UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
June 20, 2007
EPA-SAB-07-009
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Subject: Consultation on EPA's Risk and Technology Review (RTR) Assessment Plan
Dear Administrator Johnson:
The EPA's Office of Air Quality Planning and Standards (OAQPS) requested that the
Science Advisory Board (SAB) conduct a consultation to provide input on whether its proposed
assessment plan is adequate to furnish the basis for regulatory decisions concerning specific
source categories. A consultation is conducted under the normal requirements of the Federal
Advisory Committee Act (FACA), as amended (5 U.S.C., App.), which include advance notice
of the public meeting in the Federal Register. Although there will be no consensus report from
the SAB as a result of this consultation, the panel would nonetheless like to underscore several
key points that arose in the conduct of its consultation on the proposed Agency plan.
On December 7, 2006, the panel met via telephone conference where representatives of
the OAQPS offered informative presentations to the members of the SAB Risk and Technology
Review Consultative Panel. The focus of the presentations by EPA representatives for this
consultation was on the emission data, dispersion and exposure modeling, dose-response
assessment and risk characterization to be utilized in the proposed Agency plan. On December
19, 2006, the panel met again via telephone conference to discuss and deliberate on the charge
questions. A copy of the overview of the plan and the charge questions to the panel are attached
to this letter (Attachment 1). In brief, the SAB was asked to comment on the appropriateness
and adequacy of using a new approach to perform an assessment with the goal of characterizing
the exposures and risks associated with the emissions of hazardous air pollutants (HAPs) from 51
different industrial source categories. Feedback on the charge questions was provided by panel
members and a compilation of their comments and recommendations is attached (Attachment 2)
to this letter and appended to the minutes for this meeting.
On behalf of the panel members, we would like to express our sincerest gratitude to the
presenters for their expertise, perspectives and insights. Their contributions greatly increased our
-------
understanding of the Agency's current policies, methods, practices and future directions proposed
for residual risk assessment. The panel members had several concerns about the RTRII process that
are expressed in their individual comments. Highlighted in this letter are several key messages that
emerged among the panel members as a result of the Agency presentations and discussions:
• There is a concern that the expedited review (RTR II) is focused on the most feasible
sources to review, i.e., the easiest assessments to make, rather than on the sources with the
greatest potential residual risk;
• The plan needs to be revised to make it clearer, and the process more transparent, by
addition of flow charts that indicate the differences between the various RTR processes
(RTR I, RTR II, RTR III, etc.); The current version does not clearly explain the basis for
greater efficiency in the RTR II over RTR I;
• The plan should incorporate a framework for improving the National Emissions Inventory
(NET) as new/more accurate data become available. There is a concern that the use of the
voluntary NEI data base for a regulatory purpose could induce changes in reporting that
modify the data base;
• The uncertainty and/or bias of the model estimates (HEM-AERMOD) should be addressed;
models should be improved to better reflect the atmospheric transformations of emissions
that affect potential health/environmental impacts; and field measurements should be
compared to modeled estimates to help determine how well models predict actual
conditions for HAPs;
• If feasible, a probabilistic analysis of the exposure scenario and other factors should be
developed in order to: (1) provide more transparency regarding variability and uncertainty in
the exposure and risk estimates, and (2) allow a more informed judgment regarding the extent
of conservatism included in the calculation/model;
• A transparent decision framework should be developed that: (1) identifies HAPs where
ecological risks rather than human health endpoints are the basis for setting air emission
limits, and (2) addresses the importance of facility emissions relative to background
sources for naturally occurring HAPs;
• The plan should clearly and repeatedly state in any communication to the public that the
assessment should be used only for the purpose intended by the Agency and that the
estimates should not be construed as absolute estimates of residual risk for use in
population-based studies;
• The existing toxicity databases should be evaluated to determine if they are complete
enough to be used in the RTR process;
• A sensitivity analysis should be conducted to determine: (1) which inputs are the main
drivers of the Residual Risk estimates, and (2) if differences in the levels of uncertainty for
those inputs (for example, uncertainties in emissions from some sources compared to
others) may potentially result in misclassification.
-------
Finally, the SAB commends the Agency on seeking early advice to enhance their Risk
and Technology Review Assessment Plan. We look forward to working with the Agency as they
implement this plan and refine their assessments.
Sincerely,
/Signed/ /Signed/
Rogene Henderson, PhD Granger Morgan, PhD
Chair, Risk and Technology Chair, Science Advisory Board
Review (RTR) Consultative Panel
Enclosures:
ATTACHMENT 1: Overview and Charge Questions
ATTACHMENT 2: Panel Member Comments
-------
ATTACHMENT 1
Office of Air and Radiation
Risk and Technology Review (RTR) Assessment Plan
Peer Review Charge
Overview
As a part of the technical basis for rulemaking in the EPA's Risk and Technology Review
(RTR) effort, EPA seeks input on whether its proposed RTR risk assessment methodology
(emission data; dispersion and exposure modeling; risk characterization) is adequate to provide
the basis for regulatory decisions concerning specific source categories. In sum, we are using a
new approach to perform an assessment with the goal of characterizing the exposures and risks
associated with the emissions of hazardous air pollutants (HAP) from approximately 30 different
industrial source categories. These categories have previously been subjected to national
emission standards, and the purpose of characterizing their risks now is to determine whether
those emission standards (which were based on emission control technologies, work practices
and other control measures available at the time they were promulgated) are adequate to protect
public health with an ample margin of safety and prevent adverse environmental effects. The
Residual Risk section (section 112(f)(2)) of the Clean Air Act (CAA) requires EPA to make this
determination through notice-and-comment rulemaking. If EPA determines that the previous
standards do not reduce lifetime excess cancer risks to the individual most exposed to emissions
of less than one in one million, EPA must issue a standard that protects public health with an
ample margin of safety unless EPA determines a more stringent standard is necessary to prevent
an adverse environmental effect. In determining if a more stringent standard is necessary, EPA
will take into account costs, energy, safety and other relevant factors.
The overall plan for the new approach is as follows: (1) conduct a risk assessment using
currently-available source and emissions data; (2) share the source and emissions data and the
results of the assessment with the public through an Advanced Notice of Proposed Rule Making
(ANPRM), asking for public comments on the methods and the source and emissions data; (3)
receive comments; (4) reconcile comments and correct the source and emissions data as
appropriate, and; (5) re-assess the risks. The results of the revised risk assessment will be used
to support proposals and promulgations of technology- and risk-based regulatory decisions for
each of the categories through the regular notice-and-comment rulemaking process.
Since the planned risk assessment represents a significant activity of an influential nature
(i.e., it will be used for regulatory purposes), and since it departs from previous risk assessments
for the residual risk program in several important ways, we are seeking a scientific peer review
of the assessment through the Agency's Science Advisory Board.
A background on the residual risk program and the regulatory decision framework
associated with it is provided below to provide reviewers with a context for the assessment.
-------
Background ~ Regulatory Context and Decision Framework
Section 112(f)(2)(A) of the CAA requires EPA to evaluate whether previously adopted
standards applicable to emissions of HAP from source categories under the technology-based
(Maximum Achievable Control Technology, or MACT) program provide an ample margin of
safely to protect public health and prevent adverse environmental effects, taking into
consideration costs, energy, safety and other relevant factors. Any standards set under this
section are to ensure that the public health is protected to a level which provides an "ample
margin of safety" unless we determine that a more stringent standard is necessary after
considering costs and other relevant factors. (We also recognize that emissions of persistent,
bioaccumulative HAPs (PB-HAPs) may have an adverse environmental effect and need further
evaluation beyond a human inhalation risk assessment. Source categories that emit PB-HAPs
will be addressed in the RTR III assessment.) The specific language of section 112(f)(2)(A)
directs EPA to set additional standards through rulemaking if we determine that the MACT
standards for the regulated source category do "not reduce the lifetime excess cancer risks to the
individual most exposed to emissions from a source in the category or subcategory to less than
one in one million..."
