UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
October 24, 2006
EPA-CASAC-07-001
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Clean Air Scientific Advisory Committee's (CASAC) Peer Review of the
Agency's 2nd Draft Ozone Staff Paper
Dear Administrator Johnson:
EPA is in the process of reviewing the national ambient air quality standards (NAAQS)
for ozone (Os) and related photochemical oxidants, which the Agency most recently revised in
July 1997. As part of its ongoing review of the ozone NAAQS, EPA's Office of Air Quality
Planning and Standards (OAQPS) developed a 2nd Draft Ozone Staff Paper, entitled, Review of
the National Ambient Air Quality Standards for Ozone: Policy Assessment of Scientific and
Technical Information (July 2006). At the request of the Agency, EPA's Clean Air Scientific
Advisory Committee (CASAC or Committee), supplemented by subject-matter-expert panelists
— collectively referred to as the CASAC Ozone Review Panel (Ozone Panel) — met in a public
meeting in Durham, NC, on August 24-25, 2006, to conduct a peer review of this draft Ozone
Staff Paper and three related draft technical support documents.
In its summary of EPA staff conclusions on the primary (health-related) ozone NAAQS
found in Chapter 6 of the 2nd Draft Ozone Staff Paper, OAQPS set-forth two options with regard
to revising the level and the form of the standard: (1) retain the current primary eight-hour (8-hr)
NAAQS of 0.08 parts per million (ppm); or (2) consider a reduction in the level of the primary
Os NAAQS within the range of alternative 8-hr standards included in Staffs exposure and risk
assessments (which included a range from 0.064 to 0.084 ppm) with primary focus on an Os
level of 0.07 ppm with a range of forms from third- through fifth-highest daily maximum. The
Ozone Panel unanimously concludes that:
1. There is no scientific justification for retaining the current primary 8-hr NAAQS of 0.08
parts per million (ppm), and
-------
2. The primary 8-hr NAAQS needs to be substantially reduced to protect human health,
particularly in sensitive subpopulations.
Therefore, the CASAC unanimously recommends a range of 0.060 to 0.070ppm for the
primary ozone NAAQS. With regard to the secondary (welfare-related) ozone NAAQS, the
Ozone Panel is in strong agreement with the scientific and technical evidence presented in the
summary of EPA staff conclusions on the secondary ozone NAAQS found in Chapter 8 of the
draft Staff Paper in support of the alternative secondary standard of cumulative form that
extends over an entire growing season.
The Ozone Panel members agree that this letter adequately represents their views. The
chartered Clean Air Scientific Advisory Committee fully endorses the Panel's letter and hereby
forwards it to you as the Committee's consensus report on this subject. A discussion of each
chapter in the 2nd Draft Ozone Staff Paper follows this letter, and the comments of individual
Panel members on the 2nd Draft Ozone Staff Paper and three related draft technical support
documents are attached as Appendix D.
1. Background
Section 109(d)(l) of the CAA requires that the Agency periodically review and revise, as
appropriate, the air quality criteria and the NAAQS for the "criteria" air pollutants, including
ambient ozone. Pursuant to sections 108 and 109 of the Act, EPA is in the process of reviewing
the ozone NAAQS. OAQPS, within the Office of Air and Radiation (OAR), developed the 2nd
Draft Ozone Staff Paper as part of this activity. In February 2006, the Agency's National Center
for Environmental Assessment, Research Triangle Park, NC (NCEA-RTP), within the Agency's
Office of Research and Development (ORD), released its final Air Quality Criteria for Ozone
and Related Photochemical Oxidants, Volumes I, II, and III, (EPA/600/R-05/004aF-cF, Final
Ozone Air Quality Criteria Document) for this current review cycle for the ozone NAAQS. The
2nd Draft Ozone Staff Paper evaluates the policy implications of the key scientific and technical
information contained in the Final Ozone AQCD and identifies critical elements that the Agency
believes should be considered in its review of the ozone NAAQS. The Ozone Staff Paper is
intended to "bridge the gap" between the scientific review contained in the Ozone AQCD and
the public health and welfare policy judgments required of the EPA Administrator in reviewing
the ozone NAAQS.
The Ozone Panel met in a public meeting on December 8, 2005 to conduct a consultation
on EPA's 1st Draft Ozone Staff Paper and two related technical support documents. However,
given that the OAQPS' first draft Staff Paper did not contain Agency staff conclusions about
whether to retain or revise the existing primary and secondary Ozone standards, the CASAC's
activity only amounted to a technical assessment of that document. The Committee's letter to
you from that meeting (EPA-CASAC-CON-06-003), dated February 16, 2006, is posted at URL:
http://www.epa.gov/sab/pdf/casac con 06 003.pdf.
-------
2. CASAC Ozone Review Panel's Peer Review of the 2nd Draft Ozone Staff Paper and
Related Technical Support Documents
The Ozone Panel reviewed the 2nd Draft Ozone Staff Paper and found it improved over
the earlier version that had been reviewed as part of a consultation process. However, the Panel
did not agree with the EPA staff conclusions that it was appropriate to consider retaining the
current NAAQS as an option that would be protective of public health and welfare. The Ozone
Panel's recommendations for reducing the level of the primary ozone standard, and its rationale
for these recommendations, are provided immediately below. Following a detailed discussion on
the primary and secondary NAAQS are the Panel's major, chapter-specific comments. Finally,
the individual written comments of Ozone Panel members on the 2n Draft Ozone Staff Paper
and the three related draft technical support documents are attached in Appendix D. Panelists'
responses to the Agency's charge questions are included in these individual review comments.
Primary Ozone NAAQS
New evidence supports and build-upon key, health-related conclusions drawn in the 1997
Ozone NAAQS review. Indeed, in the 2nd Draft Ozone Staff Paper, EPA staff themselves arrived
at this same conclusion:
"Based on the above considerations and findings from the [Final Ozone AQCD], while being
mindful of important remaining uncertainties, staff concludes that the newly available
information generally reinforces our judgments about causal relationships between O3 exposure
and respiratory effects observed in the last review and broadens the evidence of O3 -related
associations to include additional respiratory-related endpoints, newly identified cardiovascular-
related health endpoints, and mortality. Newly available evidence also has identified increased
susceptibility in people with asthma. While recognizing that important uncertainties and research
questions remain, we also conclude that progress has been made since the last review in
advancing our understanding of potential mechanisms by which ambient O3, alone and in
combination with other pollutants, is causally linked to a range of respiratory- and cardiovascular-
related health endpoints." (Pages 6-6 and 6-7)
Several new single-city studies and large multi-city studies designed specifically to
examine the effects of ozone and other pollutants on both morbidity and mortality have provided
more evidence for adverse health effects at concentrations lower than the current standard. (See
the numerous ozone epidemiological single-city studies shown in Figure 3-4 on page 3-53 of the
2nd Draft Staff Paper and, in addition, Appendix 3B of the staff paper, which contains the
summary of effect estimates and air quality data for these studies and multi-city epidemiological
studies.) These studies are backed-up by evidence from controlled human exposure studies that
also suggest that the current primary ozone NAAQS is not adequate to protect human health
(Adams, 2002; McDonnell, 1996).
Furthermore, we have evidence from recently reported controlled clinical studies of
healthy adult human volunteers exposed for 6.6 hours to 0.08, 0.06, or 0.04 ppm ozone, or to
filtered air alone during moderate exercise (Adams, 2006). Statistically-significant decrements
in lung function were observed at the 0.08 ppm exposure level. Importantly, adverse lung
function effects were also observed in some individuals at 0.06 ppm (Adams, 2006). These
-------
results indicate that the current ozone standard ofO. 08 ppm is not sufficiently health-protective
with an adequate margin of safety. It should be noted these findings were observed in healthy
volunteers; similar studies in sensitive groups such as asthmatics have yet to be conducted.
However, people with asthma, and particularly children, have been found to be more sensitive
and to experience larger decrements in lung function in response to ozone exposures than would
healthy volunteers (Mortimer et a/., 2002).
Going beyond spirometric decrements, adverse health effects due to low-concentration
exposure to ambient ozone (that is, below the current primary 8-hour NAAQS) found in the
broad range of epidemiologic and controlled exposure studies cited above include: an increase in
school absenteeism; increases in respiratory hospital emergency department visits among
asthmatics and patients with other respiratory diseases; an increase in hospitalizations for
respiratory illnesses; an increase in symptoms associated with adverse health effects, including
chest tightness and medication usage; and an increase in mortality (non-accidental,
cardiorespiratory deaths) reported at exposure levels well below the current standard. The
CASAC considers each of these findings to be an important indicator of adverse health effects.
As demonstrated in Chapter 5 of the 2nd Draft Ozone Staff Paper (specifically, Figures 5.5, 5.7,
5.8, and 5.9), a significant decrease in adverse effects due to ozone exposures can be achieved by
lowering the exposure concentrations below the current standard, which is effectively 0.084
ppm. Beneficial effects in terms of reduction of adverse health effects were calculated to occur
at the lowest concentration considered (i.e., 0.064 ppm). (See also Figure 3-4, "Effect estimates
(with 95% confidence intervals) for associations between short-term ozone exposure and
respiratory health outcomes," on page 3-53.)
The justification provided in the 2nd Draft Ozone Staff Paper for retaining the current
level of the primary ozone standard as an option for the Administrator was based on results of
controlled human exposure studies measuring modest declines in FEVi after exposures to 0.08
ppm ozone. However, as stated in the Staff Paper (page 3-6), while average decrements in the
FEVi were relatively small, 26% of the subjects had greater than 10% decrements, which can be
clinically significant. Also, while measures of FEVi are quantitative and readily obtainable in
humans, they are not the only measures — and perhaps not the most sensitive measures — of the
adverse health effects induced by ozone exposure. As stated on page 6-32 of the Final Ozone
AQCD, "Spirometric responses to ozone are independent from inflammatory responses and
markers of epithelial injury (Balmes et a/., 1996; Bloomberg et a/., 1999; Hazucha etal., 1996;
Torres et a/., 1997). Significant inflammatory responses to ozone exposures that did not elicit
significant spirometric responses have been reported (Holz et a/., 2005; McBride et a/., 1994)."
Agency staffs analyses placed most emphasis on spirometric evidence and not enough emphasis
on serious morbidity (e.g., hospital admissions) and mortality observed in epidemiology studies
(see page 6-44).
Therefore, on the basis of the large amount of recent data evaluating adverse health
effects at levels at and below the current NAAQS for ozone, it is the unanimous opinion of the
CASAC that the current primary ozone NAAQS is not adequate to protect human health.
Furthermore, the Ozone Panel is in complete agreement both that: the EPA staff conclusion in
Section 6.3.6 arguing that "consideration could be given to retaining the current 8-hr ozone
standard" is not supported by the relevant scientific data; and that the current primary 8-hr
-------
standard of 0.08 ppm needs to be substantially reduced to be protective of human health,
particularly in sensitive subpopulations.
Additionally, we note that the understanding of the associated science has progressed to
the point that there is no longer significant scientific uncertainty regarding the CASAC 's
conclusion that the current 8-hr primary NAAQS must be lowered. A large body of data clearly
demonstrates adverse human health effects at the current level of the 8-hr primary ozone
standard. Retaining this standard would continue to put large numbers of individuals at risk for
respiratory effects and/or significant impact on quality of life including asthma exacerbations,
emergency room visits, hospital admissions and mortality. (Scientific uncertainty does exist with
regard to the lower level of ozone exposure that would be fully-protective of human health. The
Ozone Panel concludes that it is possible that there is no threshold for an ozone-induced impact
on human health and that some adverse events may occur at policy-relevant background.)
Moreover, EPA staff concluded that changes in the concentration-based form of the
standard (i.e., whether to use the third-, fourth-, or fifth-highest daily maximum 8-hr average
concentration) should also be considered. The analysis found in the 2nd Draft Ozone Staff Paper
indicates that modest changes in the form of the standard can have substantial impacts on the
frequency of adverse health effects. Therefore, the CASAC recommends that the Agency
conduct a broader evaluation of alternative concentration-based forms of the primary 8-hr ozone
standard and the implications of those alternative forms on public-health protection and stability
(i.e., with respect to yearly variability to ensure a stable target for control programs).
The CASAC further recommends that the ozone NAAQS should reflect the capability of
current monitoring technology, which allows accurate measurement of ozone concentrations
with a precision of parts per billion, or equivalently to the third decimal place on the parts-per-
million scale. In addition, given that setting a level of the ozone standard to only two decimal
places inherently reflects upward or downward "rounding," e.g., 0.07 ppm includes actual
measurements from 0.0651 ppm to 0.0749 ppm, the CASAC chooses to express its
recommended level, immediately below, to the third decimal place.
Accordingly, the CASAC unanimously recommends that the current primary ozone
NAAQS be revised and that the level that should be considered for the revised standard be from
0.060 to 0.070 ppm, with a range of concentration-based forms from the third- to the fifth-
highest daily maximum 8-hr average concentration. While data exist that adverse health effects
may occur at levels lower than 0.060 ppm, these data are less certain and achievable gains in
protecting human health can be accomplished through lowering the ozone NAAQS to a level
between 0.060 and 0.070 ppm.
Secondary Ozone NAAQS
An important difference between the effects of acute exposures to ozone on human health
and the effects of ozone exposures on welfare is that vegetation effects are more dependent on
the cumulative exposure to, and uptake of, ozone over the course of the entire growing season
(defined to be a minimum of at least three months). Therefore, there is a clear need for a
-------
secondary standard which is distinctly different from the primary standard in averaging time,
level and form. Developing a biologically-relevant ozone air quality index would be directly
responsive to the 2004 National Research Council (NRC) recommendations on Air Quality
Management in the United States (NAS, 1994) and will help support important new Agency
initiatives to enhance ecosystem-related program tracking and accountability.
In its 1996 review of the ozone NAAQS, EPA staff proposed several cumulative seasonal
ozone exposure indices, including SUM06, the concentration-weighted metric (i.e., the seasonal
sum of all hourly average concentrations > 0.06 ppm), and W126, the integrated exposure index
with a sigmoidal weighting function, as candidates for a secondary standard. The Administrator
considered a three-month, 12-hr SUM06 secondary standard at a level of 25 ppm-hr as an
appropriate, biologically-relevant secondary standard, but ultimately rejected this option in favor
of simply setting the secondary standard equal to the primary. It was rationalized that efforts to
attain the new 8-hr primary standard would also eliminate most adverse effects on vegetation,
and at that time there were uncertainties in how cumulative seasonal exposures would change
with efforts to reduce peak 8-hour concentrations. Additionally, it was assumed that future
ozone/vegetation effects research over the coming years would clarify the very uncertain
quantitative relationships between ozone exposures and vegetation/ecological responses under
ambient field conditions.
Unfortunately, however, the Agency has supported very little new vegetation/ecological
ozone effects research over the past decade. The net result is that the quantitative evidence
linking specific ozone concentrations to specific vegetation/ecological effects must continue to
be characterized as having high uncertainties due to the lack of data for verification of those
relationships. It is not surprising that substantial research needs remain, as indicated both in
Chapter 8 and in individual reviewer comments. The quantitative evidence linking specific
ozone concentrations to specific vegetation effects — especially at the complex ecosystem level
— must continue to be characterized as having high uncertainties due to the lack of data for
verification of those relationships. To a large extent, this is an unavoidable consequence of the
inherent complexities of ecosystem structure and function, interactions among biotic and abiotic
stressors and stimuli, variability among species and genotype, detoxification and compensatory
mechanisms, etc. Nevertheless, the compelling weight of evidence provided in Chapter 7 of the
2nd Draft Ozone Staff Paper results from the convergence of results from many various and
disparate assessment methods including chamber and free air exposure, crop yield and tree
seedling biomass experimental studies, foliar injury data from biomonitoring plots, and modeled
mature tree growth.
Despite limited recent research, it has become clear since the last review that adverse
effects on a wide range of vegetation including visible foliar injury are to be expected and have
been observed in areas that are below the level of the current 8-hour primary and secondary
ozone standards. Such effects are observed in areas with seasonal 12-hr SUM06 levels below 25
ppm-hr (the lower end of the range of a SUM06 secondary standard suggested in the 1996
review and the upper end of the range suggested in Chapter 8 of the 2nd Draft Ozone Staff
Paper). Seasonal SUM06 (or equivalent W126) ranges well below 25 ppm-hr were
recommended for protecting various managed and unmanaged crops and tree seedlings in the
1997 workshop on secondary ozone standards (Heck and Cowling, 1997). The absence of clear-
-------
cut lower effects thresholds for sensitive vegetation combined with the lower recent estimates of
policy-relevant background (typical range of 0.015 to 0.035 ppm) emphasizes the importance of
efforts to reduce low- to mid-range environmental exposures below 0.060 ppm.
Based on the Ozone Panel's review of Chapters 7 and 8, the CASAC unanimously agrees
that it is not appropriate to try to protect vegetation from the substantial, known or anticipated,
direct and/or indirect, adverse effects of ambient ozone by continuing to promulgate identical
primary and secondary standards for ozone. Moreover, the members of the Committee and a
substantial majority of the Ozone Panel agrees with EPA staff conclusions and encourages the
Administrator to establish an alternative cumulative secondary standard for ozone and related
photochemical oxidants that is distinctly different in averaging time, form and level from the
currently existing or potentially revised 8-hour primary standard. The suggested approach to the
secondary standard is a cumulative seasonal growing standard such as the indices SUM06 or
W126 aggregated over at least the three summer months exhibiting the highest cumulative ozone
levels and includes the ozone exposures from at least 12 daylight hours. The CASAC suggests a
range of 10 to 20 ppm-hours for the three-month growing season SUM06 index for agricultural
crops rather than the 15-25 ppm-hours proposed in Chapter 8.
However, the Ozone Panel views the three-month growing season W126 index as a
potentially more biologically-relevant index than the 3-month growing season SUM06 index.
This is because the W126 index has no absolute minimum ozone concentration threshold and
only lightly weights the lower ozone concentrations. Therefore, a three-month seasonal W126
that is the approximate equivalent of the SUM06 at 10 to 20 ppm-hr is preferred. As shown by
the references cited at the end of Chapter 8, the consensus view among expert persons in the
ecological communities of both this country and elsewhere around the world is that a secondary
standard of cumulative form and extending over an entire growing season will be far more
effective than a secondary standard that is not cumulative inform and does not include the whole
growing season.
In conclusion, the Clean Air Scientific Advisory Committee is pleased to provide its
scientific advice and recommendations to the Agency on the primary and secondary ozone
NAAQS. We recognize that our recommendation of lowering of the current primary ozone
standard would likely result in a large portion of the U.S. being in non-attainment. Nevertheless,
we take very seriously the statutory mandate in the Clean Air Act not only for the Administrator
to establish, but also for the CASAC to recommend to the Administrator, a primary standard that
provides for an "adequate margin of safety ... requisite to protect the public health. "
Finally, as announced during the Ozone Panel's August meeting, once the Agency
releases the Final Ozone Staff Paper in early January 2007, the CASAC intends to hold a public
teleconference in late January or early February 2007 for the members of the Ozone Panel to
review — and, prospectively, to offer additional, unsolicited advice to the Agency concerning —
Chapter 6 (Staff Conclusions on Primary Os NAAQS) and Chapter 8 (Staff Conclusions on
Secondary Os NAAQS) in that final Agency document. The purpose of such advice would be to
-------
inform EPA's efforts as it develops the forthcoming, proposed rule for ozone and related
photochemical oxidants. As always, the CASAC wishes EPA well in this important endeavor.
Sincerely,
/Signed/
Dr. Rogene Henderson, Chair
Clean Air Scientific Advisory Committee
Appendix A - Clean Air Scientific Advisory Committee Roster (FY 2006)
Appendix B - CASAC Ozone Review Panel Roster
Appendix C - Charge to the CASAC Ozone Review Panel
Appendix D - Review Comments from Individual CASAC Ozone Review Panel Members
-------
CASAC Chapter-Specific Discussion Comments on
EPA's 2nd Draft Ozone Staff Paper
Sub-groups of the CASAC Ozone Review Panel who led the discussion on individual
chapters of the Staff Paper summarized their comments in the following paragraphs:
Chapter 2 (Air Quality Characterization): A better introduction to the central role of
photochemical oxidation reactions as the key reactions governing the behavior of air pollutants
in the atmosphere would improve this chapter. Ozone is the key indicator of the extent of
oxidative chemistry and serves to integrate multiple pollutants. Oxidation in the atmosphere
leads to the formation of particulate matter from SO2, NOx, and volatile organic compounds
(VOCs) as well as gas phase irritants (formaldehyde, acrolein, etc). Thus, although ozone itself
has direct effects on human health and ecosystems, it can also be considered as indicator of the
mixture of photochemical oxidants and of the oxidizing potency of the atmosphere. Section 2.2.6
only briefly covers the relationship of ozone to other photochemical oxidants. It would be
beneficial to add a short paragraph outlining the role of ozone and other photochemical oxidants
in the atmospheric transformation processes that may results in the formation of more toxic
products (both in an outdoor and indoor environment), as provided in the individual comments
appended to this letter.
The section on policy-relevant background (2.7) continues to have problems. Although
the section briefly cites the results of comparison of different models and measurements, it does
not adequately address the uncertainties of the global GEOS-CHEM model, and how these
uncertainties are reflected in the health risk analysis. Since ozone health effects are observed
down to concentrations of the order of 0.04-0.05 ppm, it is important to know how the PRB is
related to the considered primary ozone standard and what uncertainties there are in the risk
attributed to controllable sources.
Chapter 3 (Policy-Relevant Assessment of Health Effects Evidence): The latest draft
of Chapter 3 is much improved over the previous draft. Efforts to respond to some of the earlier
concerns expressed by the CASAC are appreciated. While in general this chapter is well written,
and is a credible basis for the risk analyses that follow, there are inconsistencies and inaccuracies
that still need to be addressed. Typically, there is appropriate use of cautionary phrases when the
data are not as strong as they might be, but this use is inconsistent across the chapter, and there
are instances where EPA staff appear to be stretching to infer that data support their statement.
While the individual comments of Ozone Panel members attached to this letter provide specifics
on these points, some of the Panel's more significant concerns are discussed briefly below.
Discussion of measurement error is convoluted, confusing, and contains some mistakes.
The primary issue in the use of central ambient monitors for ozone in time-series
epidemiological studies is whether they provide any information at all that reflects daily personal
ozone exposure in susceptible populations. The discussion on p. 3-37 of the impact of various
types of exposure measurement error is incorrect; the difference between true and measured
ambient concentrations is an example of classical measurement error that results in bias of effect
-------
estimates to the null, not just an increase in standard error. Claiming that the difference between
average personal exposure and ambient concentrations results in "attenuation of risk" is not
appropriate.
The Ozone Panel does not completely agree with staffs conclusion that "the use of
routinely monitored ambient ozone concentrations as a surrogate for personal exposures is not
generally expected to change the principal conclusions from ozone epidemiological studies."
Indeed, Panel members have little insight as to what we would find if we had actual exposure
measurements. Personal exposures most likely correlate better with central site values for those
subpopulations that spend a good deal of time outdoors, which coincides, for example, with
children actively engaged in outdoor activities, and which happens to be a group that the ozone
risk assessment focuses upon.
Some statements about which individuals are at greatest risk of ozone-induced effects are
not adequately supported by the information discussed in the chapter. Individuals with chronic
obstructive pulmonary disease (COPD) and cardiovascular disease (CVD) are likely to be at
increased risk, but the hypothesis that such "hyper-responsiveness" can be used to identify
individuals with COPD or CVD who are at greatest risk of Os-induced health effects has not
been confirmed. A more appropriate conclusion would be that individuals with COPD and CVD
are at increased risk of O3-induced health effects.
The discussion of the ranges for changes in FEVi that are considered to be small,
moderate, or large for persons with impaired respiratory systems is not consistent. While EPA
staff state that the table values for the ranges do not need to be changed, staff indirectly
acknowledge that a 10% reduction in this variable in asthmatics could have serious
consequences, an interpretation that is used in Chapters 4-6.
The 30 subjects studied by Adams had a great influence on the analyses presented in
Chapters 5 and 6. While the discussion of the low-level exposures used in the controlled human
studies by Adams and colleagues is technically correct that no statistically significant changes
were found in FEVi for ozone at 40 to 60 ppb compared to filtered air, there were clearly a few
individuals who experienced declines in lung function at these lower concentrations. These were
healthy subjects, so the percentage of asthmatic subjects, if they had been studied, would most
likely be considerably greater.
The lack of statistical power is consistently offered in Chapter 3 for why there appears to
be an inconsistent effect seen for COPD mortality. Coherence of respiratory effects for ozone
suffers from neither no more nor no less power considerations that do those for particulate matter
(PM). Yet the Agency did not argue a lack of power when assessing PM risks, so consistency is
needed here relative to ozone effect estimates for COPD mortality.
The relatively strong and relatively consistent effect of ozone on emergency department
visits for respiratory disease, especially asthma, as evidenced in Figure 3-4 is misrepresented in
several places in the Chapter (and in Chapters 5 and 6) as "inconclusive" or "inconsistent." This
should be corrected.
10
-------
Chapter 4 (Characterization of Human Exposure to Ozone): The second draft of
Chapter 4 has responded to many of the comments made on the first draft, and is thus clearer
than before. The panel was pleased to see the reanalysis for 2002 in addition to 2004.
It would be helpful to have the estimated exposures for current (2002 and 2004) levels
displayed in Tables 4-8 & 4-9 (p. 4-32) and Figures 4-4 to 4-21 (pp. 4-33 to 4-41), in addition to
only those for just meeting the current standard and alternative more stringent standards. This
would be analogous to the way estimated effects are displayed in Chapter 5 (Figures 5-5 to 5-9
[pp.5-58 to 5-65]).
On the whole, Chapter 4 provides a clear "road map" for what was done to characterize
available knowledge about human exposure to ozone in the framework of generally accepted
modeling approaches of appropriately selected populations in 12 urban areas of the U.S. Much
of the text reads like a basic textbook on human exposure assessment using state-of-the-art
modeling approaches, such as the Air Pollutants Exposure Model (APEX), including adjustments
for lung ventilation of delivered ozone dose. This extension, beyond exposure characterization,
is particularly important for ozone where the extent of measurable human responses is very
sensitive to the amount of ozone inhaled and to where it deposits along the respiratory tract.
Further extension of the methodology to estimate dose would have important implications and
should be discussed.
There is an explicit discussion of the limitations of the APEX model in terms of
variability and the quality of the input data, which is appropriate and fine as far as it goes. There
are good reasons presented for selection of urban areas and the time periods to be modeled.
However, there was inadequate consideration of the populations selected for modeling. Those
selected were appropriate, but the omission of the elderly, the population most at risk for ozone-
associated premature daily mortality, was notable and not even mentioned in terms of why it was
not considered.
The chapter was very good at exposition and clear presentation of modeling results, but
was deficient it its discussion of seemingly counterintuitive results, and of a potentially large
influence of measurement biases. As an example of the first of these issues, the children in LA
& Houston are estimated to have far fewer exposures above 0.07 ppm (8-hr) than in most other
cities with lower ozone concentrations and fewer children. This was likely due to the greater
within-day and sampler-to-sampler variations in concentration within these two cities than in the
others, the fact that the entire year was modeled while for other sites the winter was not included
and/or the greater extent of air conditioning, especially in Houston. Whatever the reasons, there
should have been some discussion of the causes. The quadratic rollback methodology should
have been better described since this strategy has important consequences for the modeled
results.
The second issue that was presented, but left hanging without an adequate discussion is at
the bottom of page 4-47, where it was simply stated that "in general, APEX systematically
under-predicts the measured values by 0.001 to 0.02 ppm (zero to 50 percent)." If this is so, is it
due to a really serious failure of the APEX model, or to unreliable measurements? The
measurements at issue were six-day average concentrations based on the use of passive
11
-------
(diffusion) samplers, which are known to be subject to significant errors when the air velocity
across the inlet is variable. The comparison of measured and modeled concentrations depicted in
Figure 4-22 is certainly worthy of further analysis and discussion.
Chapter 5 (Characterization of Health Risks): Generally the panel found Chapter 5
and its accompanying risk assessment to be well done, balanced and reasonably communicated.
