SUMMARY REPORT OF THE REVIEW WORKSHOP
ON THE MERCURY STUDY REPORT TO CONGRESS
Andrew W. Breidenbach Environmental Research Center
Cincinnati, Ohio
January 25-26,1995
Prepared for:
U.S. Environmental Protection Agency
Environmental Criteria and Assessment Office
26 West Martin Luther King Drive
Cincinnati, OH 45268
Contract No. 68-C1-0030
Work Assignment 3-49
Prepared by:
Eastern Research Group, Inc.
110 Hartwell Avenue
Lexington, MA 02173
February 15,1995
-------
-------
SUMMARY REPORT OF THE REVIEW WORKSHOP
ON THE MERCURY STUDY REPORT TO CONGRESS
Andrew W. Breidenbach Environmental Research Center
Cincinnati, Ohio
January 25-26,1995
Prepared for:
U.S. Environmental Protection Agency
Environmental Criteria and Assessment Office
26 West Martin Luther King Drive
Cincinnati, OH 45268
Contract No. 68-C1-0030
Work Assignment 3-49
Prepared by:
Eastern Research Group, Inc.
110 Hartwell Avenue
Lexington, MA 02173
February 15,1995
-------
NOTE
This report was prepared by Eastern Research Group, Inc., a contractor of the U.S.
X"
Environmental Protection Agency (EPA), as a summary of the review workshop on the mercury
study report to Congress, held on January 25-26,1995, in Cincinnati, Ohio. As requested by
EPA, the report captures the main points of the presentations and discussions by peer reviewers,
and includes edited transcripts of comments made by observers. The report is not a complete
record of all the details presented, nor does it embellish, interpret, or enlarge upon matters that
were incomplete or unclear. None of the statements made by peer reviewers or observers
represent analyses or positions of EPA. This report will be used by EPA as one input for the
revision and finalization of the Mercury Study Report to Congress.
-------
CONTENTS
Page
1. INTRODUCTION 1
2. REPORT OF THE WORKSHOP CHAIR 3
2.1 Overview 3
2.2 General Review Panel Assessments of the Report 4
3. SUMMARY OF PREMEETING COMMENTS 7
3.1 Volume II (Emissions) and Volume in (Exposure) 7
3.1.1 Introduction 7
3.1.2 Natural Emissions 7
3.1.3 Anthropogenic Sources 8
3.1.4 Exposure to Mercury 9
3.2 Volume IV (Health Effects) 10
3.3 Volume V (Ecological Effects) 12
3.4 Volume VI (Risk Characterization) 13
4. SUMMARY OF BREAKOUT GROUP DISCUSSIONS 15
4.1 Exposure Breakout Group (Volumes II and III) 15
4.1.1 Volume II (Emissions) 15
4.1.2 Volume in (Exposure Assessment) 16
4.2 Effects Breakout Group (Volumes IV, V, and VI) 18
4.2.1 Volume IV (Health Effects) 18
4.2.2 Volume V (Ecological Assessment) 21
4.2.3 Volume VI (Risk Characterization) 23
-------
Page
5. OVERVIEW OF REVIEWER DISCUSSION IN THE PLENARY SESSION 25
5.1 Volume VI (Risk Characterization) 25
5.2 Volume VII (Risk Management) 27
6. EDITED TRANSCRIPTS OF THE PANEL DISCUSSION ON VOLUME VI
(RISK CHARACTERIZATION) 29
7. SUMMARY OF OBSERVER COMMENTS DURING PLENARY SESSIONS 47
7.1 Tom Hewson, Energy Venture Analysis, Inc., Arlington, Virginia 47
7.2 Jonathan Kiser, Director of Waste Services Programs, Integrated Waste
Services Association (IWSA), Washington, DC 48
7.3 Robert Collette, National Fisheries Institute, Inc., Arlington, Virgini 49
7.4 Evelina Norwinski, Hunton & Williams, Washington, DC 49
7.5 Arnold Kuzmack, Office of Water, U.S. EPA 49
7.6 Ralph Roberson, RMB Consulting & Research, Inc., Raleigh,
North Carolina 50
7.7 Robert Imhoff, Environmental Research Center, Air Quality Branch,
Tennessee Valley Authority, Muscle Shoals, Alabama 50
APPENDIX A WORKSHOP AGENDA
APPENDIX B LIST OF REVIEWERS
APPENDIX C LIST OF OBSERVERS
APPENDIX D REVIEWER PREMEETING COMMENTS
11
-------
1. INTRODUCTION
On January 25-26,1995, a IVi-day workshop was held at the EPA's Andrew W.
Breidenbach Environmental Research Center in Cincinnati, Ohio, to provide external review of
the draft Mercury Report to Congress. This draft report was prepared by EPA's Office of Air
Quality Planning and Standards and Office of Research and Development in response to Section
112(n)(l)(B) of the Clean Air Act Amendments of 1990, which requires EPA to submit a report
to Congress on mercury emissions. The draft report consisted of six volumes1:
n Volume II: Inventory of Anthropogenic Mercury Emissions in the United
States.
n Volume III: An Assessment of Exposure from Anthropogenic Mercury
Emissions in the United States.
n Volume IV: Health Effects of Mercury and Mercury Compounds.
n Volume V: An Ecological Assessment for Anthropogenic Mercury Emissions
in the United States.
n Volume VI: Characterization of Human Health and Wildlife Risks from
Anthropogenic Mercury Emissions in the United States.
n Volume VII: An Evaluation of Mercury Control Technologies, Costs and
Regulatory Issues.
In preparation for the workshop, Eastern Research Group, Inc., a government contractor,
identified 15 independent external scientists to review the document. The reviewers' expertise
covered a variety of subject areas relevant to the report, including mercury emissions and sources
of mercury emissions; the transport to and fate of mercury in the environment; the
physicochemical and biotic transformation among mercury forms in environmental compartments,
particularly of inorganic to methylmercury; exposure of human and ecological populations to
methylmercury and other mercurials; human and ecological toxicology; quantitative risk
assessment; and risk management. Each reviewer was asked to focus on that portion of the
Volume I, Executive Summary, is being prepared for the next edition of the draft report.
1
-------
report that matched his or her area of expertise. Reviewers prepared and submitted premeeting
comments on the report prior to the workshop.
Fourteen reviewers,210 EPA representatives involved in writing and/or revising the
mercury report, and 39 observers attended the workshop. The agenda included plenary sessions
and breakout groups. The first day of the workshop began with a presentation, by the two
breakout group chairs, of summaries of the reviewers' premeeting comments for Volumes II, in,
IV, and V. The participants then broke into two groups—one to discuss Volumes II and m, and
the other to discuss Volumes IV and V. During a plenary session at the end of the first day, the
breakout group chairs presented a summary of their groups' discussions and observers
commented on the report.
The second day of the workshop consisted of a half-day plenary session. Two reviewers
presented a summary of the premeeting comments on Volumes VI and VII, and all the reviewers
then discussed these two volumes. Following this discussion, additional observers presented their
comments on the mercury report.
This report summarizes the workshop proceedings, with a focus on the contributions of
the reviewers and observers. It includes a report of the workshop chair (Section 2), a summary
of the reviewer presentations and discussions (Sections 3, 4, and 5), an edited transcript of the
reviewer discussion of Volume VI (Section 6), and a summary of the observer comments (Section
7). Appendix A contains the workshop agenda. Lists of reviewers and observers are provided in
Appendices B and C, respectively. Appendix D contains the reviewer premeeting comments.
2One of the 15 original reviewers was unable to attend.
2
-------
2. REPORT OF THE WORKSHOP CHAIR—Paul Mushak, Ph.D.
2.1 OVERVIEW
The draft report and reviewers' comments clearly show that, while we do not know nearly
as much as we would like to about environmental mercury, we know a lot. In fact, we know
more about environmental mercury than about most contaminant metals or metalloids of
concern.
The principal challenge for both the authors and external reviewers of the draft report was
to critically evaluate the problems associated with integrating what we do and do not know into a
scientifically credible synopsis. One of these problems appears to be that the extensive database
for mercury is mainly available as discrete blocks of information within various scientific
disciplines, while the congressional mandate requires EPA to establish and quantify linkages
between these blocks of data. For example:
Information in one block tells us that the forms of mercury addressed in the
draft report—particularly methylmercury—are intrinsically toxic, with a
relatively high degree of lexicological potency to humans and various other
biological receptors. The types of toxic responses known or anticipated in both
ecological and human populations are qualitatively recognized.
Information in a second block tells us that mercury is emitted to the
environment from a variety of sources, and that one can generally determine
the relative contribution of different anthropogenic mercury source categories.
Information in a third block tells us that some fraction of the mercury emitted
to the atmosphere from a point source will eventually be deposited by
precipitation processes onto land and water bodies. Direct or indirect post-
depositional processes not only will impart mobility to the contaminant but also
wiU transform mercurial species.
Information in a fourth block tells us that inorganic ionic mercury entering
certain environmental compartments will undergo biomethylation to
methylmercury, and that methylmercury will accumulate and biomagnify in the
human food web, particularly in high-trophic-level predator fish. Data in this
block also show that mercurial forms can contaminate several environmental
media, depending on the exposure particulars.
-------
These examples of what we know clearly indicate that the difficulties in synthesizing all
this information into a coherent statement about the potential health and ecological risks posed
by mercury in the United States are rooted in uncertainties about how to quantitatively link these
blocks together. Areas of uncertainty include, for example:
n How much of current anthropogenic atmospheric emissions is deposited in
various environmental compartments?
n What is the link between natural and anthropogenic mercury in terms of
proportional contamination and subsequent impact?
n How much of this post-deposition mercury is converted to highly toxic
methylmercury?
n How much of any increased toxicity risk associated with consumption of
methylmercury-contaminated fish can be traced back to anthropogenic
atmospheric emissions of mercury?
The draft report was variably successful in dealing with the numerous complexities,
uncertainties, and data gaps connected with quantifying linkages. The essence of the reviewers'
comments concerned whether the report under- or overstated these uncertainties, particularly
with reference to risk characterization.
2.2 GENERAL REVIEW PANEL ASSESSMENTS OF THE REPORT
In their comments before and during the workshop, the peer reviewers recommended
revisions to strengthen the report's scientific credibility. Reviewers generally agreed that the
report would serve a useful purpose once it had been revised and improved in the various ways
they had suggested. Few, if any, reviewers felt the report should not be submitted at all, and no
reviewer thought the report should be transmitted without revision.
The review panel generally agreed that some portions of the report underestimated the
uncertainty associated with modeled estimates or pathway analyses. The panel suggested that
one way to better acknowledge this higher uncertainty was to use a range of values rather than
point estimates in the estimating exercise; some panel members also argued that a more refined
-------
point estimate could be presented in certain cases—for example, in deriving the reference dose
(RfD) for methylmercury.
On the other hand, the panel also generally agreed that the draft report conveyed too
much uncertainty by failing to include important peer-reviewed data available in the recent
scientific literature. For example, the exposure breakout group generally agreed that data do
exist to indicate a relationship between point-source mercury emissions and gradients in mercury
deposition consistent with a point-source contribution.
The review panel was similarly concerned about including or excluding available
information on other topics in the report. The panelists felt that the authors should revisit the
most recent scientific information to close any gaps that affect quantification of the linkages
noted above.
Reviewers also were concerned about the role of modeling in the report. However, they
had different opinions about how much data from the recent literature could be used to
complement the model estimates. Panelists generally agreed that the report volumes should be
more consistent and integrated, particularly concerning information relevant to risk
characterization.
-------
-------
3. SUMMARY OF PREMEETTNG COMMENTS
3.1 VOLUME II (EMISSIONS) AND VOLUME III (EXPOSURE)—Gerald Keeler, Ph.D.,
and Paul Mushak, Ph.D.
3.1.1 Introduction
Reviewers felt that Volume II probably was the best of the four volumes reviewed. The
approach used to characterize emissions was reasonable. However, the volume provides no
estimates of natural and baseline emissions and ignores several potentially important sources.
Specific comments on the various sections are provided below.
3.1.2 Natural Emissions
The inadequate coverage of natural sources of mercury detracts from the entire report.
Chapter 2, Natural Sources of Mercury Emissions, which consists of only a single page in Volume
II, is incomplete and misleading. The topic of natural sources of mercury is controversial and
qualitative at best. If the authors want to include this topic in the report, they should provide a
more complete and defensible assessment of natural emissions. Reviewer William Fitzgerald,
Ph.D., recommended that natural emissions could be roughly calculated using an approach
similar to that of Mason et al. (1994).3 This approach suggests that natural emissions in the
United States are approximately 20 percent of anthropogenic emissions. A recent estimate of
natural emissions in Europe gave a similar result of 25 percent of the total emissions (Axenfeld
et al., 1992).4 However, the quantitative data concerning natural emissions are very limited, and
there are numerous problems with the estimates in the literature.
3Mason, R.P., W.F. Fitzgerald, and F.M.M. Morel. 1994. Aquatic biogeochemical cycling of
elemental mercury: anthropogenic influence. Geochim. Cosmochim. Acta 58:3191-3198.
"Axenfeld, F., J. Munch, and J.M. Pacyna. 1991. Europaische Test-Emissionsdatenbasis von
Quecksilber-Komponenten fur Modellrechnungen. Dornier Report. Friedrichshafen, Germany.
-------
3.13 Anthropogenic Sources
The report's list of source categories for mercury emissions is complete with respect to the
major source categories. Many of the source categories discussed have relatively low annual
mercury emissions. For a few source categories for which insufficient information was found, the
report provides no emission estimates. Emission factors and data are missing for several
potentially important sources, including hazardous waste incinerators, primary mercury
production, mercury compounds production, by-product coke production, refineries, and mobile
sources. In addition, Volume II provides no information or discussion on emissions from iron-
steel production and primary zinc production. Emission factors and data are available for
European sources and could be used to estimate the U.S. emissions to determine their potential
importance (see page D-25 of Appendix D of this workshop summary report).
The report to Congress provides only very limited information on emissions of various
physical and chemical forms of mercury. Better information is needed on mercury speciation in
both emissions and environmental samples.
The report could be strengthened by adding maps showing the actual location of point
sources for categories like utilities (by fuel type), incinerators (sludge, municipal), iron-steel
production, coke ovens, and cement production. The spatial distribution of the gridded
emissions presented at the workshop by report author Martha Keating should also be included.
The report suffers from a general lack of recent information and actual measurement data
in the recent peer-reviewed literature. References will be provided by the reviewers, and the
Monterey Mercury Meeting Book will be provided by Donald Porcella (Electric Power Research
Institute). Inclusion of more recent information will address such comments as "There is a
general recognition of uncertainty," "So much is said about uncertainty that it appears as if we do
not know much about mercury," and "Little of the most recent knowledge has found its way into
this report."
The meaning of some key terms used in the report, such as "total emissions" and
"background," was confusing. The peer reviewers strongly recommended that the authors add to
8
-------
the report the definitions provided in the Atmospheric Mercury Expert Panel Report and that they
use the various terms consistently throughout the report based on these definitions.
Lastly, the report lacks information regarding seasonal or temporal variations in emissions
by source category. While utilities may have fairly constant emissions both diurnally and
seasonally, other sources do not. Operations involving multiple steps over different time periods
will probably have time-varying emissions.
3.1.4 Exposure to Mercury
A comprehensive quantitative assessment of the relationship between anthropogenic
mercury releases to the atmosphere and the potential exposure of people, wildlife, and terrestrial
and aqueous systems to these releases may not be possible due to the apparently limited state of
knowledge of the mercury cycle in nature and the environmental consequences from
anthropogenic emissions of mercury. The report states that the exposure assessment is a
"qualitative study based partly on quantitative analyses." As noted by reviewer William Fitzgerald
in his premeeting comments:
...this important exposure assessment provides a valuable guide for research.
Although the results and conclusions are qualitative, this extensive and essential
modeling effort provides a credible means for evaluating the present sparse data
base, and for identifying major gaps, inconsistencies and weaknesses associated with
major aspects of the biogeochemical cycle of Hg at the Earth's surface....
As the report confirms, human exposure to methylmercury is almost exclusively from
consumption of fish and fish products. Intake of methylmercury through consumption of
nonlocal fish and seafood should be evaluated. Such intake should not be considered
"background," as the mercury found in coastal environments and in saltwater fish may be of
anthropogenic origin. The report lacks an assessment of the exposure of the marine
environment—especially the coastal zone—to anthropogenic mercury emissions and of the effects
of such exposure.
-------
The modeling results should be "ground-truthed" where possible. The report's estimates
of deposition and water concentrations often are more than an order of magnitude greater than
any actually measured in the United States.
3.2 VOLUME IV (HEALTH EFFECTS)—Steven Bartell, Ph.D., and Paul Mushak, Ph.D.
Several of the reviewers' key premeeting comments concerned the major sources of intake
and exposure in human populations. Some reviewers suggested that the report should address
the contributions to human mercury exposure of dental amalgams containing mercury. Similarly,
reviewers recommended that the drinking water pathway be further examined, including the
potential human health risks associated with drinking water at locations known to have elevated
mercury concentrations in ground water. The report should clearly explain why particular papers
concerning human health endpoints are cited while others were omitted.
Reviewers also commented on the subject of mercury disposition among biological
indicators of mercury exposure, particularly exposure to methylmercury. The derivation and use
of a constant ratio of mercury in hair to mercury in blood for estimating blood levels of mercury
may require additional attention. Reviewers expressed reservations about the time-scale
differences implicit in comparing blood mercury with hair mercury—namely, that mercury
concentrations in hair reflect exposure over a longer time scale, while mercury concentrations in
blood may correspond to a shorter time frame. The reported variability may reflect
interindividual variability rather than just measurement error as Volume IV suggests. Reviewers
identified an error in the equation used to calculate the methylmercury concentration in blood;
an additional term defining blood volume is needed to make the units in this equation work out
to those stated.
The quantitative linkage of mercury intake by exposed populations and the expression of
some toxic endpoint is mediated through the toxicokinetics—i.e., the uptake, distribution, and
retention/excretion—of the particular mercurial. The modeling of the systemic behavior of
methylmercury is particularly critical in this regard. The reviewers felt that the derivation of the
parameters used in the pharmacokinetic modeling needed additional explanation and
10
-------
justification. For example, the elimination rate or half-life used to describe methylmercury
conversion to inorganic mercury and its subsequent removal from the body in feces is an
important model parameter; reviewers disagreed about the most appropriate value. Differences
in this parameter can result in appreciable variability in the modeled mercury concentrations for
the human populations of interest.
Chapter 4 on toxic effects of various mercurials, particularly methylmercury, was the
subject of several comments. The organization and presentation of toxic endpoints in the
chapter could benefit by progressing from lethal through acute effects to subchronic and chronic
effects. Distinct subsections organized along this framework would improve the presentation.
The rationale for selecting the set of core studies of toxic responses should be clarified.
Not surprisingly, many comments involved the chapter on dose-response relationships.
Several reviewers were concerned that the current RfD for methylmercury might not be
protective, particularly for more subtle neurotoxic endpoints such as neurobehavioral and
neurodevelopmental endpoints. One reviewer pointed out some confusion regarding the
interpretation and presentation of the apparent association between maternal methylmercury
exposure and abnormalities in deep tendon reflexes in their male children. Two reviewers
recorded their disagreement regarding the adjustment of No Observed Adverse Effect Levels
(NOAELs) and Lowest Observed Adverse Effect Levels (LOAELs) to lifetime exposures for
different exposure pathways (e.g., inhalation, ingestion) in the derivation of RfDs and reference
concentrations (RfCs). Exposure resulting from these pathways would be more realistically
described as intermittent, shorter-term events. There was apparent confusion regarding the
derivation and use of uncertainty factors (UFs) and modification factors (MFs). The values were
not carried through the analysis according to the usual protocols. Reviewers pointed out some
confusion and inconsistency regarding the relative sensitivity of adult and fetal developmental
toxicity used to derive overall human health assessment endpoints.
Reviewers disagreed with the presentation regarding the possible interactions between
mercury and selenium, particularly the implication that interaction with selenium may mitigate
the human toxic effects of mercury.
11
-------
3.3 VOLUME V (ECOLOGICAL EFFECTS)—Steven Bart ell, Ph.D., and Paul Mushak, Ph.D.
Reviewers were concerned with the efficacy of the overall approach to the report's
ecological assessment, which involves defining overlapping areas of potentially high mercury
exposures with the distribution of sensitive piscivorous birds and mammals. For example, the life
history and distribution of the Florida panther differ considerably from those of the mink or
kingfisher. Failure to address life history and migration patterns in developing this overall
approach might lead to inaccurate assessments of risk.
Reviewers also pointed out the report lacked a consideration of mercury effects on
organisms at lower tropic levels (e.g., plankton, invertebrates). Additional reservations were
expressed over the absence of wading birds, particularly species of declining abundance that are
known piscivores. Effects of mercury on fish and reptiles should also be explored, or their
omission should be further justified.
Reviewers were concerned about the report's dependence on assessment approaches and
data that emphasized the Great Lakes and upper midwestern lakes, for example, in developing
the bioaccumulation factors (BAFs). Concern also was expressed regarding the removal of
surface waters with pH > 5.5 from regions of concern. This approach would exclude the
circumneutral waters of the Florida Everglades that are suspected of posing mercury-related risks
to resident populations of birds and mammals.
A major review issue focused on the use of NOAELs as endpoints for developing the
wildlife criteria for the ecological assessment. This approach removes any consideration of a
dose-response relationship from the assessment. If measured or modeled mercury exposures
exceed the wildlife criteria values, we would not know the nature or magnitude of the expected
response. Also, this approach implies different time scales between the shorter-term toxicity data
used to develop the wildlife criteria and the longer-term exposure values. The fact that limited
data were used to develop NOAELs for the selected wildlife species also calls into question the
t
efficacy of the report's overall approach for estimating ecological risks.
12
-------
In developing the BAF values, the report essentially ignored the complex chemistry of
mercury in surface waters. Instead, these factors were developed using constant ratios of
methylmercury to total water column mercury. Reviewers expressed serious concerns with this
assumption, which ignores the complex environmental chemistry of mercury. Also, in developing
the BAF values, the assumption was made that the selected piscivores restrain their feeding to
specific "trophic level" fish. This assumption is certainly open to question; it remains unclear
what the impacts of this assumption are on' the resulting estimates of BAFs and wildlife criteria
used as endpoints for the assessment. The assumption of a simple linear food chain implied by
this approach was similarly of concern; the draft does not address spatial and temporal variations
in diet and feeding behavior that might increase or decrease exposures for the selected piscivores.
It was not clear what the exposure models (RELMAP, COMPMERC) really provide to
the assessment. The different spatial scales of these exposure models were not related to the
spatial scale of the distributions of the selected species.
Finally, the reviewers noted that the sensitivity/uncertainty analyses did not
comprehensively address all the components of the equations used to develop the BAFs or the
final wildlife criteria values. The reported analyses addressed some of the models' structural
uncertainties (e.g., correlations), but did not adequately address parameter uncertainty. The
results of the sensitivity analyses do not lend themselves to defining future research needs in
relation to reducing uncertainty on the endpoints of the assessment.
3.4 VOLUME VI (RISK CHARACTERIZATION)—Pamela Shubat, Ph.D., and Paul Mushak,
Ph.D.
Reviewers agreed that Volume VI fell short of expectations for a risk characterization of
health and ecological effects from mercury emissions. One reviewer felt that the necessary data
to conduct a risk assessment are lacking, considering that a risk characterization should estimate
the probability of health effects.
13
-------
Reviewers noted that the volume should have compared the measurements of fish mercury
levels and the incidence of health effects in populations to the volume's assumptions and results.
The volume assumed a body weight and a fish consumption rate for each species; it also assumed
a NOAEL and LOAEL for the selected species and derived a fish concentration that would
permit consumption without exceeding the NOAEL or LOAEL. Reviewers felt that more data
were needed to support this approach, and they expressed particular concern about the NOAEL
and LOAEL selected for each species.
Reviewers felt that the assumptions, in the relative exposure ranking, that a given lake has
only a single mercury concentration and a single trophic level were not accurate. The exposure
rankings for the eagle, kingfisher, otter, and other species should be compared to measured
values in tissue samples from these species.
14
-------
4. SUMMARY OF BREAKOUT GROUP DISCUSSIONS
4.1 EXPOSURE BREAKOUT GROUP (VOLUMES II AND HI)—Gerald Keeler, Ph.D., and
Paul Mushak, Ph.D.
4.1.1 Volume II (Emissions)
Panelists suggested that the "minor sources"—i.e., those not included in the quantitative
assessment—may contribute as much as an additional 20 percent to the total amount of mercury
emitted annually. European emission factors should be used to improve the accuracy of this
assessment of the minor sources.
Reviewers stressed that, to provide a complete picture of the atmospheric flux of mercury
and to properly assess anthropogenic contributions to environmental mercury, the report should
assess natural sources of atmospheric mercury as well as the reemission of mercury previously
deposited on both aquatic and terrestrial environments by anthropogenic emissions.
Reviewers suggested that a national network of atmospheric mercury monitoring be
established to validate emission data and to provide necessary information on trends in mercury
deposition.
The panel felt that the division of sources into point and area source categories should be
improved. For example, mercury emissions from residential heating furnaces should be defined
as area sources, while crematories and medical waste incinerators should be categorized as point
sources.
The panel agreed with the appropriateness of the emission factor approach. Many of the
emission factors are based on actual test data and measurements, which contributes to the
accuracy of the inventory. The emission estimates, when compared on a per capita basis, are
quite similar to those in selected industrialized countries in Europe (see the comment of Jozef
15
-------
M. Pacyna, Ph.D., on page D-27 of Appendix D). In addition, the total U.S. anthropogenic
mercury emissions are similar in magnitude to those of other industrialized nations in the world.
The exposure volume utilized state-of-the-art methods in investigating the relationships
between mercury emissions and exposures. Nevertheless, only plausible relationships between
anthropogenic emissions and exposure could be defined.
The draft report does not assess the impact of anthropogenic mercury emissions in coastal
environments. However, since fish consumption is the dominant exposure pathway, seafood or
saltwater fish should be included in the total exposure estimates.
The analysis presented in the report supports the conclusion that current levels of
emissions from major combustion/industrial sources result in incremental exposures above
background to both humans and wildlife through the consumption of contaminated freshwater
fish.
4.1.2 Volume III (Exposure Assessment)
The group discussed the use of exposure estimates derived from the RELMAP and
COMPMERC models. The discussants felt that the report should better describe how the model
estimates were added. After questioning the modelers directly during the breakout group, the
reviewers suggested that the authors consider alternative strategies for the risk assessment. For
example, decoupling the regional impact provided by RELMAP from the local-scale exposure
scenarios may improve the site-specific risk analysis and provide a clearer definition of the
uncertainties in the exposure estimates utilized in the risk assessment.
Reviewers recommended that actual observations (i.e., measured mercury concentrations)
could be used to "ground-truth" the model estimates or could themselves be used in the local-
scale risk assessments. Although a wealth of high-quality atmospheric mercury data or mercury
deposition data is not available, enough data are available from the Great Lakes programs to
perform a risk assessment at a similar or better level of accuracy than the models provided. The
16
-------
only drawback to this approach would be the lack of assignment of risk to specific source
categories.
Additional suggestions for improving the assessments include:
Evaluate the existing exposure to methylmercury via seafood consumption.
Base this evaluation on existing data and not the model results.
Perform the risk assessment and exposure to methylmercury from existing
freshwater fish data. (This could be time-consuming because so many data are
available.)
Utilize existing wet and dry deposition data as input to the Indirect Exposure
Model (IEM) to see what is predicted. This approach would remove two of the
greatest uncertainties from the modeling and could be used to estimate the risk
in the risk characterization.
Attempt to identify a better indicator of the central tendency (perhaps the
median) from the exposure assessment uncertainty analysis, which used the
distributions rather than the high-end (maximum) estimates.
In conclusion, the panel members felt that the accuracy of the estimates decreases as the
report moves from the initial emissions inventory through the exposure modeling using
RELMAP and COMPMERC to the risk assessment phases. This results in a risk assessment
that may have relatively large uncertainties and, therefore, may not provide a sound basis for
decision- or policy-making.
The report would be improved by providing linkage between the risk management and
the emissions inventory. The type and cost of mercury control technologies depend largely on
the form of mercury in an emission and, thus, on the source category being considered for
emission reduction.
17
-------
4.2 EFFECTS BREAKOUT GROUP (VOLUMES IV, V, AND VI)—Steven Bartell, Ph.D.,
and Paul Mushak, Ph.D.
4.2.1 Volume IV (Health Effects)
After some discussion, all or most group members generally agreed with the views and
recommendations reported below. Dissenting views on key issues, where they occurred, are
noted.
The group expressed several concerns about the organization and accuracy of Volume IV.
Chapter 4 is difficult to follow, but group members generally agreed that its goal was to provide
toxicity data for a human health risk assessment.
The description and discussion of lipophilicity of mercury compounds was not entirely
accurate. The term is simplistic and does not account for current knowledge of binding and
ligand-transfer interactions of methylmercury and other mercurials.
With respect to toxicity endpoints, the group noted that developmental impacts in the
neonatal period should not be dismissed, since neonatal effects of elemental mercury have been
reported in mice.
Differential sensitivity to mercurials among human populations is well established, and
the fetus is now assumed to be the most sensitive to effects of methylmercury. The basis of such
sensitivity includes physiological vulnerability, population variability concerning
biotransformations (e.g., demethylation of methylmercury by gut flora), and variable patterns of
exposure. Overall, sufficient data are not available to generate a highly resolved summary of
differential sensitivity.
Of concern to the reviewers was treatment of the time course of exposure-effect
relationships—i.e., are we dealing with latency or a masking phenomenon with long-term
exposures?
18
-------
Some reviewers were critical of the RfD calculation for inorganic ionic mercury (i.e.,
back-calculating from the drinking water equivalent level [DWEL]). Some also questioned how
good a surrogate the Brown Norway rat is for humans sensitive for renal effects in the form of
an autoimmune glomerulonephritis. One reviewer thought that the Integrated Risk Information
System (IRIS) document is not convincing in this regard, and recommended that the mercury
report at least reproduce the DWEL.
How UF factors were used in the analysis was not clear; the RfDs and RfCs need a
closer look. Authors should reexamine the original data to see if they can justify how they used
the numbers, and they should better explain their rationales.
The report should indicate that additional studies are under way (other than the Iraqi
data set), although it is not known when the data will be available. Basically, the message here
was to proceed with caution, but proceed.
Either Chapter 2 of Volume IV should be expanded to provide a concise summary of the
integrated exposures to mercury, or an integrating final section should be added in Volume ILL
The authors should include more information on mercury exposure from dental amalgams and
from ground waters that are or will be drinking water sources—particularly when mercury
concentrations in these waters approach or exceed the RfC or RfD. Information should be
added on how dietary components (other than methylmercury in fish) contribute to human
exposure. This should include information, however qualitative, on any linkages of nonfish
dietary mercury to atmospheric emissions.
Several comments concerned the mechanisms of mercury toxicology in humans and test
animals. Although mechanisms of toxicity are critical to understanding the plausibility of
epidemiological relationships reported for different populations and to understanding where
thresholds for toxic effects may lie, the report gave them short shrift. The report should expand
the discussion of this topic and should address how mercury forms move in and out of cells.
However, reviewers recognized that a complete mechanisms sections might require an effort
beyond the scope of the report.
19
-------
Reviewers generally agreed that the health endpoints selected for the assessment and the
dose-response relationship for each of the three forms of mercury were appropriate for the risk
assessment. However, they thought the authors should strengthen the discussion of the validity
of the endpoints and epidemiological data selected. Also, the group recommended that authors
scrutinize the numbers employed from modeling, such as the fraction that goes into blood, the
half-life, and the elimination parameter. The hainblood ratio of 250 seems to be a middle-of-the
road number and is probably acceptable. Reviewers questioned why the report did not use
distributional analysis rather than selecting point values that might result in an unknown bias.
The group's comments on Appendix C of Volume IV mainly concerned model
uncertainty and not variability in data-based parameters.
Reviewers considered the issue of selenium-mercury interactions. They felt this issue was
complicated because the data sets are isolated and have no mechanistic underpinning. The
critical question is how selenium in diet affects long-term exposures and associated chronic toxic
endpoints. Was the Iraqi population at risk because of dietary habits (i.e., because they were
grain eaters)? On the other hand, the reported selenium content of cereal grains is not vastly
different than the selenium content measured in certain fishes. Although the selenium issue may
have a bearing on which population exposed to methylmercury is valid for risk characterization,
reviewers felt it premature to use selenium intake as a criterion for selection. One problem
concerning the selenium-mercury connection is that the clearest associations are seen in gross
endpoints, such as high-dose teratogenesis.
Regarding which dose-response data to use in risk characterization, reviewers expressed
some sentiment for using at least two RfDs: one for the general adult population and one for
pregnant women. Reviewers emphasized that the methylmercury RfD used in the assessment
should be reported as an interim value, and that the assessment should be formulated to
facilitate near-term (i.e., within the next several months) modifications to the RfD.
Some comments expressed in the effects breakout group also concerned the risk
characterization volume. For example, the values of the NOAELs or LOAELs should be carried
20
-------
forth into the risk assessment instead of transforming them into permissible fish tissue
concentrations.
4.2.2 Volume V (Ecological Assessment)
The group generally agreed that the goal was to provide data for a risk assessment and
that the appropriate species were identified except for lower trophic levels and wading birds.
There was consensus that methylmercury was the compound of interest in addressing the
toxic effects of mercury on piscivores. The consensus was further evidenced by the reported
mortality of panthers, which was diagnosed as mercury toxicosis. The group also discussed the
fact that the population of wading birds in the Everglades has significantly decreased in the last 5
years. Loss of habitat and exposure to mercury were listed as the suspected causes of these
declines. One reviewer reported that loons in Minnesota also were suffering increased mortality
from mercury exposure. Analyses showed elevated mercury concentrations in the feathers of
juvenile loons. Approximately 2,500 loons died in coastal waters off Florida, in part from
mercury exposure.