Risk assessments performed in the residual risk program are designed to generate
answers to the questions posed by section 112(f)(2). Thus, the initial goal is to determine if
previous standards do not reduce lifetime excess cancer risks to the individual most exposed to
HAP emissions from a source in the category or subcategory to less than 1 in 1 million. In this
context, the "individual most exposed " is an individual who has been determined to "live" in a
census block that currently shows at least one resident (EPA is using the 2000 census data). If a
source presents lifetime excess cancer risk levels to this individual of at least 1 in 1 million, then
EPA must conduct rulemaking to ensure protection of public health with an ample margin of
safely.
While the process for determining what constitutes an "ample margin of safety" was not
explicitly specified in the 1990 CAA, section 112(f)(2)(B) expressly preserves the interpretation
of "ample margin of safety" as it was used in the pre-1990 version of section 112 and as reflected
in the 1989 rulemaking promulgating National Emissions Standards for Hazardous Air Pollutants
for sources of benzene (the Benzene NESHAP>(54 FR 38044, Sept. 14, 1989). In that rule, EPA
explained that, "in protecting public health with an ample margin of safety , (we) strive to
provide maximum feasible protection against cancer risks from hazardous air pollutants by (1)
protecting the greatest number of persons possible to an individual lifetime risk level no higher
than approximately 1 in 1 million and (2) limiting to no higher than approximately 1 in 10
thousand the estimated risk that a person living near a plant would have if they were exposed to
the maximum pollutant concentrations for 70 years." (FR 38044, Sept. 14).
In the approach used in the Benzene NESHAP rulemaking, the first step of the two-step
ample margin of safety framework is the determination of acceptability, i.e., the level of cancer
risks which can be considered "acceptable" based on health considerations only (costs, technical
feasibility and other non-health-related factors are not considered at this stage). The
determination of what represents an acceptable risk level is to be made in the context of "the
world in which we live," that is, recognizing that our world is not risk-free. In the Benzene
NESHAP, EPA determined that the cancer risk to the individual most exposed to emissions from
-------
the sources addressed in that rulemaking could not be considered acceptable unless it was at or
below approximately 100 in a million (or 1 in 10,000). This determination established a
"presumptive" acceptable level of 100 in a million cancer risk, thereby providing a benchmark
for judging the acceptability of maximum individual risk for future risk-based emission
standards, but not constituting a rigid line for making those judgments. FR 38045, Sept. 14).
The second step of the ample margin of safety framework is the determination of what
level of standard actually provides an ample margin of safety to protect public health. The
maximum level of excess lifetime cancer risk associated with this standard can be no higher than
what EPA determines to be the "acceptable" level for the particular category or subcategory, and
it can be as low as or lower than approximately 1 in a million, but the ample margin of safety
must be determined by balancing the costs associated with further reducing emissions against the
health risk reductions achieved. To inform this judgment, residual risk assessments are designed
to provide multiple metrics of risk (e.g., maximum individual risk, distribution of risks across the
exposed population, total cancer incidence, noncancer hazard indices), as well as an indication of
the limitations and uncertainties associated with the assessment.
Finally, to fulfill EPA's obligation under section 112(d)(6), every eight years we must
review and revise as necessary (taking into account developments in practices, processes and
control technologies) emission standards adopted under section 112. As recently explained in the
notice of proposed residual risk rulemaking for the Hazardous Organic NESHAP (HON), we
view costs and risk as relevant factors in determining whether it is necessary to revise standards
under section 112(d)(6). 71 (FR 34422, 34437; June 14, 2006).
Previous Relevant Peer Reviews
Several previous peer reviews have covered elements associated with this assessment or
assessments with similar scopes or contexts. A peer review of this assessment is not intended to
duplicate these previous efforts. A brief summary of each peer review is provided:
• The Residual Risk Report to Congress., a document describing the Agency's overall
analytical and policy approach to setting residual risk standards, was issued to Congress
in 1999, following a peer review by the Agency's Science Advisory Board. Many of the
design features of the RTR assessment were described in this report, although individual
elements have generally been improved over the techniques described in that document.
• Individual residual risk assessments - several internal peer reviews and one external peer
review were conducted on risk assessments for individual source categories, including
Coke Ovens, Perchloroethylene Dry Cleaning, and Halogenated Solvent Cleaners. Each
of these assessments used emission estimates from the National Emission Inventory,
human exposure modeling at the census block level, dose-response methodologies, and
risk characterization which are similar to those for the planned RTR assessment.
• The National Air Toxics Assessment, or NAT A, for 1996, was peer-reviewed by an SAB
panel in 2002. While this assessment was a comprehensive and cumulative risk
assessment (it was designed to include all mobile sources, small industrial sources, and
large industrial sources, as well as background contributions of air toxics), because of the
-------
large amount of associated uncertainty it was deemed to be not appropriate for regulatory
purposes, and it did not carry a census block-level resolution (it was performed at the
census tract level). For this reason, on EPA's NAT A web site the estimated risks are
characterized as "starting points" for developing refined assessments.
• AERMOD, a recently-developed source-to-receptor air quality dispersion model, was the
subject of significant interagency cooperation and peer review. It is now EPA's preferred
local-scale air dispersion model for industrial sources of air pollution.
• Toxicity assessments - the individual dose-response metrics used in the RTR assessment
have themselves been the subject of peer reviews through the agencies who developed
them (including EPA, through its Integrated Risk Information System, or IRIS; the
California Environmental Protection Agency, the Agency for Toxic Substances and
Disease Registry, and Environment Canada). EPA proposes to select dose-response
values from these sources in the same priority order it used for NATA (i.e., IRIS, then
ATSDR, then CA).
• It is not the intention of this consultation to duplicate or comment on these previous peer
reviews, but rather to acknowledge that we are using the most recent scientifically-
credible dose-response approaches (as determined through peer review) and that, since
dose-response science is continually-evolving, it is a source of significant, usually
unquantifiable uncertainty.
Charge Questions for Peer Consultation
The peer consultation on the plan for the assessment will focus on: (1) the development
of the source data and emissions inventory for the 30 source categories; (2) the analytical
approach for quantifying and characterizing human exposures and risks. In particular, we would
like reviewers to consider the following questions as a means of focusing their consultation:
1. Scope: Is the scope of the assessment appropriate for the stated purpose? Is the overall
approach clearly and adequately explained for review?
2. Emissions and Source Data: The National Emissions Inventory for hazardous air
pollutants represents an ongoing voluntary national effort whose creation results from the
collaborative efforts of State, local, and tribal air agencies with EPA Regional and
Headquarters staff.
a. Short of creating a federal mandate for reporting emissions to the EPA, do the
methods by which the NEI was developed, reviewed, and compiled result in a
technically-credible database that can support regulatory assessment and action?
If not, can you suggest ways to improve it?
b. Do the plans for conducting an engineering review and incorporating currently-
available refined emissions and source data into the inventory add value to the
assessment? Does the plan for soliciting public comment through an advanced
notice of rulemaking add scientific credibility to the inventory? Is the plan for
-------
reconciling comments on the inventory adequate? If not, can you suggest other
approaches for reconciling such comments?
3. Identifying Source Categories with Significant Non-inhalation Risk Potential: This
assessment is only designed to include source categories whose risks are dominated by
the inhalation pathway. Are the methods planned for selecting source categories with
potentially-significant ecological risks or multi-pathway human health risks for a
separate, more refined ecological and multipathway assessment sufficiently health-
protective? Are there ways that you might suggest for improving such screening
techniques that can make them less conservative and still scientifically-defensible?
4. Dispersion Modeling: Does the coupling of the AERMOD dispersion model with the
census block human exposure modeling (HEM) approach to estimating individual and
population exposures represent a credible approach for this goal? Are there other more
credible approaches available for the estimation of inhalation risks from the types of
source categories being examined? Is the level of accuracy of this approach acceptable
for the purposes of residual risk decision-making? Are there any specific source
categories for which this approach might be considered inadequate?