Additional text is needed at the beginning and end of the chapter to put the limited risk
assessment into the context of the much larger body of evidence of ozone health effects. The
discussion of uncertainty in these risk estimates is expanded in section 5.3.2.5. Although a
number of issues are raised, their impacts on the estimates have not been thoroughly explored.
Additional sensitivity analyses seem warranted. In particular, it is essential that the sensitivity of
the risk assessment to the shape of the dose-response curve for FEVi be evaluated. Although the
3 parameter logistic (3PL) model emulates the pattern seen in the five "data points," these points
are aggregates of the original data, and may give a misleadingly optimistic picture of the quality
of the fit. More importantly, although the problem of model uncertainty is noted it has not been
addressed even though methods exist for doing so. Even if only the linear and logistic models
were included in the analysis, the error bands around the estimated response probabilities would
likely increase to better reflect that uncertainty. In addition, a suggestion to deal with the
uncertainties surrounding estimation of PRB, particularly as related to Table 5.5 (for lung
function) and Table 5.11 (mortality), would be to change the form of the analyses to assess the
impact of the concentration change in the expected number of health effects relative to the
current standard. The key advantage of estimating the effect of concentration change is that it
does not depend on the choice of the PRB.
With regard to the controlled human exposure studies, Ozone Panel members believe that
the selection of changes in pulmonary function expressed as percent change in FEVi in children
is a fair indicator of an adverse effect at 15% change in all active children; and, in asthmatic
children, a 10% change is indicative of adverse effects. However, the presentation of the figures
showing these effects needs to be revised to indicate the uncertainties in the results used,
particularly at the lower levels of exposure. The potential mechanisms whereby these changes
are a reflection of both pain on breathing, partial inflammation of smaller airways, other effects
on airways, and potentially triggers for more significant respiratory morbidity, particularly in
asthmatic children, are not adequately discussed. In addition, some added discussion is
necessary to indicate that these measures are generally taken in areas with relatively high
background levels of ozone exposure, and that the role that tolerance may play in minimizing the
degree of adverse effect observed needs to be considered.
From the perspective of the epidemiological data, the Ozone Panel judged the selection
of: respiratory symptoms in moderate/severe asthmatic children (ages zero [birth] to 12); hospital
admissions for respiratory illness among asthmatic children; and premature total non-accidental
and cardiorespiratory mortality for inclusion in the quantitative risk assessment to be appropriate.
However, the CASAC believes that several other endpoints should be discussed qualitatively to
support the findings that these endpoints indicate that significant adverse effects are occurring at
exposure concentrations well below the current standard. Other endpoints deemed worthy of
additional discussion included respiratory emergency department visits among asthmatics and
patients with other respiratory diseases, increased medication usage, and increased
12
-------
symptomatology reported at exposure levels well below the current standard. Taken together,
members of the Ozone Panel felt strongly that these findings preclude including the current
standard as a scientifically defensible option for the Administrator (see discussion about Chapter
6 found in the main portion of the letter above).
Another problem in the health effects calculations (see Table 5-5 and 5-11) is that they
are based on computations of the form Rx - RPRB, where Rx is the risk at a given concentration x
of O3 and RPRB is the corresponding risk at policy-relevant background (PRB) for O3. As
discussed at the Ozone Panel's August meeting, the PRB is highly-problematic to calculate and
is, in some sense, "unknowable." One can avoid this problem by calculating the A = R0.8 - Rx
for various concentrations x. This form would allow focus on the question, "What is the
difference in the expected number of health effects that will occur at various concentrations of
O3, relative to the current standard of 0.08?" A key advantage of A is that it does not depend on
the choice of PRB, and thus is free of the uncertainties surrounding estimation of PRB.
Chapter 6 (Staff Conclusions on Primary Os NAAQS): See the discussion on Chapter
6 found in the main portion of the letter above. It would also be helpful to have the estimated
exposures for current (2002 and 2004) levels displayed in figures 6-1 to 6-6 (pp. 6-34 to 6-39), in
addition to only those for just meeting the current standard and alternative more stringent
standards. This would be analogous to the way estimated effects are displayed in Chapter 5
(Figures 5-5 to 5-9 [pp.5-58 to 5-65]).
Chapters 7 (Policy-Relevant Assessment of Welfare Effects Evidence) and 8 (Staff
Conclusions on Secondary Os NAAQS): Chapter 7 is a well-developed and persuasively
presented assessment of the welfare effects of ozone on vegetation, which forms the technical
basis for the range of secondary standards recommended in Chapter 8. That having been said,
the potential for significant propagation of error/uncertainty in the underlying technical
documentation draws into question the conclusions drawn by EPA Staff. As observed in the
Agency's 1989 and 1996 Ozone Staff Papers, ozone remains the most prevalent phytotoxic
compound in the ambient air "impairing crop production and injuring native vegetation and
ecosystems more than any other air pollutant" (USEPA 1989, 1996). Furthermore, as has been
noted in the current assessment of human health effects, there also appears to be no safe
threshold concentration below which ozone effects on sensitive vegetation are eliminated. See
the additional discussion on Chapter 8 found in the main portion of the letter above.
13
-------
Appendix A - Clean Air Scientific Advisory Committee Roster (FY 2006)
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, 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 - CASAC Ozone Review Panel Roster
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CASAC Ozone Review Panel
CHAIR
Dr. Rogene Henderson*, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
MEMBERS
Dr. John Balmes, Professor, Department of Medicine, University of California San Francisco,
University of California - San Francisco, San Francisco, California
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, National Jewish Medical and
Research Center, Denver, CO
Dr. William (Jim) Gauderman, Associate Professor, Preventive Medicine, Medicine,
University of Southern California, Los Angeles, CA
Dr. Henry Gong, Professor of Medicine and Preventive Medicine, Medicine and Preventive
Medicine, Keck School of Medicine, University of Southern California, Downey, CA
Dr. Paul J. Hanson, Senior Research and Development Scientist, Environmental Sciences
Division, Oak Ridge National Laboratory (ORNL), Oak Ridge, TN
Dr. Jack Harkema, Professor, Department of Pathobiology, College of Veterinary Medicine,
Michigan State University, East Lansing, MI
Dr. Philip Hopke, Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Potsdam, NY
Dr. Michael T. Kleinman, Professor, Department of Community & Environmental Medicine,
University of California - Irvine, Irvine, CA
Dr. Allan Legge, President, Biosphere Solutions, Calgary, Alberta, Canada
B-l
-------
Dr. Morton Lippmann, Professor, Nelson Institute of Environmental Medicine, New York
University School of Medicine, Tuxedo, NY
Dr. Frederick J. Miller*, Consultant, Cary, NC
Dr. Maria Morandi, Assistant Professor of Environmental Science & Occupational Health,
Department of Environmental Sciences, School of Public Health, University of Texas - Houston
Health Science Center, Houston, TX
Dr. Charles Plopper, Professor, Department of Anatomy, Physiology and Cell Biology, School
of Veterinary Medicine, University of California - Davis, Davis, California
Mr. Richard L. Poirot*, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
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. Elizabeth A. (Lianne) Sheppard, Research Professor, Biostatistics and Environmental &
Occupational Health Sciences, Public Health and Community Medicine, University of
Washington, Seattle, WA
Dr. Frank Speizer*, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA
Dr. James Ultman, Professor, Chemical Engineering, Bioengineering Program, Pennsylvania
State University, University Park, PA
Dr. Sverre Vedal, Professor of Medicine, Department of Environmental and Occupational
Health Sciences, School of Public Health and Community Medicine, University of Washington,
Seattle, WA
Dr. James (Jim) Zidek, Professor, Statistics, Science, University of British Columbia,
Vancouver, BC, Canada
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 gov)
* Members of the statutory Clean Air Scientific Advisory Committee (CASAC) appointed by the EPA
Administrator (FY 2006)
B-2
-------
Appendix C - Charge to the CASAC Ozone Review Panel
0:3 air quality information and analyses (Chapter 2):
1. To what extent are the air quality characterizations and analyses clearly communicated,
appropriately characterized, and relevant to the review of the primary and secondary Os
NAAQS?
2. Does the information in Chapter 2 provide a sufficient air quality-related basis for the
exposure, human health and environmental effects, health risk assessment, and
environmental assessment presented in later chapters?
O^-related health effects (Chapter 3):
1. To what extent is the presentation of evidence from the health studies assessed in the AQCD
and the integration of information from across the various health-related research areas
drawn from the Ch AQCD technically sound, appropriately balanced, and clearly
communicated?
2. What are the views of the Panel on the appropriateness of staffs discussion and conclusions
in Chapter 3 on key issues related to quantitative interpretation of animal toxicology and
controlled-exposure human experimental studies and epidemiologic study results, including,
for example, exposure error, the influence of alternative model specification, potential
confounding or effect modification by co-pollutants, and lag structure?
3. What are the Panel's view on the adequacy and clarity of staff discussion on the issue of
potential thresholds in concentration-response relationships discussed in Chapter 3?
Exposure Analysis (Second Draft Chapter 4 of the O^Staff Paper, draft Exposure Analysis
technical support document and OAQPS Staff Memorandum on Uncertainty Analysis):
1. To what extent are the assessment, interpretation, and presentation of the results of the
exposure analysis as presented in Chapter 4 (and in the second draft Exposure Analysis
technical support document) technically sound, appropriately balanced, and clearly
communicated?
2. Are the methods used to conduct the exposure analysis technically sound? Does the Panel
have any comments on the methods used?
3. To what extent are the uncertainties associated with the exposure analysis clearly and
appropriately characterized in Chapter 4, the Exposure Analysis technical support document,
and the uncertainty memorandum?
C-l
-------
4. To what extent is the plan for the remaining uncertainty assessment technically sound? Are
there other important uncertainties which are not covered? What are the views of the Panel
on sensitivity analyses conducted to evaluate the influence of uncertainties in the exposure
analysis?
Heath Risk Assessment (Second Draft Chapter 5 of the (X Staff Paper and draft Health Risk
Assessment technical support document):
1. To what extent are the assessment, interpretation, and presentation of the results of the
revised exposure analysis as presented in Chapter 5 (and in the second draft Risk Assessment
technical support document) technically sound, appropriately balanced, and clearly
communicated?
2. In general, is the set of health endpoints and concentration-response and exposure-response
functions used in this risk assessment appropriate for this review?
3. Are the methods used to conduct the health risk assessment technically sound? Does the
Panel have any comments on the methods used?
4. To what extent are the uncertainties associated with the health risk assessment clearly and
appropriately characterized in both the second draft Chapter 5 and the second draft Health
Risk Assessment technical support documents?
Staff Conclusions and Standard Options for the Primary O^NAAQS (Chapter 6):
1. What are the views of the Panel on the approach taken by staff (as discussed in Chapter 6) of
using both evidence-based and quantitative exposure- and risk-based considerations in
drawing conclusions and identifying options as to a range of standards to protect against
health effects associated with exposure to 63, alone and in combination with the ambient mix
of photochemical oxidants, for consideration in this review of the primary OsNAAQS?
2. Does the Panel generally agree that the range of alternative primary 63 standards identified in
Chapter 6 is generally consistent with the available scientific information and is appropriate
for consideration by the Administrator?
3. What are the views of the Panel on the key uncertainties and Os research recommendations
discussed in Chapter 6?
O3-related welfare effects and secondary standard options (Chapters 7):
1. To what extent is the presentation of evidence drawn from the vegetation effects studies
assessed in the 63 AQCD technically sound, appropriately balanced, and clearly
communicated?
C-2
-------
2. What are the views of the Panel on the appropriateness of staffs weight-of-evidence
approach which assesses information from across the various vegetation-related research
areas described in the Os AQCD, including chamber and free air exposure crop yield and tree
seedling biomass experimental studies, foliar injury data from biomonitoring plots, and
modeled mature tree growth?
3. To what extent are the methods used to conduct the exposure assessment and the
interpretation and presentation of the results of the exposure assessment in the second draft
Chapter 7 and the draft Environmental Assessment technical support document technically
sound, appropriately balanced, and clearly communicated?
4. To what extent are the uncertainties associated with the exposure analysis clearly and
appropriately characterized in the second draft Chapter 7 and the draft Environmental
Assessment technical support document?
5. To what extent are the uncertainties associated with the vegetation risk assessment clearly
and appropriately characterized in both the second draft Chapter 7 and the draft
Environmental Assessment technical support document?
6. Staff recognizes that gradients can exist between O3 levels measured at monitor probe heights
and those measured over low vegetation canopies. What are the Panel's views on the
appropriateness of applying a single adjustment factor to hourly monitoring data to account
for the range of potential gradients that can exist across sites and crop and tree seedling
canopy structures? Are there alternative approaches or adjustment values the Panel would
suggest? Are staffs planned sensitivity analyses appropriate and sufficient?
7. To what extent do the figures aid in clarifying the text? Should more or less information of
this type be included in the final Chapter 7 or its Appendices?
8. Given the lack of quantitative information on Os-related ecosystem effects, what are the
Panel's views on the appropriateness of how this topic is addressed in the second draft
Chapter 7?
Staff Conclusions and Standard Options for the Secondary O^NAAQS (Chapter 8):
1. Does the Panel generally agree that the secondary standard options identified by staff
(including indicator, averaging time, form, and level) are generally consistent with the
available scientific and technical information and are appropriate for consideration by the
Administrator?
C-2
-------
Appendix D - Review Comments from
Individual CASAC Ozone Review Panel Members
This appendix contains the preliminary and/or final written review comments of
the individual members of the Clean Air Scientific Advisory Committee (CASAC)
Ozone Review Panel on the 2nd Draft Ozone Staff Paper and three related draft technical
support documents — Ozone Health Risk Assessment for Selected Urban Areas: Draft
Report (2nd Draft Ozone Health Risk Assessment, July 2006); Ozone Population
Exposure Analysis for Selected Urban Areas: Draft Report (2nd Draft Ozone Exposure
Assessment, July 2006); and Draft Ozone Environmental Assessment: Exposure, Risk
and Benefits Assessment (Draft Ozone Environmental Assessment, July 2006) — who
submitted such comments electronically. The comments are included here to provide
both a full perspective and a range of individual views expressed by Panel members
during the review process. These comments do not represent the views of the CASAC
Ozone Review Panel, the CASAC, the EPA Science Advisory Board, or the EPA itself.
The views of the CASAC Ozone Review Panel and the CASAC as a whole are contained
in the text of the report to which this appendix is attached. Panelists providing review
comments are listed on the next page, and their individual comments follow.
D-l
-------
Panelist Page#
Dr. John Balmes D-3
Dr. Ellis Cowling D-6
Dr. James D. Crapo D-l 1
Dr. William (Jim) Gauderman D-12
Dr. Henry Gong D-13
Dr. Paul J. Hanson D-16
Dr. Jack Harkema D-24
Dr. Philip K. Hopke D-26
Dr. Michael T. Kleinman D-28
Dr. Allan Legge D-34
Dr. Morton Lippmann D-36
Dr. Frederick J. Miller D-39
Dr. Maria Morandi D-43
Mr. Charles Plopper D-46
Dr. Armistead (Ted) Russell D-47
Dr. Elizabeth A. (Lianne) Sheppard D-52
Dr. Frank Speizer D-62
Dr. James Ultman D-68
Dr. Sverre Vedal D-69
Dr. James (Jim) Zidek D-76
Dr. Barbara Zielinska D-91
D-2
-------
Dr. John Balmes
•>nd
Review of Chapter 3 of the 2n draft Ozone staff paper - John Balmes
Overall, I find this draft of Chapter 3 much improved over the previous draft. In particular, the
efforts to respond to the some of the concerns expressed in the letter written by CAS AC
subsequent to the publication of the CD are appreciated.
In general, the chapter provides careful interpretation of the available data in the controlled
human exposure, epidemiological, and animal toxicological literature. However, I do think that
several of the statements about which individuals are at greatest risk of ozone-induced effects in
the concluding paragraph of the chapter are not adequately supported by the information
discussed in the chapter. First, we do not know that individuals "with cardiorespiratory
impairment (e.g., those with COPD or cardiovascular disease) who are "hyperresponsive" to Os
exposure (i.e., exhibit much higher than normal lung function decrements and/or respiratory
symptoms) would be considered at greatest risk of Os exposure." While I agree that individuals
with COPD and cardiovascular disease (CVD) are likely to be at increased risk, I do not believe
that the hypothesis that such "hyperresponsiveness" can be used to identify individuals with
COPD or CVD who are at greatest risk of Os-induced health effects has been confirmed. The
hypothesis is certainly plausible, but for the purpose of this chapter, I suggest that it simply be
stated in the conclusion that individuals with cardiorespiratory impairment (e.g., those with
COPD or CVD) are at increased risk of adverse effects of Os exposure.
Second, the final sentence of the chapter states that "those with genetic polymorphisms for
antioxidant enzymes and inflammatory genes may be at heightened risk of effects of Oj." While
I concur with this statement, the chapter provides no supporting evidence from human studies
(either controlled exposure or epidemiological). The only relevant supporting information that is
discussed in the previous sections of the chapter is that murine studies have used strain
differences in O3-induced responses to identify candidate genes that are likely to be involved in
determining susceptibility. These studies have involved genes in the early inflammatory
response rather than antioxidant enzymes. The previous discussion also includes the caveat that
the relevant murine studies used relatively high-dose exposure protocols. Unless, the limited
human data suggesting that polymorphisms of certain antioxidant enzyme genes (e.g., GSTM1
and NQO1) are discussed in the chapter, I would revise the final sentence.
Specific Comments
3-6-21 and 22 The Adams et al. studies were not done in southern CA. They were done at UC
Davis.
3-9-1 I would add "short-term" before declines in lung function in this sentence to be
absolutely clear.
D-3
-------
3-9-4 and 5 For both clarity and consistency, I would revise this sentence as follows:
"... pulmonary function in asthma, which included increased nonspecific airway responsiveness
secondary to airway inflammation due to Os exposure."
3-12-7 and 8 This sentence is misleading as currently written. It should be revised as follows:
"Both human and animal studies indicate thatAHR is not associated with airway
inflammation..."
3-34-2 This sentence overstates the limitations of the data on ED visits for respiratory disease,
especially asthma. For example, in Figure 3-4 ED visits for asthma, only one study did not show
an increased risk. I would revise the sentence as follows: "... positive but somewhat less robust
evidence for associations with respiratory ED visits."
3-36-9 and 10 This sentence is a bit unclear as written. I would revise this sentence to end
after "...studies."
3-48-11 Spirometry cannot be done in animals so "measures of lung function" should be
substituted.
3-56-33 Controlled human exposure studies cannot evaluate effects of long-term O3
exposures.
3-57-15 through 17 In my view, this sentence misrepresents the status of the available
evidence. I would favor a statement that used "inconclusive" or "limited" to describe the
evidence on long-term (Vinduced effects on lung function.
3-60-35 Change "don't" to do not.
3-60-32 through 34 This sentence is poorly written and should be revised as follows:
"Controlled studies of dietary antioxidant supplements have shown some protective effects on
lung function decrements but not on symptoms and airway inflammatory responses."
3-64-16 through 18 The Vagaggini study cited exposed subjects to allergen before 63 and
obtained samples of respiratory tract lining fluid by sputum induction rather than by
bronchoalveolar lavage.
3-65-3 This sentence is simply untrue regarding asthmatic subjects. I would revise the
sentence to delete the words "and asthmatic".
3-67-11 This sentence is poorly written. I would revise the sentence as follows: "...the
involvement of specific genes or genetic loci in Os-induced airway hyperresponsiveness and
inflammation."
Table 3-3 For consistency, I would replace "bronchial" with "airway".
3-75-3 through 5 These bullets should be revised to ".. .Os after... "
D-4
-------
3-81-16 Use "owrs" here rather than "our."
3-82-27 through 29 This statement is again too strong in my view. I would suggest wording to
the effect that the data from epidemiological studies is "limited" or "inconclusive" while data
from clinical studies is "lacking".
3-82-30 and 31 This sentence is poorly written. I would revise the sentence as follows: "...of
Os-mduced effects with a clean air recovery period..."
References The Mortimer reference listed is a secondary paper for the study. In addition to
the 2002 Eur Respir J citation, the primary paper for this study published in 2000 in the Am J
Respir Crit Care Med should also be cited.
D-5
-------
Dr. Ellis Cowling
Dr. Ellis Cowling
North Carolina State University
August 17, 2006
General Comments on Chapters 7 and 8 of the
"Second Draft OAQPS Staff Paper on Ozone: Policy Assessment of Scientific and
Technical Information"
and the draft technical support document titled
"Technical Report on Ozone Exposure, Risk, and Impacts Assessments for Vegetation:
Draft Report"
The following General and Specific comments are focused almost entirely on the CASAC
"Charge Questions" raised in Karen Martin's letter to Fred Butterfield dated July 17, 2006.
Please note in the paragraphs below, that Karen Martin's questions are written in bold-faced
non-indented paragraphs, whereas my responses are written in indented paragraphs without
bold type.
Comments in Response to the Charge Questions for Chapter 7 in the Second Draft Staff
Paper on Ozone and the draft technical support document mentioned above:
1. To what extent is the presentation of evidence drawn from the vegetation effects studies
assessed in the ozone Criteria Document technically sound, appropriately balanced, and
clearly communicated?
I find the evidence drawn from the vegetation effects text and figures of Chapter 7 to be
technically sound, appropriately balanced, and clearly communicated.
2. What are the views of the Panel on the appropriateness of staffs weight-of-evidence
approach which assesses information from across various vegetation-related research areas
described in the ozone Criteria Document, including chamber and free air exposure crop
yield and tree seedling biomass experimental studies, foliar injury data from bio-
monitoring plots , and modeled mature tree growth?
I find the weight-of evidence approach used by EPA staff appropriate and reasonably
thorough with regard to chamber and free air exposure crop yield and tree seedling biomass
experimental studies, foliar injury data from bio-monitoring plots, and modeled mature tree
growth.
3. To what extent are the methods used to conduct the exposure assessment and the
interpretation and presentation of results of the exposure assessment in the second draft
Chapter 7 and the draft Environmental Assessment technical support document technically
sound, appropriately balanced, and clearly communicated?
D-6
-------
I find these methods, their interpretation, and presentation of results to be technically sound,
appropriately balanced, and clearly communicated in both Chapter 7 and the technical support
document mentioned above.
4. To what extent are the uncertainties associated with the exposure analysis clearly and
appropriately characterized in both the second draft Chapter 7 and the draft
Environmental Assessment technical support document?
I find these uncertainties to be both clearly and appropriately characterized in both the second
draft Chapter 7 and the draft Environmental Assessment technical support document.
5. To what extent are the uncertainties associated with the vegetation risk assessment
clearly and appropriately characterized in both the second draft Chapter 7 and the draft
Environmental Assessment technical support document?
Similarly, I find these additional uncertainties with regard to the vegetation risk assessment
to be both clearly and appropriately characterized in both the second draft Chapter 7 and the draft
Environmental Assessment technical support document.
6. Staff recognizes that gradients can exist between ozone levels measured at monitoring
probe heights and those measured over low vegetation canopies. What are the Panel's
views on the appropriateness of applying a single adjustment factor to hourly monitoring
data to account for the range of potential gradients that can exist across sites and crop and
tree seedling canopy structures? Are there alternative approaches or adjustment values
that the Panel would suggest? Are staff sensitivity analyses appropriate and sufficient?
I agree that these gradients in ambient air concentrations within vegetation canopies can and
do exist and also find that the idea of applying a single adjustment factor to account for these
gradients is satisfactory. I think the approach used in these documents is as reasonable as any
other that could be considered and that the sensitivity analyses used by staff are both appropriate
and sufficient.
7. To what extent do the figures aid in clarifying the text? Should more or less information
of this type be included in the final Chapter 7 or its appendices?
As indicated in my General Comments on the First Draft Staff Paper on Ozone on December
17, 2005, lack of figures — and especially pictures and charts displaying the important and
sometimes very severe impacts of ozone on crops, forest and shade trees, and natural ecosystems
— is a major shortcoming of Chapter 7 in this Second Draft Staff Paper on Ozone effects on
vegetation. Pictures are indeed worth a thousand words (or more!)
I assert once again that it is very important that the final draft of Chapter 7 in this Staff Paper
on Ozone should include illustrations that show:
a) The distribution of stomata on leaf surfaces, ideally also a paired set showing stomata open
and stomata closed, and a diagram showing the structure of the stomata within the palisade
layer of a leaf,
D-7
-------
b) Crop plants or floral plants grown under different concentrations of ozone,
c) Differences in grain yield with and without exposure to injurious concentrations of ozone,
d) The greater effect of ozone exposure on root growth than on shoot growth,
c) A collage of pictures showing foliar injury that decreases economic value of various leafy
vegetables like spinach, cut flowers, and Christmas trees, and most important of all:
1) Widespread ozone-induced mortality or ozone-and-biotic-pathogen induced mortality of
different species trees exposed to ozone in some of the national parks, state parks, and/or
wilderness areas in various parts of the US.
2) A picture of the cross-section of the stem of a tree showing differences in width of annual
rings in trees grown with and without exposure to injurious concentrations of ozone, and,
if possible, also, a cross-section showing the differential width of annual rings on the
same tree during years when ozone exposures were high and ozone exposures were low.
8. Given the lack of quantitative Information on ozone-related ecosystem effects, what are
the Panel's views on the appropriateness of how this topic is addressed in the second draft
Chapter 7?
I think the staff have done about as well as can be expected with this question under the
present circumstances of substantial scientific and technical ignorance about whole-ecosystem
effects.
Comments in Response to the single Charge Question for Chapter 8 in the the Second
Draft Staff Paper on Ozone:
1. Does the Panel generally agree that the secondary standard options identified by staff
(including indicator, averaging time, form, and level) are generally consistent with the
available scientific and technical information and are appropriate for consideration by the
Administrator?
I believe that EPA staff have done precisely what should be done with regard to
recommending firmly and persuasively to the Administrator of EPA that the time has come to
formulate and implement a Secondary (Welfare-Based) National Ambient Air Quality Standard
for Ozone that is distinct in averaging time, form, and level from the Primary Standard.
The scientific and technical evidence in support of the staff recommendations for serious
consideration of the alternative cumulative SUM06 and/or W126 Secondary Standards is well-
developed and persuasively presented.
As shown by the references cited at the end of Chapter 8, the consensus view among expert
persons in the ecological communities of this country and elsewhere around the world is that a
Secondary Standard of cumulative form and extending over a whole growing season — will be
far more effective than a Secondary Standard that is not cumulative in form and does not include
the whole growing season. Thus, it is simply NOT appropriate to continue to try to protect
D-8
-------
vegetation from the substantial known or anticipated, direct and/or indirect, adverse effects of
ozone by continuing to promulgate identical Primary and Secondary Standards for ozone.
This generalization is true for a wide variety of commercially important crop plants. But it
also is true for the vegetation in:
a) both intensively managed and wild-land forests,
b) scenic vistas in natural parks and wilderness areas,
c) ornamental shrubs and shade trees in urban, suburban, and rural areas, as well as
d) the vegetation in natural grasslands, rangelands, and other natural ecosystems
all over the United States.
A few final comments on the relationship between Chapter 2 and Chapters 7 and 8 in this
Second Draft Staff Paper on Ozone and the Name of the Standard to which this Second
Draft Staff Paper is addressed:
1. Relationship between Chapter 2 titled "Air Quality Characterization" and Chapter 7
titled Policy Relevant Assessment of Welfare Effects Evidence and Chapter 8 titled
Staff Conclusions on Secondary Ozone NAAQS.
There are several important parts of the Air Quality Characterization (Chapter 2) of this
Second Draft Staff Paper on Ozone that are highly relevant to the conclusions drawn in Chapters
7 and 8. Thus, in my General Comments on the First Draft Staff Paper on Ozone, I
recommended that maps be included in either Chapter 2 or Chapter 7 (or both) (and possibly also
in Chapter 8) that would show the distribution of counties in various parts of the US where
exceedances of various alternative Primary and Secondary standards have occurred in the past
and/or would be expected to occur in the future.