One reviewer pointed out that ethylmercury was measured in the Everglades, but this
compound was not expected to be environmentally or lexicologically important in the overall
assessment. Ethylmercury has not been identified in fish, for example. Dimethylmercury also
exists in nature, but is quite volatile and, based on known information and the compound's
fundamental chemistry, is not expected to represent any significant ecological threat.
Reviewers generally agreed that the report's treatment of methylmercury as a constant
fraction of total mercury in the water column was an oversimplification. Additional work might
be undertaken to determine the impacts of this assumption on the final estimates of the BAF
and wildlife criteria values developed as assessment endpoints.
The group discussed the fact that chronic toxicity tests for methylmercury are extremely
limited and that such effects are difficult to demonstrate under field conditions. For example,
21
-------
eggs can be collected from the nests of mercury-contaminated birds; however, it is not easy to
detect toxic effects of mercury (e.g., hatching success, survivorship, growth). Different histories
of exposure for adult birds may also make it difficult to establish effects in the field. As a result
the reviewers suggested that the use of toxic effects measured in the laboratory is justified,
particularly developmental effects. In other words, laboratory-to-field extrapolations should be
conserved. The group expressed concern about whether frank toxicity is the most appropriate
endpoint, but acknowledged that frank effects are the best known.
A couple of reviewers thought that the dose-response relationships were adequately
treated, the choice of a NOAEL and LOAEL was acceptable, and the limited toxicity data were
used in an appropriate manner to develop the NOAELs and LOAELs used in the assessment.
Some discussion ensued concerning the utility of toxicity data from laboratory studies on other
animals (e.g., domestic animals and birds); these data might be used to at least help define the
range of toxic exposure concentrations. The assessment needs to clarify the use of the wildlife
criteria values developed in an approach paralleling human health risks (i.e., protection of
individuals) for protecting populations of the selected wildlife species.
There was considerable discussion and concern regarding the validity of the overall
conceptual model for the ecological assessment. This relates in part to the consideration of the
complex chemistry of mercury in surface waters, where different physicochemical factors might
determine exposure. Reviewers noted that lakes located side by side might show very different
concentrations of mercury in fish. This multifactor complexity calls into question the linearity
implied in the current approach for developing the BAF and wildlife criteria values. The
concern is particularly important given the national scope of the intended assessment.
The reviewers noted the need to better articulate the uncertainty regarding the BAFs and
the selection of the mean value. They also felt the report needed better discussions of
distributions and of the nature of the uncertainty analysis.
22
-------
4.2.3 Volume VI (Risk Characterization)
The effects breakout group's primary concern regarding Volume VI was its lack of
emphasis on risk integration. Volume VI mainly reiterates and summarizes the material
presented in the first five volumes. The reviewers were disappointed to find that the wildlife
criteria values developed in Volume V were not carried directly through to the risk
characterization. Substituting fish tissue mercury concentrations that are consistent with the
wildlife criteria values is acceptable as long as the authors can clearly explain in the report why
this was done. Nevertheless, the tissue concentrations (or, preferably, the wildlife criteria),
should be developed as distributions, not single values. These distributions should be compared
with distributions of expected mercury exposures on a regional basis for each of the selected
piscivores. Such comparisons, which are more consistent with a probabilistic framework for
ecological risk, will quickly identify species and regions of concern. They also will highlight
where current information on exposure or toxic endpoints is insufficient to develop distributions
that are precise enough for an assessment. Methods such as sensitivity and uncertainty analysis
can then be used to examine the variance underlying such imprecision to pinpoint the major
factors (e.g., model structure, model parameters) contributing to the overall uncertainty.
Identifying the sources of uncertainty is important to promote efficient and effective allocation of
limited resources and to improve precision, reduce bias, and refine the overall ecological
assessment.
Reviewers felt the risk characterization might also address the risks posed by mercury to
production dynamics at lower trophic levels. Clearly, such impacts have a profound effect on fish
production that is independent of the direct accumulation and toxic effects on fish. These
indirect effects are also relevant for assessing human and piscivore exposure to contaminated
fish—fewer, smaller fish translates into reduced exposure, or at least a greater effort to obtain
fish and, thus, significant mercury exposure if a larger number of smaller fish are consumed.
The group also expressed concern regarding the report's nearly total reliance on
unverified models to produce the risk assessment. Where possible, the models that provided
estimates of regional deposition and exposure should be evaluated in relation to known mercury
concentrations. Any efforts at "ground-truthing" either the exposure or the toxicity models
23
-------
should be pursued within the resource and time constraints imposed by the overall schedule for
delivering the report.
24
-------
5. OVERVIEW OF REVIEWER DISCUSSION IN THE
PLENARY SESSION—Paul Mushak, Ph.D.
5.1 VOLUME VI (RISK CHARACTERIZATION)
Panelists noted that a considerable portion of Volume VI consisted of summaries of
Volumes II, III, IV, and V. These summaries covered human and wildlife health effects, overlay
maps of sensitive wildlife populations with predicted high mercury depositions, and the
uncertainties and assumptions in modeling emissions. Volume VI then provided a relative
exposure ranking, a relative dose-response ranking, and levels of methylmercury in fish tissue that
would be of concern for fish eaters.
The panel found the summaries to be confused and lacking; they failed to provide a
comprehensive or quantitative discussion of the uncertainties and assumptions, and they did not
discuss the extent and magnitude of the harmful exposures. Insufficient attention was given to
linkages between anthropogenic emissions and background mercury data with the risk
characterization.
One reviewer suggested that an ecological risk assessment be performed by using
distributions of the parameters used to develop Tables 4-3 and 4-4 of Volume VI. Reviewers
were impressed with the uncertainty analysis for the human RfD value found in Volume IV,
Appendix C, and were interested in a discussion of propagated uncertainties.
The methodology and results in the comparative risk chapter of Volume VI were major
areas of concern. Reviewers pointed out that the NOAELs and LOAELs are not based on the
same set of endpoints and, therefore, are not directly comparable; in fact, the NOAELs and
LOAELs may reflect a wide range of adverse responses. Another important concern was that
the human NOAEL did not account for uncertainty areas such as different sensitivities. This
indicates that use of the RfD would be more appropriate.
25
-------
Regarding the wildlife criteria, reviewers felt that use of the published rat and monkey
dose-response data would potentially capture more subtle effects in the rat. Notwithstanding the
problems, information is available to enhance the accuracy of the criteria.
Reviewers offered several caveats regarding the strength of the linkages between point
source emissions of mercury and increased levels of methylmercury in fish. Reviewers agreed
there is no doubt that fish in certain areas exceed advisory limits. One reviewer claimed that all
the conclusions in Volume VI are based on models rather than actual data. The volume would
benefit from a discussion of the pathways for which there are claimed to be no data. Reviewers
discussed the extent to which the report went beyond actual data, but did not come to a clear
consensus.
In terms of fish consumption rates, reviewers felt the estimates of the distribution of such
intakes should be improved.
Reviewers agreed that there is a significant need for systematic collection of data on
increased levels of methylmercury in exposed wildlife populations.
In the aggregate, the discussion clearly indicated a need to better integrate the exposure
and health effects data—for example, by comparing distributions of fish mercury levels with
distributions of wildlife criteria. Some reviewers argued that background (baseline)
determinations were needed to better determine increases over time. The panel also suggested
that the RfD be clearly defined as "interim" and that it be revisited periodically as new data
become available. Panelists also questioned the validity of comparing a human NOAEL to overt
toxicity-based guidelines in wildlife, and why an RfD was not used.
Several comments concerned specific chapters in Volume VI. Deposition rates drive the
overall analysis, and field verification is desirable. With reference to this, the exposure breakout
group chair reemphasized that very recent data document the linkage between anthropogenic
mercury emissions and deposition (e.g., the existence of a gradient with distance). Also,
reviewers agreed that the report should better characterize seafood consumption, since it elevates
the baseline for mercury exposure to which freshwater mercury intakes are added for the overall
26
-------
risk characterization. In addition, the panel recommended that seafood levels not be called
"background" because some fraction of mercury in seafood is likely to come from anthropogenic
sources.
5.2 VOLUME VII (RISK MANAGEMENT)
Reviewers agreed that Volume VH was generally good, but felt that it emphasized
controls and did not adequately examine pollution prevention options. Pollution prevention
could include banning products containing mercury (e.g., Minnesota's ban on alkaline batteries).
Reviewers also expressed concern about the volume's cost estimates for mercury control. For
example, could the aggregate cost of reducing mercury emissions by half be calculated?
Reviewers thought it economically inaccurate to allocate all the costs of mercury
reduction strictly to mercury, since typical reduction technologies also remove other
contaminants. They suggested that the authors lower the cost estimate for mercury reduction by
distributing reduction costs over all contaminants controlled by the technologies.
The panel felt that the absence in Volume VII of recommended actions and research
needs is a major gap that should be filled. Recommendations could include, for example,
market-based approaches, product reformulations, product bans, and recycling. The European
experience was suggested as a valuable source for information on market-based approaches.
27
-------
-------
6. EDITED TRANSCRIPTS OF THE PANEL DISCUSSION
ON VOLUME VI (RISK CHARACTERIZATION)
LEONARD LEVIN:
There is no doubt that concentrations of mercury in fish found in natural and manmade
waterways exceed state and federal advisory levels, and that mercury is emitted from several
anthropogenic sources. There are data on fish concentrations of mercury throughout the United
States. There are limited data on mercury deposition rates in the United States and worldwide.
There are good data on emissions from some anthropogenic sources, but there are no good data
on background terrestrial emissions from biogenic and paleogenic sources. EPA chose to model
those anthropogenic sources that could be quantified, but used models that have not been peer-
reviewed or tested in publicly released field evaluations.
The reader must keep in mind that all the conclusions in Volume VI are based on model
results; none is data based. The report's statement that a plausible link has been demonstrated
between anthropogenic sources and fish concentrations is not adequately supported because it is
based on the results of particular models that use a particular set of assumptions.
The models used in the report are multimedia models that include air and nonair
pathways. We do not know the intermodel variability of multimedia models. Also, while there
are a lot of validation data for air pathways, there essentially are no validation data for nonair
pathways, which are the dominant pathways for mercury exposure.
Field validation for these models will be very difficult to perform. We have searched in
vain for a copollutant for mercury that has the same biogeochemical cycle as mercury (to
simulate the behavior of mercury in the environment) but that can be more easily traced than
mercury, which is often confused because of sample and analytical bias.
Drawing national conclusions from these models at this time provides very little
substantiation. When one looks in detail at the numbers on which the report's conclusions are
29
-------
based, one finds very little or no field substantiation for those numbers that can be compared to
observation. The primary numbers of concern are the mercury deposition rates. The rates
provided in the report for a substantial part of the eastern United States are higher than any
deposition rates that have actually been observed anywhere in the world. This does not mean
that the report's rates are incorrect, but it does mean they cannot yet be substantiated. A
footnote in one of the tables indicates mercury deposition rates of 500 jtg/m2/yr. The report says
this rate is higher than any rate that has ever been measured anywhere, yet this rate has been
used as the basis for the risk assessment.
Without the proper caveats, this report may thwart future mercury research by giving the
impression that we know more about mercury than we really do. Much research is needed; most
of this is field based and, therefore, extremely expensive. Field-based mercury research must be
conducted under national guidance so that all who use the results will understand how well-
founded they are.
While this report is a very good beginning, it goes beyond what can be substantiated by
current data. If the executive summary does not strongly emphasize this and tone down the
statements about plausible links, future mercury research may be thwarted.
A simple sensitivity test of the modeling carried out for this report could be performed in
a short time. For example, one could:
n Use the same methodology to model only the natural sources and see what kind
of fish numbers that produces.
n Conduct one case study involving a real source and report it without identifying
the site. We can help you find sites for case studies.
n Conduct field measurements and use them to verify the modeling results.
Making a comprehensive suggestion in the executive summary for additional research
would itself be a major caveat that the public could understand. The executive summary is
probably the only piece that will be read by 99 percent of the report readers. Whatever is
included in there will drive mercury research.
30
-------
I would like to have seen a risk assessment in Volume VI using fish consumption rates
(for which there are a lot of data) to address the risk to sensitive populations. Instead, Volume
VI basically contained just a single sentence "Don't eat fish that have too much mercury." While
this is all one can say based on what is in the report, it was a let-down.
EDWARD SWAIN:
I see more support for the qualitative deposition pattern produced by the study than
Leonard Levin does. Confirming the pattern is difficult because data have not been collected in
a consistent manner across the country. Data collection is needed; this should be a major
recommendation in the report.
There are some data to support the deposition pattern. For example, Richard Lathrop at
the Wisconsin Department of Natural Resources was the senior author of a study that analyzed
fish data from Ontario, Wisconsin, and possibly Michigan.5 The study controlled for lake
alkalinity. The researchers stratified the data below 300 microequivalents/L. They did see a
longitudinal gradient of increasing fish concentrations from west to east. This is entirely
consistent with the Nater-Grigal transect of soil concentrations from western Minnesota to
Michigan that shows an increasing gradient.6 Both those studies tend to support the pattern of
deposition modeling.
In terms of population-level effects, some data not included in the report support the
overall conclusions of Volume VI. Linda J. Welch at the University of Maine just completed
and defended a master's thesis in December on "Contaminant Burdens and Reproductive Rates
of Bald Eagles Breeding in Maine" that shows a population-level effect on inland eagles and
reproduction levels negatively correlated with mercury both in the eagle and the fish prey. A
5Lathrop, R.C., D.W. Rasmussen, and D.R. Knauer. 1991. Mercury concentrations in
walleyes from Wisconsin (USA) lakes. Water, Air and Soil Pollution 56:295-307.
6Nater, E.A., and D.S. Grigal. 1992. Regional trends in mercury distribution across the
Great Lakes states, north central USA. Nature 358:139-141.
31
-------
study involving necropsy of loons in Minnesota found that 23 percent of the loons analyzed had
mercury liver levels above the level that Jack Ban found to be inhibiting reproduction in his
Ontario study on reproductive rates in loons.7 In Minnesota, we have otter fur concentrations
up to 75 ppm that are consistent with reproductive and neurological effects in the Ontario
study—and Minnesota is not at the high end of the deposition gradient.
We need consistent data collection across the states. Many states are collecting fish
contaminant data, but it is next to impossible to map these data and get even a sense of the
gradient. Data collection needs to be coordinated; maybe E-MAP should coordinate the state
efforts, or at least get involved.
We need a national deposition network for mercury that is just getting started—this is
being sponsored by NAPAP. I think these data will tend to support the model once they are
collected.
STEVEN BARTELL:
My comments will focus on the ecological component of the risk characterization since
that is my area of expertise. What I hoped for in Volume VI was a solid integration of the
exposure assessment with the effects developed in Volumes IV and V. Volume V showed some
very interesting maps of atmospheric deposition patterns for mercury on a national scope and of
the habitat distributions of the species of concern that were selected for the assessment—
primarily various raptor species, the Florida panther, and mink. I anticipated that the risk
characterization volume would return to those maps.
Since the authors went through the effort in the effects section to come up with wildlife
criteria, I wanted to see some sort of comparison in Volume VI, on a regional basis, of the
distribution of wildlife criteria for a species of concern—loon, for example—with the distribution
7Barr, J.S. 1986. Population dynamics of the common loon (Gavia immer) associated with
mercury-contaminated waters in northwestern Ontario. Canadian Wildlife Service Occasional
Paper No. 56, Ottawa, Canada. 23 pp.
32
-------
of mercury concentrations (Figure 1). We could do a similar comparison for the Florida
panther. If we develop this type of map for each species of concern and do our best to
characterize the distribution of wildlife criteria—as long as we recognize and, to a certain extent,
deal with the uncertainties involved—we could do the statistical mechanics to estimate the
probability of an exposure concentration exceeding a particular wildlife concentration (Figure 2).
(Some of the bioaccumulation factors in the report were developed using some field data, so that
was not entirely a modeling exercise.)
This (Figure 2), in my opinion, is what a risk assessment really boils down to. It is
advantageous to express a risk assessment in a probabilistic framework whenever possible. This
is what separates risk assessment from traditional environmental impact assessment.
This approach also affords the opportunity to go back and exploit the uncertainties to
determine where you need better information to better characterize risk. With kingfisher or
mink, for example, it is entirely possible that you might have a very broad distribution for the
wildlife criteria, reflecting the large uncertainties in the bioaccumulation factor and a fairly
narrow distribution of exposure concentrations based either on the modeling results or actual
measurements (Figure 3). In such a situation, it is difficult to assess the risks. This situation
essentially tells you that the wildlife criteria need to be better characterized. By explicitly
bringing these uncertainties forward into the calculations and presentations, you at least know
what the situation is. This approach is much more informative and makes better use of the
available data than the single numbers of mercury concentrations in fish that were presented in
the final tables in Volume VI.
PAMELA SHUBAT:
To build on Steve Bartell's comment, dose-response curves offer a terrific way to discuss
the anticipated health effects. Someone asked: "Shouldn't we be seeing health effects if these
levels are so high?" We have the tools to determine whether we should be seeing health effects;
this would be a relatively simple step, at least with the laboratory and mink data that provide a
dose-response curve.
33
-------
Exposure Wildlife Criteria
Wildlife Criteria
Exposure
Wildlife Criteria
Exposure
Figure 1. Hypothetical Mercury Risks to Loon on a National Scale.
34
-------
Example of Minimal or Zero Risk:
Exposure Wildlife Criteria
Example of High Risk:
Wildlife Criteria Exposure
Exposure=
distribution of
mercury exposure
concentrations for
a particular
species
Distribution of
wildlife criterion for
mercury for a
particular species
Risk is proportional
to the degree of
overlap of the
distributions
Figure 2. An Approach to Risk Assessment.
Exposure=distribution
of mercury exposure
concentrations for a
particular species
Distribution of wildlife
criterion for mercury
for a particular
species
Figure 3. Example of Distribution for Which Risk Assessment Is Difficult.
35
-------
The human work is more difficult. I expected to see a linkage of the emissions data to
the health data. If the report is not going to do this, then it should clearly state why this step
was not taken, so that anyone else who might attempt it in the future can be well advised of the
pitfalls. There obviously are good reasons why this was not done, so let us spell it out.
From the human perspective, background is a very important consideration that we
should be working on. The risk characterization should include information on consumption of
fish, including marine fish, and values of mercury in marine fish. We have good data on tuna;
we should use it.
The risk characterization should better describe the uncertainties and assumptions—for
example, what is lost by using a NOAEL rather than an RfD. We can use a benchmark dose
(BMD) approach in the same way that I mentioned for the animal dose-response curves to talk
about what we might expect to see with high levels of mercury in fish.
LEONARD LEVIN:
Health effects data on overseas populations that are expected to come out in the next
year may play a role in reevaluating and changing the mercury RfD presented in the report.
Even though the RfD is not mentioned in Volume VI, its presence in this comprehensive
document will cause it to stand out in the public mind as the standard dose for mercury risk
assessments that may be carried out in the near future. It is, therefore, very important to
integrate, in a timely fashion, any new data on mercury that emerge in the near future.
Important new data on mercury will be emerging soon after the report is published, so EPA's
effort to examine mercury must not slacken off after the report is published.
STEVEN BARTELL:
I am concerned about the potential impacts of mercury exposure on lower trophic levels.
Given the observed toxicity of mercury to some of the plant and animal planktonic species, I am
36
-------
concerned that impacts at those levels might translate into reduced fish populations at higher
trophic levels. There have been past attempts to look at the implications of mercury impacts on
these lower trophic levels using ecosystem models. While the effort required to implement those
models on a national scale may have been prohibitive, perhaps these models could be applied to
a few specific locations identified as "hot spots" by the kinds of maps I drew a few minutes ago.
PAMELA SHUBAT:
My main concern is that the report make it crystal clear why it is acceptable to take a
human NOAEL for reproductive or developmental effects or paresthesia and compare it to overt
toxicity in wildlife. If the rationale does not look good when it is written down, maybe we should
refine those values for the NOAELS and LOAELs in wildlife. Regarding humans, the report
also should make it crystal clear why it is acceptable to use a NOAEL—which is a 95 percent
lower confidence limit on a 10 percent adverse effect level on a dose-response curve—instead of
anRfD.
LEONARD LEVIN:
I expressed my primary concern earlier about the deposition rate as the variable that
drives the subsequent analyses and that can benefit from extensive field work to verify it.
It is worth conducting a full risk characterization on at least a preliminary basis so that
we could examine the methodology. The actual results may be treated with skepticism because
they will be driven by the models and assumptions.
STEVEN BARTELL:
We certainly can justify to a certain extent using the LOAELs and NOAELs. However,
the problem from an ecological point of view is that, if the probability that exposure will exceed
37
-------
critical thresholds is high, I still do not know the risk. I would be interested in focusing on
exposure or dose-response functions in relationship to the endpoint of concern so that,
depending on the actual magnitude of exposure and dose, I could get some direct estimate of
what the impact actually was rather than just knowing that the wildlife criterion based on the
NOAEL was exceeded. In my mind, ecological risk assessment is really a conditional exposure
or dose-response function and all the uncertainties associated with that. There should be more
of this in the risk characterization.
PAMELA. SHUBAT:
I do not think a defensible risk characterization can be done in one month. If you do it,
it will be criticized. I do think it is quite legitimate to spell out the steps you will be taking or
that you endorse taking to develop that risk characterization so that it can emerge over time.
JOANN HELD:
Currently, Volume VI contains nothing about risk characterization. These are some ideas
for risk information that could be added in one month:
n There are hard data on human exposure to mercury via seafood consumption.
This could be added and compared to the RfD, which I feel also should be in the
risk characterization.
n There also are a lot of freshwater fish consumption data that could be compared
to thresholds such as the NOAEL or RfD.
n It would nice to tease out some distributions from Volume in as a first step. This
would have to be carefully caveated.
n Wet and dry deposition field data are available and could be run through the
IEM.
Casting the RfD on a one-meal-per-week and one-meal-per-month basis would be helpful
for the states because this is how many of the state advisories are given.
38
-------
The wildlife and human data should be in separate tables.
ALAN STERN:
I agree that the contribution from marine seafood should be considered. The little data
we have suggest that this contribution will be substantial. If so, the additional exposure to
mercury from freshwater fish would potentially have greater significance for human health.
However, I have examined and modeled the data for exposure from marine fish, and
there are problems with the data. The data are limited and even the available data have gaps.
The modeling I did suggested that a significant percentage of the women of childbearing age
were above the RfD that we were using at that time, which was 0.07 /*g/kg/day. Though the
current RfD is 30 percent higher, this will not substantially affect this finding.
In my opinion, the currently available data on methylmercury exposure from marine fish
consumption support qualitative, but not quantitative, exposure estimates. Nevertheless, the
report would benefit greatly by talking about exposure estimates from marine fish.
There is good agreement on a mean of exposure of about 3 to 4 /zg/day. The problems
arise in looking at the tails of that exposure. We do not have good data on fish consumption
levels by percentile. The Food and Drug Administration (FDA) has data on the 90th percentile,
but we do not know what the distribution actually looks like. Nonetheless, the 90th percentile
data could be used as a point estimate in modeling to give an idea of methylmercury exposure
for that percentage of the population at the 90th percentile of fish exposure.
Another problem is that we do not know what kinds of fish people are actually eating. In
my opinion, assuming that people eat only tuna fish is too biased. Also, it is not clear to what
extent people who eat significant amounts of marine seafood also eat freshwater fish.
In summary, I recommend that the report acknowledge that there is a significant
background of methylmercury exposure from consumption of marine fish in the fish-eating
39
-------
population in general, to which exposure to freshwater fish that receive mercury from
anthropogenic sources would be added as incremental exposure. Considering exposure in that
way makes the "acceptable" increase in exposure from freshwater fish receiving anthropogenic
sources of mercury much lower than if people received all their exposure from freshwater fish.
Certainly, the report should acknowledge that the table in the risk characterization section could
well be an underestimate from the standpoint of looking at background exposures.
With regard to the issue of RfDs versus NOAELs, I feel they both have value and are
not mutually exclusive. Rather than substituting NOAELs with the RfD, a comparison to RfDs
could be added to the tables in Volume VI.
GERALD KEELER:
Methylmercury in seafood should not be considered "background" because some of it
could come from anthropogenic sources, though we do not yet have data to know how much of
the mercury in seafood and the coastal environment is from anthropogenic sources. If the report
implies that mercury in seafood is not from anthropogenic sources, and if seafood consumption
poses a greater risk than freshwater fish consumption, this could give the impression that
anthropogenic sources are a nonissue with regard to potential human health impacts. The report
needs to be very careful about how this is worded.
Currently, there are insufficient data on the coastal environment with respect to mercury;
this is an important data gap.
ALAN STERN:
I agree that we should look for another word beside "background"—perhaps
"nonmodeled" sources of fish exposure would be better. I also agree that marine fish
consumption could represent a very significant exposure source, especially for populations that
eat noncommercial marine fish—for example, professional and sport fishers.
40
-------
DONALD PORCELLA:
Since most people will read only the report's executive summary, there are some critical
issues that should be addressed in the executive summary. Clearly, there is much doubt about
the exposure numbers given in the exposure section, yet these drive the rest of the assessment,
which is at best a hypothesis for testing, and at worst gives an impression that the problem is
severe and there ought to be observable effects.
This is a very ambitious and complex analysis. Lindquist's Swedish report8 was nowhere
near this ambitious, yet it was very complete and is respected among scientists doing research on
mercury cycling and on the issues and policies regarding mercury. That study covers an area
where mercury emissions are somewhat similar to those in the United States. The numbers in
Lindquist's study certainly should be used to constrain the numbers you are evaluating for the
United States.
I do not agree that we do not have any data. We have lots of good data. This needs to
be reflected in the executive summary, as well as the fact that the analysis so far substantially
overestimates the mercury problem as it exists.
JOHN CICMANEC:
I authored the RfD section. Do you think we will be able to develop a usable product in
the limited time that we have?
8Lindqvist, O. et al. 1991. Mercury in the Swedish environment—Recent research on causes,
consequences and corrective methods. Water, Air and Soil Pollution 55:i-261.
41
-------
LEONARD LEVIN:
I believe you can. I think that most of the numerical work presented in the report as the
core of the analysis should be supplied as appendices to shorter technical overviews of the
problem.
In 1990, the Intergovernmental Panel on Climate Change issued a technical assessment by
experts from over 50 countries on what we know about climate change. That report included a
summary table at the beginning of each chapter. The summary tables listed beliefs about climate
change and the atmosphere dealing with greenhouse gases and ranked those beliefs using a star
system. Four stars indicated absolute certainty from a technical standpoint—in other words, a
global technical consensus. One star indicated things that were still to be proven or thought not
to be true. For example, "Carbon dioxide is increasing in the atmosphere" was given four stars,
while "It will get warmer fast" was given one star.
This report needs something similar. The chapters should focus on what the data tell us.
An appendix could contain a modeling exercise using those data. Structuring the report in this
way and focusing the executive summary on what we do know about mercury—rather than trying
to draw extensive links from source to effect, which I do not think is yet doable
nationally—would be a better approach than the current structure. If you move some of the
pieces around and draft some connective tissue, the report will better serve its original purpose,
which was to present the current state of knowledge on mercury to Congress. Appendices would
allow you to model the data but indicate that you do not know how well these models are
performing because we cannot test them.
JOHN CICMANEC:
You remarked that if we do not do this job well we will hurt mercury research because of
the attention this report will draw. Do you feel that the recommendations you just outlined will
get us to that point?
42
-------
LEONARD LEVIN:
There will still be questions, but I think that discussing what we do not know about
mercury in the executive summary—that is, providing the key research recommendations from
each chapter—would be very useful. I think most of the research recommendations the report
makes are very good and would draw a strong consensus from a broad segment of the research
community. However, a number of those recommendations are specifically model based,
including for example one recommendation that says we do not need to do research on wet
deposition because the model tells us that wet deposition is not a significant player in exposure.
I think that is an unwarranted conclusion and should not be included as a research
recommendation. I think the report should clearly and explicitly differentiate and delineate
where the data are limited from what the model tells us.
What you have attempted on Monte Carlo simulation is excellent. Both the method and
results of Monte Carlo simulations are extremely hard for nonspecialists to understand. You
need to pay very close attention to how those are presented in the report and which of those go
forward to the point analysis at the end.
KATHRYN MAHAFFEY:
While I was at FDA for about 11 years, I struggled with trying to use food consumption
data. In this report, we are trying to put together two sources of data on fish consumption—
marine and freshwater. Fish consumption is driven by people's behavior. Pulling together two
sets of behavioral data is doable but complicated. We have to acknowledge a lot of intrinsic
uncertainty, but I think it is worth trying.
We have had diverse points of view on using the severe wildlife effects, but I sense the
group's feeling is that we should do this and get some ways to construct variability in caloric
intake for different species of wildlife and some estimates on how much mercury would be in fish
consumed by these wildlife species. These things would strengthen the report.
43
-------
Interpreting models obviously is complicated. I think there are some distributional
elements we can bring to the risk characterization chapter from data on methylmercury in fish.
EPA is currently drafting guidance on how to conduct risk characterization, but much of that
guidance is not focused on a situation that is as broadly integrated over multiple species with
multimedia exposures as this report to Congress apparently requires. We understand that we do
have a challenge.
JOHN CICMANEC:
Is comparing wildlife criteria to human RfDs like comparing apples and oranges?
STEVEN BARTELL:
I share that concern. It really is an issue of risk management. If we can say, for
example, that the risk of a certain health effect in humans appears to be greater than the
reproductive impacts on loons, I think that is useful information as long as we know which fruit
we are trying to compare. Perhaps we should be less concerned about trying to bring it all to
some common denominator.
CHRISTOPHER CUBBISON:
The suggestion for including other trophic levels was excellent. We originally limited the
draft report to those species that appear to be most exposed to mercury, but species exposed to
lower concentrations could be more severely affected. Although the data are fairly limited, we
will go back and look at this.
The suggestion of looking at the dose-response curves in laboratory animals was also
excellent. We are gathering dose-response data for wildlife and will also do so for laboratory
animals. The steepness of the dose-response curve will tell us a lot about whether we can
44
-------
compare NOAELs and LOAELs from different levels of severity. Also, it will give us confidence
in making such comparisons. If it varies widely from species to species, that will tell us we have
a real problem. It also will help us quantify our level of uncertainty regarding the wildlife
criteria we have calculated.
We built in some conservatism, but we have chosen, as did the framers of the Great
Lakes Initiative (Gil) who developed that methodology, to be less conservative with wildlife than
with humans on the assumption that, as long as population levels were stable, you could live with
some morbidity and mortality. However, we are confounded by the fact that mercury does have
a lot of effects we really do not understand. I have a couple of studies that indicated that, in
both mink and Japanese quail, exposure to sublethal amounts of mercury has subclinical effects
that are not easily observable that make the animals more subject to cold stress. Cold snaps
decimated populations of these mercury-exposed animals.
We were criticized that we did not evaluate the effects of mercury on ecosystems and
whole populations. But most of you probably know that those data just are not there. It would
be hard to find a population that had not been exposed to mercury to use as a control, and it
would be politically difficult if not impossible to do a study exposing a wildlife population to
mercury. So, for the present, wildlife toxicity data will have to suffice.
45
-------
-------
7. SUMMARY OF OBSERVER COMMENTS
DURING PLENARY SESSIONS
7.1 TOM HEWSON, ENERGY VENTURE ANALYSIS, INC, ARLINGTON, VIRGINIA
Tom Hewson commented concerning the report's emissions estimate of 253 tons in 1990
from anthropogenic sources in the United States (Table ES-3 in Volume II). He felt the report
should do a better job of examining "missing" mercury emission sources, including taconite
productions, global sources beyond the United States, and natural sources. Data on global
sources, in particular, might be useful for estimating the potential effectiveness of any future U.S.
regulation of mercury emissions.
Tom Hewson expressed concern that the report's emissions estimate of about 53 tons for
the utility industry (Table ES-5 in Volume II) is too high. The report states that 48 of the 53
tons comes from coal. Yet, work that Tom Hewson did for the Electric Power Research
Institute (EPRI) suggests that coal itself contains less mercury than is reflected by the report's
estimate. This work is documented in an EPRI report that EPA has access to. The EPRI report
used a different methodology. Also, it looked at coal on a seam basis and, therefore, provided a
more accurate estimate of coal quality, which affects emissions estimates because lower sulfur
and lower pyrite coals tend to have a lower mercury content. The commenter recommended that
the report analyze coal on a seam basis.
Finally, Tom Hewson pointed out that existing sources will be more heavily controlled
and will be switching to lower sulfur coal because of the 1990 Clean Air Act Amendments.
Thus, future emissions may be very different from 1990 emissions.
47
-------
7.2 JONATHAN RISER, DIRECTOR OF WASTE SERVICES PROGRAMS, INTEGRATED
WASTE SERVICES ASSOCIATION (IWSA), WASHINGTON, DC
Jonathan Kiser felt that IWSA, as well as some other stakeholders such as the medical
waste industry and the public sector, were not given sufficient notice and time to comment on
the report. He, therefore, requested that all parties be given 30 to 60 additional days to
comment on the report.
IWSA submitted emissions data to EPA in July 1993 that are not reflected in the report.
These data, from a 1991 IWSA study, show that all municipal waste combustors in the United
States contribute 44 tons of mercury emissions annually, which is less than 1 percent of known
anthropogenic sources worldwide. IWSA urges EPA to incorporate these data throughout the
report.
Many developments since 1991 have further reduced mercury emissions, and these should
be reflected in the final report. These developments include joint industry-EPA test programs to
improve mercury collection efficiency at modern plants; further reductions of mercury in
manufactured products entering the waste stream; and closing or retrofitting facilities with
modern controls in anticipation of mercury regulations as part of the Clean Air Act Maximum
Achievable Control Technology (MACT) standards.
Finally, the report should reflect that multipathway risk assessments of both new and
existing facilities conducted over the past several years show that mercury emissions result in a
health risk to the most exposed individual that is 10 to 100 times lower than the regulatory
threshold. The bottom line is that mercury emissions from waste energy plants pose no significant
health risk.
