UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
November 30, 2005
EPA-CASAC-06-001
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Clean Air Scientific Advisory Committee (CAS AC) Peer Review of the
Federal Reference Method (FRM) for Coarse Particulate Matter (PMio-2.s)
Dear Administrator Johnson:
The Clean Air Scientific Advisory Committee's (CASAC or "Committee") Ambient Air
Monitoring and Methods (AAMM) Subcommittee ("Subcommittee") met in a public meeting on
September 21-22, 2005 in Durham, NC to conduct a peer review of EPA's proposed Federal
Reference Method (FRM) for coarse particulate matter (PMio-2.s) and a consultation on various
particulate matter (PM) monitoring-related issues.
The CAS AC hereby forwards this letter to you as the Committee's consensus report on
this subject. The current roster of the seven-member Clean Air Scientific Advisory Committee
— three members of which are also members of the Subcommittee — is attached as Appendix A
to this report, and the CASAC AAMM Subcommittee roster is contained in Appendix B. EPA's
charge to the Subcommittee is found in Appendix C of this report, and Subcommittee members'
individual review comments are provided in Appendix D.
A national monitoring program for PMio-2.5 needs to address multiple, disparate
objectives, including: timely reporting of the air quality index (AQI) and associated public health
advisories, determining compliance with (daily) standards, providing support for future studies
of coarse particle sources, chemical and biological composition and associated effects on human
health and welfare. No single sampling method can meet all of these objectives, but a critical
function of the FRM will be to provide a precise, repeatable definition of coarse PM which can
be used to evaluate the performance of and assure the quality of various Federal Equivalent
Method (FEM) samplers to be deployed in a national monitoring network.
The Committee was extremely impressed by the continuing high quality of technical
work evident in the PMio-2.5 methods evaluation field studies. In general, the CASAC agrees that
-------
there are several important scientific or operational strengths of the proposed difference method
PMio-2.5 to be used as the FRM, while noting that there are several prominent weaknesses as
well. Despite these weaknesses, no other better, currently available candidate FRM method has
been identified. A majority of the Subcommittee members expressed the opinion that the
demonstrated data quality of the PMio-2.5 difference method and its documented value in
correlations with health effects data support its being proposed as the PM Coarse FRM.
However, it is recommended that, in addition to the proposed PMio-2.5 difference method, an
FRM that actually provides a coarse particle sample should be proposed as a second FRM. The
only such sampler currently available is the dichotomous sampler. In both cases, this should be
done with the clear understanding that these manual filter-based samplers are not intended for
extensive field deployment as the basic component of the compliance network and would be
employed primarily as a benchmark for evaluating performance of continuous or dichotomous
FEM instruments. The dichotomous sampler would have the additional benefit of providing
coarse particle samples for chemical speciation. There is clearly a need for the Agency to
develop more direct coarse-particle-only sampling methods and an associated need to devote
more resources to support the necessary research and development in this important area.
1. Background
The CASAC, which comprises seven members appointed by the EPA Administrator, was
established under section 109(d)(2) of the Clean Air Act (CAA or Act) (42 U.S.C. 7409) as an
independent scientific advisory committee, in part to provide advice, information and
recommendations on the scientific and technical aspects of issues related to air quality criteria
and national ambient air quality standards (NAAQS) under sections 108 and 109 of the Act. The
CASAC, which is administratively located under EPA's Science Advisory Board (SAB) Staff
Office, is a Federal advisory committee chartered under the Federal Advisory Committee Act
(FACA), as amended, 5 U.S.C., App. The SAB Staff Office established the CASAC AAMM
Subcommittee in early 2004 as a standing subcommittee to provide the EPA Administrator,
through the CASAC, with advice and recommendations, as necessary, on topical areas related to
ambient air monitoring, methods and networks. The CASAC and the Subcommittee comply
with the provisions of FACA and all appropriate SAB Staff Office procedural policies.
Under section 108 of the CAA, the Agency is required to establish NAAQS for each
pollutant for which EPA has issued criteria, including particulate matter (PM). Section 109(d)(l)
of the CAA requires that EPA carry out a periodic review and revision, where appropriate, of the
air quality criteria, to reflect advances in scientific knowledge on the effects of the pollutant on
public health and welfare, and the NAAQS for "criteria" air pollutants such as PM. EPA is
currently reviewing the NAAQS for PM. As part of this review, the Agency is considering
potential NAAQS for coarse particulate matter (PMio-2.s).
In conjunction with the review of the NAAQS for PM, EPA is evaluating potential
monitoring methods for measurement of PMio-2.5. The Agency's Office of Air Quality Planning
and Standards (OAQPS), within EPA's Office of Air and Radiation (OAR), requested that the
CASAC conduct a peer review of the proposed Federal Reference Method (FRM) for PMio-2.5, to
provide independent scientific advice on the appropriateness of this method as a basis of
comparison in approving Federal Equivalent Method (FEM) coarse-particle monitors, which
provide better temporal (continuous) information, or provide coarse-only filter samples
-------
(dichotomous) more amenable to chemical analyses. The FRM for PMio-2.5 will establish the
basis for approval of FEM monitoring methods in a performance-based measurement system
process.
In addition, OAQPS asked the CASAC to conduct a consultation with the Agency on:
fine particle (PM2 5) FRM optimization and equivalency criteria for continuous monitors; and
PMio-2.5 methods evaluation, network data quality objectives (DQOs), and equivalency criteria
for continuous monitors. (This consultation portion of the Subcommittee's September 21-22
meeting is not covered by this report, although it is addressed by individual reviewers in
Appendix D.) The CASAC AAMM Subcommittee previously provided advice and
recommendations for this ongoing work at a July 22, 2004 consultative meeting on PMio-2.5
methods and DQOs. Prior to this meeting, OAQPS posted all relevant written review materials
on the "CASAC File Area" page of the Agency's Ambient Monitoring Technology Information
Center (AMTIC) Web site at URL: http://www.epa.gov/ttn/amtic/casacinf.html.
2. CASAC Peer Review of the FRM for PM10.2.5
A national monitoring program for PMio-2.5 needs to address multiple, disparate
objectives, including: timely reporting of the air quality index (AQI) and associated public health
advisories, determining compliance with (daily) standards, providing support for future studies
of coarse particle sources, chemical and biological composition and associated effects on human
health and welfare. No single sampling method can meet all of these objectives, but a critical
function of the FRM will be to provide a precise, repeatable definition of coarse PM which can
be used to evaluate the performance of and assure the quality of various Federal Equivalent
Method (FEM) samplers to be deployed in a national monitoring network.
With respect to the first charge question (Appendix C), CASAC AAMM Subcommittee
members generally agree that there are several important scientific or operational strengths of the
proposed difference method PMio-2.5 FRM. These include:
• Direct gravimetric measurement of mass by proven and available technology.
• Use of existing FRM equipment minimizes equipment and training costs.
• Measurements can be highly precise, even when mass concentrations are low.
• Low face velocities may reduce evaporative losses of some semi-volatile species.
• The particle size separation properties (sampling effectiveness curves) of the samplers
are better characterized and documented than other candidate methods.
• The use of reference method filters, filter handling procedures and inlets for PM2.s and
PMio make the method "accurate" by definition (although this doesn't necessarily
provide an accurate depiction of coarse particles in the ambient air).
• No need for air-conditioned shelters.
• Consistency with historical database of mass measurements avoids the need for
expensive field comparisons (although historical PMio measurements are primarily by
high-volume methods).
• Presence of PM2.s particles causes coarse particles to adhere to filter (avoiding mass
losses that may affect dichotomous coarse-only filter samples).
-------
• Filter-based samples may allow for chemical speciation (although the validity of
speciation measurements by difference methods requires further evaluation).
The Subcommittee also noted several weaknesses of the proposed method, including:
• Accuracy of the proposed filter difference method is unknown and difficult to establish
under relevant field conditions (however this is also true for the PM2.5 and PMio FRM).
• Suitability for speciation analysis (by subtraction) has not been established yet, especially
for species not predominantly in the coarse mode (dichotomous or impactor samples may
be more suitable, if sufficient sample material can be collected).
• Possible sampling artifacts from losses of volatile material (i.e., nitrate, organic
compounds) during sampling lead to inaccuracies that cannot be quantified with this
method.
• Likelihood of different sampling artifacts for the PM2.5 and PMio filters, because
reactions between fine and coarse PM species may reduce the volatility of nitrates and
other compounds, and/or because evaporative losses on PM2.5 may exceed evaporative
losses on PMio due to the pressure drop provided by the WINS (Well Impactor - Ninety-
Six) or cyclone. Positive artifacts — such as from reactions between acidic gases and
coarse alkaline crustal material might also occur more frequently on PMio filters.
• Poor time resolution: only suitable for determining compliance with 24-hour standards;
completely unsuitable for use in Air Quality Index (AQI) reports/forecasting or
investigating associations between short-term (e.g., hourly average or maximum)
concentrations and health endpoints.
• Expensive, labor-intensive, manual sample collection and laboratory analysis, requiring
great care in all aspects of method operation.
If the Agency adopts the proposed PMi0-2.5 difference method as the exclusive FRM, this
should be done with the clear understanding that it is not intended for extensive field deployment
and would be employed primarily as a benchmark for evaluating performance of other
continuous or dichotomous FEM instruments. Continued development and evaluation of
methods based on virtual impaction to collect samples of coarse-particles-only should be given a
high priority. The Agency should also consider the possibility of specifying more than one FRM
for PMio-2.5 (as it did for PMio), if one or more of the current or evolving dichotomous sampler
designs shows reasonable agreement with the difference method (assuming filter-handling
procedures can be developed to minimize losses of coarse-only particles prior to weighing).
Work should continue on development of accurate techniques for measurements of coarse
particle mass concentrations and on methods to directly quantify the accuracy. In addition,
Subcommittee members expressed their desire to see automated, time-resolved (hourly), real-
time samplers for lower operating costs, forecasting and AQI support, health studies, etc.
Since the proposed difference method FRM would be used as a basis for approval of
other methods, some of the Subcommittee members were concerned that this may close the door
for the new methods that more accurately measure the ambient PM mass (including semi-volatile
species) than the proposed difference method. Precision, historical continuity, and compatibility
between different PM methods are all desirable data quality objectives, but these objectives
should not take precedence over the "science quality objective" of providing an accurate
-------
characterization of coarse particle concentration and composition in the ambient air. EPA should
explore options to certify alternative Federal Equivalent Methods (or alternative FRM methods)
that can demonstrate superior accuracy to the difference method FRM.
Many Subcommittee members felt that more thought should be given to comparing the
responses of alternative samplers with the FRM using laboratory generated-aerosols of known
composition and size or size distribution. Such work could include calibrated generation and
sampling of known semi-volatile compounds, such as ammonium nitrate and selected organic
compounds. While this methodology might not be applicable to equivalency determinations as
specified by law, it could take us a long way towards an understanding of measurement
accuracy. The laboratory tests would enable unambiguous testing of sampler performance to
particles having known physical and chemical properties. This approach would help to improve
our understanding of measurement accuracy, and would lead to the design of improved samplers
in the future.
Some members questioned EPA's claim that a difference method FRM would provide a
sound basis for chemical analysis (i.e.., coarse chemical composition by subtraction). This is an
important issue in that the "urban" focus of EPA's proposed "UPMio-2.5" indicator is based on
assumed (but not routinely measured) differences in the chemical (and/or biological)
composition of coarse particles in urban vs. rural locations. Clearly, separate PMi0 and PM2.s
samples can be collected and analyzed, but the much higher uncertainty associated with the
chemical analyses of these speciation samples, and the probability of different, chemical or size-
specific sampling and analytical artifacts on PMio and PM2.5 filters are factors that make the
quality of such "speciation by difference" data highly uncertain. Perhaps it will yield acceptable
data, perhaps not; further studies to examine the practicality and validity of the difference
method for speciation are needed to demonstrate its utility. EPA should address some of the
questions about speciation by analyzing the already-collected speciation data from the field
studies. The need for speciation data, however, is inescapable, and the virtual impactor offers
significant advantages for speciation analysis by collecting an aerodynamically-sorted sample in
which PMio-2.5 is greatly enhanced relative to PM2.5. Moreover, the understanding of virtual
impaction has advanced significantly during the 30 years that have passed since the design used
by current instruments was originally developed, and it is likely that a much-improved virtual
impactor design could be developed if support for such research were made available.
Regarding the second charge question (Appendix C), the Subcommittee was extremely
impressed by the continuing high quality of technical work evident in the field methods
evaluation studies, in particular, by EPA's National Exposure Research Laboratory (NERL),
within the Office of Research and Development (ORD); equipment vendors; and the Jefferson
County [AL] Department of Health. It is essential that such work be continued and that the
resources necessary to sustain it are maintained or increased. A majority of the Subcommittee
members expressed the opinion that the demonstrated data quality of the PMio-2.5 difference
method in the EPA field studies performed to date supports it being proposed as a PM coarse
FRM. However, due to many weaknesses of the difference method, EPA should emphasize
deployment of continuous or dichotomous FEMs in the network and use the difference method
FRMs primarily to evaluate the performance of these alternative methods. Such performance
evaluations need to be conducted over a range of locations and seasons, but there should not be
-------
requirements for a high proportion of difference method FRM samplers in State, local and Tribal
(SLT) monitoring networks. It is also unclear if a difference method FRM will be practical for
use as a routine field audit device. A dichotomous sampler might be more suitable for this
purpose, in the event that more than one FRM is established.
It has also been proposed that since the number of PMi0-2.5 [urban] non-attainment areas
is expected to be much smaller than for PM2.5 — and many of these areas have existing PMio
compliance and Toxics monitoring programs — EPA could limit deployment of the cumbersome
difference method for PMio-2.s if it allowed PMio monitors to be used to demonstrate attainment.
U.S. EPA and SLT agencies have already invested large resources into the current regulatory
PMio, Toxics PMio and PM2.5 monitoring networks. Several states (e.g., California) have state
ambient air quality standards for PMio and do not plan to follow U.S. EPA in adopting a coarse
particle standard. It stands to reason that, if a site meets the PMio-2.s standard with PMio
monitoring data (uncorrected for the inclusion of PM2.s), then there is no need to deploy a PMi0.
2.5-specific monitor at the site for compliance determination. In urban areas where PM-coarse
concentrations are expected to be close to or above standards, continuous PMi0-2.5 sampling may
be needed at middle to neighborhood scales for purposes of determining compliance. For
purposes of supporting future health effects studies, sampling is also needed at sites which
represent neighborhood to urban scale exposures. Since PMio-2.s is often emitted at ground level
with a relatively short atmospheric lifetime, the height of the sampler inlet will be a critically
important consideration.
Although the quality of data obtained by the Jefferson County Department of Health is
very impressive, some members questioned if the same data quality could be obtained under the
resource-constrained routine compliance monitoring network conditions, typically found at many
State and local monitoring agencies. It can be noted for example that the precision of the PM2.s
measurements in the EPA and Jefferson County field studies is substantially tighter than that
which has been observed for PM2.s nationwide. These field studies indicate that precise PMio-2.s
data can be obtained by careful, expert personnel, but not necessarily that such precise data will
be obtained in more routine field operations. It was also noted that problems with the Tapered
Element Oscillating Microbalance (TEOM™), dichotomous sampler, Aerodynamic Particle
Sizer (APS™), and other units were only discovered during the ORD inter-comparison study
because multiple units were carefully collocated and operated by U.S. EPA and monitoring
industry experts. If the units were operating by themselves in an SLT agency monitoring
network, it is unlikely that the multiple instrument problems observed in the EPA study would
have been detected in a timely manner.
Without the ability to challenge a PM analyzer with a known concentration of PM, all we
have to verify proper operation of an analyzer is the "due diligence" of the site technician and
highly skilled data review, including level 2 data validation, both which occur months after data
collection - maximizing the potential for data loss. Once again, this reflects on a need for
resources — preferably by an EPA "Science to Achieve Results" (STAR) grant — to develop
traceable standards for PM. This would obviously be a very challenging undertaking, but
without such standards, PM in any size range will always be a pollutant defined by how it is
sampled, rather than by what is in the ambient air and what may potentially deposit onto the
human respiratory tract. Research priorities should also be placed on development of an
-------
improved virtual impactor (for both PM fine and coarse), and for more clearly identifying (and
eliminating) PM coarse filter handling or shipping losses.
In summary, although the proposed difference method has many flaws, no other better,
currently available candidate FRM method has been identified. A majority of the Subcommittee
members expressed the opinion that the demonstrated data quality of the PMi0-2.5 difference
method and its documented value in correlations with health effects data support its being
proposed as the PM Coarse FRM. However, it is recommended that, in addition to the proposed
PMio-2.5 difference method, an FRM that actually provides a coarse particle sample should be
proposed as a second FRM. The only such sampler currently available is the dichotomous
sampler. In both cases, this should be done with the clear understanding that these manual filter-
based samplers are not intended for extensive field deployment as the basic component of the
compliance network and would be employed primarily as a benchmark for evaluating
performance of continuous or dichotomous FEM instruments. The dichotomous sampler would
have the additional benefit of providing coarse particle samples for chemical speciation. There is
clearly a need for the Agency to develop more direct coarse-particle-only sampling methods and
an associated need to devote more resources to support the necessary research and development
in this important area. The Clean Air Scientific Advisory Committee and the CASAC AAMM
Subcommittee found it to be very valuable to advise the Agency in this extremely important task,
and we recommend that the Subcommittee continue to serve in this role. As always, we wish
EPA well and stand ready to offer additional advice as the Agency continues this process.
Sincerely,
/Signed/
Dr. Rogene Henderson, Chair
Clean Air Scientific Advisory Committee
Appendix A - Roster of the Clean Air Scientific Advisory Committee
Appendix B - Roster of the CASAC AAMM Subcommittee
Appendix C - Charge to the CASAC AAMM Subcommittee
Appendix D - Review Comments from Individual CASAC AAMM Subcommittee Members
-------
Appendix A - Roster of the Clean Air Scientific Advisory Committee
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CHAIR
Dr. Rogene Henderson, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
MEMBERS
Dr. Ellis Cowling, University Distinguished Professor-at-Large, North Carolina State
University, Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina
State University, Raleigh, NC
Dr. James D. Crapo, Professor, Department of Medicine, Biomedical Research and Patient
Care, National Jewish Medical and Research Center, Denver, CO
Dr. Frederick J. Miller, Consultant, Cary, NC
Mr. Richard L. Poirot, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
Dr. Frank Speizer, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA
Dr. Barbara Zielinska, Research Professor, Division of Atmospheric Science, Desert Research
Institute, Reno, NV
SCIENCE ADVISORY BOARD STAFF
Mr. Fred Butterfield, CASAC Designated Federal Officer, 1200 Pennsylvania Avenue, N.W.,
Washington, DC, 20460, Phone: 202-343-9994, Fax: 202-233-0643 (butterfield.fred@epa.gov)
(Physical/Courier/FedEx Address: Fred A. Butterfield, III, EPA Science Advisory Board Staff
Office (Mail Code 1400F), Woodies Building, 1025 F Street, N.W., Room 3604, Washington,
DC 20004, Telephone: 202-343-9994)
A-l
-------
Appendix B - Roster of the CASAC AAMM Subcommittee
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CASAC Ambient Air Monitoring & Methods (AAMM) Subcommittee
CHAIRS
Mr. Richard L. Poirot* (Chair - Monitoring), Environmental Analyst, Air Pollution Control
Division, Department of Environmental Conservation, Vermont Agency of Natural Resources,
Waterbury, VT
Dr. Barbara Zielinska* (Chair - Methods), Research Professor, Division of Atmospheric
Science, Desert Research Institute, Reno, NV
SUBCOMMITTEE MEMBERS
Mr. George Allen, Senior Scientist, Northeast States for Coordinated Air Use Management
(NESCAUM), Boston, MA
Dr. Judith Chow, Research Professor, Desert Research Institute, Air Resources Laboratory,
University of Nevada, Reno, NV
Dr. Ellis Cowling*, University Distinguished Professor-at-Large, North Carolina State
University, Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina
State University, Raleigh, NC
Mr. Bart Croes, Chief, Research Division, California Air Resources Board, Sacramento, CA
Dr. Kenneth Demerjian, Professor and Director, Atmospheric Sciences Research Center, State
University of New York, Albany, NY
Dr. Delbert Eatough, Professor of Chemistry, Chemistry and Biochemistry Department,
Brigham Young University, Provo, UT
Mr. Eric Edgerton, President, Atmospheric Research & Analysis, Inc., Gary, NC
Mr. Henry (Dirk) Felton, Research Scientist, Division of Air Resources, Bureau of Air Quality
Surveillance, New York State Department of Environmental Conservation, Albany, NY
Dr. Philip Hopke, Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Potsdam, NY
B-l
-------
Dr. Rudolf Husar, Professor, Mechanical Engineering, Engineering and Applied Science,
Washington University, St. Louis, MO
Dr. Kazuhiko Ito, Assistant Professor, Environmental Medicine, School of Medicine, New
York University, Tuxedo, NY
Dr. Donna Kenski, Data Analyst, Lake Michigan Air Directors Consortium, Des Plaines, IL
Dr. Thomas Lumley, Associate Professor, Biostatistics, School of Public Health and
Community Medicine, University of Washington, Seattle, WA
Dr. Peter McMurry, Professor and Head, Department of Mechanical Engineering, Institute of
Technology, University of Minnesota, Minneapolis, MN
Dr. Kimberly Prather, Professor, Department of Chemistry and Biochemistry, University of
California, San Diego, La Jolla, CA
Dr. Armistead (Ted) Russell, Georgia Power Distinguished Professor of Environmental
Engineering, Environmental Engineering Group, School of Civil and Environmental
Engineering, Georgia Institute of Technology, Atlanta, GA
Dr. Jay Turner, Visiting Professor, Crocker Nuclear Laboratory, University of California -
Davis, Davis, CA
Dr. Warren H. White, Research Professor, Crocker Nuclear Laboratory, University of
California - Davis, Davis, CA
Dr. Yousheng Zeng, Air Quality Services Director, Providence Engineering & Environmental
Group LLC, Baton Rouge, LA
SCIENCE ADVISORY BOARD STAFF
Mr. Fred Butterfield, CASAC Designated Federal Officer, 1200 Pennsylvania Avenue, N.W.,
Washington, DC, 20460, Phone: (202) 343-9994, Fax: (202) 233-0643 (butterfield.fred@epa.gov)
[Physical/Courier/FedEx Address: Fred A. Butterfield, III, EPA Science Advisory Board Staff
Office (Mail Code 1400F), Woodies Building, 1025 F Street, N.W., Room 3604, Washington,
DC 20004, Telephone: (202) 343-9994]
* Members of the statutory Clean Air Scientific Advisory Committee (CASAC) appointed by the EPA
Administrator
B-2
-------
Appendix C - Charge to the CASAC AAMM Subcommittee
Peer Review Questions:
Questions associated with Attachment 1 - Selection and technical summary of PM 10-2.5 FRM:
1. What are the scientific and operational strengths and weaknesses of the PMio-2.5
difference method relative to other options for a proposed FRM, especially when used as
the basis for approval of other methods?
2. Based on the field study report as well as any other available data, e.g., data from State
and local agencies, how does the demonstrated data quality of the PMio-2.5 difference
method support or detract from it being proposed as a FRM?
Consultation Questions:
Question associated with Attachment 2 - EPA 's Multi-Site Evaluations of Candidate
Methodologies for Determining Coarse Paniculate Matter (PM 10-2.5) Concentrations: August
2005 Updated Report Regarding Second-Generation and New PM 10-2.5 Samplers:
1. Based upon the latest available field study data, which PMi0-2.5 methods have both
sufficient utility to meet one or more important monitoring objectives and appropriate
data quality to be considered for deployment as Federal Equivalent Methods (FEMs) or
speciation samplers in a potential PMio-2.5 monitoring network?
Questions associated with Attachment 3 -Memo to PM NAAQS Review Docket (OAR-2001-
0017) - Potential changes being evaluated for the PM2.5 Federal Reference Method
2. What are the Subcommittee's views on the Very Sharp Cut Cyclone (VSCC) being
approved as an alternative second-stage impactor to the Well Impactor Ninety-Six
(WINS) for use on a PM2.5 FRM?
3. To what extent are the stated advantages of relaxing existing requirements identified for
the PM2.s FRM supported by the information cited in Attachment 3, available literature,
or good field and laboratory practices? Does the Subcommittee have additional
recommendations for the PM2.s FRM that would be neutral with respect to bias, but
would improve the performance and minimize the burden on agencies conducting the
sampling?
Questions associated with Attachment 4 - Criteria for Designation of Equivalence Methods for
Continuous Surveillance of PM2.5 Ambient Air Quality
4. Considering the statistical measures of precision, correlation, multiplicative bias, and
additive bias identified for approval of PM2.5 continuous methods, what are the
C-l
-------
Subcommittee's views on the usefulness of each measure to ensure that approved or
equivalent methods meet the monitoring network data quality objectives?
5. What are the advantages and disadvantages of using sampler precision and sample
population to help determine the minimum correlation requirement for the approval of
PM2 5 continuous methods?
6. What are the Subcommittee's views on using a PM2.5 continuous monitor approved as a
FEM, being applicable for use as part of a potential PM2.5 secondary standard for
visibility?
Question associated with Attachment 5 - Sensitivity of the PMio-2.sData Quality Objectives to
Spatially Related Uncertainties
1. To what extent have the assessments of spatial variability and the sensitivity of the DQO
process to a variety of population distributions been appropriately addressed?
Question associated with A ttachment 6 — PM] 0-2.5 Method Equivalency Development
8. What are the Subcommittee's views on the approach identified for the development of
criteria to approve continuous PMio-2.5 equivalent methods?
C-2
-------
Appendix D - Review Comments from
Individual CASAC AAMM Subcommittee Members
This appendix contains the preliminary and/or final written review comments of
the individual members of the Clean Air Scientific Advisory Committee (CASAC)
Ambient Air Monitoring & Methods (AAMM) Subcommittee who submitted such
comments electronically. The comments are included here to provide both a full
perspective and a range of individual views expressed by Subcommittee members during
the review process. These comments do not represent the views of the CASAC AAMM
Subcommittee, the CASAC, the EPA Science Advisory Board, or the EPA itself. The
views of the CASAC AAMM Subcommittee and the CASAC as a whole are contained in
the text of the report to which this appendix is attached. Subcommittee members
providing review comments are listed on the next page, and their individual comments
follow.
D-l
-------
Panelist Page#
Mr. George Allen D-3
Dr. Judith Chow D-10
Dr. Ellis Cowling D-20
Mr. Bart Croes D-25
Dr. Kenneth Demerjian D-34
Dr. Delbert Eatough D-38
Mr. Dirk Felton D-51
Dr. Philip Hopke D-57
Dr. Rudolf Husar D-60
Dr. Kazuhiko Ito D-62
Dr. Donna Kenski D-68
Dr. Thomas Lumley D-70
Dr. Peter McMurry D-74
Dr. Kimberly Prather D-78
Dr. Armistead (Ted) Russell D-83
Dr. Jay Turner D-85
Dr. Warren H. White D-89
Dr. Yousheng Zeng D-96
D-2
-------
Mr. George Allen
To: Fred Butterfield, Designated Federal Officer
EPA SAB, Clean Air Scientific Advisory Committee (CASAC)
Ambient Air Monitoring and Methods Subcommittee
From: George Allen, AAMM subcommittee member, September 30, 2005
The following are revised written comments based on discussions during the September 21-22,
2005 meeting on a peer review of the PM-coarse FRM. A copy of these comments is being sent
to Dr. Barbara Zielinska and Mr. Rich Poirot, CASAC AAMM Subcommittee Co-Chairs. These
comments address the Charge Questions in the EPA OAQPS memo to the SAB dated August 19,
2005.
Questions associated with Attachment 1 - Selection and technical summary of PM10-2.5 FRM:
1. What are the scientific and operational strengths and weaknesses of the PM10-2.5 difference
method relative to other options for a proposed FRM, especially when used as the basis for
approval of other methods?
2. Based on the field study report as well as any other available data, e.g., data from State and
local agencies, how does the demonstrated data quality of the PM10-2.5 difference method
support or detract from it being proposed as an FRM?
Attachment 1 is a clear, concise, and thorough summary of the issues involved with use of the
difference method for PM-coarse (PM10-PM2.5 or PMio-2.5 as used in this document). This
summary is well written and technically complete. My comments on these two questions based
on this document follow.
Charge Question #1.
I agree with EPA that the difference method for PM-coarse is clearly the most defendable
approach for a reference measurement method. When exactly matched pairs of samplers and
protocols are run by highly skilled staff in a well-controlled research environment, the difference
method is the most definitive PM-coarse measurement with the fewest data quality ambiguities.
As noted in the writeup, identical hardware for both systems must be used (with the exception of
the lack of a PM2.5 impactor or cyclone in the "PMlOc" sampler. Specifically, this excludes use
of a PM10 Hi-Vol sampler for this purpose, even when it is designated as an FRM.
The EPA summary recommends that any sampler pair that has been designated as a PM2.5 FRM
can be used as a PM-coarse FRM. This reviewer suggests that for PM-coarse, the acceptable
average field blank value of 30 jig for the PM2.5 FRM is too high and could lead to degraded
precision of PM-coarse data where PM2.5 is substantially greater than PM-coarse (much of the
eastern U.S.). There is a wide range in both mean and variation of field blanks across different
PM2.5 FRMs (here, a field blank is defined as a filter loaded into the sampler and left in place
for the same duration as a normal sample - typically 48 hours or more). Some methods routinely
achieve mean field blanks in the range of 5 to 10 jig; it is recommended that this specification be
D-3
-------
reduced to a maximum of 10 jig field blank mean for sampler pairs used for PM-coarse
measurements.
I do not agree with the proposal to also accept (as class I equivalent methods) pairs of dissimilar
models of FRMs that have been designated as PM2.5 FRMs without further testing. There are
potential subtle differences between models that may degrade the PM-coarse measurement. This
is especially true of sequential samplers. Issues here include the potential for different field blank
values among different sampler models, and in the case of a sequential and manual FRM pair,
different post-sample times (and thus different semi-volatile PM losses) before field collection.
Another potential variable is "effective filter face velocity", determined not only by flow and
"apparent" exposed filter surface area, but also the "effective" exposed area, determined by the
actual open area (holes) in the filter support screen. This latter parameter can vary between
different models of PM2.5 FRMs unless the same support screens are used; the resulting
difference in "effective filter face velocity" could affect loss of semivolatile PM species. In
summary, all of these parameters (physical and temporal) must be kept identical between the two
samplers in a PM-coarse pair to avoid potential biases.
This summary suggests on page 5 that the difference method for PM-coarse can be rapidly
deployed into existing compliance monitoring networks with minimal training or pilot operation
periods. Although the difference method for PM-coarse has performed very well in the EPA tests
(a highly controlled research-grade environment), that performance is likely to degrade
substantially in the real-world of resource constrained SLT monitoring programs. This reviewer
does not recommend wide deployment of this PM-coarse method without substantial further field
tests in the SLT environment across a range of agencies.
Related to this issue of SLT deployment of the difference method, as I understand it the officially
designated FRM method for PM-coarse must be used by SLTs for audit purposes. This poses a
potential problem; many agencies may have difficulty generating sufficiently precise pm-coarse
data with the difference method to meaningfully evaluate the performance of the routine
(nondifference) method, rendering the audit process useless at best and misleading at worst.
Page 5 also suggests that the difference method can provide speciated analyses of coarse mode
aerosols. This is only true in a practical sense for species that are present primarily only in the
coarse mode (typically crustal elements); for PM species that are dominant in the fine mode, the
data from this approach will usually have severely degraded coarse mode precision. The
dichotomous sampler method is much better suited for this kind of analysis. Ideally a medium
flow dichot design would be used for coarse mass speciation; sample flows would be at least 6
times those in the "classic" dichot sampler, or at least 100 LPM inlet flow. An example of such
an approach would be the custom dichot system used in the St. Louis supersite, with an inlet
flow of 113 LPM and a coarse channel flow of 11.3 LPM. These flows are almost seven times
greater than the standard dichot sampler, with corresponding increases in sample material on the
coarse channel filter.