I had in mind that these maps would show the distribution of counties that exceed: a) the
present identical primary and secondary ozone standards (8-hour) and b) and c) - the two
distinctive cumulative secondary standards that were proposed for consideration by the
Administrator of EPA in 1996 and again in this Second Draft Staff Paper in 2006 [b) SUM06
andc)W126].
At present, page 7-15 in Chapter 7 is the only place in this Second Draft Staff Paper where I
could find all three of these alternative existing and proposed ozone standards defined in way
that allows ready comparison among these three alternatives:
a) the existing identical Primary and Secondary Standards,
b) the proposed alternative cumulative SUM06 Secondary Standard, and
c) the proposed alternative cumulative W126 Secondary Standard.
But these definitions are now presented in Chapter 7 - which is a very long way away from the
maps that are presented in Chapter 2!
A map showing the distribution of counties across the US where exceedances of the existing
identical Primary and Secondary Standards were exceeded in 2002-2004 is shown in Figure 2-6
on page 2-20 in Chapter 2. And several different maps (based on two different sets of
D-9
-------
monitoring sites in 2001 and 2002) are shown in Chapter 2 for the SUM06 proposed cumulative
Secondary Standard.
But no maps at all are shown in either Chapter 2 or chapters 7 or 8 for the W126 alternative
proposed cumulative Secondary Standard. I know (and maybe some readers of the Final Staff
Paper on Ozone will also know) that the SUM06 and W126 maps will be much the same, but
that very similarity is good reason for showing how very similar they are (or by chance are not so
similar in one or another part of the US).
I believe the value of various parts of Chapter 2 for understanding the major conclusions in
Chapters 7 and 8 will be enhanced if more inter-chapter linkages and cross-references regarding
exposures of crops, forests and natural ecosystems are included in the text of both chapters.
I believe it also will be valuable to include in Chapter 2 or Chapter 7, a chart similar to those
on pages 2-28 and 2-32 in the First Draft Staff Paper on Ozone showing the diurnal patterns of
ozone concentrations at a high-elevation forest site where ozone concentrations during nighttime
hours frequently remain relatively high. These nighttime concentrations of ozone are important
because of their injurious effects on high-elevation forests as discussed on page 7-22 of the First
Draft Staff Paper on Ozone.
2) Title of this Draft Staff Paper on Ozone
In my General Comments on the First Staff Paper on December 17, 2005,1 agreed with
Philip Hopke's assertion that the title of this Staff Paper should be changed so it will be faithful
to the title of the National Ambient Air Quality Standards to which this Staff Paper applies.
Since the Standards dealt with in this Staff Paper on Ozone bear the name "Ozone and Related
Photochemical Oxidants," both Philip Hopke and I advised that the title of the staff paper should
also include "Ozone and Related Photochemical Oxidants."
Apparently, either both Philip and I were wrong in believing that "Related Photochemical
Oxidants" are a part of the present name of these NAAQS, or the authors of the Second Draft
Staff Paper considered our recommendation not to be relevant to the job they had at hand.
Section 1.2.2 is titled "History of Ozone NAAQS Reviews." But it does not explain when
the NAME of these Standards (as opposed to the specific "indicator" for these standards) was
changed. Thus, I recommend that Section 1.2.2 be revised to avoid the confusion that at least
Philip Hopke and I have had (and continue to have) about the NAME of the standards to which
this Staff Paper is relevant.
D-10
-------
Dr. James Crapo
Comments on Second Draft of the Ozone Staff Paper
Chapter 6, "Staff Conclusions on Primary Ozone NAAQS"
Dr. James Crapo, CASAC
August 22, 2006
This chapter is in general well written and describes the substantial progress that has occurred in
the science underlying our understanding of potential health effects from exposures to ozone at
levels near or below the current primary standard of 0.08 PPM on an 8 hour average. In the
interval since the current standard was set, a variety of studies have found that measurable lung
function and/or other health effects occur, particularly in sensitive populations, for exposures to
ozone that are at or below the current standard and possibly as low as 0.04 PPM. Uncertainties
increase as the level of ozone exposure decreases and the staff are appropriate in considering this
in making final recommendations.
The summary recommendations in Section 6.3.6 provide a rationale for both retaining the current
8 hour ozone standard and for decreasing the standard to a level of 0.07 PPM with a range of
forms from the third to the fifth highest daily maximum 8 hour average concentration. Given the
consistency of data from multiple sources demonstrating adverse health effects at levels at and
below the current 0.08 PPM 8 hour standard, the reasonableness of retaining the current standard
should be re-assessed. In my opinion, this position is not well justified by the CD or by the
arguments in the staff paper. Clear health benefits are estimated by the EPA staff analysis to
occur if the standard were lowered to 0.07 PPM and, although uncertainty increases, further
benefit is demonstrated to occur by the staff analysis if the standard is lowered to 0.06 PPM.
Based on staff analysis that enhanced benefit occurs by lowering the standard as low as 0.06
PPM, I would recommend that the range of consideration for modification of the standard be
expanded to include a full analysis of the potential benefit of lowering the standard to 0.06 PPM.
The staff paper should also further consider the form of the standard. The staff paper analysis
indicates that modest changes in the form of the standard can have profound impacts on the
frequency of adverse health events. This would justify a more thorough evaluation of the form
of the standard and a possible recommended change. This is an area where research
recommendations should emphasize the uncertainties and opportunity to more rigorously test the
local and regional impacts of different choices for the form of the primary ozone standard.
Overall, Section 6 is excellent and provides a strong rational basis for rule setting with regard to
the ozone NAAQS.
D-ll
-------
Dr. William (Jim) Gauderman
Comments on Staff Paper, post meeting
Gauderman
8/25/06
The health-effects calculations given for lung function (e.g. Table 5-5) and mortality (e.g. Table
5-11) are based on computations of the form Rx - RPRB where Rx is the risk at a given
concentration x of Os and RPRB is the corresponding risk at policy relevant background for Os.
As discussed at the Panel meeting, a preferable alternative is to compute A = Ro.8 - Rx for various
concentrations x. This form would allow focus on the question "What is the difference in the
expected number of health effects that will occur at various concentrations of Os, relative to the
current standard of 0.08?" A key advantage of A is that it does not depend on the choice of PRB,
and thus is free of the uncertainties surrounding estimation of PRB.
D-12
-------
Dr. Henry Gong
Post-Meeting Comments on Chapter 6 (Staff Conclusions on Primary O3 NAAQS) in Second
Draft of Ozone Staff Paper (July 2006).
Henry Gong, Jr., M.D.
August 30, 2006
I did not present pre-meeting comments for this chapter. The discussion of Chapter 6 by the
Committee as a whole helped me with my evaluation of this chapter. My comments about this
important chapter are as follows:
1. I believe that most of my comments regarding Chapter 5 also apply to Chapter 6,
especially in terms of the foundation for Staff decision-making about clinical endpoints
and their public health relevance. The essential points are that many clinical endpoints
(e.g., see Figure 3-4, page 3-53) may not be sufficiently evaluated with quantitative risk
assessment but nonetheless constitute highly relevant evidence-based ozone-related
health effects. Such evidence and its consistency "trumps" models, according to many
clinicians.
2. The "adversity" of the health effects may be described using guidelines developed by the
American Thoracic Society for healthy people and people with respiratory impairment
(see Tables 3-2 and 3-3, pages 3-70 and 3-71). These tables may be somewhat
misleading when comparing exposure responses of healthy people versus respiratory
patients. People with pre-existing respiratory disease have varying pulmonary reserves;
small percent decrements in lung function may result in more symptoms, interference or
limitation of function, use of medications, etc., than anticipated. Even pulmonary
patients with "mild" or "moderate" responses may require significantly more therapy
(e.g., medications) and/or emergency department visits or hospitalization. The exposure
to ambient ozone is typically recurrent (since ozone episodes generally persist over
several days), and adverse effects are likely experienced repeatedly in sensitive groups.
3. The weight of the scientific evidence that was compiled and summarized in the Ozone
Criterion Document and Ozone Staff Paper clearly indicates that the current ozone
standard is inadequate to protect human health. Although more data is desirable such as
at low ozone concentrations, we see adverse health effects below the current 0.08 ppm-
standard. My reading of the numerous tables and figures indicate that thousands of
sensitive children will continue to experience ozone exposures of concern and resulting
lung function decrements (and other health effects) at or below 0.07 ppm. [I note that the
use of percentages can be confusing and misleading in many of the tables (see Dr. Lianne
Sheppard's Final Comments, 8/28/06).] Thus, my overall gestalt is to support a reduction
of the 8-hour standard to 0.060-0.070 ppm (to three decimal places) to achieve greater
protection of public health. I am still concerned that even this range does not build in an
adequate margin of safety for sensitive groups, such as asthmatics. Thus, I favor a
standard in the lower end of the proposed range. I anticipate that future research studies
will demonstrate evidence of adverse effects in sensitive groups at lower ozone
concentrations.
D-13
-------
Post-Meeting Comments on Chapter 5 (Characterization of Health Risks) in Second Draft of
Ozone Staff Paper (July 2006).
Henry Gong, Jr., M.D.
August 28. 2006
My pre-meeting comments (dated 8/22/06) remain active. The discussion by the Committee as a
whole helped me extend my evaluation of this chapter. My further comments are as follows:
4. As I discussed during the meeting (August 24-25), I am concerned that a great deal of
emphasis is given to the quantitative risk assessment of lung function decrements in
school-age children. The concern about this vulnerable group is appropriate and well
supported. Nonetheless, I conclude from this chapter that the emphasis is
disproportionate in the sense that numerous other clinical endpoints are down-graded or
considered less than optimal, and, thus, should not be effectively or appropriately used to
support the adverse health risks from contemporary ozone exposures. These ozone-
related endpoints are respiratory symptoms among asthmatic children, increased
medication usage, emergency department visits and hospitalizations for respiratory
illnesses, and non-accidental, cardiorespiratory deaths. The discussion is up and down
regarding their respective causality by ozone, sufficiency of exposure-response data, data
consistency, etc. I cannot see how such a large volume of supporting evidence of
clinically relevant information can be masked and not even considered in some type of
qualitative risk assessment. Figure 3-4: "Effect estimates (with 95%) for associations
between short-term ozone exposure and respiratory health outcomes" (Chapter 3, page 3-
53) is the best overall visually convincing clinical summary of this argument. This figure
should be prominently and effectively used in subsequent communications regarding the
adverse respiratory effects of ambient ozone exposure. I believe that the clinicians
(physicians) on the CASAC Ozone Review Panel recognize and agree to this point as
well. As someone stated during the meeting, the formal risk assessment process is only a
piece of the evidence and not the entire argument for ozone-induced adverse health
effects.
5. Lung function at low ozone concentrations (0.04, 0.06, and 0.08 ppm): This type of
clinical data is very important to obtain and understand. I appreciated and supported the
inclusion of Dr. Adams' data into the Ozone Criterion Document and Staff Ozone Paper.
Nonetheless, his data (n=30 subjects) must be carefully understood and interpreted. The
probabilistic exposure-response relationships for FEV1 decrements (Figures 5-2a,b,c)
show a logistic modeled sigmoid curve that is weighted by relatively few data points at
the lower ozone concentrations. Are the FEV1 responses normal variations or real ozone
responses? Are these responses reproducible in the same subjects? How confident can we
be at this time about the lower risk estimates for lung function response with the limited
data points? Thus, I am expressing cautious review and interpretation of the small amount
of lung function data at these low ozone levels.
6. Page 5-56. The "ozone season" in different study areas appears to be a "moving target"
and is not identical from region to region. I am concerned here that we are not comparing
comparable ozone seasons. Changing meteorology is another influential factor that alters
ozone levels from year to year as well as spatially.
7. A recently published paper (Triche EW et al: Low-level ozone exposure and respiratory
symptoms in infants. Environ Health Perspect 2006.114:911-916) did not make it into the
D-14
-------
current CASAC ozone review process. Nonetheless, I would like to call attention to it as
another indication that respiratory symptoms occur at or below the current EPA
Standards.
Pre-Meeting Review Comments on Chapter 5 (Characterization of Health Risks) of Second Draft
of Ozone Staff Paper (July 2006)*
Henry Gong, Jr., M.D.
August 22, 2006
Responses to Four Queries from Dr. Martin et al. (OAQPS):
1. I generally found the assessment, interpretation, and presentation of the results of the
revised exposure analysis in the two documents to be technically sound, balanced, and
adequately communicated. The progression of discussion appears logical. I specifically
found the presentation on exposure-response functions (section 5.3.1.3) to be reasonable.
2. The set of health endpoints and concentration-response and exposure-response functions
used in the risk assessment appear appropriate, given the limited information available
for other potentially important endpoints, e.g., airways hyperreactivity, respiratory
hospital admissions. As such, it seems that we are underplaying and understating the
biological and health impacts of the non-analyzed endpoints of ambient ozone exposures.
3. As far as I can tell, the methods used to conduct the health risk assessment are technically
sound and appropriate. I look forward to hearing other reviewers comment on the
methods used.
4. The major uncertainties associated with the health risk assessment appear clearly and
appropriately characterized.
Other Comments about the Documents:
1. The value for policy relevant background such be stated early to set a reference point.
2. Page 5-12/line 3: "FEVli." must be a typo?
3. Why are there negative numbers for incidences of mortality (3 columns) in Table 5-11,
page 5-51?
4. I cannot find Figure 5-12 (see page 5-71).
5. I do not understand the significance of the statement that "the 8-hr maximum
concentrations are on average twice the 24-hour average level." I read it on page 5-73
(lines 37-38) and I believe that it was also stated previously in the chapter.
* Also includes supporting Chapter 3 (Assessment of Risk Based on Controlled Human Exposure Studies) of the
Draft Ozone Health Risk Assessment for Selected Urban Areas: Draft Report (June 2006).
D-15
-------
Dr. Paul J. Hanson
Review of Chapters 7 and 8 of the
Second Draft OAQPS Staff Paper on Ozone: Policy Assessment of Scientific and Technical
Information
and the
Technical Report on Ozone Exposure, Risk, and Impact Assessments for Vegetation: Draft
Report July 13, 2006
Final Comments from Dr. Paul J. Hanson: August 27, 2006
The staff has done a credible job of summarizing the important issues of crop and forest
tree species responses to tropospheric ozone. Statements of the uncertainties surrounding their
conclusions are presented, but some improvements could be made. I do, however, recognize that
based on the nature of the available data the statements of uncertainty will default to qualitative
statements in some cases (e.g., projected mature tree responses).
A. Responses to the Panel's Charge Questions on Chapter 7
1. To what extent is the presentation of evidence drawn from the vegetation effects studies
assessed in the O3 AQCD technically sound, appropriately balanced, and clearly communicated?
I was generally pleased with the technical content, balance, and presentation of scientific
details contained in the second draft staff paper and the associated technical
documentation. Comments and suggested changes for specific items are detailed after
the responses to the charge questions.
2. What are the views of the Panel on the appropriateness of staff s weight-of-evidence approach
which assesses information from across the various vegetation-related research areas described
in the 63 AQCD, including chamber and free air exposure crop yield and tree seedling biomass
experimental studies, foliar injury data from biomonitoring plots, and modeled mature tree
growth?
I support this approach and was impressed that a variety of disparate methods could be
used to develop a consistent picture of the sensitivity of vegetation to ozone uptake. I
have listed a number of concerns that the Staff should consider in the preparation of a
final document.
3. To what extent are the methods used to conduct the exposure assessment and the interpretation
and presentation of the results of the exposure assessment in the second draft Chapter 7 and the
draft Environmental Assessment technical support document technically sound, appropriately
balanced, and clearly communicated?
D-16
-------
The methods are appropriate to the appointed tasks, but my comments point to areas
where details were missing and where the clarity might be improved.
4. To what extent are the uncertainties associated with the exposure analysis clearly and
appropriately characterized in the second draft Chapter 7 and the draft Environmental
Assessment technical support document?
Uncertainties related to the calculation and enumeration of exposure values and statistics
are covered in the document (Page 7-29), but more attention should be paid to the
uncertainties of the biological input data with respect to how well they characterize the
full breadth of genetic variation within crops and natural species. The document does a
reasonable job of characterizing the variability of the available data for cultivated crops
(pages 7-42 and 7-43), however, no similar discussion is provided for the response of tree
species and the vegetation of ecosystems.
The document should acknowledge that available studies on non-crop species target a
combination of species of commercial importance and species known to be sensitive to
ozone. That is, much of what we know about the response of non-crop-plant response to
ozone results from work to understand why particular species or within-species
genotypes exhibited their obvious sensitivity to ozone. The fact that they were sensitive
to ozone caused them to be studied more than species that did not exhibit measurable
responses. The document should indicate that we have an inadequate quantitative
understanding of the full range of ozone sensitivity in natural vegetation. Available data
could represent either an over or an underestimate of vegetation response throughout the
United States. A dearth of information exists for key species present in many areas.
Research is needed to understand the true range of ozone sensitivity of species and
within-species genotypes distributed across the landscape.
5. To what extent are the uncertainties associated with the vegetation risk assessment clearly and
appropriately characterized in both the second draft Chapter 7 and the draft Environmental
Assessment technical support document?
No comment.
6. Staff recognizes that gradients can exist between 63 levels measured at monitor probe heights
and those measured over low vegetation canopies. What are the Panel's views on the
appropriateness of applying a single adjustment factor to hourly monitoring data to account for
the range of potential gradients that can exist across sites and crop and tree seedling canopy
structures? Are there alternative approaches or adjustment values the
Panel would suggest? Are staffs planned sensitivity analyses appropriate and sufficient?
This issue is discussed in more detail and with much greater clarity within Chapter 7
(pages 7-47 to 7-48) than in the supporting technical report. The authors did a good job
of characterizing the desirability of adjusting measured ozone concentrations to relevant
crop heights and they clearly recognize that such a procedure "Ideally should account for
D-17
-------
the exact height of each monitor, canopy roughness for each crop, and the seasonal and
diurnal nature of turbulence." Nevertheless, in their attempt to adjust ozone exposure
levels for monitor vs. crop heights they may have improved the reality of ozone exposure
for some crop species but degraded the applicability for others (e.g., tall maize crops).
Correction factors for crop species growing in windy environments with lower boundary
layers would be different than correction factors for crops growing in areas with more
stagnant air masses.
I do not suggest that Staff eliminate the completed analysis, but rather include an
additional analysis based on the POES data without the 10% adjustment. Having both
data sets available will provide a bracket of responses within which the reality probably
lies for all crops. Based on their presentation at the public meeting, I see that the Staff
has indeed begun to make these calculations, and understand that they will be included in
the final document.
7. To what extent do the figures aid in clarifying the text? Should more or less information of
this type be included in the final Chapter 7 or its Appendices?
The figures were generally appropriate and helpful, but I made some recommendations
on specific figures and tables in the detailed comments listed below.
8. Given the lack of quantitative information on Os-related ecosystem effects, what are the
Panel's views on the appropriateness of how this topic is addressed in the second draft Chapter
7?
The document provides a concise summary of the important issues to be considered at the
ecosystem scale, but appropriately concludes, "it is difficult to quantify the contribution
of 63 due to the combination of stresses present in ecosystems".
B. Responses to the Panel's Charge Questions on Chapter 8
1. Does the Panel generally agree that the secondary standard options identified by staff
(including indicator, averaging time, form, and level) are generally consistent with the available
scientific and technical information and are appropriate for consideration by the
Administrator?
Given the reality of the available data and the status of our understanding of the response
of vegetation to ozone exposure and uptake, the proposed indicators, averaging times,
and forms are appropriate. I recognize that specifying the level or levels to be applied to
the new form of the secondary standard is a difficult task underscored by substantial
uncertainties in the data. Nevertheless, I encourage the Staff to propose a level specific
to the protection of vegetation independent of the primary standard. This is an important
distinction. By providing an independent justification for the averaging time, form and
level for the secondary standard the Administrator will have the option of considering a
change to the secondary standard even if the primary standard were to be left as is.
D-18
-------
The Staffs recommendation for changing the form of the secondary standard and a range
of proposed levels should be accompanied by documented statements of the benefits that
society would gain by achieving the proposed standard. Tangible benefits that might be
calculated include increased crop yield, forest production, and tree seedling survival.
Conclusions regarding the benefits to be gained by a change in the secondary standard
should be accompanied by acknowledgement of the uncertainties involved. The resulting
summary would demonstrate to policy makers and the public the utility of considering a
new secondary NAAQS for ozone.
In the future, alternate approaches based on mechanistic models of plant and ecosystem
ozone uptake and subsequent response can and should be developed and discussed.
However, the application of such detailed approaches to the combination of species and
biomes present across the United States is not possible with current data (see page 7-4).
C. Specific comments and editorial suggestions by document
In the following section I provide comments on the items that I questioned or for which I
would like to point out possible editorial modifications. Readers of these comments should not
view the list as a negative reflection on the document as a whole. Overall, I found that Chapters
7 and 8 and the associated technical document contained many excellent conclusions and
statements. I have simply not taken the time to enumerate and point out the positive.
Chapter 7
Page 7-7 lines 21 to 30: The authors fall into the trap of using teleological concepts to describe
active rather than reactive plant responses to stress. Plants do not think about responding to
ozone stress (they can't). On line 22 the phrase "plants have a variety of compensatory
mechanisms" should probably be changed to 'plants exhibit a variety of compensatory
mechanisms'. What looks like or appears to be variable repair capacity among species or
genotypes could be an induced response to wounding, but it could also be the result of inherently
different membrane or protein turnover within those same tissues.
Page 7-7 line 4: One might replace "reallocation" with 'translocation'.
Page 7-8 last paragraph: Please reword as ".. .with the exception of a few clones of the genera
Populus., which can be highly sensitive and in some cases are as sensitive to O3 as annual
plants." Some Populus clones are also very insensitive to ozone.
Page 7-9 line 9:1 would prefer that "cause a" be changed to 'result in a'.
Page 7-9 line 32: Replace this line with '.. .in vigor may be plant death for sensitive plant
species'. The response discussed would not be general to all plants.
Page 7-12 line 16: Change ".. .which diseases are likely..." to '.. .the disease-ozone interactions
likely...'.
D-19
-------
Page 7-12 line 27: Would it be appropriate to indicate that the competitive response is driven
indirectly by changes in plant species ability to compete for water and nutrients?
Page 7-13: Change "increasing ozone flux" to 'foliar ozone uptake'.
Page 7-14 line 16: The authors might add a new heading at this location: Climatic Change.
Page 7-14 lines 24 to 28: The authors could consider citing a few modeling papers that have
dealt with offsetting CC>2 and ozone responses.
Page 7-14 lines 30 to 34: Suggested rewording: '... simulations will be important in specifying
hypotheses to be tested for the many complex interactions of ozone and various combinations of
environmental factors. The results obtained will, of course, only be as reliable as the input data
for their parameterization, and additional data from organized, systematic study will be needed to
judge the efficacy of such model runs.'
Page 7-15 lines 1 to 3: A reference or the name of the EPA program should be provided.
Page 7-16 line 25: Delete the word 'plots'.
Page 7-19: Figures 7-1 and 7-2 have inadequate captions. They don't stand on their own. I was
unable to follow these figures until they were explained at the meeting.
The caption for Figure 7-3 was inadvertently split by the figure and "Smoke" should be 'Smoky'.
Page 7-28 line 26: What are the criteria for the statement that the current NAAQS may not
provide adequate protection at this point in the text? At this point, all that has been discussed are
the exposure metrics. They have yet to be connected to biological response or a discussion of
risk. The authors are associating attainment of ozone atmospheres above the current standard
with a negative biological response. Although this may be true for some plant species (perhaps
even many plant species), it won't be true for all plant species. I would reserve the statements
regarding levels of protection for later discussion.
Page 7-41: Exposure response information can be developed from FACE exposure systems, but
one must recognize the potential for significant gradients of exposure gas concentrations
throughout the exposure rings. While the FACE protocols minimize exposure concentration
gradients, plants near the gas emitters will be exposed to larger concentrations than centrally
located plants near the air monitoring/sampling point. In the case of CC>2 studies this has often
been considered a minor point, but in the context of ozone response studies where peak
concentrations are known to disproportionately affect plant function, subtle gradients could be
very important. If growth data are averaged across the entire ring, the mean response associated
with ozone concentrations measured only at the center of the ring will lead to overestimates of
plant response when extrapolated directly to other field locations.
Page 7-43: The authors should delete the comparison of the SoyFACE results with the CR-
functions. The hail damage totally confounds the desired comparison to the point that the
D-20
-------
comparison should not have been attempted. The Morgan et al. (2006) article is missing from
the reference list.
Page 7-44 figure 7-13 versus Figure 7D-1 in the annex: When response surface data such as
these are interconverted, how is it done? That is, how exactly is the x-axis based on 12 hour
SUM06 altered to derive the exact or approximate equivalent for the "4th highest daily 8-h
average"? Is an approximate multiplier applied or are the values recalculated from raw hourly
ozone values collected at the time of the original experiment? I asked a similar question at the
public meeting and wasn't sure that I got my message across. I hope that this alternate wording
of the question helps.
Page 7-51 line 17: Should "Taylor et al. 1993" be listed as 'Taylor 1993'.
Page 7-57: What Populus clonal data were used as the base response function for the
development of this map? Can the authors characterize the extent to which that response
function is appropriate and applicable for the full range of Populus distributed nationwide?
Page 7-58 to 7-60: This section starts out with so many qualifying statements about the limited
use of foliar injury responses I began to wonder why the authors bothered to include it in the
document. Further reading of this section, however, suggested that the FIA foliar injury data has
real value as qualitative indicator of ozone exposure. If carefully measured and interpreted in the
context of available ozone concentration measurements, I believe foliar injury data may have
found a useful role as a semi-quantitative assessment of exposure. The authors do an excellent
job of characterizing the limitations of this analysis. Such data will likely never be refined
enough to help characterize appropriate levels for future standards, but they demonstrate nicely
the relative magnitude and regional nature of vegetation responses to ozone.
Appendix 7 A Page 3 and 4:
I believe that item number one leaves the reader with an inappropriate conclusion. The
real issue is that for flux-based indices to work they would need to be based on perhaps hourly
time step models to ensure that ozone concentration and foliar conductance data are
appropriately matched. I don't believe that a flux-based approach can be reduced to a simple
mathematical index. The full mechanistic model based on a 24-hour calculation would be
essential for success. In addition, a flux based model of ozone response would also need to
include model components necessary to account for the transient development of soil water
deficits that would lead to stomatal closure and result in reduced ozone uptake by foliage.
Appendix 7D Page 16:
Tables 7E-1, 7E-2, 7E-3 and similar Table in the technical report (pages 5-6 and 5-7)
would be more useful for future reference if the parameter estimates for each regression analysis
(mean and perhaps max and min) were accompanied by their respective standard deviations.
Future probabilistic estimates of response, as apposed to the deterministic approach provided in
the tables, would be made possible by the inclusion of these estimates of parameter error. In
fact, if only the parameter estimates for those curves characterizing the mean response data were
provided with the appropriate estimates of error, calculated max and min estimates could be
reproduced for defined confidence limits around the mean.
D-21
-------
Providing such data in these key documents could very well be the basis for new and
independent analyses by other researchers.
Chapter 8
Page 8-5 lines 4 to 16: The predicted value of anticipated loss would be greater if the ozone
concentration reduction of 10% had not been applied to the POES data prior to this analysis. For
reasons stated earlier, one might consider these estimates of loss to therefore be too low. Please
include a discussion of the unadjusted values as presented by Staff during the public meeting.
Page 8-11: Please include an explanation and justification for why the cumulative exposure
indices are limited to a 3-month window? Vegetation grows for much longer than 3 months in
many regions of the United States. There may be an obvious answer apparent to those associated
with the standard setting process, but I believe it should be spelled out in the Staff paper.
Page 8-14: As discussed and concluded during the public discussion, I agree that the Staff paper
should not recommend keeping the secondary standard identical to the primary. Such an option
remains a default that the Administrator could consider. It need not be listed and certainly is not
supported by the weight-of-evidence.
