median and the interior dashed red line is the arithmetic
mean. Whiskers are 10th/90th percentiles, and closed circles are 5th/95th percentiles.
Values in parentheses above each box are the number of samples in the respective
distribution. Data from May 2010 are excluded. Mass loadings were averaged for days
with collocated carbon sampling.
4-11
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
0
(a) major flo
O
0
O
O
-©
-0 ^
33 0
0 0 0
O 0 S?o° 0
(gp0^ 0
0
O
20 40 60 80 100 120
front filter OC, ug/filter
140
50
40
30
o
o
o
ra
_Q
10
(b) minor flow
O
0 O
O u
O
0
0 0
0
0 0
3 O O 0° 0
0
20 40 60 80 100 120
front filter OC, ug/filter
140 160
Figure 4-8. Dichotomous sampler OC mass loadings (jjg/filter) for paired front and back
filters: (a) major flow and (b) minor flow. The horizontal lines are the 95th percentile trip
blanks (solid) and field blanks (dashed).
As shown in Table 4-1, mean OC mass loadings on the dichot minor flow backup filters
are very similar to the mean field blank mass loadings. In contrast, mean OC mass loadings on
the dichot major flow backup filters are higher than the site-specific mean field blanks by 1.7
times for Phoenix and 2.5 times for St. Louis.
Table 4-1. Mean mass loadings on the field blanks and dichot backup filters.
Mass loadings are in jjg/filter.
Mean Mass Loading Item
Phoenix
St. Louis
Field blanks
8.1
6.0
Dichot minor flow backup filters
7.8
5.7
Dichot major flow backup filters
13.6
15.1
Figure 4-9 shows OC concentration scatter plots for collocated dichot samplers at each
site, excluding the May 2010 data. Collocated precision with no artifact correction is 0.44 |jg/m3
(14% of the average OC) for dichot PM2.5 and 0.61 |jg/m3 (23% of the average OC) for dichot
PMC. Future work could include examining the sensitivity of collocated precision to different
forms of the artifact correction. Backup filters are collected with every example, so it would be
possible to evaluate whether the collocated precision improves or degrades when using sample-
specific artifact corrections versus a single-valued correction derived from an ensemble of
backup filters or field blanks. In principle, it would be possible to examine whether the collocated
precision can be explained by the variation in the field blanks, but this is confounded by the
apparent changes in field blank OC mass loadings as the study has progressed. For dichot PMC,
there are additional considerations. The minor flow collocated precision could be calculated and
compared to the PMC collocated precision to determine the influence from applying the
correction for fine particle intrusion. However, as demonstrated in Figure 4-9b, analysis of the
dichot PMC collocated OC data is confounded by an apparent bias between the samplers; such
bias was also observed for crustal species.
4-12
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
8
6
o
o
Q_
< 4
o
o
~o
~o
Q)
ro
o
o
2
o
o
PHX
STL
0
0
2
4
6
8
8
E
O)
=5
d 6
o
in
oi
o
Q_
4
<
O
JZ
o
~o
~o
Q)
2
03
o
o
o
o
PHX
STL
0
0
2
4
6
8
primary dichot (B) PM25 0C, ug/m3 primary dichot (B) PM10_25 0C, ug/m3
Figure 4-9. OC concentrations for collocated dichot samplers: (a) PM2.5 and (b) PMC.
Data from May 2010 are excluded.
4.4.4 Summary of Carbon Artifacts
Based on TOA analysis of samples, the following conclusions are drawn. First, as
described in Appendix D, OC mass loadings for the field blanks decreased as the study
progressed, and the most recent mass loadings were comparable to the trip blanks. Second,
OC mass loadings on backup filters for the dichot minor flow are statistically indistinguishable
from the trip and field blanks. This finding, together with other analyses presented in this
section, lead to the conclusion that there is very little volatile OC in the coarse PM size range,
and thus a low capacity for negative artifacts. In contrast, fine PM exhibits enhanced OC mass
loadings on the backup filters. The study design cannot distinguish whether these higher
loadings are from adsorption of native vapors or adsorption of OC that volatilized PM on the
front filter.
4.5 Biological Data
In addition to the routine one-in-three day sampling, additional dichot samples were
collected between routine sampling days and analyzed for biomarkers. The data set includes 22
sampling events in Phoenix and 19 sampling events in St. Louis over the period from February
to May 2011. Four field blanks and two trip blanks were also taken. RTI analyzed Teflon filters
from the dichot minor flow channel (i.e., coarse particles with no correction for fine particle
intrusion) for gravimetric mass, (1,3)-p-D-glucans, endotoxins, and proteins. RTI used water to
extract the samples prior to using the assays specific to each biomarker.
• The (1-3)-p-D-glucans are found in the cell walls of many types of fungi. While this
constituent is not unique to fungi, it is commonly used as a proxy to evaluate
spatiotemporal patterns in fungal concentrations. The (1-3)-p-D-glucans were measured
using Glucatell®, a commercially available assay.
4-13
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
• Endotoxin was measured by the sample reaction with Pyrochrome®, a commercially
available Limulus Amebocyte Lysate (LAL) assay.
• Protein, an indirect marker of total biological load, was measured using the Molecular
Probes® NanoOrange® Protein Quantitation Kit. Seven of these sampling events
included the collection of collocated samples (four at Phoenix, three at St. Louis). For
these seven sample pairs, the collocated precision was 9% for dichot minor flow channel
mass concentration (PMC not corrected for fine particle intrusion).
RTI analyzed samples in two batches, with the field blanks and trip blanks data
corresponding to each batch (Table 4-2) used to blank-correct the data. Blank values were
converted to concentration units using the target air volume sampled of 24 m3. Field blanks and
trip blanks were statistically indistinguishable at the 95% confidence level. Samples with
negative biomarker concentration after blank correction were deemed to be below the
operational detection limit. Table 4-3 summarizes the glucan and endotoxin data by batch and
also pooled across the two batches. Figure 4-10 shows the concentration distributions for these
data. Protein was excluded from these summaries for reasons described below.
Table 4-2. Biomarker field and trip blank data summary. Each batch included two field
blanks and one trip blank. Field blanks and trip blanks were statistically indistinguishable
at the 95% confidence level. Concentration units were calculated using a total air volume
sampled of 24 m3. ND = not detected.
Analysis Batch
(Sample Dates)
Mean ± 1a
Glucans,
ng/m3
Endotoxin,
EU/m3
Protein,
/jg/rn3
Batch 1 (2/3/11 to 3/18/11)
0.0056 + 0.0003
ND
0.14 + 0.02
Batch 2 (3/23/11 to 5/20/11)
0.0071 +0.0019
ND
0.07 + 0.05
Pooled
0.0063 + 0.0015
ND
0.10 + 0.05
Table 4-3. Biomarker data summary for blank-corrected data. Values are the geometric
mean (GM), with the geometric standard deviation (GSD) in parentheses.
Analysis Batch
Phoenix
St. Louis
N
Glucans,
ng/m3
Endotoxin,
EU/m3
N
Glucans,
ng/m3
Endotoxin,
EU/m3
Batch 1 (2/3/11 to
3/18/11)
10
0.18 (1.3)
2.8 (2.1)
10
0.17 (1.5)
0.04 (2.6)a
Batch 2 (3/23/11
to 5/20/11)
12
0.16 (2.5)
0.30 (2.3)
9
0.21 (1.8)
0.13 (2.8)
Pooled
22
0.17 (2.0)
0.81 (4.0)
19
0.19 (1.6)
0.07 (3.1)a
a Endotoxin was not detected in one sample at STL. A concentration value of 0.01 EU/m3 was imputed, which
corresponds to one-half of the lowest observed concentration.
4-14
-------
EPA's Coarse PM Pilot Study
Analysis of Carbon
0.6
(b) endotoxin (EU/m )
o. /
0.5
0.4
0.3
0.2
PHX-2
ST)
I— EE H
0.0
(a) glucan (ng/m3)
PHX STL PHX-1 PHX-2 STL
Figure 4-10. Box-whisker plots by site of (a) glucan (ng/m3) and (b) endotoxin (EU/m3).
Endotoxin data at Phoenix are stratified by analysis batches (PHX-1, February 3 to March
18, 2011; and PHX-2, March 23 to May 20, 2011).
For protein, the collocated precision on blank-corrected data was 0.07 |jg/m3 (70%) for
the four sample pairs that had both samples above the operational detection limit. The protein
data have high uncertainty because the blanks correction is relatively large. Only six samples
(two in Phoenix, four in St. Louis) had raw protein loadings more than three times the pooled
blank level, and at both sites only about 40% of the raw protein loadings were more than two
times the pooled blank level. Geometric means and standard deviations cannot be reliably
computed because the results will be very sensitive to the method used to impute values that
are negative (i.e., smaller than the blank correction). The median blank-corrected protein
concentrations are 0.18 |jg/m3 for Phoenix and 0.07 |jg/m3 for St. Louis; these concentrations
are close to the blank values (Table 4-2). Given that the protein concentration values are
subject to large uncertainties, further interpretation of these data is not warranted.
Glucans had good detectability, with concentration values for all but one sample at least
ten times higher than the blank correction. Collocated precision on glucan blank-corrected data
was 0.04 ng/m3 (19%). Geometric mean glucan mass concentrations were 0.17 ng/m3 at
Phoenix and 0.19 ng/m3 at St. Louis.
Endotoxin was not detected in the blanks samples. Collocated precision for endotoxin
was 0.51 EU/m3 (99%), including all seven sample pairs, and 0.19 EU/m3 (48%) when excluding
the March 2, 2011, sample at Phoenix. The geometric mean endotoxin level was ten times
higher for Phoenix than for St. Louis, and also exhibited more variability for Phoenix. Samples
were analyzed in two batches, with batch-specific geometric mean endotoxin values for Phoenix
of 2.8 EU/m3 for the first batch (February to mid-March) and 0.3 EU/m3 for the second batch
(late March to May). While the between-batch variability in endotoxin was large, monthly median
PM10 endotoxin concentrations for Fresno, California, reported by Tager et al. (2010) exhibited
similar range and variability. Thus, while the Phoenix endotoxin levels exhibit a large step-
change precisely coinciding with the batches, it is not clear whether this change is real or from
measurement error. While there is not clear evidence that the data are invalid, caution should
be used when interpreting results for Phoenix endotoxin concentrations.
4-15
-------
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
5. Analysis of Ions
This section describes analyses focused on concentrations of ions. Coarse-mode ion
concentrations, including nitrate and sulfate, were very low at both sites throughout the
monitoring period, and much lower than fine-mode concentrations (Section 2 and Section 5.3).
Coarse-mode nitrate appears to be non-volatile compared to fine-mode nitrate, since the
majority of the coarse-mode nitrate was retained on the Teflon filter (Section 5.1). Ammonium
predictions based on anion concentrations were much higher than observed, indicating the
coarse-mode nitrate is predominantly paired with cations other than ammonium (Section 5.4),
though there was no consistent correlation of PMC nitrate with other ions (Appendix E).
Concentrations of both fine- and coarse-mode nitrate were similar with and without a denuder to
remove nitric acid in the sampling stream (Section 5.5). Coarse-mode XRF sulfur predominantly
consisted of sulfate, and coarse-mode XRF sodium predominantly consisted of sodium ion, so
the ion measurements are largely redundant. In contrast, most of the potassium was nonionic
(Section 5.6). Given these findings, laboratory analysis of PMC ions as part of a routine
monitoring network does not appear to justify the necessary costs, except in environments, such
as coastal areas, where coarse-mode nitrate is expected to be significant.
5.1 Approaches to Measuring Ions
Ambient PM2.5 ions are routinely measured in both the CSN and IMPROVE networks. In
both cases, there is a devoted sample collection channel to ions, with a nylon filter and an
upstream acid gas denuder to minimize positive artifacts from the uptake of such gases by the
nylon filter. The filter samples are extracted in water and analyzed by IC. The ion measurements
serve three purposes:
1. They are the only measurement of nitrate (both networks) and ammonium (CSN only),
which can be significant contributors to ambient PM2.5 burdens.
2. Together with the element concentration measurements from the Teflon filter, they allow
for a quality check on the sampling and laboratory analyses.
3. The fraction of a given element that is present in ionic form can provide insights into the
relevant sources.
In contrast to the CSN and IMPROVE networks, which feature a dedicated sampling
channel for ion measurements using a nylon filter, to reduce the total number of samplers
needed for this study, ions were measured using a nylon-behind-Teflon filter sandwich. Ion
analysis was performed on both the Teflon and nylon filters with the ions from nonvolatile salts
(e.g., ammonium sulfate) to be found only on the Teflon filters, and the ions from semivolatile
salts (e.g., ammonium nitrate) to be found on both the Teflon and nylon filters. The nylon filter is
used specifically to measure nitrate and ammonium from the volatilization of ammonium nitrate
from the Teflon filter, although other ions might be present in the nylon filter from chemistry that
takes places on the Teflon filter that displaces ions. In cases where semivolatile forms of nitrate
are low, the nylon filter sampling and analysis might be unwarranted. The Teflon and nylon
filters are in series in the same sampling channel, so although the ion and elemental
measurements (e.g., XRF sulfur and IC sulfate) do not provide a quality check on the sampling,
they do provide a quality check on the laboratory analyses.
5-1
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
5.2 Nitrate Concentrations on Teflon and Backup Nylon Filter
As described in Section 2, PMC nitrate concentrations were generally low. Some forms of
nitrate can be lost or volatilized off of Teflon filters, so backup nylon filters were used to quantify
the amount of nitrate volatilized off of the Teflon filter. Table 5-1 summarizes average
concentrations at each site, size fraction of nitrate by filter, and the average fraction of nitrate on
the Teflon filter. Most of the nitrate is retained on the PMC Teflon filter (71% to 77% on average)
while half or less of the nitrate is retained on the PMf Teflon filter (40% to 53%). This is
consistent with coarse-mode nitrate consisting of salts that are less volatile than fine-mode
nitrate, which is predominantly ammonium nitrate.
If only Teflon filters are used in routine sampling, with no backup nylon filters, then fine-
mode nitrate would be severely under-measured (Ashbaugh and Eldred, 2004), though the
majority of the coarse-mode nitrate will be captured. Nitrate is a small fraction of coarse mass in
these two locations, so the effect of not having backup nylon filter measurements on coarse
mass balance would be negligible at these and similar sites. Thus, the decision on whether to
use nylon backup filters and analyze them by IC will be dictated by the extent to which the PMC
speciation sampling network will also be used for PM2.5 speciation.
Table 5-1. Summary of nitrate concentrations via dichot on Teflon, nylon, and total
(Teflon + nylon), plus fraction of total nitrate on Teflon filter.
Site
Size
Fraction
Avg. Nylon
NOs (MQ/rn3)
Avg. Teflon
NOs (MQ/rn3)
Avg. Total
NOs (MQ/m3)
Avg. % of NO3
on Teflon
PHX
PMc
0.13
0.42
0.54
77
PHX
PMf
0.44
0.50
0.94
53
STL
PMc
0.13
0.43
0.56
71
STL
PMf
0.54
1.01
1.54
39
5.3 Nitrate Partitioning Between Fine and Coarse Modes
Figure 5-1 shows the partitioning of PM10 nitrate between the fine and coarse fractions
as measured by the dichotomous samplers (Figure 5-1 a) and PMf nitrate partitioning to the front
Teflon filter compared to the Teflon/nylon filters (Figure 5-1b). This analysis assumes negligible
loss of nitrate during XRF analysis; however, results from an EPA study on PM2.5 speciation
indicated that the vacuum applied during XRF can reduce the amount of nitrate on a Teflon filter
by as much as 40% (U.S.EPA, 2001). If nitrate measurements need to be made using a dichot
then the issue of nitrate loss from theTeflon filter due to XRF would need to be addressed. The
data set includes all Dichot A nitrate data and Dichot B nitrate data for days on which no Dichot
A nitrate data were collected. Figure 5-1 (a) shows that some samples with PMf nitrate below
about 1 |jg/m3 have more PMC nitrate than PMf nitrate. However, PMC nitrate is not elevated on
days of high PMf nitrate concentrations. Deviations below the 1:1 (diagonal) line in Figure 5-1 (b)
represent PMf nitrate volatilization loss from the front filter that is captured by the back filter. At
low nitrate concentrations, PMf nitrate volatilization losses are generally lower in Phoenix
5-2
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
compared to St. Louis (this observation is not discernible in Figure 5-1 b). This difference
between the two sites might reflect a counter-ion other than ammonium for nitrate in many of the
Phoenix PMf samples, or the differential influences of environmental conditions on nitrate
volatilization. For PMf nitrate greater than about 2.5 |jg/m3, the mean Teflon filter nitrate loss is
0.7 ± 0.3 |jg/m3 (N = 12) at St. Louis, and 1.5 ± 1.3 |jg/m3 (N = 6) at Phoenix,8 reflecting greater
and more variable absolute loss of nitrate from the Teflon filter at Phoenix compared to St. Louis
for high nitrate concentrations.
7
PHX
STL
6
5
4
3
2
1
0
0
1
2
3
5
6
7
4
7
PHX
STL
6
a>
=L
5
4
3
2
1
0
0
1
2
3
5
6
7
4
PMf nitrate, ng/m3 PMf nitrate, ^g/m3
Figure 5-1. Scatter plots fordichotomous sampler nitrate showing (a) the partitioning of
nitrate between the fine and coarse PM fractions and (b) total fine nitrate and the nitrate
on the front (Teflon) filter. Phoenix is represented by gray triangles and St. Louis by open
circles.
5.4 Ammonium Balance and Implications for Ammonium
Measurements
To understand whether sulfate (SO4) and nitrate (NO3) are associated with ammonium
(NH4), or whether they are associated with other cations, such as sodium, it is useful to compare
measured ammonium to the amount of ammonium expected based on sulfate and nitrate
concentrations.