48
-------
73 ROBERT GOLLETTE, NATIONAL FISHERIES INSTITUTE, INC., ARLINGTON,
VIRGINIA
Robert Collette commented that the report should clearly indicate that the RfD used is
interim and may change in the future based on emerging studies, such as the Seychelle Island
study.
Also, Robert Collette heard from colleagues who attended a meeting in Hot Springs,
Arkansas, that the principal investigator of the Iraqi study, on which the mercury RfD is based,
publicly stated that there were some questions about the appropriateness of using the Iraqi study
as the basis for an RfD.
Robert Collette recommended that the uncertainties and caveats stated in the body of the
report also be included in the executive summary, since many decision-makers will read only the
executive summary.
7.4 EVELINA NORWINSKI, HUNTON & WILLIAMS, WASHINGTON, DC
Hunton & Williams, which represents the Utility Air Regulatory Group, requests that
EPA have a public comment period of at least 30 days to ensure broad input from all
stakeholders. EPA is urged not to reach any conclusions that cannot be supported by sufficient
data and, in particular, to avoid stating unsupported conclusions in the executive summary or the
chapter summaries.
7.5 ARNOLD KUZMACK, OFFICE OF WATER, U.S. EPA
Arnold Kuzmack, who is currently on EPA's mercury task force, felt that the report may
focus too much on trying to quantitatively link a particular source or group of sources to
receptors and effects—an approach that appears to stretch the science and the data. He
suggested an alternative strategy that works with the available data to characterize the risk
49
-------
associated with mercury in the environment. This alternative strategy involves (1) using available
data on mercury levels in freshwater and marine fish and consumption levels of humans and
various wildlife species to estimate exposure; (2) comparing exposure levels to RfDs and other
toxicological levels of concern to ascertain where these thresholds are exceeded; and (3)
determining, at least qualitatively, how much of this total exposure is due to sources that EPA
might control.
7.6 RALPH ROBERSON, RMB CONSULTING & RESEARCH, INC., RALEIGH, NORTH
CAROLINA
Ralph Roberson, who works as a consultant to the utility industry, expressed concern that
the report gives the impression that much is already known about the removal efficiency of
activated carbon. In fact, pilot-scale tests show widely varying removal efficiencies, and no data
are available on efficiencies at full-scale operation. Assumptions about removal efficiency greatly
affect the results of the cost-benefit analysis. Ralph Roberson recommended that the
report—and particularly its executive summary—clearly state where there are uncertainties and
data gaps.
7.7 ROBERT IMHOEF, ENVIRONMENTAL RESEARCH CENTER, AIR QUALITY
BRANCH, TENNESSEE VALLEY AUTHORITY, MUSCLE SHOALS, ALABAMA
Robert Imhoff commented that the models used in the report may fail to give even a
qualitative understanding of deposition patterns because they do not have any background on the
inflow, and they assume that all the ionic mercury is in the gas phase, which is deposited very
rapidly around a given source. This assumption needs to be tested; particulate-phase mercury is
removed much more slowly and may even be transported out of the continental United States.
The levels of paniculate- and gas-phase mercury entering the United States must be carefully
considered, because they may negate the effect of any attempts to control U.S. sources.
Robert Imhoff also commented concerning the control technologies chosen in Volume
VII for utility boilers. The Pisces test program found that wet flue gas desulfurization scrubbers
50
-------
at several plants were very effective in removing mercury. Many such scrubbers will be put into
operation due to the Title IV requirements of the Clean Air Act Amendments. The report
should take into account the resulting mercury reductions. The efficiency of such scrubbers is
influenced by many factors and, therefore, should be considered on a case-by-case basis.
Volume VII assumes that all the mercury from utility boilers is in the ionic phase. This
is not borne out by any test results that the commenter is aware of. The highest result he has
seen is 80 percent, and the efficiency is frequently lower.
51
-------
-------
APPENDIX A
WORKSHOP AGENDA
A-l
-------
-------
United States
Environmental Protection Agency
Environmental Criteria and Assessment Office
Review Workshop on the Mercury
Study [Report to Congress
Andrew W. Breidenbach Environmental Research Center
U.S. Environmental Protection Agency
Cincinnati, OH
January 25-26, 1995
Final Agenda
WEDNESDAY JANUARY 25, 1995
8:30AM Introduction and Welcome
Terry Harvey, Director, Environmental Criteria and Assessment Office (ECAO), U.S. Environmental
Protection Agency (EPA), Cincinnati, OH
8:45AM Overview
Rita Schoeny, Associate Director for Science, ECAO, EPA
9:OOAM Breakout Group Chairs' Summary Presentation of Premeeting Comments
o Exposure Breakout Group
Gerald Keeler, Breakout Group Chair, University of Michigan, Ann Arbor, Ml
o Effects Breakout Group
Steven Bartell, Breakout Group Chair, SENES Oak Ridge, Inc., Oak Ridge, TN
9:30AM Charge to Participants
Paul Mushak, Workshop Chair, PB Associates, Durham, NC
9:45AM Break
10:OOAM Convene Breakout Groups
n Exposure Breakout Group
Discussion of Volume II, Inventory of Anthropogenic Mercury Emissions in the
United States
o Effects Breakout Group
Discussion of Volume V, An Ecological Assessment for Anthropogenic Mercury Emissions
in the United States
12:OOPM Lunch
1:OOPM Reconvene Breakout Groups
n Exposure Breakout Group
Discussion of Volume III, An Assessment of Exposure from Anthropogenic Mercury
Emissions in the United States
n Effects Breakout Group
Discussion of Volume IV, Health Effects of Mercury and Mercury Compounds
Printed on Recycled Paper
-------
WEDNESDAY JANUARY 25, 1995 (continued)
3:30PM Break
3:45PM Chairs' Report on Breakout Group Discussions
Steven Bartell, Effects Breakout Group
Gerald Keeler, Exposure Breakout Group
4:15PM Observer Commentary Period
5:OOPM Workshop Chair's Wrapup
Paul Mushak, Workshop Chair
5:1SPM Adjourn
THURSDAY JANUARY 26, 1995
8:30AM Plenary Session
Paul Mushak
8:45AM Volume VI, Characterization of Human Health and Wildlife Risks From
Anthropogenic Mercury Emissions in the United States
B Overview of Comments
Pamela Shubat, Minnesota Department ofHeakh, Minneapolis, MN
B Discussion Among Panelists
10:15AM Break
10:30AM Volume VII, An Evaluation of Mercury Control Technologies, Costs, and
Regulatory Issues
B Overview of Comments
Edward Swain, Minnesota Pollution Control Agency, St. Paul, MN
B Discussion Among Panelists
11:30AM Observer Commentary Period
12:30PM Summary Comments
Paul Mushak
12:45PM Adjourn
A-4
-------
APPENDIX B
LIST OF REVIEWERS
B-l
-------
-------
United States
Environmental Protection Agency
Environmental Criteria and Assessment Office
Mercury Study
Andrew W. Breidenbach, Environmental Research Center
U.S. Environmental Protection Agency
Cincinnati, OH
January 25-26, 1995
Reviewer List
Thomas D. Atkeson, Ph.D.
• Mercury Coordinator
Florida Department of Environmental
Protection
2600 Blair Stone Road (MS-6540)
Tallahassee, FL 32399-2400
904-921-0884
Fax: 904-922-2843
E-mail: atkesont@dep.state.fl.us
Steven M. Bartell, Ph.D.
Vice President & Director
SENES Oak Ridge. Inc.
102 Donner Drive
Oak Ridge, TN 37830
615-483-6111
Fax:615-481-0060
E-mail: 73362.553@compuserve.com
James P. Butler, Ph.D.
Environmental Health Scientist
University of Chicago
Argonne National Laboratory
9700 South Cass Avenue (EAD/900)
Argonne, IL 60439
708-252-9158
Fax: 708-252-4336
E-mail: jpbutler@anl.gov
Tim Eder*
Regional Executive
Great Lakes Natural Resource Center
National Wildlife Federation (NWF)
for the States of Michigan and Ohio
506 East Liberty Street
Ann Arbor, Ml 48104
313-769-3351
Fax:313-769-1449
William F. Fitzgerald, Ph.D.
Professor
Department of Marine Sciences
University of Connecticut
Avery Point
Groton, CT 06340
203-445-3465
Fax: 203-445-3484
E-mail: wfitzger@uconnvm.uconn.edu
Joann L. Held
Chief, Bureau of Air Quality Evaluation
New Jersey Department of
Environmental Protection & Energy
401 East State Street (CN-027)
Trenton. NJ 08625
609-633-1113
Fax: 609-292-7793
Gerald J. Keeler, Ph.D.
Assistant Professor
Environmental and Industrial Health/
Atmospheric, Oceanic, and Space
Science
The University of Michigan
109 Observatory Street (2518-SPll-l)
Ann Arbor, Ml 48109-2029
313-936-1836
Fax:313-764-9424
E-mail: jerry.keeler@um.ccumich.edu
Leonard Levin, Ph.D.
Manager
Risk Analysis & Issue Integration Section
Environmental Risk Analysis Program
Electric Power Research Institute (EPRI)
3412 Hillview Avenue
Palo Alto, CA 94303
415-855-7929
Fax:415-855-1069
E-mail: llevin@msm.epri.com
Paul Mushak, Ph.D.
Principal
PB Associates
714 Ninth Street
Suite G-3 - Couch Building
Durham, NC 27705
919-286-7193
Fax:919-286-7369
E-mail: 74511,3227@compuserve.com
*This reviewer submitted premeeting comments, but was unable to attend the workshop.
B-3
> Printed on Recycled Paper
-------
Jozef M. Pacyna, Ph.D.
Senior Scientist
Norwegian Institute for Air Research
Institute Street 18
N-2007 Kjeller
Norway
476-389-8155
Fax: 476-389-8050
E-mail: jozef.pacyna@nilv.no
Donald Force I la, Ph.D.
Senior Project Manager
Land and Water Resources Management
Electric Power Research Institute (EPRI)
3412 Hillview Avenue
Palo Alto, CA 94304-1395
415-855-2723
Fax:415-855-1069
E-mail: dporcell@msm.epri.com
Pamela Shubat, Ph.D.
Environmental Toxicologist
Minnesota Department of Health
925 Southeast Delaware Street
P.O. Box 59040
Minneapolis, MN 55459-0040
612-627-5048
Fax: 612-627-5479
E-mail: pamela.shubat@health.state.mn.us
Alan H. Stern, Dr. P.M.
Research Scientist
Division of Science and Research
New Jersey Department of
Environmental Protection & Energy
401 East State Street (CN-409)
Trenton, NJ 08625
609-633-2374
Fax: 609-292-7340
E-mail: ahstern@fidelio.rutgers.edu
Edward B. Swain, Ph.D.
Research Scientist
Acid Deposition Program
Minnesota Pollution Control Agency
520 Lafayette Road
SLPaul, MN 55155-4194
612-296-7800
Fax: 612-297-8701
E-mail: edward.swain@pca^tate.mn.us
M. Anthony Verity, M.D.
Professor of Pathology
(Neuropathology)
Center for Health Sciences
University of California
10833 Le Conte Avenue
Los Angeles, CA 90024-1732
310-825-7230
Fax:310-206-5178
U.S. EPA Representatives
O. Russell Bullock
Meteorologist
Air Resources Laboratory
U.S. Department of Commerce
National Oceanic and
Atmospheric Administration
(MD-80)
Research Triangle Park, NC 27711
919-541-1349
Fax:919-541-1379
John Cicmanec
Veterinary Medical Officer
Environmental Criteria and
Assessment Office
Office of Research and Development
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
(MS-190)
Cincinnati, OH 45268
513-569-7481
Fax:513-569-7916
Christopher Cubbison
Environmental Health Scientist
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7599
Fax:513-569-7916
Stanley Durkee
Office of Research and Development
U.S. Environmental Protection Agency
401 M Street, SW (8105)
Washington, DC 20460
202-260-0300
Fax: 202-260-6932
Terry Harvey
Director
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7531
Fax:513-569-7475
E-mail: harvey.terry@epamail.epa.gov
Beth Hassett-Sipple
Environmental Health Scientist
Office of Air Quality
Planning and Standards
U.S. Environmental Protection Agency
(MD-15)
Research Triangle Park, NC 27711
919-541-5346
Fax:919-541-0824
Martha Keating
Office of Air Quality
Planning and Standards
U.S. Environmental Protection Agency
(MD-15)
Research Triangle Park, NC 27711
919-541-5340
Fax:919-541-0824
Kathryn Mahaffey
Leader, Risk Characterization Team
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Room 117
Cincinnati, OH 45268
513-569-7957
Fax:513-569-7475
John Nichols
Research Toxicologist
Environmental Research Laboratory
U.S. Environmental Protection Agency
6201 Congdon Boulevard
Duluth, MN 55804
218-720-5524
Fax:218-720-5539
B-4
-------
Glenn Rice
Environmental Health Scientist
Environmental Criteria and
Assessment Office
Office of Research and Development
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
(MS-190)
Cincinnati, OH 45268
513-569-7813
Fax:513-569-7916
Rita Schoeny, Ph.D.
Associate Director for Science
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7544
Fax:513-569-7475
E-mail: schoeny.rita@epamail.epa.gov
B-5
-------
-------
APPENDIX C
LIST OF OBSERVERS
c-i
-------
-------
United States
Environmental Protection Agency
Environmental Criteria and Assessment Office
on the Mercury Stody
Review
[Report to Congress
Andrew W. Breidenbach Environmental Research Center
U.S. Environmental Protection Agency
Cincinnati, OH
January 25-26, 1995
Final Observer List
Linda Bergeron
Manager
Air-Water Quality Control Division
Long Island Lighting Company
445 Broadhollow Road
Melville, NY 11747
516-391-6181
Fax:516-391-6500
Eletha Brady-Roberts
Environmental Criteria and
Assessment Office
Environmental Scientist
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati. OH 45268
513-569-7662
Fax:513-569-7916
Robert Bruce
Toxicologist
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
10937 War Admiral Drive
Union. KY 41091
513-569-7569
Miguel Castellanos
Associate Scientist
Dyncorp
26 West Martin Luther King Drive
Cincinnati. OH 45211
513-569-7661
Harlal Choudhury
Toxicologist
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7536
Fax:513-569-7475
John Claypool
Senior Engineer
Capital Environmental
1299 Pennsylvania Avenue, NW
Washington. DC 20004
202-383-7387
Fax:202-383-6610
Robert Collette
Director of Food
Regulatory Affairs
National Fisheries Institute. Inc
1525 Wilson Boulevard
Suite 500
Arlington, VA 22209
703-524-8883
Fax:703-524-4619
Barbara Cook
Secretary
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Room 117
Cincinnati, OH 45268
513-569-7553
Fax:513-569-7475
Kevin Cullather
Regulatory and Policy Analyst
National Rural Electric
Cooperative Association
1806 Massachusetts Avenue, NW
Washington, DC 20036
202-857-9596
Fax:202-857-2152
Bennie Deaton
Senior Process Specialist
Chlor-Alkali and Olefins Division
BF Goodrich
Highway 1523-P.O. Box527
Calvert City. KY 42029
502-395-3445
Fax: 502-395-3208
C-3
) Printed on Recycled Paper
-------
Matthew DeVito
Supervisor, Support Group
Technical Services
CONSOL, Inc
4000 Brownsville Road
Library, PA 15129-9566
412-854-6679
Fax:412-854-6613
Arthur Dungan
Vice President, Safety,
Health, & Environment
Chlorine Institute
2001 LStreet, NW-#506
Washington. DC 20036
202-872-4730
Fax: 202-223-7225
Pamela Frink
Environmental lexicologist
Environmental Science Center
Syracuse Research Corporation
2159 Gilbert Avenue
Cincinnati, OH 45206
513-751-3043
Harry Hackett
Manager
Environmental Management Systems
9299 Tonnegerger
Tecumseh. Ml 49286
517-423-5174
Fax:517-423-5175
Tom Hewson
Principal
Energy Venture Analysis, Inc.
1901 North Moore Street
Suite 1200
Arlington. VA 22209
703-276-8900
Fax: 703-276-9541
John Holsapple
Manager of Environmental Affairs
New York Power Pool
5172 Western Turnpike
Altamont,NY 12009
518-356-6122
Fax:518-356-6208
Seymour Holtzman
Scientist
Department of Applied Science
Biomedical and Environmental
Assessment Group
Brookhaven National Laboratory
Building 490D
Upton, NY 11973
516-282-4992
Fax:516-282-7867
Robert Imhoff
Environmental Scientist
Research Section
Air Quality Branch
Tennessee Valley Authority
217 Chemical Engineering Building
Musde Shoals, AL 35660
205-386-3801
John Jansen
Principal Scientist
Southern Company Services
P.O. Box 2625
Birmingham, AL 35202
205-877-7698
Fax: 205-877-7294
Betty Jensen
Fuel and Environment Manager
Public Service Electric and Gas Company
80ParkPlaza(MC-T16)
Newark, NJ 07101
201-430-6633
Fax: 201-242-3962
Jonathan Kiser
Director, Waste Services Programs
Integrated Waste Services Association
1401 H Street, NW - Suite 220
Washington, DC 20005
202-467-6240
Fax: 202-467-6225
Arnold Kuzmack
Senior Science Advisor
Office of Water
U.S. Environmental Protection Agency
401 M Street, SW (4301)
Washington, DC 20460
202-260-5821
Fax: 202-260-5394
Bruce Lawrence
President
Bethlehem Apparatus Company, Inc.
890 Front Street - P.O. Box Y
Hellertown, PA 18055
610-838-7034
Fax: 610-838-6333
Ned Leonard
Manager of Communications and
Governmental Affairs
Western Fuels Association, Inc.
1625 M Street, NW
Magruder Building
Washington, DC 20036-3264
202-463-6580
Fax: 202-223-8790
Chris Maxwell
Research Associate
Oak Ridge National Laboratory
1060 Commerce Park Drive
Room 259-37 (MS-6480)
Oak Ridge, TN 37830
615-241-5794
Fax:615-241-4283
Frank McDowell
Senior Environmental Project Analyst
PSI Energy
1000 East Main Street
Plainfield, IN 46168
317-838-1749
Fax:317-838-2490
Patricia Murphy
Epidemiologist
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7226
Fax:513-569-7916
Evelina Norwinski
Attorney
Hunton & Williams
2000 Pennsylvania Avenue, NW
Suite 9000
Washington, DC 20006
202-955-1603
Fax: 202-778-2201
C-4
-------
Linda Papa
Chief, Methods Evaluation &
Development Branch
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7587
Fax:513-569-7916
Jacqueline Patterson
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7574
Fax:513-569-7916
William Pitman
Air Quality Specialist
Tennessee Valley Authority
400 West Summit Hill Drive
(WT8C-K)
Knoxville.TN 37902
615-632-6699
Fax:615-632-6855
Ralph Roberson
President
RMB Consulting & Research, Inc.
5400 Glenwood Avenue
Suite G-11
Raleigh, NC 27612
919-510-5102
Fax:919-510-5104
Margaret Round
Program Analyst
NESCAUM
129 Portland Street
Suite 501
Boston, MA 02114
617-367-8540
Fax:617-742-9162
Carolyn Smallwood
Environmental Scientist
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7425
Fax:513-569-7916
Robert Statnick
Project Coordinator
Project Development
CONSOLJnc.
4000 Brownsville Road
Library, PA 15129-9566
412-854-6758
Fax:412-854-6613
Jeff Swartout
lexicologist
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH 45268
513-569-7811
Fax:513-569-7916
Joy Taylor
Environmental Quality Analyst
Michigan Department of
Natural Resources
P.O. Box 30028
Stevens & Mason Building
Lansing, Ml 48910
517-335-6974
Fax:517-335-6993
Linda Knauf Teuschler
Mathematical Statistician
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
(MS-190)
Cincinnati, OH 45268
513-569-7573
Fax:513-569-7916
C-5
-------
-------
APPENDIX D
REVIEWER PREMEETING COMMENTS
D-l
-------
CONTENTS
CHARGE TO WORKSHOP REVIEWERS D-3
WORKSHOP BREAKOUT GROUP ASSIGNMENTS D-9
REVIEWERS' PREMEETING COMMENTS .. D-13
VOLUME II Inventory of Anthropogenic Mercury Emissions
in the United States D-15
Gerald Keeler D-17
JozefPacyna D-25
VOLUME III An Assessment of Exposure from Anthropogenic
Mercury Emissions in the United States D-31
James Butler , D-33
William Fitzgerald D-41
Joann Held D-53
Donald Porcella D-67
VOLUME W Health Effects of Mercury and Mercury Compounds D-85
Paul Mushak D-87
Alan Stern D-103
Anthony Verity D-135
VOLUME V An Ecological Assessment for Anthropogenic Mercury Emissions
in the United States D-143
Thomas Atkeson D-145
Steven Bartell D-153
VOLUME VI Characterization of Human Health and Wildlife Risks from
Anthropogenic Mercury Emissions in the United States D-161
Steven Bartell D-163
Leonard Levin D-169
Pamela Shubat D-181
VOLUME Vn An Evaluation of Mercury Control Technologies,
Costs, and Regulatory Issues D-197
Tim Eder D-199
Edward Swain D-209
D-2
-------
Charge to Workshop Reviewers
D-3
-------
-------
CHARGE TO REVIEWERS
Technical Review Guidelines
Section 112(n)(l)(B) of the Clean Air Act (CAA), as amended in 1990, requires the U.S.
Environmental Protection Agency (EPA) to submit a report on mercury emissions to Congress.
The Mercury Study Report to Congress evaluates the rate and mass of mercury emissions,
determines the health and environmental effects of these emissions, analyzes the technologies
available to control these emissions, and determines the costs of such technologies. The sources
of mercury emissions that must be addressed are electric utility steam generating units, municipal
waste combustion units, and other emission sources, including area sources.
The report does not address natural sources of mercury emissions nor all anthropogenic sources.
Limitations of resources and lack of data precluded addressing all known or suspected
anthropogenic mercury sources; for example, waste sites are not covered in this report.
Although reviewers are welcome to read and comment on the entire seven-volume report, this is
not expected of individual reviewers. Rather, please direct your attention and analysis to the
volume(s) for which you were specifically recognized as having expertise. The following are issues
or questions to be considered in preparing for your review, including the development of
premeeting comments.
All Volumes
Are additional data or analyses available that would have a major impact on the
conclusions presented in any volume of the report?
Are arguments and conclusions presented clearly and in a logical manner?
Do the Research Needs chapters of particular volumes present a program of research
projects that will address uncertainties in the evaluation of mercury impacts?
Volume I. Executive Summary
Does the summary adequately reflect the conclusions of the other volumes?
Is additional information presented in the report that should be added to the summary
for clarity or completeness?
Is the summary sufficiently clear and informative to function as a stand-alone volume, or
does it rely too heavily on familiarity with the report as a whole?
D-5
-------
Volume II. Inventory of Anthropogenic Mercury Emissions in the United States
This volume estimates emissions of mercury from area and point sources and provides
abbreviated process descriptions, control technique options, emission factors, and activity levels
for these sources. If sufficient information is available, locations by city and state are given for
point sources. The information contained in this volume will be useful in identifying source
categories that are major emitters of mercury, in selecting potential candidates for mercury
emission reductions, and in evaluating possible control technologies or materials
substitution/elimination that could be used to achieve these reductions. The emissions data
presented in this volume also serves as input data to EPA's long-range transport model, which
assesses the dispersion of mercury emissions nationwide.
Please critique the emission factors approach used in the inventory.
Are you aware of information for source categories identified as having insufficient data
for evaluation?
Volume III. Exposure Assessment of Mercury
This assessment addresses atmospheric mercury emissions from selected major anthropogenic
combustion and manufacturing sources: municipal waste combustors (MWC), medical waste
incinerations (MWI), coal- and oil-fired utility boilers, chlor-alkali plants (CAP), primary lead
smelters, and primary copper smelters. The exposure assessment draws upon available scientific
information and develops two quantitative analyses: a long-range transport analysis and a local
impact analysis. The long-range transport analysis modeled site-specific, anthropogenic emission
source data to generate mean annual atmospheric mercury concentrations and deposition values
across the continental United States. The RELMAP atmospheric model was used to model
multiple mercury emission sources. The local impact analysis was undertaken to estimate the
impacts of mercury from single emission sources. Model plants were located in hypothetical sites
intended to simulate a wide array of typical U.S. sites. Exposure estimates were then developed
through fate-and-transport modeling for a number of hypothetical human and ecological
receptors located around model plants. The contribution of regional mercury transported from
other sites also was included in the exposure estimates around the single source. The models
used included a modified version of the COMPDEP air dispersion model, called COMPMERC,
and the Indirect Exposure Methodology, which is composed of a series of fate-and-transport
models. Together these models were used to predict mercury exposure in humans through
inhalation, consumption of drinking water, and ingestion of soil, farm products, and fish. These
models also were used to predict mercury exposure in piscivorous birds and mammals through
consumption of fish.
Please critique the conclusions of the exposure modeling. Are the conclusions well
supported by the analyses presented in the text of Volume III?
Is there material in the text of Volume III that would be more appropriately presented in
an appendix?
Please critique methods used and assumptions made for the local impact analysis.
Do the appendices provide necessary supporting information concerning methods
described in the text?
D-6
-------
Volume IV. Health Effects of Mercury and Mercury Compounds
This volume deals with three forms of mercury modeled in the exposure assessment: elemental,
mercuric, and organic (primarily methyl mercury). Volume IV was not intended to be an
exhaustive review, but rather to provide sufficient information for hazard identification and dose-
response assessments for those endpoints for which EPA has established risk assessment
guidelines.
Is the information provided on pharmacokinetics sufficient for evaluating human health
effects associated with mercury?
Please critique the weight-of-evidence categorizations for carcinogenicity, developmental
toxicity, and germ cell mutagenicity. Is the level of detail in the report descriptions in
Volume IV sufficient to permit evaluation of these endpoints?
No quantitative dose-response assessment was conducted on carcinogenicity for inorganic
or methyl mercury. Are the arguments against conducting a quantitative assessment
presented cogently and are they supported by the information given in this volume?
Are the reference doses (RfDs) and reference concentrations (RfCs) properly calculated?
Were the appropriate critical study and endpoint(s) chosen? Were the proper uncertainty
factors and modifying factors used?
Are there any factors modifying mercury toxicity in humans that have not been addressed
in the volume?
Volume V. An Ecological Assessment for Anthropogenic Mercury Emissions in the United States
This volume covers the ecological effects of mercury and mercury compounds. It formulates the
nature and extent of the potential for mercury to affect wildlife and ecosystems. Piscivorous birds
and mammals are thought to be at greatest risk from mercury because mercury bioaccumulation,
which occurs in aquatic ecosystems, often results in high mercury levels in fish flesh. Of the
various forms of mercury in the environment, methyl mercury has the highest potential for
bioaccumulation. Several species of birds (kingfisher, osprey, bald eagle) and mammals (mink,
river otter, Florida panther) are selected for case studies. Regions of concern where high
mercury deposition coincides with acid surface water are defined on maps. Estimated
bioaccumulation factors (BAFs) were used in calculation of wildlife criteria values (WCVs) based
on a methodology developed for the Agency's Great Lakes Initiative (GLI).
Please critique the methods used for generating a trophic level three BAF and a trophic
level four BAF.
Please critiques the methods used for generating an uncertainty analysis.
Were appropriate endpoints and studies selected for generating wildlife RfDs?
Were appropriate assumptions used in developing wildlife water criteria?
Are there other species of concern that should be considered in this volume?
Are there other geographic areas of concern that should be included in this volume?
D-7
-------
Volume VI. Characterization of Human Health and Wildlife Risks From Anthropogenic Mercury
Emissions in the United States
In this volume the human health and wildlife risks are briefly described with an emphasis on
areas of uncertainty, assumptions made, and defaults used. The exposure assessment is described
in the same terms. Summaries of the quantitative uncertainty analyses that were done also are
presented. This volume attempts to characterize and compare the extent of the potential hazard
from mercury emissions to selected wildlife species and human subpopulations.
Are the summaries of human and wildlife risk assessment sufficient for a scientific
critique?
Are there major areas of uncertainty, defaults, or assumptions that were not discussed?
Please critique the uncertainty analyses.
Please critique both the methods and results of the comparative discussion of risk
presented in this volume.
Volume VII. An Evaluation of Mercury Control Technologies. Costs, and Regulatory Issues
This volume deals with mercury controls, including product substitution, process modification,
and flue gas treatment technologies. Also included is an estimate of the costs and impacts of
mercury controls as well as the social costs of mercury pollution. Ongoing state and federal
activities are described.
Are you aware of any quantified benefits of mercury control? Please specify.
Are you aware of data on the efficacy of materials separation programs or other pollution
prevention measures other than that presented in this volume? Please specify.
Please critique the cost analysis presented in this volume.
D-8
-------
Workshop Breakout Group Assignments
D-9
-------
-------
REVIEW WORKSHOP ON THE MERCURY STUDY REPORT TO CONGRESS
BREAKOUT GROUP ASSIGNMENTS
EXPOSURE
EFFECTS
Volume:
No. II Emissions and Trends
No. V Ecological Effects
Volume:
No. Ill Exposure Analysis
No. IV Health Effects
Chairs:
Gerald Keeler
Steven Bartell
Participants:
James Butler
William Fitzgerald
Joann Held
Leonard Levin
Jozef Pacyna
Donald Porcella
Edward Swain
Tom Atkeson
Tim Eder
Paul Mushak
Pam Shubat
Alan Stern
Anthony Verity
D-ll
-------
-------
REVIEWERS' PREMEETING COMMENTS
D-13
-------
-------
Volume II
Inventory of Anthropogenic Mercury Emissions in the United States
D-15
-------
-------
Gerald Keeler
D-17
-------
A REVIEW OF THE MERCURY STUDY REPORT TO CONGRESS
VOLUME n
INVENTORY OF ANTHROPOGENIC MERCURY EMISSIONS IN THE UNITED STATES
Draft Dated December 12,1994
Gerald J. Keeler
The University of Michigan
General Comments
An evaluation of any report must take into account the objectives of the report which are clearly
stated in the Executive Summary. As stated the Clean Air Act (CAA) required the EPA to submit
a study on atmospheric mercury to Congress. The sources of emissions that must be studied
include:
• electric utility steam generating units
• municipal waste combustion units and
• other sources, including area sources
While the report does a good job with the first two categories it is not clear which other sources
were meant to be included and which ones were omitted simply because of the lack of data.
There was an uneven coverage of the sources which left one feeling like there was more unknown
than known, which while possibly true, takes away from the good job that was done on the report
as a whole.
The emissions'data given in the report are primarily to be used for: (1) identifying source
categories that are major emitters of Hg and (2) as input to EPA's long-range transport model.
The comments presented below will then focus on whether the report meets to goal of the
mandate given in the CAA and whether the data compiled are adequate to meet the needs of the
two primary uses of the mercury emissions data. The omission of a Hg emission estimate, for
a source like coke ovens, would primarily be important for the gridded emissions used as input
to the transport model. The issues and questions suggested in the charge to reviewers will be
utilized throughout this written review.
The document discusses many source categories for which detailed discussions of the process was
D-18
-------
given and then the report states that due to the lack of information or the large uncertainty no
emission estimate for that source was provided. Many emissions source categories are discussed
with relatively small annual Hg emissions. Emission factors and data for several potentially
important sources, coke ovens, iron-steel, etc. are available for European sources and could be
used to estimate the US emissions to determine their potential importance. Was a decision to not
use the European data applied to US sources in the report made?
The report does a good job in characterizing the major sources and source types for mercury
emissions to the atmosphere. Information is presented, for many major sources, by geographical
location or by state. The report could be strengthened by the addition of maps showing the actual
location of the point sources for categories like utilities (by fuel type), incinerators (sludge,
municipal), iron-steel industry, coke ovens, and cement production. This would be useful
information for scientists and engineers working on Hg but also for policy makers looking at the
density of Hg sources in certain geographical regions such as the Great Lakes or the New England
Coastal region. Lastly, there is a general lack of information regarding any seasonal or temporal
variations in the emissions by category. While the utilities may have fairly constant emissions
both diurnally and seasonally other source type do not have constant operations. Those which
have multiple steps which take various amounts of time should have time-varying emissions. On
an annual basis these considerations are obviously not important but when one considers that these
data are utilized as input to transport and exposure models this information would be critical.
SPECIFIC COMMENTS
Volume n - Executive Summary
On page ES-1 it states that "natural Emissions are also considered" but in the Charge to
Reviewers it states "that the report does not address natural sources of mercury emissions nor all
anthropogenic sources". Inclusion of natural sources detracts from the report and the 1-page
Chapter 2. Natural Sources of Mercury Emissions is both incomplete and misleading. The
topic of natural sources of mercury is controversial and qualitative at best. Why include estimates
of natural emissions at all in the report? The last sentence of Chapter 2 states "therefore, the
estimates cited above for natural processes must be viewed as uncertain". This statement grossly
simplifies the present lack of understanding and the amount of quantitative data concerning this
category. There are numerous problems with the estimates in the literature which can be
discussed in more detail at the review meeting if the authors of the report would like to keep the
D-19
-------
Chapter in the report. Why not move it to the end with all of the other sources where there is
insufficient data for estimating the annual emissions. The report includes European estimates of
natural emissions but fails to utilize European information for sources that are not well-
characterized in the US??
Having reviewed the L&E document, I was glad to see the footnote on page ES-1 that
the mercury emissions factors are consistent with those presented earlier but that the nationwide
emission estimates may vary slightly between the two documents. It would be appropriate to
qualitatively and quantitatively discuss in the document what the changes are and how using a
baseline year of 1990 and the most recent information are different. The most drastic change
between the two documents is for the utility emissions -coal combustion, which are one of the
largest source categories. The emissions estimates are about a factor of 2 lower in the Mercury
Study Report to Congress. If the difference between the two estimates is due to better emissions
information than some insight on how the data have improved would be helpful. The use of
emission modification factors (EMF) for coal cleaning was only a 21% reduction in the Hg
emissions and could not alone account for the large difference between the L & E estimates and
this documents. Other smaller but still significant differences between the two emissions estimates
from the two reports included electric lamp breakage and chlor-alkali production.