5. Acute Exposure Screening: The plan describes a screening methodology for identifying
potentially-significant acute exposures from routine emissions. Is this method
appropriately protective? If the potential for acute exposures of concern is identified, is
the plan for refining the assessments appropriate?
6. Exposure Assessment: Beyond the use of AERMOD-HEM, is the methodology planned
for characterizing exposure commensurate with the needs of residual risk assessments?
Specifically, does the underlying theory and data used to account for the effect of
population migration on exposure make our lifetime population risk assessment more or
less defensible than assuming that the exposed population lives in the same location for a
lifetime of 70 years? Is omitting the attenuation of exposure concentrations associated
with building penetration justifiable when estimating lifetime risks for these chemicals
and these types of sources? Is omitting the impact of short-term human activity patterns
on exposures acceptable for these purposes?
7. Dose-Response Values: Is the plan for using available dose-response information (e.g.,
sources of information, prioritization scheme) appropriate for the purposes of this
assessment? If not, can you suggest ways to improve it?
8. Risk Characterization: What are the strengths and the weaknesses of the overall
conceptual approach to risk characterization planned for this assessment? Does the
characterization plan adequately cover sensitive subpopulations and early-life exposures?
Does the risk characterization plan appropriately aggregate cancer risks? Does the risk
characterization plan appropriately aggregate noncancer risks? What are the strengths
and weaknesses of the planned approach for characterizing important uncertainties,
variabilities, and limitations? Given the underlying science and the intended purposes of
the assessment, can you suggest ways that the characterization of uncertainty and
-------
variability could be improved, made more transparent, or integrated more effectively into
the risk characterization?
Overall: Has any important scientific information been omitted from this assessment
plan that could impact a subsequent regulatory decision? In your opinion, will the overall
approach for the 30 source categories provide results that will be sufficient to support
regulatory decision-making in the context of EPA's residual risk program?
-------
ATTACHMENT 2
Comments from Dr. Timothy Buckley
1. Scope: Is the scope of the assessment appropriate for the stated purpose? Is the
overall approach clearly and adequately explained for review?
The Risk and Technology Review (RTR) Assessment Plan (Draft 11/20/2006) is a well written
report. The methods and plans are well organized and clearly described. The proposed risk
assessment approach is appropriate for meeting the regulatory needs under 112(f)(2) for
assessing source category residual risks.
7. Dose-Response Values: Is the plan for using available dose-response information
(e.g., sources of information, prioritization scheme) appropriate for the purposes
of this assessment? If not, can you suggest ways to improve it?
The tiered approach for selection of hazard ID and dose-response values for use in the RTR is
appropriate and reasonable. The development of dose-response values is a complex and
challenging process that requires distillation and synthesis of an always evolving primary
literature in a systematic and transparent way. IRIS provides an appropriate 1st priority database
because it contains a large number of relevant chemicals and is scientifically credible, in recent
years including a process of peer review. The use of ATSDR and CalEPA dose response
findings as secondary and tertiary sources appears reasonable in evolving from federal to state
agencies. This section of the RTR Assessment Plan can be strengthened by more completely
describing the differences between the three databases (e.g. peer review, # of chemicals, staff and
resource allocation) that support priority selection. It would also be of interest to know what the
overlap is across the three data sets and a general assessment of how values differ (e.g. is one
more conservative than another).
The hazard identification and dose-response approach for selecting acute effects is not as clear
cut in that not all HAPs have values listed and selection is individual chemical specific based on
professional judgement. The criteria for selection including conceptual consistency and level of
peer review are appropriate.
On pg 16 it is stated that dose-response values for chronic oral exposure were obtained from
OAQPS rather than IRIS but this selection is not justified. Why wouldn't IRIS be the
appropriate selection as is the case for inhalation exposure?
9. Overall: Has any important scientific information been omitted from this
assessment plan that could impact a subsequent regulatory decision? In your
opinion, will the overall approach for the 51 source categories provide results that
will be sufficient to support regulatory decision-making in the context of EPA's
residual risk program?
-------
The report is comprehensive in its consideration of the science that underlies the RTR
Assessment Plan.
The report's scientific credibility can be strengthened by including the extensive peer review
literature that relates to NATA risk and ambient concentration estimates. Studies
comparing/validating ambient concentration measurements to modeled estimates are particularly
relevant and provide assurance of the model estimates.
The scope of the RTR II is stated to be based on practical considerations including the
availability of emissions, complexity of the assessment, etc (pg 4). Although it is understandable
that feasibility is considered, a stronger rationale for inclusion under RTRII would be an
assessment of what source categories are believed to pose the greatest risk and therefore where
will the greatest public health benefit be gained.
Section 1.2 RTR II Process could benefit from a flow chart that shows the process elements and
how they relate to one another.
-------
Comments on RTR-II Charge Questions 3&8
A.R. Schnatter
December 18,2006
Charge question 3
Are the methods planned for selecting source categories with potentially significant ecological
risks or multi-pathway human health risks for a separate, more refined ecological and multi-
pathway assessment sufficiently health-protective? Are there ways that you might suggest for
improving such screening techniques that can make them less conservative and still scientifically
defensible?
The screening risk assessments for these compounds appear, in most cases, to be sufficiently
health protective, but the selection of the compounds should be scrutinized more closely to
assure that compounds and sources that may have such effects are not missed. Comments
consider both the selection methods and the procedures used in the screening risk assessments.
Selection of PB compounds
The proposed method uses existing Agency tools/policies that are not intended to prioritize
ecological risks associated with HAPs.
The proposed method assumes that the 14 HAPs designated as "persistent and bioaccumulative"
(Appendix 5) may pose an adverse environmental effect and are thus the focus of the screening
procedure proposed. This list of priority HAPs were selected based on previous EPA priority
lists/policies:
• PBT profiler
• Great Waters pollutants of concern
• TRI PBT rule
Based on information provided on EPA's website, the PBT Profiler is a screening tool, and PBT
estimations rendered by it are not sufficient for definitive PBT determinations. The profiler is a
research rather than a regulatory tool, and is used to identify chemicals that may need further
evaluation for potential Persistence, Bioaccumulation and Toxicity characteristics. The use of the
PBT profile for prioritizing HAPs that pose ecological risks is can be questioned.
Pollutants of concern for the Great Waters policy were selected on the basis of available data on
effects and deposition. These pollutants were known to be persistent and/or bioaccumulative and
cause adverse effects in humans and the environment. All pollutants were known to occur
specifically in Great Waters with atmospheric deposition being one potential source. It should
be pointed out that not all the compounds listed in this policy were considered persistent and
bioaccumulative (e.g. cadmium, BAP). For example, BAP, and PAHs (POMs) in general, were
designated as chemicals that were not bioaccumulative chemicals of concern in EPAs Great
Lakes Water Quality Initative.
The TRI PBT rule required reporting thresholds for substances designated as persistent
bioaccumulative toxic (PBT) chemicals. These chemicals were judged to be of particular concern
-------
not only because they are toxic but also because they remain in the environment for long periods
of time, are not readily destroyed, and build up or accumulate in body tissue. While substances
that are persistent and/or bioaccumulative will have higher exposure potential, HAPs that do not
possess these attributes may still pose ecological risks.
The approach proposed by EPA also does not differentiate differences in effect endpoints that are
relevant to human health versus ecological risk assessment. For example, in developing initial
PB-HAP emissions thresholds for POM, facility specific emissions were converted to toxic
equivalents of BAP using inhalation cancer risk estimates of 8 categories. However, cancer is
not the relevant endpoint for POM's ecological effects. Rather, POM ecologic risks are based on
survival, growth and reproduction due to narcosis.
An alternative approach is to develop a prioritization strategy that focuses on a relative human
vs. ecological risk assessment process. For most HAPs, protection of human health will likely be
protective of the environment. A transparent decision framework should be developed that (1)
identifies HAPs where this is not the case (ecological risks rather than human health endpoints
drive derivation of air emission thresholds) and (2) addresses relative importance of facility
emission relative to background sources for naturally occurring HAPs in prioritization.