Figure 5-2 shows measured versus predicted ammonium for Teflon and backup nylon
dichot and FRM filters at each site for fine and coarse aerosol. Full neutralization of sulfate and
nitrate was assumed for predicted ammonium:
Predicted ammonium based on NH4N03 and (NH4)2S04 = 0.29(JVOJ) + 0.38(SO^)
8 Excluding the Phoenix value at (5.1, 1.3) in Figure 5-1 b, for PMf nitrate greater than 2.5 |jg/m3, the mean Teflon filter
nitrate loss is 1.1 +0.7 |jg/m3 (N = 5).
5-3
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
For PMf on the Teflon filter, whether dichot or FRM, at both sites, the predicted and
measured ammonium concentrations are highly correlated with quantitative agreement in
St. Louis and overprediction of the measured ammonium in Phoenix. The latter overprediction
may indicate that the sulfate aerosol is not fully neutralized by ammonium, and so ammonium is
not the sole counter-ion for sulfate and nitrate. For PMC on the Teflon filter, ammonium is grossly
overpredicted, again indicating that the sulfate and/or nitrate are associated with non-
ammonium cations. At both sites, the PMC ammonium concentrations are low (only one value
greater than 0.1 jjg/m3).
For PMf on the nylon filter, predicted and measured ammonium are highly correlated; in
this case, there is quantitative agreement at Phoenix and overprediction for St. Louis. The
reason for this overprediction is not known. For PMC on the nylon filter, ammonium is
underpredicted for the FRM in St. Louis and overpredicted for all other samplers at both sites. At
both sites, the PMC ammonium concentrations are low.
The low PMC ammonium concentrations at both sites on the Teflon and nylon filter
indicate that PMC ammonium measurements may not be warranted.
5-4
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
(a) PHX, Dichot A, Teflon
1.2
1.0 -
3- O.i
1 0.6 -
E
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
0.5
0.4
3
0.3
0.2
0.1
0 0.5 1.0 1.5 2.0 2.5 3.0
measured ammonium, jxg/m
0.1 0.2 0.3 0.4
measured ammonium, u.g/m3
(e) STL, Dichot A, Teflon
(f) STL, Dichot A, Nylon
nP
(g) STL, FRM, Teflon
measured ammonium, p.g/m
measured ammonium, jxg/m
Figure 5-2 (continued). Measured ammonium versus predicted ammonium scatter plots
at Phoenix (a-d) and St. Louis (e-h) on Teflon and nylon. Coarse particles are
represented by blue diamonds and fine particles by red squares. Diagonal lines show the
1:1 relationship.
5-6
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
5.5 Comparison of Concentrations With and Without Denuders
To test the influence of using a denuder on ammonium and nitrate concentrations, on a
subset of days, one dichot was run with a denuder and the other dichot was run without a
denuder. Multi-channel annular denuders coated with magnesium oxide were placed in the
sampler down tubes. The collocated sampler data were used to examine whether nitric acid is
adsorbing onto the Teflon filter and then being quantified as aerosol nitrate, since a denuder
strips out gaseous nitric acid. As shown in Figure 5-3, concentrations for both fine and coarse
mode ammonium and nitrate were essentially no different whether or not a denuder was used.
For comparison, sulfate and crustal species also showed no difference between measurements
with and without a denuder, as expected. Thus for dichot measurements at locations similar to
Phoenix and St. Louis, a denuder is likely not needed. If some other sites are deemed to
warrant denuders based on a denuder/no denuder comparison study, the good quantitative
agreement for PMC crustal species demonstrates that the denuder used in this study can be
used with negligible particle losses.
5-7
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
a) ammonium
0.8
0.7
~ PMf
0.5
0.4
0.3
Ptl
0.2
0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Dichot with Denuder (|ig/m3)
c) sulfate
50.8
Q
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Dichot with Denuder (ng/m3)
(d) crustal species (Ca+Fe+Si)
1 2 3 4 5
Dichot with Denuder (|ig/m3)
Figure 5-3. Collocated dichot (dichot with denuder versus dichot without denuder) scatter
plots for (a) NH4, (b) NO3, (c) SO4, and (d) Ca, Fe, and Si at Phoenix. Coarse particles
are represented by blue diamonds and fine particles by red squares. Diagonal lines show
the 1:1 relationship.
(b) nitrate
OPMc
1 0.5
0.5 1 1.5 2
Dichot with Denuder (|ig/m3)
~ PMf
5-8
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
5.6 Comparison of Ion and Corresponding Element Concentrations
Figures 5-4 and 5-5 display the correlation between total elemental concentration
(analyzed by XRF) and ion concentration (analyzed by IC) for certain elements measured on
Teflon filters from Dichot A and B and the FRM samplers. If an ion is well-correlated with the
corresponding element, then the element could feasibly be used to estimate ion concentrations
without having to conduct IC analysis for ions. At both sites, PMf potassium was primarily ionic,
while PMC potassium was primarily nonionic. Sodium was primarily ionic for both size fractions
and both sites. Sulfate ion (reported as sulfur in the figures) accounted for most of the sulfur at
both sites for both coarse and fine fractions. Thus, based on the data from these two sites, the
added value from the measurement of ions is the ability to measure nitrate and to resolve PMC
potassium ion compared to elemental potassium. Given the limited added value, at least for
sites with low PMC nitrate concentrations, and the superior detection limits for XRF compared to
IC for most of the reported ions, the added cost for ion analysis is not justified except possibly in
cases with high PMC nitrate.
5-9
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
0.4
.£
CTl
A
6 1
(a) potassium at STL, Dichot
~ PMf: [IC] = O.04[x:RF] +0.01 ; R* = 0.012
O PMc: [IC] = 0 39[XRF] + 0 00 : R2 = 0.428/
0 3
(b) potassium at STL, FRM
~ PMf: [IC] = 0.79[XRF] +001 R2 = 0.780
O PMc: [IC] = 0.2B[XRF] + 0.00 ; R2 = 0.525/
j 0.2 -
0.0 0 1 0 2 0.3 0.4
element by XRF, jj,g/m3
~
~
0.1 0.2 0.3
element by XRF, jig/m3
(c) sodium at STL, Dichot
~ PMf: [IC] = 0.91 [XRF] +0.00 ; R2 = 0.632 / O
O PMc: [IC] = 0.80IXRF] + 0.01 ; R* = 0.969
(O) excluded
(d) sodium at STL, FRM
~ PMf: [IC] = 1.01 [XRF] - 0.01 ; R2 = Of
O PMc: [IC] = 0.85[XRF] - 0.01 ; F32 = 0.897.
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
element by XRF, ^g/m3
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1 .€
element by XRF, jxg/m3
(e) sulfur at STL, Dichot
~ PMf: [IC] = 1 02[XRF] - 0.05 ; R2 = 0.981
O PMc: [IC] = 0.70|XRF] - 0.01 ; R2 = 0.811
0.0 0.5 1.0 1.5 2.0 2.5
element by XRF, jig/m3
(f) sulfur at STL, FRM
~ PMf: [IC] = 0.91 [XRF] +0.01 ; R2 = 0.982
O PMc: [IC] = 0.37[XRF] + 0.05 ; R2 = 0.129
0.5 1.0 1.5 2.0 2.5 3.0
element by XRF, p,g/m3
Figure 5-4. Comparison of ion and element concentrations (jjg/m3) at St. Louis for K and
K+, Na and Na+, and S and SO42" via dichot and FRM samplers for PMC and PMf.
5-10
-------
EPA's Coarse PM Pilot Study
Analysis of Ions
1 4
5 0.8
O
i? 0 6
(a) potassium at PHX, Dichot
~ PMf: [IC] = 0.85[XRF] - 0.03 ; R2 = 0.974
O PMc: [IC] = 0 25[XRF] - 0 03 ; R2 = 0.372 ^
iff
1.4 -
(b) potassium at PHX, FRM
~ PMf: [IC] = 0.87[XRF] - 0.04 ; R2 = 0.954
O PMc: [IC] = 0.23[XRF]- 0.04 : R2 = 0 I 78 ,
E
5 0.8 -
O
£ 0 6 -
0.0 0 2 0.4 0 6 0.8 1.0 1 2 14
element byXRF, jig/m3
~ o
0 0 0.2 04 0.6 0 8 1 0 1 2 1 4
element by XRF. )ig/m3
(c) sodium at PHX, Dichot
/ 0
~ PMf: [IC] = 0 92[XRF] +0.01 ; R2 = 0 708
o PMc. [IC] = 0.83[x:RF] + 0.01 , R2 = 0.039/
o
"e
O)
zL
o
0 o°
O
>>
si
O 00
O
c
o
o
(d) sodium at PHX, FRM
~ PMf: [IC] = 1.15[XRF] - 0 01 ; R2 = 0.974
o PMc: [IC] = 0.BOPCRF] - 0.01 , R2 = 0775
o 0
§>oo°
0.0 0.2 0.4 0.6 0.8 1.0
element by XRF, ng/m3
0.0 0.2 0.4 0.6 0.8 1.0
element by XRF. jig/m3
(e) sulfur at PHX, Dichot
~ PMf: [IC] = 0 95[XRF] - 9.91 ; R2 = 0 980
O PMc. [IC] = 0.78[XRF]- 0.01 ; F2 = 0.900/
0.2 0.4 0.6 0.8 1.0
element by XRF, ng/m3
(f) sulfur at PHX, FRM
~ PMf: [IC] = 0.95[XRF] - 0.01 ; R2 = 0.9B0
o PMc: [IC] = 0.90[XRF]- 9.91 ; R1 = 0 083 /
0.2 04 0 6 0.8 1.0
element by XRF. .g/m
Figure 5-5. Comparison of ion and element concentrations (jjg/m3) at Phoenix for K and
K+, Na and Na+, and S and SO42" via dichot and FRM samplers for PMC and PMf.
5-11
-------
-------
EPA's Coarse PM Pilot Study
Mass Balance
8, IU,5--nil
Mass balance closure reflects the ability to reconstruct the observed Teflon filter
gravimetric mass from the chemical species data. It is an uncertain process because not all
species are quantified (e.g., retained water), and some of the chemical species data are derived
from parallel samples that have different measurement artifacts (e.g., carbon data from a quartz
filter). The objective of this analysis is to account for the mass on the Teflon filter rather than to
make a best estimate of the species concentrations in the atmosphere. Thus, the data were not
corrected for nitrate volatilization measured as the nitrate taken up by the nylon backup filter.
The approach also assumes that the fraction of fine-particle nitrate loss from the dichot major
flow Teflon filter is the same as the fraction of fine-particle nitrate loss from the dichot minor flow
Teflon filter.
The organic carbon data have not corrected for OC artifacts, and a sensitivity study
showed that including an estimate of the artifact correction did not change the key findings. To
account for other elements in the organic aerosol, OM was assumed to be OC multiplied by 1.6
(i.e., [OM] = [OC] x 1.6). An OC multiplier of 1.6 was chosen because it is the midpoint of the
commonly used range of 1.4 to 1.8, but, as discussed in the next section, sensitivity studies
were performed using various OC multipliers. Other species included in the reconstructed mass
include sulfate ion, nitrate ion, major inorganic cations (ammonium [NH4+], sodium [Na+], and
potassium [K+]), CI (as measured by XRF on the Teflon filter), EC, soil oxides (estimated as
2.20[AI] + 2.49[Si] + 1.63[Ca] + 2.42[Fe] + 1.94[Ti]),9 and other metals.10 The species group
"other metals" was included to initially generate a reconstructed mass estimate that would be
conservatively high. It was deemed inconsequential to the mass balance reconstruction and,
given that such metals are factored to some extent into the above soil equation, this species
group could have been omitted from the mass reconstruction. As described in Section 3, RTI
applied XRF attenuation factors to the low atomic number elements (Z < 20).
6.1 Results for Dichotomous Samplers
Figure 6-1 shows the species distributions and mass balance closure for each sample.
Soil oxides dominate PMC, while organic matter accounts for 10% to 20% of the PMC mass.
Figure 6-2 compares the reconstructed mass and gravimetric mass for PMf and PMC with
statistical metrics summarized in Table 6-1. Ordinary least squares (OLS) regression, which
assumes there is no uncertainty in the x-axis values, was used for this analysis. This should be
a robust approach because the uncertainty in gravimetric mass is much smaller than the
uncertainty in the reconstructed mass.
9 This soil-oxides estimate is the standard formula applied to IMPROVE network data
(http://vista.cira. colostate.edu/improve/tools/aertvpeeas. htm).
10 Other metals=(silver [Ag] + arsenic [As] + barium [Ba] + bromine [Br] + cadmium [Cd] + cerium [Ce] + cobalt [Co] +
chromium [Cr] + cesium [Cs] + copper [Cu] + indium [In] + magnesium [Mg] + manganese [Mn] + nickel [Ni] +
phosphorus [P] + lead [Pb] + rubidium [Rb] + antimony [Sb] + selenium [Se] + tin [Sn] + strontium [Sr] + vanadium [V]
+ zinc [Zn] + zirconium [Zr]).
6-1
-------
PHX PMf
-i 1 r-
-r 1 1 r—
(a)
nO n |
kill \lift
25
20
1 1 1 ; 1 1 1 ! 1 1 1 p-
STL PM,
(C)
J
Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Years 2010-2011
—i 1 r*
Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Years 2010-2011
PHX PM
STL PM
° 10 -
Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Years 2010-2011
Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Years 2010-2011
Soil Oxides
!!¦—1
Nitrate
1 1
Sulfate
1 1
Metals
Cations (NH4+, Na+, K+)
EC
> i
OM (OCx1.6)
Chlorine
e
Gravimetric Mass
Figure 6-1. Mass balance closure between Teflon filter gravimetric mass and the chemical speciation data fordichot samples. All
species were measured on the Teflon filter except EC and OC, which were measured on the quartz filter from a collocated dichot
sampler. Organic matter was estimated using OM = 1.6 x OC, and the OC was not corrected for measurement artifacts.
-------
EPA's Coarse PM Pilot Study
Mass Balance
0 5 10 15 20 25 0 5 10 15 20 25 30 35 40
PMf gravimetric mass, |.ig/m3 PMC gravimetric mass, ^g/m3
Figure 6-2. Reconstructed mass versus gravimetric mass for the dichotomous sampler
Teflon filters: (a) PMf and (b) PMC for Phoenix and St. Louis. Organic matter was
estimated using OM = 1.6* OC, and the OC data were not corrected for measurement
artifacts.
Table 6-1. Measures of agreement between reconstructed and gravimetric mass for
Phoenix and St. Louis, using OM = 1.6* OC and no correction for OC artifacts.
Site and Size
N
OLS Regression, jjg/m3(a)
r2
Ratio of
Means(b)
PHX PMf
37
[recon] = 0.92 [grav] + 1.6
0.96
1.14
PHX PMc
37
[recon] = 1.18 [grav] - 1.0
0.94
1.13
STL PMf
21
[recon] = 0.90 [grav] + 0.5
0.92
0.94
STL PMC
21
[recon] = 0.87 [grav] + 1.3
0.92
1.01
a OLS regression of reconstructed mass on gravimetric mass.
b The PMc ratio of means is 1.09 for PHX PMC and 0.96 for STL PMC when applying a multiplier of 0.82 to
the OM to correct for the presumed carbon losses from the Teflon filter. This multiplier is the dichot-to-
FRM mean ratio for gravimetric mass (Table 2-1). The PMC ratios of means further decrease to 1.07
and 0.93, respectively, when also neglecting K+ and other metals, assuming these species are
present as crustal material and are included in the standard formula for soil oxides.
Several factors that may influence the quality of mass balance closure include, but are
not limited to, the assumed OM/OC ratio, OC measurement artifact corrections, the assumed
oxide forms of the crustal elements, and XRF attenuation factors applied to light elements. As
detailed in Section 3.5, the latter is particularly relevant for PMC. In this section, the attenuation
factors developed and applied by RTI were used for all data. An alternative form for the soil
oxides equation that excludes Al and increases the Si multiplier from 2.49 to 3.48 (Simon et al.,
2011) yields soil oxides contributions that are, on average, 5% higher for PM^
6-3
-------
EPA's Coarse PM Pilot Study
Mass Balance
For Phoenix data using OM/OC =1.6 and no OC artifact correction, the reconstructed
mass consistently overestimates the gravimetric mass for both PMf and PMC, with an average
difference of approximately 13% for both size ranges (9% for PMC after applying a correction for
loss of carbonaceous matter from the Teflon filter as described in a footnote to Table 2-1). The
OM/OC ratio that best closes mass balance (i.e., the minimum difference between
reconstructed and gravimetric mass, which are shown as root-mean-square (RMS) residuals in
Figure 6-3) is 1.2 to 1.3 for PMf. This ratio is consistent with the value of 1.25 reported by
Simon et al. (2011) for Phoenix IMPROVE PM2.5 data. The best-fit OM/OC ratio for PMC is 0.6
(off the scale range in Figure 6-3) with the same ratio obtained when applying a multiplier of
0.82 to the carbon data to account for losses from the Teflon filter. The PMC ratio of 0.6 is
implausibly low (OM/OC must be greater than or equal to unity), and in this case there must be
other factors driving the observed bias.
For St. Louis data using OM/OC =1.6 and no OC artifact correction, the reconstructed
mass underestimates the gravimetric mass by 6% for PMf, and the agreement is within 1% for
PMC. The OM/OC ratio that yields the best agreement for PMf is 1.8; this ratio is consistent with
the value of 1.81 reported by Bae et al. (2006) for PM2.5 data collected at the East St. Louis site.
For PMC, shown as the open circles in Figure 6-3b, the best-fit OM/OC ratio is 1.4 to 1.6 with a
ratio of 1.8 obtained when applying a multiplier of 0.82 to the carbon data (to account for losses
from the Teflon filter). However, the RMS residual varies by only 0.3 |jg/m3 over the range of
OM/OC ratios between one and two. This flat response surface means the best-estimate
OM/OC ratio of 1.4 to 1.6 for St. Louis PMC is highly uncertain.