Table ES-1 Sources of Mercury Emissions
The sources listed under natural including oceans, vegetation, and wildfires should all have a
superscript a as was given to soils8 denoting that these may also have an anthropogenic
component.
Add other natural waters (lakes, rivers, streams, etc.) to oceans in the table
I would recommend that you change Table ES-2 Mercury Sources with insufficient information...
to a table of those sources that you have reliable information as it will contain a smaller number
of entries and be positive rather than negative in approach.
On page ES-3 an entire paragraph, almost half a page, is given to natural sources Natural sources
when all of Chapter 2 is only a couple of sentences longer. It is not apparent to me how: wind-
blown dust .got stuck in under natural sources as this category is one which would really be
effected by prior atmospheric deposition.
D-20
-------
The statement that "These sources (referring to volatilization)...represent a relatively constant flux
to the atmosphere is probably inaccurate. Where is data to support this? Volatilization is likely
to be a function of season, time of day, location, temperature, and a host of other ever changing
parameters.
Should the natural emissions section be eliminated or changed in someway to better reflect how
poor the present information is for us to bound the annual estimates?
While it is may be the overall goal to be able to make a pie diagram showing the % of the annual
emissions from the various source categories I am not sure that this can be done properly with the
present Hg emissions data. For example, I am concerned that the utility information is quite new
(1994) and reflects new measurements while the chlor-alkali production data is dated 1991 with
out concern for whether the manufacturing plants are still operating in the same capacity in 1994.
These estimates are then used in the % of total inflating the chlor-alkali impact. The base year
for the L & E was 1990 but the information does not appear to be current for many of the source
types.
INTRODUCTION
Comments pertaining to Table 1-1 Sources of Mercury Emissions
Add the superscript a to oceans, vegetation, and wildfires for the same reason that soils are now
noted this way. All of these "natural" sources types have the same potential for being influenced
by previous anthropogenic emissions.
Under anthropogenic sources they are divided into combustion, manufacturing, and misc. The
»
term combustion may be problematic as it suggests that those sources falling into the other two
anthropogenic point source categories do not involve a combustion process. Perhaps a more clear
division of the sources into the categories could be considered.
What is meant by wood combustion?
Table 1-2 Mercury Sources with Insufficient Information to Estimate Emissions
D-21
-------
Add other natural waters (lakes, rivers, streams, etc.) to oceans
Footnoting Paint Usea this way with the superscript gives me the sense that you are trying to say
not to worry about this source as Hg is no longer used in the US Why include it if this was the
intent or perhaps the footnote should be changed.
1.3 Organization of the Rest of the Document
"Chapter 2 describes the natural sources of mercury emission." Why is an entire chapter devoted
to something that is poorly understood and not the focus of the report?
2. Natural Sources of Mercury Emissions
This is a difficult area to deal with scientifically as is identified in the text. However, the chapter
really does not point out how little reliable, quantitative, scientifically defensible data are available
at this time to be able to even include this in the document. To state that the "estimates cited
above for natural processes must be viewed as uncertain does not, in my opinion, reflect the true
state of knowledge on this topic. The estimates should be viewed as having uncertainties of plus
or minus 100% AT BEST!
The following are comments or questions for specific sections.
3. AREA SOURCES OF MERCURY EMISSIONS
3.1 Mobile Sources
The emissions of Hg from motor vehicles has been investigated recently by Olmez et al. at MIT.
The older work by Pierson et al. from the Allegheny Tunnel did not have Hg as a focus when
they did the experiment in 1977. The data were for particulate Hg emissions only as they were
derived from instrumental neutron activation analysis (INAA) of particulate filters. The EF
estimated looks higher than I would have estimated and perhaps is inappropriate. If Hg is emitted
from tailpipes this source would be important due to the magnitude of mobile sources. Olmez et
al. have measured Hg from non-tail pipe emissions and found the crankcase oil Hg to be more
enriched than residual fuel oil. This should be followed up and more research is needed in better
defining this sources emission. Hg emissions from boats may also be a large source of Hg to
certain waterways and coastal areas. Ambient measurements of Hg in ship exhaust were elevated
more than an order of magnitude higher than background concentrations (Keeler, unpublished
D-22
-------
data).
3.2 Electric Lamp Breakage
The amount of Hg emitted to the atmosphere when mercury-containing lamps are disposed of will
be a function of many factors. One important one is the chemical form of the Hg in the lamp and
the size of the particulate forms in the lamp powder. While the report gives a good treatment of
the life-cycle of lamps I can't believe the lamps actually make it to the land-fill unbroken. The
Hg would be lost long before the glass gets dumped from the refuse truck.
Table 3-5 gives information of crematory locations by state. A subscript b is given denoting that
a number of states are not included but the list does not include all of the omitted states, CA!
In section 3.6 how does the frequency of cremations enter into the calculation of the emission
factor?
A general question that arose when reviewing the section 4.2.1 was whether it really is a best
engineering judgment to assume that because a chor-alkali plant reports no emission in the section
114 questionnaire or in the 1991 TRI that it has no emissions. Is this a policy decision or an
engineering judgment?
Research Needs
The need for additional source testing was identified on page 5-1. In addition to secondary Hg
production and smelting operations I would rank iron-steel production and coke production as very
important source categories for which little or no data exist. The need for Hg speciation is
mentioned for stack testing but it is also critically important for ambient measurements.
Understanding the fate of emissions from all sources could be improved if more information was
available on the forms of Hg emitted and in the plumes of the point sources.
Emission of Hg from natural sources and the re-emission of Hg from terrestrial and water
environments is still poorly understood. Recent improvements in the methods to accurately
measure the flux of gaseous Hg from various surfaces (soil and lakes) could allow us to determine
the relative importance of this source to the overall emissions from anthropogenic sources. The
re-emission of Hg that was previously deposited is an important but largely unknown term in
models of Hg transport and fate. An effort to quantify the surface emission of Hg, in various
forms, under various meteorological conditions and in different geographical locations, would help
D-23
-------
us understand the environmental cycling of Hg and help us understand the impacts of
anthropogenic Hg emissions. Determinations of the flux from natural sources will help us
understand the "real" anthropogenic impacts on the global Hg cycle as well as on the impact of
Hg emissions in the United States.
Improvements in the treatment of the atmospheric chemistry of Hg in the transport and fate
models is needed. The simple models available today are a good starting point for the
development of more sophisticated models which do a better job with the aqueous phase
chemistry. These models will require that emissions data contain more detailed information on
the forms of Hg being emitted and the temporal variability of the emissions, if any.
In conclusion, an enormous amount of work was performed to put together all of the emissions
data for mercury sources in the US. Overall the document is quite good and the information has
been taken largely from the L & E report on mercury which was peer reviewed previously. Much
more could be done to accurately define the amount of mercury entering the atmosphere from
both natural and anthropogenic sources. Technologies exist or could be developed to allow
studies to be performed that would provide the data that is presently lacking and hampering our
ability to understand the fate and impact of Hg in the environment.
D-24
-------
Jozef Pacyna
D-25
-------
COMMENTS ON A REPORT ON "MERCURY STUDY REPORT TO
CONGRESS. VOLUME H: INVENTORY OF ANTHROPOGENIC
MERCURY EMISSIONS IN THE UNITED STATES
Jozef M. Pacyna
Norwegian Institute for Air Research
1. Introduction
Comments presented here discuss four major features of the reviewed emission inventory for
mercury: completeness of emission data, their transparency, comparability, and accuracy.
Completeness of any emission inventory deals with the coverage of emission source categories
presented in the emission survey and a list of physical and chemical forms of the inventoried
compound. Transparency of emission inventory can be discussed with a view to detailed
description of estimation methods, as well as presentation of technological, meteorological,
physical, and chemical conditions at which the emissions have occurred. In the case of the reviewed
report, it is a matter of transparency of emission factors used to calculate emissions. The
comparability of emission data deals with the way of presentation of emission estimates which
should be easy to compare with other emission inventories. This issue is linked with the
transparency of emission data. The accuracy of emission estimates is linked with verification of the
emission estimates. There are various methods which can be used to verify the emission data.
Recently, these methods were reviewed at EPA (see a document on "Concepts for emissions
inventory verification" by J. David Mobley and Mark Saeger of the EPA Office of Air Quality
Planning and Standards, prepared for the UN Economic Commission for Europe (ECE) Task
Force on Emission Inventories).
2. Completeness of the reviewed inventory.
The reviewed document is based on information presented in the previous EPA document
entitled "Locating and Estimating Air Emissions from Sources of Mercury and Mercury
Compounds" (EPA/454/R-93-023). This latter document, called here as L&E report, has already
been reviewed by a number of emission experts (see a Summary of Comments on Draft Report
by Tom Lapp and Dennis Wallace of Midwest Research Institute, prepared for the EPA Office of
Air Quality Planning and Standards on April 1, 1993). Their comments have been used to
improve the final version of the L&E report. Threfore, it can be concluded that a list of source
categories considered in the reviewed report is a complete one with respect to at least major source
categories for mercury emissions. However, the division of these source categories into two major
groups: area sources and point sources is somewhat confusing. For example, combustion of fuels
in commercial and residential boilers to produce heat is considered as an area source in emission
inventories prepared in Europe. Can the geographical location, a parameter describing point
sources, be assigned for these sources?
In contrary, crematories are often considered as point sources.
I like better the division presented in the L&E report. I would like to propose the following major
source category split:
D-26
-------
JozefM. Pacyna
- fossil fuel combustion processes,
- waste disposal,
- industrial production processes (industrial manufacturing processes without combustion),
- application of mercury and its compounds, and
- miscellaneous fugitve and area sources.
There are a few source categories mentioned in the reviewed report, for which insufficient
information was found, mostly on emission factors. In such cases no emission estimate was
given. These categories include:
- hazardous waste incinerators,
- primary Hg production,
- mercury compounds production,
- byproduct coke production,
- refineries, and
- mobile sources.
Obviously, it is difficult to conclude on how important the emissions from the above mentioned
sources can be compared to emissions from other sources, estimated in the report. They could be
as low as within a level of accuracy of the total emission estimates. Even so, there are sources
mentioned in the report for which emissions were calculated to be very small, e.g. batteries and
carbon black production. To be consistent with the estimates for these small sources, as well as to
ensure the completeness of source category list, it can be suggested that an assessment of
emissions from the above listed sources (with insufficient information in the U.S.) can be
approached using information on emission factors obtained from other countries. Very detailed
emission inventory for mercury has just been completed in the United Kingdom. A draft version
of the final report on "Mercury in the UK" has been prepared for the Department of the
Environment by the Environmental Resources Management (ERM). Many of the emission
factors missing the US inventory is presented in the UK inventory. I advise to contact Dr.
Andrew Jackson (ERM, Eaton House, Wallbrook Court, North Hinksey Lane, Oxford OX2 OQS,
UK, fax +44-865-793504) concerning both the applicability and official use of these factors if
further improvement of the US emission inventory for mercuryis foreseen.
Similar advice is given concerning the mercury emission factors for source categories which are
missing in the US emission inventory. These categories include:
- iron and steel production, and
- primary zinc production (are there any pyrometallurgical zinc smelters left in the United States?)
There is only a very limited information in the report on emissions of various physical and
chemical forms of mercury. I understand that this important information was outside the scope of
the project. The species resolution of the mercury emission data can be considered by EPA as the
follow-up activity. Toxicity and the chemical reactivity of various chemical forms of mercury
differ and the users of the reviewed report should be aware of what is the physical and chemical
form of mercury like from a given emission source. Preliminary approaches to present the species
resolution of emission data for atmospheric mercury have been made in Europe (e.g., Axenfeld, F.,
Munch, J., Pacyna, J.M.: 1991, Europaische Test-Emissionsdatenbasis von Quecksilber-
D-27
-------
Jozef M. Pacyna
Komponenten fur Modellrechnungen, Domier Report, Friedrichshafen, Germany). These results
can be applied as a first step in the US inventory.
3. Transparency of the reviewed inventory.
As already mentioned, the estimates presented in the reviewed report are based on emission factors
from the L&E report. Their transparency was discussed before the completion of the final version
of the L&E report. In general, the methods used to elaborate the applied emission factors have been
described clearly and in good degree of details. However, it is still unclear to me how the emission
factors for industrial combustion were used. Is it so that one universal emission factor for coal and
one for residual oil were used homogeneously for the estimation of emissions from industrial
boilers in various industries without considering the type of industry? How the industrial processes
with combustion were treated in the inventory compared to the industrial processes without
combustion? A short explanation should be given on this subject
4. Comparability of the reviewed inventory.
Presentation of emission data in the reviewed report, particularly formulation and presentation of
emission factors allows for easy comparison of the reviewed inventory with relevant data from
other emission surveys. In general, the total quantity of mercury emissions from anthropogenic
sources, estimated in the reviewed report is in a good agreement with the estimates in the L&E
report. Obviously, this similarity is not surprising taking into account the fact that the emission
factors used in both inventories are consistent, as stated in page ES-1. In fact, these factors are
often identical.
The statement on the same page ES-1 that "some of the nationwide emission estimates may vary
slightly between the two documents because this report (the reviewed report - JP) uses the most
recently available data, whereas the emission factor document mentioned above (the L&E report -
JP) is based on a baseline year of 1990" does not seem to be true for the major source of mercury
emissions, namely coal combustion. The emission estimates for this category in the reviewed report
are 2 times lower than the estimates in the L&E report. Some explanation should be given why
these estimates differ so much. I pressume that the major reason for this difference is the
application of the "emission modification factor" (EMF) and 21 % reduction of the emissions from
the utility boilers due to the coal cleaning. However, these two factors alone may not result in such
a big difference between the estimates. Are there also other factors contributing to these difference?
Are there any special reasons why the EMF for electrostatic precipitators (ESPs) differ by a factor
of 3 depends on the type of ESP and fuel?
Two other source categories for which a significant difference was found in emission estimetes
presented in the reviewed report and the L&E report are electrical uses of mercury, particularly
lamp breakage, and chlor-alkali production. Concerning the lamp breakage emission source, the
L&E emission estimates are 9 tonnes of Hg in 1990 while the estimates in the reviewed work are
1.4 tonnes of Hg per year at the beginning of the 1990s.
Mercury emissions data from chlor-alkali facilities in the reported work were obtained from Clean
Air Act section 114 questionnaires. The data reported are from 1991 and indicate emissions of 5.9
tonnes. In the 1990 TRI summary, 17 of the 18 mercury cell facilities reported air emissions of
mercury to be about 8.7 tonnes. The prerated emission data for all 18 facilities are 9.3 tonnes of
mercury. This latter estimate was used in the L&E report. An emission factor calculated on the
D-28
-------
Jozef M. Pacyna.
basis of data in the L&E report is about 2.9 g Hg/ tonne of chlor-alkali produced. This value fits
very wefl the range of emission factors for this source category in Europe, being from 2.0 to 7.0 g
Hg/ tonne. The emission factor calculated on the basis of the data from the reviewed report is much
lower.
I uderstand that the reviewed report is a revised version of the the EPA report on National
Emissions Inventory of Mercury Compounds: Interim Final Report (EPA-453/R-93-048). Is this
correct? If not, the EPA report should be acknowledged in the reviewed work.
The emission estimates can also be compared using information on the quantity of mercury emitted
per person in the United States and in some European countries, particularly those with coal being
the major source of energy. The following results were obtained:
- the United States (the reviewed work): 0.91 g Hg/person x year,
- the United Kingdom (national inventory by ERM): 0.90 g Hg/ person x year,
- Germany (national inventory for Western part): 0.75 g Hg/person x year,
- Poland (national inventory): 0.88 g Hg/ person x year.
The above presented similarity is quite striking. The emission factors for coal combustion in these
countries are quite comparable, although they were elaborated by national emission experts on the
basis of independent tests and chemical mass balance approaches.
The European average emission is about 1.2 g Hg/person x year, and the worldwide average about
0.66 g Hg/ person x year. Only anthropogenic emission of mercury is considered in all of these
comparisons.
5. Accuracy of the estimates.
A ranking of the degree of uncertainty presented in Tables ES 4 through ES 7 is assigned properly.
There is no need for more detailed division of this qualitative ranking at a present stage of the
mercury emission estimates. However, a short description of what is meant by high, medium, and
low degree of uncertainty could be added, e.g. low degree means up to 25 % inaccuracy in
emission estimates, medium between 25 and 50 %, and high above 50 %. This quantitative
accuracy assessment, although imprecise, will be very valuable in describing overall accuracy of
the mercury emission inventory for the United States, as well as very indicative with respect to
elaborating the emission reduction strategies.
6. Final remarks
The reviewed report contains very large body of information on emissions of mercury from
anthropogenic sources in the United States. The emission estimates are quite reliable at least for
major source categories, including combustion processes and waste disposal. Some explanations
rather than revisions are needed with respect to changes introduced in the reviewed report
compared to its basis, the L&E report.
One section which is clearly missing in the reviewed report is on:
D-29
-------
JozefM. Pacyna
- conclusions with respect to the application of the presented estimates in proposing the emission
reduction strategies and modeling the atmospheric transport and environmental migration of
mercury, and
- recommendations on further research which will result in the improvement of the present version
of the inventory.
The policy makers would like to obtain information on emissions of the most toxic forms of the
element The modelers need information on chemical and physical forms of mercury. In addition, a
spatial resolution of the mercury emissions is needed for the modeling purposes. A gridded
distribution of mercury emissions in the United States is clearly missing.
Very little is known about the air/surface exchange of mercury, especially deposition to and
emission from terrestrial systems. Lack of proven field methods for measurements of air/surface
exchange rates of mercury vapor has limited our ability to accurately quantify either the rates of
mercury emission from soils, or its total atmospheric deposition, making any estimates of regional
and global budgets highly uncertain. In addition, significant re-emission of deposited mercury will
seriously complicate our estimates of its tropospheric residence time as well as development of
emission control strategies. Quantifying mercury fluxes to and from the atmosphere will be crucial
to understanding the effects of emissions from man-made sources on the regional and global
atmospheric cycle of the element The quantification of mercury fluxes to the atmosphere from
terrestrial and aquatic environments in the United States can be recommended.
Currently available information on natural sources of mercury in the United States is rather
inconclusive, yet the degree to which geologic and re-emission sources contribute to the total pool
must be understood in order to evaluate man's influence on the cycle of the element The
assessment of mercury emissions from natural sources in the United States should be approached
using at the beginning the common knowledge on the subject, e.g. temperature-driven models of re-
emissioa
D-30
-------
Volume HI
An Assessment of Exposure from Anthropogenic
Mercury Emissions in the United States
D-31
-------
-------
James Butler
D-33
-------
James P. Butler
Comments on Mercury Study Report to Congress. Volume III: An Assessment
of Exposure from Anthropogenic Mercury Emissions in the United States
My comments are organized into two sections. The first part addresses the issues
raised in the Charge to Reviewers, and the second part contains specific comments and
recommendations that were not considered in the charge.
PARTI
Additional Data
The exposure assessment is a very comprehensive document, especially' the
section presenting background information on mercury in the environment and the
discussion of measurement data in environmental media and near emissions sources.
Although the authors note that this was not a critical review of the literature, most of the
relevant publications that I am aware of were cited. One very recent paper (Sorenson,
J. A., et al., 1994, "Regional Patterns of Wet Mercury Deposition," Environ. Sci. Technol.
28, 2025-2032) should be included because it is one of the first studies to clearly
demonstrate significant transport of mercury on a regional basis. These findings are
important to consider in light of the exposure assessment conclusion that mercury
deposition is dominated on the local rather than the regional scale (by factor of 20 within
3 Km of the source).
Organization of Report
Considering the amount of data and complexity of the analysis, the exposure
assessment arguments are presented in a reasonably clear and logical manner. I
recommend moving most of the deposition maps and concentration range plots to an
appendix and replacing with a few summary tables in the actual chapter. In addition,
the analysis of long-range transport and local impacts are generally kept separate; the
report needs to clarify when the results of both analyses are being combined for
assessment purposes.
D-34
-------
James P. Butler
Research Needs
In addition to the research needs identified in Volume 3,1 recommend adding the
following topics:
- There is a pressing need to collect environmental monitoring data for model input
parameters that drive the exposure assessment in order to validate the modeling
approach.
- Concurrent with sampling and analysis of environmental media, direct measurements
of mercury in blood should be obtained in order to estimate background exposures
of methylmercury in the general population. (The issue of background exposures is
discussed further in Part n of my comments.)
- In lieu of a national survey to confirm human body burdens and intakes of mercury,
it may be preferable to conduct a more focused total exposure study in which
(micro)environmental and biological samples are collected at periodic intervals for
selected populations. An example of this type of approach is the Total Human
Environmental Exposure Study (THEES), in which exposure and risk estimates were
derived for another ubiquitous environmental contaminant [Butler, J. P., et al., 1993,
"Assessment of Carcinogenic Risk from Personal Exposure to Benzo(a)pyrene in
THEES," /. Air Waste Manage. Assoc. 43, 970-977].
- Additional data is needed on the current distribution of methylmercury concentrations
in different fish species in order to (a) validate the elevated fish tissue levels that were
modeled, and (b) estimate background methylmercury exposures in sensitive
populations.
- The significance of additional background sources of mercury not considered in this
report (e.g, inorganic mercury from dental amalgams and elemental mercury from
occupational exposures), should be studied to determine their effects on increasing the
body burden of mercury in selected human populations.
D-35
-------
James P. Butler
- The bioavailability of methylmercury from contaminated water, soil, and plants needs
to be further investigated.
Exposure Modeling Conclusions
In the overall conclusions of the exposure assessment, it is stated that "...it is not
possible or advisable to attempt to defend any strong statements regarding the modeling
results..." Earlier it was stated that this is a qualitative study based on quantitative
modeling. Given the substantial uncertainties for many of the key modeling variables,
the reason for including caveats is clear. However, if the modeling results are truly
suspect because of data gaps, either they should not be included or the overall level of
confidence in the exposure modeling results should be estimated and clearly stated.
I agree with the general conclusions regarding a plausible link between Hg
emissions and Hg concentrations in environmental media and freshwater fish. The
analysis also supports the conclusion that significant incremental exposures could occur
via the consumption of freshwater fish. However, I do not feel that the relative
significance of local source contributions versus regional transport of Hg has been
conclusively demonstrated.
With respect to specific conclusions of the local impact analysis, the relative
rankings of source types and the critical variables identified are reasonable. In addition,
the conclusion that the most significant exposure to methylmercury is through fish
ingestion is warranted and serves to confirm earlier suggestions that this was the case.
For future exposure assessments of Hg, it will be possible to focus primarily on fish
ingestion by humans and wildlife. The conclusions about the uncertainty analysis results
also serve to highlight key topics and variables that should be a priority for future
studies (e.g., fish BAF parameter, fish consumption rates, Hg emission speciation
estimate, dry deposition velocity for vapor-phase divalent Hg, etc.).
Local Impact Analysis Assumptions
The compilation of scenario-independent and dependent parameters seems to
D-36
-------
James P. Butler
utilize appropriate distributions, with the values reasonably justified and referenced. The
technical basis for uncertainties with the data are noted and accounted for in developing
the range of values. In addition to fish ingestion rates for local fish consumption, the
range and distribution of ingestion rates for all types of fish and seafood that are
consumed in the general population should be summarized; these data would be used
in an assessment of background exposures, as discussed below.
I agree with the decision to perform fate and transport modeling starting with
stack emissions from anthropogenic sources, as opposed to modeling based on
environmental monitoring data. Although approaches based on direct measurements are
clearly preferable, these data have not yet been systematically collected.
Supporting Information in Appendices
In general, the appendices for the exposure assessment are extensive, provide the
rationale for parameter values, and include the appropriate supporting information for
the modeling performed.
PART II
Background Exposures
Background intake of methylmercury through the consumption of non-local fish
and seafood should be evaluated in the exposure modeling. By not considering
background exposures, the largest source of methylmercury in the diet is omitted for
most of the general population. The calculation of incremental exposures in the report
shows the potential for significant intake levels of methylmercury from a single source
for certain subpopulations. For example, subsistence fishers would usually receive most
methylmercury from locally caught freshwater fish and, therefore, have negligible
background intake from other sources. But for much of the U.S. population,
consumption of commercial seafood and fish is the primary source of methylmercury
intake. To decide whether or not an incremental exposure is acceptable, one needs to
know the current intake levels and existing body burden of mercury for the fish-eating
D-37
-------
James P. Butler
population. It has been calculated that a significant fraction of women of childbearing
age already have an unacceptably high level of methylmercury in their diets based on
estimates of seafood consumption and Hg levels in the U.S. catch [Stern, A. H., 1994, "Re-
evaluation of the Reference Dose for Methylmercury and Assessment of Current
Exposure Levels," Risk Analysis 13, 355-364]. If this is true, any increase in incremental
exposure to methylmercury would have public health implications, at least for specific
sensitive populations. For example, more stringent regulations limiting mercury
emissions could be needed if a segment of the fish-eating population is currently at risk
from background exposures to methylmercury.
Risk Characterization
The characterization of risk in Volume VI is incomplete by not directly comparing
the exposure modeling results with the benchmark dose for oral ingestion of
methylmercury by humans. Limitations in the modeling notwithstanding, the bottom
line of the assessment should be: how do the mercury intake results for the different
scenarios compare to the benchmark dose (RfD) of 1 x KF* mg/kg/day? The results of
the quantitative analyses are extensive, e.g., Appendix G contains over 30 tables of
mercury intakes for different facilities, receptors, and scenarios. By comparing mercury
intakes for different facilities, one can see that incremental exposures exceed the RfD in
a number of instances. A general conclusion of the exposure assessment is that mercury
emissions from major sources may result in significant incremental exposures to humans
and wildlife through the consumption of contaminated freshwater fish. However, by
mainly focusing on NOAELs and LOAELs for piscivorous species in the risk
characterization section, the subpopulations at greatest risk are identified, but the
significance of incremental exposures is not presented quantitatively.
Additional Sources of Mercury
The major anthropogenic sources of atmospheric mercury emissions are evaluated
in the exposure assessment. However, mercury is reported to be a contaminant of
concern at almost half of the 1300 hazardous waste sites on the National Priorities List.
In an assessment of local impacts and uncontrolled releases, these types of hotspots
D-38
-------
James P. Butler
should be included in the analysis. These are locations where mercury concentrations
in environmental media are already measured (or will be), reducing the number of
modeling steps and assumptions that would be necessary. Depending on the proximity
of the site to water bodies, sampling of aquatic biota may have already been conducted.
Similarly, land application of sludge as an increasingly common disposal technique could
result in releases of elevated levels of mercury into the environment. The main concern
at these sites would be transport of mercuric salts in water, soil, and sediments, as
opposed to volatilization of elemental mercury. Environmental concentrations can be
readily determined at these sites and should be factored into the assessment of
cumulative impacts at the local level.
Study Goals
The report is a comprehensive evaluation of health and environmental aspects of
mercury emissions to the atmosphere. In interpreting the results, however, it would
useful to know what was the overall goal of the assessment. Presumably this will be
spelled out in the executive summary (Volume I). In addition to evaluating mercury
concentrations in media and biota and the potential for human and wildlife exposures,
was the goal to support the development of a national emissions standard for mercury?
Or was the purpose to support the need for fish consumption limits for high-risk groups,
or the development of a revised fish consumption rate and reference dose for use in
subsequent assessments? The identification of high-risk subpopulations with
dramatically higher rates of freshwater fish consumption can be an environmental equity
concern, unless specifically addressed using a distribution of fish consumption rates. If
the contribution of local sources is relatively greater than long-range transport of mercury
as indicated in the study, site-specific exposure assessments may be the more appropriate
way to address the health risks of methylmercury exposure to sensitive populations. The
goal of the report needs to be spelled out in order to utilize the results in addressing
some of these issues.
D-39
-------
-------
William Fitzgerald
D-41
-------
W. F. Fitzgerald
Pre-meeting comments on Volume III, "An assessment of exposure from anthropogenic
mercury emissions in the U.S."
William F. Fitzgerald, Department of Marine Sciences, U. of Connecticut, Groton, CT 06340
Overview
There is much published literature on the Hg cycle and the quality of the data has
improved dramatically over the past decade. Nevertheless, critical information is lacking
concerning the emissions, chemical speciation and reactivity of Hg in the environment. This
report stresses, appropriately, that the our understanding of the biogeochemical cycling of Hg and
assessments of the impact from anthropic Hg releases are limited by many "uncertainties" in
current knowledge. For example, the cycling of elemental Hg (Hg°) plays a central role in
dispersing Hg at the Earth's surface. Most of the Hg in the atmosphere is Hg°, yet little is known
about the atmospheric reactions and interactions leading to the oxidation of Hg°, and its removal
in wet and dry deposition. It has become increasingly evident that the in-situ production of Hg°
in aquatic systems, and its water-air transfer are very prominent features of the natural cycle.
Indeed, the production and evasion of Hg° in aqueous systems will affect the synthesis and
bioaccumlation of monomethylmercury (MMHg) on a local to a global basis (Fitzgerald, et al.,
1991; Mason, et al., 1994). The number of papers dealing with the cycling of Hg° is limited, and
its important influence ori the behavior and fate of Hg in nature is based on broad extrapolations.
As this reports confirms, human exposure to monomethylHg (MMHg) comes almost
exclusively from consumption of fish and fish products, and prenatal life is more susceptible to
MMHg-induced brain damage than adults (Fitzgerald and Clarkson, 1991). At present, the
mechanisms associated with in-situ methylation process (processes) in fresh and marine waters are
poorly known and reaction rates show broad variation. Bioaccumulation pathways and factors
are uncertain. Moreover, the role of the watershed (i.e., wetlands) in the synthesis of MMHg and
its export to fresh and coastal waters is very poorly understood. Deposition is critically
dependent on the chemical form of Hg. Yet, there are few data on the physical and chemical
species of Hg emitted from various sources. Other weaknesses in the data base for Hg in the
environment are cited in the report.
D-42
-------
W. F. Fitzgerald
Given the apparent limited state of knowledge for the Hg cycle in nature and the
environmental consequences from human-related emissions of Hg, a comprehensive quantitative
assessment of the relationship between anthropogenic Hg releases to the atmosphere and the
potential exposure to people, wildlife, terrestrial and aqueous systems is not possible. Indeed, it
is emphatically stated hi this report that the exposure assessment is a "qualitative study based
partly on quantitative analyses." This is a substantial but accurate caveat. Accordingly, the
primary merit and usefulness of the assessment is in the establishment of a general framework
outlining the linkages between the releases of Hg to the atmosphere from anthropogenic sources
and the exposure to humans and wildlife to MMHg. The modeling shows the insidiously complex
nature of the biogeochemical cycle of Hg. It reinforces the current suggestions and hypotheses
which indicate that the synthesis of MMHg, its bioaccumulation in aquatic systems, especially
piscivorous fish, and the resultant exposure to humans and wildlife is driven by chemical
reactions and biologically mediated transformations involving ultra-trace amounts of Hg in the
atmosphere and natural waters. The assessment concludes that human-related Hg emissions are
significant. Mobilization of Hg and subsequent deposition will occur over local, regional and
global scales. As expected, near-source contamination appears related to the emissions of ionic
and paniculate forms of Hg while the farther field effects are associated with elemental Hg (Hg°).
From an academic view and practical standpoint, this important exposure assessment
provides a valuable guide for research. Although, the results and conclusions are qualitative, this
extensive and essential modeling effort provides a credible means for evaluating the present sparse
data base, for identifying major gaps, inconsistencies and weaknesses associated with major
aspects of the biogeochemical cycle of Hg at the Earth's surface (e.g., sources, source strengths,
species distribution, fluxes, atmospheric and aquatic reactions, etc.). We have a scientifically
reasonable blueprint to use in designing and conducting experimental research on Hg in the
environment.
Conclusion Section
In general, the conclusions are consistent with the qualitative nature of the report. There
is little new in the general conclusions. Indeed, one doesn't need to conduct an extensive modeling
effort to establish plausible linkages between anthropogenic combustion/industrial sources with
Hg in the environment and MMHg in fish. For example, mercury is a naturally occurring element.
D-43
-------
W. F. Fitzgerald
Thus, the biogeochemical processes and reactions leading to the presence of inorganic Hg
species and organo-Hg forms in the environment, and exposure of humans, and wildlife (at sortie
level) to MMHg bioacccumulating in aqueous systems are natural. Further, and since
anthropogenic activities add Hg at the Earth's surface, there will be a link between human-
related Hg fluxes and the amounts of Hg in the environment. Moreover, the larger the
interferences from anthropic Hg fluxes on the natural Hg cycle, the more significant the impact
and the more obvious the connection.
Mason et al., 1994 use 1000 metric tons (MT) as an estimate of annual world-wide natural
terrestrial Hg emissions. This study establishes 230 MT yr"1 as the overall flux of Hg from U.S.
anthropogenic sources to the atmosphere. This is a very large flux (ca. 23%) relative to the 1000
MT value for the natural Hg fluxes from land. On an average area! basis, U.S. anthropic Hg
emissions are about 29 \ig m"2 yr"1. This estimate is approximately 3 times the mean natural flux
from land @ 9.7 ^ig m"2 yr"1 (excluding the Antarctic). Even if the Mason et al estimate were off
by a factor of 2 or 3, the emissions from the US. represent a significant interference in the
terrestrial cycling of Hg. Given this evidence, one would arrive at the general conclusions in this
report from simple mass balance considerations.
That is, since anthropogenic emissions represent a significant interference within the
natural Hg cycle,
1. There is " a 'plausible' link between emissions from anthropogenic combustion/industrial
sources and Hg concentrations in air, water, sediments, and soil;"
2. There is " a 'plausible' link between emissions from anthropogenic combustion/industrial
sources and methylmercury (CH3*) concentrations in freshwater fish;"
3. "Current levels of emissions from major combustion/industrial sources 'may result' in
significant incremental exposures, above background, to humans and wildlife through the
consumption of contaminated freshwater fish."