Research needs:
All of the proposed research needs listed on pages 8-14 to 8-16 are justified, but they are
not prioritized and therefore represent a simple wish list of desired research. To move ozone
effects research ahead at an optimum pace, any list of proposed research needs focus,
prioritization and integration. Limited financial resources underscore the need for making tough
choices regarding the order of completion of research objectives. A long wish list is a very scary
budget prospect, but a prioritized research program to be addressed over time represents a
palatable option for sustained progress in times of limited financial resources.
My personal opinion is that baseline mechanistic research for the identification of the role
of antioxidants and tissue replacement (i.e., repair) in conveying species or species genotype
specific ozone resilience has received inadequate attention in recent years. The physics of ozone
transport to and into foliage is reasonably well understood, but improved models may be
required to characterize this phenomenon within complex canopies of mixed vegetation.
Development and implementation of a plan to characterize the full range of natural plant
sensitivity to ozone for plant species with important commercial and/or aesthetic value to society
is needed. Such data will ultimately be needed to translate plant specific responses to the
monetary or intangible values they have throughout the United States.
Notwithstanding excellent past suggestions for research to improve health and ecological
risk assessments for ozone (EPA December 2001), the EPA has made very limited investment in
ozone research since the last CD was produced. As I understand the situation, reduced emphasis
on ozone was driven by the need to refocus funds and effort on other criteria pollutants (e.g.,
PM). To advance the science in time for the next periodic evaluation of the Ozone and Related
Photochemical Oxidants Standards, however, a fundamental change in the manner in which EPA
funds research must be considered. Because such a plan must be realistic and consider the
distribution of limited fiscal resources, I recommend that EPA consider a two phase research
concept consisting of sustained funding for mechanistic investigations whereby ozone damages
D-22
-------
tissues and their functions, to be supplemented periodically by field-scale, exposure-response
campaigns to provide tangible empirical data for testing models of plant response to ozone
exposure. The timing of new expensive field campaigns could be determined or initiated when
major shifts in cultivar use are evident, are currently needed for many dominant tree species, and
might be applied to other ecosystem indicator species as their relationship to ecosystem net
production is better understood.
Whether my specific recommendations are followed is not important. It is critical,
however, that EPA make plans for sustained investment in criteria pollutant research. Without
sustained research investment we stand to lose critical institutional knowledge both within and
outside EPA necessary to make the periodic evaluation of NAAQS work. Without new research
results and sustained monitoring of indicator pollutants at relevant scales the legislated
requirement for periodic evaluation of air quality standards could become a pointless exercise.
Technical report
The draft Technical Report on Ozone Exposure, Risk, and Impact Assessments for Vegetation
dated July 13, 2006 contains useful information in support of Chapters 7 and 8 of the Staff paper.
Page 5-7: The 10% downward adjustment of ozone values is not fully explained in this
document, but appropriate text for this issue is included in Chapter 7. If the technical report is
intended to act as a stand-alone document as well as a supplement, it should contain some of
those details.
I personally found Figures 5-5 to 5-9 to be very instructive regarding the yield losses for crops
and one or more of these figures might be moved forward into Chapter 7. The analogous figures
for tree seedling responses were also very useful (Figures 5-10 to 5-14). Including such figures
in Chapter 7 would provide the policy makers with visual information about the species-to-
species variation.
The mature tree growth discussion on pages 7-1 and 7-2 should be expanded to include the
details for the western tree species simulations discussed in Chapter 7 (i.e., ponderosa pine).
D-23
-------
Dr. Jack Harkema
Comments on Second Draft of the Ozone Staff Paper
Chapter 3, "Policy-Relevant Assessment of Health Effects Evidence"
Dr. Jack R. Harkema
CASAC Ozone Review Panel
August 25, 2006
General Comments:
In general this chapter is very well written and adequately summarizes the health effects
evidence presented in Chapters 4 through 7 of the CD. The authors have done a good job of
integrating the most relevant health-effects data from epidemiologic and human/animal
controlled exposure studies in this summary chapter. Emphasis on new reports in the peer-
reviewed, scientific literature regarding subpopulations with elevated susceptibility to ozone
exposures, effects of ozone exposure on the cardiovascular system, and epidemiology studies
concerning the effects of acute exposure of ozone on mortality are most appropriate and
adequately presented in this document. The new data presented in the CD and summarized in this
chapter is strongly supportive of the current ozone standard, but also suggests that the current
standard may not be adequate to protect susceptible populations such as those with
cardiopulmonary disease, children and the elderly from adverse health effects. This needs to be
more strongly and clearly stated in this chapter.
The revised summary tables and figures in body of the text and appendix are excellent
and nicely complement the text. For the most part, more interpretation of the results of key
studies has been provided in this revised draft compared to the previous draft. There are,
however, areas in this chapter that still need more interpretation of results to help the reader
understand the plausible reasons underlying the principle health-effects findings. For example,
additional statements are needed on biological mechanisms that may be responsible for seasonal
and individual variations in response to ozone exposure in the subsections of 3.4. Numerous
animal and human controlled exposure studies have documented that individual exposure history
to ozone (or other oxidant pollutants) determines an individual's response to short-term
exposure. A person living in an ozone-polluted city will likely develop some form of tolerance to
ozone and may respond differently than a person who lives in a relatively clean city and who has
not built up a tolerance to ozone. The latter person may be at greater risk for developing adverse
health effects in response to acute ozone exposure than the former person. This concept needs to
be presented more clearly in the document and how it relates to the issue of threshold
determination.
Specific Comments:
P3-1, L29-31 Authors need to state the key findings not just what was examined.
P3-6, L7 Define "BSA" and "triangular exposure."
D-24
-------
P3-15, L15 Epithelial cell proliferation (hyperplasia) is a more prominent and well-
documented ozone-induced CAR lesion after acute exposure than is intramural fibrosis.
P3-21, L13 Monkeys exposed to 0.15 ppm (not 0.2) ozone developed airway
remodeling (Harkema et. al., Am J Pathol. 1993 Sep;143(3):857-66).
P3-23-24, L35-2 This is an example of "tolerance" in an animal study.
P3-44-45, L34-2 Study by Vedal et al. is a good example of a population with a low level of
oxidant air pollution that have a significant association between exposure and mortality at a low
ozone concentration (see general comments).
P3-46, L21 Change "inflex" to "influx."
Overall, Chapter 3 is excellent and provides strong supportive documentation for setting the
ozoneNAAQS.
D-25
-------
Dr. Philip K. Hopke
Comments on Chapter 2
P.K. Hopke
Ozone is of key importance in the atmosphere because it is the easily measured indicator of
photochemical oxidation and oxidative chemistry. This chemistry leads to particle formation and
the production of oxidize products that serve as irritants (acrolein, carbonyl compounds) and
materials that can produce other effects. Thus, this chapter should take the opportunity to
indicate the broader role of ozone and related photochemical oxidants on the overall production
of air pollution in both the gas and particle phases.
I still believe the role of other ambient photochemical oxidants in producing both health and
ecosystem effects has been underestimated in this document. The short rationale provided in
Chapter 2 for using ozone as the indicator basically dismisses the role of other oxidants, but
clearly is written with the concept that the other oxidants are in the gas phase. There should be at
least a mention of the role of ozone in particle formation and the formation of particle bound
ROS. We currently know little about the concentrations and potential health effects of these
species, but a placeholder for future better understanding of this issue should be added through a
few sentences addressing this issue.
One aspect of ozone chemistry and exposure that has not been noted in the staff paper is the
formation of indoor particles from ozone that has infiltrated into the indoor environment reacting
with organic compounds of indoor origins. I suggest a paragraph something along the lines of
that which follows be added to Chapter 2.
Formation of new PM by nucleation has been observed indoors as a result of infiltrated ambient
Os interaction with terpenes (Sarwar et al., 2003; Weschler, 2004). One class of compounds that
are active ingredients of indoor air is terpenes (Weschler and Shields, 1997). Terpenes such as a-
pinene, and p-pinene are also found in emissions from building material such as pine, hardwood,
southern pine etc., (Baumann etal., 1999). The terpene system (Fan etal., 2003; Weschler &
Shields, 2003; Weschler, 2003; Sarwar et al., 2003) has been studied and known to produce
aerosols with ozone levels normally observed indoors from exchange with outdoor air. Thus,
another route of exposure to air pollutants is indoor exposure to particles resulting from
infiltrated ambient ozone.
Baumann, G. D. M., Batterman, S.A., Zhang, G., Terpene emissions from particleboard and
medium-density fiberboard products, Forest Products Journal 49, 49-56, 1999.
Fan, Z., Lioy, P., Weschler, CJ. , Fielder, N., Kipen, H., Zhang, J., Ozone initiated reactions
with mixtures of Volatile organic compounds under simulated indoor conditions",
Environmental Science and Technology 37, 1811-1821,2003.
Sarwar G, Corsi R, Allen D, Weschler C, The significance of secondary organic aerosol
formation and growth in buildings: experimental and computational evidence,
Atmospheric Environment 37', 1365-1381, 2003
D-26
-------
Weschler CJ, Indoor/outdoor connections exemplified by processes that depend on an organic
compound's saturation vapor pressure, Atmospheric Environment 37, 5455-5465, 2003
Weschler, C. J., Chemical reactions among indoor pollutants: what we've learned in the new
millenium, Indoor Air 14 (suppl 7): 184-194, 2004.
Weschler CJ, Shields HC, Potential reactions among indoor pollutants, Atmospheric
Environmental, 3487-3495, 1997.
Weschler C J, Shields HC, Experiments probing the influence of air exchange rates on secondary
organic aerosols derived from indoor chemistry, Atmospheric Environment 37, 5621-
5631,2003.
Otherwise, this chapter provides an adequate background on the atmospheric background for the
standard
D-27
-------
Dr. Michael T. Kleinman
Chapter 4: Characterization of Human Exposure
Comments: M. Kleinman
General comments: This chapter is well written however there are some specific points that
could be addressed better to strengthen it.
Page
4-1
4-2
4-5
4-5
4-6
4-6
4-11
4-12
4-13
4-14
4-15
4-18
4-18
4-19
Line
18
29
29
30
O
10
31-32
31-32
25
17
18
3-4
Table 4-2
6
Comment
The criteria for selecting the 12 locations should be
summarized or at least a pointer to pg 4-18 (where this is
somewhat discussed) should be inserted.
Briefly discuss why dose estimation is not necessary in this
exercise.
Is the 180 ppb an average for a room? Perhaps the room size,
air exchange and other key parameters could be discussed
since neither of the sources cited are primary peer reviewed
literature.
A clear statement of the impact of outdoor O3 on indoor O3
concentration should be inserted here to provide the
foundation for statements made on pg 4-6, L 3
The basis for this needs clarification given the earlier
statement that people spend most of their time indoors.
Explain why it is not necessary to go further and estimate
dose.
Is this double-counting, or is the penetration factor computed
sans inside source contributions?
This could be a problematical concept the way it is phrased.
Isn't this moderated by the probabilistic function? i.e the
likelihood of one person getting the entire sum of occurrences
is vanishingly small.
Among the improvements should be an extension to actual
computation of dose.
The degree to which the modeled 12 sites are representative of
US populations should be described.
Were the AERs derived from the 12 modeled sites? If not the
uncertainty engendered by not matching the sites should be
discussed.
This should be rephrased since the meaning is not clear.
The lack of consistency in periods modeled among the sites is
a limitation on the generalization of the model. Some
additional justification for the partial data coverage should be
included.
Change "particularly" to "potentially"?
D-28
-------
4-32
14
Tables 4-8
and 4-9
Change " section 4.3.1" to "section 4.3.4.7"
There is an apparent logical inconsistency when children in
LA and Houston have fewer exposures above 0.07 ppm-8hr
cities with lower average O3 levels and fewer children.
Technical Support Document:
The modeling efforts are well supported in the document. The method by which the attainment
scenarios and the calculations of base cases in Chapter 5 is not obvious. There seems to be a
logical inconsistency in finding more person-days of excessive exposures in Boston than in
Houston and some explanation needs to be provided.
Additional Comments
Chapter 2
The issue of policy relevant background should be more extensively described. However, the
use of PRB as a modifier for the risk assessments leads to increased uncertainty in the estimates.
The PRB is very important in how states will develop implementation plans to achieve the O3
NAAQS, but the issue is very complex. Perhaps it would best be handled on a local level given
the key factors such as source of background O3, precursors, reducing agents that may degrade
O3 and many others will vary from locale to locale.
Chapter 3
Pg 3-39 L 1-4 Document states: "As discussed in the CD, Section 3.9, using ambient
concentrations to determine exposure generally overestimates true personal Cb exposures by
approximately 2- to 4-fold in available studies, resulting in biased descriptions of underlying
concentration-response relationships and attenuated risk estimates." Given this more emphasis
should be given to the large number of epidemiological studies summarized in Figure 3-4 that
show clinically important adverse outcomes associated with O3 exposures where most of the
exposures are at ambient concentrations BELOW the current NAAQS. This data, in concert with
risk estimates in Chapter 5, proves that the current NAAQS does not provide adequate protection
and in fact affords no margin of safety. This should be carried forward in Chapter 6 to support a
recommendation for setting a standard below the current 0.08 8hr NAAQS.
Chapter 4
General comments: This chapter is well written however there are some specific points that
could be addressed better to strengthen it.
Page
4-1
4-2
Line
18
29
Comment
The criteria for selecting the 12 locations should be
summarized or at least a pointer to pg 4-18 (where this is
somewhat discussed) should be inserted.
Briefly discuss why dose estimation is not necessary in this
exercise.
D-29
-------
4-5
4-5
4-6
4-6
4-11
4-12
4-13
4-14
4-15
4-18
4-18
4-19
4-32
29
30
O
10
31-32
31-32
25
17
18
3-4
Table 4-2
6
14
Tables 4-8
and 4-9
Is the 180 ppb an average for a room? Perhaps the room size,
air exchange and other key parameters could be discussed
since neither of the sources cited are primary peer reviewed
literature.
A clear statement of the impact of outdoor O3 on indoor O3
concentration should be inserted here to provide the
foundation for statements made on pg 4-6, L 3
The basis for this needs clarification given the earlier
statement that people spend most of their time indoors.
Explain why it is not necessary to go further and estimate
dose.
Is this double-counting, or is the penetration factor computed
sans inside source contributions?
This could be a problematical concept the way it is phrased.
Isn't this moderated by the probabilistic function? i.e the
likelihood of one person getting the entire sum of occurrences
is vanishingly small.
Among the improvements should be an extension to actual
computation of dose.
The degree to which the modeled 12 sites are representative of
US populations should be described.
Were the AER's derived from the 12 modeled sites? If not
the uncertainty engendered by not matching the sites should
be discussed.
This should be rephrased since the meaning is not clear.
The lack of consistency in periods modeled among the sites is
a limitation on the generalization of the model. Some
additional justification for the partial data coverage should be
included.
Change "particularly" to "potentially"?
Change " section 4.3.1" to "section 4.3.4.7"
There is an apparent logical inconsistency when children in
LA and Houston have fewer exposures above 0.07 ppm-8hr
cities with lower average O3 levels and fewer children.
Chapter 5
This Chapter is well written and provides a useful description of the exposure characterization
process. The chapter does lean heavily on the human exposure data at 0.04 and 0.06 ppm to
seemingly justify the findings that there are measurable effects below the currant NAAQS.
There is a tendency in the document to downplay the epidemiological findings of important
health effects in cities with average concentrations at or below the NAAQS. The point that could
be made is that the limiting the studies to days that were below the NAAQS did not change the
outcome. Also a more precise description of the fact that the average exposures (used in the epi.
Studies) might be below the NAAQS but that at some location an individual sampler could be
above (putting the city out of compliance) although very few people in the population would
actually have received an excessive exposure, needs to be addressed. A quantitative assessment
D-30
-------
of the sensitivity of average exposure to the "peak" exposure might help to put some concrete
values on margin of safety.
Chapter 6
Pg 6-2, L18 Document states: "For the purpose of evaluating the level of the Cb standard in this
review, we have placed greater weight on U.S. and Canadian studies, taking into account the
extent to which such studies have reported statistically significant associations. This is because
findings of U.S. and Canadian studies are more directly applicable for quantitative considerations
in this review as studies conducted in other countries may well reflect quite different
populations, exposure characteristics, and air pollution mixtures." Was the Canadian activities
data used in CHAD?
Pg 6-8, L 7-10 Doc. States: "however, when corrected for the effects of exercise in clean air a
small percentage (7%) of healthy adult subjects experienced moderate lung function decrements
(^ 10% FEVi) with exposure to 0.06 and 0.04 ppm Cb." Some indication that this finding has
public health significance in context with the confidence limits around the concentration
response function should be added to the paragraph.
Pg 6-11, L 29 Doc States: [in reference to ED visits] " These studies provide evidence of effects
in areas that likely would not have met the current standard but do not address the likelihood of
effects occurring in areas that likely would have met the current standard." Should be amended
to reflect that in several cases, restricting the analyses to days with concentrations below 0.08
ppm still resulted in significant effects on health.
Pg 6-13 to 6-15 Table 6-1: This table is based on the exposure assessment modeling used
previously in Chap 4 which provides "artifactual" low values for key cities such as LA and
Houston. The low values are due to the non-uniform O3 distribution with specific samplers
having exceptionally high values causing an "inflated" design value. When the rollback is
performed, the O3 levels for most of the region is pushed to unrealistically low values leading to
an inflated estimate of the reduced numbers of cases at the current standard. This leads up to the
statement on Pg 6-16 L22 "When 2004 air quality is adjusted to just meet the current 8-hr
standard, there are an estimated 50,000 person days of 8-hr, moderate exertion exposures of
concern experienced by about 50,000 school age children, which means that there are very few
occasions where there are multiple occurrences of exposures of concern." Inclusion of these
atypical cities in this analysis has provided a biased view of the value of possible alternative
standards. This analysis should be rerun after excluding the "peaky" cities or with some other
roll-back scenario that more realistically predicts effects of controls. Another improvement
might be to present the percent of the children that still show adverse effects (as presented in
Table 5-5) because this would highlight the fact that even when meeting the current standard
substantial numbers (amounting to about 10%) of these sensitive children will still experience
adverse effects.
Pp 6-23-25, Tables 6-4 and 5. These tables should be expanded to include the alternative
scenarios that will be discussed. This is where one could provide some idea of the magnitude of
excess effects that could be removed if one adopted a more stringent standard. Some statement
D-31
-------
should also be made that these tables show that there is not an adequate margin of safety in the
current N A AQS.
Pg 6-28 L31: Doc. States: "Only a few studies presented results for different Cb averaging
periods using the same data set. Two of the recent multi-city mortality studies reported
associations for multiple averaging times (Bell et al., 2004; Gryparis et al., 2004). Both reported
that the effect estimates for different averaging times were not statistically different, though the
effect estimates for associations with 1-hr daily maximum Cb concentrations were somewhat
larger than those for longer averaging times, especially 24-hr average Os." The 1 hr standard is
important for those anomalous cities with greater non-uniformity in O3 exposure. There should
be some mention that states such as California have adopted a 1 hr standard to provide a margin
of safety not afforded by the 8 hr standard.
Pg 6-31 L10: Doc states "Collectively, the epidemiologic studies are inconclusive." This is
inconsistent with the Figure 3-4.
Peep ratsry gj-^p
•••ec isi'ir ts* 4
:»c=f*«r-£ 10 _
Respiratory Symptoms:
2. &«.«: *: s . ;•:•;; r,*:\ 5r; snj
piu:; Admissions:
Emergency Department Visits:
• ',*^5?n ;1 s ICC: '.'.'.-^e?:*' *,-
- >»5 »; 2 ,S2"< f, 3 ;3
Re&pcraiory McxtaEfcy:
•>•! »s a* * H£. t" (5 ±2-1 »»=*
4 •,::«r*i.v* ?:s :-:C3 .-ar::-y^
Figure 3-4. Effect estimates (with 96% confidence intervals) for associations
between short-term ozon-e exposure and respiratory health outcomes,
r O,, 3tt ppfc *sr S.li'-O,,
r t-h' Or
Pg 6-40 L10 Doc states "In conclusion, as discussed in the previous section addressing the
adequacy of the current standard, just meeting the current 0.08 ppm, 4th daily max standard
substantially reduces estimates of exposures of concern and risks of various health effects" An
additional statement that even when meeting the standard substantial numbers of sensitive
individuals are still expected to experience adverse effects, hence it is unlikely that the present
standard provides the required margin of safety.
D-32
-------
Pg6-44 L 17 Doc States: "Such a view might place more limited weight on evidence of more
serious morbidity (e.g., associations with hospital admissions) and mortality effects derived from
time-series epidemiological studies" Again referring to fig 3-4 these strongly suggest that there
is not an adequate margin of safety in the current standard.
Pg6-44 It would be appropriate to restate the current standard to 3 significant figures which is
consistent with the precision of current monitoring devices and which will improve the margin of
safety by eliminating "rounding up" to 0.084.
D-33
-------
Dr. Allan Legge
FINAL REVIEW COMMENTS: Allan H. Legge
"Review of the National Ambient Air Quality Standards for Ozone: Policy Assessment of
Scientific and Technical Information ( Second Draft) - OAQPS Staff Paper, July 2006".
Chapter 7: Policy-Relevant Assessment of Welfare Effects Evidence and
Chapter 8: Staff Conclusions on Secondary Ozone NAAQS
Overall Comment:
It is extremely unfortunate that the state of the science regarding ambient ozone and vegetation
has not significantly changed since 1996 with the last Os AQCD. The Agency has essentially
done little in the intervening years from 1996 to the present to help solve this problem. The
Agency must be strongly criticized for this lack of action. The underlying science to be able to
confidently set a meaningful biologically based secondary NAAQS for ozone simply does still
not exist. This is especially unfortunate when there is evidence of foliar injury from ozone
exposure at the current primary NAAQS for ozone. One can only make scientific advancement
with investment and for some reason the Agency has chosen not to make that investment with
respect to the science needed to be able to establish a viable and protective secondary NAAQS
for ozone.
General Comments:
Despite the above comments, Staff is to be congratulated for attempting such a monumental task.
This is especially true given that there has been essentially little new science carried out which
has addressed the uncertainties which were evident in the last iteration of the Oj AQCD in 1996.
While the presentation of the evidence which Staff has drawn from the current Os AQCD is as
technically sound as the evidence would allow, the large scientific uncertainties remain. Some of
these uncertainties are recognized by Staff while others are downplayed as not being important.
Staff, for example, have placed major emphasis on the results of the ozone exposure research
experiments carried out in open-top chamber (OTC) under the National Crop Loss Assessment
Network (NCLAN) Program for selected agricultural crop species cultivars and for selected tree
species seedlings by EPA's National Environmental Effects Research Laboratory-Western
Ecology Division (NHEERL-WED) in the 1980's. It has never been shown that the
concentration-response (C-R) functions derived from these OTC experiments realistically reflect
the response(s) of these plant species cultivars under ambient field conditions.
The argument that one can use the same C-R functions developed during the NCLAN Program
on crop species cultivars developed in the middle to late 1970's is highly questionable. This
assumes, for example, that the current crop species cultivars in use in 2002 have the same ozone
sensitivity 25-30 years later as the crop species cultivars developed in the 1970's. Further,
agricultural crop breeders do not directly bred for Os tolerance, it happens indirectly as a result
D-34
-------
of selection for the highest yielding disease resistant plants because the ambient 63 present will
be one of the selective factors.
The argument that one can use the same C-R functions developed by NHEERL-WED for
selected tree species seedlings grown in pots and exposed to 63 in OTC's and that these data can
then extrapolated to mature trees growing under field conditions is highly questionable. Further,
one cannot assume that tree species seedlings will have the same sensitivity to Os as mature
trees.
Another very important point is that ozone exposure indices such as SUM06 and W126 which
have been used by Staff do not have a biological basis as has been implied. It is ozone uptake by
plants which is critical and not the ozone exposure.
While the Staff Paper (SP) is very logical in it's presentation, Staff does not appear to be aware
of the very real potential for significant propagation of error and hence significant uncertainty in
the analysis prepared for OAQPS by Abt Associates Inc., (2006) [ "Technical Report on Ozone
Exposure, Risk, and Impact Assessments for Vegetation: Draft Report - July 13, 2006]. This is a
critically important point because the analysis in the Abt Associates Inc. (2006) draft report
provides much of the foundation for conclusions drawn by Staff regarding potential secondary
standards for O3. Put another way, the analyses provided by Abt Associates Inc. (2006) while
quite elegant is essentially a 'house of cards' with an unstable S foundation.
Each chapter builds on the previous chapter with the uncertainties in each chapter being carried
forward to the next chapter. These data are then used to generate an economic assessment for
agriculture crop loss on a national scale. With respect to the forest trees the seedling data are
scaled up to mature trees using the TREGRO Model which includes ozone as a stress factor to
estimate yield loss. In the end of all of this one is not sure whether any of this is realistic or that it
would be better described as science fiction.
D-35
-------
Dr. Morton Lippmann
M. Lippmann - Review Comments on Second Draft of Ozone Staff Paper, Chapter 2
Specific Comments
Page Line(s) Comment
2-47 5+ Policy-relevant background (PRB) needs to be discussed in a much broader
context, with recognition that it is influenced by long-range transport and scavenging processes
(PRB of 40 ppb over the Pacific Ocean and less over nearby land areas due to reactions with
vegetation and with background NO). The uncertainties that PRB creates in apportioning ozone
to PRB and North American anthropogenic sources are not relevant to uncertainties in exposure-
response relationships at concentrations being considered for a revised ozone NAAQS.
M. Lippmann - Review Comments on Second Draft of Ozone Staff Paper, Chapter 3
Specific Comments
Page Line(s) Comment
3-8 15-19 The concentrations that produced the effects described need to be stated.
3-8 21 insert "routine" after "with".
3-8 29-35 The concentrations that produced the effects described need to be stated.
3-9 2 add "and that they occurred at concentrations below those used in chamber
studies using exercise." at end of line.
3-10 29 insert "and" before "particularly."
3-11 34 insert "at rest" after "exposure".
3-12 14 insert "engaged in moderately high exercise" after "adults".
3-15 24 insert the duration of the exposure
3-16 16 insert "pattern" after "temporal".
3-21 18 insert "an association between O3 and" before "acute".
3-26 17 provide a reference for the cited study.
3-57 27 insert "Short-Term" before "Mortality"
3-59 19,20 change "which" to "that".
3-61 27 insert "relatively low concentrations of after "to".
3-63 7 Continue with a brief discussion of the O3 being part of a mixture.
3-66 20 change "effects" to "mortality".
3-67 2 change "health" to "lung function".
D-36
-------
M. Lippmann - Review Comments on Second Draft of Ozone Staff Paper, Chapter 4
Specific Comments
Page Line(s) Comment
4-1 30 Say something here about other important at-risk groups, such as seniors for
excess daily mortality.
4-3 12 change "NOx" to "NO".
4-4 11 add ", respectively" at end of line.
4-5 24 change "NOx" to "NO".
4-13 11 add "elderly" as another population at risk (for premature mortality).
4-16 6 change "until we have sufficient information characterizing" to "insofar as other
variables have greater effects, for example"
4-18 19 clarify what the New York urban area covers, specifically how far it extends into
PA.
4-30 5 change "city' to "urban area".
4-47 9,10 Sacramento is not an inland portion of the Los Angeles area.
4-47 2-25 The text says Apex was under-predicting. This is a rare case where I would
suspect the measurements rather than the model. The measurements were made with ozone
passive monitors, which are notoriously sensitive to ambient air velocity across the inlet, and
these velocities were likely to be very different across the inlets of the personal, indoor, and
outdoor units.
M. Lippmann - Review Comments on Second Draft of Ozone Staff Paper, Chapter 5
Specific Comments
Page Line(s) Comment
5-1 26 change "or" to ", or are".
5-3 27 add "It is also important to consider that O3 in ambient air is present in a complex
mixture of air pollutants, and that some other components of the mixture may play some role in
the health-related effects". At end of line
5-6 11 add "as well as similar responses in outdoor workers and others engaged in
recreational outdoor activities" at end of line.
D-37
-------
5-6 14 insert "young" before "adults".
5-21 7 insert", but larger than" after "to".
5-26 28 change "which are not statistically significant" to "that did not yield statistically
significant results".