~ PHX - PMf
a PHX-PIVI
Q)
"TO
=3
~Q
(D
CD
STL - PMf
STL - PIVL
3 -
A A A A A
(a)
=3
o~ 9
a) z
c
ro
CD
E
o 1
o
o o o
0 8 0
• • • •
(b)
1.0 1.2 1.4 1.6
OM/OC
2.0
1.0
1.2
1.4 1.6
OM/OC
2.0
Figure 6-3. RMS residuals between reconstructed mass and gravimetric mass as a
function of the assumed OM/OC ratio: (a) Phoenix and (b) St. Louis.
6-4
-------
EPA's Coarse PM Pilot Study
Mass Balance
The sensitivity of mass balance closure to OC artifact correction was examined by
assuming a negligible negative artifact and assuming the grand mean field blank mass loading.
(Appendix D is a representative measure of the positive artifact for both the dichot major and
minor flow channel samples). Artifact-corrected OC ambient mass concentrations for PMf and
PMC were calculated after subtracting the filter- and trip-blank pooled mean OC loading of
6.6 |jg/filter from the major and minor flow channels. The best-fit OM/OC ratios were 1.4 for PMf
at Phoenix and 2.1 for St. Louis. For PMC, the best-fit OM/OC ratios were 0.5 at Phoenix and 1.6
at St. Louis. Once again, the ratio for PMC at Phoenix is implausible and the response surface
for the RMS residual at St. Louis is very flat, which means the ratio is highly uncertain.
6.2 Mass Balance Implications
Mass balance closure for PMC can be influenced by several factors, including but not
limited to OC artifact correction, the assumed OM/OC ratio, the assumed soil oxides
composition, and XRF attenuation factors used for light elements. If all else is equal, a
correction applied for a negative OC artifact decreases the reconstructed mass and requires a
higher OM/OC ratio to achieve closure compared with the case of no artifact correction.
For PMC at St. Louis, plausible OM/OC ratios were estimated when including and
excluding a correction for positive artifacts. However, the estimated OM/OC ratios are highly
uncertain.
For PMC at Phoenix, implausible OM/OC ratios were estimated for no OC artifact
correction and also for when a correction was applied for positive artifacts using field blanks
data. In both cases, the reconstructed mass is greater than the gravimetric mass for OM/OC
ratios larger than approximately 0.5. This finding demonstrates that the reconstructed mass is
being systematically overestimated, with possible explanations being bias in the XRF
attenuation factors applied to Al, Si, and Ca; or bias in the multipliers used to account for the
oxide forms of the crustal elements; or both. Mass balance closure alone cannot be used to
obtain meaningful best-fit attenuation factors and oxides multipliers. Furthermore, the best-fit
values depend on assumed OC artifact correction and OM/OC ratio.
6-5
-------
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
7. Comparison of FDMS TEOM and
* life* Me* or *
In addition to the dichot and FRM measurements, each site also had a Thermo 1405-DF
dichotomous TEOM with FDMS®. The Thermo 1405-DF dichotomous sampler with FDMS is an
approved FEM for PM2.5. The 1405-DF measures fine (PM2.5) and coarse (PM10-2.5) particulate
matter mass concentrations at six-minute time resolution. Its deployment in this study provides
insights into the climatology of coarse PM by placing the 24-hour integrated chemical speciation
samples in a broader context with daily, high-time resolved, coverage of PM mass, and also by
providing the ability to examine diurnal behavior.
The dichot TEOM includes a 16.7 LPM standard PM10 inlet followed by a virtual impactor
that splits the sample stream into major (15.0 LPM) and minor (1.67 LPM) flows. A portion of the
major flow stream and the entire minor flow stream pass through Nafion dryers; subsequently,
the entrained particles are deposited onto TEOM filters for mass determination.
The major flow includes only fine particles, whereas the minor flow includes all of the
coarse particles and 10% of the fine particles; therefore, PM2.5 mass concentrations are
calculated directly from the major flow, and PM10-2.5 mass concentrations are calculated by
correcting the minor flow for fine-particle intrusion. In both cases, the size-segregated mass is
further partitioned into volatile (reference) and nonvolatile (base) components by the dual-
channel FDMS (Meyer et al., 2002). This partitioning is operationally defined by the
measurement method. PM2.5 FDMS volatility component concentrations are the same as the
major flow volatility component concentrations. It is often not appreciated that the instrument
reports FDMS volatility component (base and reference) concentrations for the minor flow rather
than forPMio-2.5¦ The base and reference flows contain 10% of the fine particles; therefore, the
user must apply Equation 7-1 to calculate the PM10-2.5 volatile and nonvolatile mass
concentrations from the instrument-reported data.
PMw_2_5 (k) = PMmnor (k) - pm^ (k) (Eq. 7-1)
^ototal
where PM is the mass concentration, Q is the volumetric flow rate, and k is the FDMS volatility
component, i.e., volatile (reference) or nonvolatile (base).
Previous work has linked the volatile component of fine PM, as measured by the FDMS
TEOM, to semivolatile particulate matter, including ammonium nitrate and semivolatile organic
compounds (Grover et al., 2005; Faveza et al., 2007). The project team's immediate objective,
as described below, was to use the dichot TEOM data to examine the temporal behavior of the
volatile and nonvolatile components of both fine and coarse PM at the two contrasting sites.
Dichot TEOMs were deployed at the Phoenix and St. Louis sites, with data reporting
starting in July 2010 and September 2010, respectively. Data were collected on six-minute
cycles and converted to hourly averages. Raw (unvalidated) hourly fine and coarse PM mass
concentrations were automatically uploaded to AirNow-Tech (www.airnowtech.org). Data for
mass and FDMS volatility components were periodically provided by the Maricopa County Air
7-1
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
Quality Department (for Phoenix) and the Air Quality Laboratory at Washington University (for
St. Louis) to the data analysis team for further validation and interpretation.
The data were censored to remove records with unstable base and reference readings
(e.g., large fluctuations within the hour). In the case of Phoenix, this resulted in the removal of
18 days (about 10% of the total data set), including the six days for which the highest
concentrations were reported. The removal of these data does not significantly affect the
analyses presented in this report, which focuses on central tendencies of the data sets.
Non-volatile/volatile data at Phoenix after February 2011 were unstable and showed
unusually high noise, and so are not included in analyses here; data at St. Louis after
replacement of a Nafion dryer in mid-March exhibited noise as well, and are not included here.
7.1 Results
Figure 7-1 compares daily-average dichot TEOM total and nonvolatile mass
concentrations to 24-hour integrated gravimetric mass concentrations using the sequential
dichot sampler (St. Louis) and paired PM2.5 and PM10 FRM samplers (Phoenix) with Teflon
filters. The Phoenix dichot TEOM data are compared to the paired FRM samplers, rather than
the dichot sampler, because the latter suffered from significant particle losses at Phoenix. For
PM2.5 at St. Louis (Figure 7-1a), the filter-based gravimetric mass concentrations are bounded
by the TEOM-measured total and nonvolatile mass concentrations. This pattern is consistent
with the Teflon filter retaining only a portion of the volatile ambient particulate matter due to
partial loss of species such as ammonium nitrate and semivolatile organic matter. For PM10-2.5 at
St. Louis (Figure 7-1b), there is also very good agreement between the dichot TEOM and Teflon
filter data, especially at low concentrations. The volatile component of PM10-2.5 is relatively small.
For PM2.5 (Figure 7-1c), there is good agreement at low concentrations, with the Teflon filter
mass closely tracking the dichot TEOM nonvolatile mass. The FRM filter sampler retains
virtually none of the volatile mass, with the nonvolatile mass in nearly quantitative agreement
with the Teflon filter gravimetric mass data. At high concentrations, the dichot TEOM PM2.5
mass is biased high compared to the FRM PM2.5 mass. This bias is also observed at all
concentrations for the PM10-2.5 data (Figure 7-1d). For all samples, the Phoenix PM10-2.5 is
essentially nonvolatile.
The minor flow volatile mass was typically 1 |jg/m3 or smaller. PM10-2.5 volatile mass
calculated using Equation 7-1 was about 50% of the minor flow volatile mass and the
propagation of uncertainties through Equation 7-1 suggests that the PM10-2.5 volatile mass
concentration for such cases is highly uncertain. PM chemical composition data are available for
many of these days, but the dichot TEOM PM10-2.5 volatile component mass concentrations are
too small and too uncertain to support meaningful comparisons to the speciation data. The
remainder of this report focuses on time series analysis of the data collected to date. The
project team's primary interest is the PM10-2.5 trends, but PM2.5 trends are also shown for
comparison.
7-2
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
30
25
=> 20
CD
E
15 -
o
LU
10 -
o
T5
(a) STL PM
TEOM total mass
a TEOM nonvolatile mass
20
15
O
LU
10
(c) PHX PM
2.5
t TEOM total mass
* TEOM nonvolatile mass
0 5 10 15 20 25 30
dichot sampler gravimetric mass, ug/m3
0 5 10 15
FRM sampler gravimetric mass, ug/m
20 25
3
(b) STL PM10,5
d PHX PM
10-2.5
v TEOM total mass
a TEOM nonvolatile mass
TEOM total mass
a TEOM nonvolatile mass
0 5
dichot sampler gravimetric mass, ug/m
0 5
FRM sampler gravimetric mass, ug/mJ
Figure 7-1. East St. Louis (left) and Phoenix (right) 24-hour average dichot TEOM total
(red markers) and nonvolatile (black markers) PM mass concentrations versus the 24-hour
integrated PM mass concentrations from filter-based samplers with gravimetric analysis on
the Teflon filters: fine PM (top) and coarse PM (bottom). Lines connecting markers
indicate corresponding data pairs. The diagonal line is the 1:1 relationship.
Figure 7-2 shows the time series for daily-average PM10 mass concentrations, stratified
by fine and coarse PM contributions and presented as "stacked" graphs, at St. Louis
(Figure 7-2a) and Phoenix (Figure 7-2b). At St. Louis, the fine PM mass is typically greater than
the coarse PM mass (median PM10-2.5/PM2.5 ratio = 0.67), but there are periods with high coarse
PM concentrations, such as mid-October. In contrast, at Phoenix, the coarse PM mass is much
greater than the fine PM mass (median PM10-2.5/PM2.5 ratio = 2.8), but there are periods with
relatively high fine particle mass, such as late January and early February. Overall, PM10
concentrations are much higher at Phoenix compared to St. Louis (note the twofold difference in
the concentration scale ranges).
7-3
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
I0-C.5
Jul-10 Aug-10 Sep-10 Oct-10 Nov-10 Dec-10 Jarv11 Feb-11 Mar-11
Figure 7-2. Daily-average dichot TEOM time series for fine PM (yellow) and coarse PM
(black) at (a) East St. Louis, 9/2010-2/2011; and (b) Phoenix, 7/2010-2/2011. Note the
concentration scale range differences (60 i-ig/m3 for STL, 120 jxg/n# for PHX). Asterisks
below the x-axis in the PHX plot denote periods with three or more successive days of
data removed.
Figure 7-3 shows the fine PM arid coarse PM mass concentrations further stratified by
FDMS volatility components. PM2.5 volatile mass is appreciable at both St. Louis and Phoenix
(Figure 7-3a.c), with median PM2.5 volatile/total mass ratios of 0.30 and 0.21, respectively. In
contrast, the PM10-2.5 volatile mass at both sites (Figure 7-3b,d) is very small, with median PM2.5
volatile/total mass ratios of 0.05 at St. Louis and 0.01 at Phoenix. The PM10-2.5 mass
concentrations for Phoenix are twice as high as those for St. Louis, as noted by the twofold
scale range for Phoenix compared to St. Louis.
7-4
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
Sep-10 Qa-10 Not-10 Odt-ID Jan-tl Feb-11 Mar-11
JuHD A119-IO S.p-10 OtJ-lO NmlO "o«40 J>n.11 FeU I Mir 11
OeMO NOT. 10 Oec-IO J»n-1
Figure 7-3. Daily-average dichot TEOM time series for fine PM (top) and coarse PM
(bottom) at East St. Louis, 9/2010-2/2011 (left), and Phoenix, 7/2010-2/2011 (right). Daily
average mass concentrations are partitioned into FDMS volatile (red) and nonvolatile
(black) components. Note the concentration scale range is twice as high for Phoenix
PM10-2 5 (panel d). Asterisks below the x-axis in the Phoenix plots denote periods with
three or more successive days of data removed.
Figure 7-4 shows diurnal profiles for PM- s total (Figure 7-4a), nonvolatile (Figure 7-4b),
and volatile (Figure 7-4c) mass at St. Louis and Phoenix. At St. Louis, there is a shallow midday
minimum in both the total and nonvolatile mass that is consistent with dilution by daytime growth
in the mixing layer depth. In contrast to the nonvolatile component, the volatile component
exhibits a midday maximum. The data collected includes relatively low nitrate, so the volatile
component is likely semivolatile organic aerosol. The midday maximum in the diurnal profile is
consistent with the production of semivolatile organic aerosol by secondary processes in the
atmosphere (Faveza et al., 2007). At Phoenix, total and nonvolatile mass increases in the early
morning before rush hour, reaching a maximum at 0800 to 0900 local standard time (LST),
followed by a deep midday minimum. Like St. Louis, the volatile component at Phoenix exhibits
a midday maximum consistent with the secondary production of semivolatile organic aerosol.
PM; s velilile
PM,3now<>taflile
-11
7-5
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
18
E
gi
c
o
03
Q)
O
c
o
o
0
U)
CD
L_
CD
>
CU
(a) STL PM2
(b) STL PM„_ nonvolatile
II 111
VTTttTTHTt
[In
TTtt
yvvVTvy
(c) STL PM„5 volatile
0 2 4 6 8 10 12 14 16 18 20 22
hour (LST)
(a) PHX PM
iTTTTT
(b) PHX PM0- nonvolatile
(c) PHX PM„5 volatile
I 10 12 14 16 18 20 22
hour(LST)
Figure 7-4. PM2 5 diurnal profiles for East St. Louis, 9/2010 to 2/2011 (left) and Phoenix,
7/2010 to 2/2011 (right): (a) total TEOM mass; (b) nonvolatile TEOM mass; and (c)
volatile TEOM mass. Whiskers are 10th and 90th percentiles, boxes are 25th and 75th
percentiles, and the interior black and red lines are the median and arithmetic mean,
respectively.
Figure 7-5 shows diurnal profiles for PM10-2.5 total mass (Figure 7-5a) and volatile mass
(Figure 7-5b) at St. Louis and Phoenix. Because of the extremely low concentrations, the
volatile component concentrations have high uncertainty and therefore are not interpreted.
Because of the very low volatile mass in the coarse aerosol (Figure 7-5b), nonvolatile
mass concentrations closely track the total mass; therefore, profiles are not shown for the
nonvolatile mass concentrations. The PM10-2 5 total mass diurnal profile at St. Louis exhibits a
broad midday maximum, which is in contrast to the broad midday minimum exhibited for PM2.5
total mass. The Phoenix PM10-2.5 total mass diurnal profile closely tracks the PM2.5 profile with an
early morning maximum and a deep midday minimum.
7-6
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
<0
F
30
CD
24
O
03
>,
1?
3
O
.C
6
0
(a) STL PM,,
II
ttt
VVtt
rYVT
(b) STL PM10 25 volatile
TtTTTTTttttV
8 10 12 14 16 18 20 22
hour (LST)
£ 100
O)
d 80
o
^ fin
(a)PHXPM
f TT tV
(b) PHX PM10 2 5 volatile (5x zoom)
0 2 4 6 8 10 12 14 16 18 20 22
hour (LST)
Figure 7-5. PM10-2.5 diurnal profiles for East St. Louis, 9/2010 to 2/2011 (left) and
Phoenix, 7/2010 to 2/2011 (right): (a) total TEOM mass and (b) volatile TEOM mass.
Whiskers are 10th and 90th percentiles, boxes are 25th and 75th percentiles, and the
interior black and red lines are the median and arithmetic mean, respectively.
To further examine these patterns, Figure 7-6 shows PM10-2.5 total mass diurnal profiles
for weekdays (Figure 7-6a,b) and weekends (Figure 7-6c,d). For St. Louis, the broad midday
maximum is observed on weekdays but not weekends. This suggests that the dominant driver
for the elevated midday PM10-2.5 concentrations at St. Louis is related to anthropogenic activities
rather than windblown dust. For Phoenix, the early morning maximum is observed on weekdays
but not weekends. This suggests that the elevated early morning PM10-2.5 concentrations at
Phoenix also arise from anthropogenic activities rather than windblown dust.
To further demonstrate the weekday excess PM10-2.5 at both sites, Figure 7-7 shows
diurnal profiles for the difference between the median weekday mass and median weekend
mass. The St. Louis weekday excess mass (green bars) is about 5 |jg/m3 from approximately
0700 to approximately 1900 LST. The weekday excess mass at Phoenix (red bars) is more than
5 |jg/m3 at 0500 LST, reaches a maximum of approximately 25 |jg/m3 at 0700 LST, and decays
throughout the remainder of the morning hours. The weekday excess is about 4 |jg/m3 during
the afternoon, and there is a local maximum at 1900 LST. Weekday versus weekday chemical
speciation data should be examined for further insights into the origins of the weekday excess.
7-7
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
36
£
1? 30
"ca 24
S 18
2 12
¦c 6
36
(a) STL PM10.a
weekdays
TT
TT
0 2 4
10 12 14 16 18 20 22
30
co 24
2 12
CD
>
03
fT 6
(b) STL PM10 2 5
weekends
TT
TV
10 12 14 16 18 20 22
hour (LST)
120
100 -
60 -
2 40 -
CD
20 -
(c) PHX PM10_:
weekdayt
0 2 4
10 12 14 16 18 20 22
120
100
co 80
CD
CD
2 40
€ 20 -
(d) PHX pm10.25
weekends
4 6 8 10 12 14 16 18 20 22
hour (LST)
Figure 7-6. PM10-2 5 weekday (top) and weekend (bottom) diurnal profiles for East St.