Conclusions 4 and 5 [ (p.5-2)] could be inferred from simple mass balance analysis, since
they are intuitive and logical.
4. On a unit weight basis "it is 'likely' that piscivorous birds and mammals have much higher
environmental exposures to Hg than humans through consumption of contaminated fish."
(The conclusion that this situation would also apply to subsistence fishers is less obvious.)
D-44
-------
W. F. Fitzgerald
5. The Mason et al. ( 1994) mass balance pointed to the importance of local Hg deposition.
However, the conclusion "that deposition is dominated on the local rather than regional scale,
with a local source having predicted deposition rates that are 20 times the regional contribution
at receptors 3 km of the source" is model dependent, and appropriately qualified as to its
accuracy.
Significant Omission
Assessment of exposure from anthropogenic Hg emissions, and their effects on marine
environment especially the coastal zone.
An unfortunate weakness in this report is the lack of consideration given to marine
environment, especially the coastal zone. Many of the chemical and biological interactions
affecting Hg in lakes will occur in marine waters. Although the ionic strength, ligands (organic
and inorganic) and organisms differ in fresh and sea waters, much of the biogeochemical
processing and movement of Hg will be similar (Rolfhus and Fitzgerald, 1994). For example,
game fish (piscivorous) from lakes and the oceans often show elevated MMHg levels, while total
Hg concentrations in the water are commonly of the order of one part per trillion (Ing/L). The
bioaccumulation of MMHg is similar in both systems. Moreover, more seafood is consumed in
the U.S. than freshwater fish. While this report indicates that "ocean fish are important source
of mercury exposure (p.2-18)", the treatment of exposure via marine systems is inadequate. Also,
it is suggested (p.2-18) that "Hg levels in freshwater fish appear to be higher than the levels in
saltwater fish." This is probably not a valid observation. Rolfhus and Fitzgerald (1994) estimated
the average Hg concentration in ocean fish at 0.2ppm. This value is comparable to the mean Hg
concentration of 0.15 ppm estimated by J. Wiener and colleagues for the fish stock in Little Rock
Lake, WI ( Fitzgerald, et al., 1991)
It can be readily demonstrated, for example, for a important coastal embayment such as
Long Island Sound (LIS) that greater than 80% of the Hg loadings (from rivers, the atmosphere
and sewage treatment plants) are anthropogenic (Fitzgerald and Vandal, 1994). Further, even a
conservative estimate for current Hg loadings to LIS at 405-520 kg yr"1 may be sufficient to
elevate MMHg levels in piscivorous fish above safe consumption levels. For example, a total Hg
input of 5.5 to 7 [ig m"3 y"1 to LIS is obtained by normalizing the total Hg input to the volume
of LIS and assuming no Hg export by tidal exchange. These values are about 2-3x the annual
D-45
-------
W. F. Fitzgerald
Hg input (ca. 2.7 ug m"3 y"1) to Wisconsin lakes that contain fish with elevated MMHg tissue
concentrations (Watras et al., 1994). If the Hg inputs, as reported by Farrow, et al. (1986),
approach 6710 kg yf1 ( lOx greater than our estimate), then the potential for enhanced
bioaccumulation of MMHg, and possible detrimental ecological effects in LIS would be
exacerbated.
This illustration of a "plausible" and significant connection between emissions from
anthropogenic combustion/industrial sources and methylmercury (CHj4) concentrations in marine
fish is very relevant. Indeed, LIS is in the northeastern corridor of the U.S. As specific conclusion
5.b indicates: "the EPA concludes that the region of the northeastern corridor from New York City
to Maine would experience the highest amount of total annual Hg deposition."
How can coastal regions such as Long Island Sound be neglected in this important report
on exposure to anthropogenic Hg emissions? This omission is especially odd since the
assessment is model-based and qualitative.
Long Range Transport Analysis
How well do the predictions from the Regional Lagrangian Model of Air Pollution
(RELMAP) as adapted for Hg, agree with observations? The agreement with our work hi
northern Wisconsin and coastal Connecticut (Long Island Sound) is surprisingly good.
While this may be serendipitous, it does lend credibility to this exposure assessment. For example,
the following information was taken from the data summaries as shown in the various figures.
Predicted Hg° [Fig. 3-11.
Northern Wisconsin = 2.1-2.2 ng m"3.
Groton, CT = 2.2-2.5 ng m3.
Predicted Hg(p) [Fig. 3-31
Northern Wisconsin = 0.1-0.2 pg m"3.
Groton, CT = 0.02-0.05 pg nV3.
Predicted Hgf++) Wet Deposition [Fig. 3-51.
Northern Wisconsin =1-2 ug m"2 yr"1.
Groton, CT = 5-10 ug m'2 yr'1.
D-46
-------
W. F. Fitzgerald
Predicted Hg(p) Wet Deposition [Fig. 3-61
Northern Wisconsin = 0.05-.l^ig m"2 yr"1.
Groton, CT = >0.02-.05 jig m"2 yr4.
Predicted Total Hg Wet Deposition [Fig. 3-71
Northern Wisconsin = 5-10 ug m"2 yr"1. (Same for alternative emissions simulation)
Groton, CT = 10-20 ^g m"2 yr"1. (Same for alternative emissions simulation)
Our atmospheric Hg results from Groton, CT (Avery Point on Long Island Sound) and
from Northern Wisconsin [CLAMS (Crab Lake Atmospheric Mercury Station )] are
summarized in the following table.
Site
rery Point, CT
:LAMS, wi (5)
Gas-Phase
2.8+1.4 ng/m3
(1)
1.8+0.4 ng/m3
Particle-Phase
62+48 pg/m3
(2)
16 ±9 pg/m3
Precipitation-
Phase
10+5 ng/L
(3)
6 ±2 ng/L
Deposition
16±5 /xg/mVyr
(4)
7±2 /ig/m2/yr
References:
(1) Rolflms, K.R. (1995) Unpublished data
(2) Fitzgerald, W.F, R.P. Mason and G.M. Vandal (1991) Atmospheric cycling and air-water exchange
of mercury over mid-continental lacustrine regions. Water,Air,Soil,Pollut. 56: 745-767.
(3) Fogg, T.R. and W.F. Fitzgerald (1979) Mercury in southern New England rains. J.Geophys. Res.
84.C11: 6987-6989.
(4) Estimated from dry deposition velocity of 0.3 cm/s and water flux of 100 mm/yr.
(5) C.H. Lamborg, W.F. Fitzgerald, G.M. Vandal and K.R. Rolflms (1995) Atmospheric mercury in
northern Wisconsin: sources and species. Water,Air,Soil.Pollut., in press.
It is evident that the model predictions are in good agreement with the observations of
average elemental Hg concentrations and total annual Hg deposition at these two locations.The
major discrepancy between the prediction and observation is in the particulate phase. RELMAP
D-47
-------
W. F. Fitzgerald
underpredicts Hgp concentrations substantially and the distribution is reversed. About 80-160x
more Hgp is found at CLAMS, then predicted, while 1200-3 lOOx more Hgp is present in Groton
than expected. This large discrepancy points to an atmospheric process which is enhancing the
amount of Hg on particles in the atmosphere. This process is not included in the models.
The reactive Hg deposition for northern Wisconsin during 1988/89 and 1989/1990 was
2.5 ± 1.3 ug m'2 yf1 and 7.1 ± 3.5 ug m"2 yf1 , respectively (Fitzgerald, et al., 1994). The
predicted range for Hg(++) wet deposition is 1-2 g m"2 yr"1. This may or nor be a genuine
difference.
Note: In the Executive summary (ES-5) and the Conclusion Section (5-3;4), Total Hg
deposition is given in mg m"2 yr"1, rather than jig m"2 yr"1.
Local Impact Analysis
The results from the local impact analysis are captivating. Although, I am currently
reexamining this section for later comment, the merit of the combined predictions from
COMPMERC and RELMAP is evident. Indeed, this analysis shows strikingly how misleading
air measurements of HgT can be as a tool to infer exposure and environmental impact from
sources such as MWC, MWI, smelters, chlor-alkali plants, utility and industrial boilers. The
predicted increases in the amounts of Hg in air at 2.5P and 25P, (Fig. 4-6&7) present no human
health hazard, and one might conclude that the impact is insignificant. Total Hg deposition (Fig
4-S&9), however, presents quite a different story. This is made evident in the exposure model for
Hg emitted from a hypothetical small municipal waste combustion plant located at a
Humid/Midwest Complex Terrain site. This work as summarized in Table 4-29 shows clearly,
the potential adverse influence from local deposition on Hg levels in soils, surface lake water and
in tier 3 and 4 fish at 2.5km of the source. This simulation provides insight into the linkages
between sources, Hg deposition, and environmental contamination.
While this Hg exposure exercise is stimulating, insightful and thought-provoking, how
realistic is the modelling and how valid are the predictions? Unfortunately, the partial list of
research needs on p6-l&2 attests to the many weaknesses and uncertainties in the simulations and
the assessments.
D-48
-------
W. F. Fitzgerald
Additional comments:
Table 2-1 is inexact, and confusing. Firstly, the total annual Hg emissions for combustion
sources at 196,400 kg/yr doesn't agree with the total obtained from the list of sources (190,100
kg/yr). Secondly, and because the listing of combustion sources is in short tons ( 907kg), it
doesn't agree conveniently with the executive summary which uses metric tons (1000kg = MT).
Thirdly, the annual flux of Hg from medical waste incinerators is listed at 58,800 kg/yr (58.8 MT)
in Table 2-1 and at 57.5 in the Executive Summary (ES). Oil and gas-fired utility boilers yield
is 3.6 MT/yr in the ES, while the value given in Table 2-1 is 2.6 MT/yr. By making the
corrections to fit the ES and adding up the individual sources, a value of 253.7 short tons/yr is
obtained. This flux is equivalent to 230 MT/yr which is the value given on page ES-3.
Conclusion 2.d (ES-4 and 5-3) is confusing. Where does the figure 144 MT come from?
Elemental Hg Cycling:
The Brosset ( 1981) reference in Section 2.3.2 is dated. Additional references that
illustrate the importance of Hg° emissions and provide flux estimates from fresh and marine
waters are the following:
Kim, J.P. and W.F. Fitzgerald. 1986. Sea-air partitioning of mercury in the equatorial Pacific
Ocean. Science 231: 1131-1133.
Kim, J.P. and W.F. Fitzgerald. 1988. Gaseous mercury profiles in the tropical Pacific Ocean.
Geophysical Research Letters 15: 40-43.
Iverfeldt, A. (1988). Mercury in the Norwegian fjord Framvaren. Marine Chem. 23: 441-456.
Fitzgerald, W.F., G.M. Vandal and R.P. Mason. 1991. Atmospheric cycling and air-water
exchange of mercury over mid-continent lacustrine regions. Water. Air and Soil
Pollution 56: 745-767.
Vandal, G.M., W.F. Fitzgerald, C.H. Lamborg and K.R. Rolfhus. 1993. The production and
evasion of elemental mercury in lakes: A study of Pallette Lake, northern Wisconsin,
USA. In Heavy Metals in the Environment. 9th International Conference Volume 2:
297-299. R.J. Allen and J.O. Nraigu, editors. CEP Consultants Ltd., Publishers.
D-49
-------
W. F. Fitzgerald
Fitzgerald, W.F., R.P. Mason, G.M. Vandal and F. Dulac. 1994. Air-water cycling of
mercury in lakes. Ch. 1.3 in Mercury as a Global Pollutant: Towards Integration and
Synthesis. C.J. Watras and J.W. Huckabee, editors. Lewis Press, Boca Raton, FL pp.
203-220.
Mason, R.P., W.F. Fitzgerald and F.M.M. Morel. 1994. Aquatic biogeochemical cycling of
elemental mercury: anthropogenic influence. Geochim.Cosmochim. Acta. 58: 3191-
3198.
Mason, R.P., J. O'Donnell and W.F. Fitzgerald. 1994. Elemental mercury cycling within the
mixed layer of the equatorial Pacific Ocean. Ch. 1.7 in Mercury as a Global Pollutant:
Towards Integration and Synthesis. C.J. Watras and J.W. Huckabee, editors. Lewis
Press, Boca Raton, FL. pp. 83-97.
Chapter 3. Long Range Transport Analysis
The RELMAP simulations rely heavily on the work by Peterson et al (1994), which in
turn relies on the Munthe, Lindqvist, & Iverfeldt studies and hypotheses concerning ozone induced
Hg° oxidation and removal from the atmosphere. The proposed reactions are quite complicated
and include mediation by S03"2, complexing ligands, and soot. Although this approach would seem
reasonable, supporting experimental field data for this reaction scheme are sparse. Indeed many
of the comparisons between field data and predictions presented by Peterson et al (1995) illustrate
the weaknesses in the model rather than providing support and validation. Atmospheric/ aqueous
redox Hg chemistry is without question an area of needed research.
The term "background" is used to describe atmospheric Hg concentrations and deposition
that are not directly associated with an identifiable anthropogenic source. Such terminology
should be qualified because it suggests that current background concentrations and deposition
(e.g., open ocean regions) have not been perturbed by anthropogenic emissions. The Mercury
Atmospheric Processes Workshop Report, (September, 1994), made clear that the present
background atmospheric Hg distribution and deposition had been perturbed significantly by human
activities relative to pre-industrial periods.
D-50
-------
W. F. Fitzgerald
References
Mason, R.P., W.F. Fitzgerald and F.M.M. Morel. 1994. Aquatic biogeochemical cycling of
elemental mercury: anthropogenic influence. Geochim.Cosmochim. Acta. 58: 3191-
3198.
Fitzgerald, W.F., G.M. Vandal and R.P. Mason. 1991. Atmospheric cycling and air-water
exchange of mercury over mid-continent lacustrine regions. Water. Air and Soil
Pollution 56: 745-767.
Fitzgerald, W.F. and T.W. Clarkson. 1991. Mercury and monomethylmercury: present and
future concerns. Environmental Health Perspectives 96: 159-166.
Rolfhus, K. and Fitzgerald, W.F (1994). The influence of atmospheric Hg deposition on the
methylmercury content of marine fish. (Presented at the 3rd International Meeting on
Mercury as a Global Pollutant, Whistler, British Columbia, July, 1994). Water Air Soil
Pollut. (in press).
Farrow, D.G.R, Arnold, F.D, Lombardi, M.L., Main, M.B. and Eichelberger, P.D. (1986). The
National Coastal Pollutant Discharge Inventory: Estimates for Long Island Sound.
Strategic Assessment Branch, Ocean Assessments Division, National Ocean Service,
NOAA, Rockville,MD, 40p.+ appendices.
Watras, C.J., N.S. Bloom, R.J.M. Hudson, S.A. Gherini, R. Munson, S.A. Claas, K.A.
Morrison, J. Hurley, J.G. Wiener, W.F. Fitzgerald, R. Mason, G. Vandal, D. Powell,
R. Rada, L. Rislow, M. Winfrey, J. Elder, D. Krabbenhoft, A. Andren, C. Babiarz,
D.B. Porcella, J.W. Huckabee. 1994. Sources and fates of mercury and
methylmercury in Wisconsin lakes. Ch. 1-12 in Mercury as a Global Pollutant:
Towards Integration and Synthesis. C.J. Watras and J.W. Huckabee, editors. Lewis
Press, Boca Raton, FL., pp. 153-177.
D-51
-------
-------
Joann Held
D-53
-------
Joann L. Held
PREMEETING COMMENTS ON THE MERCURY STUDY REPORT TO CONGRESS
VOLUME HI:
AN ASSESSMENT OF EXPOSURE FROM ANTHROPOGENIC
MERCURY EMISSIONS IN THE UNITED STATES
COMMENTS PREPARED BY:
Joann L. Held
New Jersey Dept. of Environmental Protection
401 E. State St.
Trenton, NJ 08625-0027
January 11,1995
D-54
-------
Joann L. Held
General comments
This modeling exercise has advanced the frontier of mercury (Hg) fate and transport
modeling. It will be helpful to future researchers to be able to use some of the scenarios
and pathways which indicate low exposure (such as fruit and vegetable ingestion)
evaluated in this report and eliminate them from future consideration so that more attention
can be paid to the critical pathways.
Because so many citations were missing from the Reference Section, it was not possible
to ascertain whether the most up-to-date sources and most appropriate studies are being
cited in many parts of this report.
Editorial comments regarding missing references, typographical errors, etc. will be
submitted to Eastern Research Group (ERG) under a separate cover.
EXECUTIVE SUMMARY
p. ES-4, item 3
The numbers in parentheses in the list of relative source contributions should be
explained. Do they represent the amount of divalent Hg deposited by each source
category relative to the total amount of mercury in all forms deposited across the whole
United States?
p. ES-6, item 2a
It seems inappropriate to include three significant figures in the lower ambient
concentration for MWI (i.e. 0.0102) when all other values only have one or two significant
figures.
p. ES-9, item 9
This conclusion should begin with a sentence which gives a general description of the
uncertainty analysis, in order to put the rest of the conclusion in context.
p. ES-9, Observation from the Local Scale Modelling
This section should be deleted since it is identical to conclusion #1 on page ES-6.
1. INTRODUCTION
p. 1-3, Item 1. Mercury Emissions
It is stated that the long range analysis was designed to answer several questions related
to mercury emissions. Aren't these questions addressed by the Volume on Emission
Inventory, rather than by the long range transport analysis?
D-55
-------
Joann L. Held
2. BACKGROUND INFORMATION ON MERCURY
p. 2-2, first incomplete paragraph
The statement that "Methods that accurately and reliably measure the total mercury
concentration in environmental media have been established for some time" is a bit of an
overstatement, since reliable methods using ultra-clean laboratory techniques have only
been in use for a few short years, and unreliable new data are still being produced quite
frequently today.
p. 2-4, Table 2-1
This information on annual Hg emissions from various source categories would be clearer if
there were some notation showing which sources are included in the "Combustion
Sources - Total" category at the top of the table. Also, are any of the listed source
categories included in the "Other sources" category at the bottom of the table?
p. 2-10, Section 2.4.1 Air
It is stated that "Anthropogenic emissions currently account for 50 -75% of the total
annual input to the global atmosphere (Expert panel on Mercury, 1994), so current air
concentrations are 2 - 3 times pre-industrial levels, in agreement with the several fold
increase noted in deposition rates (Swain et al., 1992)." It is not at all clear that the
conclusion regarding pre-industrial levels can be supported by the information offered
about anthropogenic vs global emissions and increases in deposition rates. Further
elaboration is needed to support this statement if it is kept as is.
p. 2-11, Table 2-2
There are dozens of studies on mercury concentrations in air. Why were these 4 studies
selected for inclusion in this table? The basic criteria should be mentioned in the text.
Also, the percentages for MHg and Hg-ll in the Toronto study do not seem right and the
particulate Hg concentration is missing. These values should be checked.
p. 2-11, Table 2-3
There are dozens of studies on mercury concentrations in rain. Why were these 3 studies
selected for inclusion in this table? The basic criteria should be mentioned in the text.
p. 2-13, Table 2-6
There is no text, other than a foot note, that discusses the Ocean Water data in Table 2-6.
This oversight should be corrected.
D-56
-------
Joann L. Held
p. 2-17, first paragraph
The results of the NJDEPE study of Hg in fish tissue which are discussed in this
paragraph do not seem to be consistent with the results of the same study which are
summarized in Table 2-10. In addition, the units are different between the table and the
text and will lead to confusion.
p. 2-24, third paragraph
The statement that the decrease in total mercury was most notable in moss samples at "
more distant sites and that this might be related to uptake and retention of different species
during drying does not make sense. What kind of drying is this? Perhaps a connecting
sentence is missing here.
p. 2-25, third complete paragraph
The small amount of local deposition from the power plant in New Mexico may also be
related to the small amount of rainfall in that part of the country. This should be mentioned
in the last sentence of the paragraph.
3. LONG RANGE TRANSPORT ANALYSIS
This section should include a qualitative discussion of the potential for mercury that is
transported out of the model region to be deposited in the oceans and bioaccumulate in
shell fish and saltwater fish.
For this analysis, RELMAP is run using meteorological data from 1989. There should be a
brief discussion of whether meteorological data from other years could give significantly
different results.
The chemical transformations included in RELMAP are described in Appendix D. There
should also be some mention in this section of their inclusion in the model.
p. 3-1, Objectives
This section is very poorly written. Two of the questions raised in this section should
have been answered in the Volume on Emissions Inventory. Namely, "How much
mercury is emitted to the air annually over the United States?" and "What is the
contribution by source category to the total amount of mercury emitted." The last sentence
of the first paragraph should be expanded into two or more sentences to more clearly
state what type of information was needed to answer the questions about deposition
which are posed here as objectives.
p. 3-4, first complete paragraph and Table 3-2
The reader should be referred to Appendix F for an explanation of how the mercury
speciation was derived for each source type, instead of merely referring to "source
D-57
-------
Joann L. Held
dependent emission speciation estimates by OAQPS."
p. 3-4, last paragraph
It is stated that "the anthropogenic Hg(0) emissions do not greatly increase the existing
[global] Hg(0) concentration." This is not substantiated in this paragraph. Can some
reference values be given here?
p. 3-5, second complete paragraph
This paragraph describes the elemental mercury concentrations in ambient air predicted
by the RELMAP simulation. Do these values include the 2 ng/m3 assumed as
background air concentrations? If not, then the actual expected air concentrations would
be twice as high in the vast portions of the country where Figure 3-2 shows
concentrations between 2 and 3 ng/m3.
4. LOCAL IMPACT ANALYSIS
This section is extremely long and complex and would benefit greatly from some
reorganization. It would be helpful if Section 4 was replaced by three separate sections:
one devoted to the model description, one on model application, and one reporting model
results. The model application section should include the text of Section 4.2.3: Example
of Model Application, and Section 4.3: Uncertainty and Sensitivity Analysis. It is
important that these two sections come before the model results in order to put the model
results into perspective.
It is not clear from the body of the main report how the RELMAP results were incorporated
into COMPMERC. There should be a brief discussion of this relationship in the section on
model application.
p. 4-3, first complete paragraph
A reference should be provided for the washout ratio for divalent mercury. Is this
discussed in detail in one of the appendices? An indication of the magnitude of the
washout ratio would be helpful here.
p. 4-9, et seq.
The term nominal value is used frequently throughout the following subsections. Is this
the same as the term default value which is used in Appendix A? If so, this should be
stated here. Also, the way in which the nominal values were used in the model runs
should be explained.
p. 4-9, Description of Hypothetical Rural Human Exposure Scenarios
This section also discusses the Urban scenarios, so the word urban should be added to
the title. It appears that the sentences of this section have been mixed-up, so that the
D-58
-------
Joann L. Held
urban and rural discussions are interwoven in an incoherent fashion. It also appears that
some sentences may be missing.
The scenarios in this section should be checked for consistency with Table 4-4.
p. 4-9, Description of Hypothetical Human Exposure Scenarios for Individuals Using
Water Bodies Near and Distant from Anthropogenic Emission Sources
The scenarios in this section do not match those in Table 4-4. Either the text or the table
should be corrected. (Note that it is not clear from the Dec. 13 draft which is in error.)
In addition, although the title of this section mentions distant water bodies, these scenarios
are not explicitly discussed here. This discussion should be added.
p. 4-13, paragraphs four and five
These paragraphs discuss bioaccumulation factors (BAF) and bioconcentration factors
(BCF). In paragraph four, BAF is said to relate the mercury in fish tissue to "total uptake
rate from water, food and sediments...." In paragraph five, it is stated that a BAF was
developed based on "total measured mercury (all species) in water." This seems
inconsistent since in paragraph five the BAF is based only on uptake from one media. Did
the authors mean BCF instead? Or is there another factor that was left out of the
discussion which will clarify this?
p. 4-16, first complete paragraph
In addition to placing receptors downwind of the hypothetical source, they are also placed
"at 120 and 240 degrees clockwise of the prevailing wind direction." Why are these
locations of interest? Surely they would be expected to have much smaller annual
ambient air concentrations. Is it expected that there could be significant wet deposition in
one of these directions if it happens to be downwind from an ocean or other large body of
water?
p. 4-17, Table 4-11
Where is the selection of the receptor elevations for the hypothetical sites described?
What is the basis for these numbers? Why is a receptor elevation of 0 meters selected for
the Humid/Midwest/Complex terrain scenario at 120 degrees clockwise of prevailing wind
and 25 km distant? A receptor that is not elevated at all doesn't appear to be in complex
terrain.
p. 4-18, second complete paragraph
The first sentence refers to model plant and other parameters in Table 4-13, but Table 4-
13 does not contain this information. Where is this model plant information summarized?
p. 4-19, Figure 4-1
This sketch of the IEM2 Watershed Modules is not clear. Which part is the lake and which
D-59
-------
Joann L. Held
part is the rest of the watershed?
p. 4-25, second complete paragraph
It would be helpful if the results for fruits, vegetables and meat were at least briefly
mentioned here to see how low they are compared to other media concentrations. The
data should also be summarized in a manner similar to that in Table 4-15.
p. 4-27, fourth complete paragraph
The statement that the aquatic food chain exposure route is the most significant is not
demonstrated in this section. Comparisons of predicted mercury concentrations in various
media and the contribution of each media to overall ingestion should be summarized here.
p. 4-27, fifth complete paragraph and Tables 4-16 through 4-18
These provide a "breakdown of the intake of mercury via inhalation and ingestion for each
exposure scenario and for three model plants." It should be made clear in this section that
the mercury ingestion in the tables is from all media combined.
p. 4-31, first complete paragraph
It is stated that" the assumed equilibrium fraction of methylmercury in [rainwater collected
in cisterns] is 15%" and this leads to"intake of methylmercury via cistern water [to] account
for about 60% of the total methylmercury intake" in some instances. Why is the
methylmercury fraction in cistern water so high? Where is it coming from?
p. 4-32, Summary of Results for Primary Output
What is meant by "primary output" in this section?
p. 4-52, Table 4-21
This table reports the combined results of the COMPMERC AND RELMAP models. It
would be helpful if the related text explained briefly (in one sentence?) how the two
models were combined.
pp. 4-84 & 4-85, Figures 4-20 and 4-21
These two figures, which show the contribution to uncertainty in adult and child intake are
almost completely illegible.
5. OVERALL CONCLUSIONS OF ASSESSMENT
Why aren't the mercury ingestion rates for the various human scenarios reported in this
part of the report? Surely this is what everyone needs to know as the bottom line of the
Exposure Assessment.
D-60
-------
Joann L Held
p. 5-1, first paragraph
This paragraph seems to be saying that the air models used in this exercise are so
unreliable that the entire report is practically invalid. It would be better to address the
uncertainty in the models more explicitly, perhaps indicating in which direction their bias
goes (most likely these models overestimate air concentration).
p. 5-2, after item 5
An additional general conclusion should be added (#6) pointing out the very minor
contribution of vegetable and meat ingestion to overall human exposure to mercury.
p. 5-3, item 3
The numbers in parentheses in the list of relative source contribution should be explained.
Do they represent the amount of divalent Hg deposited by each source category relative
to the total amount of mercury in all forms deposited across the whole United States?
p. 5-3, item 4
The statements in this section relate to the national emission inventory, not to the model
results developed for this volume. The information is of interest, but it does not seem to
be a conclusion of this report and should be moved elsewhere.
pp. 5-5 through 5-7, Specific Conclusions of the Local Impact Analysis
Why are the source categories divided into two groups (Incinerators/combustors and
Other high temperature processes) on these pages? If there is something about this
grouping that is helpful in interpreting the data, perhaps it should be stated at the
beginning of Section 5.3 to assist the reader.
The data in this section are all reported in terms of ranges. What do the upper and lower
ends of the range represent? This information should also be provided at the beginning of
Section 5.3.
Whenever parameters which influence plume height are mentioned in this section, stack
gas temperature should be included.
p. 5-7, 7.n Water
The water concentrations for lead and copper smelters are missing for the 25 Km scenario.
p. 5-8, last paragraph of item 9
The last sentence states that an assumption is made "that the fish within different trophic
levels of a given lake are contaminated with the same concentration of methylmercury." If
this is the case, why did the modelers bother to distinguish between trophic levels 3 and
4 in the parameter justifications in Appendix A?
D-61
-------
Joann L. Held
p. 5-8, item 10
This section, which discusses the uncertainty analysis, should mention the Latin
Hypercube simulation which was done using the parameter distributions described in
Appendix A.
p. 5-9, second complete paragraph
The last sentence mentions that rank correlation coefficients above 0.1 were used to
identify the parameters which contribute to uncertainty in calculated soil concentrations.
Other parameters with lower correlation coefficients are not identified in this section. If this
is the case for the entire uncertainty discussion in item 10, then this should be mentioned
earlier.
6. RESEARCH NEEDS
The following recommendations could be added to this section.
* The use of an Eulerian model for long range transport should be explored to determine
whether this type of model will significantly change model predictions.
* A long range transport model should be used to estimate the deposition of U.S. mercury
emissions in the world oceans.
* The wet and dry deposition parameterization in the local impacts model should be
refined. This would include better estimates of washout ratios and dry deposition
velocities for each relevant form of mercury.
p. 6-1, item 6
The last sentence "Natural emission sources need to be better studied and their impacts
better evaluated." should be a separate recommendation and some detail regarding the
types of studies should be added.
APPENDIX A: PARAMETER JUSTIFICATIONS
The term default value is used regularly in this Appendix. Is this the same as the term
nominal value which is used in Section 4 in the body of the report? If so, this should be
stated here. Also, the way in which the default values, distributions, and ranges were
used in the model runs and other analyses should be explained.
p. A-iii, Distribution Notation
What is the difference between the two Lognormal distributions? They both have the
same notation here.
Point distribution should be added to this list since it is used in this Appendix.
D-62
-------
Joann L. Held
An Introductory paragraph on the development and use of distributions would help to put
this whole appendix into context.
p. A-6, Livestock Consumption of Plants Table
The major headings in this table are not completely clear. For example, what is the
difference between beef and beef liver as it relates to amount of plants consumed? Also,
the eggs category seems to be in the wrong column and sheep is missing altogether.
Why do these headings differ from those on p. A-7 which summarizes soil ingestion by
animals?
p. A-9, first paragraph below the table
A proper citation should be provided for the human food consumption rates that are
discussed in this paragraph. Who has published these data?
p. A-11, Groundwater Ingestion Rate Table
What is a Log* distribution? This should be defined on p. A-iii.
Also what does the notation TEA mean here? (see also p. A-17)
p. A-12, Fish Ingestion Rate Table
How can 0 g/day represent the lower end of the range for a subsistence fisher or the child
of a subsistence fisher? If they rely on fish for food, then does zero mean that they are
not eating? Better descriptions of the distributions are needed in this table.
Also, the Recreational Angler Default Value seems to be wrong (the text indicates it should
be 30 g/day), the distribution for the Recreational Angler is missing, and the General
Population receptor is missing altogether.
p. A-13, second paragraph
How were the children's fish consumption rates derived for the subsistence fisher
scenario?
p. A-25, Air-Plant BCF Table
What kind of distribution is represented by two numbers in brackets (e.g. [12000,24000])?
p. A-28, Animal BTF Table
What does TBD mean in the context of this table?
p. A-29, Table A-8
This table of biotransfer factors is redundant. All of the information is included in the Animal
BTF table on the previous page.
D-63
-------
Joann L. Held
p. A-30, first paragraph
The values in this section are not reflected in the tables provided for this section and
contradict the discussion on the previous page which indicates that other studies were
used to determine the animal biotransfer factors tables.
APPENDIX D: DESCRIPTION OF EXPOSURE MODELS
p. D-2, first incomplete paragraph
The manner in which observed ozoneconcentrations air are incorporated into RELMAP
should be briefly explained here.
p. D-3, first incomplete paragraph
The manner in which the mercury speciation was determined for the base and alternate
case for each point source is an important part of the analysis. This requires a better
reference than "obtained from the EPA Office of Air Quality Planning and Standards." This
should be replaced with a reference to Appendix F and a very brief discussion of the
methods used to determine the mercury speciation.
p. D-3, first paragraph below the table
An ambient atmospheric concentration of elemental mercury of 2 ng/m3 is used in
RELMAP. In which layer of the model is this mercury placed?
p. D-3, Carbon Aerosol Emissions
The two assumptions for estimating soot from minor sources seem inconsistent. Is it
proportional to gasoline combustion or to other fossil fuel combustion?
Also, how were emissions derived for major sources of soot?
p. D-9, Lagrangian Transport and Deposition
The notation (Greek letter sigma) for the pressure-based vertical coordinate should be
defined more completely and this definition should appear earlier in the paragraph, when
the notation is first introduced.
p. D-10, second complete paragraph
The second item in the list of aqueous chemical processes in this paragraph refers to a
reaction with sulfite ions. This reaction is not discussed elsewhere. Is this an error?
p. D-16, Atmospheric Mass Balance for Selected Facilities
This discussion should be moved to Section D.2 since it relates to COMPMERC, not
RELMAP.
D-64
-------
Joann L. Held
p. D-47, Figure D-6
This sketch of the IEM2 Watershed Modules is not clear. Which part is the lake and which
part is the rest of the watershed?
APPENDIX F: DESCRIPTION OF MODEL PLANTS
p. F-1, second paragraph
Mercury transport/deposition rates should not be included in the list of source
characteristics that are important for a risk assessment. This is not a source characteristic;
it is a pollutant characteristic.
Model plant parameters should include emission rate rather than stack concentration.
p. F-5, first paragraph
Several other relevant reports that distinguish between Hg(0) and Hg(2+) from Municipal
Waste Combustors (MWC) are not considered in this section. They include Bergstrom
(1986), Vogg et al. (1986), Schroeder et al. (1991), and Munthe (1992). Based on these
reports, NJDEPE (1993) came to the conclusion that the mercury emissions from MWCs
should be characterized as 30% Hg(0) and 70% Hg(2+) in the generic model. Would
incorporation of these reports (references attached) into the review in this section result in
a different conclusion regarding speciation of MWC emissions?