5-27 22 change "nitrogen dioxide, sulfur dioxide, or carbon monoxide" to "NO2, SO2, or
CO".
M. Lippmann - Review Comments on Second Draft of Ozone Staff Paper, Chapter 6
Specific Comments
Page Line(s) Comment
6-3 4 insert "of O3 exposures" after "associations".
6-3 15 insert "O3" before "air"
6-3 16 change "quality" to "concentration".
6-3 22 change "an" to "a photochemical oxidant".
6-3 32 insert "and other photochemical oxidants" after "O3".
6-4 12 insert "indices of pulmonary function, such as" before "forced."
6-4 12,13 delete "also known as lung function decrements"
6-4 24 insert", who were judged to be a sensitive subgroup of concern" after "children".
6-8 33 change "may" to "generally".
D-38
-------
Dr. Frederick J. Miller
Dr. Fred J. Miller
August 29, 2006
Policy-Relevant Assessment of Health Effects Evidence
Chapter 3, OAQPS Staff Paper - Second Draft
General Comments
Overall, Chapter 3 now represents a good assessment of the health effects evidence to serve as
the basis for characterizing the health risks in humans from exposure to ozone. The material is
well written and organized. For the most part, the interpretation of inferences the data may
support are reasonable; however, there are a few instances where this reviewer considers staff to
have "stretched" to infer the data support their statement (see specific comments below).
Probably my major points of disagreement with the material in this chapter are:
• Discussion of a policy relevant assessment of the evidence for health effects concerning
the ranges for changes in FEV1 that are considered to be small, moderate, or large for
persons with impaired respiratory systems. Discussion of this issue during the August
24-25th meeting yielded an agreement to not change the table values but to emphasize
that a 10% reduction in asthmatics or persons with COPD could have more serious
consequences than a similar reduction in a person without a compromised lung.
• The chapter comes across with the view that central monitors in time series epi studies
are giving the true picture. Yet we know from material contained in the CD that personal
monitors consistently show lower levels. Staff seem to write this off but it has
tremendous implications for Chapter 6 relative to the case for lowering the 63 standard
and for margin of safety considerations.
• While the discussion of the low level exposures used in the controlled human studies by
Adams and colleagues is technically correct that no statistically significant changes were
found in FEV1 compared to filtered air, the fact that a reasonable percent of the subjects
had large decrements is glossed over. These were healthy subjects, so the percentage of
asthmatic subjects, if they had been studied, would be considerably greater. The fact that
asthmatics respond with greater differences is shown in other studies discussed in the
chapter.
• The lack of statistical power is offered for why there appears to be an inconsistent effect
seen for COPD mortality. While the same arguments applied for PM, I don't recall an
emphasis on a lack of power. You can't have it both ways - yes for PM and no for 63
using many of the same studies.
Specific Comments
p. 3-15,1. 14
It is incorrect to claim that short-term ozone exposures cause fibrosis.
The wording needs to be changed here. I believe the authors are
intending to convey that fibrotic changes occur with short-term
exposure.
P. 3-17,1. 14
Consistently, the authors seem to push the false negative possibility for
D-39
-------
explaining discrepancies when the finding they are defending could just
as well be a false positive. Better balance here and throughout the
chapter is needed.
p. 3-18,1.24 In this section, an Austrian study is cited as showing growth-related
increases in lung function over the summer as being reduced with
exposures ranging from 32.5 to 37.3 ppb. This study begs the question
how low can you go and what is really being measured as growth
related increases. This reviewer questions whether confounders are
present and whether this kind of study can really be interpreted as
growth-related.
p. 3-32,1. 6 The authors state that a positive association was seen between long-
term (^concentrations in the warm months, but this association is not
statistically significant. The writing comes across biased. If the
association had been slightly negative with no statistical significance, I
doubt the authors would have written about this. Caution against over
stressing weak associations as there is more than enough positive data
on other endpoints to carry the day that the current standard is
inadequate to protect public health with a margin of safety.
p. 35,1. 9 The rationale and accuracy of this sentence escapes me. Please clarify
why the assumption that the chemical reactivity of 63 does not induce
strong temporal correlations is needed or is correct.
p. 3-47,1. 2 The Sexton et al. (2004) reference cited here does not appear in the list
of references.
p. 3-55,1J 3 The authors seem to be stretching how far the data can be interpreted to
imply that 03 can have a role in producing effects on the cardiovascular
system. The current paragraph is highly speculative and should be
rewritten.
p. 3-57 This section on mortality-related health endpoints includes a number of
sentences that are a stretch of the available data or that need to be made
more factual if they are to be used. For example, on p. 3-58,1. 9 there is
a sentence that says the NHNE follow up data have 20% of adults with
a reduced FEVi value that suggests impaired lung function. What is the
% reduction being referred to? The reactions of O3 with cholesterol in
lung surfactant are stated to generate products responsible for
atherosclerotic plaque formation in arterial walls. Where are the data to
support this statement?
p. 3-61,1. 36 in the description of the FEVi change in berry pickers in Canada, the
adjective "large" is used to describe a 5% decrease. I seriously doubt
that pulmonary clinicians would classify this decrease as large.
Table 3-3 This reviewer does not agree that the same levels of FEVi decrements
as for healthy individuals should be used to describe the degree of
change in this variable for those with impaired respiratory symptoms.
The statement is made in the text that there is no need to change the
criteria from the previous NAAQS review. However, staff
acknowledge on p. 3-72 that a 5 to 15% change in FEVI is considered
D-40
-------
p. 3-81,1.
p. 3-84, 1.
14
1
to have clinical importance to asthma morbidity.
Strike "our" from this sentence.
What is meant by this sentence? Provide references for the statement.
Chapter 4 OAQPS Staff Paper - Second Draft
There are numerous places in Chapter 4 where only part of the data is presented with the reader
being required to go to technical support documents for the "rest of the story". This should not be
the case. If the data for one alternative scenario are presented, the data for other scenarios that
were considered should also be presented. The number of cases expected for a given health
endpoint and a given exposure scenario is helpful but staff need to also describe what percent of
the population is represented by these cases. The reader should not have to dig this out of
supporting documents.
As part of the sensitivity analyses associated with using the APEX model, staff have provided
some information on the number of children experiencing at least 3 or more exposures above
various ozone levels for Boston and Huston with changes in the activity database, ozone decay
rate, air exchange rate, etc (pages 4-44 to 4-46). This reviewer appreciates finally receiving this
type of information, but staff have not gone far enough. One has to go to one of the supporting
documents (Ozone Population Exposure Analysis for Selected Urban Areas) to find that, when
considering exceedances above 70 ppb, for Boston the number of children having these events
with moderate exercise represents 7% of the children and about 1-2% of the children in Houston.
Moreover, the figures on pages 50-52 of that document show large numbers of children with 5 or
more exceedances.
The frequency of exceedance is at the heart of the question about the level of the standard in
order to protect against adverse health effects. This reviewer would submit that a child
experiencing only one exceedance to an ozone level slightly above the current standard is not
likely to be of significant concern from a public health perspective. However, the same child
having 3 to 5 exceedances would be of great concern due to the repetitive injury to the lungs and
the increased chance for more serious health outcomes to be experienced. Thus, for every
scenario for potential alternative standards, this reviewer wants to see the percentage of children
that will have 3 or more, 5 or more, etc exceedances.
Specific Comments
p. 4-2,1. 28
The statement is made that staff are modeling intake of 03 and not dose,
which is correct. However, staff fail to make use of the dosimetry models
that are available. This reviewer has heard that after the last NAAQS
review staff were told to forget about using dosimetry models because
certain "scientists" did not believe they could be used. Significant
improvements in ozone dosimetry modeling have been made over the last
decade. This reviewer would be glad to debate the merits of the dosimetry
models and their uncertainties compared to the uncertainties present in
D-41
-------
various biological endpoints and epidemiology studies.
More importantly, by not considering the actual dose of ozone to the
centriacinar region, we may well be putting all our eggs in the "asthmatic
child basket" when outdoor workers, who inhale much larger amounts of
ozone over longer time periods each day and over the entire ozone season
may well be the most affected subpopulation. As Dr. Charles Plopper
pointed out during the O3 CD discussions, the frequency and duration are
important determinants of the type of injury and the probability for the
development of serious pulmonary disease. Discussions with EPA staff
during the August 24-25th CASAC meeting indicate that there are not
sufficient data available to be able to model risk to outdoor workers, and
staff agreed to emphasize this point in the final staff paper.
p. 4-5,1. 26 Information is given on how older photocopiers and laser printers run
under highly unlikely scenarios can generate 150-180 ppb of ozone. Can
staff provide any data on how many such older copiers and printers are
still in use? This reviewer highly doubts than many such machines are still
in use.
P. 4-14,1. 15 Why do staff contend that a source of uncertainty results from modeling
more than one ozone season? Having more seasonal data simply provides
data on the variability over years, which is a good thing to know.
p. 4-16,1. 19 The age of the diary information is a significant source of uncertainty for
the exposure analysis. About 50% of the diary data are more than 15 years
old. This reviewer strongly feels that the extent of computer and video
game availability and popularity makes the use of these older data
meaningless.
p. 4-19,1. 16 Classifying 45% of the children as "active" seems too high. How much of
this classification depends upon the outdated diary studies?
p. 4-21,1. 10 More and more communities are going to year-round schools, so not
including school commuting may have more of an impact than staff have
supposed.
p. 4-47,1. 18 The APEX model is shown to have a negative bias, which is 30 to 50%
most of the time. The paragraph discussing this minimizes the potential
impact of this bias. Material in the technical support document should be
brought forward for an expanded discussion of the bias. This review
considers the bias to be a major weakness in the risk assessment if the
implication that significantly more exceedances than predicted are
actually occurring due to APEX always under predicting exposure
D-42
-------
Dr. Maria Morandi
POST-MEETING COMMENTS ON THE 2ND DRAFT OF THE OZONE STAFF PAPER
Maria Morandi
Comments on Chapter 6
Charge Question 1: What are the views of the Panel on the approach taken by staff (as discussed
in Chapter 6) of using both evidence-based and quantitative exposure- and risk-based
considerations in drawing conclusions and identifying options as to a range of standards to
protect against health effects associated with exposure to O3, alone and in combination with the
ambient mix of photochemical oxidants, for consideration in this review of the primary O3
NAAQS?
Answer: In general, Chapter 6 delineates clearly the rationale of the staff for the
recommendations to be proposed to the Administrator, with appropriate referencing to the
relevant sections of the CD and/or Chapters 3-5 of the staff document. Using both evidence-
based and quantitative exposure and risk considerations is a reasonable approach. The available
scientific evidence since the last CD overwhelmingly supports the staffs recommendation that
any new primary standard should be at least as protective as the current one. Clearly, there is no
evidence that the current standard is too conservative. The evidence also supports the staffs
decision that, while there is some evidence for chronic effects on lung function growth, there is
insufficient data for establishing a quantitative standard directed at this chronic effect at the
present time.
Chapter 6 elaborates in detail on the additional evidence for effects for which the evidence was
very limited at the time the current standard was set - specifically cardiovascular and mortality
outcomes and effects on asthmatics, especially children. My sense upon reading this discussion,
which reflects the CD but not the Review Panel's general sense on the robustness of these
studies, is that evidence for cardiovascular/mortality outcomes is still limited and based on
approaches that raise methodological and other questions. There is a contrast between these
concerns and the weight the mortality studies receive in this Chapter, and elsewhere in the
document, although, in the end, they are not included in the formulation of the various standards.
Charge question 2. Does the Panel generally agree that the range of alternative primary O3
standards identified in Chapter 6 is generally consistent with the available scientific information
and is appropriate for consideration by the Administrator?
On balance, the scientific evidence for effects on asthmatic children and the risk analysis in the
Draft document does not support the recommendation that the current standard is sufficiently
protective of this sensitive population, so that a lowering of the 8-hour standard appears justified
on this basis. However, it is not clear to me that the evidence for supporting 8-hour standard
would be protective of for this group without consideration of a shorter-duration standard is
presented in a convincing manner as indicated in conclusion statement (2) on page 6-43 (also
lines 14-22 on page 6-30). Clearly, this statement is appropriate for effects observed in healthy,
adult volunteers, but the evidence presented in the document does not shows that it would be as
D-43
-------
protective of sensitive asthmatic children without a shorter-duration standard in locations where
such excursions occur (i.e., cities such as Houston and LA). There is a somewhat "waving of the
hands" feel to the discussion in the Chapter that one side describes the evidence for effects on
this sensitive subgroup below the 1-hour or 8-hour standard, but appears not to follow through
in the recommendations to the Administrator (either for the 0.08/4 or 0.07/7 ppb/8hr standard).
The evidence and risk analysis presented in Chapter 6 do not support keeping the current
standard as an option.
Charge question 3. What are the views of the Panel on the key uncertainties and O3 research
recommendations discussed in Chapter 6?
Answer: The extensive discussion of the 0.08 ppm vs. 80 ppb concentration appears excessive in
light of the uncertainties inherent in the exposure and risk evaluation. The significance of the
concentration rounding can only be evaluated in light of an uncertainty analysis that includes
both exposure and risk uncertainty, and it should be placed in that context.
Other issues include:
1. Adequacy of ozone as a surrogate for effects from other oxidants: it is correct to say that there
is no evidence to support other indicator for airborne oxidants at this time. However, actual data
to support this statement are limited very limited (i.e., lack of evidence may be the reflection of
lack or research). The lack of data available to support the use of ozone as the only and best
indicator for photochemical oxidants can be linked directly to recommendation (3) on page 6-46
under research needs.
2) Addition to research needs: Past and current evidence shows that there are especially more
susceptible subgroups among otherwise healthy individuals or those with compromised health
such as asthmatics. These subgroups are generally called "responders" but the factors associated
with such increased susceptibility remain unknown and, consequently, they may still be at risk
even with a more restrictive standard. Studies are needed that can identify the underlying causes
that make a subgroup among susceptible individuals particularly sensitive to the effects of ozone
(or other agents). It would appear that there are different phenotypes within subpopulations with
specific chronic diseases, such as asthma, and that each phenotype may be more sensitive to
different specific agents, one of them being ozone. Understanding the underlying characteristics
that make, for example, a subgroup of asthmatic children especially sensitive to ozone would
provide more robust information for setting a protective standard.
Other general comments
In general, the 2nd draft reflects the scientific evidence presented in the AQCD. There are
editorial issues that can be corrected (some of these corrections have been made already) that the
Staff can address through a thorough editorial review.
Chapter 2 presents the air quality characterization and analyses in a clear manner that reflects the
AQCD. However, there is a strenuous effort that drives the analyses for supporting a higher
D-44
-------
PRB. As discussed by the Panel during the meeting, the uncertainties in model estimates of PRB
for the 12 cities included in the study remain large undetermined. Ozone levels monitored at
Trinidad Head are probably indicative of extracontinental background at that site and perhaps
along the western seaboard, but not for areas well within the continent to the eastern seaboard.
Quite clearly, the PRB has implications for compliance with the standard and, therefore, is an
important consideration, but evaluation of commonality in temporal patterns of concentrations at
the Trinity Head site and elsewhere in the continent without more fundamental support basic on
basic principles is not warranted for a reactive pollutant such as benzene.
Chapter 3 also presents the available evidence in a manner that reflects the AQCD descriptively,
but there is fussiness in the presentation as to 1) which are the more robust studies and end points
that will form the basis for supporting the standard, and 2) which is the specific sensitive
subgroup the standard is meant to protect. While the AQCD presents the plethora of studies and
their results without engaging in a critical review, the Draft document needs to be more
purposive and clear on its description of the key scientific findings (as compared to the prior
AQCD) that will form the basis for the NAAQS recommendation. As indicated in the review of
the AQCD and by Panel members in the review of the 1st and 2nd Draft, the time series
mortality studies raise a number of methodological questions beyond just ozone, but they are
discussed extensively with limited critical evaluation. Given the methodological uncertainties,
apparent consistency of results may not necessarily be evidence for the strength of an association
but perhaps reflect the similarity of assumptions inherent in the models (and which are typically
never stated in an explicit manner). Given the more robust evidence for effects found in the
human chamber studies and the more epi evidence indicating effects on asthmatic children, it
would appear that the extensive discussion of the mortality effects and the comparatively
speculative nature of the proposed mechanisms are not warranted at this time.
Chapter 4 presents a fairly reasonable assessment of exposures under a variety of scenarios. As
with any model, evaluation of internal/external consistency and model predictions is critical, but
there are very few personal exposure data available for this purpose.
D-45
-------
Mr. Charles Plopper
Comments Regarding Chapter 3-Charles G. Plopper
This chapter is a very satisfactory summation of the health effects evidence evaluated in
Chapters 4 through 7 of the CD. In almost all cases, the description of the materials is accurate
and provides strong substantiating evidence for the determinations of the subpopulations with
elevated susceptibility that may be of greatest public health interest and the understanding of the
health-related effects associated with exposures. There are two aspects of this analysis which are
germane to understanding health effects evidence that need some expansion. The first of these is
the phenomena of tolerance. What this refers to is that the exposure history for an individual will
modify their response to subsequent exposure. This is a scientifically well-documented
biological aspect of the response to repeated ozone exposures over a period of time. This will
have a tremendous impact on such issues as the determination of threshold in both epidemiologic
studies and experimental studies using controlled exposures. It is therefore difficult to address
the issue of threshold, because the threshold will vary depending on the air quality of the city in
which a study population resides. Seasonal variations in 03 levels in the environment will also
modify the determination of no effect level in the same study population depending on the time
of the season when testing is done. While this issue is emphasized, especially in section 3.4.2.3,
the discussion could use clarification in terms of the biological reasons behind the seasonal
variation in responses. This may explain some of the lack of strong positive associations
between short-term ozone exposure and respiratory mortality, such as discussed in section 3.5 .2.
An example of the impact of exposure history on analysis of response is the Vedal, et al. (2003)
study which was carried out in a region with very low 03 concentrations and found significant
associations with mortality at very low concentrations. Studies conducted in such low pollution
environments may be stronger indicators of the level of risk and the threshold for effect,
especially in more susceptible populations.
As an editorial note, in some portions of Chapter 3 there is reference to a lack of data from the
experimental animal studies on long-term exposure affects. However, in other portions of the
chapter, there is discussion of long-term exposure studies, of which there have been an
abundance.
The authors should not be hesitant about making strong inferences from the data available. The
new data is accurately summarized in the CD and in Chapter 3. It is more strongly supportive of
the current standard than was the data when the previous standard was set and indicates that the
current standard may be too high to be protective of susceptible populations such as children and
the aging. Further, there is a growing body of evidence, well documented by the CD and
summarized in Chapter 3, for cardiovascular health effects which must be taken into
consideration for evaluating a new, lower standard.
D-46
-------
Dr. Armistead (Ted) Russell
Expanded Review of the OAQPS Staff Paper - Second Draft
Armistead (Ted) Russell
The comments below deal primarily with the Second Draft of Chapter 4 dealing with
Characterization of Human Exposure to Ozone, as well as the related Ozone Population
Exposure Analysis for Selected Urban Areas report, the Analysis of Uncertainty in Ozone
Population Exposure Modeling memo, and the Uncertainty in Ozone Measurements Memo.
This document includes prior comments, additional comments after the August 24-25 meeting
(provided in italic), clarifying prior comments (provided in italic), and additional remarks (also
in italic).
Response to Charge Questions:
1. To what extent are the assessment, interpretation, and presentation of the results of the
exposure analysis as presented in Chapter 4 (and in the second draft Exposure Analysis technical
support document) technically sound, appropriately balanced, and clearly communicated?
The chapter is more clear this time around, and is reasonably balanced. It draws on the
associated report well, bringing forward the main topics and findings. It should, however,
be more clear as to how the APEX model and quadratic role-back are specifically used in
the impacts analysis.
2. Are the methods used to conduct the exposure analysis technically sound? Does the Panel
have any comments on the methods used?
For the most part, they have used appropriate methods. The addition of using 2002 is
appreciated.
3. To what extent are the uncertainties associated with the exposure analysis clearly and
appropriately characterized in Chapter 4, the Exposure Analysis technical support document, and
the uncertainty memorandum?
I appreciate all of the work they have done and are doing on the uncertainty assessment.
Specific comments are found below.
4. To what extent is the plan for the remaining uncertainty assessment technically sound? Are
there other important uncertainties which are not covered? What are the views of the Panel on
sensitivity analyses conducted to evaluate the influence of uncertainties in the exposure analysis?
They are heading in the right directions, though I recommend that they also conduct an
uncertainty apportionment as part of their studies. Try not to get too caught up in details
and aspects that are so poorly defined that it will slow down the overall assessment. They
have done well to state that they will not include some uncertainties.
D-47
-------
Remarks on the second draft of Chapter 4: The second draft of Chapter 4 has responded to
many of the comments made on the first draft, and is thus more clear than before. I was pleased
to see the reanalysis for 2002 in addition to 2004.
Specific comments:
Page 4-47: They show a bias from APEX, but do not provide information on how that impacts
their analyses, and also tend to downplay the magnitude of the bias. 50% is big, and the bias,
while smaller in other cases, is consistent. The bias shown continues to concern me, suggesting
that the risks/exposures to levels of concern based upon the exposures found by APEX are low.
If staff can address this, it would be good.
4-46:1 might refrain from calling the uncertainty analysis as being results just yet.
Chapter 2: I am pleased to see the measurement and design value uncertainty added to this
chapter. They should add to this chapter the possible bias in the measurements as presented,
noting that this may mean that in past analyses (particularly those depending upon routine
monitoring data), the levels of ozone experienced may have been less than measured, thus
increasing the risk at lower exposure levels. They should also suggest that future instruments be
designed to address this issue.
In regards to PRB, given the contentious nature of the issue, the use of the PRB should be
discussed in Chapter 2, along with other measures. However, their current approach has been
peer-reviewed, and is appropriate. They should, before the next analyses, continue to refine and
evaluate their approach to setting the PRB, and how it is used in the assessments. For example,
might they use the PRB in the roll-back formula? This could be done by analyzing CMAQ
simulations where the boundary conditions are set to those conditions representative of PRB
conditions.
Chapter 6: This chapter should deal not only with the recommended standard, but also with the
precision of the standard. They have done an analysis of the uncertainty in the observations, and
find that most of the uncertainty is due to instrument bias. They have also looked at the impact
of the precision chosen on results. It is apparent that the exposure is sensitive to the choice of
not only the standard, but also the precision chosen. It would seem unlikely that given the large
number of monitors used in the various health studies upon which the standard is based, it is
unlikely that the bias identified for a single monitor is appropriate, and this should be reflected in
the precision of the standard. This chapter should discuss this issue and make an appropriate
recommendation.
/ was pleased to hear that they are planning to suggest that the standard be given to three
decimal places (e.g., 0,080 or 0.070 ppm) as this is strongly suggested by their analyses. This
should not be just recommended, but strongly recommended.
As an engineer, I can not speak so directly about the interpretation of the health data, but they
very specifically note that (page 6-16) depending upon the year rolled back to meet the current
standard, 20,000-600,000 children are estimated to experience exposures of concern. Thus, the
D-48
-------
current standard is not protective, and would leave a large number of children at risk. It should
be noted that the range should not be interpreted as that the number of children experience
exposures of concern will vary depending upon year, e.g., from 20,000-600,000, so maintaining
the current standard will lead to a very large number of exposures of concern in some years. On
page 6-32, the SP notes that lowering the standard to 0.07 would still have, relative to just
meeting the standard, 0-10% of the population experiencing exposures of concern, suggesting
that there is a strong likelihood of continued exposure to exposures of concern. Such an analysis
argues that the maximum standard in the future should be 0.07(0). No data presented suggest
that the current standard is adequate.
Remarks on the second draft of Ozone Population Exposure Analysis for Selected Urban
Areas: Again, the draft is more clear: thanks. However, they have not "corrected" some of the
mathematical presentation (e.g., the del operator) is still used inappropriately (why?).
Page 2: Sec. 1.2: Update year.
Page 19: Eq. 2-15: Need to correct subscripting.
Page 20: You switch units between hours and minutes in the analysis.
Page 27: I still think it should be "... differentiate between people.
Page 28: Are you able to correlate between meteorology, ozone and activity together?
Page 44, Pare 1: Awkward.
Page 64: Needs a bit more explanation.
Chapter 7: They really need to explain what the low bias from APEX really means and how it
will be accounted for. Also, why use ug/m3 in this section, ppm/ppb in others?
Specific Comments:
Page 2: Section 1.2: They still have just 2004 as the application year.
Remarks on the Draft Analysis of Uncertainty in Ozone Population Exposure Analysis for
Selected Urban Areas:
Overall: This is a nice start, but obviously is not a final product, so the comments are on what is
currently available. As a general question, are you planning to attribute uncertainty to specific
parameters and inputs? If not, why not? I would strongly recommend conducting a first order
(or higher) analysis on contributor attribution.
Page 2: Add "Bias" to the list of important concepts in the list.
D-49
-------
Page 8: Paragraph 1 is a repeat.
Page 15: Using the ratio approach will give a "bias" and it is not surprising that the central value
is 1.065 (what would appear to be a 6.5% bias). Instead, use absolute values, unless you plan to
use the ratio to scale your exposures. The last line of this paragraph is opaque, and they need to
come up with a better explanation/reasoning. I would recommend that they not sue the ratio
approach, but use the absolute differences in magnitudes, as this will likely show much less bias,
and should lead to less skew ness later in their analyses. If they wish, they can calculate the
relative bias as the (average absolute bias)/(average concentration).
Page 16 (et al.): Please label all axis of the graph. Also, the last sentence needs to be better
justified. The figure caption for Figure 7 (and similar figures) is insufficient.
Page 15: No need to use additional methods for spatial interpolation... they will give different,
possibly slightly better, estimates, (they picked two of the easiest approaches, not best) but the
difference will likely be quite minor.
Page 18: Table 7 mixes uncertainty units, likely unnecessarily.
Page 19: Fig. 10: Are the standard deviations available? Is the trend for O3 significant?
Page 21: Is the ratio really 0.12-0.16? This strikes me as a very low number for locations
outside of street canyons.
Pages 22- 27: The rollback process and the resulting Figures need to be better explained in the
text.
Page 39: If you are really going to do what is shown in Diagram 1, you can do some uncertainty
apportionment. Not sure why this would be done just to find influential statistics separate of a
complete uncertainty analysis.
Page 43 et seq.: I am not sure I grab why they are saying that averaging together a set of
simulations is not as accurate as having a single simulation with the same net number of
"people". Also, it is not obvious why you need 60,000 given the trend of Fig. 20.
Page 45: The spread shown really is not very big (i.e., why do they say that the spread is
significant?).
Figs. 21 and 22: Labels on all axes.
Page 55: The statement is made that the factor model parameters have no uncertainty by
definition. Please explain. I have trouble with this statement. They could have a more
sophisticated set of factors (e.g., more than two, not all multiplicative), that would allow for
more careful uncertainty assessment. They can make the statement that they are not examining
the impact of having the current form on the results and the uncertainty assessment.
D-50
-------
Page 59... We await the results.
As presented at the meeting: I was pleased with what I saw at the meeting, and rather than
attempt to be exhaustive, and not complete all of the analyses, continue as planned to examine
the most influential. They should concentrate on the uncertainties in the "differences " in
exposures (this was presented, and one notes that the differences are small). They need to be
ready to explain the skewed results, and a figure would help.
Comments on the Cox and Camlier memo: First, thanks for doing this. I was disappointed
that a bottom line recommendation was not given as how to treat the calculated uncertainty in
terms of the precision of the standard. This would indicate that the precision of the standard
should be to the nearest 1 ppb.
Other: As discussed with staff (Langstaff), a simple diagram of the rollback approach would
help significantly.
D-51
-------
Dr. Elizabeth A. (Lianne) Sheppard
Review of the Ozone Staff Paper Second Draft
Staff Paper Chapter 5
Response to charge questions - Overall comments
Heath Risk Assessment (Second Draft Chapter 5 of the O3 Staff Paper and draft Health Risk
Assessment technical support document):
1. To what are the interpretation, of the results of the
analysis as in Chapter 5 (and in the draft Risk
technical document) technically sound, appropriately balanced, and
clearly communicated?