Louis, 9/2010 to 2/2011 (left) and Phoenix, 7/2010 to 2/2011 (right). Whiskers are 10th
and 90th percentiles, boxes are 25th and 75th percentiles, and the interior black and red
lines are the median and arithmetic mean, respectively.
PM 10 2 §¦ (weekday median) - (weekend median)
MMM
10 12 14 16 18 20 22
hour (LST)
Figure 7-7. Diurnal profiles for the difference in weekday and weekend hourly median
PM10-2.5 concentrations for East St. Louis, 9/2010 to 2/2011 (green) and Phoenix, 7/2010
to 2/2011 (red).
7-8
-------
EPA's Coarse PM Pilot Study
Comparison of FDMS TEOM and Filter Coarse PM Measurements
7.2 FDMS TEOM-to-Dichot Summary
For the seven to eight months of dichot TEOM data collected in this study, the
FDMS-defined volatile component of PM2.5 at both St. Louis and Phoenix exhibits a midday
maximum consistent with the production of semivolatile organic aerosol by secondary
processes. The governing equation is presented in Equation 7-1 for calculating the FDMS-
defined PM10-2.5 volatility components from the minor and major flow data. At both sites, the
PM10-2.5 volatile component was relatively small. PM10-2.5 total and nonvolatile mass is elevated
on weekdays compared to weekends at both sites, albeit with different diurnal patterns for the
weekday excess mass. Anthropogenic activities are the likely driver for this excess; more work
exploring speciation data is needed to determine the precise nature of the contributing activities.
7-9
-------
-------
EPA's Coarse PM Pilot Study
References
8. References
Ashbaugh L.L. and Eldred R.A. (2004) Loss of particle nitrate from Teflon sampling filters:
Effects on measured gravimetric mass in California and in the IMPROVE network. J. Air
Waste Manage., 54(1), 93-104, Jan.
Bae M.S., Schauer J.J., and Turner J.R. (2006) Estimation of the monthly average ratios of
organic mass to organic carbon for fine particulate matter at an urban site. Aerosol
Science & Technology, 40, 1123-1139, September.
Chow J.C. and Watson J.G. (2002) PM2.5 carbonate concentrations at regionally representative
Interagency Monitoring of Protected Visual Environment sites. Journal of Geophysical
Research-Atmospheres, 107(D21), Sep-Oct.
Dzubay T.G. and Stevens R.K. (1975) Ambient air analysis with dichotomous sampler and X-ray
fluorescence spectrometer. Environ. Sci. Technol., 9(7), July. Available at
http://pubs.acs.Org/doi/abs/10.1021/es60105a011.
Faveza O., Cachiera H., Sciarea J., and Moullec Y.L. (2007) Characterization and contribution
to PM2.5 of semi-volatile aerosols in Paris (France). Atmos. Environ., 41(36), 7969-7976.
Available at http://www.sciencedirect.com/science/article/pii/S13522310070Q8126.
Grover B.D., Kleinman M., Eatough N.L., Eatough D.J., Hopke P.K., Long R.W., Wilson W.E.,
Meyer M.B., and Ambs J.L. (2005) Measurement of total PM2.5 mass (nonvolatile plus
semivolatile) with the Filter Dynamic Measurement System tapered element oscillating
microbalance monitor. J. Geophys. Res., 110(D07S03), doi: 10.1029/2004JD004995.
Available at http://www.aqu.orq/pubs/crossref/2005/2004JD004995.shtml.
GutknechtW., Flanagan J., McWilliams A., Jayanty R.K., Kellogg R., Rice J., Duda P., and
Sarver R.H. (2010) Harmonization of uncertainties of X-ray fluorescence data for PM2.5
air filter analysis. J. Air Waste Manage., 60(2), 184-194, (PubMed PMID: 20222531).
Available at http://www.ncbi.nlm.nih.gov/pubmed/20222531.
Hyslop N.P. and White W.H. (2011) Identifying sources of uncertainty from the inter-species
covariance of measurement errors. Environ. Sci. Technol., 45(9), 4030-4037, doi:
10.1021/es102605x. Available at http://pubs.acs.Org/doi/abs/10.1021/es102605x.
Lee T., Yu X-Y, Ayres B., Kreidenweis S.M., Malm W.C., and Collet J.L. (2008) Observations of
fine and coarse particle nitrate at several rural locations in the United States. Atmos.
Environ., 42, 2720-2732.
Meyer M.B., Patashnick H., and Ambs J.L. (2002) Ongoing development of a continuous
reference standard particulate matter mass monitor for ambient air. In Proceedings from
the Symposium on Air Quality Measurement Methods and Technology; A&WMA:
Pittsburgh, PA. Paper No. 53.
Simon H., Bhave P.V., Swall J.L., Frank N.H., and Malm W.C. (2011) Determining the spatial
and seasonal variability in OM/OC ratios across the US using multiple regression.
Atmospheric Chemistry and Physics, 11,2933-2949, doi: 10.5194/acp-11-2933-2011.
Available at http://www.atmos-chem-phys.net/11/2933/2011/acp-11-2933-2011 .html.
Smith R.J. (2009) Use and misuse of the reduced major axis for line-fitting. American Journal of
Physical Anthropology, 140, 476-486.
8-1
-------
EPA's Coarse PM Pilot Study
References
Tager I.B., Lurmann F.W., HaightT., Alcorn S., Penfold B., and Hammond S.K. (2010)
Temporal and spatial patterns of ambient endotoxin concentrations in Fresno, California.
Environ. Health Persp., 118(10), 1490-1496, doi: 10.1289/ehp.0901602.
U.S. EPA (2001) Evaluation of PM2.5 chemical speciation samplers for use in the EPA national
PM2.5 chemical speciation network. EPA-454/R-01-005, Office of Air Quality Planning
and Standards, Research Triangle Park, NC 27711.
U.S.EPA (2010) PM10-2.5 speciation pilot monitoring program, sample analysis, and data
reporting: Quality Assurance Project Plan (QAPP). U.S. Environmental Protection
Agency, March 2010. Available at http://epa.gov/ttn/amtic/pm10pilot.html.
U.S. EPA (2011) Field evaluation of the PM2.5 federal reference method and development of a
reference method method to measure and chemically speciate ambient coarse-mode
aerosols. EPA/600/R-11/099, Office of Research and Development, National Exposure
Research Laboratory, Reseach Triangle Park, NC 27711.
8-2
-------
EPA's Coarse PM Pilot Study
Appendix A
Appendix A: Summary Statistics
Table A-1. Summary of concentrations and MDLs by species and size fraction (PMf or
PMc) at St. Louis for primary dichot, on Teflon and quartz filters; all species except OC
and EC are from Teflon filter measurements. Two species, K and Na, were analyzed via
XRF and IC; the latter analysis results are indicated by "_IC."
Species
Avg
Cone
(pg/m3)
N
N >
N >
%>
%>
3MD
L
MDL
3MDL
MDL
Avg
Uncertainty
(Mg/m3)
Ag_XRF
PMc
0.0001
0.0145
31
0
0
0
0
N/A
AI_XRF
PMc
0.2779
0.0098
31
30
29
97
94
0.0397
As_XRF
PMc
0.0001
0.0010
31
0
0
0
0
N/A
Ba_XRF
PMc
0.0142
0.0033
31
28
18
90
58
0.0025
Br_XRF
PMc
0.0011
0.0009
31
12
3
39
10
0.0006
Ca_XRF
PMc
1.1732
0.0017
31
31
31
100
100
0.1024
Cd_XRF
PMc
0.0003
0.0089
31
0
0
0
0
N/A
Ce_XRF
PMc
0.0000
0.0023
31
0
0
0
0
N/A
CI_XRF
PMc
0.2613
0.0019
31
31
31
100
100
0.0208
Co_XRF
PMc
0.0008
0.0006
31
20
2
65
6
0.0005
Cr_XRF
PMc
0.0017
0.0009
31
19
5
61
16
0.0005
Cs_XRF
PMc
0.0004
0.0050
31
2
0
6
0
0.0029
Cu_XRF
PMc
0.0050
0.0007
31
31
26
100
84
0.0006
Fe_XRF
PMc
0.3109
0.0007
31
31
31
100
100
0.0227
ln_XRF
PMc
0.0004
0.0129
31
0
0
0
0
N/A
K_IC
PMc
0.0376
0.0058
36
33
30
92
83
0.0040
K_XRF
PMc
0.0921
0.0016
31
31
31
100
100
0.0082
Mass_Grav
PMc
9.4453
0.3022
36
36
36
100
100
0.5661
Mg_XRF
PMc
0.0740
0.0039
31
30
29
97
94
0.0139
Mn_XRF
PMc
0.0071
0.0007
31
30
28
97
90
0.0007
Na_IC
PMc
0.1974
0.0117
36
35
26
97
72
0.0337
Na_XRF
PMc
0.2935
0.0113
31
29
24
94
77
0.0632
NH4_IC
PMc
0.0083
0.0069
36
22
10
61
28
0.0250
Ni_XRF
PMc
0.0003
0.0005
31
6
0
19
0
0.0002
N03JC
PMc
0.4334
0.0028
36
36
36
100
100
0.0484
P_XRF
PMc
0.0342
0.0037
31
22
22
71
71
0.0070
Pb_XRF
PMc
0.0060
0.0018
31
15
5
48
16
0.0018
Rb_XRF
PMc
0.0002
0.0010
31
1
0
3
0
0.0004
S_XRF
PMc
0.1094
0.0026
31
31
31
100
100
0.0243
Sb_XRF
PMc
0.0020
0.0206
31
1
0
3
0
0.0189
Se_XRF
PMc
0.0001
0.0010
31
0
0
0
0
N/A
A-1
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
El
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Si_XRF
PMC
0.8686
0.0048
31
31
31
100
100
0.1180
Sn_XRF
PMC
0.0016
0.0146
31
0
0
0
0
N/A
S04_IC
PMC
0.1959
0.0038
36
36
36
100
100
0.0590
Sr_XRF
PMC
0.0024
0.0014
31
21
4
68
13
0.0007
Ti_XRF
PMC
0.0118
0.0017
31
28
23
90
74
0.0016
V_XRF
PMC
0.0004
0.0012
31
4
0
13
0
0.0008
Zn_XRF
PMC
0.0663
0.0009
31
31
28
100
90
0.0051
Zr_XRF
PMC
0.0022
0.0087
31
3
0
10
0
0.0048
Ag_XRF
PMf
0.0002
0.0162
31
0
0
0
0
N/A
AI_XRF
PMf
0.0406
0.0108
31
25
12
81
39
0.0073
As_XRF
PMf
0.0005
0.0011
31
6
1
19
3
0.0009
Ba_XRF
PMf
0.0029
0.0036
31
7
1
23
3
0.0024
Br_XRF
PMf
0.0044
0.0010
31
30
19
97
61
0.0007
Ca_XRF
PMf
0.0788
0.0019
31
31
31
100
100
0.0057
Cd_XRF
PMf
0.0002
0.0099
31
0
0
0
0
N/A
Ce_XRF
PMf
0.0000
0.0025
31
0
0
0
0
N/A
CI_XRF
PMf
0.0232
0.0021
31
31
27
100
87
0.0022
Co_XRF
PMf
0.0004
0.0006
31
6
0
19
0
0.0004
Cr_XRF
PMf
0.0010
0.0010
31
9
2
29
6
0.0005
Cs_XRF
PMf
0.0007
0.0045
31
2
0
6
0
0.0028
Cu_XRF
PMf
0.0046
0.0008
31
27
20
87
65
0.0006
Fe_XRF
PMf
0.0831
0.0008
31
31
31
100
100
0.0060
ln_XRF
PMf
0.0007
0.0144
31
0
0
0
0
N/A
K_IC
PMf
0.0577
0.0065
36
36
35
100
97
0.0045
K_XRF
PMf
0.0626
0.0018
31
31
31
100
100
0.0045
Mass_Grav
PMf
10.7419
0.3342
36
36
36
100
100
0.5517
Mg_XRF
PMf
0.0068
0.0044
31
17
5
55
16
0.0025
Mn_XRF
PMf
0.0025
0.0008
31
26
12
84
39
0.0004
Na_IC
PMf
0.0505
0.0129
36
33
17
92
47
0.0252
Na_XRF
PMf
0.0593
0.0119
31
30
17
97
55
0.0088
NH4_IC
PMf
1.2044
0.0076
36
36
36
100
100
0.0853
Ni_XRF
PMf
0.0002
0.0006
31
2
0
6
0
0.0003
N03_IC
PMf
1.0070
0.0031
36
36
36
100
100
0.0720
P_XRF
PMf
0.0001
0.0040
31
0
0
0
0
N/A
Pb_XRF
PMf
0.0070
0.0020
31
22
7
71
23
0.0016
Rb_XRF
PMf
0.0001
0.0012
31
0
0
0
0
N/A
A-2
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
H
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
S_XRF
PMf
0.8575
0.0028
31
31
31
100
100
0.0608
Sb_XRF
PMf
0.0031
0.0228
31
0
0
0
0
N/A
Se_XRF
PMf
0.0006
0.0011
31
5
0
16
0
0.0004
Si_XRF
PMf
0.1005
0.0054
31
31
29
100
94
0.0092
Sn_XRF
PMf
0.0008
0.0162
31
0
0
0
0
N/A
S04_IC
PMf
2.4379
0.0042
36
36
36
100
100
0.1734
Sr_XRF
PMf
0.0004
0.0016
31
0
0
0
0
N/A
Ti_XRF
PMf
0.0026
0.0019
31
12
3
39
10
0.0011
V_XRF
PMf
0.0009
0.0013
31
9
1
29
3
0.0006
Zn_XRF
PMf
0.0308
0.0010
31
31
31
100
100
0.0023
Zr_XRF
PMf
0.0004
0.0097
31
0
0
0
0
N/A
EC_TOT
PMC
0.3576
0.0640
41
29
25
71
61
N/A
OC_TOT
PMC
2.0415
0.0640
41
40
40
98
98
N/A
EC_TOT
PMf
0.5671
0.0640
41
40
37
98
90
N/A
OC_TOT
PMf
3.1828
0.0640
41
41
41
100
100
N/A
A-3
-------
EPA's Coarse PM Pilot Study
Appendix A
Table A-2. Summary of concentrations and MDLs by species and size fraction (PMf or
PMc) at St. Louis for FRM sampler, on Teflon and quartz filters; all species except OC
and EC are from Teflon filter measurements. Two species, K and Na, were analyzed via
XRF and IC; the latter analysis results are indicated by "_IC."