Multi-pathway screening risks
Initial review of the preliminary assessment of non-inhalation human exposure to HAPs using a
multi-pathway modeling approach raises two general concerns. Firstly, there is a lack of
transparency in the model algorithms used to calculate exposure via non-inhalation routes. The
selection and use of particular model input parameters is unclear and undefended. Without
greater transparency, it is difficult to offer a comprehensive review of the modeling aspects.
Second, a simpler modeling approach may help reduce conservatism, while maintaining a high
degree of health protection and scientific integrity, by focusing attention on the most important
processes driving chemical fate and human exposure relative to the non-inhalation pathway.
The text and attachments of Appendix 5 do not clearly identify the model equations used to
estimate human exposure by multimedia pathways. While there are a number of model
parameters listed in Attachment A, a description of how these inputs are used is not included. It
is also important to note that there is no explanation as to why the listed values have been
selected. For example, on page 5-16, a soil mixing zone depth of 20 cm is selected, while a
value of 2 cm is used in the HHRAP default. There also appears to be two separate sets of food
consumption rate data (page 5-17), however, it is not indicated which (or if these) apply to adults
versus children. Moreover, a listed value of 0.074 kg kg"1 day"1 for milk consumption seems
unlikely (5.2 kg day"1 for a 70 kg adult or 1.1 kg day"1 for a 15 kg child).
An ambient air temperature of 11 °C is used in this analysis while the HHRAP default is 25 °C.
It is not clear why this value has been used and more importantly, whether or not physical-
chemical properties have been adjusted accordingly (i.e. physical-chemical properties are
commonly measured at 25 °C and need to be adjusted for application at different temperatures).
There is also concern regarding the selected environmental parameters representing the worst
-------
case scenario. For example, the water flow rate from the lake has been set to zero, which is
probably not truly realistic.
The current approach for assessing multipathway exposure to HAPs requires a large number of
model input parameters. At a screening level, the goal should be to reduce these inputs to the
least number possible. One method may include the use of sensitivity analyses; which help to
highlight the parameters that drive chemical exposure. By indicating the most sensitive
parameters, efforts can be focused on obtaining the best estimates or measurements for these
inputs. This approach can be used to reduce model conservatism. Furthermore, and maybe most
useful, would be to use a multimedia fate model to help identify the environmental and physical-
chemical properties for which non-inhalation exposure pathways become most important (i.e. the
ratio of the intake fraction in water, food or both to the intake fraction by inhalation is greater
than 1). Not only are models, such as EQC or RAIDAR more transparent and widely used and
accepted, they offer simpler approaches to determine important environmental and physical-
chemical parameters.
Charge question 8
8aj What are the strengths and weaknesses of the overall conceptual approach to risk
characterization planned for this assessment?
Strengths:
1. attention is given to using validated emissions data as the basis for estimating exposure.
2. an attempt is made to provide a consistent approach for assessing risks for different
sources.
3. an attempt is made at estimating the effect of uncertainties in risk characterization
Weaknesses:
1. use of a single value (e.g. slope factor or reference concentration) to characterize potency.
The uncertainty in these values is not made transparent in final decision making
2. use of potency values from different agencies
3. non-transparent methods to incorporate uncertainties or variabilities into final steps of the
process.
Overall, many of the techniques proposed seem most appropriate for a screening risk assessment
- i.e. to produce a ranking of source categories that would require more in-depth risk
assessments that account more directly for many of the scientific uncertainties inherent in use of
slope factors, population estimates, exposure modelling, emission estimates, etc.
8b) Does the characterization plan adequately cover sensitive populations and early life
exposures?
It is difficult to answer this question generically. For some HAP's (and therefore some source
categories), the cancer slope factors or reference concentrations may have uncertainty factors or
other health protective assumptions that would be expected to adequately cover sensitive sub-
-------
populations. Indeed, in some cases, the dose-response information could have been derived from
these sensitive sub-populations, and therefore would cover them. There is always the possibility
that a more sensitive, unknown subgroup exists. However, in the absence of plausible scientific
evidence that these subgroups exist, it would be difficult to account for their "possible"
existence.
When known sensitive subgroups exist of a reasonable size, the RTR should account for this
sensitivity. On page 24, the RTR-II document states that risk characterizations will assess
physiologically susceptible demographic groups or life stages, if known. This is certainly
appropriate and should be encouraged.
Page 24 also states that for HAP's acting by a mutagenic mode of action, extra factors of 10 (for
children aged 0-1), 3 (for children aged 2-15), or 1.6 (for 70 years beginning at birth) will be
applied. This is likely to be overly conservative, since cancer is primarily a disease of the elderly
and the longest latent periods (e.g. for asbestos and mesothelioma) are on the order of 40 years.
The extra factors proposed for "early life sensitivity" are likely to result in higher risk estimates
than are truly present. They may be justified in a screening exercise, (to perhaps make them
somewhat comparable to the conservative uncertainty factors used to derive RfC's), but are
probably not justified in final risk assessments of source categories.
8c) Does the risk characterization plan appropriately aggregate (cancer and non-cancer) risks?
Page 25 correctly notes that non-cancer HQ's that act by similar modes of action can be
aggregated. It's also noted that when this information is absent, HAP's that affect the same target
organ will be aggregated to form a target organ specific hazard index. However, some HAP's are
likely to act on the same target organ through different modes of action. This situation (same
target organ/different MO A) is not a possibility for all unknown MOA's. Since a large number of
TOSHI's could be made up of unknown, aggregated MOA's, EPA should consider balancing
this conservative assumption with relaxed judgments on the evidence needed to apply a known
MO A. The strategy of developing target organ specific hazard indices when an MOA is not
known is better justified in the screening assessments rather than the final risk characterizations.
8d) What are the strengths and weaknesses of the planned approach for characterizing
important uncertainties, variabilities and limitations! Given the underlying science and the
intended purposes of the assessment, can you suggest ways that the characterization of
uncertainty and variability could be improved, made more transparent, or integrated more
effectively into the risk characterization?
This is perhaps the most important, but least clear aspect of the RTR-II assessment plan. Table 2
begins to provide a good framework for assessing various uncertainties (and variabilities) in the
RTR-II plan. The outputs listed in section 4.5 include a table of "generic sources of uncertainty
and variability for all source categories and for each specific source category. Both of these are
strengths. While this is good practice, it is still unclear how uncertainties are factored into final
decisions such as whether a new control technology is indicated for a given source, or for a given
set of HAP's. This is the primary weakness.
-------
The characterization of uncertainty and variability could be improved and made more transparent
by including estimates, where possibly quantifiable in the outputs mentioned in section 4.5. EPA
should consider regarding the 'first pass' of the RTR-II plan a screening assessment to rank high
priority sources and/or HAP's. Here, uncertainty could be characterized as proposed by the
agency. Then, a second step could be made for higher priority sources and/or HAP's identified
through the first pass. In this step, a more complete and transparent assessment of variability and
uncertainty could be conducted for the sources or HAP's that are ranked high by the screening
assessment.
A couple of examples of what this more complete assessment of uncertainty and variability
might entail follow. In the exposure assessment area, centroid locations for receptors should be
supplemented minimal and maximal distances within blocks. Exposure estimates with and
without plume depletion assumptions should be made.
Another example involves dose response values. Rather than point estimates of dose response
values, ranges for these should be given. For example, the IRIS URE for benzene is 2.2 to 7.8 x
10"6, and within that range any calculated URE has equal scientific validity. Yet only the 7.8
value is used in risk characterization. Uncertainty in reference concentrations are thought to
cover one order of magnitude. Thus a reference concentration that shows this uncertainty
explicitly should be provided.
While this process may ultimately be more resource intensive, it involves a better scientific
assessment, which should focus attention on the sources and/or HAP's that may be truly
affecting population health.