This summary correctly reflects the concerns related to use of a virtual impactor-based
(dichotomous) sampler. This reviewer agrees with the EPA summary that the dichotomous
sampler is less suitable as a PM-coarse (mass) reference method. The loss of large particles from
D-4
-------
the coarse mode filter during shipment and the potential for excessive amounts of coarse
particles being carried over to the fine mode filter are very real concerns.
With regard to the first dichot issue above, a meaningful dichot coarse filter mass shipping loss
test has still not yet been performed; given that the existing literature shows losses (for PM-15)
in the range of 30 to 50% (Spengler and Thurston, JAPCA December 1983, 33:12; and Dzubay
and Barbour, JAPCA August 1983, 33:7), this is a critical test.
The second issue is one of virtual impactor design. All the systems tested (except one version of
the Kimoto dichot) use essentially the same virtual impactor design - the Loo and Cork design
from the mid 1970's, described in Loo and Cork, Aerosol Science and Technology 9:167-176
(1988). The state of the virtual impactor aerosol science has advanced dramatically over the last
30 years, and it is likely that a much improved virtual impactor design (minimizing the problems
observed with the current design) could be developed if support for such research were made
available.
It is worth noting here that any potential legal constraints against using a difference method for
the PM-coarse FRM would not only prohibit the PM10-PM2.5 method, but may also limit the
use of the dichotomous sampler, thus leaving no practical method for a filter-based (with
gravimetric analysis) PM-coarse FRM. The dichot is also a difference method, although in a
somewhat different way from paired samplers. PM-coarse for the dichot sampler is calculated as
follows:
[a] calculate PM10 using the net mass from the sum of both the fine and coarse channel
filters and the total sampler (inlet) flow;
[b] calculate PM2.5 using the net mass from the fine channel filter and the fine channel
flow;
[c] calculate PM-coarse by subtracting PM2.5 in [b] from PM10 in [a].
The final step [c] is clearly a calculation of PM-coarse by difference. This dichot PM-coarse
calculation description is not the one normally used in instrument manuals or network data
reduction procedures, but is mathematically identical to those calculation procedures and is
fundamentally clearer and simpler to implement. Any future documentation on how to calculate
PM-coarse from dichot samplers should use this simpler approach.
There is potential for a PM-coarse method that avoids the limitations of both the difference and
dichot methods that has not been evaluated. Collection by impaction is in principle a simple and
direct measurement of PM-coarse. An example of such a method would be the existing PM2.5
FRM, with the WINS impactor substrate replaced with a suitable (weigh-able) filter media that
retains the coarse-mode particles (e.g., does not have a significant "coarse particle bounce"
problem). This is worth evaluating, since the approach has several advantages over present
methods:
1. It uses existing technology already deployed in state/local networks.
2. Collection by impaction minimizes chemical artifacts relative to filtration methods.
3. It is the only way to get a direct measurement of PM-coarse.
The largest concern in an impaction-based PM-coarse method is particle bounce (loss) from the
impaction substrate. The test results from an FRM with crystalized Dow 704 impactor oil as the
D-5
-------
WINS surface (Vanderpool et al., Aerosol Science and Technology., 34(5): 465-476, May 2001)
imply that a suitable filter substrate (perhaps Fluoropore) is practical and worth evaluating. The
Vanderpool tests with crystalized Dow oil showed that "...no large particle bounce from the
crystallized oil surface was observed" and "...there appears to be no adverse effect of crystallized
oil on the overall performance of the WINS separator nor on the PM2.5 concentration
measurement." This reviewer strongly encourages that this approach be considered and properly
evaluated as a candidate for routine field deployment.
Charge Question #2.
The demonstrated data quality of the PM10-2.5 difference method in the EPA field studies
performed to date clearly supports it being proposed as a PM-coarse FRM; other than the greater
than typical resources required to generate data of the high quality of this study, there are no
detractions to this approach for PM-coarse measurement. There are limited studies done by state
or local air agencies; the work done in the Birmingham AL area by the Jefferson County Dept. of
Health is the only one available to this reviewer. This study included some sites (the Providence
site for example, with ratios of mean PM2.5 to mean PM10 of 0.7) where the PM2.5 to PM-
coarse ratios were substantially greater than one and mean PM-coarse concentrations were under
10 |ig/ni3. These concentrations are typical of much of the eastern U.S., even in neighborhood
urban sampling locations. This scenario is a tougher test of the PM-coarse difference method
than the sites used in the EPA field test studies, but is essential in understanding the performance
of the method under this common condition.
The Jefferson County tests were clearly performed under carefully controlled operating
conditions; the mean CV for both collocated PM2.5 and PM10 samples was 1.4% for a single
sampler (this needs clarification in the report; the CV reported there is from a pair of samplers,
not a single sampler), an excellent result for the mean PM2.5 and PM10 concentrations sampled
(15 and 24 jig/ms respectively). Despite the excellent PM2.5 and PM10 sampler precision as
indicated by the reported CVs, the single-sampler PM-coarse CV was much higher at 5.7%.
While still acceptable, this degradation in PM-coarse precision (relative to the PM2.5 and PM10
data used to generate the PM-coarse data) at sites with higher PM2.5 than PM-coarse is expected
with the difference method. This effect is explained in detail in Allen et al. (J. Air & Waste
Manage. Assoc., 49:PM, September 1999, pages 133-141), available at:
http://www.awma.org/journal/special/Sept99/allen.pdf
As noted above, there has not been any assessment of the difference method for PM-coarse under
the resource-constrained routine compliance monitoring network environments typically found at
many State and local monitoring agencies. The reported PM-coarse difference method precision
(CV) of 5.7% is likely to be much higher in these network environments, especially where
PM2.5 dominates the PM10 concentrations.
During the panel discussions on other candidates for PM-coarse FRMs, the R&P coarse TEOM
method was mentioned. Assuming proprietary methods are not candidates for designation for
FRMs, that rules out this approach. One panelist commented that patents can always be worked
around. In this case, that is not likely. The claims in US patent # 6,829,919 issued December
D-6
-------
14, 2004 to Costas Sioutas and Paul Solomon for a "High-quality continuous particulate matter
monitor" cover inlet flows over the range of 5 LPM to 100 LPM and a virtual impactor flow ratio
of 2 to 50 (an inlet flow of 50 LPM and a flow ratio of 25 is used in the commercial version of
the method). These claims are sufficiently broad to make it essentially impossible to create a
public domain version of this method that does not infringe on this patent. As such, this method
should not be considered as a candidate FRM.
To: Fred Butterfield, Designated Federal Officer
EPA SAB, Clean Air Scientific Advisory Committee (CASAC)
Ambient Air Monitoring and Methods Subcommittee
From: George Allen, AAMM subcommittee member, September 30, 2005
The following are revised written comments based on discussions during the September 21-22,
2005 meeting on a consultation on the development of criteria for a PM-coarse FEM. A copy of
these comments is being sent to Dr. Barbara Zielinska and Mr. Rich Poirot, CASAC AAMM
Subcommittee Co-Chairs. As requested in the memo from the AAMM co-chairs dated August
25, 2005, these comments address Question 8 in the EPA OAQPS memo to the SAB dated
August 19, 2005. An additional comment on an approach to quantitatively defining "urban" PM-
coarse is included at the end of these comments.
Consultation Question 8, associated with Attachment 6:
"What are the Subcommittee's views on the approach identified for the development of criteria
to approve continuous PMio-2.5 equivalent methods?"
Attachment 6 is titled "Technical Report on PM10-2.5 Method Equivalency Development", by
OAQPS and Battelle, dated August 19, 2005. As noted in the executive summary, the objective
is to develop standards or criteria to compare candidate PM-coarse methods (continuous or
"direct") to the PM-coarse FRM. The goal of this comparison is to insure that the quality of
decisions made with regard to compliance with a PM-coarse NAAQS are as good as if the
measurements were made with the FRM.
The process presented in this report is based on and is parallel to the process used to develop
equivalency requirements for PM2.5 continuous FEMs (presently available in draft form and
expected to be published for comment on December 20, 2005). My comments on this approach
for the use of this technique for PM-coarse FEMs follow.
Section 2.1, Data Assumptions. This section details the collocated data set needed to determine
equivalency. It appropriately requires multiple sites (the number not yet established), but does
not discuss the need for a range of aerosol size distributions or concentrations for those sites.
Only one site would be required to have tests performed during more than a single season; this
may not be sufficient unless seasons are identified in advance and chosen to provide the most
challenging aerosol for the candidate method. An example is Phoenix AZ; if a single season is
D-7
-------
chosen it would have to be summer, not winter. For each "season" test period, samplers will be
run for 30 days, with 75% (23 days) of valid data required (23 days where data from at least 2 of
the three collocated pairs are valid). An acceptable concentration range of collected data will be
set; this may further reduce the available number of days for analysis below 23. This reviewer
recommends a target sample period of 40 collection days, not 30. That is more likely to produce
a valid and useful comparison data set of 30 or more days, resulting in a more robust and
meaningful statistical analysis.
General comments:
Among other metrics, the DQO acceptance method relies on the compensating effect of additive
bias (regression intercept) and multiplicative bias (regression slope), resulting in an example
acceptance range "window" for regression slopes and intercepts for a given FEM candidate as
shown in figure 4. While this approach has some merit for qualifying FEMs for measurements to
determine compliance with the PM2.5 annual mean standard metric, it does not work as well for
a daily standard form such as the expected PM-coarse NAAQS. This is explained in the
following paragraphs.
A 3-year annual mean metric is composed of up to 1095 daily samples with a population that has
a large dynamic range - samples close to zero and samples 4 to 5 times the "bright line" that is
the NAAQS standard. The additive and multiplicative biases are combined in ways that tend to
average out the method's errors with respect to the 3-year annual mean NAAQS metric
(currently 15 jig/ms).
A daily PM-coarse standard similar to that proposed in the final EPA PM staff paper (the 3-year
mean of the 98th or 99thpercentile annual values) is based on 3 samples that are typically
clustered in a range of concentrations near the standard's value; as such, the sample population is
very small and has a very limited dynamic range. In this situation, the only performance metric
that is important in the context of NAAQS compliance is the method's response for samples
slightly above or below the standard — typically in the range of 80 to 90 jig/ms — a relatively
small dynamic range. Unlike an annual standard, how the method performs at levels well below
the standard has no meaning in this daily standard compliance context. The question boils down
to "how accurately does the FEM measure 85 jig/ms".
To demonstrate the limitations of the example in Figure 4, if the PM-coarse standard were 85
(the upper range of the staff paper), a method that gave values ranging from 65 to 107 when the
actual value was 85 would fit within the box and be an acceptable FEM. 107 ^ 65 is 1.65 - a
rather wide range when determination of compliance with the NAAQS is the goal. It is
recognized that the example given in Figure 4 is not necessarily the EPA's recommendation for
PM-coarse FEM performance, but the underlying concerns remain.
Of course, determining compliance with NAAQS is not the only goal of PM-coarse
measurements. To be useful to the health-effects community for future PM-health studies and for
modeling purposes, a PM-coarse method must produce reasonably accurate data over the entire
range of ambient values, and do this ideally on a time-frame much shorter than 24-hours. This is
not the goal of the currently proposed FEM evaluation process, and beyond the scope of this
D-8
-------
charge question. Still, this additional data use objective must be kept in mind when determining
how to characterize what level of performance is adequate for designation of a method as a PM-
coarse FEM. If we end up with a network of PM-coarse FEMs that do not generate data of
sufficient quality at the relatively low (with respect to a likely standard) ambient concentration
levels frequently observed in the eastern U.S., then we limit future progress in better
understanding the health effects of PM-coarse.
There was some discussion during the second day of the meeting regarding how EPA might
define "urban" PM-coarse for regulatory purposes. This reviewer would like to offer an approach
to quantitatively define "urban" PM-coarse that was not discussed at the meeting. One of the
underlying assumptions behind having PM-coarse be an urban-only standard is that urban PM-
coarse particles are more harmful than a simple wind-blown "clean dust" because they have a
mobile-source related chemical composition on the particle surface. A simple and quantitative
assessment of the extent of how "urban" (mobile-source influenced) a coarse mode aerosol is can
be made based on the color of the coarse mode aerosol. An example of this is readily observed
by looking at the coarse-mode particles collected in the PM2.5 FRM WINS impactor. In core
urban areas, those particles are black (very black, not just grey). In rural areas (areas without
substantial local mobile source influence), the color is somewhere between a sandy color and a
greyish-sandy color. There are optical reflectance methods that can quantify how "black" a
particle deposit is. Thus, a quantitative assessment of how "urban" a site is can be made if
sufficient samples are analyzed in this manner.
D-9
-------
Dr. Judith Chow
September 29, 2005
To: Fred Butterfield, Designated Federal Officer
Clean Air Scientific Advisory Committee (CASAC)
Rich Poirot, Co-Chair - Monitoring
Barbara Zielinska, Co-Chair - Methods
CASAC Ambient Air Monitoring and Methods (AAMM) Subcommittee
From: Judith C. Chow, CASAC AAMM Subcommittee Member
Subject: CASAC Review of the Particle Methods and Data Quality Objectives
This memo addresses the questions on which the Subcommittee members were asked to
comment regarding Attachment 1 ("Summary and Rationale for the PMio-2.5 FRM") and
Attachment 3 ["Memo to PM NAAQS Review Docket (OAR-2001-0017): Potential Changes
being Evaluated for the PM2.5 FRM"].
Questions on Attachment 1 (Summary and Rationale for the PMio-2.5 FRM)
Question 1: What are the scientific and operational strengths and weaknesses of the PM 10-2.5
difference method relative to other options for a proposed FRM, especially when
used as the basis for approval of other methods?
The PMio-2.5FRM is selected based on: 1) its ability to provide credible, reliable, and validated
monitoring data for NAAQS attainment, and 2) its practicality in comparison with alternative
methods to determine their qualifications as Federal Equivalent Methods (FEMs).
I agree with the statements in the "Summary and Rationale for the PMio-2.5 FRM" (Attachmentl)
in that: 1) the difference method provides maximum comparability (e.g., filter medium, sample
collection, gravimetric analysis, quality assurance procedures) to new or existing PM data sets,
2) PM2.5 and PMio sampling inlets have been wind tunnel tested to sustain a wide range of wind
speeds and directions, 3) the PM2.5 and PMioc (PMio reference sampler with replacement of the
PM2.5 FRM WINS impactor with straight downtube adaptor)
FRM samplers have been laboratory and field tested and evaluated, and 4) they are commercially
available and network operators are familiar with them.
The statement (second paragraph on page 4) that: "An inherent advantage of a difference method
is that some (additive) biases may be eliminated or substantially reduced by the subtraction" is
not entirely true. There are several aspects of the proposed difference method that need to be
considered:
• Potential variations in the changes of flow rates during the 24-hour sampling
duration. Even though each PM2.5 and PMio sampler is equipped with a volumetric
D-10
-------
flow control, as particle loading increases, the flow rate may alter differently with
PM2.5 and PMio particles. The alteration of flow rates may not affect the cut points by
very much (unless the flow is restricted as happens under certain pollution episodes),
but it may bias the PMio-2.5 measurements.
Filter equilibration conditions need to be reconciled. For PMio FRM weighing, filters
are equilibrated at a set temperature between 15 °C and 30 °C with a variability not
more than ± 3 °C and a set relative humidity (RH) between 20% and 45% with a
variability not more than ± 5% for 24 hours prior to weighing. For PM2.5 FRM
weighing, filters are equilibrated for 24 hours at a set temperature between 20 °C and
23°C with a variability not more than ± 2 °C and a set RH between 30% and 40%
with a variability not more than ± 5% (U.S. EPA, 1998). To minimize volatilization
and water retention, a temperature of < 20 °C and RH < 20% would be preferable,
although the RH may be difficult to control in humid environments. To ensure
comparability, the weighing conditions for PMio-2.5 need to conform to those of the
PM2.5FRM.
Sampling and reporting conditions need to be consistent. Currently, PMio is adjusted
to sea level pressure and a 25 °C temperature, whereas PM2.5 sample volumes are
intended to represent sampling conditions. PMio samplers are intended to operate in
actual conditions, but this isn't always the case because the calibrator conditions
(usually standard) need to be adjusted to the actual conditions (which vary by
season), then reconverted to the sea level conditions. Try to figure out the three
temperature and pressure set-points for the PMio TEOM sampler and see how often it
isn't correct. This doesn't make a big difference in moderate coastal areas, but it is
significant in mountainous areas with seasonal (and even diurnal) temperature
extremes.
Separate timers and pumps need to be synchronized. A single timer, temperature, and
pressure sensor could be used to start, stop, and perform temperature and pressure
corrections for both parallel units. A retrofit for existing samplers might be possible.
The use of two samplers requires increased initial capital investment and perhaps site
modification to accommodate the additional sampler. Also, it requires more of the
operator's time for sampling and maintenance. Newer designs might combine both
channels in a single unit with a common set of controls.
If chemical analysis, such as X-ray fluorescence (XRF), is to be performed on these
samples, it will add additional cost and uncertainties, as large-particle corrections
(Dzubay and Nelson, 1975) only apply to the PMio-2.5 fraction.
There is less volatilized nitrate on PMio samples than on PM2.5 samples, probably
owing to the adherence of nitric acid to alkaline soil particles and sea salt (Wu and
Okada, 1994, Galy-Lacaux, 2001, Underwood et al., 2001). The different amounts of
evaporation will make the coarse mass appear higher than it is in the atmosphere.
D-ll
-------
• At times and places where PM2.5 constitutes most of the PMio mass, negative values
for PMio-2.5 are possible, even though they may be within measurement error. The
uncertainty of the difference [i.e., • pMio-pM25= (* ?Mio2+ * ?M252)/2, Bevington, 1969]
should be estimated and compared with the difference to determine its significance.
To overcome the poor time-resolution of integrated 24-hour samples, EPA is
encouraged to examine the difference between PM25 and PMio BAM and TEOM
mass and their equivalence or comparability with filter-based PMio-2.5, Since many
state and local agencies already own a PM2.5 or PMio B AMs or TEOMs, additional
units can be added to acquire hourly PMio-2.5 mass. For areas or seasons for which
coarse particle volatization is not significant, the difference between the PMio or
PM2.5 BAM or TEOM may be considered as a candidate for the equivalency method.
Question 2: Based on the field study report as well as any other available data e.g., data from
State and local agencies, how does the demonstrated data quality of the PMi 0-2.5
difference method support or detract from it being proposed as an FRM?
Data presented in the report on "Network Operations of the PMio-2.5 Difference Method" by
Vanderpool and Dillard (2004) demonstrate that when procedures are followed good precision
can be obtained with the difference method for the Jefferson County, AL, aerosol. It would be
useful to report the PMio-2.5 mass concentrations with the propagated precisions, as described
above, to confirm that measurements are identical within the measurement uncertainties. These
tests represent seven sites in Jefferson County, AL, using BGIPQ200 samplers. PM in Jefferson
County shows a high sulfate and moderate crustal aerosol. More tests are needed in dry, high
crustal environments (e.g., Phoenix, Las Vegas) and at sites that have large fractions of nitrate
and crustal material (e.g., Rubidoux, Fresno).
Questions on Attachment 3 [Memo to PM NAAQS Review Docket (OAR-2001-0017):
Potential Changes being Evaluated for the PM2.5 FRM]
Consultation Question 2: What are the Subcommittee's views on the Very Sharp Cut Cyclone
(VSCC) being approved as an alternate, second-stage impactor to the Well
Impactor Ninety-Six (WINS) for use on a PM2.sFRM?
Field experiments (e.g., Kenny et al., 2004) show that WINS impactors in PM2.5 FRMs need
frequent cleaning to retain their cut-points. The oil can also freeze at low temperatures, although
alternatives are available (Hunike, 2000, Vanderpool et al., 2004). The oil can also get on the
filters (Pitchford et al., 1997). The WINS was not intended for continuous PM2.5 monitors, and
may be impractical when used with them. Cyclone inlets don't use oil, have a high loading
capacity, and can be easily cleaned (Chan and Lippmann, 1977; Gussman et al., 2002; Kenny
and Gussman, 1997, 2000; John and Reischl, 1980; Kenny et al, 2000, 2004; Peters et al.,
2001a). Sharp-cut cyclones (Kenny et al., 2000) have sampling effectiveness curves only slightly
flatter than the WINS. The very sharp cut cyclone (VSCC, Kenny et al., 2004) has a sharper
effectiveness curve and retains its Dso and effectiveness curve even under conditions with heavy
loadings of 150 jig/ms (Kenny and Thorpe, 2001; Kenny et al., 2004).
D-12
-------
Given the nature of the WINS impactor and the burden of frequent cleaning and oiling in the
field (typically once every five runs) as compared to VSCC (once every 30 runs), it is reasonable
to substitute the VSCC for the WINS on the FRM. More frequent cleaning than once every 30
runs (e.g., every 10 to 15 runs) for the VSCC should be required to ensure data quality.
Replacement of WINS with VSCC in PM2.5 FRM is approved as a Class IIFEM (Federal
Register, 2002). A slight loosening of the inlet effectiveness requirement would allow sharp cut
cyclones to be used as well, probably with a negligible effect on the PM2.5 mass measurement.
Past experience shows that some of the EPA-designated methods may perform poorly if the
sampler is not well-maintained and the inlet is not frequently cleaned (Chow, 1995). While it is
important to maintain consistency and data quality, the EPA should encourage rather than
discourage vendors to apply for the Class III equivalent method designation for in-situ
continuous monitors.
Consultation Question 3: To what extent are the stated advantages of relaxing existing
requirements identified for the PM2.5 FRM supported by the information cited in
Attachment 3, available literature, or good field and laboratory practices? Does
the Subcommittee have additional recommendations for the PM2.5 FRM that would
be neutral with respect to bias, but would improve the performance and minimize
the burden on agencies conducting the sampling?
There are several published comparison studies of PM2.sFRMs among themselves and with other
PM2.5 samplers (Allen et al., 1997: Babich et al., 2000; Bardsley and Dal Sasso, 2005; Chung et
al., 2001; Kenny et al., 2000; Lee et al., 2005a, 2005b; Long et al., 2003; Motallebi et al., 2003;
Peters et al., 2001a, 2001b; Pitchford et al., 1997; Poor et al., 2002; Rizzo et al., 2003; Russell et
al., 2004; Tanner and Parkhurst, 2000; Tropp et al., 1998; Yanosky et al., 2002). As one might
expect, some show better agreement than others. Comparisons are poorer for environments with:
1) low concentrations, 2) a larger fraction of coarse particles, and/or 3) plentiful ammonium
nitrate.
FRM prototype field tests in Birmingham, AL (Pitchford et al., 1997) showed oil drops on the
filter due to the oil used in WINS splashing out of the impactor well. In addition, accumulation
of particles in the impactor produced a cone-shaped deposit at a point just below the impactor jet;
a slender needle developed at the top of the cone that extended above the oil surface. When the
needle broke off, it contaminated downstream filter measurements. The alternative oil, diocty
sebacate (DOS) seems to perform better than the previously specified diffusion oil,
tetramethyltetraphenyltrisiloxane, CAS3982-82-9 (i.e., DOW 704) that overcomes crystallization
under extreme atmospheric conditions (Vanderpool et al., 2004). I agree with the
recommendation to use WINS with DOS oil as an approved equivalent method. This should be
an independent FEM, irrespective of the FRM status of VSCC.
The recommendation of changing filter recovery time from 96 hours to 177 hours is favorable
and will ease the burden of site visits from nearly twice to once per week. Field tests by Papp et
al. (2002) did not report field blanks. Field blanks should be acquired and evaluated for passive
deposition and gas adsorption over the different time periods in the sampler and in storage.
Passive deposition may differ by sampler type. Papp et al. (2002) sampled different aerosols at
D-13
-------
Seattle, Rubidoux, Austin, Athens, Augusta, and RTF during different seasons. However, five
days per calendar quarter does not necessarily address the issues of positive and negative organic
artifact (e.g., Eatough et al., 1990; McDow and Huntzicker, 1990; Gundel et al., 1995; Chow et
al., 2005a) or nitrate volatilization (e.g., Hering and Cass, 1999; Chow et al., 2005b) with the
additional 81 hours in the field. Chow et al. (2002) showed that nitrate volatilization is
approximately 20% for 24 hours of filter recovery time after sampling, and approximately 44%
for 72 hours of filter recovery time in Mexico City. For areas with high nitrate concentrations,
such as Rubidoux, Fresno, and Bakersfield, CA, more tests during fall and winter are needed.
With regard to the "Filter Transport Temperature and Post Sampling Recovery Time," EPA
recommended the extension of post-sampling gravimetric analysis up to 30 days after the end of
the sample period (i.e., assume the day of sample recovery), provided that samples are
maintained at < 4 °C during transport from the field to the laboratory. The current equation
(Mobley, 2002) to calculate the elapsed time for post-weighing is:
D(Number of Days) = 34 - Tave (Average Temperature in °C) (1)
where:
Tave=(Tmax+Tmin)/2 (2)
Equation 2 is inexact since the guidance does not specify the method for measuring Tmax or Tmin
when field operators pack the PIVh.sFRM samples and record the temperatures. Our past
experience with Texas' FRM PM2.5 samples shows that Tmax represents ambient temperature of
the open cooler (i.e., Tmax is recorded immediately after the thermometer is turned on and before
the top layers of packed ice have covered the thermometer), rather than the cooler temperature at
the time that field operator is packing the cooler (Tropp et al., 2003). This issue can be resolved
if the procedure specifies that the thermometer is to be conditioned for 15 minutes before the
Tmax reading is taken.
During the past three years, on many occasions when Tmax was 20 °C and Tmin was -2 °C as
recorded by the field operator before shipping, the cooler temperature remained at approximately
3 °C when the cooler was opened in our laboratory. According to Equations 1 and 2, the
allocated days to perform gravimetric analysis would be 25 days instead of 31 days. Specific
procedures to record Tmax and Tmin should be clarified to implement such a calculation.
Under the current procedure, one can retrieve, pack, and ship the filters within one day, and have
them arrive at the laboratory the next day at a temperature of < 4 °C. One can then unload these
filters from the cooler, and equilibrate the filters in the weighing laboratory (at approximately 21
°C) for up to 28 days before weighing, since the regulations only state that they must be
conditioned at certain environmental conditions for at least 24 hours prior to weighing.
Additional tests in our laboratory during past years have shown that as long as filters are sealed
(airtight), and stored at < 4 °C, post-gravimetric weights remained within the tolerance of
microbalance precision even after two years of storage.
D-14
-------
Additional Recommendations for the PMi.5 FRM
• The current guidelines do not specify the chain-of-custody as field data (e.g., temperature,
flow rate) are downloaded. To prevent alteration of the raw data file, the EPA is
recommended to issue guidance for field data retrieval similar to that in U.S. EPA (1995).
• With respect to the Teflon-membrane filter, we have found the Pall Sciences (Ann Arbor,
MI) PTFE (polytetrafluoroethylene) Teflon-membrane with PMP (polymethylpentene)
support ring to be more versatile than the Whatman (Hillsboro, OR) PTFE Teflon-membrane
with polypropylene support ring (Catalogue # 7592-104). Whatman Teflon filters (40 jim
thickness) are 60% thicker than those from Pall Sciences (25 |im thickness), which results in
poorer minimum detection limits (MDLs) for elements by X-Ray Fluorescence (XRF)
analysis. Sometimes the FRM samples are used for elemental analysis. The advantage of
lower MDLs outweighs the cost difference between the two types of Teflon-membrane
filters. In the PMio-2.5 fraction, elements will be the dominant component. Sampling on
thinner filters for XRF analysis is desirable.
For the Rupprecht and Patashnick (R&P) 2025 Partisol-Plus Sequential Filter Sampler
(Rupprecht and Patashnick, Albany, NY), which is widely used in the PM2.5 FRM network,
the filter cassettes and magazines should be pre-labeled to minimize filter switching. This
will be more of an issue when two samplers are used for PMio-2.5, where misplacing the filter
cassette or magazine can result in erroneous PMio-2.5 masses. We have also observed that
filter cassettes can flip over within the magazine if they are jostled too much. The FRM
support grid is also an effective filter, as little mass gain was observed when air was drawn
through the holder in the wrong direction.
References
Allen, G. A., Sioutas, C., Koutrakis, P., Reiss, R., Lurmann, F. W., and Roberts, P. T. (1997).
Evaluation of the TEOM method for measurement of ambient particulate mass in urban
areas. Journal of the Air & Waste Management Association 47(6), 682-689.
Babich, P., Davey, M., Allen, G., and Koutrakis, P. (2000). Method comparisons for particulate
nitrate, elemental carbon, and PM2.5 mass in seven U.S. cities. Journal of the Air & Waste
Management Association 50(7), 1095-1105.
Bardsley, T. B. and Dal Sasso, N. A. (2005). Field trial of a TEOM®, FDMS TEOM® and
manual gravimetric reference method for determination of PM2.5 mass concentration. 1-5.
2005. Hobart, Tasmania. Proceedings, 17th Annual Meeting of the Clean Air Society of
Australia and New Zealand. 5-3-2005.
Bevington, P. R. (1969). Data Reduction and Error Analysis for the Physical Sciences, McGraw
Hill, New York, NY.
Chan, T. and Lippmann, M. (1977). Particle collection efficiencies of sampling cyclones: An
empirical theory. Enivron. Sci. Technol. 11(4), 377-386.
D-15
-------
Chow, J. C. (1995). Critical review: Measurement methods to determine compliance with
ambient air quality standards for suspended particles. Journal of the Air & Waste
Management Association 45(5), 320-382.
Chow, J. C., Watson, J. G., Edgerton, S. A., Vega, E., and Ortiz, E. (2002). Spatial differences in
outdoor PMio mass and aerosol composition in Mexico City. Journal of the Air & Waste
Management Association 52(4), 423-434.
Chow, J. C., Watson, J. G., Lowenthal, D. H., Chen, L.-W. A., and Magliano, K. (2005a).
Particulate carbon measurements in California's San Joaquin Valley. Chemosphere: in
press.
Chow, J. C., Watson, J. G., Lowenthal, D. H., and Magliano, K. (2005b). Loss of PM2.5 nitrate
from filter samples in central California. Journal of the Air & Waste Management
Association 55(8), 1158-1168.
Chung, A., Chang, D. P. Y., Kleeman, M. J., Perry, K. D., Cahill, T. A., Dutcher, D.,
McDougall, E. M., and Stroud, K. (2001). Comparison of real-time instruments used to
monitor airborne particulate matter. Journal of the Air & Waste Management Association
51(1), 109-120.
Dzubay, T. G. and Nelson, R.O. (1975). Self absorption corrections for x-ray fluorescence
analysis of aerosols, in Advances in X-Ray Analysis, Vol. 18, edited by W. L. Pickles, C.
S. Barrett, J. B. Newkirk, and C. O. Rund, pp. 619-631, Plenum Publishing Corporation,
New York, NY.
Eatough, D. J., Aghdaie, N., Cottam, M., Gammon, T., Hansen, L. D., Lewis, E. A., and Farber,
R. J. (1990). Loss of semi-volatile organic compounds from particles during sampling on
filters, in Transactions, Visibility and Fine Particles, edited by C. V. Mathai, pp. 146-
156, Air & Waste Management Association, Pittsburgh, PA.
Federal Register. (2002). Final Rule: Consolidated emissions reporting, 40 CFRPart 51. Federal
Register 67(111), 39602-39616. 6-10-2002.
Galy-Lacaux, C., Carmichael, G. R., Song, C. H., Lacaux, J. P., Al Ourabi, H., and Modi, A. I.
(2001). Heterogeneous processes involving nitrogenous compounds and Saharan dust
inferred from measurements and model calculations. Journal of Geophysical Research
106(D12), 12559-12578.
Gundel, L. A., Stevens, R. K., Daisey, J. M., Lee, V. C., Mahanama, K. R. R., and Cancel-Velez,
H. G. (1995). Direct determination of the phase distributions of semivolatile polycyclic
aromatic hydrocarbons using annular denuders. Atmospheric Environment 29(14), 1719-
1733.
D-16
-------
Gussman, R. A., Kenny, L. C., Labickas, M., and Norton, P. (2002). Design, calibration, and
field test of a cyclone for PMi ambient air sampling. Aerosol Science & Technology
36(3), 361-365.
Hering, S. V. and Cass, G. R. (1999). The magnitude of bias in the measurement of PM2.5 arising
from volatilization of particulate nitrate from Teflon filters. Journal of the Air & Waste
Management Association 49(6), 725-733.