Species
Size
Avg
Cone
(pg/m3)
HI
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Ag_XRF
PMc
0.0002
0.0136
38
0
0
0
0
N/A
AI_XRF
PMc
0.3543
0.0094
38
38
38
100
100
0.0473
As_XRF
PMc
0.0003
0.0010
38
6
2
16
5
0.0011
Ba_XRF
PMc
0.0239
0.0034
38
37
32
97
84
0.0039
Br_XRF
PMc
0.0019
0.0009
38
16
7
42
18
0.0013
Ca_XRF
PMc
1.5866
0.0019
38
38
38
100
100
0.1361
Cd_XRF
PMc
0.0007
0.0087
38
1
0
3
0
0.0083
Ce_XRF
PMc
0.0024
0.0025
38
4
4
11
11
0.0035
CI_XRF
PMc
0.1622
0.0021
38
38
38
100
100
0.0138
Co_XRF
PMc
0.0008
0.0006
38
25
3
66
8
0.0007
Cr_XRF
PMc
0.0015
0.0009
38
25
7
66
18
0.0006
Cs_XRF
PMc
-0.0001
0.0041
38
1
0
3
0
0.0042
Cu_XRF
PMc
0.0071
0.0007
38
38
36
100
95
0.0011
Fe_XRF
PMc
0.4297
0.0007
38
38
38
100
100
0.0380
ln_XRF
PMc
0.0001
0.0123
38
0
0
0
0
N/A
K_IC
PMc
0.0391
0.0058
38
38
31
100
82
0.0091
K_XRF
PMc
0.1274
0.0019
38
38
38
100
100
0.0164
Mass_Grav
PMc
13.4626
0.3026
38
38
38
100
100
1.3528
Mg_XRF
PMc
0.0908
0.0042
38
36
35
95
92
0.0162
Mn_XRF
PMc
0.0104
0.0008
38
38
37
100
97
0.0012
Na_IC
PMc
0.1445
0.0117
38
35
23
92
61
0.0376
Na_XRF
PMc
0.1781
0.0116
38
37
26
97
68
0.0420
NH4_IC
PMc
-0.0399
0.0069
38
10
5
26
13
0.1082
Ni_XRF
PMc
0.0003
0.0005
38
14
0
37
0
0.0003
N03_IC
PMc
0.5952
0.0028
38
38
38
100
100
0.1253
P_XRF
PMc
0.0526
0.0040
38
36
34
95
89
0.0067
Pb_XRF
PMc
0.0048
0.0017
38
24
10
63
26
0.0026
Rb_XRF
PMc
0.0002
0.0010
38
2
0
5
0
0.0006
S_XRF
PMc
0.0661
0.0027
38
31
31
82
82
0.0787
Sb_XRF
PMc
0.0000
0.0201
38
1
0
3
0
0.0201
Se_XRF
PMc
0.0002
0.0009
38
3
0
8
0
0.0006
Si_XRF
PMc
1.2516
0.0049
38
38
38
100
100
0.1676
Sn_XRF
PMc
0.0007
0.0140
38
0
0
0
0
N/A
A-4
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
HI
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
S04JC
PMC
0.2262
0.0037
38
38
38
100
100
0.2505
Sr_XRF
PMC
0.0031
0.0013
38
33
10
87
26
0.0014
Ti_XRF
PMC
0.0179
0.0018
38
38
36
100
95
0.0022
V_XRF
PMC
0.0013
0.0012
38
9
4
24
11
0.0012
Zn_XRF
PMC
0.0326
0.0009
38
38
35
100
92
0.0043
Zr_XRF
PMC
0.0017
0.0080
38
3
0
8
0
0.0070
Ag_XRF
PMf
0.0001
0.0137
40
0
0
0
0
N/A
AI_XRF
PMf
0.0387
0.0094
40
38
18
95
45
0.0064
As_XRF
PMf
0.0006
0.0010
40
12
2
30
5
0.0008
Ba_XRF
PMf
0.0042
0.0033
40
17
3
43
8
0.0023
Br_XRF
PMf
0.0047
0.0009
40
40
31
100
78
0.0007
Ca_XRF
PMf
0.0971
0.0018
40
40
40
100
100
0.0070
Cd_XRF
PMf
0.0005
0.0087
40
1
0
3
0
0.0071
Ce_XRF
PMf
0.0001
0.0024
40
0
0
0
0
N/A
CI_XRF
PMf
0.0167
0.0020
40
39
29
98
73
0.0017
Co_XRF
PMf
0.0004
0.0006
40
9
0
23
0
0.0003
Cr_XRF
PMf
0.0008
0.0009
40
12
2
30
5
0.0004
Cs_XRF
PMf
0.0004
0.0040
40
0
0
0
0
N/A
Cu_XRF
PMf
0.0052
0.0007
40
38
28
95
70
0.0005
Fe_XRF
PMf
0.0956
0.0007
40
40
40
100
100
0.0068
ln_XRF
PMf
0.0009
0.0124
40
0
0
0
0
N/A
K_IC
PMf
0.0632
0.0059
40
39
39
98
98
0.0050
K_XRF
PMf
0.0717
0.0017
40
40
40
100
100
0.0052
Mass_Grav
PMf
10.9770
0.3025
40
40
40
100
100
0.5600
Mg_XRF
PMf
0.0088
0.0040
40
26
14
65
35
0.0024
Mn_XRF
PMf
0.0029
0.0008
40
39
21
98
53
0.0004
Na_IC
PMf
0.0433
0.0116
40
38
19
95
48
0.0207
Na_XRF
PMf
0.0507
0.0110
40
40
24
100
60
0.0077
NH4_IC
PMf
1.1790
0.0070
40
40
40
100
100
0.0836
Ni_XRF
PMf
0.0002
0.0005
40
3
0
8
0
0.0002
N03_IC
PMf
0.8343
0.0028
40
40
40
100
100
0.0597
P_XRF
PMf
0.0003
0.0039
40
1
0
3
0
0.0025
Pb_XRF
PMf
0.0076
0.0018
40
35
17
88
43
0.0015
Rb_XRF
PMf
0.0001
0.0010
40
1
0
3
0
0.0005
S_XRF
PMf
0.8726
0.0027
40
40
40
100
100
0.0619
Sb_XRF
PMf
0.0044
0.0202
40
2
0
5
0
0.0182
Se_XRF
PMf
0.0006
0.0009
40
9
0
23
0
0.0004
A-5
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
BIB
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Si_XRF
PMf
0.1107
0.0047
40
40
40
100
100
0.0099
Sn_XRF
PMf
0.0001
0.0140
40
0
0
0
0
N/A
S04_IC
PMf
2.4086
0.0036
40
40
40
100
100
0.1712
Sr_XRF
PMf
0.0004
0.0013
40
2
0
5
0
0.0006
Ti_XRF
PMf
0.0025
0.0017
40
19
2
48
5
0.0010
V_XRF
PMf
0.0009
0.0012
40
10
2
25
5
0.0006
Zn_XRF
PMf
0.0215
0.0009
40
40
40
100
100
0.0016
Zr_XRF
PMf
0.0004
0.0081
40
0
0
0
0
N/A
EC_TOT
PMC
0.3834
0.0640
26
24
20
92
77
N/A
OC_TOT
PMC
1.8947
0.0640
26
24
24
92
92
N/A
EC_TOT
PMf
0.5907
0.0640
29
29
27
100
93
N/A
OC_TOT
PMf
3.3402
0.0640
29
29
29
100
100
N/A
A-6
-------
EPA's Coarse PM Pilot Study
Appendix A
Table A-3. Summary of concentrations and MDLs by species and size fraction (PMf or
PMc) at Phoenix for primary dichot, on Teflon and quartz filters; all species except OC
and EC are from Teflon filter measurements. Two species, K and Na, were analyzed via
XRF and IC; the latter analysis results are indicated by "_IC."
Species
Size
Avg
Cone
(pg/m3)
Avg
MDL
(pg/m3
)
N
N>
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Ag_XRF
PMc
0.0001
0.0138
51
0
0
0
0
N/A
AI_XRF
PMc
1.1182
0.0094
51
51
51
100
100
0.1420
As_XRF
PMc
0.0003
0.0010
51
5
0
10
0
0.0007
Ba_XRF
PMc
0.0281
0.0033
51
51
47
100
92
0.0037
Br_XRF
PMc
0.0007
0.0009
51
22
1
43
2
0.0005
Ca_XRF
PMc
1.1447
0.0019
51
51
51
100
100
0.0982
Cd_XRF
PMc
0.0004
0.0087
51
0
0
0
0
N/A
Ce_XRF
PMc
0.0010
0.0024
51
5
1
10
2
0.0024
CI_XRF
PMc
0.2680
0.0021
51
51
51
100
100
0.0216
Co_XRF
PMc
0.0018
0.0006
51
44
27
86
53
0.0007
Cr_XRF
PMc
0.0027
0.0009
51
47
21
92
41
0.0005
Cs_XRF
PMc
0.0009
0.0043
51
7
0
14
0
0.0038
Cu_XRF
PMc
0.0151
0.0007
51
51
51
100
100
0.0013
Fe_XRF
PMc
0.7905
0.0007
51
51
51
100
100
0.0575
ln_XRF
PMc
0.0004
0.0124
51
0
0
0
0
N/A
K_IC
PMc
0.0732
0.0057
61
61
61
100
100
0.0064
K_XRF
PMc
0.3853
0.0018
51
51
51
100
100
0.0321
Mass_Grav
PMc
18.7492
0.3013
61
61
61
100
100
0.9885
Mg_XRF
PMc
0.1221
0.0040
51
51
51
100
100
0.0211
Mn_XRF
PMc
0.0156
0.0008
51
51
51
100
100
0.0013
Na_IC
PMc
0.2472
0.0120
61
61
61
100
100
0.0349
Na_XRF
PMc
0.2731
0.0113
51
51
51
100
100
0.0547
NH4JC
PMc
0.0039
0.0068
61
29
6
48
10
0.0076
Ni_XRF
PMc
0.0010
0.0005
51
40
10
78
20
0.0003
N03_IC
PMc
0.4154
0.0028
61
61
61
100
100
0.0368
P_XRF
PMc
0.0414
0.0039
51
51
46
100
90
0.0063
Pb_XRF
PMc
0.0026
0.0017
51
26
3
51
6
0.0013
Rb_XRF
PMc
0.0012
0.0010
51
29
1
57
2
0.0005
S_XRF
PMc
0.1274
0.0027
51
51
51
100
100
0.0151
Sb_XRF
PMc
0.0036
0.0202
51
4
0
8
0
0.0187
Se_XRF
PMc
0.0001
0.0009
51
0
0
0
0
N/A
Si_XRF
PMc
3.3396
0.0048
51
51
51
100
100
0.4380
A-7
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
Avg
MDL
(pg/m3
)
N
N>
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Sn_XRF
PMC
0.0016
0.0140
51
0
0
0
0
N/A
S04_IC
PMC
0.2836
0.0036
61
61
61
100
100
0.0324
Sr_XRF
PMC
0.0068
0.0013
51
47
39
92
76
0.0011
Ti_XRF
PMC
0.0575
0.0018
51
51
51
100
100
0.0045
V_XRF
PMC
0.0009
0.0012
51
12
2
24
4
0.0010
Zn_XRF
PMC
0.0354
0.0009
51
51
51
100
100
0.0029
Zr_XRF
PMC
0.0026
0.0081
51
3
0
6
0
0.0056
Ag_XRF
PMf
0.0002
0.0153
51
0
0
0
0
N/A
AI_XRF
PMf
0.1286
0.0105
51
51
51
100
100
0.0139
As_XRF
PMf
0.0006
0.0011
51
12
1
24
2
0.0008
Ba_XRF
PMf
0.0064
0.0037
51
29
11
57
22
0.0026
Br_XRF
PMf
0.0034
0.0010
51
51
32
100
63
0.0006
Ca_XRF
PMf
0.1543
0.0021
51
51
51
100
100
0.0114
Cd_XRF
PMf
0.0007
0.0097
51
0
0
0
0
N/A
Ce_XRF
PMf
0.0000
0.0027
51
0
0
0
0
N/A
CI_XRF
PMf
0.0792
0.0022
51
51
43
100
84
0.0060
Co_XRF
PMf
0.0006
0.0006
51
20
1
39
2
0.0005
Cr_XRF
PMf
0.0015
0.0010
51
29
4
57
8
0.0005
Cs_XRF
PMf
0.0010
0.0043
51
5
0
10
0
0.0029
Cu_XRF
PMf
0.0060
0.0008
51
51
43
100
84
0.0006
Fe_XRF
PMf
0.1990
0.0008
51
51
51
100
100
0.0141
ln_XRF
PMf
0.0001
0.0138
51
0
0
0
0
N/A
K_IC
PMf
0.0801
0.0063
61
57
56
93
92
0.0063
K_XRF
PMf
0.1336
0.0019
51
51
51
100
100
0.0097
Mass_Grav
PMf
7.6092
0.3325
61
61
61
100
100
0.3985
Mg_XRF
PMf
0.0264
0.0044
51
46
37
90
73
0.0041
Mn_XRF
PMf
0.0051
0.0008
51
49
43
96
84
0.0006
Na_IC
PMf
0.1084
0.0133
61
61
51
100
84
0.0264
Na_XRF
PMf
0.1013
0.0120
51
51
42
100
82
0.0130
NH4_IC
PMf
0.3586
0.0075
61
61
61
100
100
0.0254
Ni_XRF
PMf
0.0007
0.0006
51
20
3
39
6
0.0003
N03JC
PMf
0.5000
0.0031
61
61
61
100
100
0.0361
P_XRF
PMf
0.0005
0.0043
51
2
0
4
0
0.0032
Pb_XRF
PMf
0.0041
0.0019
51
26
9
51
18
0.0015
Rb_XRF
PMf
0.0002
0.0012
51
1
0
2
0
0.0006
S_XRF
PMf
0.3052
0.0030
51
51
51
100
100
0.0221
A-8
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
Avg
MDL
(pg/m3
)
N
N>
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Sb_XRF
PMf
0.0036
0.0224
51
2
0
4
0
0.0206
Se_XRF
PMf
0.0002
0.0010
51
2
0
4
0
0.0004
Si_XRF
PMf
0.3880
0.0053
51
51
51
100
100
0.0363
Sn_XRF
PMf
0.0013
0.0155
51
0
0
0
0
N/A
S04_IC
PMf
0.8465
0.0040
61
61
61
100
100
0.0608
Sr_XRF
PMf
0.0012
0.0015
51
13
1
25
2
0.0010
Ti_XRF
PMf
0.0083
0.0020
51
51
31
100
61
0.0013
V_XRF
PMf
0.0004
0.0013
51
5
0
10
0
0.0007
Zn_XRF
PMf
0.0227
0.0009
51
51
51
100
100
0.0017
Zr_XRF
PMf
0.0009
0.0090
51
0
0
0
0
N/A
EC_TOT
PMC
0.1919
0.0640
48
36
21
75
44
N/A
OC_TOT
PMC
2.3576
0.0640
48
48
48
100
100
N/A
EC_TOT
PMf
0.7180
0.0640
48
47
45
98
94
N/A
OC_TOT
PMf
2.4693
0.0640
48
48
48
100
100
N/A
A-9
-------
EPA's Coarse PM Pilot Study
Appendix A
Table A-4. Summary of concentrations and MDLs by species and size fraction (PMf or
PMc) at Phoenix for FRM sampler, on Teflon and quartz filters; all species except OC and
EC are from Teflon filter measurements. Two species, K and Na, were analyzed via XRF
and IC; the latter analysis results are indicated by "_IC."
Species
Size
Avg
Cone
(pg/m3)
m
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Ag_XRF
PMc
-0.0001
0.0136
44
0
0
0
0
N/A
AI_XRF
PMc
1.3564
0.0094
44
44
44
100
100
0.1738
As_XRF
PMc
0.0005
0.0010
44
13
1
30
2
0.0011
Ba_XRF
PMc
0.0369
0.0034
44
44
43
100
98
0.0056
Br_XRF
PMc
0.0012
0.0009
44
28
2
64
5
0.0009
Ca_XRF
PMc
1.4398
0.0020
44
44
44
100
100
0.1339
Cd_XRF
PMc
0.0005
0.0087
44
0
0
0
0
N/A
Ce_XRF
PMc
0.0013
0.0025
44
4
2
9
5
0.0029
CI_XRF
PMc
0.2782
0.0022
44
44
44
100
100
0.0279
Co_XRF
PMc
0.0019
0.0006
44
36
23
82
52
0.0010
Cr_XRF
PMc
0.0030
0.0009
44
37
22
84
50
0.0008
Cs_XRF
PMc
-0.0005
0.0042
44
4
1
9
2
0.0049
Cu_XRF
PMc
0.0186
0.0007
44
44
44
100
100
0.0021
Fe_XRF
PMc
0.9827
0.0007
44
44
44
100
100
0.0900
ln_XRF
PMc
0.0004
0.0123
44
0
0
0
0
N/A
K_IC
PMc
0.0773
0.0059
44
44
43
100
98
0.0144
K_XRF
PMc
0.4816
0.0020
44
44
44
100
100
0.0502
Mass_Grav
PMc
22.6664
0.3032
44
44
44
100
100
1.6672
Mg_XRF
PMc
0.1480
0.0043
44
44
44
100
100
0.0259
Mn_XRF
PMc
0.0194
0.0008
44
44
44
100
100
0.0021
Na_IC
PMc
0.2424
0.0116
44
44
44
100
100
0.0475
Na_XRF
PMc
0.3140
0.0121
44
44
44
100
100
0.0695
NH4JC
PMc
0.0030
0.0070
44
21
14
48
32
0.0314
Ni_XRF
PMc
0.0013
0.0005
44
37
16
84
36
0.0004
N03_IC
PMc
0.4946
0.0028
44
44
44
100
100
0.1030
P_XRF
PMc
0.0593
0.0040
44
43
42
98
95
0.0071
Pb_XRF
PMc
0.0036
0.0018
44
33
6
75
14
0.0021
Rb_XRF
PMc
0.0017
0.0010
44
33
6
75
14
0.0007
S_XRF
PMc
0.1212
0.0028
44
44
44
100
100
0.0375
Sb_XRF
PMc
0.0003
0.0202
44
0
0
0
0
N/A
Se_XRF
PMc
0.0001
0.0010
44
1
0
2
0
0.0006
Si_XRF
PMc
4.3101
0.0050
44
44
44
100
100
0.5878
Sn_XRF
PMc
-0.0014
0.0140
44
0
0
0
0
N/A
A-10
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
HI
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
S04JC
PMC
0.3097
0.0037
44
44
44
100
100
0.1026
Sr_XRF
PMC
0.0092
0.0013
44
42
40
95
91
0.0016
Ti_XRF
PMC
0.0728
0.0018
44
44
44
100
100
0.0065
V_XRF
PMC
0.0014
0.0012
44
16
3
36
7
0.0013
Zn_XRF
PMC
0.0530
0.0009
44
44
44
100
100
0.0053
Zr_XRF
PMC
0.0032
0.0080
44
6
2
14
5
0.0067
Ag_XRF
PMf
0.0003
0.0136
45
0
0
0
0
N/A
AI_XRF
PMf
0.1526
0.0094
45
45
45
100
100
0.0144
As_XRF
PMf
0.0005
0.0010
45
9
1
20
2
0.0008
Ba_XRF
PMf
0.0106
0.0033
45
31
22
69
49
0.0028
Br_XRF
PMf
0.0032
0.0009
45
44
25
98
56
0.0006
Ca_XRF
PMf
0.2033
0.0018
45
45
45
100
100
0.0145
Cd_XRF
PMf
0.0005
0.0087
45
1
0
2
0
0.0080
Ce_XRF
PMf
0.0000
0.0024
45
0
0
0
0
N/A
CI_XRF
PMf
0.0827
0.0020
45
45
43
100
96
0.0061
Co_XRF
PMf
0.0009
0.0006
45
29
3
64
7
0.0005
Cr_XRF
PMf
0.0018
0.0009
45
29
10
64
22
0.0005
Cs_XRF
PMf
0.0016
0.0041
45
9
0
20
0
0.0028
Cu_XRF
PMf
0.0077
0.0007
45
45
43
100
96
0.0007
Fe_XRF
PMf
0.2564
0.0007
45
45
45
100
100
0.0182
ln_XRF
PMf
0.0008
0.0123
45
0
0
0
0
N/A
K_IC
PMf
0.0920
0.0059
45
43
42
96
93
0.0072
K_XRF
PMf
0.1478
0.0017
45
45
45
100
100
0.0105
Mass_Grav
PMf
9.0100
0.3024
45
45
45
100
100
0.4649
Mg_XRF
PMf
0.0255
0.0040
45
43
35
96
78
0.0033
Mn_XRF
PMf
0.0063
0.0008
45
44
42
98
93
0.0006
Na_IC
PMf
0.1092
0.0116
45
45
42
100
93
0.0243
Na_XRF
PMf
0.0962
0.0110
45
45
39
100
87
0.0110
NH4_IC
PMf
0.4103
0.0069
45
45
45
100
100
0.0292
Ni_XRF
PMf
0.0009
0.0005
45
23
5
51
11
0.0003
N03_IC
PMf
0.7160
0.0028
45
45
45
100
100
0.0513
P_XRF
PMf
0.0017
0.0039
45
5
2
11
4
0.0032
Pb_XRF
PMf
0.0047
0.0018
45
23
10
51
22
0.0015
Rb_XRF
PMf
0.0002
0.0010
45
2
0
4
0
0.0005
S_XRF
PMf
0.3032
0.0027
45
45
45
100
100
0.0216
Sb_XRF
PMf
0.0021
0.0201
45
1
0
2
0
0.0156
Se_XRF
PMf
0.0002
0.0009
45
1
0
2
0
0.0005
A-11
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Size
Avg
Cone
(pg/m3)
HI
N
N >
MDL
N >
3MDL
%>
MDL
%>
3MD
L
Avg
Uncertainty
(Mg/m3)
Si_XRF
PMf
0.4651
0.0047
45
45
45
100
100
0.0388
Sn_XRF
PMf
0.0017
0.0140
45
0
0
0
0
N/A
S04_IC
PMf
0.8495
0.0037
45
45
45
100
100
0.0609
Sr_XRF
PMf
0.0017
0.0013
45
24
4
53
9
0.0007
Ti_XRF
PMf
0.0117
0.0017
45
45
38
100
84
0.0014
V_XRF
PMf
0.0004
0.0012
45
4
0
9
0
0.0007
Zn_XRF
PMf
0.0173
0.0009
45
45
45
100
100
0.0013
Zr_XRF
PMf
0.0013
0.0080
45
1
0
2
0
0.0047
EC_TOT
PMC
0.2818
0.0640
29
24
17
83
59
N/A
OC_TOT
PMC
2.1730
0.0640
29
28
27
97
93
N/A
EC_TOT
PMf
0.6420
0.0640
30
28
25
93
83
N/A
OC_TOT
PMf
2.3259
0.0640
30
30
30
100
100
N/A
A-12
-------
EPA's Coarse PM Pilot Study
Appendix A
Table A-5. Summary of concentrations by species and size fraction (PM10 or PM2 5) at St.