At least one line appears to be missing from this paragraph (beginning after the fifth line,
which ends: "60 percenf). Perhaps other studies are mentioned in the missing line(s).
p. F-5, Figure F-1: Distribution of Hg in EPA Method 29 Sampling Train
The Y-axis is labeled "Percent." What is this a percentage is of? Is it percent of test
runs?
p. F-9, Table F-2
For Carbon beds, how can the median removal efficiency be known when the range is
unknown? Was this information not reported in the Hartenstein paper?
Also, Coal washing appears to have a range of removal efficiency which begins at -200.
Is this a typo? If not, what does this negative removal efficiency represent? Perhaps it
could be explained in a footnote.
D-65
-------
Joann L. Held
Attachment
References for consideration in Appendix F, Section F.2.1.2: Summary of Available Data
on Emissions and Controls, page F-5, first paragraph - speciation of mercury in MWC
emissions (copies have been forwarded to ERG)
Bergstrom, J. 1986. Mercury Behavior in Flue Gases. Waste Management and Research.
4:57-63.
Munthe J., and McElroy, W.J., 1992. Some aqueous reaction of potential importance in
the atmospheric chemistry of mercury. Atmospheric Environment. 26A: 553-557.
New Jersey Dept. of Environmental Protection and Energy 1993. Final Report on
Municipal solid Waste Incineration, Vol. II: Environmental and Health Issues. Prepared by
the Task Force on Mercury Emissions Standard Setting, July 1993. pp.67-68.
Schroeder, W.H., Yarwood, G. and Niki, H. 1991. Transformation processes involving
mercury species in the atmosphere -- Results from a literature survey. Water, Air and Soil
Pollution. 56:654-666.
Vogg, H., Braun, H., Metzger, M. and Schneider, J. 1986. The Specific Role of Cadmium
and Mercury in Municipal solid Waste Incineration. Waste Management and Research.
4:65-74.
D-66
-------
Donald Porcella
D-67
-------
D. B. Porcella
REVIEW OF USEPA DRAFT REPORT TO CONGRESS ON MERCURY
D. B. Porcella, Ph.D
Electric Power Research Institute, 3412 Hillview Ave. Palo Alto, CA 94304
I. Perspective and Summary
USEPA was directed under Section 112(n)(l)(B) of the Clean Air Act Amendments
of 1990 (CAAA) to submit a Report to Congress on mercury emissions, their health
and environmental effects, and technologies and costs to control such emissions.
USEPA has prepared a draft of the report, and reviewers have been asked to
comment on whether the conclusions of the report are adequately supported,
whether other studies impact the conclusions of the report, whether the
assumptions and approaches are adequate to support the development of accurate
conclusions, whether identified research needs lead to a reduction in the
uncertainties of the assessment, and to comment on the organization and
presentation of the report. The focus of my review was directed at the exposure
assessment (Volume HI).
Mercury has complex behavior because of its potential to exist in many chemical
and physical states. This makes discussion of its behavior and effects difficult, and
causes difficulty in understanding whether there is a need to control mercury
while complicating finding the means to control it. In the process of reviewing my
assignment, Volume M on Exposure, I found that I eventually had to read the
entire report to arrive at conclusions, because within each volume there were
excessive uncertainties, leaps of faith rather than logical progression of thought,
ambiguities, improper references, and incorrect statements. Furthermore, the
references are frequently outdated; I realize that many of the most important
studies are still in progress or planned, but often the authors did not even refer to
well known literature that has been in print for 3-4 years.
Congress assigned USEPA a formidible task, which they have largely failed. As a
final Draft Report to Congress, I expected a much better document. The Draft
Report to Congress consists of 7 volumes, of which the reviewers received 6; the
Executive Summary (Vol. I) will be prepared after the review. Without the overall
January 12,1995
D-68
-------
D. B. Porcella
Executive Summary, I cannot assess the overall conclusions. Of the 6 volumes,
they are logically organized, but vary widely in quality and accuracy. The
emissions volume (Vol. n) is probably the best, utilizing data and a reasonable
approach to characterize emissions; however, no estimate of natural and baseline
emissions were made and several important potential sources were ignored. The
Exposure Assessment (Vol. ffl) depended almost exclusively on models and
assumptions, avoided comparisons with field measurements and had limited
coverage of the information available in the scientific community; assumptions
were always excessively conservative leading to very large extimates of exposure
that exceed field observations. In addition, they did not ground-truth any of their
model results. Health Effects (Vol. IV) was very deficient, depending on old
information gained in the 1970's from an acute poisoning by methylmercury
contaminated grain. The Ecological Assessment (Vol. V) is merely a review, as
there is little information on ecological effects of mercury. The Risk Assessment
(Vol. VI) has many of the same problems as the exposure assessment, and uses
the exposure assessment to begin the analysis of risk. The Control
Technologies/Costs (Vol. VII) is probably the second best section, but still relies
too much on off-the-cuff assumptions; this volume contains the economic
assessment of natural resource values which provides a valuable perspective on
potential economic risks of mercury effects. However, the authors do not consider
whether damages to the economy have occurred nor what the relation to damages
might be.
Specific comments to address these summary comments follow in the rest of my
review. I hope my comments are addressed by the authors, because otherwise the
report will embarrass the generally excellent science of EPA. I am presuming that
the purpose of the review is to help the authors to provide the best science to
Congress, and I will make a strong case that there are some errors in analysis
which bias the results, and furthermore could lead to regressive policy decisions.
Generally, there are four rules that govern whether there is an environmental
problem and what to do about the problem: a) There must be demonstration of an
effect that is significant; 2) There must be sources that contribute to the effect; 3)
The problem should be getting worse; 4) There should be cost-effective means of
controlling the sources of the problem. This Draft Report to Congress does not
provide evidence that any of these rules have been met. Although there are US
January 12,1995
-------
D. B. Porcella
sources which they quantified, they did not consider that other sources (natural
and global) might affect their results. USEPA has not established that there are
any significant health or ecological effects in the US. Furthermore, there is no
evidence that the problem is getting worse. At least in some areas the mercury
deposition is much less (one-third) than deposition in the 1960's. Finally, there are
significant difficulties in controlling some sources.
To arrive at a risk assessment, an estimate of exposure and bioaccumulation of
methylmercury is needed, and this was the subject of Volume HI. Obviously, there
is input to aquatic systems of US anthropogenic mercury and other sources. A
central principle for environmental assessment is that loading (input into a specific
system like a lake or watershed) of a pollutant will control the effect. If the total
loading produces little effect, the problem is small and perhaps negligible. The
relative loading is the input from a specific source relative to the loading from all
sources. If the relative loading is small for a particular source, the effect of
controlling that source will be small and perhaps negligible. If present loading of
non-degradable pollutants is small compared to past loadings, then it is likely that
past loading will govern the problem. Often, sediments are a reservoir for
pollutants in a lake, especially those created during past loadings. Sediment
release of the pollutant often called internal loading, can dominate effects even
when external sources are eliminated. Because past anthropogenic activities with
mercury have contaminated watersheds and lake sediments and most mercury
ends up in these compartments, internal loading seems to dominate mercury
biogeochemistry, and source controls will not have significant effects.
Consequently, the benefits of controlling mercury will be difficult to see.
This review is organized as follows: I. Perspective and Summary, n. Response to
Charge to Reviewers, ffl. General Comments, and IV. Specific Comments.
References cited in the Draft Report to Congress are given only as citations. Supra-
numerals indicate references or sets of references that apply. These are listed at the
end of this review and will be completed later.
January 12,1995
D-70 3
-------
D. B. Porcella
II. Charge To Reviewers Of Volume HI
•
Specific questions were asked of the reviewers:
1. Are conclusions well supported by the analyses?
The conclusions are not well supported by the analyses. The authors correctly
point out major uncertainties in the RELMAP /Local Analysis models - none of the
models have been validated, they do not have documentation for review, nor have
they been published in the open literature. There are large uncertainties in the
parameters, and furthermore, the authors recognize that model predictions of
deposition of mercury from the modeled US anthropogenic sources are almost
treble the highest measured total deposition in the US. How can you draw
conclusions from results obtained from such models? The Expert Panel on
Mercury Atmospheric Processes (EPMAP, 1994) concluded that there remain
major uncertainties in making estimates of the relative contribution of local,
regional, and global deposition estimates.
In addition there are quite a few multiplicative errors caused by conservative
assumptions - not all of which I can estimate - that overestimate the impact of
present-day US mercury emissions. For the RELMAP analysis, the authors use a
different emissions data set than gained from Volume n and listed in Volume HI,
Table 2-1. They did not include medical emissions and overestimated utility
emissions by a factor of about 20 percent. In addition, not all emissions were
included in the emission inventory in Volume n. In Volume n they overestimated
mercury emissions from oil-fired power plants by a factor of 4, and that represents
about a 10 percent overall overestimate of total utility emissions.^ They did not
include natural/baseline emissions which could amount to as much as 50-100
percent additional mercury emitted within the United States. If one accounts for
these multiplicative overestimates, the overestimate was on the order of a factor of
4-5 for utility emissions.
They ignored the historical aspect of mercury emissions. Within the United States
and other OECD (Organization of Economic and Community Development)
countries, there has been a general decline in uses of mercury for industrial and
commercial products since the late 1960-early 1970's. Todays emissions are
January 12,1995
D-71
-------
D. B. Porcella
considerably less than previous emissions^/ 3, and deposition has decreased
substantially since the 1960's, at least in Minnesota, even though today's deposition
still is a factor of three greater than the pre-industrial era^/S. Furthermore, there
was little perspective given to the regional/global sources of mercury including
our Canadian and Mexican neighbors. Despite neglecting these other sources, they
still overestimated deposition by more than a factor of two. If the other sources are
taken into account, I suggest they overestimated deposition by factors of 5-6!
In addition the authors assumed that all of the Hgll emissions from combustion
sources remained as gaseous Hgll and that it deposited as if it were nitric acid. So
far, gaseous Hgll cannot be measured and no one knows yet whether it deposits
like nitric acid. The more likely fate of gaseous Hgll is its incorporation into
particles that deposit at a much slower rate than does nitric acid^. In fact for
certain particle size ranges, rain-scavenging has little effect on particle removal,
and without rainfall, fine particles can be transmitted globally'7. Thus, both
RELMAP and the local models overestimate local and regional Hgll deposition.
The authors overestimated the rate of oxidation of gaseous mercury via
homogeneous reactions^. However, Swedish researchers have described
heterogeneous reactions which can produce potentially more deposition than
estimated in this volume^.
The authors used a background level of 2 ng/m3. This is probably high even in
the Northern Hemisphere. I would use a value about 1.6 ng/m3 as observed in
Wisconsin, Florida, the central north Patific, Michigan and other areas^. However,
the 1.6 ng/m3 contains a small fraction of particulate Hgn that could add
substantially to deposition from regional/global sources^.
If the authors had more carefully ground-truthed their data using recent estimates
of mercury deposition, they would have realized that they were overestimating
deposition and the effects of within-US mercury sources. Instead they carried their
analysis further to calculate mercury accumulation in fish from the supposed
mercury deposition obtained from this overestimate of deposition. In this further
analysis, they assumed that all lake mercury came from atmosperic deposition, the
lake came to a steady-state in 30 years, the bioaccumulation factor (BAF) was on
the order of 350,000 - which is a very conservative assumption -, and then
D_?2 January 12,1995
-------
D. B. Porcella
calculated the fish bioaccumulation. This result was compared to a rather arbitrary
new methylmercury reference dose (RfD, Volume IV). Using the estimated
mercury in water, and the high BAF, they obtained fish concentrations that greatly
exceed almost all measured background fish mercury concentrations measured by
a large variety of surveys". In addition, the only source considered for mercury
uptake was from the point source emission. In such a case, one must expect that
fish concentrations be less than the highest literature values. In reality, many of
the modeled fish concentrations substantially exceeded the highest values^.
Furthermore, given that they overestimated mercury deposition from
combustion/industrial sources and ignored other sources of mercury to surface
waters, including water borne sources like wastewater treatment plants and runoff
from mercury dump sites such as Superfund sites, they overestimated fish
concentrations from emission sources by an incredible amount.
Again, they failed to ground-truth their results. Then, they combined the fish
concentrations with the RfD, to show that they exceeded the human RfD by
factors of 10 or more in some cases. This is despite the fact that several other
published studies showed that - for power plant combustion sources at least - fish
mercury provides only fractions of the RfD^O. Since their results are very different
from earlier studies^, ft is their obligation to show why their numbers are
different. The over-estimate of deposition and the other conservative assumptions
suggests an explanation for the error.
2. With respect to questions about organization, most of the material in Chapter 2
should be put into an appendix. Sections 2.5 and 2.6 should be inserted within
Sections 3 and 4 respectively. Table 2.1 should be placed within Section 3, and the
authors should explain why the RELMAP analysis varies from the data in Table
2.1.
3. Reviewers were requested to provide a specific critique of the local source
impacts' methods and analysis. Specific comments are in my section IV. The n-1
comments apply, also. The overall approach seems okay (given that I am
unfamiliar with the model) and generally agrees with other risk analyses.
However, the selected coefficients are very conservative. They produce extremely
high deposition numbers that exceed published data, and they have never been
verified. For comparison, the Maryland Power Plant Team sponsored a study of
D_?3 January 12,1995
-------
D. B. Porcella
mercury levels in fish from freshwater ponds that showed no measurable effect of
local deposition of mercury.
4. The appendices are relatively complete, and support the information in the
Volume. See below for cases where I disagree with data, etc. However,
assumptions and parameters are very conservative, and not suitable even for a
screening analysis or scoping study.
m. General Comments
1. The authors failed to put US emissions into a global perspective. The total
anthropogenic mercury emissions from global sources are on the order of 2300-
4500 metric tons/year (10^ grams or megagrams). US emissions are 5-10 percent
of the total global emissions as currently estimated. US utility emissions are on the
order of 1-2 percent of global anthropogenic. Global natural emissions have been
estimated as being about 3000 metric tons/year. Thus, total US anthropogenic
emissions are 3-5 percent of total global emissions. This comparison shows how
important non-US sources can be.
2. The authors failed to consider the historical pattern of mercury use in this
country or globally. This has important bearing, since an increasing problem
requires more attention than one that is decreasing. Because of global sources, it
may be that deposition in the US may begin to increase in the near future.
3. The authors have not demonstrated that mercury is currently a problem in the
US. I have sought information on human health effects in the US from chronic
exposures, and in most cases the researchers have concluded that only a few
people have exhibited high levels of mercury and these were due to special
circumstances. One case involved a family that ate sea bass and other fish 4-5
times a week who had elevated mercury levels but no obvious mercury-related
symptoms^.
4. Because the methylmercury RiD is a key part of the overall mercury assessment,
and it ultimately determines the safe intake of mercury by identifying whether or
not there is a potential problem, I read Volume IV on Human Health Effects of the
Draft Report to Congress. There is a great quantity of information there, much of
January 12,1995
D-74
-------
D. B. Porcella
which seems irrelevant. There are errors of fact. But the greatest problem is
reliance on the more than 20 year old data set from the Iraqi methylmercury
poisoning episode^. Because of the role of the RfD, and its apparently arbitrary
development, Kenny Crump, Harvey Clewell, and Annette Shipp of ICF-Kaiser, K.
S. Crump Division, a contractor to EPRI on health effects of mercury were asked to
comment on the report. As background, the reader should look at the review of
human health studies in Table 5-1 of EPRI's Mercury Synthesis Report, which
describes many of the deficiencies and value of the different human health studies
(in EPRI, 1994; Trace Metals Synthesis Report). The following is taken from their
analysis: The USEPA has not relied on the most recent relevant epidemiological
data upon which to base the RfD and has failed to appropriately include recent
work in physiologically based pharmacokinetic modeling of methylmercury and
dose-response analyses of the epidemiological data. The major elements of the
methyl mercury work conducted by ICF Kaiser, K. S. Crump group have been
directed toward assessing the chronic intake level of methylmercury that would be
expected to be without an adverse impact on human health or pose a public health
concern. These major elements have included a reassessment of the dose-response
data for populations exposed to methyl mercury using the New Zealand database,
application of the most recent benchmark methodology to these data,24 and
development of a physiologically-based pharmacokinetic model for methyl
mercury24. This work has been presented at two international workshops and
submitted for publication, and preprints of these publications were provided to the
EPA. The papers show that the calculations in Volume IV are much more
conservative than needed to protect health.
5. In the discussion within all the volumes, there is a general recognition of
uncertainty. In fact they say so much about uncertainty that it appears as if we do
not know much about Hg. In fact we know a great deal. Unfortunately, little of
the most recent knowledge has found its way into this report. The authors have
used very little of the recent information (after 1990), especially in Volume m. In
addition there are several volumes of mercury papers from three international
conferences that are available or will become available soon [1.1991. Water, Air,
Soil Pollut, Vol 55 (Swedish Final Mercury Report) and 56 (Gavle, Sweden First
International Mercury Conference); 2.1994. Lewis Publishers: Mercury Pollution:
Integration and Synthesis (Monterey US Second International Mercury
January 12,1995
D-75 J
-------
D. B. Porcella
Conference); 3. 1995 Water, Air, Soil Pollut., (In press for February). (Whistler,
BC, Canada, Third International Mercury Conference)].
6. Although there are many typos (for example, the executive summary for
Volume HI shows atmospheric deposition rates in units of milligrams instead of
micrograms, a difference of 3 orders of magnitude) and errors in the texts of the
Draft Report to Congress, many errors involve literature citations. There has been
very uneven coverage of the literature. Sometimes, the coverage is good
(discussion of RELMAP) and other times old citations and reviews and reports
dominate (Section 2). Moreover, many of the text citations are missing from the
reference list or are mis-cited or cited different ways. The old citations have two
problems: many times the results are wrong and mislead the investigator. Also,
the older work does not have the same frame of reference and understanding that
modern investigators have. Section 2 is particularly bad this way and should be
rewritten and placed in the appendix.
IV. Specific Comments
1. p. 1-5,1.1. It is proper to assess environmental problems using process
modeling, but it is important to ground-truth your modeling projections. Para. 4.
Although total mercury exposure is not the aim of this assessment, the relative
exposure is significant when one is talking about the effectiveness of controls. If
controls produce no measurable effect on exposure, then one might decide to
spend money more wisely. The last sentence is irrelevant for the para.
2. p. 2-3,1. 7. Total (includes 'natural' + anthropogenic) not 'natural' (which
includes natural + re-emissions of previously deposited). Note also, the reference
should be to the "Expert Panel on Mercury Atmospheric Processes"; there were
many sponsors.
3. p. 2-6, Para. 2. This discussion of global sources is very important. I suggest
you see other references^, because there is great uncertainty about how much
mercury deposition is distributed on a local, regional, and global basis**. It is
necessary to use only the latest estimates for these numbers, e.g., 12. por example,
the estimate of 10^ g in the atmosphere (Nriagu, 1979, first paragraph in section
2.3) is closer to 10? g based on Fitzgerald (1986) if one assumes a 1.5 km
atmospheric layer.
D ?6 January 12,1995
-------
D. B. Porcella
4. p. 2-7, para, beginning at top of page. This is a mixture of old and some more
recent information. Soils remain fairly uncharacterized, especially regarding
speciation. One bit of information that has recently become available has been the
importance of wetlands and sediments as reservoirs and sources of both mercury
and methylmercurylS. in other words, methylmercury appears to be produced in
wetlands. Measurements of fish from waters adjacent to wetlands seem to have
higher mercury than those that do not, and methylmercury fractions of total
mercury in water are higher than other areas. Consequently, calculation of
methylmercury as a simple ratio to total mercury is erroneous. Many site specific
factors appear to govern how much methylmercury is formed for uptake by fish^,
and that is why models like the Mercury Cycling Model were developed 15. This
allows one to calculate production of methylmercury is a particular site, and its
accumulation by fish. One result of importance shows that virtually half the
mercury entering fish comes out of the sediments, even in a seepage lake which is
dominated by atmospheric deposition^. The sediments represent past activities.
Thus, previous higher deposition of mercury could lead to higher fish
concentrations today as mercury comes out of the sediments. In Minnesota, even
though the present deposition of mercury is about 1/3 of that in the 1960's^, there
has been no apparent decrease of fish mercury concentrations-^. These results
suggest that control of mercury emissions will provide little benefit in reducing
fish mercury concentrations. With regard to that question, Swedish investigators
have shown a response to a marked reduction in mercury emissions from central
Europe which led to about 10-20 percent reduction in total mercury water
concentrations-^. At present we do not know the response of fish. This was
observed after the complete cessation of emissions amounting to 300 metric
tons/yr from sources in former East Germany. These emissions are more than the
total US emissions, and moreover arise from a relatively small area. The rssults
show that cessation of a large source can have measureable benefit. It is unlikely
that many relatively small sources spread over a large area will show similar
responses.
5. p. 2-7, Section 2.3.4. Translocation in plants is generally considered to be
minimal. Again there is great uncertainty in plant mercury dynamics. The authors
have come up with what I consider a reasonable conclusion, but the science is not
clearly stated, especially reflecting the above-ground plant processes that may
affect mercury exposure^-
January 12,1995
D-77
-------
D. B. PorceUa
6- p. 2-8, para. 4 (also, end of para. 4, p. 2-9). In the last sentence the authors state
that methylmercury which is lost is made up by additional methylation. I am not
sure what this means. If they are saying methylmercury is at steady state
concentrations, I would agree and this result is consistent with field evidence-^.
Such a result would explain why Minnesota fish have not responded to a 3-fold
reduction in mercury deposition. On this same page, two paragraphs should be
reorganized: Para. 2 should be inserted in section 2.3.4; and the last para, should
be inserted into para. 1 of 2.3.5.
7. p. 2-9, para. 3. The Akagi work is out of date, and moreover, the sentence does
not make sense except for a system that is highly polluted and which does not
have a food chain. Generally, in remote lakes, fish contain more methylmercury
than the rest of the entire ecosystem with the exception of sediments.
8. Bullets and summary para., p. 2-10. Bullet 1 has the same error referred to in
Comment IV-7. Bullet 3 is unclear - see Comment 6. Bullet 4 has not been
discussed previously. I agree with the statement but additional discussion is
needed before putting it in a summary (see 20). The paragraph is very unclear. It
seems to restate Bullet 3. The consequences of these bullets appears tobe that
atmospheric deposition is not a major contributor to mercury accumulation by
fish. Rather site-specific factors are the important drivers of mercury accumulation
by fish14.
9. p. 2-10, last two sentences of last para. That atmospheric mercury is increasing
is an hypothesis20'21/22. The hypothesis is based on global mercury not on
regional mercury. For at least one part of the US, Minnesota, mercury deposition
appears to be decreasing.
10. Section 2.4.2. The discussion of water concentrations is out of date. Rain often
has more than ten times the mercury concentration of normal drinking water. The
Michigan Environmental Science Board report has a better summary of analytical
problems of old data, as do many others Table 2-6 provides an example of what I
refer to. The data by WHO and especially Seritti appear high. More recent
compilations are available^.
11. Section 2.4.4. The authors do not address a central question about fish
mercury concentrations and diets. Generally, humans eat only a small portion of
freshwater fish (11 percent in New Jersey^), and the analysis of US mercury
emission sources is directed only at freshwater fish. Shouldn't there be a factor to
correct exposure based on actual diets (see Lipfert et al. 1994). Marine fish
presumably respond to global mercury. However, deposition in the central
January 12,1995
D-78
-------
D. B. Porcella
northern Pacific is equivalent to lake deposition in northern Wisconsin^, and the
major factor associated with mid-ocean deposition appears to involve particles^.
The last paragraph on p. 2-15 mentions that paper mills were located near areas
where the highest fish mercury was found, and presumably, this was an
association because of the use of methylmercury as a fungicide; more discussion is
needed. In Table 2-9 there is no notation for point source contamination,
Superfund sites, or other such sources even though there have been references to
the fact that the highest fish mercury levels were associated with paper mills and
wastewater treatment plants (p. 2-15).
• RELMAP specific:
12. Table 3-2. Note: The RELMAP simulations were performed with different
data than the emissions inventory (Table 2-1). Medical waste combustion was not
identified, and electric utility boilers were overestimated by about 20 percent.
Why were non-utility (industrial boilers) fossil fuel combustion sources not given
the same speciation as electric utility boilers. Presumably, these properties are a
function of coal composition and combustion temperature. They should be the
same. Moreover, in Volume n, several potentially large sources have not been
assessed (ore roasting, paints, natural sources, and re-emissions).
13. p. 3-4, last sentence, para. 2. The authors refer to particulate formation of Hgll
in the stack plume as speculative. The idea that Hglf remains as a gaseous
component is certainly more speculative.
14. Figure 3-1, discussion on p. 3-5. The base case is 1989. It seems logical that the
scenario of 2010, when SOx and NOx controls are implemented under the 1990
CAAA, should be considered, also. I was curious why LosAngeles, San Francisco,
and south Florida showed up on these maps when they do not presently burn coal.
Is it because of municipal waste combustion? Certainly, the high deposition in
California must represent excessive oxidation of HgO.
15. Section 3.2.5. The importance of speciation of emissions cannot be
overestimated. Very little data are available on the relative proportion of Hgll and
which factors control mercury speciation. 1 Better understanding might lead to
better methods to control mercury. Unfortunately, this section dwells on
simulations that give excessively high deposition compared for anthropogenic
sources to actual numbers, and which do not accurately include all the sources.
16. p. 3-22, last para. After many warnings about the uncertainty in the RELMAP
modeling, the authors provide the good news that an improved model is in the
January 12,1995
D-79
-------
D. B. Porcella
works. Based on the results presented in Volume HI, I conclude that it is
premature to draw conclusions from the long-range transport analysis.
• Local Impact Analysis
17. p. 4-2, para. beg. line 3. The discussion of uncertainty is accurate, and applies
strongly to the entirety of section 4. The authors refer to the analysis as a scoping
study in Table 4-11, and I think that accurately describes the effort. Moreover, the
analysis of uncertainty is described as limited, with a view to influence future
research directions. This is entirely appropriate for the level of detail in this report.
It is obvious to this reviewer that the uncertainty in this report has served
primarily to direct where research is needed. The conclusions listed in the
Executive Summary are unjustified given the conservatism of the assumptions and
the errors in analysis.
18. p. 4-3, para. 3. The vapor/particle ratio is described as being constant from
stack through plume, and correctly state that this is a simplification of reality.
Beginning on line 25, they further state that the assumption of no plume chemistry
is a particularly important source of uncertainty. This is correct.
19. Table 4-4 is confusing, perhaps due to typos. Should the RSF notes be UR
notes under the urban location?
20. p. 4-12, para. 4,5. The fish consumption data seem to be irrelevant. Why use
fish consumers in the Columbia River, where salmon consumption is considerably
greater than other parts of the US where the high deposition occurs. This
overestimates the impact of deposition. Perhaps, a more realistic approach would
be to use the data shown in Table 4-9 as exemplified by Lipfert et al. 1994.
21. Section 4.1.5.5. This section describes a large, very shallow lake, with an
extensive watershed. This is unlike most of the lakes listed in the fish surveys, and
would tend to maximize the amount of mercury exposure to fish.
22. Section 4.1.6. The purpose of this section is very unclear; in addition, it is
redundant with the introduction. It describes the modelling efforts, but after
reading it, I remain unclear about what was done.
23. p. 4-19,1.5. This discussion describes the mass balance cycle for mercury from
the IEM2 model. It would help the reader to have a presentation of the mass
balance which shows how the mercury emissions are distributed according to
Figure 4-1. I am curious what percentage of mercury falling on a watershed
remains in the watershed.
24. Section 4.2, p. 4-20. The authors assume that steady state exists after 30 years
January 12,1995
D-80
-------
D. B. Porcella
However they provide not evidence for their data, and do not dearly discuss loss
processes to soil (sediments), biota, or the atmosphere.
25. Figure 4-2ff.. What is the P in 2.5 P? Is this the same as 2.5 km? Avoid jargon!
26. Table 4-14. Typo where boilers and copper smelters have same air values.
27. Table 4-15. How do these values compare to actual measurements. It should
be possible to groundtruth these data using this process. For example, in
Maryland, a study by the Maryland Power Plant Team shows no differences in fish
mercury concentrations for ponds adjacent to power plants and far from power
plants. Soil mercury concentrations might show a difference if it exists. If
designed correctly, such an experiment could provide a means of testing their
hypothesis of such high local deposition.
28. p. 4-27,1.20. The divalent mercury emission rate of 50 percent is reasonable,
and cannot be classed as a "... generally low assumed divalent mercury emission
rate,...".
29. Tables 4-16 to 4-18. I am unsure of what was done here. WereCOMPMECH
and RELMAP values added together? Both model studies were mentioned in the
table titles. The discussion is very unclear. On page 4-32, there is a statement that
the RELMAP results were added to the local impact data (para. 3). This should be
clearly explained since the grid scales are different and the concentrations,
especially in soil, appear quite high compared to sites studied by Nater and Grigal.
30. p. 4-31,1.14. Is the cistern concentration of methylmercury based on the
arbitrary ratio of 15 percent. Where did that number come from; it is highly
suspect since rainwater is usually no more than 1 percent methylmercury. As an
aside: the methylmercury in air does not seem to come from stacks, and remains a
mystery. Recent Swedish measurements found no methylmercury in FGD sludge
or other solid wastes from combustion sources.
31. Figures 4-8 and 4-9. The deposition rates appear extremely high. These rates
should be easy to detect. I gather from previous discussion that it is not easy to
detect such deposition. Again, there is no attempt to use data to check highly
uncertain analyses - to groundtruth the predictions.
32. Table 4-20. Comments similar to 29-31. Except for the low range, the values
are almost all greater than measured in this country. This is an important area for
research and should afford the opportunity to make accurate and precise tests of
this hypothesis.
33. Figures 4-18,4-19. The water concentrations are more than an order of
magnitude greater than any concentration measured in US surface waters where a
January 12,1995
D-81
-------
D. B. Porcella
point source was not involved. The values exceed by more than a factor of 10 the
Onondaga Lake concentrations, a Superfund site involving a former chlor-alkali
plant. Thus, any conclusions drawn from this analysis have no value.
34. p. 4-3, Fish section. The BAF is a concept that probably has little value except
for making ballpark estimates. Mercury concentrations in seepage lakes fish vary
by a factor of 10 in areas receiving very uniform deposition and having relatively
similar lake types. Water quality and other limnological variables seem
35. p. 4-52, para. 2. The fish concentrations are excessive when compared to actual
field data, e.g., 26 ug/g (26ppm). Such a value exceeds the highest value in the
literature, and supposedly this value comes from a single emission source.
35. Figure 4-19a. Most field studies, even for lakes adjacent to power plants^ do
not exceed a range of 1-2 ppm, values considerably below the averages for several
of the examples studies shown.
36. Table 4-24. Although I agree that fish are the most likely source of risk to
humans, the construction of the scenarios has resulted in non-credible results. No
one should accept these results. A similar comment applies to Table 4-25
regarding wildlife.
37. p. 4-56, 1. 35. The authors use the expression "Again, in this modelling effort,
..." suggesting again that this is a scoping study. However, it is not a good scoping
study since no attempt was made to groundtruth the results or to carry the
assessment further.
38. Section 4.2.2.5. There have been several attempts to use bird feathers to
measure methylmercury exposure in birds - analogous to using hair samples to
evaluate human methylmercury dynamics. The opportunity to test hypotheses
gained from the wildlife estimates should be taken.
39. p. 4-59. Para. 2,3. These paragraphs contain new information that is not
discussed, and are so poorly written that the reader cannot understand them.
Similarly, Table 4-26 and 4-27 need discussion that helps explain what the results
mean and how they were derived. Perhaps, breaking the information into several
tables would help.
40. Section 4.2.3 and Table 4-29. As in previous comments, the excessive
predictions of mercury in fish show that the analysis has serious faults.
January 12,1995
D-82 3
-------
Porcella, D. B.
List of supra-numeral citations for References (in following section).
1995 citations refer to presentations at the Whistler Meeting on Hg: In Press. Water
Air, and Soil Pollution.
1. Chu and Porcella, 1995
2. New Jersey DEPE. Report on Mercury. No reference available. Contact Joanne
Held.
3. Lindqvist et al. 1991; USEPA, 1990.
4. Benoit et al. 1994.
5. Engstrom Personal Communication 1994.
6. Fitzgerald et al. 1991; Munthe, 1991; Seigneur et al. 1994, EPMAP1994.
7. Fitzgerald 1986,1988; Porcella, 1994; Watras et al. 1994.
8. EPMAP, 1994, Sorenson et al. 1994.
9. Fish surveys: McMurtry et al. 1988; Spry and Wiener 1991; Wiener and Spry
1994; Swain and Helwig, 1989; Grieb et al. 1990; Gloss et al. 1990; New Jersey
DEPE 1994; DOE-FDA-EPA Workshop, 1994.; Driscoll et al. 1994; there are many
others.
10. Seigneur et al., 1994; Lipfert et al. 1994; EPRI 1994; Constantinou et al. 1994.
11. Engstrom et al. 1994; EPMAP, 1994.
12. Lindqvist et al. 1991; Nriagu and Pacyna 1988, Nriagu 1993,1994, Nriagu et al.
1992; Fitzgerald and Clarkson 1991; Porcella, 1994.
13. Zillioux et al. 1993. Rudd 1983, St. Louis et al.1994, and Rudd and colleagues at
Whistler Hg conference. Driscoll et al. 1994. Lee et al. 1985.
14. Watras et al. 1995. Driscoll et al. 1994.
15. Hudson et al.1994.
16. Porcella 1994, Hoffman, 1994.
17. Swain and Helwig, 1989.
18. Hultberg Personal Communication 1994.
19. Lindberg 1992. Hansen et al. 1995.
20. Mason et al. 1994.
21. Lamborg et al. 1995.
22. EPMAP 1994.
23. Mason and Fitzgerald, 1990.
24. Crump, 1994; Crump et al. 1994, Gearhart et al. 1994.
25. Knobeloch,et al.1994.
D-83
-------
Porcella, D. B.