There is one issue in the Table 2, where the magnitude and direction of uncertainties are
mentioned. It is suggested that the largest uncertainty is that resulting from not considering
background exposures. Yet, as the agency has stated in other places, the goal of the RTR-II plan
is to estimate the additional risk from specific sources. Thus, background exposures are
appropriately not included. While this could result in underestimation of total exposure, it does
not result in underestimation of exposure or risk from the sources included in RTR-II. Since
Table 2 includes the influence on risk estimates from sources or HAP's, "background risk"
should not be included, or if it is, the influence on risk estimates (from sources) should be
"mixed".
-------
Final Comments from Dr. Jeff Fisher
Charge Question 5. Plan describes screening methods for identifying important short term
exposures. Is the method protective? Can the method be refined to support acute exposure
assessments?
ACUTE EXPOSURE APPROACH
Page 7, fourth paragraph, 'It will be assumed that the maximum one-hour emission rate from
a source is ten times the average annual emission rate for that source.' The factor of 10 is
based on a paper by Allen et al. (2004) using VOC data. More specific information may be
considered during the ANRPM comment period, especially if the screening identifies acute
exposures of concern.
ACUTE EXPOSURE DATA BASE FOR TOXICITY SCREENS
Appendix 6 contains a list of AEGL-1 and AEGL-2 values, ERPG-1 and ERPG-2, MRL, and
REL values.
Comments
It is unclear to me what the relationship between increasing the emission rate by a factor often
and the resultant model predicted exposures or the health risks. The authors need to edit their text
to include more information about the nature of the data they receive for use in dispersion/risk
modeling and explain how they perform exposure simulations after using the lOx factor. The
authors do more computational work than is described in the text of this document. My
questions below reflect some of my uncertainty about what is actually carried out with the
calculations.
The Allen et al. (2004) paper, as cited by the authors of this Plan, suggest that short term
emissions are greater than annualized emissions by a factor of 2-9 fold. It appears that the
resulting exposures can vary many fold, depending on assumptions such as weather conditions. I
think a better description of the consequences of the lOx factor is needed. If this assumption
holds for many source categories, this appears to be a public health protective measure. It would
be better to obtain stronger justification for this phenomenon. Are there monitoring data that can
support this? Is the under-prediction of short term exposures by the modeling programs a
function of the assumptions or model parameters? It appears that the data used in the model is in
units that do not lend itself to short term analysis? Please explain this for people who do not do
this type of computational work. Why do you need to stay with hourly intervals and annualized
information? Can the models be changed to predict an 8 hr exposure with 5 min increments?
The implementation of the acute exposure toxicity data base is not well defined. I think some
effort is needed to determine best how to use existing data bases (Appendix 6), a priori. I have
copied and pasted definitions of the toxicity values for the various data bases (see below). There
are several considerations for using these data bases, such as how current is the chemical value in
the data base, does the toxicity value makes sense relative to what you are doing (professional
judgment) and how is the toxicity value interpreted relative to how it will be used in the Risk and
Technology Review Assessment Plan. I am on the NAS AEGL subcommittee and most familiar
with this program. I think the REL values, by definition, may be most useful for this situation.
-------
These toxicity values represent hot spots from which the public may have one hour intermittent
exposures, where AEGL and ERPG values represent emergency situations and one time
exposure (in theory). MRL values may be useful if you are interested in longer term exposures.
I am unsure if much is known about temporal aspects of atmospheric exposures. AEGL values
for carcinogens contain a cancer risk calculation to ensure that the theoretical risks of cancer are
not exceeded for a short term high exposure. I believe this rarely happens.
AEGLs are developed for emergency situations and one time exposure. Thus, when these values
are finally approved by the NAS/AEGL subcommittee, single exposure data is much preferred
over repeated exposure data.
AEGL-? lepresent thteshoki exposure limits lor the general public and aie applicable to
emergency e\po>ure pencils ranging Ironi 10 minutes to X hours. Three levels AHtil-1.
AtUL-2 and AtUL-3 are developed lor each ol" five exposure periods (|() and 30 minutes. 1
hour. 4 hour>. and 8 hour*) and are distinguished by varying degrees of severity of loxie effects.
The three AF.Gl.s aie defined as follows:
AF.Cil.-1 is the airborne concentration (expressed us pails per million or milligram* per
cubic meter [ppm or mg/nr j) of a substance above which it is predicted that the general
population, me hiding susceptible indhidtials. could experience notable discomfort, irritation, or
cerium asymptomatic, non-sensory effects. However, ihe ell eels are not disabling and are
transient and reversible upon cessation ul exposure.
AEGL-2 is the airborne concentration (expressed as ppm or mg;nv'l of a substance above
which il is predicted that the genera! population, including .susceptible individuals, could
experience irreversible or other serious, long-lasting adverse health effects or an impaired ability
to escape.
AIX iL-3 is the airborne concent ration (expressed as ppm or nig'nv) ol a substance above
\\hieh n is predicted that the .general population, including .susceptible individuals, could
experience lit'c-tliieatctnng health effects or death.
Airborne concentrations below the Al'.Cil.-t represent exposure levels that could produce
mild and progressively increasing but transient and nondisabling odor, taste, and sensory
irritation or certain asymptomatic, non-sensory effects. With increasing airborne concentrations
above each AL(,'
-------
air and that results from routine operation of a facility or that is predictable, including, but not
limited to continuous and intermittent releases and predictable process upsets or leaks."
1.1.1 Definition of Reference Exposure Level (REL)
The concentration level at or below which no adverse health effects are anticipated for a
specified exposure duration is termed the reference exposure level (REL). RELs are based on the
most sensitive, relevant, adverse health effect reported in the medical and toxicological literature.
RELs are designed to protect the most sensitive individuals in the population by the inclusion of
margins of safety. Since margins of safety are incorporated to address data gaps and
uncertainties, exceeding the REL does not automatically indicate an adverse health impact.
ERPG
The Emergency Response Planning Guidelines (ERPGs) were developed by the ERPG
committee of the American Industrial Hygiene Association. The ERPGs were developed as
planning guidelines, to anticipate human adverse health effects caused by exposure to toxic
chemicals. The ERPGs are three-tiered guidelines with one common denominator: a 1-hour
contact duration (Figure 1). Each guideline identifies the substance, its chemical and structural
properties, animal toxicology data, human experience, existing exposure guidelines, the rationale
behind the selected value, and a list of references.
ERPG-3
Ihe irax'rr-L-rr flirt-cm* contenlralicn b-Iow «?/hich- H is- believed
neaHy all.'ndividi.!«ilscoi,!d be- exposed tar <.ip to I hour without
reiiVRq ord?ve:Opinq lil'e4i'ir«afgriirK| heaifh efeets.'
ERPG-2
=-.s the iM>:;fT»brp jirbcn^coneeniMlion b«low which -s is-bsltevftd
I -rial nt?a'1y aM Jids'/iduabcokld be eKcotedfar aa 1.3 I hour with cut
w.psrei-Krp.c; ordweiopinq irrewrs'bie of -Dlhe< serious hea;(h f'lsc's or
syrnploms ",«»tifc!"i could- impair ai> individual's aM;!$> lo like protecsiva action.'
'•.s ths trvay-rnun1 airboR? eon^ginralion bslo'.v k«,tiicl' :-t i'-. believed
IM! nearly aM-'ridi-yidueK could be e:<-s-tti&dfer jp lo I iiour-wi:h<>ul
w.psrB-iicfi'-u alhf-i "l'«in -'riilt! '.rj"iii«-ti>.i.«wfy»h«-jllh'iff>:ife- w ptfK
<> deaHy acfinid osjjtvl-orvdlj-'c uilur.'
FIGURE 1 . The three-tiered ERPG public exposure guidelines. The definitions and format are
from the ERPG publication.