Louis for FRM sampler, on Teflon and quartz filters; all species except OC and EC are
from Teflon filter measurements. Two species, K and Na, were analyzed via XRF and IC;
the latter analysis results are indicated by "_IC."
Species
Avg PM25
(pg/m3)
Avg PM10
(Mg/m3)
Avg (PM10/PM25)
(Mg/m3)
Avg PM10-2.5
(Mg/m3)
EC_TOR
0.7776
1.244
1.8529
0.4664
OC_TOR
3.114
4.9257
1.7234
1.8117
Ag_XRF
0.0001
0.0003
0.6316
0.0002
AI_XRF
0.0389
0.3931
15.7212
0.3543
As_XRF
0.0006
0.0009
2.865
0.0003
Ba_XRF
0.0036
0.0275
8.5088
0.0239
Br_XRF
0.0047
0.0066
1.342
0.0019
Ca_XRF
0.0979
1.6845
17.6537
1.5866
Cd_XRF
0.0005
0.0012
0.7959
0.0007
Ce_XRF
0.0001
0.0025
10.0014
0.0024
CI_XRF
0.0173
0.1795
9.2252
0.1622
Co_XRF
0.0003
0.0012
3.6365
0.0008
Cr_XRF
0.0008
0.0023
14.3764
0.0015
Cs_XRF
0.0004
0.0004
2.1755
-0.0001
Cu_XRF
0.0052
0.0123
2.6465
0.0071
Fe_XRF
0.0977
0.5273
5.7266
0.4297
ln_XRF
0.0009
0.0011
0.8065
0.0001
K_IC
0.064
0.1031
1.7057
0.0391
K_XRF
0.0723
0.1997
2.9407
0.1274
Mass_Grav
10.9587
24.4213
2.3083
13.4626
Mg_XRF
0.0087
0.0995
24.7169
0.0908
Mn_XRF
0.003
0.0133
5.1863
0.0104
Na_IC
0.0444
0.1889
3.8968
0.1445
Na_XRF
0.0509
0.229
4.2644
0.1781
NH4_IC
1.1763
1.1363
0.9748
-0.0399
Ni_XRF
0.0002
0.0005
3.0465
0.0003
N03_IC
0.8753
1.4705
5.5111
0.5952
P_XRF
0.0003
0.0529
360.3697
0.0526
Pb_XRF
0.0077
0.0125
1.713
0.0048
Rb_XRF
0.0001
0.0004
1.1014
0.0002
S_XRF
0.859
0.9252
1.1038
0.0661
Sb_XRF
0.0043
0.0043
1.7227
0
Se_XRF
0.0006
0.0008
1.4583
0.0002
Si_XRF
0.1107
1.3622
14.1381
1.2516
A-13
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Avg PM25
(pg/m3)
Avg PM10
(Mg/m3)
Avg (PM10/PM25)
(Mg/m3)
Avg PM10-2.5
(Mg/m3)
Sn_XRF
0.0001
0.0007
1.4
0.0007
S04_IC
2.3774
2.6035
1.1025
0.2262
Sr_XRF
0.0004
0.0034
11.1222
0.0031
Ti_XRF
0.0025
0.0204
10.9788
0.0179
V_XRF
0.0009
0.0023
2.9389
0.0013
Zn_XRF
0.0222
0.0548
2.1138
0.0326
Zr_XRF
0.0004
0.0021
0.8464
0.0017
A-14
-------
EPA's Coarse PM Pilot Study
Appendix A
Table A-6. Summary of concentrations by species and size fraction (PM10 or PM2 5) at
Phoenix for FRM sampler, on Teflon and quartz filters; all species except OC and EC are
from Teflon filter measurements. Two species, K and Na, were analyzed via XRF and IC;
the latter analysis results are indicated by "_IC."
Species
Avg PM25
(pg/m3)
Avg PM10
(Mg/m3)
Avg
(PM10/PM25)
(Mg/m3)
Avg PM10-
2.5 (|jg/m3)
EC_TOR
0.705
0.9455
1.5653
0.2405
OC_TOR
2.2175
4.4318
2.0787
2.2143
Ag_XRF
0.0003
0.0002
0
-0.0001
AI_XRF
0.1548
1.5112
12.3484
1.3564
As_XRF
0.0004
0.0009
2.3656
0.0005
Ba_XRF
0.0105
0.0475
6.8446
0.0369
Br_XRF
0.0032
0.0044
1.5075
0.0012
Ca_XRF
0.2067
1.6465
9.2552
1.4398
Cd_XRF
0.0005
0.001
1.2469
0.0005
Ce_XRF
0
0.0014
0
0.0013
CI_XRF
0.0841
0.3623
7.2231
0.2782
Co_XRF
0.0009
0.0028
9.68
0.0019
Cr_XRF
0.0019
0.0048
5.2488
0.003
Cs_XRF
0.0016
0.0011
1.5787
-0.0005
Cu_XRF
0.0078
0.0263
3.7401
0.0185
Fe_XRF
0.2592
1.2418
5.5261
0.9827
ln_XRF
0.0009
0.0012
1.0127
0.0004
K_IC
0.0927
0.17
2.4991
0.0773
K_XRF
0.1492
0.6308
5.6951
0.4816
Mass_Grav
9.0859
31.7523
3.9565
22.6664
Mg_XRF
0.026
0.174
12.627
0.148
Mn_XRF
0.0063
0.0258
5.1943
0.0194
Na_IC
0.1105
0.3529
3.5033
0.2424
Na_XRF
0.0974
0.4115
4.8287
0.314
NH4JC
0.4099
0.4128
1.0612
0.003
Ni_XRF
0.0009
0.0022
4.3738
0.0013
N03_IC
0.7237
1.2184
2.8001
0.4946
P_XRF
0.0017
0.061
29.5247
0.0593
Pb_XRF
0.0048
0.0084
3.6215
0.0036
Rb_XRF
0.0002
0.002
6.657
0.0017
S_XRF
0.3023
0.4234
1.5213
0.1212
Sb_XRF
0.0021
0.0025
2.4151
0.0003
Se_XRF
0.0002
0.0002
1.5455
0
A-15
-------
EPA's Coarse PM Pilot Study
Appendix A
Species
Avg PM25
(pg/m3)
Avg PM10
(Mg/m3)
Avg
(PM10/PM25)
(Mg/m3)
Avg PM10-
2.s (|jg/m3)
Si_XRF
0.4725
4.7826
11.8616
4.3101
Sn_XRF
0.0018
0.0004
0.018
-0.0014
S04_IC
0.8447
1.1544
1.4591
0.3097
Sr_XRF
0.0017
0.0109
8.7813
0.0092
Ti_XRF
0.0118
0.0846
8.6226
0.0728
V_XRF
0.0004
0.0018
4.2143
0.0014
Zn_XRF
0.0175
0.0705
5.1821
0.053
Zr_XRF
0.0013
0.0045
2.8155
0.0032
A-16
-------
EPA's Coarse PM Pilot Study
Appendix B
Appendix B: Summary Ratios of
Collocated Dichot Measurements
Table B-1. Dichot B-to-Dichot A concentration ratios were calculated for each sample
and the means and medians of these distributions by species and size fraction (PMf or
PMc) at Phoenix and St. Louis on Teflon and quartz filters are reported below; all species
except OC and EC are from Teflon filter measurements. Two species, K and Na, were
analyzed via XRF and IC; the latter analysis results are indicated by "_IC." Species that
have at least 25% of samples above MDL are indicated in the last two columns.
Site ID
Size
Species
N >
MDL
Avg
(DichotB/
DichotA)
Med
(DichotB/
DichotA)
%>MDL
(Dichot
A)
%>MDL
(Dichot
B)
PHX
PMc
EC_TOT
5
0.95
0.68
68
75
PHX
PMc
OC_TOT
14
1.22
1.20
100
100
PHX
PMc
AI_XRF
22
0.95
0.94
100
100
PHX
PMc
As_XRF
1
0.50
0.50
PHX
PMc
Ba_XRF
22
0.94
0.93
100
100
PHX
PMc
Br_XRF
3
0.92
0.95
43
PHX
PMc
Ca_XRF
22
0.96
0.95
100
100
PHX
PMc
Ce_XRF
2
1.03
1.03
PHX
PMc
CI_XRF
22
0.94
0.93
100
100
PHX
PMc
Co_XRF
18
1.18
0.86
86
78
PHX
PMc
Cr_XRF
19
0.99
1.00
92
88
PHX
PMc
Cs_XRF
1
1.34
1.34
PHX
PMc
Cu_XRF
22
0.95
0.95
100
100
PHX
PMc
Fe_XRF
22
0.96
0.97
100
100
PHX
PMc
K_IC
31
0.98
0.98
100
100
PHX
PMc
K_XRF
22
0.96
0.96
100
100
PHX
PMc
Mass_Grav
31
0.97
0.95
100
100
PHX
PMc
Mg_XRF
22
0.97
1.01
100
100
PHX
PMc
Mn_XRF
22
0.96
0.97
100
100
PHX
PMc
Na_IC
31
1.01
0.95
100
100
PHX
PMc
Na_XRF
22
0.92
0.86
100
100
PHX
PMc
NH4JC
6
1.05
1.11
48
32
PHX
PMc
Ni_XRF
14
0.81
0.83
78
75
PHX
PMc
N03_IC
31
0.96
0.95
100
100
PHX
PMc
P_XRF
20
0.91
0.93
100
91
PHX
PMc
Pb_XRF
8
1.11
1.08
51
59
PHX
PMc
Rb_XRF
8
1.14
1.11
57
53
PHX
PMc
S_XRF
22
0.96
0.95
100
100
PHX
PMc
Si_XRF
22
0.96
0.96
100
100
B-1
-------
EPA's Coarse PM Pilot Study
Appendix B
N >
MDL
Avg
Med
%>MDL
%>MDL
Site ID
Size
Species
(DichotB/
(DichotB/
(Dichot
(Dichot
DichotA)
DichotA)
A)
B)
PHX
PMC
S04JC
31
0.99
0.94
100
100
PHX
PMC
Sr_XRF
20
0.94
0.94
92
91
PHX
PMC
Ti_XRF
22
0.97
0.95
100
100
PHX
PMC
V_XRF
3
1.04
1.11
28
PHX
PMC
Zn_XRF
22
0.98
0.98
100
100
PHX
PMf
Na_IC
3
1.21
1.28
PHX
PMf
NH4JC
19
0.98
0.99
56
62
PHX
PMf
N03JC
31
0.96
0.98
100
100
PHX
PMf
S04_IC
27
1.07
0.98
87
90
PHX
PMf
EC_TOT
13
1.18
1.07
93
98
PHX
PMf
OC_TOT
14
1.04
1.01
100
100
PHX
PMf
AI_XRF
22
1.13
1.08
100
97
PHX
PMf
As_XRF
2
0.89
0.89
PHX
PMf
Ba_XRF
11
1.03
0.85
57
62
PHX
PMf
Br_XRF
21
1.07
1.07
100
94
PHX
PMf
Ca_XRF
22
1.14
1.11
100
100
PHX
PMf
CI_XRF
22
1.11
1.09
100
97
PHX
PMf
Co_XRF
6
0.88
0.81
39
44
PHX
PMf
Cr_XRF
9
0.89
0.81
57
50
PHX
PMf
Cs_XRF
1
1.25
1.25
PHX
PMf
Cu_XRF
22
1.08
1.04
100
97
PHX
PMf
Fe_XRF
22
1.08
1.07
100
100
PHX
PMf
K_IC
28
1.00
1.00
93
93
PHX
PMf
K_XRF
22
1.07
1.07
100
100
PHX
PMf
Mass_Grav
31
1.06
1.01
100
100
PHX
PMf
Mg_XRF
18
1.15
1.04
90
88
PHX
PMf
Mn_XRF
21
1.11
1.13
96
94
PHX
PMf
Na_IC
31
1.04
1.02
100
100
PHX
PMf
Na_XRF
22
1.13
1.13
100
100
PHX
PMf
NH4JC
31
0.99
0.98
100
100
PHX
PMf
Ni_XRF
8
1.13
1.13
39
38
PHX
PMf
N03_IC
31
1.05
1.03
100
100
PHX
PMf
Pb_XRF
9
1.21
1.13
51
56
PHX
PMf
S_XRF
22
1.00
1.00
100
100
PHX
PMf
Si_XRF
22
1.12
1.12
100
100
PHX
PMf
S04_IC
31
1.00
0.99
100
100
PHX
PMf
Sr_XRF
5
1.30
1.17
25
47
B-2
-------
EPA's Coarse PM Pilot Study
Appendix B
N >
MDL
Avg
Med
%>MDL
%>MDL
Site ID
Size
Species
(DichotB/
(DichotB/
(Dichot
(Dichot
DichotA)
DichotA)
A)
B)
PHX
PMf
Ti_XRF
21
1.29
1.23
100
94
PHX
PMf
Zn_XRF
22
1.08
1.05
100
100
STL
PMC
EC_TOT
7
0.77
0.62
88
71
STL
PMC
OC_TOT
13
0.97
0.90
100
98
STL
PMC
AI_XRF
11
0.81
0.83
97
100
STL
PMC
Ba_XRF
11
0.93
0.94
90
93
STL
PMC
Br_XRF
4
1.19
1.15
39
36
STL
PMC
Ca_XRF
11
0.90
0.92
100
100
STL
PMC
CI_XRF
11
0.84
0.85
100
100
STL
PMC
Co_XRF
4
1.27
0.98
65
61
STL
PMC
Cr_XRF
4
1.06
1.05
61
68
STL
PMC
Cu_XRF
11
0.85
0.86
100
100
STL
PMC
Fe_XRF
11
0.88
0.91
100
100
STL
PMC
K_IC
8
0.89
0.93
92
90
STL
PMC
K_XRF
11
0.88
0.87
100
100
STL
PMC
Mass_Grav
11
0.84
0.89
100
100
STL
PMC
Mg_XRF
11
0.78
0.88
97
100
STL
PMC
Mn_XRF
10
0.90
0.89
97
100
STL
PMC
Na_IC
11
0.91
0.87
97
97
STL
PMC
Na_XRF
10
0.92
0.81
94
82
STL
PMC
NH4_IC
3
0.54
0.63
61
45
STL
PMC
Ni_XRF
1
1.10
1.10
STL
PMC
N03_IC
11
0.97
0.94
100
100
STL
PMC
P_XRF
8
0.83
0.85
71
82
STL
PMC
Pb_XRF
5
0.88
0.92
48
54
STL
PMC
S_XRF
11
0.92
0.86
100
100
STL
PMC
Si_XRF
11
0.90
0.91
100
100
STL
PMC
S04_IC
11
0.86
0.83
100
100
STL
PMC
Sr_XRF
6
0.83
0.82
68
64
STL
PMC
Ti_XRF
10
0.86
0.87
90
100
STL
PMC
V_XRF
1
1.10
1.10
STL
PMC
Zn_XRF
11
0.92
0.88
100
100
STL
PMf
Na_IC
1
1.08
1.08
STL
PMf
NH4_IC
9
1.10
1.09
83
87
STL
PMf
N03_IC
11
1.02
1.01
100
100
STL
PMf
S04_IC
11
1.27
1.08
100
97
STL
PMf
EC_TOT
13
0.82
0.80
100
98
B-3
-------
EPA's Coarse PM Pilot Study
Appendix B
N >
MDL
Avg
Med
%>MDL
%>MDL
Site ID
Size
Species
(DichotB/
(DichotB/
(Dichot
(Dichot
DichotA)
DichotA)
A)
B)
STL
PMf
OC_TOT
13
0.87
0.89
100
100
STL
PMf
AI_XRF
8
1.14
1.08
81
82
STL
PMf
As_XRF
1
1.90
1.90
29
STL
PMf
Ba_XRF
2
1.20
1.20
39
STL
PMf
Br_XRF
10
1.00
1.08
97
100
STL
PMf
Ca_XRF
11
1.23
1.15
100
100
STL
PMf
CI_XRF
11
0.91
0.94
100
96
STL
PMf
Cr_XRF
2
0.44
0.45
29
39
STL
PMf
Cs_XRF
1
1.28
1.28
STL
PMf
Cu_XRF
10
1.51
1.30
87
96
STL
PMf
Fe_XRF
11
1.11
1.04
100
100
STL
PMf
K_IC
11
0.98
0.99
100
97
STL
PMf
K_XRF
11
1.00
1.00
100
100
STL
PMf
Mass_Grav
11
1.00
1.02
100
100
STL
PMf
Mg_XRF
4
0.89
0.79
55
64
STL
PMf
Mn_XRF
7
1.05
1.02
84
86
STL
PMf
Na_IC
10
1.06
1.07
92
93
STL
PMf
Na_XRF
10
1.24
1.21
97
100
STL
PMf
NH4_IC
11
0.96
0.96
100
100
STL
PMf
N03JC
11
0.93
0.96
100
100
STL
PMf
Pb_XRF
6
0.96
0.89
71
82
STL
PMf
S_XRF
11
0.96
0.97
100
100
STL
PMf
Si_XRF
11
1.18
1.15
100
100
STL
PMf
S04_IC
11
0.98
0.98
100
100
STL
PMf
Ti_XRF
3
1.40
1.10
39
50
STL
PMf
V_XRF
1
0.74
0.74
29
STL
PMf
Zn_XRF
11
1.15
1.03
100
100
B-4
-------
EPA's Coarse PM Pilot Study
Appendix C
Appendix C: Summary of Dichot-to-FRM Comparisons
Table C-1. Dichot-to-FRM concentration ratios were calculated for each sample and the
means and medians of these distributions by species and size fraction (PMf or PMC) at
Phoenix and St. Louis on Teflon and quartz filters are reported below; all species except
OC and EC are from Teflon filter measurements. Two species, K and Na, were analyzed
via XRF and IC; the latter analysis results are indicated by "_IC." Species that have at
least 25% of samples above MDL are indicated in the last two columns. Elements were
analyzed by XRF, and ions were analyzed by IC. Potassium and sodium were analyzed
by both methods.