References (Please cross reference names with numbers):
Benoit, J. Mv W. F. Fitzgerald, and A. W. H. Damman. 1994. "Historical
Atmospheric Mercury Deposition in the Mid-Continental United States as
Recorded in an Ombrotrophic Peat Bog." In Mercury as a Global Pollutant. Edited
by C J. Watras and J. W. Huckabee. Ann Arbor: Lewis Publishers, pp. 187-202.
Chu, P. and D. B. Porcella. 1995. Mercury stack emissions from U. S. electric utility
power plants. Water, Air, Soil Pollut. In Press.
Constantinou, E., M. Gerath, D. Mitchell, C. Seigneur, and L. Levin. 1995. Mercury
from power plants: Environmental cycling and health effects. Water Air Soil
Pollut. In press.
Crump, K. 1994. Calculation of benchmark doses from continuous data. Risk
Anal. In press.
Crump, K., J. Viren, A. Silvers, H. Clewell. J. Gearheart, and A. Shipp. 1994. Re-
analysis of dose-response data from the Iraqi methylmercury poisoning episode.
DOE/FDA/EPA. 1994. Workshop on methylmercury and human health. 1994.
Brookhaven National Laboratory. Upton, NY. Conf. 9403156.140 pp.
Driscoll, C. D., C. Van, C. L. Schofield, R. Munson, and J. Holsapple. 1994. The
Chemistry and Bioavailability of Mercury in Remote Adirondack Lakes. Environ.
Sci. Tech.. 28:136A-143A.
Engstrom, D. E. 1994. Personal Communication. University of Minnesota.
Engstrom, D. R., E. B. Swain, T. A. Henning, M. E. Brigham, and P. L. Brezonik.
1994. Atmospheric mercury deposition to lakes and watersheds: a quantitative
reconstruction from multiple sediment cores. In Environmental Chemistry of
Lakes and Reservoirs. (L. A. Baker, ed.) ACS, Advances in Chemistry Series,
Washington, DC. pp. 33-66.
EPMAP (Expert Panel on Mercury Atmospheric Processes). 1994. Mercury
Atmospheric Processes: A Synthesis Report. EPRI/TR-104214. 23 p.
EPRI (Electric Power Research Institute). 1994. Electric utility trace substances
synthesis report. EPRI TR-104614-V1-4. EPRI, Palo Alto CA. 4 volumes.
Fitzgerald, W. F. 1986. "Cycling of Mercury Between the Atmosphere and
Oceans." In The Role of Air-Sea Exchange in Geochemical Cycling. NATO
Advanced Science Institutes Series. Edited by P. Buat-Menard. Dordrecht,
Netherlands: Reidel, pp. 363-408.
Fitzgerald, W. F. 1989. "Atmospheric and Oceanic Cycling of Mercury." In
Chemical Oceanography. Edited by J. P. Riley and R. Chester, guest edited by R.
A. Duce. New York: Academic Press, Ltd., Vol. 10, pp. 152-185.
Fitzgerald. W. F. and T. W. Clarkson. 1991. Mercury and Monomethylmercury:
D-84
-------
Porcella, D. B.
Present and Future Concerns. Environ. Health Persp. 96:159-166.
Fitzgerald, W. E, R. P. Mason, and G. M. Vandal. 1991. Atmospheric Cycling and
Air-Water Exchange of Mercury Over Mid-Continental Lacustrine Regions. Water Air
Soil Pollut. 56:745-768.
Gearhart, J. M., H. J. Clewell m, K. Crump, A. Shipp, and A. Silvers. 1995.
Pharmacokinetic dose estimates of mercury in children and dose-response curves of
performance tests in a large epidemiological study. Water, Air, Soil Pollut. In press.
Gloss, S. P., T. M. Grieb, C. T. Driscoll, C. L. Schofield, J. P. Baker, D. H. Landers, and
D. B. Porcella. 1990. Mercury levels in fish from the Upper Peninsula of Michigan
(ELS Subregion 2B) in relation to lake acidity. USEPA. Corvallis OR..
Grieb, T. M., C. T. Driscoll, S. P. Gloss, C. L. Schofield, G. L. Bowie, and D. B. Porcella.
1990. Factors Affecting Mercury Accumulation in Fish in the Upper Michigan
Peninsula. Environ. Toxicol. Chem. 9:919-930.
Hanson, P. J., S. E. Lindberg, K. H. Kim, J. G. Owens, and T. A. Tabberer. 1995.
Air/surface exchange of mercury vapor in the forest canopy -1. Laboratory studies of
foliar Hg vapor exchange. Water, Air, Soil Pollut. In Press.
Hoffman, R. Proceedings document for the National Forum on Mercury in Fish,
New Orleans, September 1994. USEPA, Washington D. C. In Press
Hudson, R. J. M., S. A. Gherini, C. J. Watras, and D. B. Porcella. 1994. "A
Mechanistic Model of the Biogeochemical Cycle of Mercury in Lakes." In Mercury
as a Global Pollutant. Edited by C. J. Watras and J. W. Huckabee. Ann Arbor:
Lewis Publishers, pp. 473-523.
Hultberg, H. 1994. Personal Communication. Swedish Environmental Research
Institute.
Knobeloch, L. M., M. Ziarnik, H. A. Anderson. Draft. Imported seabass as a source
of mercury exposure: A Wisconsin case study. Bureau of Public Health. Madison
WI. MS 17pp.
Lamborg, C. H., W. F. Fitzgerald, G. M. Vandal, and K. R. Rolfhus. 1995.
Atmospheric mercury in northern Wisconsin: sources and species. Water, Air, Soil
Pollut. In press.
Lee, Y. H., H. Hultberg, and I. Andersson. 1985. Catalytic Effect of Various Metal
Ions on the Methylation of Mercury in the Presence of Humic Substances. Water,
Air, Soil Pollut. 25:391-400.
Lindberg, S. E., 1992. Atmosphere-Surface Exchange of Mercury in a Forest:
Results of Model and Gradient Approaches. J. Geophys. Res. 97D2:2519-2528.
Lindqvist, O., K. Johansson, M. Astrup, A. Andersson, L. Bringmark, G. Hovsenius,
A. Iverfeldt, M. Mieli, and B. Timm. 1991. Mercury in the Swedish Environment—
Recent Research on Causes, Consequences and Corrective Methods. Water Air Soil
Pollut. 55: i-261.
D-84(a)
-------
Porcella, D. B.
Lipfert, F. W., P. D. Moskowitz, V. M. Fthenakis, M. P. DePhillips, J. Viren, and L.
Saroff. 1994. Assessment of mercury health risks to adults from coal combustion.
Brookhaven National Laboratory. Upton, NY.
Mason, R. P. and W. F. Fitzgerald. 1990. Alkylmercury Species in the Equatorial
Pacific. Nature 347:457-459.
Mason, R. P., W. F. Fitzgerald, and F. M. M. Morel. 1994. The Biogeochemical
Cycling of Elemental Mercury: Anthropogenic Influences. Geochim. Cosmochim.
Acta. 58:3191-3198.
McMurtry, M. J., D. L. Wales, W. A. Scheider, G. L. Beggs and P. E. Dimond. 1988.
Relationship of mercury concentrations in lake trout (Salvelinus namaycush) and
smallmouth bass (Micropterus dolomieui) to the physical and chemical
characteristics of Ontario lakes. Can. J. Fish. Aq. Sci. 46:426-434.
Munthe,J. 1991. "The Redox Cycling of Mercury in the Atmosphere." PhD
dissertation, Department of Inorganic Chemistry, University of Goteborg,
Goteborg, Sweden.
New Jersey Department of Environmental Protection and Energy. 1994.
Preliminary assessment of total mercury concentrations in fishes from rivers, lakes
and reservoirs of New Jersey. Rept. No. 93-15f. Trenton, NJ. 92 pp.
Nriagu, J. O. 1993. Legacy of Mercury Pollution. Nature 363:589.
Nriagu, J. O. 1994. Mercury Pollution From the Past Mining of Gold and Silver in the
Americas. Sci. Total Environ., in press.
Nriagu, J. O. and J. M. Pacyna. 1988. Quantitative Assessment of Worldwide
Contamination of Air, Water and Soils by Trace Metals. Nature 333:134-139.
Nriagu, J. O., W. C. Pfeiffer, O. Malm, C. M. M. de Souza, and G. Mierle. 1992.
Mercury Pollution in Brazil. Nature 356:389.
Porcella, D. B. 1994. "Mercury in the Environment: Biogeochemistry." In Mercury
as a Global Pollutant. Watras. Edited by C. J.Watras and J. W. Huckabee. Ann
Arbor: Lewis Publishers, pp. 1-19.
Rudd, J. W. M., M. A. Turner, A. Furutani, A. Swick, and B. E. Townsend. 1983. I.
A Synthesis of Recent Research With a View Towards Mercury Amelioration. Can.
J. Fish. Aquat. Sci. 40:2206-2217.
St. Louis, V. L., J. W. M. Rudd, C. A. Kelly, K. G. Beaty, N. S. Bloom, and R. J. Flett.
1994. Importance of Wetlands as Sources of Methyl Mercury to Boreal Forest
Ecosystems. Can. J. Fish. Aq. Sci., 51:1065-1076.
Seigneur, C., E. Constantinou, and T. Permutt. 1994. Uncertainty analysis of helath
risk estimates. In Science and judgment in risk assessment. National Research
Council. National Academy Press. Washington, DC. Appendix F, pp. 453-478.
Seigneur, C., J. Wrobel, and E. Constantinou. 1994. A Chemical Kinetic
Mechanism for Atmospheric Mercury. Environ. Sci. & Technol. 28:1589-1597.
D-84(b)
-------
Porcella, D. B.
Sorensen, J. A., G. E. Glass, and K. W. Schmidt. 1994. Regional patterns of wet
mercury deposition. Environ. Sci. Technol. 28:2025-2032.
Spry, D. J. and J. G. Wiener. 1991. Metal Unavailability and Toxicity to Fish in
Low-Alkalinity Lakes: a Critical Review. Environ. Poll. 71:243-304.
Swain, E. B. and D. D. Helwig. 1989. Mercury in fish from northeastern Minnesota
Lakes: Historical trends, environmental correlates, and potential sources. J.
Minnesota Acad. Sci. 55:103-109.
USEPA. 1990. Characterization of products containing mercury in municipal solid
wastes in the US 1970-2000. EPA Contract No. 68 W9-0040.
Watras C. J., N. S. Bloom, R. J. M. Hudson, S. A. Gherini, R. Munson, S. A. Claas, K.
A. Morrison, J. Hurley, J. G. Wiener, W. F. Fitzgerald, R. Mason, G. Vandal, D.
Powell, R. Rada, L. Rislove, M. Winfrey, J. Elder, D. Krabbenhoft, A. W. Andren, C.
Babiarz, and D. B. Porcella. 1994. "Sources and Fates of Mercury and
Methylmercury in Remote Temperate Lakes. " In Mercury as a Global Pollutant.
Watras. Edited by C. J.Watras and J. W. Huckabee. Ann Arbor: Lewis Publishers,
pp. 153-177.
Watras, C. J., K. A. Morrison and N. S. Bloom. 1995. Mercury in remote Rocky
Mountain lakes of Glacier National Park (Montana) in comparison with other
temperate North American Regions. Can. J. Fish. Aq. Sci. In press.
Wiener, J. G. and D. J. Spry. 1995. lexicological significance of mercury in
freshwater fish. In Interpreting envionmental contaminants in animal tissues. (G.
Heinz and N. Beyer, eds.) Lewis Publishers, In press.
ZiUioux, E. J., D. B. Porcella, and J. M. Benoit. 1993. Mercury Cycling and Effects
in Freshwater Wetland Ecosystems. Environ. Toxic. Chem. 2:2245-2264.
D-84(c)
-------
-------
Volume IV
Health Effects of Mercury and Mercury Compounds
D-85
-------
-------
Paul Mushak
D-87
-------
Mushak
REVIEW COMMENTS ON TEXT & APPENDIX B OF VOLUME 4:
EPA STUDY REPORT TO CONGRESS
Paul Mushak, Ph.D., Principal
PB Associates
Ste G-3, Couch Bldg.
714 Ninth St.
Durham, NC 27705
Review comments are presented as (1) general comments on the volume, to include
IRIS data, (2) general comments on individual chapters in the volume, to include IRIS
derivations and values in Appendix B and (3) specific comments on each chapter.
GENERAL COMMENTS ON VOLUME IV
The material in this volume in tandem with Volume 3 drives, in large measure, the
effort for quantitative human health risk assessment done in this EPA report to Congress. This
volume's quality and credibility, therefore, helps determine the relative quality of the report.
Preparation and Organization of the Volume
The sequence of chapters in Volume 4 is generally acceptable, with some notable '
exceptions. I also have a number of problems with how the chapters were organized.
It is not clear to me why a very short exposure summary, that omits adequate
discussion of important sources, appears in this volume on adverse health effects. A better
explanation in the Introduction as to why there is an exposure section would be helpful. It is
equally unclear why chapters on interactions and risk populations appear after, and seemingly
uncoupled from, the critical Chapter 5. Whether explicitly subsumed in the exercises in
Chapter 5 or not, such modifying factors should precede chapter 5. These factors help
determine the relevance and credibility of the quantitative risk measures. Finally, what's the
point of the current Chapter 8 as it is now structured? Tabulated grantee information as it is
now presented, is largely inadequate to say what the researchers are doing or how well they
are doing it. The discussions of the ongoing work with the Seychelles and Faroe islanders
should remain.
A number of the grants noted in the table are for a topic which appears under-
evaluated in this chapter, that of elemental mercury release from Ag-Hg dental amalgams,
particularly among individuals with large numbers of fillings and/or who chew gum or grind
D-88
-------
Mushak
their teeth (bruxism). There are now quite a few credible scientific studies from around the
world that show (1) elemental mercury release from amalgams in proportion to fillings, (2) an
increase in body burden with fillings present and a decline in mercury body burden with
filling removal, and (3) intake/retention levels can greatly exceed all other environmental
sources of inorganic, elemental mercury and can even approach intakes that occur with certain
occupational exposures to Hg vapor. WHO, in its environmental health criteria documents
for both methylmercury (#100) and inorganic mercury (#118) acknowledge the potential
problems with this exposure source. Chapter 2 should discuss this critically as well.
The chapters themselves are quite uneven as to their clarity, ease of reading, and
handling of the key data sets. Various authors may have prepared the different sections.
Chapter 4 is a jumble hi the current draft and requires major reorganization, although the key
information appears to be largely present. It is diffuse in its focus, confusing or arcane in its
text, and almost hopelessly complex as to any clear subsectional interconnections. The reason
is simple: the authors have attempted to have chapter 4 do two things: to present a core of
pre-selected key studies for later risk assessment sections and to attempt to present a fairly
comprehensive stand-alone compendium of health effects information but without the critical
evaluation. It does not work. We wind up with tables all over the place and out of place,
obscurely listed information, information that is not obviously assigned to a clear heading,
etc. The whole chapter is currently hostage to this intent of a dual purpose. What's more, the
selection of the key studies to get the most attention is left now to EPA's author(s). That
process of pre-selection may give a set of studies that is or is not apt to be the same as the
peer review group's selection^). Since EPA has shortened its evaluation and discussion of the
many remaining studies, the reader can't easily judge from the skimpy discussions how good
this preselection was. The process of study selection should consist of chapter 4 giving all the
key information arrayed among major headings and subheadings, with tables of the core
studies to be used presented in chapter 5, along with cross-referencing to chapter 4.
I also have a problem with the present classification of toxic effects, a classification
which should be set forth in easily comprehended language at the beginning of chapter 4. For
example, the current focus on "developmental" effects of, e.g., methylmercury, lead, and
other developmental toxins are on subtle toxic effects such as 'developmental neurotoxicity'
and 'neurobehavioral teratology.' The chapter needs a better way to deal with these sorts of
D-89
-------
Mushak
subtle neurotoxic effects that are imparted at critical stages of development. The present text's
use of the term to mean overt/severe effects confined largely in-utero, is narrower than the
classical definition and is of limited usefulness for subtle developmental effects of toxins
operating pre-, peri-, and postnatally early in life. I have further comments below.
Level of Critical Evaluation of Data
The level of critical evaluation given data from various studies is uneven across
chapters and even within chapters. This may be due in part to variable familiarity of given
authors with different chapters or even different authors of the chapters. Some of the section
text is given critical discussion in depth, while other portions are offered with little critical
comment, even to the point of simply offering the original author's conclusions. Readers and
reviewers are less interested in authors' conclusions-which in any case are just as likely to be
incorrect as correct- as they are in what the report authors conclude about the data.
Fortunately, there appears to be more critical evaluation given to the more significant studies,
particularly those that are used in chapter 5.
Gaps in the Data Base
Overall, the volume is reasonably thorough in terms of the amount of published, peer-
reviewed material that was included for evaluations critical to the purposes of the volume.
However, there are published, peer-reviewed citations that do not appear in the various
chapters of the volume, particularly Chapter 4, the adverse health effects chapter. Part of the
omission may be papers appearing after the cut-off time for report preparation, a time line I
don't readily find in any sections. If time permits, I will provide a citation list to ERG of any
of these papers, if any appear necessary for inclusion. Based on years of experience co-
authoring or peer-reviewing assessment documents on metals and metalloids for Federal and
international agencies, I would say that only a fraction of the many papers published for a
substance is actually useful for risk assessment. This report, given its purpose, should not
attempt to be an exhaustive archive on the mercury data base. An explanation in the
introduction to the report should note the criteria for inclusion, evaluation and use of
published work.
Integration of the Material Across Chapters
Overall, the volume appears to be well integrated within and across chapters,
whatever the sequence of the chapters and their form. Chapter 5 employs the material
D-90
-------
Mushak
provided in earlier chapters in a reasonable fashion. The chapter (3) on toxicokinetics
integrates exposure data from known important intake routes to in-vivo kinetics, and from
there provides the toxicokinetic underpinning for the use of the valid biomarkers of exposure
and the dose portion of dose-response relationships in chapter 5.
GENERAL COMMENTS ON EACH OF THE REPORT CHAPTERS
Executive Summary
The executive summary is generally good in describing the various chapters in the
volume, acceptable hi terms of providing concise description of the key conclusions to be
drawn, but not very clear or focused about one critical element, the data in Table ES-1, or a
critical look at the various chapters.
Executive summaries are intended to be stand alone sections that should integrate the
data and give the critical points in plain language, for the nonexpert policy maker, legislator,
or regulator. If each chapter has a well-done Executive Summary, then these summaries
could be simply combined, with added overview, to provide the report summary.
Table ES-1, and the associated text that introduces it, needs more explanation and
discussion. The current executive summary simply states: here are the RfDs, the RfC, and the
cancer risk rankings. What does that mean to the general, nonexpert reader? Defining these
terms and placing them hi relative exposure context are serious gaps that require closure.
The Executive Summary should be more critical and interpretive about the
health effects information and the associated risk assessment information that follows it. It's
more than a little ridiculous that the whole focus of this volume is health effects, but the
actual text covering toxic effects at environmental exposures in the Executive Summary is no
more than part of a skimpy five-line paragraph. Worse yet, the key toxic effects summary is
part of a two-line sentence. By contrast, the toxicokinetics section gets a whole 13-line
paragraph and the interactions section gets a whole paragraph. These portions of the
Executive Summary need rewriting for both balance and useful summary information for the
general reader. The discussion of the interactions and the risk populations need more critical
comment and pruning as to which risk populations are at particular risk and which interactions
are most germane for public health. None of this level of interpretation appears. For example,
D-91
-------
Mushak
the text does not indicate to the reader that it is the interaction with selenium that is
particularly critical, yet the paragraph lumps Se with other substances indiscriminately, such
that the former gets the same play as tellurium and atrazine and the general reader will
assume equivalent importance among them.
The last portion of the Executive Summary, research needs and current research
activity, is of particular importance for the policy maker or others holding the purse strings
for research funding. Yet the text is rather skimpy and need more specifics. Wording such as
"well designed studies are needed to clarify exposure levels... other toxic effects occur" does
not even make grammatical sense. Toxic effects other than what toxic effects????
1. Introduction
This is not an Introduction; it's more like a glorified Table of Contents. The Chapter
should be rewritten with an expansion to provide the overall rationale for why the volume
contains what it does contain and its purpose in the report. The chapter should define and put
the jargon in context, e.g., IRIS, for the otherwise informed scientist who does not follow
EPA regulatory risk assessment shorthand. The detailed rationale for the organization of the
chapter is critical for the topic of health effects, because it is one that is much more multi-
faceted in terms of disciplinary focus than the others. Text for the introduction can simply
begin with material lifted from the introductory parts in each of the following chapters. If
there were different authors for this volume, perhaps they can combine their efforts to provide
a coherent rationale to the overall volume.
2. Summary of Human Exposure to Mercury Compounds
It is not clear to me that this very brief, 3-page summary of exposure adds much in
the form it is in. It should include more comparative exposure analysis text for dental
amalgam Hg in the section on general population exposures. Presently, there is an iolated
paragraph summarizing reported releases and statements that it's tough to quantify such
releases. Actually, releases from a person's fillings can potentially be estimated to some
extent by use of his/her urinary Hg level, an approach described by Roels et al. (1987) for
workers, and based on various studies showing a good urinary Hg-amalgam Hg relationship.
The complexities and uncertainties in the case of amalgam Hg monitoring are hardly any
more problematic than they are for, say, modelling approaches for long-distance transport and
deposition estimates for stationary-source emissions of elemental Hg to ambient air. Yet, the
D-92
-------
Mushak
latter's complexities have not impeded the authors of Volumes 2 and 3 of the report. More
logic and consistency in the process, please.
Table 2-1 makes it clear that amalgams can be the dominant source of elemental Hg
and can even rival occupational exposures. As noted already, there are a number of reliable
studies that establish Hg release from amalgam fillings and contribution of this source to body
burden Hg. Recent data, e.g., those of Begerow et al. (Int. Arch. Occup. Environ. Health
66:209, 1994), also show a mercury half-life (Hg-U) of 90-plus days, the same value in the
toxicokinetic profile for occupational exposure. This indicates a steady-state, common
distribution and transfer pool for both sources.
Besides saying little hi this volume about amalgam mercury releases, this report to
Congress says nothing on the topic hi the exposure volume, volume 3, and says little in
Volume 5. A huge number of Americans have amalgam fillings; the population that is
therefore affected in terms of risk assessment scenarios is enormous. The critical question
with Hg-amalgam that is required for discussion in this report is not whether release and
uptake by tissues occurs, something which is scientifically established and accepted by those
informed on the topic, but rather are releases sufficient to produce subtle or overt forms of
toxic effects. The available data do not indicate to me that any overt/clinical effects have been
convincingly documented. A number of clinical disorders have been claimed for Hg in
amalgams, but there are no parallel findings hi workers exposed to high amounts of mercury
vapor, where rates should be obviously elevated. For example, I am not aware that the
prevalence or incidence of multiple sclerosis, claimed by some to be related to amalgam
fillings, is seen in chloralkali or other Hg-exposed workers at a rate higher than in the general
population. We would expect higher rates in workers, on simple dose-response grounds.
Similarly, on toxicokinetic grounds, one cannot rationalize an immediate relief from
symptoms when amalgam fillings are removed. The half-life for Hg removal from various
tissues would be relatively much longer than the quick response times that are reported by
individuals claiming improvement after removal of fillings.
The current RfC for elemental Hg, based on a triad of neurotoxic effects seen in
occupational epidemiology studies of workers, is 3 x 10"4 mg/m3. It is useful to compare this
value with elemental Hg releases from dental amalgam fillings. If one selects a rough mid-
point for elemental Hg intake of about 15 /zg/day from the data used to develop intake and
D-93
-------
Mushak
retained Hg from amalgam elemental Hg release in Table 2-1, the corresponding
concentration of buccal cavity air Hg would be —0.8 /ig/m3, assuming 20 mVday combined
mouth/nasal breathing. This oral air level exceeds the RfC of 0.3 /ig/m3 by about three-fold.
Retention of mercury atoms released from amalgam would be at least the 80% accepted for
ambient air Hg° and possibly higher, factoring in trans-gingival/trans-buccal mucosal diffusion
as well as pulmonary deposition and retention. Similarly, the upper value of the intake range
hi Table 2-1 for amalgam Hg, 27 /zg/day corresponds to ~1.4 /tg/m3, about five-fold higher.
These intakes also greatly exceed typical ambient air elemental Hg intakes. Table 2-1
indicates a two-to-three-orders-of-magnitude difference. If one compares the maximum intake
of 27 fig/day from fillings in Table 2-1 with the maximum modelled ambient air elemental Hg
concentration for the 25 km emitter distance scenario for full-time adult exposures in Table 4-
21 of Volume 3 of the report, the difference is about 1,000-fold higher for the amalgam
releases as well.
The above comparisons permit several conclusions. The relative impact of ambient air
vs dental amalgam Hg release is considerably less. But such very high baseline release rates
for individuals in this exposure risk population who have a lot of dental fillings also clearly
indicate that no additional sources of mercury vapor intake, i.e., from ambient air, are
permissible, especially since the RfC can already be significantly exceeded via the amalgam
source alone. On the other hand, ambient air Hg remains an environmental healdi issue
regardless of such comparisons since Hg in this medium enters biogeochemical cycling with
neurotoxic methylmercury formation, a phenomenon unlikely to be quantitatively significant
for Hg releases from amalgam fillings remaining in-situ.
Chapter 3. Toxicokinetics
Overall, this is a good chapter and has a valuable purpose in the report. It is clearly
written and is organized in a logical sequence. Inclusion of a section on exposure biomarker
monitoring is a good idea, since the relative merits of biological monitoring have both
measurement and toxicokinetic components. Monitoring in biological media requires a
toxicokinetic basis for useful dose-response relationships. Superb measurement methodology is
still useless if the numbers obtained are not diagnostically or lexicologically relevant.
However, this should be stated in the monitoring section.
D-94
-------
Mushak
The methodology section is too brief and diffuse to be of much use. I would
recommend a major expansion of this section, to include discussion of the various media for
measurement as well. For example, hair Hg as total Hg is rather useless for monitoring
endogenous Hg disposition, if there is high likelihood of surface contamination. Hair Hg can
still be used for external monitoring, even with this limitation. It just is not reliable for
systemic dose evaluation. Mercury measurement of total element in hair lengths as described
by Clarkson's group for methylHg exposure is reasonable, since atmospheric Hg
contamination of hair surfaces in the Iraqis and other populations was not a concern.
A major gap in this chapter is absence of a section on toxicokinetic modelling, with
discussion of the appropriateness of the various models proposed over the years. A discussion
of PB-PK modelling applied to the mercurials should be included. The basic compartment
model approach described hi the earlier Task Group on Metal Accumulation documents with
later refinements should be included.
A second major gap is absence of a subsection on bioavailability of mercurial forms
when deposited in the receiving body compartments, i.e., the GI and pulmonary tracts. Both
bio-physicochemical and biological determinants of such bioavailability should be noted. For
example, are mercury levels in certain matrices as bioavailable as those in media such as food
and water?
Chapter 4. Biological Effects
This chapter has a number of problems, a major problem being the way it is
organized. The current arrangement with early appearance of the major studies used in
chapter 5 creates confusion with the rest of the chapter. I suggest that the sections in this
chapter be arranged by the major headings, with the key studies identified as such and
appearing in the first part of each subsection. For example, darting back and forth between
several sets of neurotoxicology texts creates a chaotic flow. The tables should appear within
the subsection where the corresponding text appears.
Studies which are not preselected for detailed discussion are, unfortunately, given too
little discussion. In some cases, it is not clear that the studies given mere mention are all that
marginal, to begin with. Authors need to look at a complete reorganization, beginning with
lethal, acute effects and then proceeding to subchronic and chronic effects. Under the latter,
the major headings by system or effect type can be appropriately placed sequentially. In all
D-95
-------
Mushak
three cases of form of mercury, it is useful to indicate what is a lethal dose/exposure. While
the focus of the report is environmental chronic exposures, it is useful to remind the reader
that these substances can be lethal if the dose is high enough and the route of administration is
effective.
As noted earlier, there is too frequent occurrence of uncritical narrative, with the text
offering little or no critical evaluation or interpretation. It is difficult to determine whether
this is due to writing by individuals with little expertise in the areas or what. In no case
should the text state as this report's only conclusions that the authors of the paper cited
conclude this or that. I am not interested in what the authors say, I'm interested in what the
critical evaluation says. An example of more required critical assessment is the study of
Gotelli et al., 1985, showing nephrotoxicity in 509 infants exposed to phenyl mercury, surely
a major exposure and toxicity episode. What does the study say about likely absorption rates,
body burdens, thresholds for kidney effects in infants, etc. The general reader and the
informed scientist will both be perplexed by such dismissive treatment of this paper.
As noted earlier, some of the sub-section headings could be tightened to reflect where
the current thinking is about such issues as potential subtle developmental effects of, e.g.,
methylmercury, on cognition and behavior, which are in fact subtle outcomes of
developmental neurotoxic insult.
The sections on carcinogenicity and mutagenicity are proportionate in terms of text
and discussion to the significance of these effects relative to developmental and other
neurotoxic effects.
Authors of this chapter and those of Chapters 6 and 7 should consider combining the
latter two into chapter 4. The risk population chapter is for effect-based risk, as opposed to
exposure-based higher risk, and can reasonably fit as a section in 4 or between present 4 and
5.
The interactions chapter should also appear before chapter 5, perhaps as a section of
chapter 4. There are two types of outcomes for interactions, one that modulates toxic effect
and one that modifies exposure, e.g., gastrointestinal uptake. Certainly, interactions that
influence toxic endpoints can be a section of chapter 4.
D-96
-------
Mushak
Chapter 5. Hazard Identification and Dose-Response Assessment
Overall, this important chapter is reasonably organized and presents the critical data
and the quantitative dose-response (imprecisely given the broader term, risk assessment)
calculations in a manner that is well-grounded in both chapter 4 and Appendix B. It would be
useful for the authors to introduce and generally discuss the elements of quantitative risk
assessment as they apply to chapter 5. It could be a summary of intro material in volume 5.
Those of us who do quantitative risk assessment on metals and metalloids have no problem
with the chapter, but a lot of people need a descriptive map for terminology and the process.
The author (s) should point out, in the beginning Sec. 5.1, that the chapter deals with
two generic components of the four typical components of an overall risk assessment as
defined by NAS and modified by EPA in such documents as the Risk Assessment Guidelines
for Superfimd (RAGS): Hazard Identification and Dose-Response Assessment. The remaining
two components are site-specific and require site-specific exposure assessment to do an overall
risk characterization.
Discussions of carcinogenicity and mutagenicity for the three mercurials are
appropriate to the current state of knowledge about such effects in humans or animals tested
via the standard NTP protocols.
I would agree that the current RfD for methylHg as given in IRIS and as applied in
chapter 5 may be problematic for protection against the more subtle neurotoxic endpoints such
as neurobehavioral teratology and other neurodevelopmental endpoints. The current IRIS RfD,
it should be noted, is based on the overt/clinically defined neurological and developmental
endpoints in March et al., 1987: cerebral palsy, altered muscle tone and deep tendon reflexes,
delays in walking and talking. Accommodating the probability of population-wide, subclinical
effects such as subtle neurobehavioral deficits is certainly justified, even if not using a specific
data base for that purpose. There is precedent in public health and regulatory policy. It is the
latter type of neurotoxic effects that drives the current regulatory and public health policies
for lead exposures in-utero, and in infants and toddlers as well. Ratcheting down the
acceptable exposure levels for methylmercury to accommodate newer data and an expanded
toxicity spectrum would parallel what has already been done for lead.
The revised RfD being proposed, 0.0001 mg/kg/day, is three-fold lower than the
current one. It appears reasonably grounded via use of an acceptable combined adjustment
D-97
-------
Mushak
factor of 10 on the 10% BMD of 1.0 /xg/kg/day. Other supporting data were presented to add
comfort in the selected revised estimate. Does this value protect against any potential subtle,
in-utero effects reflected in postnatal behavior and cognitive deficits? That may be difficult to
discern. The major prospective studies now under way among the Seychelles and Faroe
islanders entail methylHg from fish oor other marine life, a source of selenium as well. To
the extent that Se actually prevents effects rather than just delays them (a distinction lost on a
lot of people, apparently), these may not be the highest risk groups. The Iraqi poisonings may
well represent a more vulnerable risk group in terms of diet and the Se-Hg connection, but
this view should consider that grains, including organomercury-treated grain, can have Se
levels as well, especially if produced in seleniferous soils. For example, the 1976 Se report of
the NAS, using data of Morris and Levander, 1970, indicates 0.4 ppm (rounding off) wet
weight as the average for cereal products and 0.5 ppm wet weight (rounding off) for seafood.
The RfC for elemental mercury being proposed appears to be well grounded in the
various occupational epidemiology studies of workers showing subtle but still objectively
quantifiable effects. Absence of reproductive effects data in the selected studies for the RfC
merits use of a M.F. of 3 and sensitivity of individuals requires coverage with the U.F. of 10.
A DWEL of 10 pg/L and associated RfD of 0.0003 mg/kg/day appear to be
appropriate. In the calculations, use of the very sensitive Brown-Norway rat as an effect
species surrogate for autoimmune glomerulonephritis in a fraction of humans appears to be
reasonably protective for the latter. Assuming that this rat species is no less sensitive than the
most sensitive humans, this,in turn, requires only a 10-fold U.F. for interspecies and sensitive
humans combined.
Chapter 6. Risk Populations
One problem with this chapter, again, is its placement. It belongs before Chapter 5.