The ERPG guidelines do not protect everyone. Hypersensitive individuals would suffer adverse
reactions to concentrations far below those suggested in the guidelines. In addition, ERPGs, like
other exposure guidelines, are based mostly on animal studies, thus raising the question of
applicability to humans. The guidelines are focused on one period of time: 1 hour. Exposure in
the field may be longer or shorter. However, the ERPG committee strongly advises against trying
to extrapolate ERPG values to longer periods of time.
10
-------
The most important point to remember about the ERPGs is that they do not contain safety factors
usually incorporated into exposure guidelines such as the TLV. Rather, they estimate how the
general public would react to chemical exposure. Just below the ERPG-1, for example, most
people would detect the chemical and may experience temporary mild effects. Just below the
ERPG-3, on the other hand, it is estimated that the effects would be severe, although not life-
threatening. The TLV, on the other hand, incorporate a safety factor into their guidelines, to
prevent ill effects. The ERPG should serve as a planning tool, not a standard to protect the
public.
MRL
MRLs are derived when ATSDR determines that reliable and sufficient data exist to identify the
target organ(s) of effect or the most sensitive health effect(s) for a specific duration for a given
route of exposure to the substance. MRLs are based on noncancer health effects only and are not
based on a consideration of cancer effects. Inhalation MRLs are exposure concentrations
expressed in units of p ATSDR uses the no observed adverse effect level/uncertainty factor
(NOAEL/UF) approach to derive MRLs for hazardous substances. They are set below levels
that, based on current information, might cause adverse health effects in the people most
sensitive to such substance induced effects. MRLs are derived for acute (1-14 days),
intermediate (>14 364 days), and chronic (365 days and longer) exposure durations, and for the
oral and inhalation routes of exposure Parts per million (ppm) for gases and volatiles, or
milligrams per cubic meter (mg/m3) for particles. ATSDR does not use serious health effects
(such as irreparable damage to the liver or kidneys, or birth defects) as a basis for establishing
MRLs. Exposure to a level above the MRL does not mean that adverse health effects will occur.
Charge Question 6. I think the idea of using dispersion models and risk assessment procedures
for technology control is an important step toward national methods and procedures. Keep up the
good work! The question of providing sufficient details to refine the risk estimates by accounting
for human daily activity/terrain, may go beyond the screening level. I would prefer to say, for a
given location, that the emission restrictions are sufficiently protective to protect some who
resides in the area of concern without leaving the region. If this is the 'worst case exposure'
calculation, then this becomes a bench mark, which is probably not realistic, but provides a worst
case scenario for the protection of public health. The use of population movement patterns is
desirable. The text should articulate that the calculations do not account for risks achieved from
living in other regions of the US or world and vise versa, when someone leave the region of
concern.
Response to SAB Charge Question Two
Responses Provided by Dr. Mark J. Rood, Primary Reviewer
December 19, 2006
Overall, the "Risk and Technology Review Assessment Plan" (RTRII) is well written when
discussing the emissions and source data analyses.
11
-------
2. Emissions and Source Data: The NEIfor hazardous air pollutants represents an ongoing
voluntary national effort whose creation results from the collaborative efforts of State, local, and
tribal air agencies with EPA Regional and Headquarters staff.
a. Short of creating a federal mandate for reporting emissions to the EPA, do the
methods by which the NEIwas developed, reviewed, and compiled result in a technically-
credible database that can support regulatory assessment and action? If not, can you
suggest ways to improve it?
Yes, the methods by which the NEI were developed, compiled, and reviewed will result
in a technically-credible database that can support regulatory assessment and action
The report could be strengthened by describing the sources of "voluntary" data and what
is done when the data are not provided for relevant sources. There is also a reference to
"EPA's National Emissions Inventory (NEI) containing] 2002 emissions data and source
characterization information for sources of HAP emissions3" in the RTRII on page 4.
However, reference 3 refers to a report about "industrial VOC emissions and their impact
on ozone formation." It would be useful to include a brief description of the source(s) of
the emission data that were used for the draft 2002 national inventory and a list of the
relevant references that describe those sources, instead of referring to the 1999 NEI and
the TRI as mentioned in the RTRII (page 9).
b. Do the plans for conducting an engineering review and incorporating currently-
available refined emissions and source data into the inventory add value to the
assessment?
Completion of a careful engineering review of the parameters used as inputs to dispersion
models (e.g. existence of the sources, emission inventories, physical stack parameters,
stack gas properties, temporal variability of the source strength, and location of stack
with respect to plant boundary) is an important part of RTR II. The reported intent of the
review is to screen the data to readily identify short-comings and problems. However, the
example RTR questionnaire and summary spreadsheet, that is located in Appendix 1,
refers to "looks accurate," "looks correct," and "looks reasonable." There is no guidance
as to how to interpret those qualitative assessments for a quantitative emission inventory.
There is also reference to significant discrepancies between two inventories (i.e. > 50%)
that will initiate additional actions. It would be constructive to justify the magnitude of
such difference. Inclusion of select sensitivity analyses will strengthen the report when
describing criteria used to make decisions about the database.
The maximum one hour emission rate is assumed to be ten times the annual average
emission rate, and such assumption is based on reference 3 (Allen et al., 2004). Short-
term emissions of pollutants related to ozone formation (e.g. NOx, VOC, HRVOC, 1,3-
butadiene, butene, propene, and ethene) were compared to their annual average emissions
in that report. It was unusual (e.g. a few events per year) for the short-term emission rate
to be greater than 10 times the annual average emission rate for the four county region.
However, the actual short-term emissions could have local impact on air quality over
limited durations, but overall it appears that such short-term emissions do not add
12
-------
significantly impact to the annual based inventories^^ the studied region (four counties
around Galveston TX).
The draft RTRII reports that the highest peak emission event was < 8.5 times the annual
average. However, Figures 9, 10 (i.e. 1,500 Ib 1,3-butadiene/event compared to 97 Ib/hr
annual average), 12, and 13 appears to report events where short term emission rates were
more than 10 times the annual averaged emission rate. Overall, the assumption of using
10 times the annual average for short term events appears reasonable, assuming that the
acute risk assessments are applicable for a region, not a specific source.
Does the plan for soliciting public comment through an advanced notice ofrulemaking
add scientific credibility to the inventory?
Soliciting public comment through the advanced notice of proposed rulemaking
(ANPRM) is an important aspect of the draft RTR II, especially due to the voluntary
nature of the emission inventory that is used as an input to the dispersion models.
Potential oversights that could occur during the development and review of the emission
inventories by USEPA could be overcome by taking advantage of information provided
during the public comment period, and such effort is expected to improve the scientific
credibility of the emission inventory. Facilities and States that have not provided
emission data could be more receptive to providing such information during the public
comment period.
Is the plan for reconciling comments on the inventory adequate?
The methodology used to reconcile differences between comments provided during the
ANPRN and the draft 2002 national emission inventory are described in detail in Section
2.1.3, on p. 8 of the RTR II Plan. The Plan includes descriptions of how USEPA will
review the comments about the emission inventory, update the database, and provide
responses from the ANPRN to the original?/new? data providers. There is reference to a
detailed Quality Assurance document that describes quality assurance issues for the
development of the draft 2002 national emission inventory. Such report complements the
discussion about how to resolve differences between the comments obtained during the
ANPRN and the national emission inventory, but it is not apparent how the quality
assurance issues will be considered when updating the emission inventory based on
comments from the ANPRN.
If not, can you suggest other approaches for reconciling such comments?
The approach proposed by USEPA appears to be reasonable.
13
-------
Response to SAB Charge Question Four
Responses Provided by Dr. Mark J. Rood, Secondary Reviewer
December 19, 2006
4. Dispersion Modeling: Does the coupling of the AERMOD dispersion model with the census
block human exposure modeling (HEM) approach to estimating individual and population
exposures represent a credible approach for this goal?