Hunike, E. (2000). Memorandum: Alternative WINS oil. U.S. Environmental Protection Agency,
Research Triangle Park, NC.
John, W. and Reischl, G. (1980). A cyclone for size-selective sampling of ambient air. Journal of
the Air Pollution Control Association 30(8), 872-876.
Kenny, L. C. and Gussman, R. A. (1997). Characterization and modeling of a family of cyclone
aerosol preseparators. J. Aerosol Sci. 28(4), 677-688.
Kenny, L. C. and Gussman, R. A. (2000). A direct approach to the design of cyclones for
aerosol-monitoring applications. J. Aerosol Sci. 31(12), 1407-1420.
Kenny, L. C., Gussman, R. A., and Meyer, M. B. (2000). Development of a sharp-cut cyclone for
ambient aerosol monitoring applications. Aerosol Science & Technology 32(4), 338-358.
Kenny, L. C. and Thorpe, A. (2001). Evaluation of VSCC cyclones. IR/L/EXM/01/01. Health
and Safety Laboratory, Sheffield, UK.
Kenny, L. C., Merrifield, T., Mark, D., Gussman, R., and Thorpe, A. (2004). The development
and designation testing of a new USEPA-approved fine particle inlet: A study of the
USEPA designation process. Aerosol Science & Technology 38(Suppl.2), 15-22.
Lee, J. H., Hopke, P. K., Holsen, T. M., Polissar, A. V., Lee, D. W., Edgerton, E. S., Ondov, J.
M., and Allen, G. (2005a). Measurements of fine particle mass concentrations using
continuous and integrated monitors in eastern US cities. Aerosol Science & Technology
39(3), 261-275.
Lee, J. H., Hopke, P. K., Holsen, T. M., Lee, D. W., Jaques, P. A., Sioutas, C., and Ambs, J. R.
L. (2005b). Performance evaluation of continuous PM2.5mass concentration monitors. J.
Aerosol Sci. 36(1), 95-109.
Long, R. W., Eatough, N. L., Mangelson, N. F., Thompson, W., Fiet, K., Smith, S., Smith, R.,
Eatough, D. J., Pope, C. A., and Wilson, W. E. (2003). The measurement of PM2.5,
including semi-volatile components, in the EMPACT program: Results from the Salt
Lake City Study. Atmospheric Environment 37(31), 4407-4417.
McDow, S. R. and Huntzicker, J. J. (1990). Vapor adsorption artifact in the sampling of organic
aerosol: Face velocity effects. Atmospheric Environment 24A(10), 2563-2571.
D-17
-------
Mobley, J. D. (2002). Memorandum: Extension of filter retrieval time period for PM2.5 samples.
U.S. Environmental Protection Agency, Research Triangle Park, NC. Motallebi, N.,
Taylor, C. A., Turkiewicz, K., and Croes, B. E. (2003). Particulate matter in California:
Part 1 - Intercomparison of several PM2.5, PMio-2.5, and PMio monitoring networks.
Journal of the Air & Waste Management Association 53(12), 1509-1516.
Papp, M., Eberly, S. I, Hanley, T., Watkins, N., Barden, H., Noah, G., Bermudez, R., Eden, R.,
Franks, B., Johnson, A., Marriner, R., and Michel, E. (2002). Evaluation of filter
recovery period for the determination of fine particulate matter as PM2.5 in the
atmosphere. U.S. Environmental Protection Agency, Research Triangle Park, NC.
Peters, T. M., Gussman, R. A., Kenny, L. C., and Vanderpool, R. W. (2001a). Evaluation of
PM2.5 size selectors used in speciation samplers. Aerosol Science & Technology 34(5),
422-429.
Peters, T. M., Vanderpool, R. W., and Wiener, R. W. (2001b). Design and calibration of the EPA
PM2.swell impactor ninety-six (WINS). Aerosol Science & Technology 34(5), 389-397.
Pitchford, M. L., Chow, J. C., Watson, J. G., Moore, C. T., Campbell, D. E., Eldred, R. A.,
Vanderpool, R. W., Ouchida, P., Hering, S. V., and Frank, N. H. (1997). Prototype PM2.5
federal reference method field studies report -An EPA staff report. U.S. Environmental
Protection Agency, Las Vegas, NV.
Poor, N., Clark, T., Nye, L., Tamanini, T., Tate, K., Stevens, R., and Atkeson, T. (2002). Field
performance of dichotomous sequential PM air samplers. Atmospheric Environment
36(20), 3289-3298.
Rizzo, M., Scheff, P. A., and Kaldy, W. (2003). Adjusting tapered element oscillating
microbalance data for comparison with Federal Reference Method PM2.5 measurements
in Region 5. Journal of the Air & Waste Management Association 53(5), 596-607.
Russell, M., Allen, D. T., Collins, D. R., and Fraser, M. P. (2004). Daily, seasonal, and spatial
trends in PM2.5 mass and composition in Southeast Texas. Aerosol Science &
Technology 38(Suppl. 1), 14-26.
Tanner, R. L. and Parkhurst, W. J. (2000). Chemical composition of fine particles in the
Tennessee Valley region. Journal of the Air & Waste Management Association 50(8),
1299-1307.
Tropp, R. J., Jones, K., Kuhn, G., and Berg, Jr., N. J. (1998). Comparison of PM2.5 saturation
samplers with prototype PM2.5 Federal Reference Method Samplers, in Proceedings,
PM2.s: A Fine Particle Standard, edited by J. C. Chow and P. Koutrakis, pp. 215-225,
Air & Waste Management Association, Pittsburgh, PA.
D-18
-------
Tropp, R. 1, Engelbrecht, J. P., Lowenthal, D. H., Kohl, S. D., Dickerson, A. L., DuBois, D. W.,
Chow, J. C., Watson, J. G., Countess, R. J., and Countess, S. J. (2003). Assessment of
PM2.5 chemical speciation results for Texas: Part 1. 6470-665-2925.1D1. Desert Research
Institute, Reno, NV.
U.S. EPA. (1995). Good automated laboratory practices: Principles and guidance to regulations
for ensuring data integrity in automated laboratory operations. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
U.S. EPA. (1998). Quality assurance guidance document 2.12: Monitoring PM2.5 in ambient air
using designated reference or class I equivalent methods. U.S. Environmental Protection
Agency, Research Triangle Park, NC.
Underwood, G. M., Song, C. H., Phadnis, M., Carmichael, G. R., and Grassian, V. H. (2001).
Heterogeneous reactions of NCh and HNCb on oxides and mineral dust: A combined
laboratory and modeling study. Journal of Geophysical Research 106(D16), 18055-
18066.
Vanderpool, R. W., Byrd, L., Wiener, R. W., Hunike, E., Labickas, M., Leston, A., Tolocka, M.
P., McElroy, F. C., Murdoch, R. W., Natarajan, S., and Noble, C. A. (2004). Laboratory
and field evaluation of crystallized Dow 704 oil on the performance of the WINS PM2.5
fractionator, in Proceedings, Symposium on Air Quality Measurement Methods and
Technology-2004, pp. 23-1-23-17, Air and Waste Management Association, Pittsburgh,
PA.
Vanderpool, R. W. and Diller, D. J. (2005). Network operation of the PMio-2.5 difference method.
U.S. Environmental Protection Agency, Research Triangle Park, NC. Wu, P. M. and
Okada, K. (1994). Nature of coarse nitrate particles in the atmosphere -A single particle
approach. Atmospheric Environment 28(12), 2053-2060.
Yanosky, J. D., Williams, P. L., and Macintosh, D. L. (2002). A comparison of two directreading
aerosol monitors with the Federal Reference Method for PM2.5 in indoor air. Atmospheric
Environment 36, 107-113.
D-19
-------
Dr. Ellis Cowling
Comments by Ellis Cowling in Connection with the CASAC Ambient Air Monitoring and
Methods (AAMM) Subcommittee Peer Review of the PMio_2.s Federal Reference Method
and Consultation on Field Evaluation of PMi0-2.s Methods, and Related Matters as
Discussed During the CASAC AAMM Subcommittee Meeting in Durham, NC on
September 21-22, 2005
After reviewing the documents provided and listening carefully to the presentations made
during this Peer Review and Consultation with the CASAC AAMM Subcommittee, I offer the
following two general recommendations, the first of which was mentioned in an earlier CASAC
peer review and consultation on EPA's National Air Monitoring Strategy.
Recommendation 1:
EPA should guard against the tendency to give undue emphasis to "Data Quality
Objectives" in the selection and evaluation of instruments and subsequent implementation
of field monitoring programs to the exclusion of concern about "Science Quality
Objectives" and "Policy Relevancy Objectives."
Experience within the Southern Oxidants Study and other large-scale field measurement
and monitoring campaigns has demonstrated repeatedly that undue emphasis on "Data Quality
Objectives" often leads to:
1) Serious lack of attention to the scientific hypotheses and assumptions that are inherent in
any choice of scientific instruments, the appropriateness of the ground-based sites at
which the instruments are located, the skills of the instrument operators, the data
processing and data-display programs used, and especially the scientific quality of the
conclusions and statements of findings that are drawn from analysis and interpretation of
the measurements that are made; and
2) Equally serious lack of attention to the policy relevancy of the measurements being made
— relevancy to the general or specific enhancements of environmental protection that are
the real reason behind the public health and public welfare concerns that led to the
decision to establish a monitoring program in the first place.
A very important example of the need for inclusion of "Science Quality Objectives"
together with "Data Quality Objectives" in implementation plans for air quality monitoring
programs was provided during the verbal presentation during the meeting in Durham on
September 21, 2005. As Tim Hanley pointed out, a very prevalent worry about use of
subtraction methods in reporting ambient air quality measurements was the widely believed
frequency with which negative numbers show up in reports of air concentrations where
subtraction methods are used. When Hanley and his colleagues did very careful analyses of
the actual numbers used in making the subtractions that resulted in negative air concentration
numbers, it became clear that the data actually used frequently were from instruments that
were not co-located and thus were not measuring pollutant concentrations at the same
location, or were for measurements made at different times of the day, so that the two
numbers being subtracted were not for the same air parcel! Thus Hanley et al concluded that
D-20
-------
the prevalent worry about negative numbers was largely a myth - which he suggested (for
our enjoyment!) was similar to reports of "observations" of a "Loch Ness Monster" in
Scottish fairy tales!
The many years of experience accumulated by the scientists and engineer in the
Southern Oxidants Study show that careful analysis, careful interpretation, and careful
formulation of statements of findings from air quality measurements is the best quality
assurance method of all.
This is why I suggest that any program of ambient air quality measurements
(including the program for measurement and reporting of air concentrations of PMio-
2.5) should have explicitly stated "Science Quality Objectives" and well as explicitly
stated "Data Quality Objectives."
In this latter connection, permit me to call attention to the attached "Guidelines for the
Formulation of Scientific Findings to be Used for Policy Purposes." These guidelines were
developed originally by the NAPAP Oversight Review Board led by Milton Russell, former
Assistant Administrator for EPA. Please find attached on page 4 below, an electronic version
of these Guidelines which the Southern Oxidants Study adopted and very slightly adapted for
use in formulating policy relevant scientific findings from our research in 1988-2005.
The original version of these Guidelines was published as Appendix III of the April 1999
Report titled "The Experience and Legacy of NAPAP." This was a Report to the Joint Chairs
Council of the Interagency Task Force on Acidic Deposition of the Oversight Review Board
(ORB) of the National Acid Precipitation Assessment Program. As indicated in Appendix
III:
"The following guidelines in the form of checklist questions were developed by the
Oversight Review Board to assist scientists in formulating presentations of research
results to be used in policy decision processes. These guidelines may have broader utility
in other programs at the interface of science and public policy and are presented here
with that potential use in mind."
Recommendation 2:
This second recommendation derives from the general principle that '"''The words we use often
show the quality of our understanding." This general principle of communication about
public policy matters leads me to join with Peter McMurry and others in the AAMM
Subcommittee in recommending that care should be taken in the choice of words that EPA
uses (and all the rest of us use!) to describe:
1) The PMi0-2.5 particles themselves,
2) The PMi0-2.5 monitoring network established to measure and report air
concentrations of these particles in various parts of the country, and
3) The PMio-2.5 National Ambient Air Quality Standard(s) that are established to help
decrease human-health and public-welfare risks associated with exposure to PMio-2.5
particles.
D-21
-------
The following words often (or occasionally) are used more-or-less interchangeably in
describing these three different entities even though these various words are not necessarily
equivalent and in many cases do not portray the same understanding about these important
entities:
1) Words used to describe the PMm-i s particles themselves:
• coarse particles (CP),
• coarse aerosol particles (CAP),
• urban coarse particles (UCP),
• coarse particulate matter (CPM),
• urban coarse particulate matter (UCPM),
• thoracic particles (TP),
• urban thoracic particles (UTP),
• urban thoracic coarse particles (UTCP),
• urban thoracic coarse particulate matter (UTCPM),
• PMio-2.5 particles (PMi0-2.5P),
• urban PMi0-2.5 particulate matter (Uio-2.sPM);
2) Words used to describe the PMm-i s monitoring network:
• Coarse Particle Network (CPN),
• Coarse Aerosol Monitoring Network (CAMN),
• Urban Coarse Particle Network (UCPMN),
• Coarse Particulate Matter Network (CPMN),
• Urban Coarse Particle Network UCPN),
• Thoracic Particle Network (TPN),
• Thoracic Coarse Particle Network (TCPN),
• Urban Thoracic Coarse Particle Network (UTCPN),
• Urban Thoracic Coarse Particulate Matter Network (UTCPMN),
• PMio-2.5 Network (PMi0-2.5N),
• Urbanio-2.5Particulate Matter Network (Uio-2.sPMN);
3) Words used to describe the PMi0-2.5 National Ambient Air Quality Standard
(NAAOS):
• Coarse Particle Standard (CPS),
• Coarse Aerosol Standard (CAS),
• Urban Coarse Particle Standard (UCPS),
• Coarse Particulate Matter Standard (CPMS),
• Urban Coarse Particulate Matter Standard (UCPMS),
• Thoracic Particle Standard (TPS),
• Urban Thoracic Particle Standard (UTPS),
• Urban Thoracic Coarse Particle Standard (UTCPS),
• Urban Thoracic Coarse Particulate Matter Standard (UTCPMS),
• PMio-2.5 Particle Standard (PMio-2.5PS),
• Urban PMi0-2.5 Particulate Matter Standard (Ui0-2.sPMS).
D-22
-------
GUIDELINES FOR THE FORMULATION OF STATEMENTS OF SCIENTIFIC
FINDINGS TO BE USED FOR POLICY PURPOSES
The following guidelines in the form of checklist questions were developed by the NAPAP
Oversight Review Board to assist scientists and engineers in formulating statements of research findings
to be used in policy decision processes.
1) IS THE STATEMENT SOUND? Have the central issues been clearly identified? Does each
statement contain the distilled essence of present scientific and technical understanding of the
phenomenon or process to which it applies? Is the statement consistent with all relevant evidence-
evidence developed either through NAPAP [or SOS] research or through analysis of research
conducted outside of NAPAP [or SOS]? Is the statement contradicted by any important evidence
developed through research inside or outside of NAPAP [or SOS]? Have apparent contradictions or
interpretations of available evidence been considered in formulating the statement of principal
findings?
2) IS THE STATEMENT DIRECTIONAL AND, WHERE APPROPRIATE, QUANTITATIVE?
Does the statement correctly quantify both the direction and magnitude of trends and relationships in
the phenomenon or process to which the statement is relevant? When possible, is a range of
uncertainty given for each quantitative result? Have various sources of uncertainty been identified and
quantified, for example, does the statement include or acknowledge errors in actual measurements,
standard errors of estimate, possible biases in the availability of data, extrapolation of results beyond
the mathematical, geographical, or temporal relevancy of available information, etc. In short, are there
numbers in the statement? Are the numbers correct? Are the numbers relevant to the general meaning
of the statement?
3) IS THE DEGREE OF CERTAINTY OR UNCERTAINTY OF THE STATEMENT
INDICATED CLEARLY? Have appropriate statistical tests been applied to the data used in drawing
the conclusion set forth in the statement? If the statement is based on a mathematical or novel
conceptual model, has the model or concept been validated? Does the statement describe the model or
concept on which it is based and the degree of validity of that model or concept?
4) IS THE STATEMENT CORRECT WITHOUT QUALIFICATION? Are there limitations of
time, space, or other special circumstances in which the statement is true? If the statement is true only
in some circumstances, are these limitations described adequately and briefly?
5) IS THE STATEMENT CLEAR AND UNAMBIGUOUS? Are the words and phrases used in the
statement understandable by the decision makers of our society? Is the statement free of specialized
jargon? Will too many people misunderstand its meaning?
6) IS THE STATEMENT AS CONCISE AS IT CAN BE MADE WITHOUT RISK OF
MISUNDERSTANDING? Are there any excess words, phrases, or ideas in the statement that are not
necessary to communicate the meaning of the statement? Are there so many caveats in the statement
that the statement itself is trivial, confusing, or ambiguous?
7) IS THE STATEMENT FREE OF SCIENTIFIC OR OTHER BIASES OR IMPLICATIONS OF
SOCIETAL VALUE JUDGMENTS? Is the statement free of influence by specific schools of
scientific thought? Is the statement also free of words, phrases, or concepts that have political,
economic, ideological, religious, moral, or other personal-, agency-, or organization-specific values,
overtones, or implications? Does the choice of how the statement is expressed rather than its specific
words suggest underlying biases or value judgments? Is the tone impartial and free of special
pleading? If societal value judgments have been discussed, have these judgments been identified as
such and described both clearly and objectively?
8) HAVE SOCIETAL IMPLICATIONS BEEN DESCRIBED OBJECTIVELY? Consideration of
alternative courses of action and their consequences inherently involves judgments of their feasibility
and the importance of effects. For this reason, it is important to ask if a reasonable range of alternative
policies or courses of action have been evaluated? Have societal implications of alternative courses of
action been stated in the following general form?:
D-23
-------
"If this [particular option] were adopted then that [particular outcome] would be expected."
9) HAVE THE PROFESSIONAL BIASES OF AUTHORS AND REVIEWERS BEEN
DESCRIBED OPENLY? Acknowledgment of potential sources of bias is important so that readers
can judge for themselves the credibility of reports and assessments.
D-24
-------
Mr. Bart Croes
U.S. EPA's Coarse PM FRM and Other PM Monitoring Issues
September 21-22, 2005 Peer Review and Consultation Meeting
CASAC AAMM Subcommittee Review Comments, Bart Croes
Overall, the documents the Subcommittee reviewed continue the impressive initiative by U.S.
EPA to take a systematic approach towards implementation of a likely coarse particle (PMio-2.s)
National Ambient Air Quality Standard (NAAQS). I appreciate the opportunity to comment
during this near-final stage of the process, about a year after we provided input on intermediate
results. The documents provide a good description of the basis for a PMi0-2.5 Federal Reference
Method (FRM), clearly explain the new results from the multi-site evaluation of candidate
methods, and provide a reasonable rationale for changes to the PM2.5 FRM and development of a
PMio-2.5 Federal Equivalent Method (FEM) process. I agree with the basic approach taken by
U.S. EPA, and offer comments on several aspects that need further attention. My comments
address the two peer review and eight consultation questions posed by Phil Lorang in his August
19, 2005 memo to Fred Butterfield. These comments reflect considerable input from California
Air Resources Board (CARB) staff responsible for implementing U.S. EPA monitoring
requirements and using the data in source apportionment and health studies.
Peer Review Questions:
1. What are the scientific and operational strengths and weaknesses of the PMio-2.5 difference
method relative to other options for a proposed FRM, especially when used as the basis for
approval of other methods?
The difference method has several major strengths over other FRM options; namely that it uses
monitoring equipment already deployed as part of the Nation's large investment in the PM2.5
FRM. It provides a consistent basis for comparison and forces cutpoint curves in candidate
samplers to conform to those of the existing PM2.s FRM. From an operational perspective, the
equipment and procedures to operate a PMio low-volume sampler are almost identical to those of
the PM2.5 FRM filter sampler, therefore the PMio and PM2.5 FRM manual filter methods should
integrate easily into existing PM2.s monitoring networks. California re-established its PMio
ambient air quality standards in 2002 (and set a new annual-average PM2.s standard). The use of
the PMio low-volume sampler for the PMio-2.5 FRM will enable monitoring agencies such as
CARB to maintain compliance with current State PMio monitoring requirements. Another
advantage is that filters can be analyzed for particle composition, although it is unclear whether
or not errors in the difference method for key tracer species are sufficiently small to allow
receptor models to be applied, in contrast to the dichot method where direct chemical analysis of
coarse particles is possible.
The major drawbacks of the difference method are the same as those of other filter-based
approaches, namely, the lack of 24-hour time resolution and expensive, manual filter handling
and analysis. Time-resolved, real-time availability of PM data are necessary for use air quality
index (AQI) forecasting and burn allocations, and can lead to a better understanding of emission
D-25
-------
sources, transport, background levels, deposition, and health effects of PM, although U.S. EPA is
addressing this need through the FEM process. As in previous monitoring programs for
airborne particles (i.e., TSP, PMio, PM^.s), it will not be advantageous to consider a method that
is only useful to answer questions of attainment if it limits one in making other important
observations. The difference method will also prove cumbersome and expensive because it
requires great care in shipping and filter handling by experienced personnel. This means that
widespread deployment of the difference method in field operations is likely to result in less data
being collected at fewer sites then for a real-time instrument. The sampler difference method
requires high quality data from collocated instruments, and despite the excellent precision results
obtained in the multi-site sampler inter-comparison study by U.S. EPA, is subject to more errors
deriving from both instrument operations and operator error than single-instrument approaches.
It has yet to be demonstrated that the difference method can function in low-PM environments
and produce useful data, although this will not be an issue if only a 24-hour (and not an annual-
average) PMio-2.5 standard is promulgated.
All of these issues are addressable if the FEM process qualifies continuous, real-time methods(s)
and perhaps a single-instrument filter sampler for speciation analyses. I am in agreement with
U.S. EPA on using the difference method as an FRM. After all, a difference method is already
successfully used to determine NC>2 levels. However, since the number of PMio-2.5 non-
attainment areas is expected to be much smaller than for PM2.5 and many of these areas have
existing PMio monitoring programs, U.S. EPA could limit deployment of the cumbersome
difference method for PMio-2.5 if it allowed PMio monitors to be used to determine attainment.
U.S. EPA and SLT agencies have already invested huge resources into the current PMio and
PM2.s monitoring networks. Several states (i.e., California) have State ambient air quality
standards for PMio and do not plan to follow U.S. EPA in adopting a coarse particle standard.
Surely if a site meets the PMio-2.5 standard with PMio monitoring data (uncorrected), then there is
no need to deploy a PMio-2.s-specific monitor at the site.
The potential scope of a national PMio-2.5 monitoring network should be defined. While U.S.
EPA has not yet promulgated a PMio-2.5 NAAQS, it has released a Staff Paper with a proposed
range of possible standards for PM2.5 and PMio-2.5. As a first-order estimate, data from the
existing PMio monitoring network should be compared to the proposed lower and upper ranges
of the coarse particle recommendations to determine if the potential scope of a PMio-2.5
monitoring network would be national in scale or restricted to a few states. In these likely non-
attainment areas, PMio would primarily consist of the coarse fraction. Sites that have collocated
PM2.s and PMio monitors, or SLT agencies that have operated dichot samplers (see Motallebi, et
al., 2003ab for California) provide more relevant data. A list and map of sites with PMio only,
PM2.5 only, and both would be a useful summary.
On a parallel basis, the U.S. EPA should devote resources (preferably a STAR grant) to
developing a traceable standard for PM. Problems with the TEOM, APS, and other units were
only discovered during the inter-comparison study because multiple units were carefully
operated by U.S. EPA and monitoring industry experts. If the units were operating by
themselves in an SLT agency monitoring station, it is unlikely that instrument drift and other
problems would have been noticed. Without the ability to challenge a PM analyzer with a know
concentration of PM, all you have to verify proper operation of an analyzer is the "due diligence"
D-26
-------
of the site technician. If U.S. EPA took a dozen of the candidate samplers and sent them to 12
randomly selected SLT agencies, after four months they would get a dozen different regressions
and correlations, no matter how consistent the analyzers performed in the controlled, three-city
study.
Other continuous, criteria pollutant monitors (Os, NC>2, CO, 802) are challenged with a known
concentration each day (the in-station zero and span checks) and at six- and twelve-month
intervals (independent transfer standards). For filter samplers the micro-balance used to weigh
the filter is similarly "zeroed and spanned" with NIST-traceable standard weights each weighing
session. Ozone does not come in a bottle, but accurate and precise quantities are generated on
demand to challenge ozone analyzers. Resources should be devoted to research an accurate and
precise PM generation system. I realize this would be very difficult to do with PM, but perhaps
something similar to an aerosol inhaler (used for administering asthma medication) could be
developed.
When simple comparisons of the FRM-like difference method to other approaches is undertaken
it is very easy to forget that the FRM does not really report on PM as it is in the real world, rather
it represents the result of highly controlled sampling and filter handling that is known to, in some
conditions, to over- or under-report ambient PM. It may not serve the reader to include
statements such as are in Attachment 1 as follows: ".. .the proposed PMi0-2.5 FRM utilizes the
same fundamentally sound integrated sample, filter-collection, and mass-based gravimetric
measurement technology that has been basic to all previous FRMs..." It is known that organic
and nitrate compounds are not retained fully in these "fundamentally sound integrated samples".
It is also well known that moisture either on particles or retained by the filter can be extremely
influential in PM mass measurements. In general, the text presented on page four of this
attachment is far too simplistic.
2. Based on the field study report as well as any other available data, e.g., data from State and
local agencies, how does the demonstrated data quality of the PMio-2.5 difference method
support or detract from it being proposed as a FRM?
For sites operating both PMio and PM2.s FRM samplers, the sampler difference method will be
very cost-effective, but it is dependent on successful collection of both samples, thus sample
recovery and data completeness would be hostage to the degree of success of operation of the
other two samplers. For sites lacking one or the other sampler (e.g., established attainment
areas), it may be preferable to operate a "stand-alone" coarse particle sampler, thus adopting a
difference-based FRM should be accompanied by adoption of a single-instrument equivalent
method.
Of the candidate technologies, only the difference method and the dichot method offer the
significant advantage of traceability to primary standards for both the mass and volume (flow x
time) components of the PMio-2.5 mass concentration. However, our experience with
dichotomous samplers has shown that they can be more difficult to calibrate, maintain, and audit
than the FRM samplers. This is due to the fact that both the PM2.s and PMio-2.5 flowrates are
interdependent and changes/adjustments to one affects the other. Given this, and the reality that
the laboratory resources are identical for both the difference method (two pre-weights, two post
D-27
-------
weights) and the dichot method (two pre-weights, two post weights), we support selection of the
difference method as the PMi0-2.5 FRM.
Decisions regarding the deployment of the various continuous PMi0-2.5 samplers should not be
made solely on equivalence to the FRM difference approach. There may be varying valid
reasons to select one or more of these methods beyond simple equivalency. For example, the
ability to address coarse or fine PM levels on an hourly or sub-hourly time interval could allow
experimental investigations of short-term health impacts and chemical/photochemical processes
that produce ambient PM. It is probably more important to understand the fundamental reasons
why the various alternative methods vary from the FRM than to simply establish a few points of
difference. Application of rigorous criteria for designation of equivalence for PMio-2.5 methods,
such as are presented in Attachment 4 (for PM2.s), may be quite cumbersome and problematic.
By definition, it would seem that only the FRM difference method can be expected to produce
fully equivalent results. Perhaps a functional equivalency based on practical considerations
would be more useful. However, discussions of functional equivalency issues are quite sparse in
the documents provided.
The attachment contains lengthy considerations related to the need to avoid selection of
proprietary sampler technologies. However, if one views the nature of the current PM2.5
network, we find that only two or three companies sell FRMs and most of these are sold by R&P.
Few companies have the resources to develop, test, and comply with all the factors needed to
reach acceptance and those that do consider the probable market when they determine to propose
devices. Should a beta gauge technology be chosen, there are no more than three manufacturers
worldwide (with Metone being the dominant U.S. vendor), and there is only one for the TEOM
or the APS. What this means is that it is likely that a proprietary sampler technology is almost
inevitable should any continuous device be selected or certified.
If the difference method is selected as the FRM for PMio-2.5, how will negative numbers,
however infrequently they occur, be handled? Will negative values be reported, invalidated,
corrected to zero, or corrected to detection limit values?
Consultation Questions:
1. Based upon the latest available field study data, which PMio-2.5 methods have both sufficient
utility to meet one or more important monitoring objectives and appropriate data quality to be
considered for deployment as Federal Equivalent Methods (FEMs) or speciation in a
potential PMio-2.5 monitoring network?
Simply documenting exceedances of PM standards, without collecting a sample that can by
subjected to chemical or physical analysis, is a poor allocation of resources. Physical samples
need not be analyzed for more than mass on a routine basis, but an archived set of samples
available for further analyses can be an invaluable aid to determining causes of past exceedances
and as a cost- and time-saving resource for a wide range of atmospheric process studies.
Methods that do not collect samples (e.g., TEOM) should not be used in areas of marginal
attainment or non-attainment of standards where regulators can reasonably anticipate that source-
related analyses will be needed. These considerations all point to a next-generation (i.e., tighter
D-28
-------
2.5 |j,m outpoint) dichotomous sampler as the most flexible measurement method for attainment
determination and source assessment.
Only the manual dichot sampler appears capable of meeting the objective of "high FRM
comparability". Though not in itself a speciation method, the dichot is the best method to collect
PMio-2.5 samples for subsequent laboratory speciation. At this time, there does not appear to be a
validated candidate for providing highly time-resolved data to support a PMio-2.5 standard,
although several methods appear promising and U.S. EPA's efforts to evaluate candidate devices
should continue.
The field study document indicated that the next (and final) intercomparsion study would take
place at Birmingham, AL. Two alternative sites (in likely nonattainment areas) would present
more challenging environments. Riverside, CA has high levels of PMio-2.5 and PM2.5 nitrate
during the late summer and fall, and nitrate volatilization losses could be evaluated for the
candidate samplers. The remaining PM Supersite in Fresno, CA also has high PMio-2.5 and PM2.5
nitrate during the fall harvest season. SLT agencies should be involved to duplicate "real-world"
operation. Perhaps the existing dichot monitor and high-volume PMio and low-volume PM2.5
monitors (that are already deployed by some SLT agencies) should be included to determine
their suitability to determine if an area meets the NAAQS (e.g., they do not have negative
biases).
The initial text for this attachment should clearly delineate the objectives for the tests conducted.
The question requests input on whether methods tested ".. .meet one or more important
monitoring objectives.. .to be considered for deployment as Federal Equivalent Methods..." It is
somewhat unclear whether the text describes the results of comparison studies conducted to
determine the performance of devices that may be considered as PMio-2.5 FRMs or as FEM
monitors. The document or other text should contain descriptions of these possible monitoring
objectives as a part of framing the testing approach and presentation of results. For example, it
should describe the data needs of health researchers, air modelers, those performing compliance
ascertainment, or others. While the researchers/planners and the U.S. EPA authors of
Attachment 2 may have an idea of data needs and how various monitors might meet these needs,
this would be useful information for the regulated community and the regulatory agencies
charged with carrying out the monitoring program.
2. What are the Subcommittee's views on the Very Sharp Cut Cyclone (VSCC) being approved
as an alternate second-stage impactor to the WINS for use on a PM2.5 FRM?
Replacing the WINS with the VSCC has strong technical merit. For the BGI, R&P, and
Andersen PM2.5 FRM, the U.S. EPA has already designated the BGI VSCC PM2.5 size inlet
modification as Federal Equivalent Methods. The VSCC requires less maintenance than the
WINS impactor. The VSCC also provides improved cut point characteristics and does not use
any oil. However, mandated network-wide retrofit of installed FRM samplers should be avoided
unless new data indicate that the "old" methods pose problems (i.e., let the local operators update
at their discretion).
D-29
-------
3. To what extent are the stated advantages of relaxing existing requirements identified for the
PM2 5 FRM supported by the information cited in Attachment 3, available literature, or good
field and laboratory practices? Does the Subcommittee have additional recommendations for
the PM2.5 FRM that would be neutral with respect to bias, but would improve the
performance and minimize the burden on agencies conducting the sampling?