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
PHX-Dichot_A
PMC
EC_TOT
13
0.92
0.86
83
68
PHX-Dichot_A
PMC
OC_TOT
20
1.13
0.98
97
100
PHX-Dichot_A
PMC
AI_XRF
35
0.90
0.93
100
100
PHX-Dichot_A
PMC
As_XRF
2
0.86
0.87
30
PHX-Dichot_A
PMC
Ba_XRF
35
0.85
0.80
100
100
PHX-Dichot_A
PMC
Br_XRF
8
0.74
0.63
64
43
PHX-Dichot_A
PMC
Ca_XRF
35
0.86
0.87
100
100
PHX-Dichot_A
PMC
Ce_XRF
3
0.78
0.86
PHX-Dichot_A
PMC
CI_XRF
35
0.92
0.94
100
100
PHX-Dichot_A
PMC
Co_XRF
28
1.06
0.93
82
86
PHX-Dichot_A
PMC
Cr_XRF
29
0.99
0.96
84
92
PHX-Dichot_A
PMC
Cs_XRF
2
0.80
0.81
PHX-Dichot_A
PMC
Cu_XRF
35
0.99
0.93
100
100
PHX-Dichot_A
PMC
Fe_XRF
35
0.85
0.85
100
100
PHX-Dichot_A
PMC
K_IC
35
1.01
0.86
100
100
PHX-Dichot_A
PMC
K_XRF
35
0.83
0.82
100
100
PHX-Dichot_A
PMC
Mass_Grav
35
0.83
0.85
100
100
PHX-Dichot_A
PMC
Mg_XRF
35
0.91
0.90
100
100
PHX-Dichot_A
PMC
Mn_XRF
35
0.84
0.84
100
100
PHX-Dichot_A
PMC
Na_IC
35
0.90
0.93
100
100
PHX-Dichot_A
PMC
Na_XRF
35
0.87
0.89
100
100
PHX-Dichot_A
PMC
NH4_IC
11
0.56
0.59
48
48
PHX-Dichot_A
PMC
Ni_XRF
24
0.85
0.82
84
78
PHX-Dichot_A
PMC
N03_IC
35
0.83
0.83
100
100
PHX-Dichot_A
PMC
P_XRF
34
0.68
0.66
98
100
PHX-Dichot_A
PMC
Pb_XRF
18
0.84
0.71
75
51
PHX-Dichot_A
PMC
Rb_XRF
15
0.75
0.69
75
57
C-1
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
PHX-Dichot_A
PMC
S_XRF
35
1.17
1.08
100
100
PHX-Dichot_A
PMC
Si_XRF
35
0.85
0.87
100
100
PHX-Dichot_A
PMC
S04JC
35
0.93
0.93
100
100
PHX-Dichot_A
PMC
Sr_XRF
33
0.76
0.77
95
92
PHX-Dichot_A
PMC
Ti_XRF
35
0.85
0.86
100
100
PHX-Dichot_A
PMC
V_XRF
7
0.92
0.71
36
PHX-Dichot_A
PMC
Zn_XRF
35
0.83
0.77
100
100
PHX-Dichot_A
PMC
Zr_XRF
1
1.47
1.47
PHX-Dichot_A
PMf
Na_IC
1
0.98
0.98
PHX-Dichot_A
PMf
NH4_IC
22
1.00
0.99
60
56
PHX-Dichot_A
PMf
N03_IC
36
1.03
1.03
100
100
PHX-Dichot_A
PMf
S04_IC
22
1.49
1.36
73
87
PHX-Dichot_A
PMf
EC_TOT
18
1.06
0.96
93
93
PHX-Dichot_A
PMf
OC_TOT
21
1.03
0.99
100
100
PHX-Dichot_A
PMf
AI_XRF
36
0.94
0.93
100
100
PHX-Dichot_A
PMf
As_XRF
2
0.90
0.90
PHX-Dichot_A
PMf
Ba_XRF
18
0.88
0.98
69
57
PHX-Dichot_A
PMf
Br_XRF
35
1.07
1.09
98
100
PHX-Dichot_A
PMf
Ca_XRF
36
0.89
0.89
100
100
PHX-Dichot_A
PMf
CI_XRF
36
0.94
0.87
100
100
PHX-Dichot_A
PMf
Co_XRF
12
0.99
0.83
64
39
PHX-Dichot_A
PMf
Cr_XRF
18
0.87
0.90
64
57
PHX-Dichot_A
PMf
Cs_XRF
1
0.93
0.93
PHX-Dichot_A
PMf
Cu_XRF
36
0.87
0.85
100
100
PHX-Dichot_A
PMf
Fe_XRF
36
0.89
0.91
100
100
PHX-Dichot_A
PMf
K_IC
34
1.00
1.00
96
93
PHX-Dichot_A
PMf
K_XRF
36
0.93
0.94
100
100
PHX-Dichot_A
PMf
Mass_Grav
36
0.92
0.94
100
100
PHX-Dichot_A
PMf
Mg_XRF
30
1.48
0.98
96
90
PHX-Dichot_A
PMf
Mn_XRF
34
0.91
0.94
98
96
PHX-Dichot_A
PMf
Na_IC
36
0.89
0.85
100
100
PHX-Dichot_A
PMf
Na_XRF
36
0.97
0.88
100
100
PHX-Dichot_A
PMf
NH4_IC
36
0.97
0.99
100
100
PHX-Dichot_A
PMf
Ni_XRF
12
0.92
0.88
51
39
PHX-Dichot_A
PMf
NQ3JC
36
0.94
0.93
100
100
C-2
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
PHX-Dichot_A
PMf
P_XRF
1
0.79
0.79
PHX-Dichot_A
PMf
Pb_XRF
14
0.91
0.87
51
51
PHX-Dichot_A
PMf
Rb_XRF
1
1.27
1.27
PHX-Dichot_A
PMf
S_XRF
36
1.00
1.02
100
100
PHX-Dichot_A
PMf
Si_XRF
36
0.92
0.89
100
100
PHX-Dichot_A
PMf
S04_IC
36
0.97
0.99
100
100
PHX-Dichot_A
PMf
Sr_XRF
8
1.12
1.09
53
25
PHX-Dichot_A
PMf
Ti_XRF
36
0.84
0.81
100
100
PHX-Dichot_A
PMf
V_XRF
2
1.43
1.43
PHX-Dichot_A
PMf
Zn_XRF
36
1.01
0.97
100
100
PHX-Dichot_B
PMC
EC_TOT
11
1.18
0.63
83
75
PHX-Dichot_B
PMC
OC_TOT
22
1.40
1.13
97
100
PHX-Dichot_B
PMC
AI_XRF
25
0.84
0.87
100
100
PHX-Dichot_B
PMC
As_XRF
4
0.96
0.95
30
PHX-Dichot_B
PMC
Ba_XRF
25
0.79
0.76
100
100
PHX-Dichot_B
PMC
Br_XRF
4
0.84
0.74
64
PHX-Dichot_B
PMC
Ca_XRF
25
0.81
0.82
100
100
PHX-Dichot_B
PMC
Ce_XRF
3
0.70
0.64
PHX-Dichot_B
PMC
CI_XRF
25
0.89
0.88
100
100
PHX-Dichot_B
PMC
Co_XRF
17
1.07
0.84
82
78
PHX-Dichot_B
PMC
Cr_XRF
18
0.81
0.84
84
88
PHX-Dichot_B
PMC
Cs_XRF
1
0.83
0.83
PHX-Dichot_B
PMC
Cu_XRF
25
0.90
0.89
100
100
PHX-Dichot_B
PMC
Fe_XRF
25
0.80
0.81
100
100
PHX-Dichot_B
PMC
K_IC
25
1.02
0.92
100
100
PHX-Dichot_B
PMC
K_XRF
25
0.79
0.78
100
100
PHX-Dichot_B
PMC
Mass_Grav
25
0.81
0.83
100
100
PHX-Dichot_B
PMC
Mg_XRF
25
0.87
0.93
100
100
PHX-Dichot_B
PMC
Mn_XRF
25
0.81
0.82
100
100
PHX-Dichot_B
PMC
Na_IC
25
0.88
0.89
100
100
PHX-Dichot_B
PMC
Na_XRF
25
0.76
0.85
100
100
PHX-Dichot_B
PMC
NH4_IC
7
0.44
0.42
48
32
PHX-Dichot_B
PMC
Ni_XRF
16
0.75
0.71
84
75
PHX-Dichot_B
PMC
N03_IC
25
0.80
0.76
100
100
PHX-Dichot_B
PMC
P_XRF
23
0.65
0.63
98
91
C-3
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
PHX-Dichot_B
PMC
Pb_XRF
11
0.71
0.54
75
59
PHX-Dichot_B
PMC
Rb_XRF
12
0.87
0.87
75
53
PHX-Dichot_B
PMC
S_XRF
25
1.11
0.96
100
100
PHX-Dichot_B
PMC
Si_XRF
25
0.79
0.82
100
100
PHX-Dichot_B
PMC
S04_IC
25
0.89
0.90
100
100
PHX-Dichot_B
PMC
Sr_XRF
23
0.84
0.81
95
91
PHX-Dichot_B
PMC
Ti_XRF
25
0.79
0.80
100
100
PHX-Dichot_B
PMC
V_XRF
5
0.78
0.64
36
28
PHX-Dichot_B
PMC
Zn_XRF
25
0.80
0.80
100
100
PHX-Dichot_B
PMf
NH4JC
14
1.01
1.05
60
62
PHX-Dichot_B
PMf
N03_IC
25
1.04
1.04
100
100
PHX-Dichot_B
PMf
S04_IC
18
1.32
1.14
73
90
PHX-Dichot_B
PMf
EC_TOT
20
1.14
1.05
93
98
PHX-Dichot_B
PMf
OC_TOT
22
1.05
1.03
100
100
PHX-Dichot_B
PMf
AI_XRF
25
0.99
0.94
100
97
PHX-Dichot_B
PMf
As_XRF
3
1.44
0.82
PHX-Dichot_B
PMf
Ba_XRF
13
0.73
0.66
69
62
PHX-Dichot_B
PMf
Br_XRF
24
1.09
1.04
98
94
PHX-Dichot_B
PMf
Ca_XRF
25
0.89
0.91
100
100
PHX-Dichot_B
PMf
CI_XRF
25
0.93
0.89
100
97
PHX-Dichot_B
PMf
Co_XRF
9
1.01
1.03
64
44
PHX-Dichot_B
PMf
Cr_XRF
11
0.90
0.72
64
50
PHX-Dichot_B
PMf
Cs_XRF
2
0.84
0.84
PHX-Dichot_B
PMf
Cu_XRF
25
0.82
0.82
100
97
PHX-Dichot_B
PMf
Fe_XRF
25
0.88
0.89
100
100
PHX-Dichot_B
PMf
K_IC
23
0.99
0.98
96
93
PHX-Dichot_B
PMf
K_XRF
25
0.91
0.93
100
100
PHX-Dichot_B
PMf
Mass_Grav
25
0.92
0.95
100
100
PHX-Dichot_B
PMf
Mg_XRF
22
1.07
0.96
96
88
PHX-Dichot_B
PMf
Mn_XRF
24
0.86
0.86
98
94
PHX-Dichot_B
PMf
Na_IC
25
0.91
0.87
100
100
PHX-Dichot_B
PMf
Na_XRF
25
1.08
0.95
100
100
PHX-Dichot_B
PMf
NH4_IC
25
0.94
0.97
100
100
PHX-Dichot_B
PMf
Ni_XRF
9
0.85
0.76
51
38
PHX-Dichot_B
PMf
NQ3JC
25
0.91
0.89
100
100
C-4
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
PHX-Dichot_B
PMf
Pb_XRF
12
1.00
0.94
51
56
PHX-Dichot_B
PMf
Rb_XRF
1
1.25
1.25
PHX-Dichot_B
PMf
S_XRF
25
0.96
0.95
100
100
PHX-Dichot_B
PMf
Se_XRF
1
1.09
1.09
PHX-Dichot_B
PMf
Si_XRF
25
0.92
0.92
100
100
PHX-Dichot_B
PMf
S04_IC
25
0.95
0.97
100
100
PHX-Dichot_B
PMf
Sr_XRF
9
1.00
1.06
53
47
PHX-Dichot_B
PMf
Ti_XRF
24
0.91
0.86
100
94
PHX-Dichot_B
PMf
Zn_XRF
25
1.03
0.97
100
100
STL-Dichot_A
PMC
EC_TOT
10
0.91
0.56
92
88
STL-Dichot_A
PMC
OC_TOT
11
1.08
1.04
92
100
STL-Dichot_A
PMC
AI_XRF
18
0.99
0.98
100
97
STL-Dichot_A
PMC
Ba_XRF
16
0.85
0.76
97
90
STL-Dichot_A
PMC
Br_XRF
5
0.77
0.86
42
39
STL-Dichot_A
PMC
Ca_XRF
18
0.97
0.96
100
100
STL-Dichot_A
PMC
CI_XRF
18
1.04
1.03
100
100
STL-Dichot_A
PMC
Co_XRF
10
0.97
0.99
66
65
STL-Dichot_A
PMC
Cr_XRF
9
1.38
1.14
66
61
STL-Dichot_A
PMC
Cu_XRF
18
1.05
0.99
100
100
STL-Dichot_A
PMC
Fe_XRF
18
0.94
0.90
100
100
STL-Dichot_A
PMC
K_IC
18
1.46
1.15
100
92
STL-Dichot_A
PMC
K_XRF
18
0.90
0.88
100
100
STL-Dichot_A
PMC
Mass_Grav
18
0.90
0.93
100
100
STL-Dichot_A
PMC
Mg_XRF
16
1.04
0.97
95
97
STL-Dichot_A
PMC
Mn_XRF
18
0.96
0.91
100
97
STL-Dichot_A
PMC
Na_IC
17
0.99
0.94
92
97
STL-Dichot_A
PMC
Na_XRF
16
0.94
0.88
97
94
STL-Dichot_A
PMC
NH4_IC
2
0.16
0.16
26
61
STL-Dichot_A
PMC
Ni_XRF
1
0.95
0.95
37
STL-Dichot_A
PMC
N03JC
18
0.82
0.86
100
100
STL-Dichot_A
PMC
P_XRF
14
0.86
0.91
95
71
STL-Dichot_A
PMC
Pb_XRF
6
1.23
0.93
63
48
STL-Dichot_A
PMC
S_XRF
15
1.12
0.89
82
100
STL-Dichot_A
PMC
Si_XRF
18
0.96
0.96
100
100
STL-Dichot_A
PMC
SQ4JC
18
0.77
0.74
100
100
C-5
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
STL-Dichot_A
PMC
Sr_XRF
13
1.01
0.87
87
68
STL-Dichot_A
PMC
Ti_XRF
17
0.95
0.93
100
90
STL-Dichot_A
PMC
V_XRF
2
0.78
0.79
STL-Dichot_A
PMC
Zn_XRF
18
1.01
0.99
100
100
STL-Dichot_A
PMC
Zr_XRF
2
1.00
1.01
STL-Dichot_A
PMf
Na_IC
2
1.15
1.16
STL-Dichot_A
PMf
NH4_IC
14
1.05
0.96
85
83
STL-Dichot_A
PMf
N03_IC
19
1.06
1.02
100
100
STL-Dichot_A
PMf
S04_IC
18
1.96
1.36
88
100
STL-Dichot_A
PMf
EC_TOT
16
1.17
1.16
100
100
STL-Dichot_A
PMf
OC_TOT
16
1.11
1.11
100
100
STL-Dichot_A
PMf
AI_XRF
14
1.15
1.12
95
81
STL-Dichot_A
PMf
Ba_XRF
5
1.05
1.06
42
STL-Dichot_A
PMf
Br_XRF
18
0.98
1.01
100
97
STL-Dichot_A
PMf
Ca_XRF
19
1.07
0.99
100
100
STL-Dichot_A
PMf
CI_XRF
18
1.29
1.02
98
100
STL-Dichot_A
PMf
Co_XRF
1
0.89
0.89
STL-Dichot_A
PMf
Cr_XRF
3
1.05
1.20
30
29
STL-Dichot_A
PMf
Cu_XRF
15
1.16
1.21
95
87
STL-Dichot_A
PMf
Fe_XRF
19
1.05
1.02
100
100
STL-Dichot_A
PMf
K_IC
19
1.04
1.01
98
100
STL-Dichot_A
PMf
K_XRF
19
1.00
1.02
100
100
STL-Dichot_A
PMf
Mass_Grav
19
1.01
1.00
100
100
STL-Dichot_A
PMf
Mg_XRF
7
1.38
1.02
65
55
STL-Dichot_A
PMf
Mn_XRF
15
1.04
1.01
98
84
STL-Dichot_A
PMf
Na_IC
18
1.01
0.99
95
92
STL-Dichot_A
PMf
Na_XRF
19
0.88
0.83
100
97
STL-Dichot_A
PMf
NH4_IC
19
0.98
0.98
100
100
STL-Dichot_A
PMf
Ni_XRF
1
1.98
1.98
STL-Dichot_A
PMf
N03JC
19
1.01
1.00
100
100
STL-Dichot_A
PMf
Pb_XRF
13
1.06
1.12
88
71
STL-Dichot_A
PMf
S_XRF
19
0.99
1.01
100
100
STL-Dichot_A
PMf
Se_XRF
2
1.21
1.21
STL-Dichot_A
PMf
Si_XRF
19
1.02
0.98
100
100
STL-Dichot_A
PMf
SQ4JC
19
1.01
1.00
100
100
C-6
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
STL-Dichot_A
PMf
Ti_XRF
3
0.80
0.82
48
39
STL-Dichot_A
PMf
V_XRF
5
0.98
0.93
29
STL-Dichot_A
PMf
Zn_XRF
19
0.99
0.94
100
100
STL-Dichot_B
PMC
EC_TOT
10
0.99
0.92
92
71
STL-Dichot_B
PMC
OC_TOT
15
1.11
1.10
92
98
STL-Dichot_B
PMC
AI_XRF
21
0.86
0.84
100
100
STL-Dichot_B
PMC
Ba_XRF
19
0.84
0.71
97
93
STL-Dichot_B
PMC
Br_XRF
4
1.49
1.15
42
36
STL-Dichot_B
PMC
Ca_XRF
21
0.88
0.90
100
100
STL-Dichot_B
PMC
Ce_XRF
2
0.69
0.69
STL-Dichot_B
PMC
CI_XRF
21
1.10
0.90
100
100
STL-Dichot_B
PMC
Co_XRF
7
1.45
1.77
66
61
STL-Dichot_B
PMC
Cr_XRF
8
1.17
1.12
66
68
STL-Dichot_B
PMC
Cu_XRF
21
0.81
0.86
100
100
STL-Dichot_B
PMC
Fe_XRF
21
0.85
0.82
100
100
STL-Dichot_B
PMC
K_IC
20
1.08
0.93
100
90
STL-Dichot_B
PMC
K_XRF
21
0.84
0.83
100
100
STL-Dichot_B
PMC
Mass_Grav
21
0.80
0.79
100
100
STL-Dichot_B
PMC
Mg_XRF
21
0.93
0.85
95
100
STL-Dichot_B
PMC
Mn_XRF
21
0.85
0.83
100
100
STL-Dichot_B
PMC
Na_IC
20
0.90
0.86
92
97
STL-Dichot_B
PMC
Na_XRF
16
0.76
0.75
97
82
STL-Dichot_B
PMC
NH4_IC
3
1.33
0.97
26
45
STL-Dichot_B
PMC
Ni_XRF
3
1.10
1.10
37
STL-Dichot_B
PMC
N03JC
21
0.69
0.80
100
100
STL-Dichot_B
PMC
P_XRF
18
0.77
0.80
95
82
STL-Dichot_B
PMC
Pb_XRF
11
0.96
0.86
63
54
STL-Dichot_B
PMC
S_XRF
17
1.75
1.17
82
100
STL-Dichot_B
PMC
Si_XRF
21
0.86
0.85
100
100
STL-Dichot_B
PMC
S04JC
21
0.73
0.69
100
100
STL-Dichot_B
PMC
Sr_XRF
13
0.79
0.74
87
64
STL-Dichot_B
PMC
Ti_XRF
21
0.85
0.82
100
100
STL-Dichot_B
PMC
V_XRF
2
0.57
0.57
STL-Dichot_B
PMC
Zn_XRF
21
0.97
0.90
100
100
STL-Dichot_B
PMf
Na_IC
2
1.06
1.07
C-7
-------
EPA's Coarse PM Pilot Study
Appendix C
SamplerJD
Size
Species
N >
MDL
Avg
(Dichot/
FRM)
Med
(Dichot/
FRM)
%> MDL
(FRM)
%> MDL