I found generally the work in this chapter and the accompanying risk assessment to be well done,
balanced, and reasonably communicated. There are a few places where I suggest
improvements and I have included these in my detailed comments. Most important, the
presentation and summarization of the clinical evidence for the FEVi decrement dose-
response implies far greater certainty than I believe is warranted by the data. Figure 5-2
must be completely revised to reflect the uncertainty in the data. A new table
summarizing the derived study results used for this figure should be added to the Staff
Paper. The method for deriving the dose-response calculation, and the use of the logistic
function must be re-evaluated. Furthermore the sensitivity of the results to the dose-
response assumption must be assessed. The apparent certainty of the FEVi decrement
dose-response has major implications, particularly since the estimated Os-induced
decreases in lung function are lower than were estimated in the last review. Because of
the currently hidden uncertainties, this conclusion may not be correct and should be
revisited after gaining new perspective on the certainty of the estimated dose-response
function.
2. In general, is the set of health endpoints and concentration-response exposure-
functions in this risk for this review?
Given the literature to date, the set of health endpoints are appropriate. I find it more difficult to
infer causality from the epidemiological studies than the staff, but I agree that the
endpoints and estimates used are reasonable choices given the current state of the science.
It is important to state clearly in both Chapters 5 and 6 that there are many endpoints with
good evidence of effects cited in Chapter 3 that are not incorporated into the risk
assessment in Chapter 5.
3. Are the to conduct the risk technically sound? the
Panel have any on the used9
D-52
-------
My main concern is the use of the 3-parameter logistic function for the derived lung function
response results. Not only is the function problematic, but also the summarization of the
data prior to selecting this function allows the perception of greater certainty than is
warranted by the data to prevail. Because it has a large impact on the use of the clinical
study estimates from the risk assessment, the approach to fitting this function and its
sensitivity must be evaluated. See my detailed comments.
4. To what are the uncertainties associated with the health risk clearly and
appropriately characterized in both the Chapter 5 the draft
Health Risk support documents9
I was pleased with the effort to document the uncertainties. While it is difficult to know whether
or not all uncertainties are well enough understood to even document, the lists appear
thorough and it is clear that this issue was thoughtfully considered. In terms of risk
communication, be careful that thoughtful scientific consideration of the uncertainties
doesn't dilute the overall message.
Detailed comments
Please review all tables and figures (including captions) with an eye towards better labeling,
particularly of figure axes and table columns. Always include units and enough details so the
figure or table can stand alone out of context. For instance there are a number of figures that
have the ozone concentrations/standards listed and also include "Recent (2004)" as one of the
items. This is confusing out of context. It also may be helpful to give a detailed key of the first
figure in a series.
Thorough cross-referencing with the technical Health Risk Assessment document would be
helpful.
My concern is that focus on the risk assessment demotes in importance all other evidence in
favor of 63 health effects. Highlight a better in this chapter that the effects and endpoints used
for the risk assessment are a small subset of the effects that were summarized in Chapter 3, as
well as a subset of populations, locations, etc. An aspect of this is noted briefly on 5-7 9-12. It
is too easy to focus only on the risk assessment results and overlook the other large body of
evidence that is not included in the risk assessment. The write-up should help readers avoid
doing this. This point also needs to be brought forward into Chapter 6.
Make sure percentages are clearly defined. Avoid whenever possible using number of events in
the denominator. Instead use total population. If events are needed for some purposes, take care
that counter-intuitive comparisons and results don't get highlighted.
5-4 lines 1-11: I'm not convinced that we can characterize the uncertainty of risk assessment in
epidemiological studies as "considerably" greater due to location. I don't think we have great
enough understanding of the sources of uncertainty to quantify its relative magnitude. Similar
comment for the use of "considerably" in line 18.
D-53
-------
5-5 and throughout: While observing consistent results across epidemiological studies may be a
necessary condition for inferring a causal effect, I don't think that it is a sufficient condition.
While it may only be semantic, I find the judgments about causality as stated in this document,
particularly with respect to the epidemiological studies, difficult to accept. One reason is that the
epidemiological study relative risk estimates come from the time series study design and we have
limited understanding of the effect of this study design, particularly for such a highly reactive
pollutant such as ozone that does not penetrate well indoors. However, based on the perspective
of the precautionary principle, I do not object to the selection of endpoints and their use in the
risk assessment.
5-12 and throughout: Notation in this document is at times confusing. Anything staff can do to
minimize the variants of notation through the whole document, clarify the definitions, and unify
the presentation, would be helpful. One suggestion is to prepare a glossary of symbols for
Chapter 5 and/or for the TSD. (Note: eventually I found a glossary for some terms in Appendix
B.2. However B is not defined there. At a minimum refer readers to this appendix in Chapter 5.)
In the process of preparing the glossary, instances of redundant and confusing notation will be
more readily identified and can then be corrected.
5-14: It is unclear to me why the categories for "as is" and "background" would ever be
different (i.e. why distinguish between /' andy?). Is it possible to simplify the equations from the
most general forms and still capture all of the work that was done? Or are there examples where
this isn't possible?
5-16: I have difficulty with the use of the 3-parameter logistic function for estimating the dose-
response curve. While from the existing figure it appears to be better than the previously used
linear dose-response curve, it is not apparent to me how sensitive the results are to this
assumption and how necessary it is to make a parametric assumption, particularly from a 3-
parameter model fit using only 5 data points. The biggest problem with the logistic function is
that it is parametric so that a small change anywhere on the curve may have a large impact in the
region of primary interest - at the low ozone concentrations. In particular, note that the threshold
response rate is estimated from these data although there are no study results at the higher
exposures to verify this estimate. (As an additional point, I don't recommend excluding data that
could be informative for fitting this model and thus suggest the dataset be expanded to include
studies conducted at higher concentrations - e.g., the Follinsbee studies cited in the 1996 CD,
particularly if the logistic function will continue to be used. Take heed from the O-ring data
analysis on a restricted dataset used to justify the decision to launch prior to the Challenger
disaster.) The logistic function is by definition symmetric, so small changes to the threshold
estimate could have a large impact on the shape of the curve at low concentrations and
consequently on the risk assessment estimates that are so heavily weighted towards the smaller
concentrations. I have several follow-up suggestions: 1. Consider using a more non-parametric
approach to smoothing as the main analysis, and certainly at a minimum as a sensitivity analysis.
I suggest considering a natural cubic spline with one knot or pure linear interpolation. 2.
Consider weighting the contributions of each data point by their uncertainties. Better yet, start
from the raw data, not data averaged at each exposure level. 3. Particularly if a parametric
function is to be used, include additional studies to better pin down the height of the curve. 4.
D-54
-------
Conduct a sensitivity analysis for a few different curves, including a linear function if indeed that
function is consistent with the data when the uncertainties are included. If the risk assessment
isn't particularly sensitive to changes in the approach to smoothing, then it will be less important
to focus additional attention on the approach to smoothing at this late date. Based on the
comparison of these new results to the previous review, the risk assessment is very sensitive to
the shape of the selected dose-response function. Therefore more needs to be done to address
that sensitivity.
Figure 5-2 (and 3-3): Completely revise this figure. Show study-specific data, not summarized
data. Include uncertainty estimates on the figures. Show the point estimates and uncertainties
for all the studies (by e.g. jittering around the target concentrations). Also add a table that gives
all the results used to create the data in these figures (or cite the appropriate table in the CD).
Chapter 3 (Appendix Table C-2) has the data, but it is not in the same form as the data used in
this figure. There are no counts given in the table. Thus a new table needs to be developed and
included in Chapter 5 to show the data used in the figure.
5-18: The "Bayesian" approach appears to be only Bayesian in flavor. For one thing the "data"
in the "Bayesian" analysis are not data, but predictions from the logistic curves. While it is a
commendable effort to quantify uncertainty in these dose-response curves, there is no
recognition of the uncertainty due to the selection of the function itself. Also the data are not
used directly.
5-22 13-23: This is a place where including the notation along with the words will help the
reader immensely. I'm still not sure I'm completely straight in this section as to when "baseline
incidence rate" refers to B (the more typical use outside of this document) and when it refers to_y.
On first reading it was only on page 5-29 line 30 that I finally began to understand that^ is the
baseline incidence.
5-24 17-30: Drop all this and just give the definition of notation, i.e. exp(pz/x)=RRZ(x.
5-35 16-18: I suggest starting that sentence with "One" rather than "The". I think this is one
possible explanation but not the only possible explanation as is currently implied.
5-49 21-22: The reason to use the highest 8-hour monitor when comparing to the 24-hour
average is not clear. Epidemiological time series studies with 8-hour average ozone as the
exposure do not use the highest monitor over a spatial region. These appendix plots are
examples of plots that need to be more clearly labeled and described in the caption. Also the
distribution of the maximum is related to the total number of monitors. Include this information
(the number of possible monitors) in each plot.
5-56 21-23: It is helpful to know that the risk estimates are based on the accumulation of events
over seasons that are different lengths. This means from one perspective, comparisons across
cities are "unfair". I think this information belongs in the table as a footnote or extra column
(summarizing the percent of year for the ozone season) as well. Another idea is to add an
additional table in the TSD to allow cross-city comparisons for the single highest average
D-55
-------
exposure month (or some other reasonable choice that would make the estimates from all cities
temporally comparable).
5-71 9-11 and 31-34: This comment elevates the importance of my concerns regarding choice of
the logistic function.
5-7115: Figure 5-12 missing. Clarification: This is a typo; should be Figure 5-2.
5-73 8-10: It is unclear to me why New York was singled out for discussion.
5-76: I think there are also uncertainties due to the features/behavior of particular study designs,
particularly the time series study design that relies on aggregate data and does not incorporate a
one-to-one link between exposure and outcome on individuals.
Minor comments:
5-123: Why is there an /' subscript included for FEVi?
5-24 13: Replace this line by: subtraction equation (5-3) from equation (5-2) and multiplying by
y/y which yields:
5-33: An asterisk appears in the table without a matching footnote.
5-36 and throughout: I would prefer the term "sampling variability" over "sample size
considerations".
5.57 16, 17; 5-73 22: Periods missing.
5-76 43: Replace line with: model uncertainty (i.e., uncertainty about the selected model, e.g.,
the shape and magnitude of the
5-78 25-27: More information is needed to easily access this paper.
Staff Paper Chapter 6
General comments:
I found this chapter less well done than chapter 5. I found several errors, suggesting to me this
chapter contains more than an acceptable number of errors. I felt that the composite body of
evidence of 63 effects summarized in chapter 3 was discounted, perhaps because many of the
results cited in Chapter 3 aren't easily quantified in a risk assessment. The force of scientific
evidence should not be lost with focus on a limited risk assessment. In contrast, a few studies
(e.g. the Peters et al 2001 pilot study results) were given inappropriate attention relative to other
evidence in the literature.
D-56
-------
Most important, I did not find the evaluation of the risk assessment well grounded in public
health considerations. I think the first order of business for this chapter is to clearly define and
justify the rationale for criteria for public health benefits of any standard. I suggest public health
significance should be based on absolute rather than relative criteria. While the risk assessment
is admittedly restricted to a subset of endpoints, population groups, and geographic locations,
numbers from the risk assessment can be used for quantifiable estimates of public health
benefits.
The second problem with the focus in this chapter is that it gives relatively too much weight to
the necessarily restricted risk assessment (restricted to specific populations, endpoints, locations)
compared to the full body of scientific evidence for Os health effects summarized in Chapter 3.
This problem starts in Chapter 5 and is carried into Chapter 6.
One of the effects of the focus of this chapter is that the conclusion of possibly maintaining the
current standard was presented as defensible from some points of view. I think the body of
evidence only supports lowering the current standard. The chapter should be revised to focus on
a vision for acceptable public health impact. Then all the evidence should be brought to bear on
this vision without overweighting those results that have an accompanying risk assessment. This
will produce different conclusions.
I found the reliance on relative comparisons using cases as the denominators misleading (e.g. see
Table 6-3; also Figures 6-1 through 6-6). Percents based on number of cases are sensitive to
variation in the number of cases from year to year and this creates apparent differences between
percentages that are not real.
• For instance for incidence in Table 6-3 (see also 6-21 lines 17-18) the incidence
differences for recent vs. just meets the standard are identical but the percentages make it
appear that it is better to reduce O3 in a clean than a dirty year when in fact the absolute
improvement is identical in both years.
• The figures show percent changes relative to the current standard.
o It does have value in normalizing the data and allowing comparisons between
locations. However, since many locations already don't meet the current
standard, this point of reference can be questioned. At a minimum keep the point
of reference the same but also include the actual data as an additional exposure on
the figure.
o Percents can also mask the public health significance by magnifying small
absolute differences and the converse. For instance, going back to Table 3-11 in
the TSD and comparing FEV decrements in Atlanta, note that the relative change
for 0.084/4 to 0.80/4 is 11% for FEV greater than 10% and 24% for FEV greater
than 15%. However, for FEV greater than 15% affects 11,000 occurrences while
for FEV greater than 10% affects 65,000 occurrences. This is so much larger in
absolute magnitude that it cannot be ignored. See page 6-33 lines 13-15 for an
example of where the relative comparison ignores the large differences in
absolute effects.
Add some discussion in this chapter about atmospheric chemistry and the importance of reducing
oxidants on reducing overall pollution. Incorporate recognition that O3 is an indicator.
D-57
-------
Review the language used in this chapter for clarity.
Details:
6-1030: The O?, data used in Sheppard 2003 and the original 1999 paper are predominantly for
the warm season. Only one year of 8 years of data included winter measurements. Also correct
the reference.
6-11 8-10: This conclusion is confusing. Based on the 98th percentile ppb averages, clearly most
or all of the monitors in the Burnett et al 1997b paper met the standard.
6-11 29-31: Since the epidemiological studies estimate a linear dose-response relationship, and
most when they check this assumption find that it is well supported by the data, does it really
make a difference whether the areas in which the studies were done met the standard? Is this an
opportunity to incorporate the concept of tolerance?
6-12 22-23: Can this be stated more clearly? Define the relevant air quality statistics.
6-12 27: Revise/add to the sentence as follows: ".. .likely to produce different estimates for
reasons such as seasonal differences in pollution or confounding.
6-15, Table 6-1: Here is an example of where the percentages are defined and their use is helpful
for the interpretation.
6-17 15: Drop "for"
6-18 Table cell "just meeting" and "occurrences for FEV >20" the percent should be 90% not
85%.
6-21 17-18: Note that the absolute changes here are identical while the relative changes are
different! From a public health point of view, there is no difference between the reductions in
the two years.
6-22 17-18: Note that the estimate from Bell et al is for all year even though the exposure data
are restricted to the 63 season.
6-22 Table 6-3: Does it make sense the 2002 incidence per 100,000 is higher than the recent air
quality incidence rate?
6-28 11-12: Improve wording, e.g. drop the second "important".
6-30 4: Citing only the pilot study results ignores the main study results which should be viewed
as scientifically better grounded.
D-58
-------
6-32 1-3: This sentence is unclear. Would this argument benefit from the evidence in the most
recent Bell et al paper (2006) that shows the same effect estimate as days are excluded at
progressively lower exposure cut-offs.
6-32 20-27: Here is an example where the relative values may not give a full perspective.
6-33 12-15: As noted above, the relative comparisons hide some important information. At a
minimum make it clear these comparisons are relative by e.g. adding "relative" after "greater" in
line 13. Better yet, revise to acknowledge the absolute impacts as well. Incorporating absolute
changes will also force the discussion to be more mindful of the very limited subsets on which
the risk assessments were done.
6-33 29: Define reduction. Reduction in what? This is an example of a more pervasive issue in
this chapter.
6-42 13: Figures 6-1 and 6-2 don't have a and b.
6-49 36: Line missing.
Staff Paper Chapter 3
Specific comments:
Exposure error section 3.4.2.1, particularly page 3-37: This page needs improvement to convey
a clear message. As a preamble, it can be helpful conceptually to partition total personal
exposure into ambient and non-ambient sources. The ambient source can be simply assumed to
be a product of ambient concentration and attenuation due to the building filter (call this a).
Absent residual measurement error and confounding, an individual study such as a panel study
using personal exposure to a pollutant will estimate toxicity of the pollutant (e.g. P). Assuming
measurable exposure, a panel study using ambient source exposure, will also estimate toxicity
(P). The same panel study using ambient concentration instead of ambient source exposure will
estimate attenuated toxicity (toxicity times a=pa). Similarly, a time series study using ambient
concentration instead of personal exposure will also estimate attenuated toxicity (Pa). This is the
point being made in Sheppard et al (2005). Lines 5-9 don't clearly summarize Sheppard et al.'s
simulation study results. A time series study does not estimate the same effect as an individual
study with personal exposure would estimate. Sheppard et al (2005) showed no further
attenuation as one might anticipate be present due to the poor correlation between personal
exposure and ambient concentration. Their results showed that when total personal exposure
was the true exposure, a time series study estimates the effect of concentration, which is the
product of toxicity of the exposure and attenuation due to the building filter.
3-37 24-25: I note that point (2) from the Zeger et al paper also results in the point made above
(i.e. that estimates of P aren't the same as estimates of ap). The way this page is written it
appears that these two points conflict.
D-59
-------
3-38 10: The Brauer et al. reference is missing.
3-39 lines 1-4: Drop the rest of the sentence after the comma. Not only does this confuse the
reader, but it is irrelevant to the Staff Paper since the epidemiological risk assessments always
use concentrations as the exposure. Line 5: change "these" to "the epidemiologic study". It is
correct that the health effect parameter in air pollution time series studies doesn't estimate
toxicity (called potency in line 5). However, the relative risks for increments of concentration,
while not directly interpretable as potency of O3, are interpretable.
3-43 2: This Peters pilot study result should not be cited over the primary study result.
AX6-4: The >6 hr exposure results are not summarized in a way that is used later in the SP.
This should be rectified here or in a later table in Chapter 5.
General Staff Paper Comments
Where possible, enhance the document with comments to help the reader navigate and look
forward and backward in the document. Add cross-references whenever possible. To look
forward, mention planned uses of calculations. To look backward, mention source numbers from
previous chapters. Make sure every figure and table has a thorough caption and good labeling so
it can stand alone out of context as much as is feasible. Most readers will not digest this
document from start to finish. Therefore anything that can be done to help a reader starting in
the middle find relevant information and assumptions from other parts of the document will be
very helpful.
Health Risk Assessment TSD
Comments that weren't completely covered above in SP Chapter 5 comments:
3-4, section 3.1.3: This section does not adequately document the details used for the
calculations. The raw data must be presented. If the averages for each exposure are to be
used for deriving the dose-response function, then the calculations and intermediate steps taken
to obtain those averages must be presented. (I would prefer that the analysis be redone using
only the raw data presented as estimated probabilities of response at the different exposures.)
3-4: Document that the logistic function fits clearly better than a linear function once the
uncertainty in the data is included. Once the uncertainties in the study results are incorporated, I
suspect this conclusion is not correct. If the logistic function is to be retained, extending the plots
(Figure 3-3) to include higher exposures in order to show the threshold level would help with
physical interpretation of the dose-response function.
4-40: I agree the lack of daily incidence data shouldn't be a problem, but can't you use the city-
specific mortality or admissions data comparable to those used in the epi studies?
D-60
-------
References: Here is one study I found from a web search that includes >6 hr exposure above
0.12 ppm. I understand from my clinical colleagues that such studies were reviewed in the
previous CD.
Kehrl HR, Peden DB, Ball B, Folinsbee LJ, Horstman D. Increased specific airway reactivity of
persons with mild allergic asthma after 7.6 hours of exposure to 0.16 ppm ozone. J Allergy Clin
Immunol. 1999 Dec; 104(6): 1198-204.
D-61
-------
Dr. Frank Speizer
Review of second draft of staff paper on Ozone Dated July 2006
Submitted by: Frank E. Speizer, CAS AC
Chapter 1
Notably on page 1-3, line 25-30 the reading of Section 109(d)(2) states that an
independent committee .... shall recommend to the Administrator and new ... .or change in the
standard. That committee since the early 1980s has been CAS AC and it therefore seems to be
more than an advisory committee that only approves or rejects what EPA staff does.
Chapter 2
Page 2-5, line 17-19, in describing the findings in table 2.1 for NOx emissions it is clear
that emissions have come down, in the face of increasing cars on the roads, and it probably worth
indicating the reasons,, e.g. Better controls, catalytic converters, fuel processing, change in
reference methods??? Say something. Same applies to table 2.2.
The box plots in Figures 2.2-2.4 of the 1, 8 and 24 hours show striking uniformity across
the country. This is indicated by suggestion in the text that the expected excesses in California
are not localized to California alone, but it might be more explicit to state that the average levels
are similar and for the 8 hour the levels above .07 represent 10% of the levels recorded in all
parts of the country. (I note this is roughly seen in Figure 2.6).
Text to page 2.40, mostly describes what is in figures, without providing much in the way
of interpretation. If one just focuses on the 8 hour data it seems pretty clear that in many of the
cities, and for many of the time trends there are a substantial number of excedences each year (
e.g. Figure 2.17 shows monthly averages of 8 hr 4th largest over .08 ppm with no particular
improvement with temporal trend. This is not discussed (maybe it is discussed elsewhere). Even
if it is, it should be mentioned here.
Page 2.48, last sentence. Would it not be worth a statement that each of the cities have
remarkably consistent finding for both month and time of day? Again this might be discussed
further in later chapters, but certainly no reason not to say it here since the subject is raised.
With regard to Martin Questions:
See above. Although it appears that the data are presented and brief descriptive analyses are
presented in the tables and figures, some interpretation needs to be added to text.
Chapter 3
The set of appendix tables (Appendix 3B and 3C) are excellent. They are laid out in a
logical fashion, and should be a model for the future.
D-62
-------
Page 3.6, lines 7-12. The terms triangular exposure and square wave exposure represent
jargon of the field, and may not be understood by most readers. Suggest definitions be supplied.
Page 3.9, line 27-28: For clarity the odds ratio of 1.37 should be indicated in this sentence
to be per 30ppb (if it is).
Page 3.12, line 7-8. Need to reconcile this statement with the first paragraph of section 3.3.1.1.3
on same page. It seems inconceivable that inflammation is not involved in airways
responsiveness and is involved in exposure in the lung. Certainly some part of the airway
system, albeit smaller airways, which contribute less to responsiveness, are involved in the
inflammatory response of the lung. In fact rest of page and next goes on to indicate that some
inflammation must be present in the airways.
Section 3.4 starting on page 3.34-4.47, title is a little confusing. Certainly what has come
before is much of the epidemiological as well as other evidence and I would have called it an
assessment thereof. To label this section "Assessment of..." seems odd. After reading it I
come away with the feeling that it is an attempt to justify what has come before on the basis of
Hill's postulates. That's ok, and maybe a little more thought to make the title of the section more
appropriate is in order. There is also a style issue in this section. Each of the subsections
concludes with essentially a quote from the CD. This really isn't an interpretive statement but
more a restating of what the CD states.
Page 3.62, line 14: Suggest change for clarity to: ... 2 fold greater decline in...
Martin Questions
1. The evidence is presented in a sound effective manner. In fact it is not clear to me
what will be different in the subsequent chapters on health as much of the information
that will be contained can only be what is here in greater detail.
2. The discussion of key issue is also well presented. It is organized effectively,
however, if anything toxicology and controlled human exposure studies seem to be
played down somewhat. Other epi issues are well discussed.
3. Threshold is adequately discussed.
Chapter 4
The data in table 4.4 suggest that a relatively small number of diary-days, particularly for
the children aged 5-18, were actually obtained. In addition, the range of days varies from 42 to
4332. Yet, as far as I can tell when these data are used each city is given equal weight. When
these data are used shouldn't the results be weighted by something like an inverse square term to
take into account the relative uncertainty of the data? Maybe this is done in the generation of
values but it needs to be specified somewhere.
There is mislabeling of the Figures in the text (Figure 4.1 and Figure 4.2 in text on page
4.30, lineS instead of 4.2 and 4.3. This continues as a problem on line 16.
Martin questions
I found this chapter relatively easy to read and quite informative in its use of the other
related documents from which it draws its information. However, I have a problem in that much
D-63
-------
of the information used is not from the CD but from the other documents, some of which are still
called drafts. I know that we recommended the use of "grey literature" but I would not consider
the ancillary documents here that grey literature. This is an issue I think we need to discuss as a
group and choose and go on record as endorsing the approach. I think it is ok but I think
CASAC needs to sign on.
The methods used are reasonably sound. Uncertainty is referred to, in citing another
document, but probably needs to be explored further. For example there are no error bars on any
of the later figures that give estimates of population exposed to specific levels. I do not think
these must be calculated for each point but some general examples need to be given. Because
there are such differences in the data by each city, are there ways they can be combined that
would allow for more robust assessments, perhaps by using dummy variables for city while
considering all data together. Other more sophisticated statistical approaches might be
considered.
Chapter 5
Page 5.7, lines 10-12: This is an important caveat. The suggestion is that the current
quantitative risk assessment gives only a minimum estimate of true short term risk. This needs
to be discussed further.
Page 5.8, Para beginning line 8: This suggests that by not using emergency room visits and
school absences in current risk assessment, the risk assessment is leaving out large potentially
important health as well as economic impact of exposures.
Page 5.20, Figure 5.3a, b, c: In contrast to Figure 5.2, each of these components extend the
curves beyond the data (0.12ppm) and seem to imply the exposure-response curve is
approaching an asymptote. I do not think this is justified and suggest curves be truncated at
0.12ppm.
Martin questions
Assessment seems technically sound and methods seem appropriate. The question that
comes up is to what degree is the concern about uncertainty a reflection of the lack of data on
endpoints that are in fact not used in the assessment. E.g. all might agree that cardiovascular end
points lack specificity and lead to uncertainty but in fact they are not used in the calculations of
the assessment. With regard to the technology used, to the degree it is dependent on the Draft
Risk Assessment technical support document, I would think that a critical evaluation of that
report and its final status will need to be documented somewhere in the process before it can
officially be used as a basis of the assessment.
Another point of concern is the details on the suggestion that the effective mortality is
related to tolerance. The issue is raised but not adequately discussed.
D-64
-------
Chapter 6
The chapter is reasonably well written and essentially summarizes what comes before.
One issue mentioned below seems to me not adequately addressed.
Page 6.6, sentence beginning line 23. This needs rephrasing or could simply be left out. One
could just as easily argue that the no threshold was found as to neither supports or refutes, as it is
all a matter of the level of exposure used in measures of effects. The rest of the paragraph
discusses this issue adequately.
Page 6.17, line 6, typo: 2004
Page 6.30, end of paragraph at line 13. Two issues that are not adequately mentioned in this
chapter but might be important modifiers in a discussion of the averaging time 1) are the
quenching of O3 by NOx, that may be occurring in heavily traffic situations and 2) the natural
night time drop of O3 that occurs as sea level sites (thus making the 24 hour measure an
underestimate of high exposure for shorter periods). Surely these are part of the logic of
selection of 8 hours as the measure of choice.
6.47, line 16, typo: microenvironments
Martin questions
The combination of evidence-based and quantitative exposure-and risk-based consideration
seems appropriate, but as mentioned in discussion of chapter 3, much of the risk assessment is
based on draft documents that have not had adequate peer review. I know that a request has been
made to have CAS AC review (and some of us will clearly get to that) but I am not sure that will
be adequate and we will need to discuss the issue.
The presentation on Page 6.44 on whether to retain or change the current standard is appropriate
but as presented results in an unbalanced form. Paragraph a indicates the standard could be
retained but bases this decision on a very narrow view of the data. This could be interpreted as
saying that only controlled human study data with a 15% decrement in lung function is an
adverse effect that should be considered. (The paragraph goes on to say that this would ignore
much other evidence, which the administrator could choose to do). However, there must be a
way to indicate some of the relevance of the other disciplines that need to be considered in
making a decision to retain the existing standard. Whereas paragraph b is more thorough in
considering the data, and emphasizes the potential importance of albeit less precise but perhaps
as relevant, population based data.
One way out of this dilemma would be to reverse the order of the presentation and give the
alternatives for changing the standard before the idea of retaining the standard. (I recognize that
this would be different than the way this is done in the past, but the logic of not using all of the
data seems inappropriate otherwise).