The other changes recommended by U.S. EPA staff also have strong technical merit.
WINS Oil: The use of the alternative DOS oil is reasonable, although utilization of the VSCC
(see above) eliminates the need for any impaction oil.
Filter Recovery Time: Although the 177-hour recovery time may ease operator time schedules,
the effort to pick up exposed filters as soon as possible after sampling should continue.
Concerns about greater opportunities for changes on exposed filters should not be over-weighted
since most of the artifacts of concern (e.g., nitrate loss, water loss/gain) are not adequately
prevented by the present procedures.
Filter Transport Temperature and Post Sampling Recovery Time: PMi0 and PM2 5 filters are
equilibrated at 25°C for a minimum of 24 hours prior to post weighing. Given this, is
maintaining 4°C from post sampling pickup and transportation necessary? Is there evidence to
support the significance of maintaining 4°C or less during transport of filter samples? Shipping
costs would lessen if transportation conditions were relaxed to 25°C or below.
4. Considering the statistical measures of precision, correlation, multiplicative bias, and
additive bias identified for approval of PM2.5 continuous methods, what are the
Subcommittee's views on the usefulness of each measure to ensure that approved or
equivalent methods meet the monitoring network data quality objectives?
The methods presented in the September 30, 2004 draft are reasonable. However, U.S. EPA
should bear in mind that there are three goals of equivalency:
• Equal protection of public health across a wide range of administrative and regulatory
settings.
• Equal treatment of the regulated community.
• Flexibility for local regulators to tailor monitoring to local needs.
So long as the uncertainty in sampling is well below the uncertainty in other aspects of the
regulatory process, then reasonable goals for "good measurement practice" are sufficient. The
DQO software tool approach is a good start in linking monitoring specifications to their
regulatory consequences. However, the current criteria for PM2 5 FEM designation are
significantly more straightforward and much easier to generate and duplicate. Once FRM and
candidate data sets are validated and paired into a software program such as Microsoft EXCEL,
graphs can be generated quickly, with linear regression results of slope (m), intercept (b) and
correlation coefficient (r). These linear regression values are universally understood by most
data users.
D-30
-------
The proposed criteria call for a range of candidate samplers. This should be a fixed number for
all FEM testing. Further, the proposed criteria call for averaging candidate sampler data. This
would have the tendency to smooth data, and average out noisy candidate samplers. This criteria
does not directly reflect real world applications where most monitoring stations have only one
sampler or monitor and data is not averaged.
The minimum daily FRM sample time should remain at 24 hours +/- 1 hour (23 to 25 hours),
adhering to current CFR criteria. There is no need to shorten this criterion.
5. What are the advantages and disadvantages of using sampler precision and sample population
to help determine the minimum correlation requirement for the approval of PM2.5 continuous
methods?
The measurement community has long recognized that local conditions can sometimes cause
otherwise "acceptable" methods to behave poorly. Specifications based on assumptions about
the composition of the atmosphere should be, where practical, replaced by specifications based
on actual local conditions, or at least, interpretation of QA data should be sensitive to local
variation. See response to Question 4 (above). In this context, the analysis presented should be
expanded to show how real data from a range of sites compare to the synthetic data presented in
the September 30, 2004 draft.
Moreover, when treating both real and synthetic data, they should be recognized as composed of
continuous size distributions (i.e., the tail of the "coarse" overlaps the "fine", and vice-versa).
This would permit evaluation of the meaning of discrepancies among samplers, rather than flat
specification of the FRM as "right" and the candidate methods being "over" or "under".
If a "loosening" of FEM criteria is desired, then it would be more prudent to simply alter the
slope, intercept, and/or correlation coefficient. In the proposed criteria, target values for the
correlation coefficient and additive bias remain unknown until all calculations are completed.
Maintaining current linear regression calculations is significantly more beneficial when
presenting comparison data or meeting government regulations.
With the proposed FEM calculations, the mixing of "population" and "sample" based equations
limits the use to manual or macro type calculations, significantly increasing the effort and time
involved in generating the proposed results. The proposed FEM method increases the possibility
of error while making the final results more difficult to understand. It appears that the proposed
FEM criteria is meant to skew the comparison of candidate samplers to FRM samplers so that
specific biases may pass while opposite but equal biases may fail.
6. What are the Subcommittee's views on using a PM2.5 continuous monitor approved as a
FEM, being applicable for use as part of a potential PM2.5 secondary standard for visibility?
Continuous PM2.5 monitoring is suggested as a reasonable surrogate for anthropogenic visibility
degradation. Since PM2.5 is generally correlated with reduced visibility, the appeal is obvious.
However, the problem with this approach is that there is no accompanying proposed mechanism
for making use of short-term PM2.5 data as a basis for regulating. The literature of human
D-31
-------
perception of haze and "smog" suggests that responses can be highly idiosyncratic. Given that
there is no policy guidance on how frequent or persistent low visibility needs to be to have a
"deleterious effect" on public welfare, it is unclear how specifying measurements now is useful
to U.S. EPA's regulatory process. It may be that visibility protection could be achieved based
solely on a program using current and expected health-based PM monitoring data.
7. To what extent have the assessments of spatial variability and the sensitivity of the DQO
process to a variety of population distributions been appropriately addressed?
Spatial and temporal variation and autocorrelation need to be treated more realistically. The
reality of sampler exposure in urban settings can include strong diurnal cycles that mix particles
from one source (e.g. road dust - with very high spatial and temporal variation) with those from
entirely different ones (e.g., aged secondary aerosol - a more regionally distributed pollutant)
within a time-integrated sample. The spatial and temporal structure of ambient concentrations,
size distributions, volatility, hygroscopicity, and other factors can all influence sampler
performance and how well sample values represent the ambient environment's real variability.
The analysis presented is a good start, but only a start: the next step is to add continuous size
distributions and to populate this analysis with data based on a wide range of realistic exposure
cases. See response to Question 5 (above).
The results from special studies could be analyzed to determine the spatial distributions
2.5 and provide guidance to the number of monitors that need to be sited in potential PMio-2.s non-
attainment areas to properly represent population exposure. The California Air Resources Board
and perhaps other SLT agencies have conducted such special studies. One example is a PM
saturation study conducted by DRI with mini-vols during 2000 in Corcoran, an agricultural
community in the San Joaquin Valley with high dust levels. Similar studies may have been
conducted in Las Vegas and Phoenix.
8. What are the Subcommittee's views on the approach identified for the development of
criteria to approve continuous PMio-2.5 equivalent methods?
The approach laid out in the August 19, 2005 draft is not adequately sensitive to "real -world"
variables. Specifically, the requirement for seasonal comparison should apply to a range of sites,
just as the "given season" is applied across a range of sites. Think of this as equalizing the
sampling matrix, with consequent potential for better statistical characterization not only of inter-
sampler agreement, but site- and season- linked variation as well.
(A few examples: winter in Fresno challenges samplers with high nitrate and high humidity;
summer in Riverside presents high nitrate and a high potential for sampling artifacts; winter in
St. Paul tests sampler resistance to freezing and internal ice formation; summer in Atlanta tests
for possible effects of ambient nitric acid and unsaturated aerosol sulfate ion).
References
Motallebi, N., C. A. Taylor, Jr., K. Turkiewicz, and B. E. Croes (2003a) Particulate matter in
California: Part 1 - Intercomparison of several PM2.5, PMio-25, and PMio monitoring networks.
J. Air Waste Manage. Assoc., 53: 1509-1516.
D-32
-------
Motallebi, N., C. A. Taylor, Jr., and B. E. Croes (2003b) Particulate matter in California: Part 2
- Spatial, temporal, and compositional patterns of PM2.5, PMio-25, and PMio. J. Air Waste
Manage. Assoc., 53: 1517-1530.
D-33
-------
Dr. Kenneth Demerjian
Final Comments by Demerjian. K.L.
September 25, 2005
U.S. EPA Clean Air Scientific Advisory Committee (CASAC)
CASAC Ambient Air Monitoring & Methods (AAMM) Subcommittee
Meeting September 21, 2005 - September 22, 2005
Peer Review of PM10-2.5 Federal Reference Method (FRM); and Consultation on Field
Evaluation of PM10-2.5 Methods, Optimization of the PM2.5 FRM, Equivalency Criteria
for
PM2.5 Continuous Methods, Monitoring Data Quality Objectives for PM10-2.5, and
Equivalency Criteria for PM10-2.5 Continuous Methods
Response to charge Questions:
Attachment 1 - Selection and technical summary of PMio-i.s FRM:
What are the scientific and operational strengths and weaknesses of the PMi0-2.5
difference method relative to other options for a proposed FRM, especially when used
as the basis for approval of other methods?
The technique does not address volatile losses and may introduce a systematic bias in mass
measurements which are likely source, composition and temperature sensitive. The health
consequences of PM exposures are likely size and chemical composition related. There is
evidence that FRM methods are subject to seasonal bias due to losses of volatile species. It is
also likely that such losses will have diurnal characteristics, emphasizing the importance of
higher time resolved measurements.
Based on the field study report as well as any other available data, e.g., data from State
and local agencies, how does the demonstrated data quality of the PMi0-2.5 difference
method support or detract from it being proposed as a FRM?
The precision reported in these studies certainly supports EPA 's rationale and desired objective
to establish PM 10-2.5 FRM that has a "high degree of fidelity and faithfulness" but unfortunately
it may not be the correct realization of PM mass in the atmosphere and therefore may not be
representative of inhalation exposures of greatest potential harm. Therefore in developing and
defining the FRM, EPA must acknowledge the likelihood of volatile losses biasing the FRM
measurement and provide the latitude for embracing emerging instrumentation technologies that
are proving to provide more accurate measurement of ambient PM mass.
Attachment 2 - EPA's Multi-Site Evaluations of Candidate Methodologies for Determining
Coarse Particulate Matter (PMio 2.5) Concentrations: August 2005 Updated Report
Regarding Second-Generation and New PMio-i.s Samplers
Based upon the latest available field study data, which PMio-2.s methods have both
sufficient utility to meet one or more important monitoring objectives and appropriate
D-34
-------
data quality to be considered for deployment as Federal Equivalent Methods (FEMs) or
speciation samplers in a potential PMi0-2.5 monitoring network?
The subject work is a more detailed look at stage one evaluations of PM course methods (2003 -
2004) as presented to the CASAC AAMMin July 22, 2004 and follow-up work on second
generation samplers performed in Phoenix, AZ (May & April 2005). As much as I find these
results interesting, they seem not to address fundamental questions about PMi 0-2.5 mass
measurement and the uncertainties associated with mass losses as a function of chemical
composition, source mix, and season. This issue was raised at the July 22, 2004 meeting and
seems to have been ignored. The fact is if the PM2.5, PMioandPMio-2.sFRMs as mandated may
not be the correct realization of PM mass in the atmosphere and therefore may not be
representative of inhalation exposures of greatest potential harm. At the July 22, 2004 meeting in
discussing the first draft, the suggestion was made that the chemical content of the samples
should be studied for the various sites and the potential for volatile losses assessed for the
respective sites. The current report provides no information on the chemical composition ofPM
at these sites. Why not?
I understand EPA 's rationale for wanting to establish PMi 0-2.5 FRM (attachment I) because of its
"high degree of fidelity and faithfulness ", but I am concerned that its regulatory mandate has
put blinders on its monitoring strategy. The health consequences of PM exposures are likely size
and chemical composition related. There is evidence that FRM methods are subject to bias due
to losses of volatile species. It is also likely that such losses will have diurnal characteristics,
emphasizing the importance of higher time resolved measurements. Some of the second
generation techniques tested may more effectively capture PM masses as compared to the FRMs.
I don't doubt that differences identified in the subject study with regard to size cutoff issues for
some second generation techniques evaluated are important and real, but the study has
discounted volatile losses and made no attempt to test it presence/contribution to the observed
differences experienced. Progress in determining the health effects ofPM (and likely important
chemical constituents) will remain hampered if our only main monitoring resource is 24-hr FRM
mass measurements.
Based on the performance evaluations reported in the subject document and assuming the inlet
and engineering issues identified are effectively resolved and robust operational requirements
met, I would considered the R&P Model 2025 sequential dichotomous sampler, the R&P
continuous Coarse TEOM monitor, the R&P single event dichotomous sampler, the Sierra-
Anderson 241 dichotomous sampler, and the R&P dichotomous TEOM sampler for deployment
as Federal Equivalent Methods (FEMs). Further testing and resolution of issues associated with
the Kimoto SPM-613D and the BGI Omni samplers are required prior to their consideration.
Given their dependencies of aerosol density and shape factors, it is not clear that the TSI Model
3321 APS or the GRIMM ET Model 1.107 can achieve FEM status. Further analysis of the
Kimoto SPM-613D and GRIMM ET Model 1.107 - FRM comparisons should be considered in
conjunction with available chemical speciation data. Both methods suggest a high bias that may
reflect FRM species volatility issues.
Additional comments on next steps
D-35
-------
Although one can provide some rationale for the sites selected in the preliminary evaluations,
the proposed follow-on study in Birmingham, AL does not seem justified. Why not consider
deployments in major cities like Houston or New York where PMpopulation exposures are high
and the sites represent diverse source mixes and climate differences that might well accentuate
the FRMbias issues raised concerning PM volatility.
Attachment 3 - Memo to PM NAAQS Review Docket (OAR-2001-0017) - Potential
changes being evaluated for the PMi.s Federal Reference Method
What are the Subcommittee's views on the Very Sharp Cut Cyclone (VSCC) being
approved as an alternative second-stage impactor to the Well Impactor Ninety-Six
(WINS) for use on a PlVh.s FRM?
Technically this is an acceptable alternative, but one does wonder if the minimal testing that has
occurred under these studies has identified all sources of error that might result in more
representative multi-site, year around operational testing.
To what extent are the stated advantages of relaxing existing requirements identified for
the
PM2.5 FRM supported by the information cited in Attachment 3, available literature, or
good field and laboratory practices? Does the Subcommittee have additional
recommendations for the
PM2.5 FRM that would be neutral with respect to bias, but would improve the
performance and minimize the burden on agencies conducting the sampling?
The three suggested changes in FRM procedures (in addition to the use of the Very Sharp Cut
Cyclone), i.e. 1) the use of alternative DOS WINS oil, 2) extension in filter recovery time to 177
hrs, and 3) maintenance of filter transport temperature and post sampling recovery temperature,
are acceptable changes that should not effect the precision and accuracy of the PM2.5 FRM and
will facilitate field operations, improving efficiencies in technician support.
Attachment 4 - Criteria for Designation of Equivalence Methods for Continuous
Surveillance of PMi.5 Ambient Air Quality
Considering the statistical measures of precision, correlation, multiplicative bias, and
additive bias identified for approval of PM2.5 continuous methods, what are the
Subcommittee's views on the usefulness of each measure to ensure that approved or
equivalent methods meet the monitoring network data quality objectives?
Each statistical measure contributes to basic understanding of the performance of equivalent
method under study. It should be noted that observed bias in comparisons may not be an
indictment of the equivalent method, but that of the FRM. Given the expected bias due to FRM
volatile losses, the statistical measure with respect to additive bias should be adjusted
accordingly.
D-36
-------
What are the advantages and disadvantages of using sampler precision and sample
population to help determine the minimum correlation requirement for the approval of
PM2.5 continuous methods? Sampler precision and sample populations are one element in
developing approval O/PM2.5 continuous methods. Establishment of absolute accuracy of the
FRM method must be the ultimate goal and if the techniques to address the additive bias as
discussed in the question above are accomplished, these techniques will prove a useful tool in
evaluating PM2.5 continuous methods.
What are the Subcommittee's views on using a PlVb.s continuous monitor approved as a
FEM, being applicable for use as part of a potential PM2.5 secondary standard for
visibility?
PM2.5 continuous mass monitors can provide viable data for assessing compliance of a
secondary visibility standard if the water content ofPMcan be adequately accounted for. The
R&P FDMS PM2.5 continuous mass monitor shows great promise in the measurement of ambient
PMwith its associated water.
Attachment 5 - Sensitivity of the PMi0-2.5 Data Quality Objectives to Spatially Related
Uncertainties
To what extent have the assessments of spatial variability and the sensitivity of the
DQO process to a variety of population distributions been appropriately addressed?
The techniques applied to address the effects of spatial variability and multi-modal distributions
on PMI 0-2.5 DQO process are reasonable. The assessment findings indicated that for the daily
standard, the performance curves were most sensitive to sampling frequency, followed by data
completeness and population. The effect of multi-modal distributions was observed to be very
small as was the effect of the spatial variability. The question remains as to whether or not PM
volatility and spatial gradients issues have been adequately address in the subject sensitivity
issues.
Question associated with Attachment 6 - PMio-2.s Method Equivalency Development
What are the Subcommittee's views on the approach identified for the development of
criteria to approve continuous PMio-2.s equivalent methods?
The proposed approach is reasonable.
D-37
-------
Dr. Delbert Eatough
Comments on Summary and Rationale lor the I'M,0., 5 FRM
Delbert J, Eatough
15 September 2005
A. Overall.
The rational for the choice of the proposed PM „,_,_, FRM method is well laid out.
Assumptions which are made, and a summary of what the proposed method will and will not
measure are given with reasonable thought. A few comments on assumptions which may not be
valid arc given in the next sections.
B. Semi-volatile Components.
First, the summary does recognize that semi-volatile species will be lost from the collected
particles with the proposed FRM. The recognition that semi-volatile species will not be properly
sampled is a step forward in the science. The assumption is made that semi-volatile losses will be
the same for both the PM1U and PM2J portions of the FRM. The assumption is also made that
losses will be greater with the proposed FRM than with an a It cm ate virtual impactor sampler
because of the higher How rate of the FRM sampler. There is no data that 1 am aware of that
would validate or invalidate the first statement. It is presently untested. There are data which
indicate that the second assumption is not correct.
With respect to the losses of semi-volatile material from a PMMI as compared to a PM, 5
sampler, this assumption has not been directly tested. The great body of data which exist on
sampling for semi-volatile species are limited to the study of the sampling of fine particles. There
is a limited body of older data which suggest that the loss of ammonium nitrate will be
substantially less for the PM,() sampler than for the PM, 5 sampler in some environments. Hence,
for these environments PM,(I.,5 would be overestimated with the FRM. This phenomenon is
apparently associated, at least in part, with the reduction in the loss of ammonium nitrate in fine
aerosols due to the neutralization of some acidic species during the collection of PMIO, PM1(,.25
usually being less acidic than PM, 5 in an urban environment. The reduction in acidity reduces the
loss of ammonium nitrate due to reactions such as ;
NH4NO,(s) + NH4HS04(s) • (NH4),SO4(s) + HNO,(g),
as fine particles of different compositions are mixed on the filter during the sampling process.
This mixing is exacerbated by 24-h sampling, with the possible changes in aerosol composition
over the source of a day. Such chemistry has been suggested, for example, in studies in the Los
Angeles Basin. Similar effects could be observed for semi-volatile organic material where the
volatility of a given compound in fine particles is altered by mixing with absorptive coarse particle-
material during sampling and volatilization is reduced. While this is a very feasible mechanism for
reduction in the loss of semi-volatile organic material on a PMm, compared to a PM,, sample, 1
DJE, 15 Sep2005
D-38
-------
am not aware of any data examining this question.
The assumption is also made that the loss of semi-volatile organic material and ammonium
nitrate from a PM,, collecting filter sampled at a flow rate of 17 L/min (the FRM) will be higher
than for a flow rate of 2 L/min ( a virtual impactor) for the same air mass and collection filter.
This will probably not be the ease based on published data. The question is one of relative
kinetics and whether or not the losses occurring over a 24-h period arc a function of flow rate.
There is some data indicating that losses are a function of flow rate for relatively short sample
collection time periods (hour or less). But we have conducted many studies comparing
undenuded and denuded quartz and Teflon filters over time periods of about an hour up to 24-h.
In all cases, the data indicate losses are about the same. This is even a more stressful test than the
comparison of different sampling rates since the gas phase species are significantly reduced in
concentration. It appears that the loss of semi-volatile organic material is fast enough that for
multi-hour sampling periods, the end result is not dependent on the flow rate (or vapor pressure
of the gas phase compound), e.g, the kinetics of loss are much faster than the sampling time
period. This results is also consistent with rates of loss of organic compounds from denuded
particles measured in chamber experiments by Kamens. If this is the case, semi-volatile losses will
not be sensitive to the filter flow rates for an FRM or FEM.
Incidently, both these observations would tend to reduce the incidence of negative reading
in a difference calculation of mass for either a FRM or virtual impactor based FEM system.
C, Loss of Coarse Particles During Sampling.
The assumption is made that use of a difference method is preferable to use of a virtual
impactor because of the reduced loss of coarse particles in a PM1(I, as compared to a PMM,.,5
collected sample. 1 concur with the arguments associated with that assumption. Incidently, the
effects which reduce particle loss are related to the effects which probably reduce semi-volatile
material loss in the same comparison as discussed in the proceeding section.
D. Automated Semi-Continuous Methods.
Using a continuous sampler as an FEM sampler is fraught with many problem related to
equivalency of the sampling procedures. Included in these problems are the question of semi-
volatile loss in the FEM method. While the sampling time is not a factor in semi-volatile loss for
the FRM, it can be for the FEM. including increased losses due to heating of the sample in the
continuous sampler (e.g, a T'EOM), or reduced volatile losses during sample collection and
analysis (e.g. a GRIMM or FDMS TEOM monitor). The very nature of the FRM sampler will
make developing a true continuous FEM sampler difficult, if not impossible. In addition, forcing a
continuous method to mimic the FRM under some conditions will not guarantee it will mimic
those conditions under all conditions. And finally, such forcing moves us away from the scientific
advantage that can be gained from correctly measuring the semi-volatile paniculate matter in the
more advanced semi-continuous methods. I have additional comments on this point in my review
DJE, 15 Sep2005
D-39
-------
of the Multi-Site Evaluations manuscript. Ultimately, EPA needs to come to gripes with a
balance of regulatory equivalence forced on us by the chosen FRM and the scientific advances
which can be made with the more accurate sampling of fine paniculate (and possible coarse
participate) material with more advanced semi-continuous methods. They have made a start in
this direction in the philosophy outlined in the future national sampling network design which we
have previously reviewed. EPA needs to be certain this vital scientific question is not completely
sacrificed to the choice of an FRM sampler.
Delbcrt J. Eatough
Professor of Chemistry
Brigham Young University
DJE, 15 Scp2005
D-40
-------
Comments on Multi-Site Evaluations of Candidate Methodologies - August 2005 Undated Report
Delbert J, Eatough
15 September 2005
A. Overall.
This report includes both the original results from the 2003 - 2004 studies at Gary,
Riverside and Phoenix (2 studies) plus the results obtained this year in Phoenix with Second-
Generation samplers modified based on the 2003 - 2004 results. Samplers evaluated include the
proposed FRM (a PMl() and PM, 5 difference calculation), an R&P Dichot Sampler, a Beta Gauge,
the APS, a BGI Saturation Sampler, the GRIMM Monitor and a Diehot FDMS TEOM sampler,
In addition to the above, I also requested and obtained from Dr. Vanderpool the results
for PM25 measured with an R&P FDMS TEOM in Gary and Riverside and the PM, 5 data
measured with a Beta Gauge in the Riverside study. My observations on the comparisons of the
various samplers with the FRM results follows,
B. Specific Comments on the Results.
1. FRM Results.
'The results of the study clearly indicate that all PMMJ and PM, 5 samplers by the three
different manufactiirers are equivalent and that the precision among the various instruments is all
such that precise PMlfl., 5 values can be calculated from the collocated results. Based on the
results obtained in the 5 different studies, the decision to use the two FRM samplers to obtain
data with reliable precision for comparison of results with other potential FEM samplers seems
well justified.
2. FRM Compared to Dichot Sampler Results.
The data in this report clarifying that the PMM) on the Dichots should be identical to that
for the PMM, FRM are helpful, compared to the first report. Please verify that is indeed the case
and 1 am not reading anything into the wording. This being the case, differences between the two
samplers must be attributed to difference after the inlet,
It seems to be well established from the data presented that the sequential R&P Dichot has
a problem with retention of coarse particles in the PMM,_, 5 channel. The fact that the problem is
much more apparent in the Arizona data than in the studies at the other two sites is quite
reasonable. Based on the site descriptions, the coarse particles sampled at the Gary and Riverside
sites will include substantial organic material, hi contrast, the Phoenix coarse particles will be
typified by crustal material, dominated by dust from construction projects. It would be expected
that the Phoenix coarse material will be more subject to particle bounce. This problem seems to
DJE, 15 Sep20()5
D-41
-------
be well controlled in the modified single event R&P dichot sampler, based on the 2005 data, but
still not solved in the sequential sampler. Of course, the improvement in results for the single
event Dichot is obtained with some sacrifice in sampler operation simplicity.
However, the data also point to another difference between the Dichot and the FRM
samplers. In the Phoenix data there is a consistent difference in the PM, 5 results obtained with
the two samplers, with the Dichot giving the higher results. Both for the 2003-2004 initial studies
and for the 2005 studies with the single event and sequential Dichots. This effect is not seen in the
data at Gary and Riverside. It seems to me that this is suggestive of a difference in particle size
distribution near the 2,5 cut point. It is expected that the shape of the curve around the 2.5 cut
point is not the same for the two samplers. It can also be expected that the higher concentrations
of crustal material at Phoenix may well lead to a larger tail of coarse material in the aerosol
sampled at the Phoenix site being present in the PM,5 sample. At the time of the last review of
the initial study it was suggested that information on this point could be obtained using the APS
data. After all, these data will give information on the particle size distribution around the 2.5 cut
point for the aerosol present at the different sites. That has still not been done. This is an
important potential factor in explaining the data which must be explored before the precision
between the two samplers can really be established. I urge EPA to inform us on the results of that
analysis.
This suggested analysis (APS exploration of reasons for the difference in the FRM and
Dichot) will be important because if the difference is due to the difference in the curve for the 2.5
WINS and the virtual impactor, EPA must make a decision as to whether this difference is
acceptable, or whether the virtual impactor, as presently configured, is not suitable for a FEM.
3. FRM Compared to Coarse TEOM Results:
There are three important ways in which the Coarse TEOM differs operationally from the
PMlo.,i5 FRM measurement:
1, The inlets are different. The inlet on the TEOM is a modified PM1(1 inlet designed to
provide a comparable cut to the 16.7 1pm with a flow of 50 1pm. Design on this inlet was changed
between the 2003-2004 and the 2005 tests. Furthermore, the inlet was assumed to have 9%
particle losses in the 2005 study, but not in the earlier studies. No validation tests have been
performed on this assumed loss. Nor is the loss as a function of the coarse particle size
distribution known.
2. The PM,, cuts are different. In the FRM the 2.5 cut is made with a WINS, A virtual
impactor is used with the TEOM. Furthermore, the design (and main to major flow ratios) of the
virtual impactor on the TEOM is different from that used in the Dichot. The data suggest that in
Phoenix, sample is preferentially directed to the tine mode in the Dichot, as compared to the FRM
cut. No information is available on this point for the TEOM virtual impactor.
DJE, 15 Sep2005
D-42
-------
3. The collection filter is heated in the TEOM sampler, but not for the FRM samplers.
Loss of semi-volatile material, to the extent it is present in coarse particles, would be expected.
The temperature was changed for the Phoenix 2005 study from 50°C to 40°. However, losses
would still be expected. For example, it has been shown that the reduction of the filter
temperature to 30°C in the PM,5 SES TEOM reduces slightly, but does not eliminate semi-
volatile loss.
The challenge in the interpretation of the Coarse TEOM data is to try and deduce the
relative effect of these three operational differences.
The slope of FRM vs Coarse TEOM PM1(I.2; data for the various studies indicates that the
Coarse TEOM gave a concentration an average of about 25% lower than the FRM. The slope of
the Phoenix 2003 comparison was similar to that of the other two initial studies, but the intercept
was substantial, 13 ug;nr\ It would certainly be informative if the data for these studies were
made available in the report. Is there any explanation for the apparent significant bias in the
Phoenix 2003, but not the other data sets. Examination of the APS data indicated that the cut
point of the Coarse TEOM initial inlet accounted for about 10% lower values. But no indication
is given as to what the other 15% was due to. Is it the effect of the shape for the curve for the
lower cut point (see for example the FRM vs Dichot data)? Is it the effect of loss of semi-
volatiles? Neither of these problems will be corrected with the changes in design of the initial
inlet.
APS data with the modified instrument used in the Phoenix 2005 study indicate the new
cut point in the modified instrument is 10, rather than 9, urn However, an equally important
question is, what does the shape of the curve at the cut point look like compared to the FRM
curve? The Phoenix data give a regression slope of 1.09 for the 2005 Coarse TEOM vs FRM
PM,,,., 5 results. This slope is conveniently played down in the discussion. However, looking
only at ratios docs not give a solid statistical look at the data. The ratio as a function of
concentrations is equally important. It is interesting that the higher slope matches the assume loss
correction. Clearly data on the actual losses as a function of environment are needed if the
instrument is to be designated FEM. The higher concentrations seen for the Coarse TEOM in the
2005 Phoenix study must be due to one of lour things: I. Differences in the upper cut point for
the two instruments being compared. 2. Differences in the lower cut point for the samples
collected, 3. The absence of semi-volatile material for the coarse TEOM in the 2005 study, 4. An
incorrect assumption on losses in the system. Examination of the APS data may shed some light
into the first two possibilities. Clearly, the Coarse TEOM is a promising instrument, but is far
from being ready for FEM designation.
A question not yet tackled by EPA is, at the end of the day if the Coarse TEOM is shown
to have semi-volatile losses different from the FRM, but other than that is OK, what will EPA do?
A converse question coming up is what if the FDMS Coarse TEOM is shown to measure higher
that the FRM due to semi-volatile losses, what will EPA do about FEM designation. This, of
course is a box we are in for consideration of what to do for PM,5 FRM and FEM samplers as the
DJE, 15 Sep2005
D-43
-------
EPA considers the move to semi-continuous monitors,
4. FRM Compared to Beta Gauge Results.
'['he FRM and Beta Gauge I'M,, data do not agree well, with the Beta Gauge data being
consistently higher than the FRM data for all studies. The EPA report attributes this to the
presence of coarse particles in the fine particle mode as a result of the characteristics of the virtual
impactor used in the Beta Gauge, similar to the problem noted above for the Die hot sampler.
However, if this were the case, one would expect PMMI measurements with the two instruments to
be comparable and the PMl(u;5 concentrations to be correspondingly lower. This however, is not
the case. Instead, the PMU).,iS results are in reasonable agreement (if anything, the Beta Gauge
data are still biased a little high) and the PM1(, results arc biased high. This bias is evident in the
slopes near unity and a significant positive intercept for the Gary, and Phoenix 2003, 2004 and
2005 data. Although statistic are not given in a Table in the report for the 2005 data, the results
given in Figures 22-24, indicate the results are comparable to the 2003 and 2004 studies.
An interesting anomaly is the Riverside data, where the PMlt) correlation shows a slope
much greater that unity and a negative intercept. Similar results are also seen for the fine and
coarse correlations. Unfortunately, the data themselves are not given in the report. But Dr.
Vanderpool has provided me with the PM, 5 Beta Gauge results and with the PM,; FDMS TEOM
results for Riverside. Both data sets are biased higher than the FRM. 'The results are also all
highly correlated. The Beta gauge results average 65% higher than the FRM. The FDMS results
average 31% higher. It is possible the Beta Gauge results are higher than both the FRM and the
FDMS TEOM in part because of the shape of the 2.5 cut point for the virtual impactor.
However, all of these results are entirely consistent with that which would be expected if the
sampled aerosols contained measurable concentrations of semi-volatile species.
If it is assumed that the differences between the two measurements is due to aerosol semi-
volatile species, then the conclusion would be reached that the semi-volatile species are highest in
the summer studies in both Phoenix and Riverside, lower in the winter study in Phoenix and
lowest in the early spring study (with much cooler temperatures) in Gary. This is exactly the
trend expected for the formation of secondary semi-volatile fine paniculate material at the various
sites. In this connection, the PM,5 FDMS TEOM results (provided to me by Dr, Vanderpool)
agree with the PM,5 FRM data, consistent with semi-volatile fine paniculate material not being
important in the Gary study.