(Dichot)
STL-Dichot_B
PMf
NH4JC
18
1.00
0.97
85
87
STL-Dichot_B
PMf
N03JC
21
1.05
1.02
100
100
STL-Dichot_B
PMf
S04JC
18
2.33
1.79
88
97
STL-Dichot_B
PMf
EC_TOT
19
0.96
0.98
100
98
STL-Dichot_B
PMf
OC_TOT
19
0.98
1.05
100
100
STL-Dichot_B
PMf
AI_XRF
17
1.34
1.11
95
82
STL-Dichot_B
PMf
As_XRF
5
1.37
1.53
30
29
STL-Dichot_B
PMf
Ba_XRF
8
1.40
1.46
42
39
STL-Dichot_B
PMf
Br_XRF
21
0.93
0.94
100
100
STL-Dichot_B
PMf
Ca_XRF
21
1.17
1.17
100
100
STL-Dichot_B
PMf
CI_XRF
20
1.04
1.03
98
96
STL-Dichot_B
PMf
Co_XRF
3
1.41
1.32
STL-Dichot_B
PMf
Cr_XRF
3
1.04
0.86
30
39
STL-Dichot_B
PMf
Cu_XRF
20
1.26
1.02
95
96
STL-Dichot_B
PMf
Fe_XRF
21
1.08
1.09
100
100
STL-Dichot_B
PMf
K_IC
20
1.02
1.03
98
97
STL-Dichot_B
PMf
K_XRF
21
0.99
0.99
100
100
STL-Dichot_B
PMf
Mass_Grav
21
1.01
1.01
100
100
STL-Dichot_B
PMf
Mg_XRF
13
1.02
0.90
65
64
STL-Dichot_B
PMf
Mn_XRF
18
1.17
1.15
98
86
STL-Dichot_B
PMf
Na_IC
20
1.06
1.02
95
93
STL-Dichot_B
PMf
Na_XRF
21
0.99
0.94
100
100
STL-Dichot_B
PMf
NH4_IC
21
0.96
0.96
100
100
STL-Dichot_B
PMf
N03_IC
21
1.18
1.08
100
100
STL-Dichot_B
PMf
Pb_XRF
15
0.96
0.83
88
82
STL-Dichot_B
PMf
S_XRF
21
0.97
0.96
100
100
STL-Dichot_B
PMf
Se_XRF
1
0.91
0.91
STL-Dichot_B
PMf
Si_XRF
21
1.14
1.12
100
100
STL-Dichot_B
PMf
S04_IC
21
0.99
0.99
100
100
STL-Dichot_B
PMf
Ti_XRF
7
0.94
0.90
48
50
STL-Dichot_B
PMf
V_XRF
3
1.07
0.81
STL-Dichot_B
PMf
Zn_XRF
21
1.06
1.00
100
100
C-8
-------
EPA's Coarse PM Pilot Study
Appendix D
Appendix D: Quartz Fiber Filter Carbon Blanks
Seven sets of trip blanks and field blanks (collected in May, July, September, and
November 2010, and January, April, and May 2011) were analyzed by thermal optical
reflectance (TOR) using the IMPROVE_A thermal-optical analysis (TOA) method. Each set
included two to four trip blank filters and four to twelve field blank filters per site. OC loadings on
these trip blanks and field blanks filters provide a context for characterizing and interpreting OC
on the front and backup filters. The OC minimum detection limit (MDL) for the IMPROVE_A
method is 2.03 |jg/filter. As summarized in this section, the trip blanks and field blanks data can
be used to estimate the lower quantifiable limits (LQL) for particulate matter OC. LQL is a metric
that complements the MDL estimate and aids in data interpretation.
Figure D-1 shows the mean and standard deviation of OC mass loadings for each trip
blank and field blank collection event at each site. For the trip blanks (Figure D-1 a), the OC
mass loadings were highest for the first collection event (May 2010) and relatively low and
consistent during the remainder of the study. Exceptions were the April and May 2011 samples
for St. Louis, for which the trip blank mass loadings exhibited relatively high mean values. At St.
Louis a single sample (mass loading 34 |jg/filter) was responsible for the high mean value in
April 2011; by excluding this sample, the April 2011 mean loading decreased from 10 |jg/filter to
2 |jg/filter. In contrast, the May 2011 mean loading of 10 |jg/filter at St. Louis included two high
samples (16 and 12 |jg/filter), and two low samples (6 and 5 |jg/filter).
The field blanks (Figure D-1b) initially decreased with each event, exhibited nearly
constant values from November 2010 to April 2011, and increased in May 2011.
30
CD
= 25
o>
=L
g 20 -
cc
o
3 15
(U
E
c
_Q
a
CO
CD
E
10
5 -
~ PHX-Trip
~ STL-Trip
A*
I
^ op ^ ^ ^ ^
(a)
30
tu
S 25
O)
i
20 -
aj
o
s15
03
E
c
_Q
10
CO
CD
E
5 -
~ PHX - Field
~ STL - Field
i
i
i
i
,NQ
^ ^ of # Vs ^
(b)
Figure D-1. Mean and standard deviation of OC (a) trip blanks data and (b) field blanks
data. Data are stratified by site and by the seven collection events between May 2010
and May 2011.
D-1
-------
EPA's Coarse PM Pilot Study
Appendix D
Statistical analyses were performed separately for the trip blanks and field blanks data
sets. In each case, unbalanced two-way analysis of variance (ANOVA) was performed on log-
transformed OC with site and date as factors. All hypothesis testing was conducted for the 95%
confidence level. For the trip blank data, site had an insignificant effect (p = 0.62), whereas date
had a significant effect (p < 0.01) on the OC mass loadings. Trip blank mass loading values
measured in May 2010, the first month of the study, were relatively high compared to
subsequent months. ANOVA with the May 2010 data excluded resulted in insignificant effects
for both the site and date (p = 0.51 and p = 0.18, respectively). Pooling all the trip blanks data
(but excluding the May 2010 samples) yields mean and median OC mass loadings of 5.4 ± 5.7
|jg/filter and 4.2 |jg/filter, respectively (N = 44).
The same statistical analyses were performed for the seven sets of field blanks, which
were collected at the same times as the trip blanks. Each set included four to twelve filters per
site, with the maximum case being front and back filters in the PM2.5 FRM, PM10 FRM, and both
channels of both dichot samplers. While in principle the sampler type and filter position (front
versus back) could be treated as additional factors, their effects were deemed insignificant, and
samples pooled across these factors were treated as pseudo-replicates. Site had an
insignificant effect (p = 0.28), whereas date had a significant effect (p < 0.01). After removing
the May 2010 data, site still had an insignificant effect (p = 0.21) and date a significant effect
(p < 0.01) on the OC mass loadings. Pooling all the field blanks data (but excluding the May
2010 samples) yields mean and median OC mass loadings of 6.6 ± 5.0 and 5.2 |jg/filter,
respectively (N = 130).
The trip and field blanks ANOVA analyses demonstrated statistically insignificant
dependencies on site. Therefore, the Phoenix and St. Louis data were pooled. Figure D-2
shows the mean and standard deviation of OC mass loadings for each trip and field blanks
collection event. ANOVA on the site pooled data yielded significant effects (p < 0.01) for both
date and blanks type (trip versus field). This result was obtained both including and excluding
the May 2010 samples.
30
i i Trip Blanks
ix-XyXj Field Blanks
25
O)
15
10
5
0
S
V
£
Figure D-2. Mean and standard deviation of OC trip and field blanks data stratified by
date and combined across the Phoenix and St. Louis sites.
D-2
-------
EPA's Coarse PM Pilot Study
Appendix D
In summary, one year of data is too short a time series to conclusively interpret the trip
and field blank trends. ANOVA analysis suggests that date has an insignificant effect on the
values. However, the high trip blank values observed in the first month of the study (May 2010)
may be the driver for the high field blank values observed during that collection event. For
subsequent months, it is not clear whether the field blanks exhibit an intrinsic seasonal behavior
or are significantly influenced by the trip blanks. Large differences between trip and field blank
values in July 2010 suggest some decoupling, but the higher field blank values towards the end
of the study coincided with higher trip blank values.
The possibility of temporal behavior in blank loadings confounds using the 95th percentile
OC mass loading as a robust estimate of the LQL, but it can be used as a conservative
estimate. The 95th percentile trip blank and field blanks mass loadings are each 19 |a,g/filter for
all samples and 15 |jg/filter excluding the May 2010 samples.
D-3
-------
-------
Appendix E; Nit1 r.to CckAjtions with Otho1'
Nitrate (NO3) can be difficult to accurately measure, since much of it can be volatilized
off of the Teflon filter. Lee et al., (2008) found that it was not uncommon at rural locations to find
a significant fraction of particulate NO3 present in forms other than ammonium nitrate. In order
to explore whether nitrate was associated with species other than ammonium, correlations
between PMf and PMc nitrate and the other measured species were evaluated.
Tables E-1 and E-2 display the correlation between NO3 (Table E-1 for Teflon filter NO3
and Table E-2 for Teflon/nylon filter NO3) and a subset of species that are primarily above
minimum detection limit (MDL) for the primary dichotomous sampler (Dichot A). No consistently
high correlations were observed between NO3 and the other parameters at either St. Louis or
Phoenix, except for ammonium in the PMf fraction. Moderate correlations (around 0.6) were
found for some elements in St. Louis for the PMc fraction. Visual inspection of scatter plot
matrices showed that there were no outliers biasing the correlation coefficients low, and
confirmed the correlation coefficients reported in Tables E-1 and E-2.
-------
Table E-1. Correlation between Teflon filter NO3 and a subset of species that are
primarily above MDL for the primary dichotomous sampler (A) in PHX and STL.
Correlations are colored from high (red) to low (green).
Correlation
with Species
PHX N03 PMc
PHX N03 PMf
STL N03 PMc
STL N03 PMf
S04
0.50
-0.15
0.56
0.13
K IC
0.25
0.27
0.32
0.21
Na_IC
0.29
-0.12
0.34
-0.07
nh4
0.33
0.84
0.22
0.80
Al
0.50
-0.15
0.59
-0.24
Ba
0.40
0.37
0.46
0.15
Br
0.22
0.35
0.11
0.31
Ca
0.44
-0.11
0.56
-0.12
CI
0.15
0.21
0.13
0.49
Co
0.18
-0.09
0.52
-0.02
Cr
0.38
0.05
0.15
0.10
Cu
-0.11
0.54
0.25
0.11
Fe
0.33
0.02
0.54
0.05
K
0.47
0.23
0.60
-0.19
Mg
0.53
-0.10
0.65
-0.36
Mn
0.41
0.10
0.47
0.27
Na
0.29
-0.12
0.38
-0.13
Ni
0.21
-0.01
0.15
-0.26
P
0.42
-0.07
0.40
-0.11
Pb
-0.02
0.12
0.12
-0.04
Rb
0.31
-0.02
0.15
-0.18
S
0.54
-0.13
0.58
0.12
Si
0.48
-0.15
0.66
-0.33
Sr
0.24
0.47
0.56
0.08
Ti
0.37
-0.06
0.60
-0.23
Zn
-0.14
0.02
0.08
-0.05
Gravimetric
Mass
0.36
0.69
0.64
0.60
-------
Table E-2. Correlation between Teflon/nylon filter NO3 and a subset of species that are
primarily above MDL for the primary dichotomous sampler (A) in PHX and STL.
Correlations are colored from high (red) to low (green).
Correlation
with Species
PHX N03 PMc
PHX N03 PMf
STL N03 PMc
STL N03 PMf
S04
0.37
-0.24
0.55
0.15
K IC
0.14
0.22
0.31
0.25
Na_IC
0.10
-0.26
0.30
-0.07
nh4
0.25
0.91
0.22
0.82
Al
0.37
-0.18
0.60
-0.25
Ba
0.36
0.35
0.49
0.17
Br
0.13
0.34
0.13
0.41
Ca
0.36
-0.14
0.57
-0.10
CI
-0.01
0.07
0.09
0.51
Co
0.10
-0.07
0.53
0.01
Cr
0.32
0.10
0.17
0.10
Cu
-0.12
0.47
0.28
0.13
Fe
0.26
0.05
0.56
0.04
K
0.35
0.18
0.60
-0.14
Mg
0.40
-0.21
0.65
-0.35
Mn
0.34
0.18
0.48
0.26
Na
0.14
-0.26
0.34
-0.09
Ni
0.15
-0.03
0.16
-0.26
P
0.31
-0.06
0.41
-0.12
Pb
-0.06
0.10
0.11
0.03
Rb
0.26
-0.06
0.14
-0.16
S
0.43
-0.22
0.63
0.12
Si
0.36
-0.15
0.67
-0.34
Sr
0.16
0.37
0.57
0.08
Ti
0.30
-0.07
0.61
-0.26
Zn
-0.14
-0.01
0.08
0.03
Gravimetric
Mass
0.29
0.63
0.64
0.62
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-15-001
Environmental Protection Air Quality Assessment Division February 2015
Agency Research Triangle Park, NC
------- |