AERMOD is a Gaussian based dispersion model that is very useful, but also has its
limitations (e.g. estimation of plume dispersion coefficients, treatment of chemically
reactive species during transport, extent of modeling domain (50 km radius), surface
roughness, and plume reflection). However, I am unaware of the current status of
regional or global models that could be used for the same purposes as AERMOD. Use of
modeled short-term (e.g. 1 hr) exposures to estimate actual short-term exposures could
provide a wide range of differences in those values. Text in the RTRII indicates that
pollutant concentrations will be over-estimated for materials that exhibit transformations
during transport, but such transformations should be very small during transport in
AERMOD's modeling domain. Such approach does not consider reactions such as the
formation of pollutants (e.g. ozone, gas to particle conversion, and oxygenated organic
species). These pathways will result in elevated concentrations of pollutants when
compared to the results provided by AERMOD, especially outside of AERMOD's
modeling domain, or within AERMOD's modeling domain when stagnant meteorological
conditions exist. Comparison of measured field results to modeled results will help to
better understand how well the models predict actual conditions for HAPs.
Effort to couple AERMOD with census block human exposure modeling (HEM) appears
reasonable, with sufficient precision (finest resolution is comprised of about 40 people or
10 households). However, the input meteorology is based on 1991 calendar year data and
the census data are from 2000. Averaging of multiple years of meteorological data
before during and after 2000 appears to be a more representative approach. The report
would be strengthened to provide a few sentences for the justification of the time period
used for the meteorological data compared to the census data.
Overall, the description of using AERMOD, AERMET, and HEM is well written to
describe their trade-offs. The approach described in the RTR II appears credible, but the
errors caused by invoking the assumptions in the models are difficult to quantify.
Are there other more credible approaches available for the estimation of inhalation risks from
the types of source categories being examined?
Not to my knowledge.
Is the level of accuracy of this approach acceptable for the purposes of residual risk decision
making?
I am not able to locate quantitative analyses about the accuracy of using AERMOD with
AERMET, and HEM. However, there are useful discussions describing qualitative issues
14
-------
that will influence the accuracy of this approach (p. 12-13). Hence, I am not able to
comment on the level of accuracy of this approach as it pertains to residual risk decision
making.
Are there any specific source categories, sources, or pollutants for which this approach might be
considered inadequate?
Pollutants that experience chemical reactions (e.g. SO2, photochemical reactants (e.g.
NOX, VOCs)), exhibit gas to particle conversion (e.g. H2SO4, NHs), hygroscopic
materials that are removed from the gas phase by clouds or precipitation (e.g. alcohols,
organic acids, SC>2 and HNOs), and are lost from the atmosphere due to dry deposition
(e.g. paniculate material) are examples of materials that appear to be inadequately
considered by AERMOD. There does not appear to be adequate treatment of materials
that experience transformations once they are deposited from the atmosphere (e.g.
elemental and ionic Hg), but such multi-pathway considerations are expected to be
treated in a follow-up RTR.
15
-------
Comments from Dr. M. Morandi. Charge Question 1.
Charge Question 1. Scope: Is the scope of the assessment appropriate for the stated purpose? Is
the overall approach clearly and adequately explained for review?
The scope of the assessment appears appropriate for the purpose of estimating source category-
specific residual risks as a tool for prioritizing the Agency's rule making activities. While there
are considerable uncertainties inherent in the proposed approach - some of them resulting from
limitations of the underlying data (e.g., source emissions) but others deriving from the state of
scientific knowledge (e.g., health impact from pollutant mixes) - unless the uncertainties impact
differentially the estimates of residual risk for some source categories compared to others, the
proposed assessment method will not result in misclassification of the source-specific residual
risks in terms of priorities (i.e., misclassifying a source as not having a significant impact on
residual risk compared to others, when it fact it does). Given the process proposed, which is
conservative, it is unlikely that misclassification of this type will occur, but the Agency should
undertake a sensitivity analysis to determine which inputs are the main drivers of the RR
estimates and if differences in level of uncertainties for those inputs (for example, uncertainties
in emissions from some sources compared to others) could potentially result in such
misclassification.
There is some concern that this type of effort, as it has happened with NATA to some extent,
acquires a life of its own once "number" are available to the general public and even the broader
scientific community, so that they may be misinterpreted and applied for inappropriate purposes.
While this is not the fault of the Agency, it is important that the Agency clearly and repeatedly
states in any communication to the public that the assessment is intended only for the purpose
intended by the Agency and that the estimates should not be construed as absolute estimates of
residual risk for use in population-based studies. The current documentation contains caveats, but
these should be listed up-front, and perhaps with some further elaboration on the limitations of
the approach for uses other than the intended purpose.
The narrative of the approach is presented in a clear manner but it may be useful to add one or
more diagrams with decision-making nodes clearly indicated so that it is easier to follow the
text. It would also be advisable to add a statement regarding what the Agency proposes to do if
and when input data may be insufficient in quantity or quality to make the estimates sufficiently
reliable for a specific source type (see the comments about the potential for misclassification.)
Performing the sensitivity analysis suggested above may also help in the prioritization in cases
where rankings are very close.
16
-------
Comments from Dr. M. Morandi: Charge question 9.
9. Overall: Has any important scientific information been omitted from this
assessment plan that could impact a subsequent regulatory decision? In your
opinion, will the overall approach for the 51 source categories provide results that
will be sufficient to support regulatory decision-making in the context of EPA's
residual risk program?
In general, the methodology incorporates current scientific understanding. One area that needs
further discussion is how fugitive emissions, especially those from large facilities, will be treated
and what the impact of uncertainties in the emission estimates or the identification/location of a
source of fugitive emissions might have in potentially misclassifying sources or strongly
underestimating exposures for some categories. It may be possible to identify some of these
situations and determine if this is the case. Overall, the methodology will be useful for
supporting decision making, but it is not clear at this time if that will be the case for all 51 source
categories.
17
-------
Comments from Dr. Randy Maddalena
Charge question #1:
Scope: Is the scope of the assessment appropriate for the stated purpose?
The scope of the assessment is stated in Section 1.1 of the RTR Assessment Plan in terms of the
choice of source categories. Essentially, the scope of the RTR II will be limited to source
categories that already have readily available post-MACT emission data in the most recent
National Emissions Inventory (NET) where the primary/dominant route of exposure is inhalation.
The purpose of the assessment is not stated very clearly but the last paragraph of the introduction
implies that the purpose of this "new approach" is to save time and money while improving the
consistency of the residual risk process and to ultimately satisfy the regulatory requirements of
the CAA. This will be done by grouping together source categories and assessing the residual
risks concurrently using available emissions data. If this is the "stated purpose" then the scope of
the study as indicated in Section 1.1 makes a lot of sense.
Beyond the need for streamlining the residual risk process, the RTR assessment plan (and the
residual risk process in general) seems to have two more overarching purposes. The first is to
determine whether additional risk reductions are necessary to protect public health and the
environment from industrial source categories after they have come into compliance with
Maximum Achievable Control Technology (MACT). The second, if necessary, is for the RTR
assessment to provide the basis for regulatory decisions concerning specific source categories. It
may be that these two overarching purposes are implied by stating that the approach will satisfy
the regulatory requirements of the CAA but it would be helpful if the purpose of the assessment
were stated clearly stated in the introduction.
Is the overall approach clearly and adequately explained for review?
The report is well written and reasonably concise.
As indicated above, I think it would be helpful if there was a section before the current section
1.1 that clearly stated the purpose of the new approach in the context of the overall residual risk
process. This is currently in the introduction but it does not come across very clearly.
Section 1.2, describes the RTR II process but it is difficult to pull out the specific steps and
chronology of events/tasks. For example, the section talks a lot about the collection and
18
-------
evaluation of emissions data but does not talk about the screen for PBTs and multipathway
HAPS, which seems to be an important part of the assessment. It would be helpful if this section
were written to present each step or task in the process and identify key decision points along the
way but without necessarily focusing on the details. A time-line or flow chart might be useful
here.
I would like to see a list of acronyms.
In addition to the model summaries provided in the text and appendices, it might be helpful for
reviewers if each of the individual models used in the assessment were linked to the relevant
page in the EPA's CREM Models Knowledge Base.