5. FRM Compared to APS Results.
The key to comparison of the results obtained using this two different samplers is related
to confidence in the uniform density and shape factors used to convert the APS particle count
data to mass. The agreement given in the figures looks in general good. However, 1 do not think
1 can judge if the treatment is really reasonable and unbiased.
DJE. 15 Sep2005
D-44
-------
6. FRM Compared to BCI Saturation Sampler Results
1 consider it most likely that the observed differences between these two samplers is a
reflection of the lower precision of the BG1 because of the lower How rates and lower measured
mass, coupled with a difference in the inlet curves. Are details on the shape of the curve for the
BGI low flow unit available which could indicate if this is true? Given these two considerations,
the comparison is not unreasonable.
7. FRM Compared to GRIMM Monitor.
The GRIMM data provide the second sampler comparison where the possibility exists for
the retention of semi-volatile species in the comparison sampler measurement. The first such
possibility was seen in the Beta Gauge results. While number are not given which will allow a
direct comparison of the Beta gauge and GRIMM sampler results in the 2005 Phoenix study,
comparison of the data in Figures 22 and 29 suggests that reasonable agreement was seen
between these two samples, with both being biased higher than the FRM data by about 35 to
40%.
We have published two reports of measurement of fine participate material with the
GRIMM sampler, one being a summer study and Riverside and the other a winter study at Fresno.
These studies included comparison with Beta gauge measurements (at Fresno) and with several
independent measurements of semi-volatile fine participate matter at both the sites. In both
studies, the GRIMM was biased high than the FRM measurements and the difference was directly
attributable to the presence of semi-volatile fine participate species not measured by the FRM.
8. FRM Compared to the R&P FDMs Dichot TEOM Sampler.
The results for this comparison are not at all expected based on other data reported to the
present with FDMS TEOM samplers. Conversations with EPA and with R&P have lead me to
believe the R&P instruments in this comparison were rushed to the field before they were working
correctly and the data in this section should be completely discounted.
C. Comment on the Possible Implications for Continuous Monitors.
A key focus of the NAAMS is a shift from the use of integrated monitors to continuous monitors
at the Level 11 sampling sites to be included in the NCORE Program of the NAAMS. The
NAAMS document correctly points out in several sections the value of such data in providing
input for use by the scientific community in the understanding of peak exposures, atmospheric
processes and diurnal variations in the atmosphere. These data will all be potentially valuable to
the both the SLTs and the scientific community which will also use the data. Improving public
access to and developing initial interpretation of this data is a key part of the Implementation Plan.
A strong push for moving in this direction is the economic advantage which can be obtained using
continuous monitors, coupled with the increased understanding of atmospheric processes using
DJE, 15 Scp2005
D-45
-------
the data, a potential win-win situation,
As pointed out in the NAAMS, there is also a clash between the monitoring needs of the
NAAMS as outlined in the current regulations which define the fine participate FRM as the basis
of monitoring for attainment and the problems in the FRM which are addressed, but not carefully
considered in the NAAMS. What will and will not be obtained using the new suite of continuous
PM monitors is not considered in the NAAMS, however, the implication is there that the expected
advances which might accrue from the availability of continuous data will accrue. The
Continuous Monitoring Implementation Plan, on the other hand, outlines carefully the protocols
which will be followed as the NAAMS Level II program is implemented, 1 have a great concern
that the philosophy given in the Continuous Monitoring Implementation plan and the objectives
for continuous monitors as outlined in the NAAMS cannot both be achieved.
The Continuous Monitoring Implementation Plan compared FRM and TEOM results
across the country, and noted areas of agreement and disagreement and suggests that aerosol
composition is responsible for the disagreements seen. The plan then proceeds to outline
protocols intended to assure that the continuous monitors to be implemented will be consistent
with data which would have been obtained with an FRM sampler. Substantial research has been
reported since the Version 2 Draft of the Continuous Monitoring Implementation Plan which
sheds additional light on implications of this approach.
EPA is probably correct in identifying the presence of semi-volatile material in fine
paniculate matter as being related to and responsible for the agreement or lack of agreement
between an FRM and a TEOM. We now understand that both nitrate and organic material can
contribute substantially to the SVM. Studies which have obtained FRM and TEOM data, as well
as correctly measuring both nitrate and organic SVM have shown that agreement between the
FRM and TEOM monitors will occur only when either there is no SVM, or when losses of SVM
is comparable for the two monitors. For example, in summer there is a tendency lor both
samplers to loss botli nitrate and semi-volatile organic material and for there to be agreement
between the two in 24-h average data. However, in w inter, the FRM is often higher than the
TEOM because of better retention of SVM. Moreover, on a 24-h comparison basis, there is
usually good correlation in the two data sets over a given season or meteorological condition.
These correlations tend to exist between the FRM and TEOM on a 24-h average even when the
slope of such a comparison is different from unity. These effects probably account for some of
the observed seasonal variations and generally reasonable regression comparisons in the EPA
Continuous Monitoring Implementation Plan report. With such results, there is a temptation to
"correct" the 'TEOM data so that agreement between the two systems is generally seen. While
this is, potentially, an acceptable solution lor monitoring purposes, serious problems arc
introduced with respect to the uses of continuous data as proposed by EPA in the NAAMS.
Even when there is reasonable correlation in 24-h data, comparison of 1-h average TEOM
and more state-of-the art instruments (such as the FDMS TEOM) show quite different diurnal
patterns. This is because the diurnal changes in the chemistry of the atmosphere which lead to
DJE, 15 Sop 2005
D-46
-------
SVM not well measured by the TEOM (and often by the FRM) are averaged out on a day-to-day
24-h comparison. The events which occur leading to significant atmospheric chemistry (and
potential health risk and maximum exposure conditions) occur on a frequency which can be seen
in 1-h data but not the 24-h data. These events are usually missed by a TEOM. Thus, using
modified TEOM data will lead to the worst possible situation, believing we have valid data on
diurnal patterns because of agreement with 24-h FRM data, but completely missing the diurnal
features which will to aid in the understanding of atmospheric processes of importance to
exposure and possible risk. However, using an instrument which will measure SVM (and hence
catch these significant events} will produce data which do not meet the agreement protocols
(particularly the =10 % slope agreement) given in the Continuous Monitoring Implementation
Plan, precisely because the atmosphere is better monitored.
EPA may well be constrained to not use these newer techniques because they will not
agree with the FRM and thus fall monitoring legal requirements, especially in locations where
SVM is important and variable. However, a consequence is that we will be producing data which
give us an inaccurate picture of the atmosphere and thus lead to incorrect decisions based on
continuous monitoring data. EPA needs to find a way around this problem which can be applied
to the various Level 11 monitoring sites, allowing for the advances in understanding which form
one of the major arguments for moving toward continuous monitors. At a minimum, some (if not
many) of the Level II sites should have both a FRM equivalent continuous monitor and a state-of-
the art continuous monitor which allows us to better understand atmospheric processes. The
comparison between the two will give valuable insights on SVM in the atmosphere and aid the
health community in obtaining a better understanding of exposure.
Delbert J. Eatough
Professor of Chemistry
Brigham Young University
DJE, 15 Scp2005
D-47
-------
Comments on the 21-22 CAS AC AAMM Review and Comment Meeting
Delbert J. Eatough
29 September 2005
This will provide my observations on what 1 consider to be the more important points
touched on at the Durham meeting. 1 have previously provided details on my thoughts on:
1. Comments on Summary and Rationale for the PM,,,., 5 FRM.
2, Comments on Multi-Site Evaluations of Candidate Methodologies - August 2005 Undated
Report,
Which were provided before the meeting. 1 believe the points made therein are unchanged by the
input obtained at the meeting and will not repeat those thoughts here. There are a lew items
which seem to me to be key points discussed at the meeting which EPA should consider as the
FRM and associated regulations are put in place for both the PM, 5 and anticipated PMLU_,5
standards.
1. Consideration of the PM25 Cut Point in the Proposed FRM and in Candidate FEM
Samplers Which Measure Both PM2 5 and PMl().,5.
As pointed out in my previous remarks, on the results of the EPA studies covered in
Attachment 2, there is a consistent tendency for potential FEM samplers which use an FRM PM1(I
standard inlet and then a virtual impactor to make the cut between the coarse and line
measurements to agree with the proposed FRM on PMIU, but to have a bias between the too
fractions with the fine being higher and the coarse lower than results obtained with the FRM. The
EPA description of the data talks about the loss of coarse into the fine channel. However,
looking at the total data it seems to me the difference may be due solely to the shape of the cut
point curve for the two PM, 5 separations. Thus for example, this tendency in seen in the Phoenix,
but not the Gary data. EPA can evaluate this possibility using the APS and''or GRIMM size data.
1 have encourage that evaluation in my earlier writcup.
However, there is another point which may be equally important. The argument for using
the proposed difference FRM samplers is the tie of filter pack data to the past health data.
However, the very sharp cut point inlet proposed for the new PM,, FRM is not the same as that
used in the past health studies. In fact, the virtual impactor inlets used by potential FEM samplers
which make both PM,5 and PM|{,_,5 measurements in a single instrument is closer to that which is
used. Thus the only reason for using the proposed difference measurement with the fine FRM
impactor or a very sharp cut cyclone is consistency with the current FRM, not tie to the health
data. I see no reason to tie the hands of potential FEM candidates which can indepenently
measure both fractions based on a rather arbitrary choice of a PM,, cut, tied to what we think is
good science (a sharp cut point curve), but not particularly to the health data. The health science
can certainly allow for either choice. Perhaps EPA should also allow for this latitude of choice in
the requirement for FEM instruments.
D-48
-------
2. Consideration of an FRM or FEM Which Allows for Direct Chemical Speciation.
As was discussed frequently during the meeting, the proposed difference FRM method
does not allow for the collection of a separate PM2.5 and PM1(,_I5 cut for potential chemical
characterization. This is a disadvantage of the prosed FRM which should he further considered
by EPA. Assuming the analysis of the PM,5 and PM1(I cuts colected by the two samplers of the
FRM will provide solid data on the composition of the PMH)_25 cut is fraught with many
assumptions and unknowns. These include the uncertainties in a difference measurement to
obtain the coarse particle composition and the probable alteration in the coarse particle
composition as it is collected with fine material of quite different composition. This concern
further supports the need to make allowance for the use of virtual impactors in ideally an FRM,
but certainly FEMs, recognizing the difference seen between the two samplers is not a measure of
relevance to the health data set on which the standards are based. EPA should not discard the
logical decision to create two FRMs, one based on a difference measurement and one based on
technology such as that present in the dichol. This opens the door for reasonable chemical
speciation data being obtained on the coarse particle traction in the future.
3. Consideration of Allowing for Both a Monitoring and Scientific Use for Data Obtained
with A Continuous FEM Sampler.
We talked about the importance of being sure that the requirements put in place for
certification of a continuous FEM based on comparison with the FRM be loose enough that
samplers which do a better job of measuring the semi-volatile material in collected particles not be
excluded for monitoring use. One possible way to do this is to make the window for comparison
quite large, as discussed at the meting. Another way to approach this is to let those samplers
which can measure both the nonvolatile and the semi-volatile fractions separately be allowed to
use the former for FRM measurement (perhaps treated as fine T'EOM data is now treated) and the
latter for scientific input. This will open the door for the health community to begin asking
questions which are currently not possible because of the limitations of the FRM data.
Instruments which have the capability of providing this input include the FDMS TEOM
instruments and the GRIMM monitor.
There is ample data in the current data set to demonstrate the presence of SVM in the
monitored PM2 5 fraction based on the comparison of proposed FRM and the combination of
FDMS TEOM, GRIMM and Beta Gauge data. These data emphasize the need for consideration
of a dual monitoring and a scientific role for the real-time instruments which EPA would like to
incorporate in the national monitoring network.
There currently is not yet data in the EPA set which answers the question of the presence
of semi-volatile material in the coarse particle fraction. The GRIMM data suggest there may be.
Confirmatory data from the FDMS T'EOM monitor arc not yet available. Hopefully this will be
addressed in the ongoing study.
D-49
-------
Comments on Potential changes to the PM, 5 FRM
Delbert J. Eatough
15 September 2005
A, Overall.
Four changes are being considered to the fine participate FRM method:
I, Replacement of the WINS with a Very Sharp Cut Cyclone,
2. Use of a different WINS oil.
3. Increase in the time allowed for filter recovery.
4, Changes in the transport temperature and post sampling recover}' time.
The reasons for each of these proposed changes are well spelled out and the
recommendations seem reasonable to me.
D-50
-------
Mr. Dirk Felton
Responses: Dirk Felton, NYSDEC (Submitted Sept 16)
Charge to the CASAC AAMM Subcommittee
September 21-22, 2005 Meeting
Questions associated with Attachment 1:
What are the scientific and operational strengths and weaknesses of the PMi0-2.s difference
method relative to other options for a proposed FRM, especially when used as the basis for
the approval of other methods?
The FRM difference method uses the same size selective inlets as the FRM for PM-2.5
and PM-10, the same sampling conditions, the same filters and the same mass
determination. Many of these characteristics of the FRM(s) for PM-2.5 and for PM-10
create differences in how these methods work in various regions of the country. For
instance the volatile components of PM-2.5 are only partially retained on the FRM and
this retained fraction varies seasonally and geographically. It is unlikely that a proposed
automated method would be able to mimic the behavior of both the FRM for PM-2.5 and
PM-10 in all of the expected conditions where it would be required to work and be
consistent with existing FRM measurements.
The principle disadvantage of the FRM difference method is the length of time between
sample collection and data availability. This is primarily a problem if in the future air
monitoring entities are required to produce public health related notices of PMi0-2.s
concentrations in near-real time. The length of time between sampling and data
availability is not really an issue for an FRMi0-2.sthat is primarily intended as a
benchmark for potential equivalent automated methods.
Another often stated disadvantage of the difference method is the expense of servicing
a manual sampler and the associated lab and shipping costs with making gravimetric
mass determinations. It is likely that these costs are minimized because the field staff
required to service the PM-10 sampler would already be assigned to service an existing
PM-2.5 sampler. Other costs such as shipping would be minimized by capitalizing on
the existing PM-2.5 network operation. These costs savings are based on a future
network of PM10-2.s samplers that is similar to or smaller in scope than the existing PM-
2.5 network.
Operationally, a primary advantage of the difference method is that it utilizes existing
technologies that air monitoring agencies are comfortable with and have been able to
attain adequate precision and data availability. Many agencies will also have surplus
samplers if NCORE monitoring initiatives allow a portion of the existing PM2.5 network
to close.
D-51
-------
Another, often overlooked advantage of the difference method is that it utilizes samplers
that do not need an external environmental enclosure. In built up urban areas it is
difficult to site equipment that needs both a temperature controlled environment and an
inlet suitable for aerosols that meets monitoring siting criteria.
Based on the field study report as well as any other available data, e.g. data from
other State and local agencies, how does the demonstrated data quality of the
PMw-2.5 difference method support or detract from it being proposed as a FRM?
Many of the original complaints about difference data came from early data analysis that
used high volume PM-10 FRM data and PM-2.5 FRM data. This is not appropriate
because the high volume samples include less of the volatile material that is included as
part of the PM-2.5 FRM measurement. Subtracting PM-2.5 FRM data from high volume
PM-10 data results in biased and often negative Coarse mass determinations.
The proposed PMi0-2.5 difference method utilizing collocated identical samplers one with
the WINs impactor removed, does produce robust measurements suitable for use as an
FRM. The New York State Department of Environmental Conservation (NYSDEC) has
operated PMi0-2.5 difference method samplers in Manhattan and in Niagara Falls for
more than three years. The following data summary includes data from 2002 through
2004 on a one day in three schedule. The high values apparent from the "Max" row
were due to the smoke from Canadian wildfires.
Max
Min
Average
Std Dev
Manhattan, NYC (344 Samples: 94°/
PM-2.5 PM-10 Coarse Ratio
82.71 88.79 26.71 0.93
3.96 6.17 1.71 0.22
15.41 25.48 10.07 0.59
9.31 11.98 4.73 0.12
-------
The NYSDEC PMio-2.s sampling program does not include a precision sampler but the overall
results seem to be similar to what EPA ORD found in Birmingham in 2003 - 2004. This is not
surprising since the Ratios of PM2.5/PM10 are similar. It is more reassuring that the results
from the Multi-Site Evaluation in 2003 - 2005 demonstrated acceptable precision in each area
and season where the PMi0-2.5 difference method was evaluated.
100
90
Ratio of (PM2.5/PM-10) Sorted 2002 - 2004 1/3 Day
PM-2.5 FRM and LowVol PM-10
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Ratio (PM-2.5/PM-10)
0.8
0.9
1.0
• Manhattan, NYC
Niagara Falls, NY
D-53
-------
Coarse: (PM-10 - PM-2.5), Low Vol PM-10,
PM-2.5 FRM Sorted 2002-2004 Manhattan, NYC
100
0 5 10 15 20 25 30 35 40 45 50 55 60
40
30
20
10
.Coarse PM-10 PM-2.5
Coarse: (PM-10 - PM-2.5), Low Vol PM-10,
PM-2.5 FRM Sorted 2002-2004 Niagara Falls, NY
&
0 5 10 15 20 25 30 35 40 45 50 55 60
30
20
10
.Coarse PM-10 PM-2.5
D-54
-------
Question associated with Attachment 2:
Based upon the latest available field study data, which PM? 0-2.5 methods have both
sufficient utility to meet one or more important monitoring objectives and
appropriate data quality to be considered for deployment as Federal Equivalent
Methods (FEMs) or speciation samplers in a potential PM? 0-2.5 monitoring
network?
Currently it looks like the Manual Dicot and perhaps the re-designed sequential Dicots
have the potential to be an FEM with their current configurations. The fact that a portion
of the Coarse mode particles are deposited on the fine filter could cause some minor
geographic and seasonal discrepancies in comparisons with the FRM difference
method. The Dicot methods also seem to suffer somewhat from the loss of Coarse
mode particles on Teflon filters. If filters could be substituted that are more suited to
retaining the Coarse mode particles without other artifact issues than the method could
be further improved.
Running Dicot FEMs may not be much of an advantage to air monitoring Agencies
versus running two collocated FRM samplers. The Dicot PM2.5 data may not be good
enough to replace the instrument used for FRM PM-2.5 monitoring and the cost of filter
and equipment maintenance for a Dicot is nearly identical to that of two stand alone
instruments.
Questions associated with Attachment 3:
What are the Sub-committee's views on the Very Sharp Cut Cyclone (VSCC) being approved as
an alternative second-stage impactor to the WINS for use on a PM-2.5 FRM?
The attachment did not include the data showing the comparisons between the WINS
and the VSCC nor any indications that there were regional or seasonal differences
between the two inlets. This data must have been quite convincing as the VSCC has
already been designated an FEM. The NYSDEC evaluated a SCC collocated with a
WINS in 1999 in NYC over several months. The regression results (SCC = 1.017 FRM
- 0.174 and R2 = 0.994) showed that at least in New York City, a well maintained SCC
could justifiably be used in place of a WINS impactor.
The advantages of the VSCC over the WINS and the similarity of the resulting data
support the decision to approve the use of the VSSC in an FRM. The caveat is that
even though the VSCC has a longer service interval than the WINS, a neglected VSCC
can cause problems with data integrity. The WINS must be serviced after five 24-Hr
runs. This service includes cleaning, filter exchange, re-oiling and inspection of the
inlet. If there is a problem found during the service of the WINS such as an obstruction
in the flow path (spider or fluff) or a leak due to a loose fitting or worn o-ring, no more
than 5 sample days have to be invalidated. If a problem is found with a VSSC during a
less frequent service, a greater amount of data may have to be invalidated.
D-55
-------
A minimum service interval should be specified for the VSCC as part of the approval
designation. A service interval of Monthly or after every ten 24-Hr sample days would
be a reasonable compromise between the necessary cleaning frequency and the
amount of data that would have to be invalidated if there was a problem. To reduce the
burden on the agencies performing the work, this VSCC service interval also coincides
with when the operator would normally perform instrument audits.
The cost of a VSCC is significant and monitoring Agencies that opt not to switch inlets
should not be penalized for using a WINs. This should not be a problem as long as
there is no requirement for collocating a FRM and an FEM. Allowing both the WINs and
the VSCC to be an FRM would eliminate this potential problem.
To what extent are the stated advantages of relaxing existing requirements identified for the PM-
2.5 FRM supported by the information cited in attachment 3, available literature, or good field
and laboratory practices? Does the Subcommittee have additional recommendations for the
PM2.5 FRM that would be neutral with respect to bias, but would improve the performance and
minimize the burden on agencies conducting the sampling?
Recommendation number 2 (WINs oil) is fine but it needs a reference to the original
requirement and description for the oil in the WINs impactor. Presumably the new oil
also has a low vapor pressure and will not interfere with analyses other than gravimetric
that may be performed on the filters.
Recommendation number 3 is fine and the supporting draft study performed by
members of the EPA and several air monitoring Agencies justifies the change in filter
recovery time. The one caveat is that there is significant negative bias at two sites that
may be related to the composition of the aerosol at those locations. For that reason,
there is not enough information to go ahead with recommendation number 4. A study
should be designed to look for significant negative or positive bias resulting from the
change in the temperature requirements prior to final weighing of the filters. This study
should include several geographic areas and all seasons to insure that no areas of the
country would have data that become biased high or low once the changes to the
regulations were enacted.
Attachment 3 has a summary that in part states that these recommendations "would
provide more consistency with other filter-based networks such as IMPROVE". This
statement particularly with respect to recommendation 4 is not entirely accurate. The
IMPROVE program has inconsistent filter handling at many of their monitoring locations.
Some IMPROVE filters are stored for 30 days in enclosed creosote coated plywood
buildings while at other sites the filters are stored in air conditioned labs. This type of
inconsistency will become more apparent if IMPROVE is used in more urban monitoring
locations.
D-56
-------
Dr. Philip Hopke
Comments by
Philip K, Hopke
Question 1: What arc the scientific and operational strengths and weaknesses of the /""A/,,,.,,
difference method relative to other options for a proposed FRM, especially when used as the
basis for approval of other methods?
I am very disappointed by the effort to keep pushing the difference method as an FRM, We have
had several prior discussion about the difference method and the committees involved in those
reviews have consistently indicated their dissatisfaction with the difference method as the basis of
an FRM We understood the potential necessity for it when we were looking at the possibility of
a standard being promulgated 3 years ago, but there were serious concerns and there remain
serious concerns regarding the chemical interactions between PMH) and PM,, and not having a
sample of PM(10_, 5| available for chemical or other analysis. The claim is that the paired sampler
provides useful samples for chemical analysis, in fact, the combination of the PM]0 and the PM25
make the chemical analysis more difficult. There is the claim that they need the PM, 5 to keep the
PM(]0_2 s( t° sta>' on tne filter. However, there are other approaches that can be used to affix the
coarse particles to the sampler such as small quantities of fluorocarbon or silicon grease. This
appears to be a case of having made up your minds as to the sampling approach and doing what
you can to prove it is what you wanted. It is simply not possible to support this proposal. It is
not the best monitoring science and represents a major step backward.
I would suggest that it will be quite possible to provide a set of performance specification that
require a continuous monitoring system that provides adequate separation of the fine and coarse
particles that there is no need for a subtraction correction to be made. It is not necessary to
specify any specific technology to make this separation and/or enrichment of the coarse particles
nor the measurement technology. Only the precision of the measurements needs to be specified.
It is clear 1 am in the minority on this issue, so 1 would suggest that the Subcommittee at least
recommend that the modem dichotomous sampler be approved as a second FRM. It is
unfortunate that EPA has not even developed a simple two-stage cascade impactor to be the
FRM. It would provide clean samples of coarse and fine particles at a standard flow rate of 16.7
LPM. One could have easily designed, built and tested such a unit within the funds that have
been expended for testing commercial units. A major opportunity has been lost to move away
from the difference method and still provide integrated samples for analysis. I would prefer the
performance standards for a continuous monitor as an FRM, but at least require a sampler that
actually collects coarse particles!
The logic of suggestion that more finely time resolved data is not needed escapes me. If there had
been no PM, 5 data before there was a standard, there would have been no basis for creation of
the PM, 5 NAAQS, Without complete and time resolved data, we are never going to be able to
sort out the time course of the effects of PM on health. The proposed FRM again only provide
data on 33% of the days and only in 24 hour increments. Yes, it permits testing attainment of the
D-57
-------
NAAQS, but we arc quite certain that the current NAAQS has significant flaws because we do
not yet really understand the relationship between exposure and adverse health effects. With
complete data that covers a much larger percentage of the days of the year and provides time
resolved data, we will not be able to better understand the time course of the onset of the adverse
cHeels and the useful insights into potential mechanisms that such an understanding will provide.
Question 2: Based on the field study report as well as anv other available data e.g., data from
Stale and local agencies, how does the demonstrated data quality of the PM,,,_-, 5 difference
method support or detract from it being proposed as an FRM?
The test in Jefferson County, Alabama does not really tell us very much about the ability of the
average SLT agency to operate paired samplers. All of the samplers were the relatively limited
use BG1 samplers rather than the more commonly employed MetOne or more automated samplers
like the R&P where there is more movement of the filters into and out of the sampling position.
Thus, it is not clear that this is a meaningful reflection of the ability of the likely endusers to
produce non-negative data.
A key part of the past tests that have been made should be the careful examination of the chemical
composition data that have been accumulated from the instrument intcrcomparison field studies.
Many of the samples have been analyzed and the data provided to the subcommittee, but there are
now filters from the latest field campaigns that need to be analyzed and more importantly, time
needs to be set aside to examine and interpret these results. This should be done before going
ahead with designation of the difference method so that one can really see if the samples can
actually be used for chemical characterization of the trace constituents of the coarse mode
particles.
Consultation Question I: Based upon the latest available field study data, which PM 10-2.5
methods have both sufficient utility to meet one or more important monitoring objectives and
appropriate data quality to be considered for deployment as Federal Equivalent Methods
(FEMsJ or speciation samplers in a potential PMia-2t monitoring network?
The performance of the continuous monitors in the latest trials was actually quite encouraging. It
is clear that several of the units would have the chance to pass a set of rigorous performance
standards that could be set for an FRM. Thus, it suggests that a more forward thinking FRM
could be designated to provide the data we need to begin to really understand the health effects of
PM,
'HO-2.5)
Consultation Question 2; What are the Subcommittee's vie\vs on the Very Sharp Cut Cyclone
(ySCC) being approved as an alternate, second-stage impactor to the Well linpactor Ninetv-Six
(WINS) for use on a PM,< FRM?
This is a positive step. The problems identified with the WINS in terms of maintenance issues are
clear and the small loss in the sharpness of the cut relative to the ease of operation of the VSC'C
relative to the WINS, clearly is an improvement.
D-58
-------
Consultation Question 3: To what extent are the slated advantages of relaxing existing
requirements identified for the PM? ? FRM supported hy the information cited in Attachment 3,
available literature, or good field and lahoratoiy practices? Does the Subcommittee have
additional recommendations for the PM:.> FRM that would he neutral with respect to bias, but
would improve the performance and minimize the burden on agencies conducting the sampling?
There is the critical problem remaining with respect to defining what is to be measured. We know
that the PM, 5 FRM tails to accurate measure the actual airborne PM mass with substantial loss of
semivolatile species such as nitrate and organics. It would be useful to make more accurate
measurements of the actual PM2; mass as has been demonstrated to be possible using the current
state-of-the-art systems such as the R&P filter dynamics measurement system (FDMS). It is time
to move beyond the past and get into measurements that better represent what is actually in the
air and with the completeness and time resolution to permit a better understanding of the
relationships between PM,5 mass and health effects.
The DQO tool for examining the performance of continuous monitors as FEMs for both fine and
coarse should be modified to include an asymmetric multiplicative "bias" factor. It is unfortunate
to call the slope above zero to be "bias" since it implies that the current flawed FRMs are accurate
which has been clearly demonstrated to be untrue. However, to permit more accurate
measurements to be included that will produce larger mass concentration values, an asymmetric
interval would permit better measurement tools to qualify as FEMs. Given the past performance,
it would probably be wise to couple the wider range of positive multiplicative factors with a
narrower range of additive factors. It has been seen in the supersite and other studies that
instruments like the FDMS have an intercept that is close to zero when compared to FRM mass
values and thus, we should not permit as wide a range of additive "bias" terms for the larger
positive side multiplicative "bias" side of the compliance parallelogram. It would make sense to
look at the sites where there are co-located FRM and FDMS systems to explore an appropriate
upper bound on the multiplicative bias term. It would be useful to permit the monitoring agency
to either report the best estimate of mass (including the semivolatilcs) as well as the base mass
values. The base mass values will more closely agree with the FRM values and may be more
acceptable to the local authorities who may not want to use the higher mass values that arise from
a more accurate measurement of the airborne paniculate matter mass. However, although they
can use the base mass measurements for attainment determinations, they should be required to
report the more accurate mass measurements so that they are available for other scientific
purposes.
In terms of site choices for testing equivalency, it may be good to look at a matrix of effects to
which one wants to challenge the samplers.
High PM(10.,.J/PM2.J
Low PM(]0_, .s,/PM,j
High semivolatiles
Low semivolatiles
D-59
-------
Dr. Rudolf Husar
Comments on
Particle Methods and Data Quality Objectives
by Rudolf Husar, Washington University, Sept 18, 2005
ATTACH i: Summary and Rationale for the PM10-2.5 FRM
For the accurate and reliable determination of daily average coarse mass concentration the
suggested PM10-PM25 difference method is quite compelling. The problem lies in the marginal
utility of the PM10-PM25 difference method for estimating the potential health effects, visibility
effects or for that matter learning more about the nature of coarse PM. Without speciation, it is
virtually impossible to separate the small fraction of anthropogenic coarse PM from the bulk of
benign, mostly natural and uncontrollable soil dust. Clearly, the difference method is not suitable
for coarse PM speciation analysis.
A closer coordination with the NAAQS - setting processes would seem highly beneficial. In
particular, it would seem logical to recommend the relevant FRM after the gross features of the
standard have been set.
ATTACH 2: Evaluations of Candidate Methodologies for Coarse PM
The field studies for the evaluation of candidate coarse PM methodologies constitute the pillar of
this preparatory activity prior to the promulgation of a standard. The multi-site, multi instrument,
multi season and multi-stage (repeated after feedback) comparison was commendable. The
presentation of results is clear and useful.
ATTACH 4: Equivalence Criteria for Continuous PM25
In addition to the statistical intercomparison measures given in the report, it would be helpful to
incorporate and weigh qualitative differences, between the methods, e.g. unattended
operation/operating cost; availability of filters for subsequent speciation analysis etc. The idea is
to make the methods-intercomparison as complete as possible. That way, methods-evaluations
and decisions are made using compatible metrics.
ATTACH 5: DQO Sensitivity to Spatial Uncertainties
The spatial uncertainties of the DQO remain to be a weak part of the analysis package. In
response to the CASAC Subcommittee review, the spatio-temporal model has been updated with
a few cosmetic changes but the procedures are still inadquate.
A well designed and tested model of cPM pattern could be of great utility for standard setting,
network design and general DQO analysis. However, the basic approach of estimating the role
of different design/operational parameters on the uncertainty is weak. In fact, it leads the authors
D-60
-------
to the unreasonable conclusion that spatial variation of coarse PM is of marginal importance
compared to the other factors. This is hardly defendable, particularly in urban settings with
strong spatial texture of both cPM source and transport pattern.
- The model is still based on an untested set of assumptions about the aerosol pattern.
Verification of the model with historical data at a few characteristic sites was probably done
by the model developers, but for reasons unknown, they were not shared in the summary.
- A particularly poor assumption is that there is a mean and a constant Coefficient of
Variation (CV). The coarse particle concentration pattern is highly episodic. Short-term but
rare high concentration events are the norm at most locations and seasons. Such pattern can
not be adequately modeled by a normal or even log-normal distribution.
- The inclusion of spatial variability into the model using Design Values, is obscure at best.
Why not looking at spatial variations and spatial correlations using actual observations that
are available in many cities.