With regard to the key uncertainties and research recommendations it is disappointing to realize
that we have made so little progress since 1997, although I am not so sure I would totally agree
D-65
-------
in that conclusion. In fact, because of the work that has been done, particularly in working out
the methodologies for PM, a lot of the methods questions have been answered. New issues have
arisen, and these have been summarized reasonably well. Perhaps more could be made of the
approaches that might be used to explore mechanisms and potential gene-environmental
interactions that may be important in identifying subgroups of susceptible populations.
Review of Analysis of Uncertainty in O3 population exposure modeling
Dated July 24, 2006
Submitted by Frank Speizer, 8/23/06
General Comment:
This document claims that in its next iteration it be included in its final form to inform
the final Ozone Staff Paper. It seems clear that unless a substantially revised document already
exists (and this one claims it is waiting for in put from CASAC and the public) that this will be
an impossible task. Much of the document is a promissory note of what the author is planning to
do. Simply reviewing the existing literature cannot be accomplished without a Herculean task to
have some thing ready in the next month that will be informative for the final staff paper. This is
another example of too little too late and my suggestion would be to put this aside and find
something more useful to do with the remaining time before the court ordered deadline is upon
us. I certainly would not be able to sign off on this document and certainly there will be no time
to review the next iteration.
Specific comments:
Page 5, para 4, line 5: Recognizing that the author is referring to another document, it is not
clear where these SD are coming from. Need to indicate that these are simulated models with
arbitrary selection (if that is what they are).
Page 8, first para: Editing: This says the same thing with the same words as that on the top of
page 4
Page 11, para 2, end: If 2003 and 2004 are not correlated, then what is the meaning of indicating
the averages from the table (I assume author means all the numbers) have average values as
given? Further what does it mean that the range of bias for 2003 is -7.35 to +11.28!?!
Page 13, first paragraph. It is not clear how these statements will be used. In fact meaning is not
clear.
From here on out the promissory nature of the discussion becomes increasing frustrating and
indicates to me that the tasks suggested will never be completed in association with the time
frame proposed.
Page 15, end of first paragraph. What is the use of indication of what is said and is it logical to
assume that the interpolation will be unbiased?
D-66
-------
Next paragraph: We plan to perform... It hasn't been done, will it be done for this report or
future reports?
Figures 6-8: Should there be error bars on these curves?
Page 19, Table 8: Not clear why all upper bounds are equal to 1.0
Figure 10, error bars?
Page 20, Table 9: Does this make sense that all values save in-vehicle, interstates are virtually
the same?
Bottom of page 20, top of page 21, first 2 para. Too much work is being planned for an October
deadline.
Because the remainder of the document suggests that work will be done, I see no way to
integrate this into the staff paper at this time.
D-67
-------
Dr. James Ultman
Chapter 6 in the OAQPS Staff Paper for Ozone (2nd draft)
Reviewer: JS Ultman
Date: 8-25-06 (revised)
This chapter is well-written and logically presented.
Based on the scientific evidence, exposure analysis and risk assessment, I conclude that the
current 0.08/4 standard does not adequately protect asthmatic children from respiratory adversity
due to ozone exposure. In addition to exposures above the "level-of-concern" and adverse FEV
decrements predicted to occur when just-meeting the current standard, a number of independent
studies indicate a significant increase in emergency department visits and hospitalizations due to
the exacerbation of asthma (fig. 3-4, pg 3-53).
Inflammatory processes are an important marker of tissue remodeling and potential long-term
lung injury due to ozone inhalation. The use of 0.08 ppm as a "level-of-concern" in this chapter
is consistent with the ozone concentration at which inflammatory mediators have so-far been
observed in human studies. Clearly the selection of this level-of-concern has an important effect
on the number of individuals and the number of accidences per individual that is predicted by the
exposure-activity model, and therefore the rationale of using 0.08 ppm should be better
explained in this chapter.
The chapter summary states that the quantitative risk assessment is now appreciably lower than
in the last review (pg. 6-44, lines 13-16). This conclusion is based on an exposure-response
curve for FEVi that is my view has a great deal of uncertainty. Thus, this statement is
misleading and should be highly qualified, or entirely removed from the chapter.
Some other comments:
Page, line
6-4, 10 Delete first "human
6-7, 24 Delete "also"
6-13,35 It would be useful to give your rationale for choosing these subgroups of the
population for analysis.
6-22 In table 6-3 and tables that follow, it is not clear what incidence means in the
context of "incidence per 100,000" and "percent of total incidence."
6-27,10-15 What are the "less serious" lung function effects you referring
to? Why are they "less serious?"
6-40, 18-21 This sentence asserts that there is a 90-100% reduction in exposures-of-
concern estimated at the 0.07/4 standard. Are these percent reductions
obtained by rolling back the actual air quality for 2002 and/or 2004 to the
0.07/4 standard. If so, it appears from table 6-1 that similar reductions
are possible using the 0.08/4 standard. I find this hard to believe.
D-68
-------
Dr. Sverre Vedal
August 2006
Critique of 2nd Ozone Staff Paper draft (chapters 3-6)
Sverre Vedal
Chapter 3 (Health effects evidence)
While in general this chapter is a credible basis for the risk analysis that follows, there are
inconsistencies and some inaccuracies, some of which are due to the necessary reliance on a too
hastily prepared CD. There is appropriate use of cautions when the data are not as strong as they
might be, but this is inconsistent. A few points:
1. Measurement error discussion is convoluted and confusing, and contains some mistakes.
The primary issue in the use of central ambient monitors for ozone in time series
epidemiological studies is whether they provide any information at all in reflecting daily
personal ozone exposure in the susceptible populations, especially that of the debilitated elderly
in the case of the mortality findings. The evidence on this issue is split, but as opposed to what is
claimed in this chapter (p. 3-38, lines 26-29), there is information on this. The two Sarnat
papers, one from Baltimore and the other from Boston, assessed relationships between personal
and central ambient concentrations in elderly subjects with COPD.
The discussion on the impact of various types of exposure measurement error is incorrect (p.
3-37, lines 22-30) or confusing. The difference between true and measured ambient
concentrations is an example of classical measurement error and hence results in bias of effect
estimates to the null, not just an increase in standard error. Further, claiming that the difference
between average personal exposure and ambient concentrations results in "attenuation of risk"
(line 29) is confusing (also p. 3-38, line 2 and 8, and p. 3-39, line 4 and 18). This statement is
generally used when, due to classical exposure measurement error, the effect estimate is biased
downwards, and hence attenuated or underestimating true risk. The effect of ambient
concentrations themselves is underestimated. Here, as in the case of PM, the effect estimate is
estimating effects of changes in ambient concentrations. But, what is being referred to here is
the difference between ambient and personal concentrations, with personal concentrations often
being substantially lower. In this case, the effect of ambient concentrations is not being
attenuated, at least not in the same sense. They are only attenuated if one is trying to estimate
the effect of personal exposures for concentrations that are similar to ambient concentrations,
which is in fact not the case. This is not, therefore, a reason why we might be "underestimating
true public health risk" (p. 3-39, line 18).
The assessment of plausibility still does not adequately discuss the issue of ozone exposure in
the mortality time series studies (referred to on p. 3-35, but does not come up again). It is again
claimed that central monitoring concentrations "may serve as adequate surrogate measures for
mean personal exposures experienced by the population" (p. 3-38, lines 21-22). How can this be
if in some settings there is no relationship between individual personal concentrations and central
concentrations? If mean personal concentrations correlate with central concentrations, and
individual personal concentrations do not, then I presume some sort of ecologic error is at play
here. I dispute the conclusion that "the use of routinely monitored ambient ozone concentrations
as a surrogate for personal exposures is not generally expected to change the principal
D-69
-------
conclusions from ozone epidemiological studies" (p.3-39, line 10-11). On the contrary, we have
little insight as to what we would find had we actual exposure measurements.
The argument that central monitoring sites provide more valid measures of personal exposure
in respiratory admission time series studies, specifically, because most such admissions are due
to asthma admissions, most asthmatics are children and children spend more time outdoors, is
especially convoluted and unconvincing.
2. Cardiovascular mechanisms and effects.
The discussion on plausibility of potential mechanisms underlying cardiovascular effects
should begin with an acknowledgement that we don't know much about these. The current
presentation (p. 3-55) comes off as a hodgepodge of "the usual suspects," some of which are a
real stretch and serve largely to undermine credibility of the more realistic possibilities. The
discussion on the role of ozone in PM-related cardiovascular effects is hypothetical (note the
ultrafme discussion, in particular), yet are presented as though these interactions are actually
playing a role (p.3-59, lines 4-7).
3. Low concentration ozone effects on lung function.
First, Adams (2002, 2006) concludes that 0.04 and 0.06 ppm responses were no different
than filtered air responses. It is necessary to make it clear (especially in Ch. 6) that the estimated
effects in Fig. 3-1 (p. 3-7) are derived from Adams' data, but are not in fact reported in his paper.
It now becomes clear that the percent of the relevant population estimated to experience FEV1
declines at these low concentrations, estimates that are central to the risk analyses in chapters 5
and 6, are based on a total of the 30 subjects in Adams' papers (i.e., 3 subjects for 10%). This
area of low concentration responses is an area that clearly needs more research.
Because these low-level effects have a very prominent role in the risk analysis in chapters 5
and 6, the summary table (p. 3-78, table 3-4) should include them in some fashion.
4. Chronic effects on lung function: overstatements and inconsistencies.
The conclusion on p. 3-20 (lines 16 and 32) is correct: we have "little evidence" for these.
Why then do we get statements such as, "these long-term exposures may be related to changes in
lung function in children" (p. 3-56, line 30), "reduced lung function development in children
which have been observed in epidemiologic studies" (p.5-56, line 25), and "recent studies of
lung function changes in children living in cities with high ozone levels support the conclusion
that long-term ozone exposure may play a role in causing irreversible lung damage" (p. 3-84,
lines 6-8)?
5. Coherence of respiratory effects.
It is repeatedly claimed that the lack of consistently seen in the increased effect estimates for
COPD mortality may be due to inadequate power (e.g., p. 3-29 [Iinel6], p.3-52 [line 18]). Note
that this is in stark contrast to the case for PM where such relatively larger effects are almost
always seen; and the data used for the PM analyses have no better power than these for ozone.
Also note that the large meta-analysis (Bell 2005) found smaller effect estimates for respiratory
causes than for total mortality (p. 3-30). This has implications for the argument for plausibility
based on coherence of the respiratory findings. Parenthetically, why are not the studies used in
the Bell meta-analysis (2005) included in the respiratory mortality studies in Figure 3-4 (p.3-53)?
D-70
-------
Smaller or detailed points:
Page
3-7 (Figures 3-1 A-^C). These do not indicate "the portion of subjects tested having FEV1
decrements in excess of 10%," but rather give the distributions of % decrement. Also,
clearly the findings are somewhat dependent on the subjects included, since the distribution
of decrements in the McDonnell subjects for filtered air is about the same as that in the
Adams subjects for 0.04 ppm ozone.
3-24 (line 12) and 3-79 (linelS). Note that Gong did in fact find an effect on increased heart
rate, as stated correctly on p.
3-25 Correct the referencing in the last paragraph - the extension to NMMAPS analyses is only
in the Bell 2004 reference
3-26 Regarding adequate control for temperature, control for single-day effects or matching on a
day (as in the case-crossover approach, lines 28-31) is probably not an adequate assessment
of the adequacy of temperature control.
3-27 Effects using 1-hr and 8-hr averages in NMMAPS are not "distinctly" larger (line 13) than
those for 24-hr averages; "slightly" larger would be more descriptive.
3-32 The ozone effect in the ACS II study are not always nil (line 18). For one of the two ozone
monitoring periods, the effect was clearly positive.
3-35 (line 18) There is an interesting allusion to the utility of time series studies here which
doesn't seem to come up subsequently in the SP.
3-38 Reference to Brauer 2002 (linelO) is needed in the reference list, if it has not already been
added.
3-44 (line 6) Note that Bell (2006) found no difference in estimated effect even when all days
with 24-hr ozone concentrations < 20 ppb were excluded.
3-45 (lines 3-5) Note that the evidence of no clear threshold comes largely from epidemiological
studies, and partly from clinical studies, but not from toxicological studies.
3-56 (lines 26-27) Another possibility, of course, is that in fact there are no effects on humans of
long-term exposure.
3-61 (Iine25) These (5% decrease per 40 ppb increase) are not large decrements in lung function.
Chapter 4 (Human exposure)
1. It would be helpful to have the estimated exposures for current (2002 & 2004) levels
displayed in Tables 4-8 & 4-9 (p. 4-32) and Figures 4-4 -» 4-21 (pp. 4-33-^4-41), in addition to
only those for just meeting the current standard and alternative more stringent standards. This
would be analogous to the way estimated effects are displayed in Ch.5 (Figures 5-5 -> 5-9 [pp.5-
58 -» 5-65]).
Small points:
4-30 (line 8 and 16) Correct the figure reference numbers.
4-31 (figure 4-3) It is still not entirely clear why the estimated exposures are so low in Los
Angeles.
D-71
-------
Chapter 5 (Risk analysis)
1. Multi-city mortality estimates vs. single-city hospitalization estimates.
Note that estimated multi-city mortality effects are qualitatively different from
hospitalization effects, which are largely single city estimates (p. 5-28, table 5-1). Not only is
there the well known phenomenon of publication bias to contend with in using single-city
estimates, but, the Bell (2004) estimates for a single city are Bayesian shrunk estimates which
are not the same as crude estimates of single-city effects, even from a multi-city study.
Therefore, in using the mortality and hospitalization effect estimates in a parallel manner, it
should be made clear that these are in fact qualitatively (and therefore quantitatively) different.
2. Confusion on the effect of exposure measurement error.
Because ambient concentrations are two- to four-fold higher than personal exposures does
not mean risk estimates are attenuated (p. 5-36) (see my comments on this point for Ch. 3 [point
1]). Yes, potency is underestimated, but underestimation of risk is due to classical measurement
error, not this type of differential measurement error. Risk would only be underestimated if
exposure were to equivalent ambient concentrations of correct potency, which is not the case.
The confusion occurs in making use of the analogy with PM.
3. Striking findings.
It is striking how much all of the estimated health impacts (from lung function decline to
mortality) are due to exposures at concentrations below 80 ppb. Obviously this is due to the
predominance of days with these concentrations.
4. Relationship between 24-hr average concentrations and those from the monitor with highest
maximum 8-hr average (p.5-73 [lines 38-40] and Appendix 5A.2).
I'm not sure what point is being made here. While it is true that these are about half as high,
it is likely that 8-hr and 1-hr averages are not nearly so different from the 24-hr average. Of
more importance, the estimates of effect from the mortality time series studies using these
different ozone metrics are not much different, as demonstrated in the Bell study (2004).
Small points:
Page
5-32 (table5-3) Note that "respiratory" as a single category is defined differently from
"respiratory" in "cardio-respiratory"; also, this restricted definition of "cardiovascular"
(limited number of ICD-9 categories) probably explains the relatively smaller percentage of
total deaths due to cardiovascular causes than is typical.
Ch. 6 (Conclusions on primary ozone NAAQS)
This is generally a well-reasoned chapter in its attempt to ultimately motivate specific alternative
recommendations for revisions to the ozone NAAQS. It is appropriately cautious when
necessary. A few points merit comment.
1. Suggestion for more informative graphs.
In order to provide a more complete perspective to help judge the relative estimated
impacts of alternative standards, I would like to see figures 6-l->6-6 (pp. 6-34 -> 6-39) display
the estimated decrease in effects of the progressively more stringent standards relative to recent
D-72
-------
levels for 2002 and 2004, rather than only compared to just meeting the current standard (84/4).
Exposures for many in the population are estimated to be reduced dramatically from recent
exposure scenarios by just meeting the current standard.
2. Apparently puzzling (but probably correct) results of the risk-based approach.
Some of the results of the risk-based approach seem paradoxical and could benefit from
some further explanation. For example, in Table 6-2 (p. 6-18) it is strange that twice as many
individuals as are exposed to >80 ppb are estimated to experience a 10% fall in FEV1, when only
10% of population reacts this much at 60 ppb (vs. 25% at 80 ppb); therefore, most of these
estimated falls in FEV1 must be due to exposure to <80 ppb. But, this is not what is shown by
the "just meeting the standard" estimates, since these only result in decreasing the number
estimated to have these FEV1 falls in half. Also, in Table 6-1 (p. 6-15), it seems strange that the
% reduction in exposure due to more stringent standards is less in the worst case scenario (2002)
than in the cleaner year (2004).
3. The form of the standard.
While it is painful to seriously consider substantially different forms of the standard, they
might nevertheless be worthy of consideration. I have been impressed by the differences in
population ozone exposures across cities (more so than for PM, for example) that have either
always or sometimes been out of compliance with the ozone standard. For example, Denver has
infrequent and relatively brief summertime episodes when concentrations skirt the ozone
standard, sometimes resulting in Denver being out of compliance. In contrast, some cities
experience sustained and repeated periods of elevated ozone concentrations resulting in chronic
noncompliance. These contrasting patterns are not distinguished by the current form of the
standard, yet likely have substantially different implications for the health of the populations in
these cities. Why not at least consider some incremental form of the standard?
Another issue of relevance in considering a change in the form of the standard is the
anomalies in estimated impacts on population exposures produced by meeting the current form
of the standard (Table 4-9 [p. 4-32]). Specifically, there are marked disparities across cities, with
dramatic effects on reducing estimated exposures in Houston and Los Angeles compared to those
on other cities, simply due to the particular form of the standard.
On a small point, it is unclear to me how, in distinguishing it from the earlier form, the
current form of the standard also "is not just whether the concentration is above a specified
level" (p. 6-41).
4. Concentration-response and thresholds.
Evidence for thresholds based on controlled human experiments and epidemiological
studies is qualitatively and quantitatively different, as reflected in the different ways of using this
evidence in the risk-based approach.
At one extreme, it is argued (p.6-7 [line 18]) that it is difficult to find epidemiological
studies that demonstrate associations at levels below the current standard, implying a threshold at
levels that are only experienced in settings just in compliance with the current standard. On the
contrary, I think it is easy to find studies that provide evidence that effects are estimated to occur
in settings where the current standard would be met (most notably, see the Bell 2006 paper that is
cited). Further, it appears that an attempt is being made to argue that because studies are done in
areas that would likely not have been in compliance with the ozone standard (pp.6-9 -> 6-11),
D-73
-------
we can't know whether the findings of these studies would have been seen if the studies had
been done when the standard was met. In fact, it is almost certain that most of the data that is
responsible for estimating the effects seen is from days when ozone concentrations are far below
the standard. When analyses are done on subsets of data that seem to have clearly met the
standard, results are unchanged (again, see Bell 2006). Further, there are in fact studies that have
been carried out when the current standard was not exceeded: for example, the Vedal et al
(2003) mortality time series study from Vancouver, the Brauer et al studies of berry pickers
(1996 and 1999 [J Environ Med]), and some respiratory hospitalizations studies (alluded to in
Ch. 3-17 [lines 13-14]). And, likely there are years in some of the time series studies when the
standard is met, even though, using the full study period, the standard would not be met.
Therefore, there is ample evidence that associations in observational studies are present below
the current standard. The real issue is one of how we interpret these associations, not whether
they are present.
At the other extreme, it is stated (p.6-6 [line 31]) that the human experimental data
provide evidence for a threshold at "near the lower limit [my italics] of ambient ozone
concentrations in the US (CD, p. 8-44)." This is not correct. The statement needs to be dropped,
or at least clarified if something else is intended.
5. The level of the standard and the level of certainty.
An argument is made that we are less certain of the findings of effects at lower ozone
concentrations, and that this uncertainty should inform our choice of alternative standards. It is
clearly true for the human experimental findings that we are less certain of effects at lower
concentrations (e.g., 40 - 60 ppb vs. 80 ppb and above). This is not necessarily the case for the
epidemiological findings, at least based on the epidemiological studies alone. That is, we are not
less certain because the studies themselves are less certain. I find little difference in the certainty
expressed about findings in studies done in higher vs. lower ozone concentration settings. Our
level of certainty concerning epidemiological findings is based on information largely from
outside of these studies, and therefore takes on a Bayesian flavor. Specifically, there are many
reasons for being uncertain about epidemiological findings at low concentrations that are not
directly related to the findings of the studies themselves. Two of the most important are: 1)
there is justifiable concern that there is little relevance of central site ambient ozone
concentrations to individual exposures, with exposure concentrations being much lower and
poorly (or at least variably) correlated with monitored concentrations, a problem that is likely
more acute in low concentration settings, and 2) there is questionable plausibility to the
contention that the concentrations to which people are actually exposed in these lower
concentrations studies can be responsible for significant (or any, for that matter) adverse effects,
most notably death.
My recommendation for the level of the standard is in the range of 0.060 to 0.070 ppm.
This recommendation is informed by clear demonstration, both experimentally and
observationally, of effects below the level of the current standard. In 1997, we were aware of
effects in experimental settings at concentrations of 0.080, concentrations already below the
current level of 0.084. My lower limit is informed by my level of certainty in study findings,
both experimental and observational, of effects at lower concentrations. Within my preferred
range, I would select a standard closer to 0.070 than 0.060. The basis for this preference is
twofold: 1) the impracticability of imposing a standard where the vast majority of the country
would be out of compliance, and 2) the risk analyses demonstrating the relative impacts of
D-74
-------
meeting the current standard and of lower proposed standards, with the majority of the benefits
to public health achieved merely by meeting the current level of the standard. After all,
standards that result in achievable benefits to public health are much preferred to those that
might achieve these benefits only in an imaginary world.
6. Suggestions for uncertainties/research needs.
The overwhelming deficiency is the paucity of experimental human data on lung function
decline at concentrations below 80 ppb. Since these data are very influential in both the
exposure analysis and the risk analysis, increasing the number of subjects in such experiments,
and assessing the reproducibility of such responses in individuals, should be a priority.
In regard to point 3 (p. 6-46) of the uncertainties and recommendations for future research,
which touches on exposure, I would emphasize the need to have better exposure estimates for the
population potentially at risk of dying from ozone exposure. Currently, a case can be made that
exposures here are too low to plausibly be responsible for the dire effects estimated in the time
series studies mortality studies. Point 6 expands on this point well. On point 2,1 would like to
see more specific recommendations for figuring out how and whether exposure measurement
error is having an effect on the shape of the concentration-response relationship.
7. Minor points
page
6-13. What of exposure measurement error in influencing our level of certainty?
6-30. Reference to the Peters pilot study on very short term effects of ozone on MI gives it too
much play. Remember that the more prominent very short term effects of PM in the pilot
study were not seen in a larger and better designed study by these same investigators (HEI
report 2004), although they did not specifically report estimates of very short term effects of
ozone in this later study.
6-42. Figure references need correcting.
D-75
-------
Dr. James (Jim) Zidek
Revised Comments of the Staff Report 2nd Draft
James V Zidek, August, 2006
CHARGE QUESTIONS
My detailed comments respond to a number of the issues raised below and only summaries are
provided here.
O3 air 2):
1. To what are the air quality characterizations analyses clearly communicated,
appropriately characterized, relevant to the review of the primary and secondary O3
NAAQS?
The results are generally clear and very well communicated. In particular, the Report
provides a good characterization of the concentration field of the current criteria metrics.
2. Does the information in Chapter 2 provide a sufficient air quality-related basis for the
exposure, environmental effects, health risk environmental
in chapters?
I remained concerned about the level of level of uncertainty associated with the estimates of
the PRB field. A sensitivity analysis of the estimated levels would be desirable. In a
a of be a
for the It
••• vn- 'i' rH'.•;..am by by in
3):
2. What are the views of the Panel on the appropriateness of staff s discussion conclusions in
Chapter 3 on key to quantitative interpretation of toxicology
controlled-exposure studies and epidemiologic results, including, for
example, exposure error, the influence of alternative model specification, potential confounding
or modification by co-pollutants, lag structure?
My earlier concerns here have been adequately addressed.
4 O3
on
1. To what are the interpretation, presentation of the results of the
as in Chapter 4 (and in the draft Exposure Analysis technical
support document) technically sound, appropriately balanced, clearly communicated?
Generally the results are clearly communicated and well presented although the interpretation for
setting standards of some of the APEX outputs is difficult. A number of specific concerns appear
in the comments below.
D-76
-------
2. Are the to conduct the analysis technically sound? Does the Panel
have any on the methods used?
Based on my experience with a similar model, I believe APEX to be a sound methodology
although too little assessment of its predictive accuracy has been carried out, especially given its
central role in scenario analysis for standard setting.
3. To what are the uncertainties with the exposure analysis clearly and
appropriately characterized in Chapter 4, the Exposure Analysis technical document,
the uncertainty memorandum9
Overall, a very thorough well-balanced account of potential sensitivities and weaknesses of the
models used has been given. In so far as these can be checked they seem correct and
comprehensive. In some instances a quantification of uncertainty estimates is needed as indicated
in my comments. More discussion and interpretation of the outputs is needed.
4. To what is the plan for the remaining uncertainty technically sound? Are
there other important uncertainties which are not covered? What are the views of the Panel on
sensitivity conducted to evaluate the influence of uncertainties in the exposure analysis?
This sensitivity analysis has been well done within the limitations of available data and the
obvious sensitivity to AER has been discovered. However, the whole analysis of output
uncertainty could have been carried out within the framework of a fractional factorial design to
look for interaction effects between the different parameters. That might have been informative.
5 of the O3
1. To what are the assessment, interpretation, of the results of the
revised analysis as in Chapter 5 in the
technical document) technically sound, appropriately balanced, and clearly
communicated?
These are generally sound and very well communicated.
4. To what are the uncertainties with the health risk clearly
appropriately in the Chapter 5 the
Assessment technical support documents?
The uncertainties are very well characterized. A central problem here and elsewhere derives
from the problem of quantifying qualitative uncertainties such as model uncertainty.
D-77
-------
GENERAL COMMENTS
1 Overall the Second Draft Staff Report is well written.
2 Some early general discussion of the ideas underlying the APEX approach should be
included in the Staff document, given its central role in this exercise. After all, it is a big leap
from the regulation of ambient levels to personal exposures. That discussion should in particular
describe how later in Chapter 5 it is used in conjunction with an exposure response function and
used in scenario analysis.
3 Some concerns that need to be highlighted:
• The lack of assessment of APEX'S predictive accuracy. This seems surprising, given
that this methodology along with its predecessor, pNEM, have been under development for well
over a decade and a principal tool for such analysis.
• APEXs assumption that reducing ozone concentrations would not lead to behavioral
changes. The Langstaff memorandum raises the same concern (Langstaff page 56).
• Uncertainty of accuracy of APEX predictions. Providing estimates of the uncertainty
for things such as person-days of overexposure should be straightforward as APEX is a
predictive probability distribution and "error bars" could have been generated and included in
figures such as 5-11. Note: page 6-16, line 31 promises such estimates in the Final draft.
4. The quadratic rollback procedure is not very well described in the AQCD, the Staff
Reports or the supplementary material. Considering its vital role in the scenario analyses and
exposure response modeling, I suggest expanding lines 22-28, page 4-28 to include the formula,
how it is used and the material on page 5-10. As now written, the reader learns for the first time
in this document that it shrinks the actual ambient hourly concentration measurements into
attainment.
5 While the discussion of APEX uncertainties is remarkably thorough and comprehensive,
they can only be partially quantified leaving concerns about the accuracy of the eventual
predictions of such things as person - hours of exposure above a prescribed threshold. However,
I see no practical solution with present knowledge. In future much more model assessment
should be done to get a better handle on its accuracy.
D-78
-------
DETAILED COMMENTS
Chapter 2:
2-2 line 16: The sentence beginning on this line should be the last sentence in its paragraph
following the sentence defining NOy.
Appendix 2-A: July (month 7) and August seem to have the lowest PRB and
Sacramento has a large month to month variation. Some rationale should be included for what
some reviewers of this document might find curious findings.
CHAPTER 3
3-36 Line 9: Is 40ppb the "standardized unit" of change referred to in line 10? If so that same
unit should be used again because the sentence is ambiguous as it stands.