19
-------
FINAL COMMENTS
Richard A. Fenske, Ph.D., MPH
University of Washington
U.S. EPA Science Advisory Board member
#3. Identifying Source Categories with Significant Non-inhalation Risk Potential: This assessment
is only designed to include source categories whose risks are dominated by the inhalation pathway.
Are the methods planned for selecting source categories with potentially significant ecological risks
or multi-pathway human health risks for a separate, more refined ecological and multi-pathway
assessment sufficiently health-protective? Are there ways that you might suggest for improving such
screening techniques that can make them less conservative and still scientifically defensible?
The Agency indicated that it has already completed Part 1 of the screening process. This initial
screening was based primarily on a "modeling assessment of emissions of 13 PB-HAPs (excluding
dioxins/furans) from a hypothetical facility emitting into a domain that included a 'worst-case'
subsistence receptor population, constituting a conservative exposure group." The screening process
used an approach that back-calculated the emission rate of each PB-HAP that produced either a
lifetime cancer risk of 1 in one million, or a hazard quotient (HQ) of 1 for non-cancer effects at the
worst-case modeled receptor. The Agency has referred to this emission rate as the "threshold
emission rate" for the PB-HAP.
For the estimation of ingestion exposures, in lieu of site-specific data, the Agency incorporated a
health-protective farming/fishing scenario that was believed to represent individuals most exposed to
HAPs emitted from the model facility. The estimated exposures were combined with dose-response
values for chronic oral exposure recommended by QAQPS for screening-level risk assessments.
Since this was a screening process, the use of "worst-case" assumptions, standard factors, and
scenarios would appear to provide an ample margin of safety to protect public health.
The Part 2 screening process will take place after the ANPRM, and will employ a screening version
of the Total Risk Integrated Methodology (TRIM) model. The Agency indicates that this screening
process will be similar to the Part 1 process, but will take advantage of an improved emission
inventory, and will be using an "improved modeling platform". Presumably, this improved modeling
platform is the TRIM model. The TRIM model appears to be a fate and transport model that will
produce estimates of pollutant concentrations in soil, water, and biota that may be ingested.
The process described in the RTR-II document describes a later, more refined multi-pathway
assessment for those source categories deemed to hold the potential for significant non-inhalation
risks. However, this second-tier analysis appears to be focused primarily on refined fate and
transport modeling (TRIM model). It is not clear how the Agency will refine the human exposure
aspects of the estimates; e.g., the receptor population, intake rates.
The Agency should ensure that risk estimates would not be overly conservative in regard to
assumptions and scenarios. It would be helpful for the Agency to provide a more detailed discussion
as to whether/how it plans to modify the assumptions, standard factors, and scenarios in the second-
tier analysis.
20
-------
Appendix 5 refers to 51 source categories that were subjected to non-inhalation screening thresholds.
However, Attachment B in the appendix lists 33 source categories, and 51 sub-categories.
Presumably the authors have applied their analysis to sub-categories rather than categories. This
point should be clarified.
The criteria used to separate the five source 'categories' with "a significant number of exceedances"
from the 17 source 'categories' that indicated a "limited potential" for non-inhalation risks were not
clear. And again, the terminology is confusing. For example, the authors list "pulp and paper" as one
of the 5 source categories in the exceedance group, but the "pulp and paper MACT I and III" sub-
category has only 1 of 127 facilities that exceed the threshold (<1%), whereas "pulp and paper
MACT II has 13 of 134 facilities that exceed the threshold (9.7%), and the third pulp and paper sub-
category has 8 of 348 facilities in exceedance (2.3%). Are distinctions across these sub-categories
important in this screening process? If not, then why is the analysis conducted at the sub-category
level? And what should we conclude about the source category "ferroalloys" that has an exceedance
rate of 36% (4 of 11), or several other source categories that have double-digit exceedance rates?
The Agency could improve their explanation of the Part I process by providing the criteria used to
sort source categories into "significant" and "limited potential".
If feasible, a probabilistic analysis of the exposure scenario and other factors would provide more
transparency regarding variability and uncertainty in the exposure and risk estimates, and would
allow a more informed judgment regarding the extent of conservatism included in the
calculation/model.
#5. Acute Exposure Screening: The plan describes a screening methodology for identifying
potentially significant acute exposures from routine emissions. Is this method appropriately
protective? Can you suggest ways to refine the proposed acute exposure assessment process to
enable it to support decision-making?
Section 3.2 of plan (pp. 20-22) describes sources of acute dose-response information. It is not clear
why, on pages 21 and 22, there are discussions of URE values for formaldehyde, nickel, 2-
nitropropane, and POM. Perhaps this information belongs elsewhere.
The authors' judgment that the multiple sources of acute hazard information are probably not
comparable seems correct. Each of the four sources have used different criteria for calculating
hazardous air concentrations, and within each source values have probably been generated over time
with different methodologies. All of these sources are reputable, and the information they provide is
probably the "best available" science or professional judgment. The Agency has not delved into
these sources to determine what value would be most appropriate for a specific pollutant. It is hard to
judge whether these values are appropriately protective without a more careful examination of how
the Appendix 6 acute exposure values were derived by the individual sources.
21
-------
USEPA Science Advisory Board
Risk and Technology Review (RTR) Assessment Plan
Response to Charge Question 2
Bryan W. Shaw, Secondary Reviewer
2. Emissions and Source Data: The NEI for hazardous air pollutants represents an ongoing
voluntary national effort whose creation results from the collaborative efforts of State, local, and
tribal air agencies with EPA Regional and Headquarters staff.
a. Short of creating a federal mandate for reporting emissions to the EPA, do the methods by
which the NEI was developed, reviewed, and compiled result in a technically-credible
database that can support regulatory assessment and action? If not, can you suggest ways to
improve it?
Development of accurate NEI estimates is always challenging. Incorporating a framework
for improving the NEI as more accurate data becomes available is critical.
b. Do the plans for conducting an engineering review and incorporating currently-available
refined emissions and source data into the inventory add value to the assessment? Does the
plan for soliciting public comment through an advanced notice of rulemaking add scientific
credibility to the inventory? Is the plan for reconciling comments on the inventory adequate?
If not, can you suggest other approaches for reconciling such comments?
The plans for conducting an engineering review adds value. However, to maximize the
benefit, the assessment must be conducted in a thoughtful manner. Furthermore, the
subjective nature of the assessment may not provide the level of accuracy or precision
needed to be useful.
Public Comment Yes this is crucial and adequate.
Response to Charge Question 2
Bryan W. Shaw, Primary Reviewer
4. Dispersion Modeling: Does the coupling of the AERMOD dispersion model with
the census block human exposure modeling (HEM) approach to estimating individual and
population exposures represent a credible approach for this goal?
There are inherent shortcomings associated with any model. In this instance the model is a
Gaussian based model. The challenges are primarily associated with dispersion parameters and
chemical reactions/deposition. The chemical reactions can be critical as the concentration of
pollutant of interest at the receptor can change dramatically due to chemical change or
deposition (this can be an increase or decrease). As models are improved to better reflect the
atmospheric chemistry, this will improve models usefulness for estimating potential
health/environmental impacts associated with emissions of interest.
22
-------
Attempt to couple AERMOD and HEM seems to be appropriate so long as effort is made to
ensure compatibility of data.
Are there other more credible approaches available for the estimation of inhalation risks
from the types of source categories being examined?
Not that I am aware of at this time.
Is the level of accuracy of this approach acceptable for the purposes of residual risk
decision making?
I do not have quantitative assessment of the accuracy of this approach. However, this does seem
to be a reasonable approach so long as the uncertainty and/or bias of the model and disparities
between AERMOD and HEM data compatibility are addressed.
Are there any specific source categories, sources, or pollutants for which this approach
might be considered inadequate?
Potential source categories that may not be well addressed by this approach include any source
where significant chemical reaction is not well defined or where deposition is not well described
by the model. I have not identified any such categories.
23
------- |