Finally, the conclusion that the spatial variability of cPM is much less significant than sampling
frequency, data completeness, the CV is unreasonable. Looking at any urban cPM pattern (the
real data from actual monitors, not a model) shows that cPM is spatially heterogeneous and has a
large 'gray zone'. A possible reason that sampling frequency was found to be a major
contributor to uncertainty is the input assumption of the (long) 1 day or 3 day sampling.
Continuous monitors virtually eliminate the temporal uncertainty at a monitoring site, so the
remaining uncertainty is dominated by the spatial texture and the other factors listed
ATTACH 6: PM10-2.5 Methods Equivalency Development
The evaluation of equivalency through dedicated inter-comparison of instruments is a sound
beginning. In addition to the statistical intercomparison measures given in the report, it would be
helpful to incorporate and weigh qualitative differences, between the methods, e.g. unattended
operation/operating cost; availability of filters for subsequent speciation analysis etc. The idea
is to make the methods-intercomparison as complete as possible. That way, methods-evaluations
and decisions are made using compatible metrics.
D-61
-------
Dr. Kazuhiko Ito
Comments on Particle Methods and Data Quality Objectives.
9/30/05
Kazuhiko Ito, NYU
The implication of the letter from the CASAC PM Review Panel
Before responding to the charge question, I need to briefly discuss the implication of the letter
from the CASAC PM Review Panel (distributed during the review meeting but dated September
15, 2005; EPA-SAB-CASAC-05-012) on the process of establishing PMi0-2.5 FRM methods and
planning. After reading this letter, it became clearer to me that, while the toxicity of the rural
coarse particles remains to be determined, the focus of the coarse particle standard appears to be
the "urban" PMio-2.5, or UPMio-2.5, and the great emphasis will be placed on identification of the
compositions of UPMio-2.5. For example, the letter says:
".. .and there is a need for more data that relate the composition of the paniculate matter to adverse health
effects. We anticipate that future coarse- and fine-mode paniculate standards will give greater weight to
paniculate composition as a critical element in defining the nsk of adverse health effects. Data are needed
on ambient concentrations in each size range in terms of mass concentrations and speciation. Continuous
monitors for mass, as well as for key components or source-related tracers, will provide the best and most
cost-effective means of collecting such data for both epidemiologic research and compliance monitoring.
..." (from 1st paragraph on page 3, EPA-SAB-CASAC-05-012)
Thus, the planned FRM and FEM methods need to accommodate the need to collect speciation
data (or some specific component of PMio-2.s). With this emphasis plus some of the comments I
heard during the meeting, I revised some of my initial answers to the charge questions.
I also learned, from the EPA presentation during the meeting, a PMi0-2.5 network design "similar
in concept to PM2.5 monitoring for daily standard is being considered" by the EPA staff. I think
it is very important to start considering the network design for PMi0-2.5 chemical speciation data
now. Though we don't have data from multiple monitors' PMio-2.5 chemical speciation monitors
yet, we may be able to at least develop a conceptual framework for the PMio-2.5 chemical
speciation monitoring based on what we learn from the PM2.s chemical speciation data collected
from multiple monitors within cities so far. I have learned, based on the data from three PM2 5
chemical speciation monitors in New York City, that the extent of spatial correlation varies
across species (Ito K, Xue N, Thurston GD. Spatial variation of PM2.5 chemical species and
source-apportioned mass concentrations in New York City. Atmospheric Environment, 2004; 38:
5269-5282). During the meeting the EPA asked for the sub-committee members' opinions on
the candidate cities where more test PMio-2.5 data would be collected. I suggest that EPA collect
more PMio-2.5 data using the candidate methods (and trial PMio-2.5 speciation data, if possible at
all) in the cities where multiple PM2.s speciation monitors are collecting data, so that some
relationship between spatial variation of PM2 5 vs. PMio-2.5 could be examined and applied to
future PMio-2.5 chemical speciation monitoring network design.
D-62
-------
Questions associated with Attachment 1 - Selection and technical summary ofPMio-2.sFRM:
1. What are the scientific and operational strengths and weaknesses of the PMio-2.5 difference
method relative to other options for a proposed FRM, especially when used as the basis for
approval of other methods?
Strength:
Since the existing FRM PMio and PM2.5 are also measured using the same principle, samplers,
and operating procedures, there is continuity in interpreting/comparing the past PMio, PM2.s data
with the future PMio-2.5- Its filter samples should accommodate subsequent size-specific
chemical analysis (PMio-2.5 and PM2.5 speciation). The multi-site evaluations of the difference
methods using the samplers from different manufacturers showed very high precision, which is
promising.
Weakness:
If the operational and maintenance cost for the difference method samplers (requiring two
samplers and daily changing of filters for mass measurement alone) is much higher than those
for the automated continuous and semi-continuous samplers, the cost can be a weakness. This is
particularly so if spatial non-uniformity of PMio-2.5 within a city requires more monitors for
PMio-2.5 (than for PM2.5).
There are at least two issues with the difference method that were raised by the sub-committee
members during the meeting. The first one was the fact that even the current FRM method for
PM2.5 has the potential loss of volatile compounds, and the proposed difference method would
have the same problem. I think the seriousness of this issue would depend on: (1) how much of
the volatile compounds we are missing from the samples, and (2) the health effects of the volatile
fraction (mostly nitrate, I imagine). I am not aware of studies that identified nitrate as important
component of PM that are associated with health outcomes, but then again few studies had
available data on nitrate. The second issue with the difference method is, for chemical speciation
purpose, collecting particles on the PMio filter would be potentially chemically mixing PMio-2.5
and PM2.5 species on the PMio filter. Thus, subtracting the speciated PM2.5 data from the
speciated PMio data may not give us adequate speciated PMio-2.5 data that would represent the
PMio-2.5 chemical components in actual ambient air. I am not familiar with concrete examples of
this issue, but this may need to be examined in the future testing.
My overall impression is that the difference method was reasonable for FRM, but considering
the expected "large" (compared to the precision/accuracy of the alternative monitoring
instruments) spatial variation of PMio-2.5 within a city, we may need to consider putting multiple,
cheaper, FEM's in addition to one or two FRM's within a city.
2. Based on the field study report as well as any other available data, e.g., data from State and
local agencies, how does the demonstrated data quality of the PMio-2.5 difference method
support or detract from it being proposed as a FRM?
From the "data user" point of view (for epidemiological studies), the difference method has the
desired very high precision and continuity to the past and current PM2.s FRM data. Since the
D-63
-------
variation due to spatial heterogeneity is a far bigger concern to me than the precision of the
proposed samplers, I would be satisfied with the proposed difference method FRM. The study in
Birmingham, Al seems to suggest that PM2.5 is more uniformly distributed than PMio-2.5 (37%
mean level difference for PM2.5 vs. a factor of three difference for PMio-2.s).
Consultation Questions:
Question associated with Attachment 2 — EPA's Multi-Site Evaluations of Candidate
Methodologies for Determining Coarse Particulate Matter (PMio-2.s) Concentrations:
August 2005 Updated Report Regarding Second-Generation and New PMi0-2.5 Samplers:
1. Eased upon the latest available field study data, which PMio-2.5 methods have both
sufficient utility to meet one or more important monitoring objectives and appropriate data
quality to be considered for deployment as Federal Equivalent Methods (FEMs) or speciation
samplers in a potential PMio-2.5 monitoring network?
With so many instruments and field study results (and not being an "instrument" expert
myself), I could not compare the pros and cons of these methods by just reading Attachment 2.
The way the results were presented for these samplers were not always the same across the
sampling campaigns. Therefore, I constructed a table below, extracting the ratios of each
instrument to FRM, R2 from regression of each method's values on those of FRM, and
statements to summarize the results for myself. The R2's of these methods were mostly very
high (> 0.9), except the methods that measure size distributions, and the main issue appears to be
the constant bias (over- or under-estimation). However, based on the 2003 and 2004 studies, this
"constant bias" appears to also vary across locations (regions) for a given method, perhaps due to
the regional difference in chemical compositions or size distributions. Thus, the alternative
sampler to FRM mean ratios obtained in the 2005 Phoenix, AZ study may be "snapshots" and
these may also vary, and we don't have the data on these region-specific variations for the newer
instruments introduced in the 2005 study. In this situation, it may be necessary to consider site-
specific (or city- or region-specific) calibration of FEM samplers to FRM samplers. Obviously,
since the apparent objectives of these samplers vary (e.g., near real-time measurement, size
distribution measurement, speciation), the choice of methods will need to be discussed for each
objective.
For routine FEM (and possibly for speciation) purposes, I could not see major differences
among the R&P dichot (sequential, sequential/manual mode, single-event) and Sierra-Andersen
dichot samplers. Since the discrepancy between the FEM and FRM samplers may be region
specific, it seems necessary to co-locate the FRM and FEM samplers in each region, at least
initially. If the operational cost for the FEM samplers were significantly lower than that for the
FRM samplers, then the FEM samplers may be used to measure the spatial variability of PMio-2.5
in the city of interest, at least initially, and reduce the number of such monitors as appropriate.
Three methods (Kimoto dichot beta gauge, R&P Coarse TEOM, and R&P dichot TEOM)
are available for near real-time mass measurements. The Kimoto dichot beta gauge showed a
major over-estimation of PM2.5 even after design modification, whereas the R&P dichot TEOM
sampler showed a major under-estimation of PM2.5, with many negative values. Thus, these
monitors may require further modifications. However, practically speaking, we may not need
D-64
-------
near real-time dichot monitors that accurately measure PM2.5 as long as a PM2.5 TEOM monitor
is running in the area (note that PM2 5 is expected to be more spatially uniform and therefore we
do not need as many PM2 5 monitors as we do PMio-2.s monitors). In that sense, the R&P Coarse
TEOM monitor may be a convenient choice for monitoring short-term excursions as well as
spatial variations (if the cost is low enough so that multiple coarse TEOM monitors could be
operated).
There were two methods (TSI APS and Grimm EnviroCheck) that measure size
distributions in real time. Both seem to suffer from R2's that are lower than those for the other
methods, and may require further modifications.
So far, we are dealing with accuracy and precision in terms of PMi0.2.5 mass
concentrations. We may also have to start thinking about the issues associated with chemical
speciation of PMi0.2.5 data.
Table 1. Cursory comparison of various methods from the field study results
Sampler
Collocated PM
and PM FRM
Samplers
R&P Model
2025 Sequential
Dichotomous
Sampler
Kimoto Inc.
Model SPM-
613D
Dichotomous
Beta Gauge
R&P
Continuous
Coarse TEOM
Monitor
TSI Inc. Model
3321
Aerodynamic
Particle Sizer
(APS)
R&P Single-
Event
Main purpose
FRM, speciation
Unattended
multi-day
operation
possible with a
filter exchange
system
Near real-time
measurement
Near real-time
measurement of
PM10-2.5
Size distribution
of particles
(larger than >
0.7 |im) in real
time
Unlike R&P
Model 2025 ,
Comparison of PM10_2.5 with
FRM (the ratio to FRM and R2)
NA
Small but consistently under-
measured PM10.2.5 in 2003 test
R&P to FRM ratio = 0.80 to
0.96; 0.89 in 2004 test; 0.93 in
2005 test. R2 ranged from 0.968
to 0.979.
The ratios of Kimoto to FRM
for PMio-2.5, the ratios ranged
from 0.91 to 1.08 for 2003 and
2004 tests; after modification,
the ratios ranged ~1 .05 to 1.13
in 2005 tests; R ranged from
0.957 to 0.995.
The ratios to the TEOM to FRM
mostly low but varied across
sites ranging 0.69 to 1.05 in
2003 and 2004 tests; but design
modification appeared to have
improved the ratio (~ 1 .04) in
2005 tests. R2 ranged from 0.926
to 0.999.
The 2004 tests showed a factor
of two under-prediction with
TSI to FRM ratios, but using an
alternative specific gravity and a
shape factor, the ratio are now
much better (0.76 to 1.02). The
2005 tests with design
modification resulted in the ratio
of -0.86 without invalidated
data. R ranged from 0.53 to
0.99.
R&P to FRM ratio = 0.99 in
2005 tests. R2 = 0.995.
Other comments
High precision; high data
capture rate/ few functional
problems
Some operational problems in
the field tests; high precision;
some intrusion of coarse into the
fine channel; The R&P to FRM
ratio for PM2.5 ranged 1.00 to
1.08.
Consistent overestimation for
PM2 5. The ratios of Kimoto to
FRM for PM2.5 were
consistently high (1.26 to 1.70)
in 2003 and 2004 tests
Few operational problems
Assumption of the specific
gravity and shape factor makes a
big difference in results. There
appear to be some functionality
problems (to be tested in the
next tests in 2005).
Not for routine use?
D-65
-------
Dichotomous
Sampler
Sierra-
Andersen
Model 241
Dichotomous
Sampler
BGI frmOMNI
Ambient Air
Sampler (Filter
Reference
Method)
Grimm
EnviroCheck
Model 1.1 07
Sampler
R&P
Dichotomous
TEOM Sampler
the potential for
post- sampling
loss of large
particles is
minimized
Routine
monitoring
For short-term
saturation
sampling at a
relatively low
cost.
Size distribution
in real time.
Near real-time
measurement of
PM10_2 5 and
PM2.5
Sierra-Andersen to FRM ratio =
0.95 in 2004 tests. R2 = 0.995.
OMINI to FRM ratio for PM10.
2.5 ~0. 85. R2 = 0.949.
The ratios for PM10_2 5 averaged
-1.53. R2 = 0.847.
Mean dichot to FRM ratios for
PMio-2.5 were 0.85 and 0.89 for
two units used. R2 = 0.992 for
the average of two units vs.
FRM.
The inlet has been fully wind
tunnel evaluated. Some intrusion
of coarse into the fine channel;
No active volumetric control
Not for routine use? Some
functionality problems in 2005
tests, reducing the data capture
rate. OMINI to FRM ratio for
PM2.5~1.07% (with somewhat
low R2 = 0.808).
Grimm to FRM ratios for PM2 5
averaged -1.37.
No operational problems during
the 2005 tests; negative PM2 5
values; Mean dichot to FRM
ratios for PM2 5 were 0.80 and
0.63 for two units used.
Questions associated with Attachment 3 - Memo to PMNAAQS Review Docket (OAR-
2001-0017) — Potential changes being evaluated for the PM2.5 Federal Reference Method
2. What are the Subcommittee's views on the Very Sharp Cut Cyclone (VSCC) being approved
as an alternative second-stage impactor to the Well Impactor Ninety-Six (WINS) for use on a
PM25FRM?
I am not familiar enough with the background information and support data for VSCC vs. WINS
performance to form an opinion on this.
3. To what extent are the stated advantages of relaxing existing requirements identified for the
PM2.5 FRM supported by the information cited in Attachment 3, available literature, or good
field and laboratory practices? Does the Subcommittee have additional recommendations for
the PM2. s FRM that would be neutral with respect to bias, but would improve the performance
and minimize the burden on agencies conducting the sampling?
The justifications explained in Attachment 3 all seem reasonable to me, but since I did not read
the original references cited, I refrain from commenting on this.
Question associated with Attachment 5 - Sensitivity of the PM 10-2.5 Data Quality
Objectives to Spatially Related Uncertainties
7. To what extent have the assessments of spatial variability and the sensitivity of the DQO
process to a variety of population distributions been appropriately addressed?
The document states "The DQO development used preliminary data collected from sites
D-66
-------
providing coarse particulate estimates from around the country as well as data from multi-site
performance evaluations..." (page 1), but the document does not give us the sense of what a
typical (or any example) spatial distribution of PMio-2.5 would be like. For example, were there
cases in the database in which PMi0-2.5 variation could be depicted in as small as an 8 km x 8 km
grid (like Figure 1 on page 4)? One of the documents distributed for this meeting (Network
Operation of the Difference Method: An Independent Study Conducted by the Jefferson County
Department of Health In Birmingham, AL, by Vanderpool and Dillard) shows spatial variations
of PMio-2.5 as measured with seven monitors in Jefferson County, which could be contained in a
50 km x 50 km grid. The data showed that there was a factor of 4 difference in the mean PMio-2.5
concentrations across the monitors during 2004. Given a case like this, my question would be:
how many monitors are needed to estimate the distribution of ambient PMio-2.5 levels the
residents in this county are exposed to? I must be misunderstanding the intended use of the
DQO model, but the question being asked in Attachment 5 seems to be using somewhat
unrealistic scenario, though I am not sure if using more realistic scenarios would make any
difference. For example, Attachment 5 states,
"Hence, the comparison is indicating how well a single monitor does in predicting the
true mean design value across the grid area. Since the day-to-day shape of the surface is
not fixed, on average, throughout the three-year period, the center should be an unbiased
indicator of the mean. Consequently, there is no inherent bias at any site being simulated,
unless a strong autocorrelation is used to "fix" the shape of the surface." (page 4)
This (that the center should be an unbiased indicator of the mean) seems unrealistic to me. My
impression of the spatial distribution of coarse particles is that they are strongly influenced by
local emissions, and such local sources (e.g., some industry complex) are not spatially uniformly
distributed within a city or metropolitan area. As a result, there would be some concentration
gradient (i.e., spatial autocorrelation) within the city, as in the case of the Birmingham study.
The scenarios used in the simulation seem to assume that there is no constant gradient.
For a 8 km x 8 km grid, this may be a reasonable assumption, but I am wondering about the
usefulness of this scale. Isn't this DQO also a part of network design? If so, different scales of
grids need to be considered. Then, scenarios with a constant gradient will need to be considered.
This would mean that the variance (and likely CV) would also spatially vary.
D-67
-------
Dr. Donna Kenski
Comments on Attachment 1: Summary and Rationale for the PMio-2.5 FRM
Donna Kenski
Lake Michigan Air Directors Consortium
Sept. 20, 2005
Questions: 1. What are the scientific and operational strengths and weaknesses of the PMio-2.5
difference method relative to other options for a proposed FRM, especially when used as the
basis for approval of other methods?
2. Based on the field study report as well as any other available data, e.g., data from state and
local agencies, how does the demonstrated data quality of the PMio-2.5 difference method support
or detract from it being proposed as a FRM?
This document accurately summarized the case for the PMio-2.5 difference method. Both the
multisite evaluation (Att. 2) and the Jefferson County, Alabama, data indicate that the method
provides high quality data under carefully controlled conditions. None of the other candidate
methods can provide the same quality data and other significant advantages inherent in this
method. And practically speaking, because the existing FRM for PM2.5 defines it by the
methodology used to collect it, the FRM for PM coarse is constrained by that definition because
it incorporates PM2.5 as a lower bound. State and local site operators are already familiar with
the technology, and instruments now being taken out of service because of downsizing the PM25
network can be redeployed for this purpose. Thus the economic advantages alone are significant,
perhaps more so in the face of looming budget cuts.
Despite the clear advantages of the difference method, it is not suitable for providing the real-
time data that is critical for public health awareness. I agree with EPA's intention to emphasize
deployment of continuous FEMs in the network and use the gravimetric FRMs as audit devices.
EPA should clearly specify the QA requirements for collocated FRM instruments and be careful
to minimize the burden on state agencies implementing the new network.
I disagree with the statement that the proposed FRM will provide aerosol samples for chemical
analysis. Undoubtedly the samples can be collected and analyzed, but the nature of the
difference method, and the much higher uncertainty associated with the chemical analyses of
these speciation samples, makes the quality of such 'speciation by difference' data highly
uncertain. Perhaps it will yield acceptable data, perhaps not; further studies to examine the
practicality of the difference method for speciation are needed to demonstrate its utility. The
need for speciation data, however, is inescapable, and the virtual impactor has some significant
advantages for providing samples for speciation analysis, although its incomplete separation of
PM2.5 and PMio is problematic. As such, it seems premature to cite the possibility of speciating
the difference samples as an advantage of the proposed method. Other candidate methods may
be more well suited to providing speciation samples. EPA could and should address these
questions about speciation by analyzing the already-collected speciation data from the field
studies.
D-68
-------
A minor shortcoming in this summary was that it did not address losses of volatile species from
the proposed PMio-2.5 FRM, except to note that losses would be equal on both filters and thus
unbiased. Because the chemical composition of PM on the two filters will be different, the
volatility should not be assumed to be the same; differences in acidity or hygroscopicity of the
collected particles in each size fraction could affect losses (or gains) on the filters. The data
analysis suggested above could also begin to address questions about differing volatility of the
two size fractions.
Both the PMio-2.5 and PM2 5 methods would greatly benefit from a through-the-probe audit
system that generates known quantities and sizes of aerosols. Efforts to develop such equipment
should be pursued, either in EPA's own facilities or by funding other researchers.
Comments on Attachment 5: Sensitivity of the PMIO-2.5 Data Quality Objectives to Spatially
Related Uncertainties
Donna Kenski
Lake Michigan Air Directors Consortium
Sept. 20, 2005
Question: To what extent have the assessments of spatial variability and the sensitivity of the
DQO process to a variety of population distributions been appropriately addressed?
This revision of the DQO model seems to have adequately addressed the concern about
multimodal distributions of PM2.5 that was raised at the July 2004 meeting of the subcommittee.
By introducing and testing the effects of phase shifts and biannual concentration peaks, the
authors have shown that these distributions have negligent effects on the decision zone
boundaries.
The report (and the accompanying documentation on the CASAC web site) was not clear about
the significance or adequacy of the grid size used to simulate spatial variability. The choice of
an 8-km grid seems too small to represent spatial variability. A larger grid more representative
of an urban area (50 km?) seems more appropriate, although that's just a guess due to the lack of
any real data. Perhaps the Jefferson County study could be used for validation, at least of spatial
variability on the urban scale. Data on spatial variability at the much smaller neighborhood scale
should also be collected (either in the field or from a literature review) and data from both scales
used to validate the model results. Further clarification on this part of the model would be
helpful, including an explanation of the choice of grid size and whether it is influential. It wasn't
possible to tell how sensitive the results were to grid size. This analysis, while statistically
interesting, seemed to raise more questions than it answered (which was a useful finding in and
of itself). In addition to the questions above, issues raised in the subcommittee discussion
included the influence of monitor height on variability, and the expectation that spatial variability
of the components of PMio-2.5 will be quite different from mass.
D-69
-------
Dr. Thomas Lumley
Thomas Lumley, Ph.D.
09/16/2005
Comments on Attachment 1
The proposed FRM appears reasonable and well justified.
An important feature of PM10 and PM2.5, as noted in the document, is that they are to a
significant extent defined operationally, by the characteristics of inlets, impactors, &c. There are
definite advantages in maintaining the same effective definition of PM2.5 and PM10 when
defining the coarse subfraction. This operational definition cannot be endorsed without
restrictions, of course, as it would imply that improvements in measurement are impossible by
definition.
Assuming that measurement error cannot be completely eliminated, no method of measuring
PM2.5, PM10 and coarse mode PM can simultaneously guarantee that both subfractions will
have non-negative measurements and that their sum will be equal to the PM10 measurement.
The proposed FRM guarantees equality with the PM10 measurement but not non-negativity.
This is reasonable for NAAQS attainment, as noted in the document, and is also reasonable for
research, where occasional negative measurements need not cause any great consternation.
The summary and rationale notes that a difference method will tend to cancel out biases that are
common to the PM10 and PM25 subfractions. This is certainly true, but the price to be paid is a
magnification of bias in the comparison of coarse and fine subfractions, where biases will no
longer tend to cancel. The reasons for not proposing a continuous or semicontinuous monitoring
technology as the FRM are cogent. It is still important that development of these monitoring
technologies is not retarded by the specfications of the FRM. This is especially true because the
adoption of a daily rather than hourly standard for coarse PM is not due to any evidence that
daily averages are more relevant to public health than are shorter time intervals. Rather, the
availability of daily data has encouraged research using daily average concentrations as a
measure of exposure.
An important possibility is that a continuous or semicontinuous measurement technology would
give results comparable to the FRM at some sites but not at others [perhaps because they are
sensitive to temperature, humidity or aerosol characteristics]. Local validation and use of such
technologies for monitoring beyond that required by the standard should be encouraged, even if
they cannot qualify nationally as FEM.
Comments on Attachment 4
The proposed FEM criteria seem to be well thought out and well justified. However, it is not
clear to me exactly how the criteria relate to the DQO.
D-70
-------
The proposed criteria ensure that a new method must agree with the FRM and must have been
tested over a reasonably wide range of PM concentrations in at least two seasons. The proposed
bounds on error appear reasonable.
One issue that is not discussed is the choice of sites. The performance of (in particular) non-
gravimetric methods may well be affected by differences in aerosol characteristics and in
temperature and humidity. For example, nephelometry provides high time resolution and
excellent agreement with the FRM in Seattle, but does not perform as well in many other
locations. There do not seem to be any criteria other than range of PM2.5 concentrations for
choosing the sites, and agreement at one set of sites may not guarantee agreement under different
conditions at other sites.
It seems that the criteria should include some reason to expect the chosen sites to be
representative (or at least the absence of reasons to expect the contrary).
An additional non-technical note: the use of the term "precision" for a quantity that has large
values for less precise measurements is unfortunate, resulting in potentially confusing statements
such as: "The precision of the FRM sampler is required to be no greater than 7 percent" (p5). In
addition to conflicting with ordinary language use of the term, this conflicts with the technical
use of "precision" in statistics as the reciprocal of a variance.
D-71
-------
Knnlier reservations on atladitnem 4
I had previoiislv behoved iluii ihi' coeilicieni ol variation (precision! constraint together with the
roeression bias constraints would necessarily provide a siilficieni: constraint on the error in the
[•"KM. \YaiToll While's example su^osIS lltal this is Dot the case and that a more careful analysis
is needed.
'['he basic daia lor determining an FKM are (TUM I'HM}. the dillerence between a single FF.M
monitor and tin1 best estimate (,f |||(. const flic I measured by idle I-']{M. "II li' Use' o|' <•( >c||ici(>nl s 111"
variation .i;ivcs lc^ vreiujii in tlK-se crrnrs at hi;j,li levels than at Inw levels. This does rmt seem
sensible either lor I'eelllalnrv purpose;-, ( \vhel'e acclll'ai'V ill lli.^ll colicelil Pat iiUlS Is illl])i )11 alll J or t( >r
scjelllilic puppost's. The coi'llicielll of Variation is \vi<.l<'lv used as a simple deS(-pi])t ive sltmmar\' of
variation, but i liis does noi ini]>ly iliat it is an N>ness of data lo
lite regression line \vere i|iiaiitilied by the residual mean squared error. 1'nlike the coprelaiion. this
Is tiol sensitive lo the Palllle of I he FKM.
Ii may be paPticiilaply itsetnl tpeafs unavoidable thai the FKM excludes these components, lull it
would be very helpful if [''KMst'ould include them. This v.-oitld ppoviile an opportunity lor transition
lo A ditfereiii indicator as seiontilic iindepstaiidine; develojis.
\\1tile chaiiein;4 the precision snnnnapy to use variance pathcT than coedii ieni of variance would
be snlticieiit. the l-"l'A should consider1 for fiiiiiiv de\-e|opmeni a set of summaries ihat is more
lralls|)arenllv related to the dillerences between the 1'1'iM and candidate FKM ineastirilients. ( )ne
approach is as iolloivs:
A simple SltminaPV o) the error's is the a'I'el'a'jed sillliipeil e]'|op
Idlis dil'ecl ly measures the dilfel'ellce bet Ween the [-'KM a lid I he FKM . wit 11 Ho asSlllilpl in US a I mill
regression stpllctupe of be! Weeli-illstriimenl homogeneity.
[{enpessioii sipiicture is still impopi;uii for extpa])olat ion from levels of PM encountered durin.u;
testing to the presinnalilv higher levels peli'vnnt fur XAA(.»S attainment. Tlie ASK does coiisii-ain
I lie acclllMcV of I lie reeressinll line, since
D-72
-------
where r2 is the variance nf the FKM data, tT2 is the residual variance in the regression.
-------
Dr. Peter McMurry
Peter H. McMurry
September 28, 2005
RE: CASAC Ambient Air Monitoring & Methods (AAMM) Subcommittee Meeting: Comments
on Materials provided for review
Attachment 1 : Summary and Rationale for the PMu^ FRM
Proposed Sampling methodology: It is proposed that compliance measurements of coarse
particle (2.5-10 jim) concentrations be determined by the difference between mass
concentrations obtained using PMi0 and PM2 5 samplers. Both samplers would be based on the
current FRM PM2.5 instrument with the exception that an impactor or cyclone is used to remove
particles larger than 2.5 jim aerodynamic diameter prior to sample collection on the filter for
PM2.5. Both samplers operate at low volume, thereby reducing evaporative losses, and both
collect integrated 24-hour samples. Finally, this standard could be implemented by adopting
existing equipment and without additional training of field operators.
Data collected by EPA show that when operated with proper care, the PM2.5 and PM10
instruments provide high precision data. Furthermore, EPA showed that even when mass
concentrations are low, PMio-PM2.s is nearly always positive. EPA argues that these qualities
support the validity of this methodology for compliance measurements.
I am very impressed with the high quality field work that was carried out by Robert Vanderpool
and coworkers when collecting data used to assess proposed sampling techniques for PMio-2.s.
This work reflects a high level of commitment, thought and energy. It is essential that such work
be continued, and that resources dedicated to it be maintained or increased. It is only through
field observations that the relative merits of different measurement methods can be assessed. The
contributions of this team need to be recognized and valued by the agency. They are
extraordinary.
My Assessment: I feel that this document focuses primarily on the strengths of the proposed
methodology while inadequately acknowledging its weaknesses. Strengths include:
-Proven technology;
-Direct gravimetric measurement of mass;
-Uses existing FRM equipment (minimizes equipment and training costs);
-Presence of sub-2.5 jam particles causes coarse particles to adhere to filter;
-Field tests show that coarse mass measured with dichot = coarse mass measured with
proposed methodology, which suggests that evaporative losses are minimal;
-Consistency with historical database of mass measurement avoids the need for expensive field
comparisons;
-Measurements are precise, and measured coarse mass concentrations are positive even when
mass concentrations are low;
-Low face velocities keep evaporative losses low.
D-74
-------
Weaknesses include:
-Unknown measurement accuracy due to positive and negative adsorption on filter;
-Likelihood that measurement artifacts will be different for the PM2.5 and PMio filters because
(1) reactions between fine and coarse PM species will affect the volatility of nitrates and
other compounds, (2) evaporative losses on PM2.5 will exceed evaporative losses on PMio
due to the pressure drop provided by the WINS or cyclone. Is this one reason that PMio was
systematically higher than PM2.5 in field trials?;
-Sampling artifacts lead to inaccuracies that cannot be quantified. Therefore, measurement
accuracy is unknown, (does this lead to data that have "a high degree of fidelity and
faithfulness?);
-The extent to which find particles helps coarse particles adhere to filters is not quantified.
While it is plausible that such adhesion should occur, it is likely to depend on mass collected
and on particle composition, and these effects are not discussed, although it became clear in
our meeting that empirical evidence supporting the problem with coarse particle losses from
dichot data is available;
-inherently poor time resolution (24 hour) with no possibility for use in forecasting; and
-expensive manual operation.
In summary, the recommended choice is pragmatic and precise but its accuracy is unknown. My
views have not changed since I wrote my preliminary evaluation on July 23, 2004. I would
probably not be enthusiastic about any filter-based method, although I would conceptually prefer
a dichotomous sampler to the proposed PMi0-PM2 5 approach. Nevertheless, it is not clear to me
that benefits of using a dichotomous sampler would justify the additional funds required to equip
sampling stations and to train operators. If the proposed methodology is adopted, it should be
done so with the understanding that it should be replaced when more suitable methodologies
become available. Furthermore, work should be continued on the development of accurate
techniques for measurements of coarse particle mass concentrations.
D-75
-------
Peter H. McMurry
September 28, 2005
RE: CASAC Ambient Air Monitoring & Methods (AAMM) Subcommittee Meeting: Comments
on Materials provided for review
Attachment 4: Criteria for Designating PM2_5 Equivalence
A statistical methodology for determining whether a measurement method is "equivalent" to the
PM 2.5 FRM is proposed. The statistical methodology involves the analysis of data acquired
with 3 to 5 "equivalent" and 3 FRM co-located samplers. Four statistical measures must be
satisfied to meet the equivalency designation. These include:
(1) Precision: must not exceed 15%;
(2) Correlation with FRM data (r): must range from 0.93 to 0.95;
(3) Multiplicative Bias (b) (slope of linear least squares fit of equivalent sampler to FRM
data): Must range from 0.9 to 1.1;
(4) Additive bias (a) (intercept of linear least squares fit of equivalent sampler to FRM data):
must fall between 15.05-17.31*b and 15.05-13.20*b. (-3.99 to 0.53 |ig/m3 for b=l.l and
-0.53 to 3.17 |ig/m3 forb=0.9).