3-37 Line 3: Should not this be "exposures to ambient concentrations"?
Line 18: This goes to a comment below and the 3rd concern above about APEX which assumes
behavior would not change if ambient ozone concentrations were lowered.
3-39 Line 8: Some emphasis should be placed in future epidemiological studies in using a
model like APEX to reduce errors in estimated population level exposures to the criteria
pollutants, given the critical role APEX is playing in setting standards and given that it is freely
available.
Line 19:1 remain concerned about the observation reported in this line and ask that the rationale
for it be included in the Report.
CHAPTER 4
4-1 Line 21: The report should say something like "based on 2002 and 2004, for reasons
given below in Section 4.3.4.1,...". However, two seems like too few years to really characterize
the "year-to-year variation and more importantly bias, since 2004 had lower than typical O3
concentration levels.
Line 24: No doubt lowering ozone levels would lead people, especially susceptibles, to spend
more time out of doors and thereby to sustain higher rather than lower exposures to ozone, a
point not discussed in limitations of the APEX approach. (Runs condition on outdoor
temperatures, not ozone levels, thereby missing this important interaction.)
4-2 Line 28: Something should be said here about "dose" (or a reference given). The text
leaves the impression without explicitly saying so, that dose is important. Exposure (intake) is
then estimated because it is an "important determinant" of dose. Why not estimate dose itself? If
dose cannot be estimated, does the science at least say qualitative about the relationship between
the surrogate (intake) and the thing it represents (dose) other than that the relationship is
monotonic? For example is the relationship linear? That is would a 10% reduction in exposure
lead to a 10% reduction in dose regardless of the level of that exposure?
D-79
-------
4-3 Line 8: To avoid ambiguity, AQDC would be better than CD here and elsewhere in the
report.
4-4 Line 17: Generally monitors are sited primarily to detect non-compliance and therefore,
tend to overestimate ozone concentrations. Nevertheless the data they produce are used for other
purposes as noted here in the report. However, it seems doubtful that they could be used to
determine their own "representativeness" without the help of Line 27. Therefore these sentences
should be linked in the text for clarity.
Line 20: For consistency the same units of measurement should be used throughout the report.
Preferably, this should be ppb since numbers like 0.085 ppm for example, seem so small that
seem unimportant.
4-6 As noted in my earlier comments, APEX is a valuable tool for exposure assessment as
well as predicting the benefits of a pollution abatement program. In particular, it can answer
"what if questions under roll back scenarios. However, APEX is only a model, an analogue of
the real world therefore its outputs need to be assessed. I was pleased to see on page 4-13 Line
23 some recognition of this fact and a plan to do something about it in the future and at least a
crude assessment on page 4-47.
4-11 Line 12: The sentence added here in the revision seems to be a partial response to my
comment on the first draft:
"randomness of model parameters needs some explanation in the Report. Their distributions
could be interpreted by readers as Bayesian priors, reflecting (epistemic) model uncertainty,
albeit with an empirical Bayes twist. In that interpretation, variability of the predictive output
distribution would reflect that epistemic uncertainty as it propagates through the model, rather
than only process (aleatory) uncertainty. It might be worth stating that these parameters are
actually random effects so that variability in the predictive distribution represents aleatory
uncertainty. Of course, that then pushes the issue of model uncertainty to a higher level on that
staircase of infinite regress, that Mosteller and Tukey famously referred to. (See my comments
below re Table 5.5.)"
Consider for example, the model in Equation (2-15) of "Exposure Analysis" technical
attachment. The stdev's for it in Table 2-4 do reflect "uncertainties in the parameter estimates".
In fact they are the standard errors of the estimates computed in a regression analysis of data
from 32 clinical trials contrary to the assertion of that sentence. Moreover the issue of model
uncertainty has been ignored even though it could well be the dominant source. In that model, for
example, it seems quite plausible that a gender*age interaction term should be added. (In other
words the coefficient b2 should be different from women than for men.)
4-12 Line 7: My earlier suggestion to interpret model outputs in terms of predictive
probabilities and expected values has not been implemented. However that would make them
easier to understand and make inter-city comparisons straightforward. Moreover, this approach
would be more natural since APEX is a joint predictive probability distribution expressed as the
product of various conditional distributions. Finally the use of probabilities and expectations in
quantitative risk analysis is standard.
In any case the value of the city specific number of "person hours" in standards setting needs to
D-80
-------
be clarified and their standard errors stated as they may not be negligible even if the number of
APEX runs is large. (See comments on page 53 below.)
4-14 Line 9: No mention is made of the possible bias that might arise in the
characterization of urban ambient ozone fields due to monitor placement for non-
compliance detection. How were the monitors placed in the selected cities?
Line 21. Using the multivariate spatial predictive distributions now available in modern space
time process modeling, the variability of the random ozone field over the USA for a single or
multiple years could well be captured and used in conjunction with APEX. (See Le and Zidek
2006. "Statistical analysis of environmental space-time processes." Springer.) This would
overcome the difficulty associated with relying only on specified years.
That would not address the other kinds of uncertainties there regarding say changing
demographic patterns and hence time activity patterns. This difficulty might be addressed
partially by running APEX conditional on the demographic group as is done in pCNEM (referred
to in my earlier comments). The predictive population estimates at that level could then be
combined with different weights to determine the possible effects of demographic trends. Users
of APEX could then predict exposures for population subgroups such as commuters or
individuals with heart disease, by representing them probabilistically as a combination of
demographic groups and combining the predictive estimates appropriately.
4-15 Line 6: The meaning of the sentence beginning on this line should be clarified. To many,
"uncertainty" means "variability". Only things that do not vary are certain.
4-18 Line 16: As noted above, two is too few years to characterize variability due to year.
4-21 Line 30: Using LM=1 and LA =0 would seem more realistic than setting both equal to 0
as has been done. Anyway, rerunning with these alternatives would give some idea of the
sensitivity of results to these choices.
4-29 Line 4. The caption could be misleading since "design value" commonly refers to a
specified cut-off criterion. "The measured values of the current design metric for the modeled
areas" would avoid the interpretation that different criteria are imposed on the different cities. By
the way, I am not sure why the ratio in the last column was included. Would not the difference
be preferable, if anything was really needed there? (By the way, notice how much simpler this
table would be if ppb were used instead of ppm.)
Line 11: The graphs do not show the numbers but rather the percentages right in line with my
suggestion above for page 4-12. These can be interpreted as the chances of the specified event
occurring for a randomly selected individual from the specified sup-population.
4-30 Line 5: The figure captions are really complicated. How about inserting an example in
the text to help the reader interpret them: "For example, Figure 4-2, shows a 42% chance that a
randomly selected asthmatic child in Cleveland experienced repeated 8 hour exposures
D-81
-------
exceeding 70ppb (Scenario 74/4) during 2002-2004 while undergoing moderate or greater levels
of exertion. In contrast...."
4-44 Line 20: The sensitivity observed here poses a dilemma for developing an optimum
National air quality standard since any one standard would result in varying exposures across the
Country due to varying human behavior patterns. How should a National standard compromise
between competing local optimum standards?
4-45 Line 14: the conclusion of "moderately sensitive" may need to be more clearly
explained, given the -57% change we see in Table 4-15. The fundamental issue here is the
reliability of the stochastic decay rate model. If it is deemed to be valid, choosing the 10th and
90th percentiles (the worst case scenarios) in place of the randomly distributed decay rate seems
unduly severe. More natural: shift the center of the decay rate distribution to the 25th and 75th
percentiles, thus preserving the uncertainty being expressed by that distribution while making it
stochastically larger and smaller, respectively.
4-47 Line 8: Tailoring the model to fit the California situation assesses whether the model is
ideally capable of accurately forecasting exposure. However, APEX as used to set National
standards is a different model and it too should be assessed. Maybe it's AERs are higher than
those in Sacramento and it does not underestimate exposure like its tailored counterpart.
Line 18: The shifty unit again, this time going from the ng/m3 to ppm.
ATTACHMENT: "Ozone population exposure analysis for selected urban areas".
10 Available predictive distribution methodologies (and software) would enable ambient
levels to be randomly generated down to the tract level and thereby reflect the uncertainty in
some urban areas about the ozone concentrations in the outside air being exchanged by the
APEX model. Staff should consider this refinement. The result would be in line with the way
APEX has been developed.
12 The claim about the stability of A "over cohorts" seems surprising if this means
"between" rather than "within cohorts". For example, one would have thought the
autocorrelation for institutionalized seniors would be far higher than for young, working persons.
Page C-S12 states that auto-correlations within cohorts "may differ greatly" among its members.
Some clarification is needed here. (I note that the model outputs are insensitive to "small
changes" whatever that means exactly.)
18 This page suggests, seemingly in contraction to page 4-2, that dose is being simulated in
addition to inhalation. Some reconciliation of language would seem desirable.
21 Clock hour "I" rather than "I" seems correct.
22 Can some estimate be made of the effect of using the school children's D and A for all
individuals? In particular is this path more liberal or conservative than that of drawing the
sequence of daily time activity patterns at random?
D-82
-------
23 See the comment about page 10 above re the imputation of local concentrations from
those at the monitoring sites.
40 There does not seem to be a Section 3.8.2 so it was not clear what fractions were actually
taken from Appendix A and used.
46 Some discussion of the results in this section would have been helpful. How are the plots
shown here to be interpreted and used, especially those about "person - days"? The same can be
said about the next section, although there at least some discussion of the findings is given.
53 It is not clear how to interpret the person-day estimates. More discussion is needed about
how exactly these person-day measures of population impact are to be interpreted for setting
standards.
No explicit statement is given about how the aggregates were computed. In the case of person-
days, was this done by: first calculating the expected number of days of exposure per person
based on a number of APEX runs; and then multiplying by sub-population size? If so the
standard error of these estimates could be quite large even if a large number of APEX runs were
made. Some statement of the reliability of these estimates needs to be given. A very relevant
analysis of this reliability is given in the Langstaff memorandum.
73 While this assessment is valuable and encouraging, nevertheless the suggestion that
APEX underestimates true exposure is of concern. So is the use of weekly the aggregation of
exposures. These could give quite a misleading impression of hourly exposures because of the
ecologic effect.
84 The technical appendices that follow on from this page seem very comprehensive and
thorough. Where I was able to check the details, I found them correct. Moreover, relative to the
complexity of their content, they seemed commendably clear.
A-l No information is given about the design of the survey used to select the 37 residences in
the (smallest) RTF study. Nothing is said about the weights if any used in the estimates to adjust
for unequal selection probabilities. Could the eventual estimates of population parameters be
biased?
E-6 I would have thought the regression model on this page should have been fitted after
logarithm transformation and the resulting model on this page expressed at a product. That
would have had the advantage of insuring amongst other things that the predicted AER is
positive, whereas it could in principle be negative as presented.
D-83
-------
DRAFT MEMORANDUM: "Analysis of uncertainty in ozone population exposure
modeling" by John Langstaff
This memorandum is a generally clear and well-written account of some very technical
issues. Some specific comments for the specified pages follow:
2 "Variability" also leads to "uncertainty" so nowadays these two are commonly referred to
as "aleatory uncertainty" and "epistemic uncertainty". The first represents the uncertainty that
derives from the various random sampling processes built into the APEX simulator. The second,
the more subtle of the two, represents lack of knowledge about the underlying stochastic models.
In APEX this means uncertainty about the stochastic model parameters such as the one reflecting
the variation in AERs. However, it should also reflect uncertainty about that model, say whether
it is lognormal or any of the logstudents-t distributions with their varying degrees of freedom.
Why include model uncertainty? The answer: to buy insurance against model mis-specification
that can lead to APEX outputs that are biased away from the actual exposures. Better to have a
heavier tailed distribution that covers the exposures the model aspires to predict than an
optimistically narrow one that misses the boat.
Incidentally, a proper assessment would determine the calibration of the APEX prediction
intervals. Too much uncertainty would mean these say 95% intervals would be too broad
relative to the observations and hence cover more than 95% of the observed exposures. That
would point to the need for additional knowledge and hence reduction in epistemic uncertainty.
Pages 4 and 8 provide good discussion on this general issue about uncertainty.
3 2nd last paragraph: Shouldn't this be "between input uncertainty" rather than "between
model input uncertainty"?
8 Log-normal distribution. Analysts usually log - transform data that are lognormally
distributed, perform inference on that scale and then transform back. This leads immediately to
confidence intervals for the GM and GSD since these are by definition, monotonic
transformations or their transformed counterparts: GM = exp (AM) and GSD = exp (SD). This
fact would lead to asymmetric confidence intervals for both. For the geometric mean for
example, the 95% confidence limits would be (GM/GSD2' GM*GSD2).
Since the normal theory estimates, AM and SD, are stochastically independent, I am puzzled by
Figures 17 and 18 that show them to be correlated. Is a different definition of GSD being used?
(The mean and standard deviation of the lognormal distribution are related in a mathematically
explicit way.) If so that definition should be stated for clarity. Or could this mean the
lognormality assumption is inappropriate?
On this page and elsewhere bias in the ambient monitoring measurements in urban areas has
been ignored.
D-84
-------
9 Indeed spatial variation in ozone concentrations can be considerable. For that reason,
SHEDS has been applied with the pollution field interpolated down to the tract level. Obviously
that would be impractical for APEX however.
10 As noted above, the quadratic roll back process is not very well described anywhere in
the documentation including the Staff paper.
13 First line: Typo: "at a given monitor as uncorrelated".
The imputations of missing values have quantifiable errors that could be quantified and reflected
in the uncertainty distribution for APEX outputs. However, the method proposed as an
alternative seems satisfactory.
Incidentally, that method is the "cross-validation" approach rather than the "jack-knife"
approach. The latter and its competitor, the "boot-strap" are used to correct bias in estimators and
estimate their standard errors. That term should be changed on this and subsequent pages.
14 Would not including all the monitors lead to an underestimation not overestimation of the
prediction errors? As noted in the last line some of these will be zero.
15 Why is the ratio rather than difference used as a measure of prediction accuracy. The
former unlike that latter would have a very non-normal distribution.
The cross-validatory assessment should have been made by removing and not using the target
site in constructing the spatial interpolator to get a better indicator of prediction error.
The basis for the assumption made at the end of the first paragraph on this page seems tenuous.
Further analysis would eliminate the need for it.
A well-tested method for spatial prediction has been developed for space time fields and
software is freely available online (http://enviro.stat.ubc.ca also includes a demo). The method
overcomes the deficiency in geostatistical methods that ignore the stream of temporal data
available for estimating the various parameters required for the interpolator. It also avoids the
assumption of covariance stationarity made in kriging.
Incidentally, kriging, being a linear method, de facto assumes the underlying field is Gaussian.
Its performance for ozone can therefore be improved by taking a square root transformation of
the field and transforming back after completing the interpolation.
21 APEX averages the working and home district ambient levels only in as much the
microenvironments through which the sampled composite commuters go from home to work to
home districts during weekdays. The first paragraph in the subsection beginning on this page
gives a misleading impression that the two ozone fields are directly averaged. Furthermore the
claim about ambient levels used for commuters who work outside the study area disagrees with
what is claimed on page 4-21 of the Staff Report. Reconciliation is needed here.
22 The section beginning on this page does not make clear what uncertainties associated
D-85
-------
with roll - back are being assessed. Presumably the reductions called by the quadratic formula
are not uncertain. Is it the quadratic method? Is the uncertainty about how well this approach
actual replicates what would happen under a real rollback?
First sentence, 2nd paragraph of section. Shouldn't "reflect" be changed to "attain"? In this
same sentence, two sentences address how uncertainty is to be assessed by the proposed method.
Is one redundant? Should they be merged? In any case, I could not understand why the first one
is true. Why do the differences between them reflect the uncertainty that is being considered in
this section?
3rd paragraph of section. If I understand its intended meaning correctly, the sentence beginning
"A 3-year.." should be rewritten as: "A 3-year period with wide range ozone concentrations will
typically have higher population exposures than one with a narrow range even though both
periods have the same design value (or both are just in attainment)."
34 Why are these values constrained to lie between 0.95 and 8.05? Why is that even
mentioned at this point since the actual estimates are well within that range?
CHAPTER 5
5-18 Line 20 - : This imaginative but somewhat ad hoc approach is not quite Bayesian
because of the way in which synthetic data are created. A likelihood based alternative:
D-86
-------
Qtirtiililying Response Probability
Jim Zidi.'k
August 2U. 2«W(»
1 Credibility Bands
uppow-- ,1 i»ri'/'r d I>IT ibi it i ••.']] lor H r.Lui ba> ,j prior »1<%ii?itv ~!«t = -|io)T>) ")";>' ! wb«%i\-
. -> /, ;Vi^'| ,.TS) (1 I If li,«!K11) - .,' i - I. X („- M. rT ^ i HFlil '. ^ /. A"i^;i - fT:i). \\''. <-i\\t \-,<- >/-t i-'',|n«l I--,-1 h'/ir iiiaxini'ini likH\\v f.nl i--tiinrti'v.;- fi> j^t mi
i-Hipirk-nl Bnye> prufi'iimv-) T(u-i> ll»- pu.iivritn>iiy Imiiliui; •..•] 0 i>
Wv- '..Mil ~>impk- itt:-ii>^ iVoiii tin' pu^li.-rior ii?i!i^ ;ii! MC.\K.' .ip]>r'..'.ifii ilia I loads In .1 dmiu
0\,0±,... whi;-li nfli-T a "biirii-itf" p«->i'ii>'d, -ay a I h>.-ii>an cosidiUoual on
il> M>()iti'l- Tlm> givni 0., nrnw ^.. , hi -I'-'p^ by (k;>.«i)ii; 'i:,|, ')',.. i- «nn ';.i f.'Sir- a\ » i.iir>r-
in order-
D-87
-------
Ideally 7( slsniiM lii' i hows! u< Timk'.' Ui>' sv-u-pliiin i- ml'1 jn-l under "il.'H'. Draw li _ j in >i
i-iiirilar la.diioii nil IT r««]>inrit!,n «>. by ;. |,
As i !M' M-i)ueii<:«' >ih...'V>:- i> Ivin^, griii-rnir-iL ilii- »>-<|i.ifiiri' {IK > ;fVj iti>iy In- H|«I. h'- jj,i.-ritTnu-d
for ati.v tixi-d r:X)W)>i.ir<% !>.HV| ->iy .1 ii."-'j" afirr the burn-m pr-riod. .\lukiirj; a liHup.rrtiii
i;r i-(|iiiv:iK'in )»\iliiy rli-t^iiy i'»tirnat>- IVuin i ht> hiitn- MX|Ui-ii'-r ivill n-vcnl ! h-.-Nh»|><* uf
thai ik-ijMiy. ,-\i 11)'.- ,-tiiiK' liriK- Isnditijr. n.- ~>:" "tid n.V- ix'tvrislilt-r- will yii-ld iiii i->titn>ili'd
'.'~,<:\ iT('!>i•.> |tv«!>;it>ili(y.
ii!'!>»-;ii in^thi- S>F'--'< i-durt' ninivi' lur ^ay 20 (lil'iTt-tit '.'xixj^iirc li-vi-h:, j mid sniuDlltiy juhihij;
i h(.] upprr fiiriil^ riiid I hs-'E 1 \\i- luwt^r hsini- ul 1 hx~- rrrdi3>ihty lirii]d> will li^hd N* n xTrur II;IIKJ>
rtl'ijuiii! thi- <'-ti:ii«t-->d ri'>|,H«]i>i-' fiirvc.
2 Model Uncertainty
I." > un 1 ly n ii.nk'l •; >.-p bt i.-m !•.-) m ieert a: nS y will be * ul>t n in i« I at id ill'.' •.•fxl ii.>i I it y ba i id <.\'i i )-
1,'Utisl abnvc. l>>\"<'d >,-ti ii --iti^li' Hymned mid Uxifi iiindi'b will In- tni, luirruw. In I.K-I.. i.iji'
inif iiiud'.'l i- ••-•'..'lisph-'tt'ly unknown MI\V jK'i'ii.ip^ lur ih'.1 hi'i liuu it ituir-i bt- iiK-ihiU.-ih'
iiiiT!-a>iu<;. M«H!-»V<-I-. i IK- very limiti-d iiiuonnl ul dntn imiilabir ai the- hit iiaiitlt-iul ul tin-
T>v.iff- til r-s|x^!trr Ic-vc-h i:- !•)(> • iH.-tiil siutdi-- r.> a prtratrii--"rii: tortn. lliiy
'it \v«ul>{ b».' !>i\-li-rab)i' i>,n-du tic- >ihu>,v >j»a!y.-j.- i'--,r a ramu- >4'j)lai^ibli' iiiudi-k, iiiclndiiv.;
t he k.•.!••; i>t »• and liiii-xr on i--. The lai!..t v,\.4ild b«.- ^iven by
D-88
-------
:\c'3i!i k..vii'. Ti!s;il di.^tt ib»:l iuii^ .di'.-uld U- iiM-.l 1---T i In- <<> mid ssonn.iii- Im I In- fi> nidi-. .'Midi
wild '> |):iraiii<'t'M> MIH rno-k-! will he I he mo-.) > litVienil i« fit with iii;>.\iimim lii>eiihu'..d. In
!!li'!'<-elrtbur.tte n»al>>i> i: WHild I-...- pi-nail/'." I k laek -..'i |Kir>J)!i'->iiy bill (itnl F-.'fil !'-'!«.,••! it
liti^JM be Irlt I'.'rliu' hlHiri- 311 view si) IIHII- C':iii>trs>iTil > UTI I lie pr?iee.-*^-
ilcAV '_',)!) ')•'• Ull'x-rlrtitltV al'uUl tl(!->'-" ItUl'-t .>til<1V; iilu'M-t I.-..' o Otlbi!(«.!'.' O»i- n>;n'h ivnuM !>»> to \\i'i^.dl t hem iX-rnrdiu;,; tu l>i»ii)^.ir';\i !,'lai:>ib:lily (which maki'* i.ly
lisuw !si-..-;k«l i|iiili- unn'nii>ii<- nl she ri,-,h! Itatid end a! iea-i; ~>\v
Riiuniii^ Use MCM(." }iiial\>i.-s lur i hv pu^irriur ili»lrib>ili'.-']i? t»?i.:<->,jii:tl wild i/aeli ul the
isiuiii-l-i will yield sixjiiciio1- siiy >•! !'j,uu« iieratii>ti> iifti-r Uivti-in. j ii-,(r\(i-\}. HJ(.I: ft!,...}]
i-:r.;»keiisi irk. "I"do IOjH.«i Viihir-. k1''-s c-an
thi-tl hi- r-f-.tnbmed \\irli I lie Irtst .'i.niii'i rtlld -l.'Kii'l \Vililef- f.;-r ":,( llse liljertr (Dl'1 liri.kelj
^tk-k uu^lt-l-. <•>.>.-|>!-r-(.ii'el_v, to yfl n >itis)'.' e..,!iil.4tied -ot •>!" valui-4. lir- .irit hinot i-v iri"ari ;4iV',v
an e-.thtinle c>| ihi- '-i.'ijibmed re-|H>t!sr- |)r«.'l>ubilrt)' while \\* .""• iind ''*'" |n'rr("iii.ilr«- w«..u]d
yii-lcl H 'ii)!'; erfdibiiily luind fi»r tlmt miknuwH r',~p..'U I dlf b:llici \V,»!lld lie ivid>'! I ll:ill l.dfolir
is" if the !"}.',i^ti''- tliudel niiitle. i't-lii-r-i jjij^ r pi virus :f \tnrniartitit. It.- \i\\-ff :-\y\- \v> in She- liMik<-n >t!«-k a-i lunnulaied \vuuld ssirtke i! ditlic nil if Iil wii Is i.he Tvhiively Miinll
rtui...'um of claui. enj|-ibusizinc: anew the uiiceviainiy c%t tl)<.- l-;-l! haiitl VIK! of the <.-i.ivvv. Slse
:il-i'i lli:idv NligO-'^t intl« I d.tl ilnpt>-ii.-'d ill'-' I-'A|'>: >.-.il if ,|l ill llli> i f-'e| HI jfitl «i-'et ie-tl i i[ it >1I! 11 l'Jtlt*.
D-89
-------
5-21 Line 28. A very good description of variability or what is commonly referred to as
aleatory uncertainty.
5-22 Line 20: Word missing here.
CHAPTER 6
6-13 Line 6: A potentially important bias could arise from the changes in the time-activity
patterns of members of the susceptible groups, such as those used to characterize the exposure -
response function and as noted above, this source should be included in the list of uncertainties.
6-34 Figure 6-1: In this and the falling figures we see a surprising amount of variability
between cities for 74/5 compared with 74/4 and that is less than 74/3. That seems curious. In any
case, it would argue for 74/4 on the basis of robust of the predicted benefits across cities.
6-47 Line 1: Environmental epidemiologist could well use APEX - type models in their
assessments of association of health risks and pollution. Such models in conjunction with
interpolation of ambient fields to local areas could help overcome the measurement error effect
in the use of ambient concentration fields as surrogates for exposure. It would also provide an
estimate of the exposure response as against the concentration response. Moreover, based on my
research work with an APEX relative (pCNEM) non-significant associations can be turned into
significant associations through the de-attentuation of the relative risk.
D-90
-------
Dr. Barbara Zielinska
Comments of the 2nd Draft Ozone Staff Paper
Chapter 2
Barbara Zielinska
DRI
In overall, this second draft of the ozone Staff Paper represents significant improvement over the
1st version. Below are my responses to the charge questions to the CASAC 63 panel from Dr.
Martin's memo (July 17, 2006), regarding Chapter 2.
1. To what extend are the air quality characterizations and analyses clearly communicated,
appropriately characterized and relevant to the review of the primary and secondary Os
NAAQS?
In my opinion, Chapter 2 is well written and presents a concise summary of the information
contained in Chapter 2 and 3 of the final 63 AQCD. The ambient ozone levels, 63 precursors,
their sources and emissions, temporal and spatial variability, long-term trends, and
characterization of ozone episodes are adequately summarized in this chapter. This information
is relevant to the review of primary and secondary O3 standards.
2. Does the information in Chapter 2 provide a sufficient air quality-related basis for the
exposure, human health and environmental effects, health risk assessment, and
environmental assessment presented in later chapters?
Section 2.2.6 covers briefly the relationship of ozone to other photochemical oxidants. However,
there is no information in this chapter regarding the role of ozone and other photochemical
oxidants in the atmospheric transformation processes that may results in the formation of more
toxic products (both in outdoor and indoor environments), either in gas or particle phases. This
topic, covered in Os AQCD, seems to be relevant to human health and environmental effects.
The policy relevant background (PRB) section (2.7) continues to have problems. Although this
section very briefly cites the results of the comparison of different models and measurements, it
does not adequately address the uncertainties of the global GEOS-CHEM model and how these
uncertainties are reflected in the health risk analysis. The authors say (page 2-48, line 1-4) that
the GEOS-CHEM model overestimates PRB ozone values for afternoon surface air over
southeast in summer and it is accurate within 5 ppbv in other regions and seasons. If the PRB
concentrations are predicted to be generally in 15 to 35 ppb range, 10 ppb overestimation is
substantial. Besides, it is not clear how this information was derived. Since ozone health effects
are observed down to concentrations of the order of 0.04 - 0.05 ppm, it is important to know
how the PRB compares to the considered primary ozone standard and what uncertainties there
are in the risk attributed to controllable sources.
D-91
-------
Minor comments:
1. Page 2-2, lines 16-21. The order of sentences should be changed. The sentence "NOx is
considered a good surrogate for NOy" should be the last sentence, after NOy is defined.
2. Table 2-1. "Total without fire" for 2003 and 2004 should be moved up one line.
3. Page 2-5, line 30. Formaldehyde should be spelled out (H2CO may not be clear to
everybody)
4. Page 2-15, lines 27-29. Transport is not evident from Figure 2-6
5. Page 2-48, line 8: The text comments very briefly that the PRB O3 concentrations are the
highest during spring and decline into summer. However, as evident from Appendix 2A,
diurnal ozone PRB profiles are very different for some city, for example for Houston or
Sacramento. Any explanation?
D-92
-------
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:
------- |