My comments: The focus of these criteria is the ability to duplicate data acquired by the FRM,
regardless of its faults or inaccuracies. The aim is to identify samplers that can be used for
compliance measurements, not necessarily for accurate measurements. It is entirely conceivable
that an alternative sampler could measure PM2.5 with high precision and accuracy but would be
excluded because, for example, it was able to detect semivolatile compounds that are not
sampled effectively with the FRM. Such a sampler might have "r" values below the designated
values, and might produce "a" and "b" values that differ significantly with season and from one
location to the next. It is also possible that some PM2.5 health and other effects are associated
with those semivolatile compounds.
It would appear that the law and our legal system are constraining us to invest most of our energy
and financial resources into developing and certifying samplers that reproduce FRM results. The
proposed criteria will probably achieve this goal for most practical situations. Furthermore,
these equivalent methods may offer significant advantages, such as high time resolution,
automated data collection, real-time measurements, etc. It would appear to me, however, that
this proposal does not address all issues that pertain to improved measurement accuracy for
PM2.5.
I feel that more thought ought to be given to comparing the responses of alternative samplers
with the FRM using laboratory generated-aerosols of known composition and size or size
distribution. Such work could include sampling of known semivolatile compounds, such as
ammonium sulfate and selected organic compounds. While this methodology would not be
applicable to equivalency determinations as specified by law, they could take us a long way
towards an understanding of measurement accuracy. Because the atmospheric aerosol is so
complex, there are many processes that could lead to discrepancies when samplers are used for
D-76
-------
atmospheric sampling, even though the samplers operate with identical inlet characteristics. The
laboratory tests would, enable unambiguous testing of sampler performance to particles having
known physical and chemical properties. This approach would help us improve our
understanding of measurement accuracy, and would lead to the design of improved samplers in
the future.
I am also concerned about EPA's narrow focus on FRM measurements. There is clearly a need
for more sophisticated measurements designed to support epidemiology studies at selected sites.
Such measurements would also help to refine our understanding of emissions control measures
that are currently being implemented, and would help with the development and evaluation of
process models for atmospheric aerosols. Substantial progress was made during the supersite
program to develop such measurement methodologies and to demonstrate ways in which they
can be used most effectively. It is not clear to me that EPA is working systematically to build on
what was learned over the past decade, and I regard this as very unfortunate and shortsighted.
Examples of measurements that can be carried out routinely and continuously include particle
size distributions and size-resolved composition.
D-77
-------
Dr. Kimberly Prather
October 1, 2005
To: Fred Butterfield, Designated Federal Officer
Clean Air Scientific Advisory Committee (CASAC)
Rich Poirot, Co-Chair - Monitoring
Barbara Zielinska, Co-Chair - Methods
CASAC Ambient Air Monitoring and Methods (AAMM) Subcommittee
From: Kim Prather
Subject: CASAC Review of the Particle Methods and Data Quality Objectives
In general, the documents provide an excellent description of the plan for establishing a PMio-2.s
Federal Reference Method (FRM). However, based on the descriptions provided, it appears the
main reason for choosing the difference method is because it is the easiest to implement given
the current suite of instruments located at current sampling sites. Is this reason enough? This is
an important question to address up front as it is one EPA will no doubt have to defend once the
new standard is implemented. It would be helpful if support and further justification were given
(i.e. health data) that led EPA to make this choice.
Peer Review Questions:
3. What are the scientific and operational strengths and weaknesses of the PMio-2.s difference
method relative to other options for a proposed FRM, especially when used as the basis for
approval of other methods?
The strengths are well laid out. The weaknesses are not. The difference method's major strength
is that it will use currently existing instrumentation already deployed to support the current PM2.5
FRM. It will allow comparison with previous FRM measurements; at the same time, this is also
it's major weakness in that the main basis for choosing the new sampling methodologies is that
they must agree with a technique (FRM) that has been shown to have serious flaws (i.e. RH
effects, volatilization losses). Yes, it will offer consistency (precision); but is a consistently
incorrect answer the one we want? Accuracy should be considered as well and in order to make
this a requirement, an effort needs to be made to create lab standards which can be used for
validating potential FRM methods. We really need to think towards the future (as much as
possible) and envision telling the public the reasons for going this route: right now we would
have to say, we chose the new standard because we have these methods for measuring PM with
high precision that we have used for a long time. Yes, we know the results they provide are not
really representative of PM in the atmosphere, we're not really sure how the numbers are
correlated with health effects, but they are what we currently have and we have invested lots of
resources in them so we need to continue to use them.
D-78
-------
Other major drawbacks include high analysis costs, requirement of skilled personnel to operate
and perform filter handling protocols which will require "experts", and limited (24 hour) time
resolution. PM, particularly in the coarse size range, show concentration excursions on
timescales as short as minutes and as long as a couple of hours; these will be completely missed
(averaged out) by the proposed 24 hour sampling times. See figure at end of report as an
example. This figure shows data taken in Rubidoux CA where PMi0 concentrations repeatedly
got as high as 158 |J,g/m3 at 6-7 am for 1.5 hours every day. However, the 24 hour average
PM10 concentrations for these dates of 60 |J,g/m3 were well within the proposed limit for the
standard. These seem like high excursions that could potentially affect health and they would be
missed. However, without a better understanding of the health effects of PM, is it wise to choose
an FRM protocol which completely misses such concentration variations? Also, better time
resolution often offers insights into the source/s leading to high concentrations of PM. With 24
hour samples, one will get a single number which will not tell us anything about the PM causing
the problem. Also, the errors associated with the difference method could be substantial,
particularly for chemical analysis (this has not even begun to be explored).
4. Based on the field study report as well as any other available data, e.g., data from State and
local agencies, how does the demonstrated data quality of the PMio-2.s difference method
support or detract from it being proposed as a FRM?
Choosing a method based on equivalency to the FRM has potential problems as stated. It still
isn't clear why these particular size cuts were chosen other than "convenience". Are there health
studies that support separating PM into these 2 size ranges? To rule out a new technique as cited
in the report because it deviates from the FRM is dangerous. The newer instrument may be
providing the "right" answer. Furthermore, when the results of one technique deviate from the
FRM, this could be telling one something about chemical differences. Maybe the deviations are
due to problems with the FRM instead of the method it is being compared against. By forcing a
technique to be "equivalent" to the FRM, you're forcing it to have the same biases. This is quite
problematic.
Even though the field studies reported in Appendix 2 were carefully thought out and conducted,
they only represent 3 data points. One should think about expanding these sites to include other
seasons. For example, Riverside in summer has low nitrate concentrations and less "issues" with
SVOC since it is so hot. Going back to Riverside in the Fall (November) and comparing the
results should be highly informative.
The peripheral data obtained in the field studies (APS size distributions, chemical speciation)
need to be exploited to understand discrepancies between methods (see notes at end). As
discussed in the meeting, it would be helpful to complement the field studies conducted to date
with lab studies using standard particles of known size and composition to test new methods that
are being proposed. More effort needs to be put into choosing an FRM that will provide accurate
answers on PM (as well as precise).
They go into huge detail about not choosing proprietary samplers. This seems like weak logic
and one that shouldn't be used. We need to step back and ask ourselves the goals of this
exercise.
D-79
-------
Consultation Questions:
9. What are the Subcommittee's views on the Very Sharp Cut Cyclone (VSCC) being approved
as an alternate second-stage impactor to the WINS for use on a PM2.5 FRM?
This change seems justified.
2. To what extent are the stated advantages of relaxing existing requirements identified for the
PM2.5 FRM supported by the information cited in Attachment 3, available literature, or good
field and laboratory practices? Does the Subcommittee have additional recommendations for
the PM2.5 FRM that would be neutral with respect to bias, but would improve the
performance and minimize the burden on agencies conducting the sampling?
No evidence is supplied to show why it is necessary to maintain 4 deg. C temperatures. There
will be a serious cost associated with it, so it needs to be better justified.
5. To what extent have the assessments of spatial variability and the sensitivity of the DQO
process to a variety of population distributions been appropriately addressed?
Real ambient data should be used to test the DQO. There are many factors that will affect this
range that have yet to be addressed. The current analysis is too simplistic and needs to be
expanded. There are large amounts of data available to do this from multiple sites, locations, and
extremes. One doesn't have to be "hypothetical".
6. What are the Subcommittee's views on the approach identified for the development of
criteria to approve continuous PMio-2.5 equivalent methods?
Not enough detail is given here. To judge this, it would be helpful to know how many sites and
seasons will be studied.
Summary:
1. As pointed out, the purpose of having the APS located at the sites during the study was
not to measure FRM mass, but to understand how changes in the size distributions could
play a role in measured differences in the FRM methods being tested. In the December
2004 meeting, we asked why EPA was using a PM10 cut on the front of the APS and
basically throwing away information that could be used to understand discrepancies.
They indicated they would take it off for future testing but this hasn't happened. It would
be really straightforward to use software to add a cut-point (sharp, 50:50, etc.) and
directly measure the effect on PM2.5, PMio, and PMio-2.5- Then the issues of "bias",
"error", and "zero values" could be directly addressed using real size distribution data
acquired at the actual sampling locations. Also, "events" where the data from 2
techniques deviate from one another could be explored further using the measured size
distributions to understand if it is playing a role.
2. Along the same line, one of the reasons for urging EPA to collect speciation data was so
when 2 methods didn't agree with each other and/or the FRM, the composition at the
different sites might be used to shed some light on this. I was quite surprised when
D-80
-------
reviewing the documents for the meeting, there was not a single piece of information
given on composition differences between the sites. All we were shown were time series
of 24 hour mass concentrations collected by multiple instruments. Impressive data were
acquired and great comparisons were done, but very little analysis and interpretation of
the data was done. Now we are being asked to choose a method—but we are missing key
information that would help us decide.
Suggestions:
1) In the current tests being conducted in Birmingham, EPA should ask TSI to deploy
another APS without a PM10 cut point and use this alongside the other instruments for
the entire study. This could provide key insight into the data and the size distribution
data could be used to address "bias" concerns regarding cut points.
2) Someone should begin looking at the chemical speciation data ASAP. Without this, it
would be very difficult to make a fully educated decision on the new FRM. EPA has
done a series of excellent field studies to date; however, it appears most of the work has
gone into collecting data and measuring differences and very little has gone into
understanding these differences.
3) Choices of sites for determining "equivalency": the details on this are lacking in the
report. This was purposely left as open to allow people flexibility in using new
techniques. However, some amount of guidance and restrictions needs to be made or this
could result in open-ended testing. More hardline guidance needs to be given: perhaps
stating the instruments need to be tested in a region that has a certain range of coarse
concentrations (giving a range of low to high values), a specific range of PM2.5/PMi0, etc.
For example, the state of CA as well as most other states with high PM levels have long
historical data on PM mass concentrations for PM2 5 and PMi0 obtained with TEOM's
and beta attenuation monitors (which don't use heating), etc. These can be studied for
different seasons in previous years to better understand measurement challenges and set
realistic time and locations for sampling and establishing equivalency. The figure below
shows mass concentration measurements for PM10 and PM2.5 acquired in Rubidoux CA
in spring 2005 generated from the AQMD web site. One can see the range of variations
of coarse PM, and how the relative amount of PM2.5/PMi0 varies over time. Such
information can be used when choosing locations for equivalency testing.
D-81
-------
PH
•O
PH
O
160-
140-
120-
100-
60-
40-
20-
Rubidoux CA
• PM10 (TEOM-heated)
PM2 5 (BAM--non-heated)
PM2yPM10
-3.0
-2.5
Date
D-82
-------
Dr. Armistead (Ted) Russell
September 19, 2005
To: Fred Butterfield, Designated Federal Officer
Clean Air Scientific Advisory Committee (CASAC)
CASAC Ambient Air Monitoring and Methods (AAMM) Subcommittee
From: Ted Russell
Subject: CASAC Review of the Coarse Particle Methods and Data Quality Objectives
This memo provides the comments by Ted Russell on two documents: Summary and Rationale
for the PMio-2.5 FRM") and ["Memo to PM NAAQS Review Docket (OAR-2001-0017):
Potential Changes being Evaluated for the PM2.5 FRM". It should be noted that I am primarily
an air quality modeler and data analyst, so I am not particularly able to address the issues
involved in the monitoring, per se, but more with the use the resulting data.
Summary and Rationale for the PMi0-2.5 FRM
Question 1: What are the scientific and operational strengths and weaknesses of the PMio-2.5
difference method relative to other options for a proposed FRM, especially when
used as the basis for approval of other methods?
In this regard, I am primarily concerned with the possible decision to use a method (i.e., the
proposed approach) that will provide very little information for data analysis and model
evaluation. A continuous method would be greatly preferred. At present, there appears to be an
overwhelming desire to continue to use filter-based sampling for the FRM. The main reason to
use a filter, as I see it, is to do speciation, for which there are no current sub-standards. A
continuous method will provide more data and can also capture 24-hour exceedences that would
otherwise get split between two days. The increased operating expenses and error introduction is
also an issue in using a filter-based method. Looking down the road, if the real health impact is
due to a more acute exposure, say over an hour, you need to have the continuous monitors to
help identify the impacts and also for future regulation. We have determined that shorter term
standards are appropriate for gas phase species (in part, since the method to measure them at
shorter time scales were there), and I see no real reason it might not turn out that some of the
effects of PM exposure (coarse, fine or whatever) are also due to shorter term events. All of the
methods proposed, to date, have biases and artifacts, so use a method(s) that provide additional
information.
Memo to PM NAAQS Review Docket (OAR-2001-0017): Potential Changes being Evaluated
for the PM2.5 FRM
Consultation Question 2: What are the Subcommittee's views on the Very Sharp Cut Cyclone
(VSCC) being approved as an alternate, second-stage impactor to the Well Impactor
Ninety-Six (WINS) for use on a PM2.5 FRM?
D-83
-------
Having run a couple of sampling systems, I prefer operating a cyclone vs. the WINS impactor
(my system uses both). My understanding is that the differences in the results are
very minor, so I would be prone to using a VSCC.
Consultation Question 3: To what extent are the stated advantages of relaxing existing
requirements identified for the PM2 5 FRM supported by the information cited in Attachment 3,
available literature, or good field and laboratory practices? Does the Subcommittee have
additional recommendations for the PM2.5 FRM that would be neutral with respect to bias, but
would improve the performance and minimize the burden on agencies conducting the sampling?
My interpretation of the current results from the continuous monitors suggests that they can
provide reliable results. (Outside of my area.)
D-84
-------
Dr. Jay Turner
Comments in Response to Attachment #6: PM'10-2.5 Method Equivalency Development
Submitted by Jay R. Turner (September 19, 2005)
The overarching philosophy set forth in this attachment is "for making
comparisons to the NAAQS, the decision quality [from the equivalent method] is as good as
if they had used filter-based reference methods". This statement is opcrationalizcd by
matching the upper- and lower-bounds for the gray zone of the decision performance curve
for the candidate equivalent method to the respective bounds for the gray zone of the
decision performance curve for the DQO. The equivalency criteria are assumed to have a
fixed form (i.e. specified thresholds/ranges for precision, correlation, multiplicative bias, and
additive bias) and will be developed based on the aforementioned matching. For equivalent
methods which provide both high sampling frequency and high data completeness (e.g.
continuous monitors) the equivalency criteria can actually be relaxed compared to the criteria
embedded in the DQO.
Constitution Ones!i<>/i. What are the Subcommittee's views on the approach identified for
the development of criteria to approve continuous PM 10-2.5 equivalent methods?
Comment*. To the extent the approach is presented at this time, it does appear reasonable.
I like the explicit linkage between the DQOs and the equivalency criteria. My comments
focus on what is not presented - how the results from this DQO/equivaleney matching are
used to actually arrive at the equivalency criteria.
1. Section 2.1 states "data collection will be replicated at multiple sites to ensure the
sampling is representative of different aerosol types [...]". It will be important to provide
additional guidance for site selection. Not only should the aerosol types be different but
rather both the aerosol types and environmental conditions should capture those
conditions which are most likely to challenge the candidate method. Such conditions
may be different for instruments which are based on different operating principles. There
will also be issues concerning the on-site physical location of instruments - ambient
conditions versus being housed in HYAC-controlled environmental shelters - which will
need to be addressed in deciding what deployment characteristics actually constitute the
equivalency test conditions.
2. In the proposed approach, the equivalency criteria are coupled. The example presented in
Attachment 6 shows how starting with a prescribed precision and multiplicative bias one
can bound the criteria for correlation and additive bias. Indeed. PNb.s equivalency
criteria presented in Attachment 4 include an additive bias criterion which is functionally
dependent on the multiplicative bias. It is not clear whether the approach presented for
the PM 10-2.5 equivalency criteria includes solely fixed forms or relational forms (like the
PMi.s equivalency criteria of Attachment 4). The text states (Section 2.0) "a fixed form
for the equivalency requirements was assumed" but later states "the entire process
follows that [PM?.5 equivalency requirements] development". Relational forms provide
the greatest flexibility towards meeting the equivalency requirements whereas fixed
forms place additional and potentially important - constraints. The example in
Attachment 6 is quite relevant to this discussion. Figure 4 shows the acceptances ranges
for additive bias and multiplicative bias for the example presented. Following
J.R. Turner ' Attachment 6 Pa»e I of 3
D-85
-------
Attachment 4, LI relational criterion for the additive bias would use the upper and lower
lines of the parallelogram to bracket the allowable range for additive bias for a given
(observed) multiplicative bias. For this particular example - which is presumably
realistic (note the assumed daily standard of 60 jug/m* and l)Hlh percentile falls in the
range set forth in the EPA Staff Paper) the acceptable ranges for the additive bias at the
extreme values for the multiplicative bias arc quite large (-7 to +22 ug/nr1 at -15%, -31 to
+ () ug/nr at + 15%), The text subsequently states that for this example a candidate
monitor additive bias criterion of ^5 ug nr' would be reasonable. This implies a
rectangle would be drawn within the parallelogram which would define a fixed form for
the additive bias, unlike the relational form set forth in Attachment 4 for PMj.s
equivalency,1 The decision on whether to use a fixed form or relational form could well
be very important in terms of what is deemed acceptable instrument performance. If the
DQO gray zone is very broad (as is the case with this example), at some point common
sense dictates that a monitor exhibiting. -31 (.ig-'nv additive bias (lower-right corner of the
parallelogram in Figure 4) should not be acceptable for a standard of 60 |ug;nf even
though it falls within the acceptance range for a relational form of the equivalency
requirements.
3. As a follow-up to comment =2, each cuinlidnii' \iinij>!cr in the i'c/iiirnlciicv toting should
meet the (/cfuieti Cijuimle/u-v criteria (rather than the mean performance over multiple
samplers meeting the defined equivalency criteria). The actual deployments for
compliance monitoring will most certainly not have the luxury of collocated, identical
samplers and thus the equivalency testing should be consistent with the real scenario.
Since equivalency designations are based on performance specifications rather than
design specifications, each individual instrument must be able to meet the equivalency
requirements.
4, If a fixed form for the equivalency criteria is chosen, how does one get started in using
the approach to generate the equivalency requirements? That is, will precision and
multiplicative bias from current/emerging monitors be used to set these parameters with
the correlation and additive bias criteria developed accordingly?
5, While perhaps modestly off-topic. I am curious how post-deployment performance
evaluations of continuous, equivalent methods would be conducted. For filter-based PM
measurements the flow rate is routinely cheeked in the field and the analyzer zero and
span (in this case, for the microbalancc) are routinely checked in the laboratory. Thus,
this approach is analogous to on-line gas measurements where the analyzer is in the field
but still routinely subjected to performance texts. For continuous PM monitors, we can
check flow rates and analyzer zeros but we don't have a convenient, through-thc-probe
method for routinely checking the "span". We can monitor parameters which indirectly
give us information on the performance of subcomponents of the measurement but this is
different from an actual span which would require a through-thc-probe aerosol. Thus,
there is a need for bench testing (prior to initial deployment) and/or acceptance testing
(immediately upon deployment) of each instrument as well as routine performance
checks. The bench'acceptance testing could be approached by aerosol challenge tests in
[ enjoyed running version Beta 2 ol'the [)(,M ) Companion lor PM^WSC (downloaded from
lillp: www.epa.uov tlnamtil diiotool.htmli and look fonvard lo getting access to the newer version which,
according to AlUiehmcnl (\ can generate the acceptance rant:e plots as well as the decision performance curves.
J.R. Turner; Attachment 6 Pa»e 2 of 3
D-86
-------
the laboratory: in the field, it could include col location with a continuous monitor with
documented performance characteristics and,or collocation with filter-based
measurements. The routine field testing (audits) will likely require collocation with
filter-based measurements (especially in the case where the continuous monitor requires
an environmental shelter with HVAC controls which would complicate installation of the
audit monitor). The bar is potentially much higher than conventional audits of filter-
based samplers because there is no routine performance check on the continuous
monitor's "span". Logistically this could be quite challenging and I look forward to
learning how (methods and frequency) the field performance of deployed continuous
monitors will be determined.
Again, these comments do not draw into question the overall approach of making the
equivalency criteria consistent with the DQOs nor docs it question the specific DQO tool
being used to study the interactions between acceptance criteria. These comments are
addressed at how the information would subsequently be used to actually arrive at the criteria
and whether there should be some level of performance testing for each instrument deployed
beyond the equivalency testing on a few instruments of that make and model.
In closing, even though it is beyond the scope of the charge question I am compelled to
comment on the example provided in this attachment. To the extent the DQO example does
represent a realistic worst-case scenario, the gray zone for the example 24-hour standard
(which is within the range stated in the EPA Staff Paper) is very large! The difference
between the lower bound and the standard is 37% of the standard and the difference between
the upper bound and the standard is ?<•)% of the standard. It will be interesting to observe
what DQOs are actually established for PM 10-2.5 im«J how this will subsequently impact the
development of the equivalency requirements.
.R. Turner / Attachment 6 Page 3 of 3
D-87
-------
Comments in Response to Attachment#1: Summary and Rationale for PM 10-2.5 FRM
Submitted by Jay R. Turner (September 19, 2005)
Ihickgnntiul. A PM 10-2.5 Federal Reference Method is proposed which is based on the PMjo
mass concentration minus the PM2.5 muss concentration us determined by independent
measurements.
Peer Review Otieslions, (1) What arc the scientific and operational strengths and weaknesses
of the PM 10-2.5 difference method relative to other options for a proposed FRM, especially
when used as the basis for approval of other methods? (2) Based on the field study report as
well as other available data, e.g.. data from State and local agencies, how does the
demonstrated data quality of the PM 10-2.5 difference method support or detract from it being
proposed as a FRM?
Comment*.
1. A compelling case is made for a filter-based FRM (be it the proposed difference
method of a virtual impactor) compared to a continuous monitor being designated as
an FRM.
2. The virtual impactor (compared to the difference method) offers no advantage in
terms of analysis costs since the fine channel (PMi.s) mass concentration must be
measured to correct the coarse (PM 10-2.5) channel mass. That said, there may be
advantages to the virtual impactor approach in terms of chemical analysis of the
coarse fraction, especially for species which would have large propagated
uncertainties using the difference method. The overwhelming drawback to the virtual
impactor appears to be the presumed high potential for particle losses during sample
handling. Are there any contemporary studies that have investigated this issue? I
recall the field evaluations described in Attachment 2 used on-site gravimetric
measurements? Were some of the coarse particle filters also shipped to a laboratory
for analysis to mimic typical handling procedures for a routine deployment by a State
or local agency? In any event, I would be very interested in seeing the evidence
which demonstrates that PM 10-2.5 - laden filters are susceptible to particle losses
during shipment; this is certainly the conventional wisdom but I have never seen the
data.
.1. The proposed advantage of the FRM providing "readily accessible aerosol samples
for subsequent chemical analysis" is overstated. Following the example of the PMj.;
FRM and PM?.5 speciation sampler, one would need .s/.v samplers to characterize the
aerosol (PMio with quartz filter, PM?.5 with quartz filter, PMio with Teflon filter,
PM2.5 with Teflon filter, PMio with Nylon filter, PM?.5 with Nylon filter).
4. While EPA anticipated a shift to continuous monitors as such instruments obtain
equivalency status, the extent to which routine on-site performance evaluations will
be required is not clear. Given there are no tractable tests of the true instrument
"span" for such monitors, such testing goes beyond the role of a traditional audit in
that it is the sole ongoing evaluation of the instrument span. These evaluations will
most likely require filter-based measurements. If the required frequency is high, the
intrinsic value of deploying continuous monitors will be diminished.
J.R. Turner Attachment 1 Pa»c I of
D-88
-------
Dr. Warren H. White
ATTACHMENT 1: Summary and Rationale for the PMi0-2.5 FRM
Revised comments by Warren H. White, 9/26/05
In its September 15 letter to the Administrator on EPA staff recommendations for a thoracic
coarse PM standard, CAS AC observed (http://www.epa.gov/sab/pdf/sab-casac-05-012.pdf, page
6) that
By use of the indicator UPMio-2.5, the Agency is taking a next step toward
including composition as well as size in its regulations of ambient air PM.
Let us not forget that the designation of "urban" PMi0-2.5 as the "indicator species" was presented
- and accepted by most of the Committee - as a surrogate for an as-yet-undefined composition.
At our September 21-22 meeting, Agency staff presented an effective case for the difference
method as a reference against which dichotomous samplers and continuous PMi0-2.5 monitors
could subsequently be tested. Alone among candidate methods, the difference measurement is
directly comparable with PM2.5 measurement and already available in final form. Beyond these
logistical advantages, Agency emphasis focused on the demonstrated precision of the difference
as a repeatable measurement of PMio-2.s. I am left, however, with three concerns that have not
yet been addressed to my satisfaction:
1 The proposed FRM has demonstrated its precision for PMi0-2.s, but there is
absolutely no evidence that it provides a repeatable measurement of "those
thoracic coarse particles that are generally present in urban
environments." The Staff Paper (page 5-56), for example, specifically
highlights episodes of high urban concentrations of PMi0-2.5 in Spokane WA that it
concludes did not represent elevated UPMio-2.5 levels. In other words, it reports
that days with comparable UPMio-2.5 concentrations would have produced disparate
FRM readings. As a measurement of the proposed NAAQS indicator, the proposed
FRM is wildly imprecise.
2 The proposed FRM will not meet the planned data quality objective for UPMi0-2.s.
The DQO will require daily monitoring, which is impractical with the FRM. Isn't it a bit
... well, awkward, that the Federal Reference Method itself can't deliver data of the
quality needed for decision-making? Examples were given at our meeting of other FRMs
that are not widely deployed, but are any of those incapable of satisfying their DQOs?
3. The proposed FRM is incompatible with the historical rationale for filter-based PM
measurements. The Agency has always favored filter sampling as the basis of its
reference methods for PM on the ground the collected deposit is then available for
potential chemical analysis or toxicological testing. This consideration carries particular
weight in the case of UPMio-2.5, where we are "taking a next step toward including
composition" as noted above. Clearly, however, collection of UPMio-2.5 for chemical and
D-89
-------
toxicological analyses is best done by virtual impaction to minimize contamination of the
sample by fine particles.
The third point merits elaboration. The PMi0c sample is an undifferentiated mix of coarse
and fine PM. Agency staff note that the composition and toxicity of the PMio-2.5 portion can be
inferred indirectly, by comparisons with the associated PM2 5 sample. Consider, however, the
Staff Paper's descriptions (page 5-57) contrasting "toxic" UPMio-2.5,
resuspended crustalparticles may be contaminated with toxic trace elements and
other components from previously deposited fine PM, e.g., metals from smelters
(Phoenix) or steel mills (Steubenville, Utah Valley), PAHsfrom automobile
exhaust, or pesticides from agricultural lands. (CD, p. 8-344),
with "less toxic" PMi0-2.5:
particles of crustal origin ... are relatively non-toxic under most circumstances,
UPMio-2.5, in other words, may just be ordinary crustal dust that has been contaminated by fine
particulate matter! How can a method that necessarily contaminates PMio-2.5 with fine PM, in
the very process of collecting it, yield useful information on ambient UPMio-2.5?
D-90
-------
ATTACHMENT 4: Criteria for designation of equivalence methods for continuous surveillance
of PM2 5 ambient air quality
Comments by Warren H. White, 9/15/05
A lot of careful thought and analysis have obviously gone into this paper, and I am inclined to
trust the authors' and Agency's practical judgments on what constitute appropriate equivalence
tests. I don't feel like I fully understand its overall rationale, however, and it seems like it should
be possible to lay that out more clearly. In particular, it's not clear to me how the proposed
criteria connect to the PM2 5 DQO.
The proposed criteria involve four measures: precision, correlation, multiplicative bias, and
additive bias. Precision is the only one of these that gives any information on the repeatability of
the candidate measurement. The remaining three all come from a standard bivariate regression
of the daily means from 2-5 candidate measurements on those from 2-3 FRM measurements.
The question I have is: How relevant is (A) the performance of means from collocated testing to
(B) meeting the DQO in actual deployments of individual monitors? The table below
summarizes some exploratory calculations based on the simulated data listed in Table 1. I used
the exact formulae given in the attachment, which yield results identical to those obtained from
the standard Excel functions CORREL, SLOPE, and INTERCEPT. I assume the differences
between the two right-most columns, comparing my bias results to their reported values (e.g.
1.060 vs. 1.058 for multiplicative bias), arise from truncations in the attachment's listing of the
simulated data. The full data set is included on the next page as an embedded Excel spreadsheet
so the calculations can be checked (in the electronic version of these comments) by double-
clicking on any cell.
r
b
a
a bound
a bound
individual candidate
vs mean FRM
0.966
-2.249
-4.1
0.4
0.938 0.963
1.020 1.050
-1.330-1.568
-2.6
1.6
-3.1
1.2
mean vs mean
WHW
0.981
1.060
-1.728
-3.3
1.1
Battelle
0.981
1.058
-1.692
-3.3
1.1
The three columns under the heading "individual candidate vs. mean FRM" show the correlation
r (Eq. 11), multiplicative bias b (Eq. 13), and additive bias a (Eqs. 14-16) of the three individual
candidate monitors relative to the mean daily FRM measurement. Should we be concerned that
two out of three of the candidate monitors fall outside the proposed criteria when tested as
individuals?
D-91
-------
Table 1
FRM samplers
Run #1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
mean
RMS
6.4
6.9
5.6
4.4
7.9
8.0
9.0
16.9
10.3
11.1
10.9
11.2
10.2
5.3
7.8
5.6
11.3
7.0
6.6
6.1
6.7
9.8
12.5
16.4
26.7
12.5
15.6
20.8
19.7
5.4
7.0
17.1
12.5
9.7
14.8
19.4
17.1
14.1
11.6
12.5
11.1
14.2
22.8
14.4
18.2
17.2
11.92
5.17
#2 #3
5.8
6.4
5.9
4.4
8.9
7.7
10.3
18.6
10.2
11.7
10.7
10.9
10.2
5.6
8.8
6.4
10.2
7.9
6.0
6.3
6.3
9.1
13.5
16.1
22.2
10.5
15.0
22.2
20.3
5.2
7.3
14.9
12.6
10.1
15.4
19.7
15.7
14.2
10.8
12.7
11.6
14.3
23.5
17.3
18.9
20.6
6.6
7.2
5.6
4.5
8.4
8.2
9.1
15.8
11.5
11.6
10.5
11.0
10.2
5.5
8.5
6.2
10.8
7.1
6.0
6.4
7.3
9.4
14.1
15.0
24.4
10.0
14.9
20.4
20.0
5.3
6.9
16.1
11.3
10.1
16.3
19.8
17.0
14.0
11.0
13.6
12.0
14.9
20.3
16.6
17.1
19.5
12.02 11.91
5.17
4.92
candidates
#1
5.5
4.4
4.5
2.7
7.5
7.2
5.4
16.2
9.6
10.9
11.9
9.1
9.2
3.6
5.0
5.3
7.8
6.3
4.7
5.7
4.3
7.4
15.4
15.0
21.2
10.8
11.5
20.8
22.8
3.6
4.8
15.9
10.9
9.1
13.4
18.5
14.0
17.2
10.2
11.3
10.8
13.4
24.3
17.1
19.7
19.5
10.99
5.78
#2
5.0
6.5
3.5
2.4
6.0
5.8
10.4
17.8
10.8
8.6
8.6
9.5
8.5
4.2
6.7
4.5
8.7
4.5
5.4
5.4
5.6
7.7
12.4
18.1
20.8
10.1
16.1
17.3
17.6
3.1
7.2
17.4
8.4
9.6
14.6
15.9
18.1
16.9
8.1
12.7
8.5
13.6
17.9
19.7
22.2
17.5
10.87
5.49
#3
3.7
4.3
4.3
2.2
7.6
6.3
8.3
15.6
7.6
12.5
10.1
12.0
8.4
2.8
5.5
5.3
9.3
6.9
4.0
4.3
7.6
10.3
10.3
19.4
20.0
11.0
13.9
19.9
22.4
4.3
5.9
15.5
11.5
7.9
14.1
18.5
18.2
14.0
10.6
10.6
11.9
14.6
22.0
15.9
17.0
17.0
10.98
5.50
FRM
Mean
6.
6.
5.
4.
8
8.
9.
17.
10.
11.
10.
11.
10.
5.
8.
6.
10.
7.
6.
6.
6.
9.
13.
15.
24.
11.
15.
21.
20
5.
7.
16.
12.
10.
15.
19.
16.
14.
11.
12.
11.
14.
22
16.
18.
19.
3
8
7
4
4
0
5
1
7
5
7
0
2
5
4
1
8
3
2
3
8
4
4
8
4
0
2
1
0
3
1
0
1
0
5
6
6
1
1
9
6
5
2
1
1
1
CV
0.066
0.059
0.030
0.013
0.060
0.032
0.076
0.082
0.068
0.028
0.019
0.014
0.000
0.028
0.061
0.069
0.051
0.067
0.056
0.024
0.074
0.037
0.060
0.047
0.092
0.120
0.025
0.045
0.015
0.019
0.029
0.069
0.060
0.023
0.049
0.011
0.047
0.007
0.037
0.045
0.039
0.026
0.076
0.094
0.050
0.091
11.95
5.04
individual candidate
r
b
a
a bound
a bound
I
vs
0.966
1.107
-2.249
-4.1
0.4
mean FRM
0.938
| 1.020
-1 .330
-2.6
1.6
0.963
1.050
-1 .568
-3.1
1.2
mean vs
WHW
0.981
1.060
-1 .728
-3.3
1.1
candidates
Mean CV
4.7
5.1
4.1
2.4
7.0
6.4
8.0
16.5
9.3
10.7
10.2
10.2
8.7
3.5
5.7
5.0
8.6
5.9
4.7
5.1
5.8
8.5
12.7
17.5
20.7
10.6
13.8
19.3
20.9
3.7
6.0
16.3
10.3
8.9
14.0
17.6
16.8
16.0
9.6
11.5
10.4
13.9
21.4
17.6
19.6
18.0
10.94
5.45
mean
Battelle
0.981
1.058
-1 .692
-3.3
1.1
0.196
0.245
0.129
0.103
0.127
0.110
0.313
0.069
0.173
0.184
0.162
0.154
0.050
0.199
0.152
0.092
0.088
0.212
0.149
0.144
0.285
0.188
0.202
0.129
0.030
0.044
0.166
0.094
0.138
0.164
0.201
0.062
0.160
0.099
0.043
0.085
0.143
0.110
0.139
0.093
0.167
0.046
0.151
0.111
0.132
0.073
D-92
-------
Editorial notes (both on page A-4):
Equation A-l is slightly garbled. A radical sign should replace the extra brackets on the right
side of the denominator.
Five lines below the equation, I don't understand how Figure A-4 "shows ... the relative
insensitivity to the number of FRM samplers."
I would like to conclude by expressing appreciation for the care that went into this document.
While it was a tough slog in places, I came to trust that the difficulty arose from the subject itself
and not from any misdirection or carelessness by the authors. My compliments to them!
D-93
-------
An example of a successful FEM demonstration. Warren H. White, 9/19/05
The accompanying plot shows (hypothetical) test data from three candidate monitors that would
satisfy the FEM requirements proposed in Attachment 4.
45
40
35
30
25
20
o
E
JD
"co
;o
T3
8 15
10
D#1
X#2
A #3
A
A
10
15
20
25
30
mean FRM, ug/m3
The FRM concentrations are exactly those listed in Table 1 of Attachment 4. All three candidate
monitors are accurate (relative to the FRM) at concentrations below 16.5 ng/m3, but #1 and #2
become "pegged" at 16.5 ng/m3 when the true concentration goes higher. Fortunately for the
manufacturer, candidate #3 is three times more sensitive than it should be to increments above
|j,g/m3, so the mean of the three candidates is still accurate at high concentrations. The marketing
slogan is
"Buy three of our instruments and keep one in your garage!"
D-94
-------
Here is an embedded Excel file showing the calculations. They can be verified (or played with)
by double-clicking to get into the spreadsheet. Note that all four test statistics are better than
those of the original example in Attachment 4: precision = 14%, correlation =1, and absolutely
no multiplicative or additive bias.
candidates
un #1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
6.3
6.8
5.7
4.4
8.4
8
9.5
16.5
10.7
11.5
10.7
11
10.2
5.5
8.4
6.1
10.8
7.3
6.2
6.3
6.8
9.4
13.4
15.8
16.5
11
15.2
16.5
16.5
5.3
7.1
16
12.1
10
15.5
16.5
16.5
14.1
11.1
12.9
11.6
14.5
16.5
16.1
16.5
16.5
#2 #3
6.3
6.8
5.7
4.4
8.4
8
9.5
16.5
10.7
11.5
10.7
11
10.2
5.5
8.4
6.1
10.8
7.3
6.2
6.3
6.8
9.4
13.4
15.8
16.5
11
15.2
16.5
16.5
5.3
7.1
16
12.1
10
15.5
16.5
16.5
14.1
11.1
12.9
11.6
14.5
16.5
16.1
16.5
16.5
mean
RMS
6.3
6.8
5.7
4.4
8.4
8
9.5
18.3
10.7
11.5
10.7
11
10.2
5.5
8.4
6.1
10.8
7.3
6.2
6.3
6.8
9.4
13.4
15.8
40.2
11
15.2
30.3
27
5.3
7.1
16
12.1
10
15.5
25.8
16.8
14.1
11.1
12.9
11.6
14.5
33.6
16.1
21.3
24.3
FRM
Mean
6.3
6.8
5.7
4.4
8.4
8
9.5
17.1
10.7
11.5
10.7
11
10.2
5.5
8.4
6.1
10.8
7.3
6.2
6.3
6.8
9.4
13.4
15.8
24.4
11
15.2
21.1
20
5.3
7.1
16
12.1
10
15.5
19.6
16.6
14.1
11.1
12.9
11.6
14.5
22.2
16.1
18.1
19.1
11.95
5.04
CV
0.066
0.059
0.030
0.013
0.060
0.032
0.076
0.082
0.068
0.028
0.019
0.014
0.000
0.028
0.061
0.069
0.051
0.067
0.056
0.024
0.074
0.037
0.060
0.047
0.092
0.120
0.025
0.045
0.015
0.019
0.029
0.069
0.060
0.023
0.049
0.011
0.047
0.007
0.037
0.045
0.039
0.026
0.076
0.094
0.050
0.091
candidates maximum value for candidates 1 & 2: 16.5
Mean
6.3
6.8
5.7
4.4
8.4
8
9.5
17.1
10.7
11.5
10.7
11
10.2
5.5
8.4
6.1
10.8
7.3
6.2
6.3
6.8
9.4
13.4
15.8
24.4
11
15.2
21.1
20
5.3
7.1
16
12.1
10
15.5
19.6
16.6
14.1
11.1
12.9
11.6
14.5
22.2
16.1
18.1
19.1
11.95
5.04
CV
o n#1 X#2 A #3
0
0 45-,
0
0
0
0 40-
0.061
0
0
0 35-
0
0
0
0 30-
0 co
o t
0 =>
0 2 25 -
o
0 |
0 °
E
0 15 20-
T3
T3
0 §
0.561
0 15 "
0
0.378
0.303
0 10 "
0
0
0
0 5 "
0
0.274
0.010
A
A
A
A
A
A
A
A
Jff^
JjP-^
^jff^
*i)R*
jjp-"
jtf
^jjjjJT1
«K
Xr^
0 ' ' ' ' ' '
0 0 5 10 15 20 25 3
0 mean FRM, ug/m3
0
0
0.445
0 FRM candidates
0.153 precision 0.054 0.140
0.236
r 1 .000
b 1 .000
a 0.000
D-95
-------
Dr. Yousheng Zeng
Peer Review Comments on PMio_2.s FRM
By Yousheng Zeng
For Sept. 21-22, 2005 AAMM Subcommittee Peer Review and Consultation Meeting
Comments Revised after the Meeting
Peer Review Charge Question 1:
What are the scientific and operational strengths and weaknesses of the PMw-2.5
difference method relative to other options for a proposed FRM, especially when used
as the basis for approval of other methods?
Comment:
The PMio-2.5 difference method has significant operational strengths as summarized in
Attachment 1 in the provided review materials. I agree with EPA's analysis in
Attachment 1 and consider the difference method superior to other options. Its major
shortcoming is its inability to monitor continuously and automatically. Hopefully this
shortcoming can be overcome by future equivalent continuous monitors that are more
robust and more accurate than currently available continuous monitors. Eventually these
equivalent continuous monitors, rather than the reference method, will be widely
deployed in a similar fashion as continuous sulfur dioxide (862) monitors in relation to
the SO2 FRM.
By definition, the difference method has to be the most accurate one among all candidate
methods because all methods are compared against the PM metrics that is defined by the
underlying methods of the difference method. In this sense, it has scientific strength
within the context of current PM metrics or definition. However, from a fundamental
scientific viewpoint, the difference method is fairly weak.
The current PM metrics is defined by sampling methods rather than nature of the
pollutant (i.e., PM). PM2.5 is defined as the PM mass concentration measured by PM2.5
FRM and PMi0 by PMi0 FRM. Therefore, it makes most sense to define PMi0-2.5 as the
difference between the two. However, I would like to bring a different perspective into
this discussion. Although I recognize that EPA is in the final stage of promulgating new
PM standards and it is probably unrealistic to evaluate a new and fundamentally different
approach in this round of the rulemaking process, I recommend that EPA evaluate it in
the next cycle of the PM standard review.
We should rethink the conventional approach of defining PM by sampling methods. The
PM defined in this way appears to be a very poor indicator of PM as a pollutant. This is
not limited to PMio-2.5- It applies to PMio and PM2.s as well. The following examples
demonstrate this point.
D-96
-------
In these examples, I used simulated ambient PM samples similar to the ones specified in
40 CFR 53, Subpart F, Table F-3 and the fractionation curves of PM2.5 FRM and PMi0
FRM to calculate expected PM2.5, PMio, and PMio-2.s values as if the FRM samplers are
used to collect these simulated samples. I also used the deposition curves of PM in
human respiratory system in the EPA PM Criteria Document (CD), Oct. 2004 version,
Figure 6-13, to see what portion of the PM will deposit onto the respiratory system if a
person is inhaling the same ambient air. A more detailed description of this approach is
expected to be available soon as a paper is currently under peer review.1 An abbreviated
description of the approach is provided as Appendix A to my comment.
Figure 1 includes one set of examples showing the changes of different PM metrics as the
ambient PM condition changes (in this case, change of coarse mode PM concentration
while fine mode holds constant, which causes change in PM2.s/PMio ratio). This example
(and many more not presented here due to space consideration) illustrate that currently
used PM metrics (PM2.5 and PMio) and proposed PMio-2.s do not track the level of PM
that can potentially deposit onto the human respiratory system. Therefore, they do not
seem to be good indicators of PM as a pollutant. The scientific community is aware of
the mismatch between what is measured and what can be deposited onto human
respiratory system. The reason for the current PM metrics is due to lack of a system,
including measurement metrics and corresponding method, that can accurately measure
what can be deposited onto human respiratory system. A recently proposed system called
Comprehensive Particulate Matter Measurement System (CPMMS) and corresponding
dosimetry-based PM metrics have shown promising features in resolving this
fundamental issue l (also see Appendix A). The results from CPMMS are also included
in Figure 1 for comparison.
D-97
-------
(a) FRM Defined PM vs. Dosimetry-Based PM Metrics
- Total Respiratory System (TOT), Nasal Breathing (NB)
f 60
1
8 40
s
t M-
°V
~"\
^
o^ ^~*^^- ^~°~~~-~~~~—a
<^$^S-« . ^>^fe .
~~~—o-— ___
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0
PM2.5to PM10 Ratio (PM2.5 relatively constant)
iO
-4-PM-TOT(NB)True
PM-TOT(NB) CPMMS
-•-PM2.5FRM
-0-PMcFRM
-0-PM10FRM
C3) FRM Defined PM vs. Dosimetry-Based PM Metrics
- Extrathoracic (ET) region, Nasal Breathing (NB)
n
= 50
0
t M
Ck^^
~^ ^-"-^
\,
<^ ^^*^^» ^~D~ — " — °
~- ^- , n n i^^- ,„ , 0
^~~~~~^--^ *^--^.
•—-<>—-__
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.
PM2.5 to PM10 Ratio (PM2.5 relatively constant)
-A-PM-ET(NB)True
-«-PM-ET(NB) CPMMS
-•-PM2.5FRM
-o-PMcFRM
-0-PM10FRM
30
(c) FRM Defined PM vs. Dosimetry-Based PM Metrics
- Tracheobronchial (TB) region, Nasal Breathing (NB)
t 60-
1»
= 40
3
0.
^
0^ ° — - Q
t "n--,, nOOTt a a
~~---o^_
"""-—• o--___^
• — *- — •— • n • , .
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0
PM2.5 to PM10 Ratio (PM2.5 relatively constant)
to
-*-PM-TB(NB)True
-•- PM-TB(NB) CPMMS
-•— PM2.5FRM
-0— PMc FRM
-a— PM10 FRM
(d) FRM Defined PM vs. Dosimetry-Based PM Metrics
-Alveolar (A) region, Nasal Breathing (NB)
n
a
I80
0^ ~~° ; n
>-3~~.^-^. „
~° ^___
" —-. ____o
• * • — • • » • •
-*— PM-A(hB) True
-•— PM-A(NB) CPMMS
-•— PM2.5FRM
-*-PMcFRM
-D-PM10FRM
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80
PM2.5 to PM1 0 Ratio (PM2.5 relatively constant)
Figure 1. Changes of various PM metrics with changes in ambient PM25/PMi0 ratio. The legend "PM-TOT(NB)
True" in (a) stands for simulated true ambient PM that has potential to be deposited onto total respiratory system
based on the deposition curve for nasal breathing in EPA PM CD, Figure 6-13, Oct. 2004. PM-TOT(NB) CPMMS
stand for the same PM metrics obtained by a method called Comprehensive PM Measurement System.1 Similar
notations are used for different regions of the respiratory system, i.e., extrathoracic (ET) region, tracheobronchial
(TB) region, or alveolar (A) region, in (b), (c), and (d), respectively.
Figure 2 is another example of how proposed PMi0-2.5 can be a misleading indicator for
PM. Two ambient PM conditions, Case 1 and Case 2, are presented in Figure 2. The two
cases differ in their size distribution [see Figure 2 (a) and (b)]. If the proposed PMi0-2.5
FRM is used to monitor the two ambient conditions, the PMio-2.5 (as well as PMio and
PM2.s) will be the same. However, the PM levels that correspond to the portions that
have potential to be deposited onto human respiratory system [as a whole (i.e., in TOT)
or in extrathoracic (ET) region, tracheobronchial (TB) region, or alveolar (A) region] are
significantly different between the two cases, especially in the ET region or the total
respiratory system, which is the region that PMi0-2.5 intend to cover. In this example,
PMio-2.5 is about 32 |ig/m3 for both cases; but the real health effect [represented by PM-
ET(NB)] is equivalent to PM level of about 52 |ig/m3 for case 1 and about 78 |ig/m3 for
case 2.
D-98
-------
"
\
Particle Aerodynamic Diameter (um)
(b)
"
\
Particle Aerodynamic Diameter (um)
(c)
I Case 1 (Baseline)
I Case 2
90.0
80.0
FM-PM-FM-FM-FM-FM-PM-FM- FM2.5 FMc FFiM PM10
ET(NB) ET(I\B) TB(NB) TB(I\B) A(NB) A(I\B) TOT(NB) TOT(NB) FFiM FFiM
True CFWMS True CFWMS True CFWMS True CFWMS
Figure 2. Illustration of two ambient PM cases with the same PM25, PM10, and PM10.2.5 but different health effects;
(a) is particle size distribution for Case 1 and (b) for Case 2; (c) is comparison of different PM metrics for Cases 1
and 2.
During the September 21-22 meeting, significant amount of discussions were about the
evaporative loss from these candidate methods. As illustrated in Figure 2, the bias with
respect to health effects due to PM definitions is as a significant (or larger) problem as
evaporation loss. Most of us believe that we should measure what people in the health
effect field want to see. They probably would not want to see PM2.5, PMio, or PMio-2.s if
they were given dosimetry-based PM metrics.
We have become accustomed to think that PM metrics can only be defined by sampling
methods. I am afraid that we are missing the target (i.e., the PM that deposit onto the
respiratory system as illustrated in the above examples). Our current rules are based on
PM defined by FRM, but FRM itself seems an inadequate representative of the health
related PM portion. For this reason, I think a PM metrics I refer to as "dosimetry-based
PM metrics" is much more meaningful and valuable, especially this concept has be
demonstrated to be feasible through CPMMS1 (also see Appendix A). This approach
D-99
-------
would represent a major paradigm shift away from conventional PM metrics such as
PM2.s, PMio, and PMio-2.s. More thorough evaluation is needed and it will take some
time to fully develop the system. However, it makes much more scientific sense than the
current PM metrics system where cut points and the shape of fractionation curves are
somewhat arbitrarily set and do not represent health impact.
It is generally accepted that the steeper the sampler fractionation curve is, the better the
sampler is. This means that a perfectly vertical line at 2.5 ji or 10 ji would be ideal. If
we use a perfectly vertical line at 2.5 |i without any sampling errors (theoretic
conditions), the current PM2 5 FRM can produce up to 10% errors purely due to ambient
PM size distribution shift alone, which is expected to happen.1 On the other hand, there
is no reason to make a perfectly vertical line at 2.5 ji or 10 ji because it does not represent
anything. Nothing in the nature has such a sharp cut. Therefore, the question is what
slope will be desirable. The only answer is dosimetry-based PM metrics because in such
a system the "sampler fractionation curve" resembles the deposition curve of the human
respiratory system.
Peer Review Charge Question 2:
Based on the field study report as well as any other available data, e.g., data from State
and local agencies, how does the demonstrated data quality of the PMw-2.5 difference
method support or detract from it being proposed as a FRM?
Comment:
Based on review of the provided data, I support EPA's proposal to use the PMi0-2.5
difference method as a FRM, again within the context of current PM metrics and
regulatory framework (see my comment on Question 1). The data quality of the
difference method is higher than that of other candidate methods. The main disadvantage
of the difference method is its manual and non-continuous operation. If the EPA
envisions that the majority of deployed monitors will be automated continuous equivalent
methods rather than the FRM (similar to the current deployment of SO2 monitors), this
disadvantage can be minimized. As far as evaporative loss is concerned, the issue is not
unique to this method. If this issue needs to be addressed, the PM2.s FRM should also be
changed.
References
1. Y. Zeng, A Comprehensive Particulate Matter Monitoring System and Dosimetry-Based
Ambient Parti culate Matter Standards, J. of Air & Waste Management Association, in review.
D-100
-------
Appendix A to Yousheng Zeng's Comments
Dosimetry-Based PM Metrics
This Appendix contains a brief description of dosimetry -based PM metrics. A more detailed
description can be found in Reference 1 .
Dosimetry-Based PM Metrics
Currently ambient PM level is measured as PM2 5 and PMi0. They are supposed to be surrogate
parameters to indicate potential human exposure to the pollutant (i.e., PM). The definitions of
PM2.s and PMi0 are based on sampler cut point diameter (D50). How well they represent the PM
that actually deposits in the human respiratory system is questionable.
A dosimetry-based PM definition is proposed to reflect the ambient PM level that may cause
human health effects due to deposition of PM in the respiratory system. It is expressed as:
Where
CD = Dosimetry-based ambient PM concentration, |ig/m3
d(i) = Respiratory tract (or a region of it) deposition fraction on a mass basis for
size interval i, unitless.
C(i) = Ambient PM interval mass concentration for size interval i, |ig/m3
The parameter d(i) is selected depending on the interest of a study or the measurement objective
and is fix for that purpose. For example, if the measurement is about PM deposition in the total
respiratory tract using adult male nasal breathing as a benchmark, the corresponding total particle
deposition fraction [TOT(NB)J as documented in the recent EPA Criteria Document for PM 2
and illustrated in Figure 1 can be used for d(i). The resulting dosimetry-based PM concentration
(CD) can be designated as PM-TOT(NB).
D-101
-------
1.0 10.0
Particle Aerodynamic Diameter (urn)
Figure 1. Total deposition in respiratory tract (TOT) for an adult male, nasal breathing (data
based on Figure 6-13 in Ref. 2).
The other parameter in the right-hand side of equation (1), CQ, is ambient PM mass size
distribution. The following PM measurement system is proposed to significantly improve the
accuracy of PM mass size distribution measurement with more tolerable monitoring instruments
(therefore promote a wide spread use of ambient PM size distribution).
Comprehensive Particulate Matter Measurement System (CPMMS)
A system called Comprehensive Particulate Matter Measurement System (CPMMS) is developed
and is illustrated in Figure 2. It consists of the following hardware and software:
• An aerosol particle sizing device, e.g., Aerodynamic Particle Sizer (APS) Model 3321
manufactured by TSI Inc., or some other particle size distribution measurement methods;
• A mass-based PM sampler, e.g., PM2.5 Federal Reference Method (FRM), PMio FRM, or
a continuous dichotomous sampler; and
• The algorithm described below.
D-102
-------
PM in
Ambient
Air
Mass-based PM Sampling
Particle
Sizing
PM
Sampling Device
Simulator
PM
Cone.
Off-scale
PM
Con
Density
Correction
i
Initial
Particle
Size
Distribution
PM Mass
Size
Distribution
Figure!. Schematic of CPMMS.
The particle sizing device and the PM sampler are co-located and measuring the same ambient
PM sample. The PM sampler measures the mass concentration (typically in |ig/m3) of the PM
fraction that can be collected by the sampler based on the sampling effectiveness curve
(aspiration curve or fractionation curve) of the sampler. The result is an aggregated mass
concentration over the sampler's designed size range (e.g., 0-2.5 jim for FRM PM2.5).
The particle sizing device measures initial particle size distribution in |ig/m3 for each particle
size interval (e.g., 0-0.523 |im, 0.523-0.542 |im, 0.542-0.583 |im, etc.). The measurement is
based on aerodynamic properties of the particles and the results in |ig/m3 are derived based on an
assumed particle density. This arbitrary assumption makes the results unreliable in the absolute
sense of |ig/m3 and unsuitable for estimating an aggregated mass concentration over a particle
size range per regulatory definition. However, the relative |ig/m3 values between particle size
intervals are much more reliable and can represent the relative particle size mass distribution. In
CPMMS, the accurate aggregated mass concentration result obtained by the mass-based PM
sampler is used to "calibrate" the results of the particle sizing device to produce accurate PM
monitoring data in terms of mass size distribution and aggregation over any size ranges. In order
to accomplish the "calibration", the following model is used to simulate the mass-based PM
samplers.
A PM sampler can be characterized by its sampling effectiveness curve. The sampling
effectiveness curves of PM2.5 FRM and PMio FRM based on the data points specified in 40 CFR
53 3 are depicted in Figure 3 (a).
D-103
-------
(a)
C
o
I 70%
m
CL
Q)
c 40A
m -30*
LU
•»AAAA
*
A «
.
4 »
A FRM PM2.5
A *
A
A
A
0.1 1.0 10.0 100.0
Particle Aerodynamic Diameter (um)
(b)
» FRMPM10
— Simulated FRMPM10
A FRMPM2.5
Simulated FRM PM2.5
1.0 10.0
Particle Aerodynamic Diameter (um)
100.0
Figure 3. Sampling effectiveness curves of mass-based PM samplers: (a) PMio FRM and PM2.5
FRM based on specifications in 40 CFR 53 Subpart D Table D-3 and Subpart F Table F-4,
respectively; (b) data in (a) plus sampling effectiveness curves simulated using the model
presented in this Appendix.
D-104
-------
The sampling effectiveness curves can be modeled using the following equation.
(2)
Where
PQ = Sampling effectiveness (or penetration) for particles in size interval i, % or
fraction
DQ = Representative aerodynamic diameter of particles in size interval i, |im
D50 = Sampler cutpoint diameter (particle aerodynamic diameter corresponding to
50% penetration), |im
n = Parameter determining the steepness of the sampling effectiveness curve
PM samplers can be simulated by the above model by selecting proper values of two parameters,
DSQ and n, using the least squares (LS) approach. The accuracy of the simulation can be
evaluated by virtually passing the idealized ambient particles as specified in 40 CFR 53 3 through
the model (i.e., simulated sampler) and calculating the sum of the squared deviations of the
model-predicted values from the true values specified in 40 CFR 53. The LS results are
normalized to be expressed as percent of the total PM mass of idealized ambient particle
samples. The model parameters and simulation accuracies are summarized in Table 1. The
accuracy of 0.043% in Table 1 suggests that the model represents FRM PM2.5 very well. This
excellent fit between the model and the true values can be seen graphically in Figure 3 (b). The
simulated FRM PMio sampler is less ideal and the accuracy is 3.864% [also see the fit curve in
Figure 3 (b)]. This is consistent with the well known fact that the FRM PMio is loosely defined
compared to FRM PM2.5.
Table 1. Summary of modeled FRM PM2 5 and FRM PMio samplers
Modeled FRM PMio Sampler
Modeled FRM PM2.5 Sampler
Inlet (10 |im)
Fractionator (2.5 |im)
Model Parameters
D50 (|im)
10.0
10.2
2.5
n
4.6
4.8
10
Simulation
Accuracy a
3.864%
0.043%
a. The sum of the squared deviations of the true values specified in 40 CFR 53 and the model-predicted
values, expressed as percent of the total PM mass of idealized ambient particle samples.
D-105
-------
With the PM sampling device simulator established as eq (2), the initial particle size distribution
data can be processed by the simulator. The initial particle size distribution data consist of
estimated mass concentrations in |ig/m3 for each particle size interval. However, the |ig/m3
value may not be correct because it is based on an assumed density for PM. At this step of the
process, we can ignore the accuracy issue of the assumed density and virtually put the PM
sample through the simulator. The simulator will predict the PM mass collected on the simulated
sampler based on the assumed density as follows:
(i)cKi) (3)
i i
Where
M = Predicted PM mass concentration seen by the simulated PM sampler, |ig/m3
m(i) = Predicted PM interval mass concentration as collected on the sampler filter
(i.e., passed sampler inlet cut) for particle size interval i, |ig/m3
Ci(i) = Initial interval mass concentration measured by the particle sizing device
(i.e., ambient PM size distribution before passing through the mass based
PM sampler) for size interval i, |ig/m3
If the initial particle size distribution, CIQ, is accurate (i.e., the assumption of particle density
happens to be correct), the value of M should match the PM concentration actually measured by
the co-located PM sampler. Otherwise, the correct particle density can be derived using the
following equations.
Where
p = Correct particle density, g/cm
PA = Assumed particle density in the initial particle mass size distribution, g/cm3
C = Actual PM concentration measured by the PM sampler, |ig/m3
When APS is used to measure particle size distributions, it actually measures the aerodynamic
diameters of individual particles using time of flight technology. The aerodynamic diameters are
used to calculate the volumes of the particles to obtain the volume size distribution. The mass
size distribution is finally calculated from the volume size distribution by assuming a particle
density:
cI(l)=106v(l)pA (5)
Where
106 = Conversion factor, 106 |ig/g
VQ = Interval volume concentration measured by the particle sizing device for
size interval i, cm3/m3 (vol. of parti cles/vol. of air sample)
D-106
-------
If the actual density of particles is known (instead of assumed), an accurate mass size distribution
should be:
c(l)=106v(i)p (6)
Where
C(i) = Actual interval mass concentration for size interval i, |ig/m3
Combining eqs (4), (5), and (6) can yield the following equation that can be used to calculate
actual mass size distributions:
Once an accurate mass size distribution (c;) is obtained, any aggregated ambient PM monitoring
data can be derived by summing up the c; over the desired size range:
For example, if the above summation is done for intervals of mass c; representing particle
diameters from 0 to 2.5 |im, the resulting C is for PM^.s; if the diameter range is from 2.5 to 10
|im, the result is PMc or PMi0-2.5, etc.
If it is desired to have PM results that mimic a particular PM sampler, the accurate mass size
distribution (c;) produced by CPMMS should again be put through a simulator of a desired PM
sampler with sampling effectiveness curve represented by p; in the same fashion as described in
eq (3), i.e.,
CPMMS relies on field measurements by a particle sizing device and a mass-based PM sampler.
If the sampler is a FRM sampler, the CPMMS is referred to as CPMMS(FRM). If the sampler is
a Dichot sampler, the CPMMS is called CPMMS(Dichot).
As shown above, once accurate mass size distribution is obtained [eq. (7)], PM measurement
metrics can be defined (and reconstructed) by exact cut point [eq. (8)], sampler sampling
effectiveness curve [eq. (9)], or PM deposition fraction of human respiratory system [eq. (1)].
The last one is dosimetry-based PM metrics and is most relevant and meaningful.
If CPMMS is implemented as a new way of managing ambient air quality for PM, the results
will be high quality mass size distribution data. The monitoring data will be available to
regenerate various dosimetry-based PM concentrations for different research purposes, e.g., an
epidemiological study on ambient PM level defined by particle deposition in the
tracheobronchial region of the respiratory tract using eq (1) and the related health impact. The
D-107
-------
study based on such PM data will be more meaningful than current PM data. Regulatory
authority can also establish dosimetry-based PM standards using human respiratory PM
deposition data that are representative of and protective to selected groups of population (e.g.,
children).
Major Features of CPMMS and Dosimetry-Based PM Metrics
The major features of the above system include:
• It closely matches human health effects and there are no arbitrary cut points. This PM
metrics system is expected to have much stronger correlation with epidemiological data
than PMio, PM2.5, or PMi0-2.5.
• Unlike past changes from TSP to PMio, to PM2.5, etc.), the monitoring data obtained
through CPMMS dose not lose its value when regulatory definition of PM changes. The
sampling devices do not need to be changed even when the PM definition changes.
• It is significantly less vulnerable to imperfection in sampler design, manufacturing, and
operating conditions.
Specific discussions and supporting examples can be found in Ref. 1.
References
1. Y. Zeng, A Comprehensive Particulate Matter Monitoring System and Dosimetry-Based
Ambient Parti culate Matter Standards, J. of Air & Waste Management Association, in review.
2. U.S. Environmental Protection Agency. Air Quality Criteria for Parti culate Matter.
EPA/600/P-99/002bF, October 2004.
3. U.S. Environmental Protection Agency. Ambient Air Monitoring Reference and Equivalent
Methods. Code of Federal Regulations 40, Part 53.
D-108
-------
NOTICE
This report has been written as part of the activities of the U.S. Environmental
Protection Agency's (EPA) Clean Air Scientific Advisory Committee (CASAC), a
Federal advisory committee administratively located under the EPA Science Advisory
Board (SAB) Staff Office that is chartered to provide extramural scientific information
and advice to the Administrator and other officials of the EPA. The CAS AC is
structured to provide balanced, expert assessment of scientific matters related to issue
and problems facing the Agency. This report has not been reviewed for approval by the
Agency and, hence, the contents of this report do not necessarily represent the views and
policies of the EPA, nor of other agencies in the Executive Branch of the Federal
government, nor does mention of trade names or commercial products constitute a
recommendation for use. CASAC reports are posted on the SAB Web site at:
http://www.epa.gov/sab.
D-109
------- |