In the Intro material, risk populations should be defined as being at risk for either
intrinsic or extrinsic reasons. Extrinsic reasons include exposures or relationships to exposures
and perhaps nutritional status as well. For example, children at age 4 or less are at risk for
heightened uptake of toxicants and consequently more toxic effect if certain nutritional
deficiencies, common in that age band, permit higher uptake. Preschool children have a high-
risk relationship to their exposure environment, in terms of normal mouthing activity and
ingestion of potentially large amounts of toxicant-contaminated dusts and soils. Susceptibility
D-98
-------
Mushak
as a criterion for intrinsic risk definition includes both the normal developmental and related
vulnerability for effects as well as genetic susceptibility among members of a human
population.
The chapter should be expanded to include a separate section for risk groups by
exposure, if this is not to be done in Volume 3, and risk groups by virtue of gender,
developmental, genetic, etc., sensitivity.
One major gap hi this chapter in terms of exposure is omission of any discussion of
individuals with amalgam fillings, since such individuals cannot be ignored in any risk
assessment among high-risk groups. The fraction of such Hg intake dwarfs any other
contributor to aggregate intake among general populations environmentally exposed.
I have shown earlier, via simple estimates, that individuals with such fillings are exposed
above the RfC for elemental Hg. In extreme cases, exposure is substantially above this
reference value.
The chapter in a later portion should rank the various risk groups as to likely
significance of the factors producing the elevated risk. Acrodynia, which is more related to
exposure level probably than any inherent, genetic sensitivity, rarely occurs in the general
pediatric population.
Chapter 7. Interactions
In general, this chapter covers the major biological interactions of importance in
modifying the toxicity or toxicokinetics (e.g., uptake) of the mercury forms. I agree with the
focus on interactions being restricted to those with selenium, either as the inorganic or
bioorganic forms.
Chapter 8. Ongoing Research and Research Needs
This chapter needs some expansion to say a sentence or two about what's happening
with the research projects in Table 8-1. Furthermore, the section 8.2, Research Needs, needs
to be made specific instead of being boiler plate.
Appendix B. IRIS Summaries
Some of my comments on this Appendix appear earlier for Chapter 5 as well. Per
request for comments from reviewers of this chapter, via an EPA 12/20/94 memo, comments
respond to specific questions.
D-99
-------
Mushak
1. Was the weight-of-evidence characterization for carcinogenicity of the three relevant
mercurial forms done in accord with EPA guidelines and good science?
Overall, the characterization of carcinogenicity for the three forms was appropriately
done. Available data on mercurial carcinogenicity that I am aware of but not included in the
summary would not materially affect the weight of evidence decisions. The fact is that the
critical health issues for mercury risk assessments remain the noncancer endpoints.
2. Are the RfDs and the RfC properly calculated?
Assuming this refers to calculations based on the actual choice of studies, etc. for
evaluation, yes they were. I do not however, concur in the RfD for methylmercury, and
believe that the revised RfD described in Chapter 5 should be used.
3. Were the right critical studies and endpoints chosen?
Overall, yes. In the case of methylmercury, the potential effects of this form on subtle
neurobehavioral endpoints remains an environmental health issue and the revised RfD better
accommodates this possibility, even though the same data base was used..
4. Were the proper uncertainty and modifying factors used?
Generally, yes they were.
5. Was there sufficient detail provided for studies and evaluations supporting the various
risk values?
Overall, the level of detail is adequate.
SPECIFIC COMMENTS
Executive Summary
p. ES-1
The statement, par. 2, that small amounts of Hg are released from amalgam fillings, misleads
the reader to think this source is trivial compared to others. Not true. Revise wording. In par.
3, include those with amalgam fillings.
p. ES-2
The 1st full par, is too skimpy. Expand with better descriptions of effects.
D-100
-------
Mushak
Also, it should be revised to either define the number that constitutes a "high dose" or drop
it. Also, how serious do "serious" effects have to be? This paragraph is too flabby to be
meaningful.
The second par, should provide brief definitions of RfDs and RfCs.
The first sentence in par. 3 refers to susceptibility to mercury effects, as opposed to exposure.
The last par, does not provide helpful information as to exactly what further research and
studies are required.
Chapter 2
On p. 2-2. top portion, the statement about the chlorides as the environmentally dominant
forms is technically incorrect or, at best, misleading. Does this refer to the atmosphere
specifically and are there reliable speciation data to document this? The chlorides of mercuric
and methylmercuric ion otherwise would only predictably occur in the stomach of mammals
during hydrochloric acid hydrolysis of ingested materials. In sediments, water, plants, etc.,
these forms are bound up in complex ways. Needs rewriting.
Chapter 3
On p. 3-1. 3rd par., the statement that Hg entering the brain cannot return requires better
more precise writing for accuracy. Hursh et al. (1976) showed that adults inhaling Hg vapor
showed a 21-day clearance half-time from the brain area. While we know that brain Hg
remains elevated in workers after exposure ceases, this does not say that all mercury that
enters brain never leaves.
The last statement on p. 3-1 needs rewriting. What is this coating of mercuric sulfide?! On p.
3-2, 1st par., is a 1918 reference the only data on dermal uptake of elemental Hg in animals?
On p. 3-3. the dermal uptake rate should be clarified as to what the authors are calling
"absorption,"- uptake into blood or transdermal binding? The GI uptake of methylmercury
occurs via initial but short-lived formation of the chloride followed by uptake via binding
through sulfhydryl groups. "Lipophilicity" is a rather primitive concept nowadays; the text
should discuss relative ligand exchange equilibria between halide and sulfhydryl groups in
gastric and intestinal epithelium.
On p. 3-11. Scalp Hair, a better discussion of the hazards of contamination are needed. It is
an acceptable medium in cases where methylHg is to be measured in populations not having
D-101
-------
Mushak
significant atmospheric Hg deposition onto hair. The ratio of 250:1 is for hair-to-blood, not
vice versa.
Chapter 4
On p. 4-2. the last par, describes a study, Singer et al., 1987, that belongs with the inorganic
ionic mercury section, since it covers exposures to mercuric species existing as such.
On p. 4-3. line 3 from bottom, should be /xg/m3 , not mg.
On p. 4-4, 2nd par., what is a "preclinical psychomotor dysfunction"? When do psychomotor
dysfunctions become clinical? The material in the 4th par. seems to contradict that in par. 3,
since the extrapolated value of 0.025 for renal and neurotoxic effects is the same.
On p. 4-11. unboxed text, what does "...unexposed controls at comparable levels..." mean?
On p. 20. last par., what's "intermediate-duration" exposure as used in this study?
On p. 4-20. line 7. bottom, et seq. Is EPA proposing that magnetic fields are competitive with
elemental Hg exposure as a risk factor for ischemic heart and cerebrovascular disease in
workers?
On p. 4-25. Hematological. more should be said about this paper, given its apparent
suggestion of toxicity due to presence of amalgam fillings, or it should be deleted. This
particular author has published a series of papers claiming a variety of adverse effects due to
amalgam fillings that are difficult to accept without replication and plausible rationales.
On p. 4-35. and elsewhere, the ionic forms of mercury are called "inorganic mercury."
Elemental mercury, technically speaking, is also inorganic mercury. The entire report should
use a consistent designation for the two forms, e.g., "elemental mercury" and "ionic inorganic
mercury"
On p. 4-60. why so much detail on Kjellstrom's New Zealand study?
On p. 4-61. first par., what are the positive control exposure ranges used in the follow-up?
In the follow-up, Kjellstrom used the TOLD, the WISC and the McCarthy Scales. What IQ
tests does the author think would be better?
On p. 4-76. the box with proposed mechanisms needs to be updated and expanded. For
example, during the laying down and elaboration of the brain in-utero, the ability of
methylHg to disaggregate microtubulin, and presumably affect cell division and cytostructure,
should be noted as well.
D-102
-------
Alan Stern
D-103
-------
A. H. STERN
Vol. IV
General Comments
I have limited my written review to chapter IV of the Report to Congress on Mercury.
Within that chapter, I have provided specific comments on the Report proper as well as separate
comments on the IRIS documents. In some cases there is overlap between these two parts of
my review.
The Report in general and chapter IV in particular, is a huge undertaking and those
responsible for the work deserve praise. In general, I find the chapter to be very well written and
for the most part clear and well organized. In some places there appears to be some
unnecessary duplication of material and I have attempted to note these in my comments. I
believe that large parts of this chapter will serve as a basic reference for mercury health effects
for some time to come. It is, in some ways, unfortunate that this field is now in active flux and
that other parts of this chapter may need to be updated in the foreseeable future. I hope that
provisions can be made to integrate such changes into the existing structure.
I have found the IRIS documents to be generally somewhat less well organized and less
clear than the Report. I believe that, given the opportunity, they would benefit from editorial
review. An overall problem that I have found with these IRIS documents is that the rationale for
key decisions often remains unstated, or only briefly glossed over. These include choice of
sensitive species, choice of conversion factors (e.g. injection dose to ingestion dose), and
choices of input values to models. If the goal is to keep the entries on the IRIS database
concise, consideration should be given to preparing a supplement or appendix to the database
entry which presents detailed discussions of assumptions and rationale.
D-104
-------
A. H. STERN
Vol. IV
Specific Comments on the Report
pg. 2-1, f 2 Unless the basis for the apparent change over time is clearly shown and
judged to be real, rather than artifactual, this statement should be made only with very broad
qualifiers, or better, not at all.
pg. 2-1, table 2-1 The estimation of "retention" is confusing. Judging from the organic/fish
category and from the inorganic/non-fish category, it appears that "retention" is actually being
used to mean absorption. True retention is mediated by half-life and presence, or absence of
steady-state conditions as well as by absorption. The calculation of true retention in the average
adult would be a non-trivial undertaking and is not likely to be close to absorption. This applies
to organic as well as inorganic Hg.
pg. 2-1 ff.(2-2), last ^ Since most of the monomethyl Hg in the environment is tied up in
biological systems, it is unlikely that methyl mercuric chloride is the major organic compound.
pg. 2-2, 1 2 While Hg exposure from drinking water is generally quite low in most locations, it
should be noted that in at least some locations there is Hg contamination of groundwater which
can lead to a significant contribution to the overall exposure budget. In some affected areas of
southern NJ, total Hg in groundwater has an average concentration of 8.42 //g/l (median = 4.6
fjg/l). If 2 I/day consumption is assumed, then total mean intake from water would be 16.8
/yg/day (median = 9.2 yug.day). Most of this is reactive (inorganic Hg), but based on limited
speciation data, the mean organic Hg (in locations where organic Hg has been detected), is 0.02
jjg/\ and the mean volatile (probably elemental) Hg is 0.2 /jgfl These values are much higher
than that cited.
D-105
-------
A. H. STERN
pg. 3-1,12 What is meant by "rapid diffusion through the Gl tract"? The kinetics of diffusion
are not the issue, rather, the fraction absorbed.
pg. 3-1,^3 Methyl Hg is not particularly lipophilic. Although I have not been able to locate a
clear reference for solubility, its lack of lipophilicity is seen by the fact that in whale, methyl Hg
is found primarily in the muscle tissue rather than the blubber. This is also true for the fatty
portions of fish compared to muscle. Its rapid uptake and distribution is probably due to its
uptake by the methionine transport mechanism (Clarkson, T.W. Env. Health Perspec. 100:31-38
(1992); Mokrzan, E.M. et al. abstract presented at Whistler Conf. 1994)
It is true that methyl Hg accumulates in the brain and fetus, but the major site for
accumulation is the liver (Suzuki, T. et al. Arch. Env. Health 48:221-229 (1993).
pg. 3-2, 1 3 WHO (Env. Health Criteria 118, (1991) cites Binder for an estimate of 5%
absorption of inorg. Hg. This is a more recent reference than those cited for 7% absorption.
pg. 3-3, 5 4 As per previous comment, the statement about high lipophilicity of methyl Hg is
not correct. I do not have immediate access to the Halbach, 1985 reference cited for this, but
from the title I suspect that the octanol water partition coefficient cited is for methyl mercuric
chloride. As discussed previously, this compound is probably not a significant environmental
species (see Clarkson's 1992 Env. Health Perspec. ref. above).
pg. 3-3, ISA recent paper (Smith, J.C. et al. Tox. Appl. Pharm. 128:251-256 (1994)) suggests
an mean value of 7.7% of absorbed dose in blood (see my comment on pg. 5-19, 12).
pg. 3-5, 1 1 More recent estimates of methyl Hg half-life in blood give a range of about 35-70
days (Smith, J.C. et al. Tox. Appl. Pharm. 128:251-256 (1994); Cox, C. et al. Env. Res. 49:318-332
(1989); Kershaw, T.G. et al. Arch. Env. Health 35:28-36 (1980); Sherlock, J. et al. Human Toxicol.
3:117-131 (1984)) with a mean of about 50 days.
pg. 3-5, ^ 2 It is not clear that this ratio would be all that useful in distinguishing among organic
D-106
-------
A. H. STERN
Hq exposures since the ratio can also be affected by exposure to inorganic Hg. In fact, this
would seem to be the most likely cause of a change in the ratio. Thus, a more reasonable
"clinical" use of this ratio would be to distinguish inorganic from methyl Hg exposure. A decrease
in the RBC/plasma Hg ratio would be indicative of inorganic Hg exposure.
pg. 3-5, 1 3 There appears to be some disagreement within the literature regarding the
relationship between maternal and cord Hg levels. There appears to be a general trend toward
higher cord levels compared to maternal levels (WHO Env. Health Criteria 101, (1990). The
Grandjean reference cited actually says that cord blood levels are generally 20-65% higher than
maternal levels. Some studies, however, fail to find a significant difference. It may be that the
difference becomes more pronounced at higher levels of exposure.
pg. 3-5, ^ 5 Increased accumulation of Hg in neonates is probably due to inability to
demethylate methyl Hg during the first year of life (humans). This is probably due to absence
of necessary gut flora (Grandjean, P. et al. Env. Health Perspec. 102:64-77 (1994).
pg. 3-6, 1 1 The last sentence is unclear. Perhaps it should read "Blood and tissue levels of
mercuric Hg following exposure to high concentrations..."
pg. 3-8, t 1 Skerfving (Bull. Environ. Contam. Toxicol. 41:475-482 (1988) can also be added
as support for milk Hg reflecting plasma levels of inorg. Hg.
pg. 3-8,1 2 The third to last sentence should be changed to reflect that "...90% of the absorbed
dose of methyl Hg is excreted in the feces as mercuric Hg."
pg. 3-8, f 4 The Grandjean ref. regarding lack of neonatal excretion of methyl Hg is also
relevant here.
pg. 3-8,1 5 It is not clear to which ref. the statement that 60% of Hg in milk is methyl Hg refers.
Perhaps this refers to mice (Greenwood) and that may be correct. For humans, however, this
does not appear to be the case. Skerfving (see above) measured only about 16% methyl Hg in
D-107
-------
A. H. STERN
breast milk from Swedish mothers. This is critical in evaluating risk to neonates. The Skerfving
ref. is also appropriate relative to the relationship between Hg content of plasma and Hg in milk
in humans.
pg. 3-9, table 3-1 To the references for whole blood should be added Brune et al. Sci. Total
Env. 100:235-282 (1991). Estimates of total Hg for whole blood from pooled studies are made.
Means stratified by number of fish meals per week and varied from 2.0 /jgl\ for no fish
consumption, 4.8 jjg/l for <2 meals/week, 8.4 for 2-4 meals/week, to 44.4 jjg/\ for >4 meals/week.
Unknown fish consumption corresponded to a mean of 5.8 jjg/\.
pg. 3-10, 1 5 ff. The equation as given is incorrect (i.e. the units do not match on both sides
of the equation). A term for the volume of blood (in liters) is needed. Equation 3 in the Kershaw
paper is correct. However, that equation is somewhat cumbersome and perhaps confusing. The
term b is often used to denote the kinetic rate constant while C is more usually used for the
concentration. The term T1/2/ln2 is simply the kinetic rate constant. Perhaps a more useful form
(based on Kershaw) would therefore be
C = d(f/V)(b)
where C = concentration (pg/l)
d = intake (pg/day)
f = the fraction of absorbed dose in the blood (unitless)
V = volume of blood (I)
b = the kinetic rate constant (day1).
pg. 3-11,^2 The ratio of 250:1 should be given with the appropriate units i.e. fjg Hg/g hair/mg
Hg/l blood.
I do not have the Birke, or Skerfving references. However, I do not believe that variability
in estimates of the blood to hair ratio results, for the most part, from measurement difficulties, or
problems. Perhaps in these early studies analytical problems were more significant. In general,
the major sources of variability appear to be true interindividual variability in this ratio. Another
possible source stems from the temporal discrepancy between blood levels, reflecting relatively
D-108
-------
A. H. STERN
recent exposures and hair levels, reflecting largely historic exposures. Populations whose intake
varies significantly over time (e.g. seasonally) will have a greater discrepancy than populations
whose intake is constant.
pg. 4-9, table 4-1 ff. The collum labeled "Exposure Frequency" in this and similar tables should
more properly be labeled Exposure Duration. Also, the collum labeled dose actually refers to
concentration. Quantification of the dose requires information on the inhalation volume.
pg. 4-35,1 2 Phenyl mercuric acetate and mercuric acetate appear to be organic rather than
inorganic compounds.
pg. 4-36, f 2 The results summary refers to a 0.66 mg Hg/kg/day group, but does not list this
dose among those in the experimental description.
It is not clear why this is considered to be "the only chronic oral study designed to
evaluate the oral toxicity of mercury salts." The 1993 NTP studies appear to be equally valid as
chronic toxicity studies.
pg.4-57, t 3 and ff. It is not clear whether the possibility of maternal effects as opposed to true
dominant lethal effects is raised here with respect to the Suter study only, or with respect to the
Zasukhina study as well.
pg. 4-59 and ff. I think it would be helpful to have distinct subsections for neurotoxic effects
(adult) and developmental neurotoxic effects for methylmercury.
pg. 4-59, f 1 Strictly speaking, there is no such compound as methylmercury. Mono
methylmercury is an ion (CH3-Hg"1'). It is therefore confusing to list methylmercuric chloride, or
methylmercuric hydroxide as "other organic mercury compounds," since they are merely salts of
the methylmercury ion.
Phenylmercuric acetate is included here as an organic compound and on page 4-35 as
D-109
-------
A. H. STERN
an inorganic compound.
pg. 4-59 and ff. I agree that the in utero developmental effects are the most sensitive of the
methylmercury endpoints. However, the studies which concentrated on adult effects in Japan
and Iraq should be given more consideration and should include Clarkson's 1976 paper (Fed.
Proc. 35:2395-2399).
pg. 4-59, \ 3 I have looked again at the key papers and it does not seem to me that
neurological symptoms were actually "observed" in subjects with Hg blood levels as low as 200
/jgl\. That value is the estimated lower concentration limit corresponding to effects. It was
derived from several different epidemiologic and pharmacokinetic approaches. It is, however, an
historical back extrapolation. A more accurate statement might be that CNS effects have been
estimated to be associated with blood concentrations perhaps as low as 200 /jg/\.
pg. 4-59, ^ 4 There seems to be an implication that maternal paraesthesia was associated with
developmental effects in fetuses. This is true for some effects, but not for others. Even where
statistical associations are significant the developmental effects are associated with maternal
paraesthesia only slightly more than 50% of the time.
Strictly speaking, fetuses were not exposed to maternal blood levels of Hg since maternal
blood does not cross the placenta and because Hg from maternal blood may be preferentially
retained in the fetus. This relationship should more properly be referred to as an association
between fetal exposure and maternal levels.
pg 4-59, t 5 and ff. The term "indices of exposure" as used here refers to maternal hair Hg
level, however this is not clear in the text.
The summary of the association between maternal methylmercury exposure and
abnormalities in deep tendon reflexes in boys seems confused. In particular, I cannot find the
source of the 7.2% incidence rate for boys cited in the text. Upon rereading the paper in some
detail, I think that the eighth sentence in this paragraph (The investigators found that when...),
D-110
-------
A. H. STERN
might be replaced with a clearer summary of the association as follows: Abnormality of muscle
tone or reflexes showed a significant positive association with maternal Hg exposure for boys,
but not for girls. A consistent dose-response relationship for this effect was not observed.
However the greatest prevalence of the effect in boys occurred for those with mothers in the
highest exposure group (13.0-23.9 ppm Hg in hair). No other measure of abnormal or
decreased neurologic function or development showed a significant positive associated with
maternal Hg exposure. The text might then continue with the sentence beginning The
discriminant analysis..."
The statement about the possible influence of smoking and alcohol consumption does
not appear to be supported by the paper. In fact, the discriminant analysis referred to in the
report text, suggests that these potential influences were not confounding factors in this analysis.
pg. 4-60, \ 1 The statement that in 17 of the 31 matched pairs in the first phase of this study,
the referent child was older than the exposed child by as much as 0.7 years has, unfortunately,
often been repeated as a criticism of this study. While there appears to be some confusion, this
does not, in fact, seem to be correct. The report of Kjellstromo et al. states (pg. 43) that "The
children's (matched pairs) birthdates were always within 30 days of each other." The report refers
the reader to the table in appendix 10 which presents the difference in days and confirms the
statement in the text. Based on appendix 10 in only 10 of the 31 pairs was the referent child
older than the exposed child and the maximum difference is given as 29 days (mean 12.4 days,
median 11 days). The confusion arises from appendix 11 which presents the summary DDST
results. Here, the ages of the exposed and referent matches are recapitulated, this time in years.
In this table the referent match is, indeed, older than the exposed match by a maximum of 0.7
years. Clearly, one of these tables in incorrect. While any such inconsistency is cause for some
doubt, the agreement of the explicitly stated maximum difference of 30 days in the text and the
table in appendix 10 indicates that it is appendix 11 is incorrect.
pg. 4-61, f 1 The high exposure children were each compared with three matched control
children, not, as stated, three control groups.
D-lll
-------
A. H. STERN
I do not believe that the last two sentences in this paragraph, which attempt to summarize
the weaknesses and limitations of both phase of the Kjellstrom et al. study, are necessarily valid
or meaningful, particularly with respect to the second phase of the study. First, based on the
description of the tests provided by the authors, it is not correct to characterize all the tests as
intelligence tests. In addition to intelligence tests perse, the battery also contained tests of fine
and gross motor coordination, social adjustment and language development. The two tests
which were ultimately selected for complete multiple regression analysis, were an intelligence test
(WISC-R) and a test of language development (TOLD-SL). Second, I agree that "these tests may
not be the most appropriate for defining the effects of methylmercury." However, I believe that
this limitation applies to all the endpoints investigated in each of the studies in the literature
(human and animal). This is because methylmercury's effect on neurologic development is to
alter brain architecture, apparently on a broad scale. The evidence indicates that the effects of
this alteration include physical CNS functions (e.g.(motor control, reflex response) as well
cognitive functions (e.g. language development, learning). Given this broad range of effects, it
is not clear that any of the tests reported in the literature can individually be considered most
appropriate for defining, or quantifying the effects of methylmercury. The situation is like the
parable of the blind men and the elephant. Each feels a different part of the animal and comes
to a different conclusion about the nature of the beast. Given this situation, I do not believe that
this is a particular limitation of the Kjellstrom study. Furthermore, I believe that the understanding
of this problem provides a useful context for interpreting the fact that methylmercury accounted
for only a small fraction of the variability in the overall model (2.0-2.5%). The fact that there does
not appear to be a single test which is specific for the developmental effects of methylmercury,
makes the observation of a statistically significant effect of any magnitude on a (partially)
indicative test qualitatively indicative of larger-scale effects. Finally, the observation that "greater
significance was seem in differences of cultural origin of the children than'the differences in
maternal hair methylmercury concentrations" is not a limitation in the investigation of the effect
of a given independent variable in a multiple regression analysis. There is no a priori reason to
assume, or expect, that the variable under study will be the one with the largest effect in the
model. That is, after all, the reason for carrying out multiple regression analyses. The only
criteria for concluding that an independent variable is associated with a qualitative effect on a
dependent variable should be that the model is legitimate and that the observed effect in the
D-112
-------
A. H. STERN
model is significantly different from zero. The same situation applies in studies of the effect of
lead on behavior and cognition.
pg. 4-61, 5 2. The Tsubaki et al. paper should be used as part of a more complete discussion
of adult health effects. In particular, the observation that disease symptoms were associated with
hair levels as low as 50 fjg/g should be noted.
There appears to be some confusion about the Tsubaki et al. reference. The reference
is given relative to a discussion on this paragraph about disease in Minimata is Tsubaki 1977.
However, the references in the report do not list a 1977 paper, but a 1978 paper by Tsubaki. The
1978 paper presents data from Niigata and not from Minimata.
The recent paper by Delgard et al. (Env. Health Perspec. 102:548-550 (1994)) should be
mentioned here. They use data generated in the Faroe Islands on the relationship between
mercury levels in umbilical cords and mercury levels in maternal hair, to estimate the maternal
exposure in diagnosed cases of congenital Minimata disease based on mercury levels in
preserved umbilical cords. This work has several significant limitations which are discussed in
the paper. However, it is of interest because 42% of the estimated hair levels of mothers of
children diagnosed with congenital Minimata disease are in the range of 10-20 ppm and are thus
consistent with the other human studies of developmental effects discussed in the report.
pg. 4-61, 1 3 The 1994 Marsh et al. paper does not appear in the references. To the best of
my knowledge, it has not yet been published and has been made public only as an abstract at
the 1994 Hot Springs conference.
pg. 4-62, f 1 and ff. The previous comment also applies to the references to the Davidson et
al.; Myers et al.; Cox et al.; and Shamlaye et al. 1994 papers.
pg. 4-62, 5 5 The description of the Faroe Islands study is somewhat misleading. The study
was not on the effects of methylmercury and PCBs on neonates, but on neurologic
developmental effects of methylmercury and PCB exposure in utero.
D-113
-------
A. H. STERN
No reference is given and, as above, I am not aware of a published paper.
pg. 4-63, f 2 and ff. Two other studies of methylmercury exposure in macaque monkeys
should be discussed. Burbacher et al. (Neurotox. and Terat. 12:65-71 (1990) found effects on
social behavior in macaques exposed in utero at a maternal dose of 0.05 mg/kg/day. Burbacher
et al. (Developmental Psychol. 22:771-776 (1986) found retardation in object permanence
development in macaques exposed in utero at a maternal dose of 0.05 mg/kg/day. This appears
to be a measure of cognitive development.
pg. 4-63, 1 4 The earlier study of Musch et al. (Arch. Toxicol. 40:103-108 (1978)) should be
noted as supporting an effect at a maternal dose of 50 mg/kg/day in this test system.
pg. 4-66, 5 2 The lowest dose as mg Hg/kg/day for the males should probably be 0.03 rather
than 0.3.
pg. 4-70,1f 3 Igata's recent paper (Igata, A. Env. Research 63:157-169 (1993)) should be added
to the references for symptoms of Minimata disease and also to the discussion on pg 4-72 ff.
pg. 4-73, ^ 3 The summary of thresholds given here is taken from WHO 1990. The WHO 1976
document (Env. Health Criteria 1) has a slightly different summary based on the Japanese and
Iraqi studies. There, the threshold levels in blood are estimated at 200-500 fjg/m\. It is not clear
whether the difference is based on a specific reassessment in the later document.
This discussion should also mention the uncertainty (at least in the Iraqi data) resulting
from the back-calculation of biological indicator levels to the time of exposure.
pg. 4-78,11 The narrative for renal effects of methylmercury gives quantitative information only
for the two studies with the highest LOAELS. The information in table 4-60 provides a very
different picture.
pg. 4-83, ^ 1 The reports of the Kjellstrom studies are available from the Swedish government.
D-114
-------
A. H. STERN
Particularly because they are controversial, they should not be reviewed or cited on the basis of
the summary in the WHO document.
pg. 4-83, ^ 1 Strictly speaking, it is correct that a dose-response relationship was not observed
in the McKeown-Eyssen study. However, the association between maternal Hg exposure and
neurologic abnormalities in boys was seen only in the group with the highest levels of maternal
exposure. This is consistent with a dose-response relationship.
pg. 4-85 While not strictly a controlled study, the work of Burton et al. (Env. Research 14:30-34
(1977)) deserves mention. Wild mice of the same species from four geographic populations in
Utah were trapped and characterized by Hg levels in their fur. Each population had a
characteristic mean Hg level reflecting differential consumption of dipterous insects which
accumulate Hg (presumably methylmercury). Swimming ability and open field performance
demonstrated a clear dose-response relationship with respect to Hg hair levels.
pg. 5-3, 1 3 I do not believe that the statement that derivation of RfDs and RfCs requires
adjustment of NOAELs and LOAELs to lifetime exposure conditions (24 hr/day, 70 years) is
correct. Both RfDs and RfCs specifically refer to chronic exposure and as discussed
subsequently (pg. 5-4, box) adjustment of less-than-chronic studies is accomplished through the
use of uncertainty adjustments. Furthermore, ingestion RfDs inherently assume that while
exposure may occur daily, it does not occur continuously. Inhalation exposures may occur
continuously. However, I do not believe that RfCs (unlike unit cancer risk slopes) are based on
continuous exposure. Rather, they are intended to reflect chronic exposure.
pg. 5-3, \ 5 The explanation of why the UF for extrapolation from animals to humans for the RfC
has been set to 3 does not follow logically. The fact that half of a UF of 10 on the log scale is
3, does not explain why it is appropriate to divide the previously used UF in half.
pg. 5-8, ^1-21 have read the Fredriksson et al. paper in some detail and it is unclear to me
why the conclusion in the second paragraph applies to that paper. It does not appear to me that
the Fredriksson paper suffers from incomplete reporting. The meaning of the term "lack of a
D-115
-------
A. H. STERN
replicative study design" is not apparent to me, but I do not see any particular weaknesses in the
Fredriksson study design. The only significant criticism of that paper which I can envision is the
fact that Hg vapor exposure occurred to neonatal rats rather than to fetal/embryonic rats.
However, the rationale given by the authors for this design - that the greatest period of brain
growth occurs postnatally in rats, as opposed to in utero in humans - seerr.s reasonable and
should not change the interpretation of this study as a developmental study. Although I am not
qualified to judge the appropriateness of the developmental tests or their interpretation, it appears
to me that this paper can reasonably provide the basis for a positive evaluation of developmental
toxicity for elemental mercury. At very least, a mojje thorough treatment and critique of this paper
is warranted.
pg. 5-14, 5 6 (last) It is stated that proteinuria occurred for all doses greater than 0.1 mg/kg,
however, this dose is not listed among those administered in this study. Since the LOAEL is
reported in terms of mg/kg/day, it would be useful to report the doses in this form for
comparison.
pg.5-15, 5 1 The only (non-control) dose level recorded in this study is 3 mg/kg/week. This
corresponds to 0.428 mg/kg/day. It is not clear how a LOAEL of 0.317 mg/kg/day could be
derived in this study.
pg. 5-17, 1 3 Since the term "benchmark dose" has no precise quantitative definition, it is
necessary to state the percent response corresponding to the specific benchmark. In this case,
I believe that this benchmark represents a 1 0% response.
pg. 5-1 7, 1 4 ff C the maternal blood concentration corresponding to the benchmark dose in
hair is given here as 44 /jgl\. This is indeed the correct value given a hair concentration of 1 1
pg/g and a hair/blood ratio of 250 /jg/g/mg/l. However, the RfD document employs a value of
47
pg. 5-17, 1 5 (formula) The formula for calculating C, the concentration of methylmercury in
blood, is incorrect. As given, the units do not reduce to /jg/l. For this to occur, rf must be in
D-116
-------
A. H. STERN
units of //g/day, rather than fjg and bw must be eliminated from the equation (the assumed body
weight is reflected in the assumptions of blood volume). This gives the following equation:
i
C = A x f x d (ag/dav) = /jg/liter
b (days) °xV (liters)
Solving for d gives
d = Cx bx V = //g/day
Axf
To convert this to a dose (//g/kg/day) requires that this quantity be divided by the assumed body
weight. Alternatively, c/can be expressed in terms of //g/kg/day although I have not found this
to be the usual practice.
pg. 5-18, 1 1 As suggested in the discussion of the Phelps et al. study, 1 think that a major
reason for variability in the hair-blood relationship for mercury concentration is the fact that
unsegmented hair analysis gives a time-weighted average of mercury exposure while analysis
of mercury in blood reflects a much shorter period average of exposure. Where seasonal, or
episodic variations in fish intake occur, the value of this ratio will vary. From my reading of the
literature, I do not believe that location on the body from which the hair sample was taken is a
significant factor in this variability. I am not aware of any important study in which hair was
sampled from any location other than the head. In some studies, hair was sampled from the
back of the head, while in other studies hair was taken from the sides, or front. However, these
differences are not likely to result in significant variability in this ratio among the studies.
It is important for this ratio to be given with the proper units (jjg Hg/g hair//>g Hg/ml
blood) because these units are not necessarily intuitive and because these units are not
necessarily the ones reported in the literature, particularly for blood.
One could reasonably derive a value of 250:1 (hainblood) from the Phelps et al. study.
D-117
-------
A. H. STERN
However, I think that a stronger argument supporting the same value can be made from
consideration of the studies cited in WHO'S 1990 criteria document. There, the average of the
mean values of the ratio for total mercury for the 10 studies cited is 265 (S.E. = 63.8). Assuming
that (based on Phelps et al.) this ratio is about 6% smaller if only organic mercury is considered,
the average of the means of these studies is 249.
pg. 5-19, 5 1 Yannai and Sachs (Food Chem. Toxicol. 31:351-355 (1993)) recently reported on
mercury absorption in rats fed fish/seafood containing intrinsic mercury. Based on broad data
on the fraction of mercury in fish/seafood present as methylmercury, this can be assumed to be
essentially 100% methylmercury. Rats fed a commercial fish meal preparation (Hg 0.04 ppm)
absorbed 75% of the mercury and rats fed an experimental preparation of fish and shellfish (Hg
0.2 ppm) absorbed 93% of the mercury. The lower absorption of the commercial fish meal is of
interest and may be related to the quite low mercury concentration. The absorption of mercury
from the higher concentration diet is consistent with the value assumed in the workgroup
document.
pg. 5-19, 1 2 A recent paper by Smith et al. (Toxicol. Appl. Pharm. 128:251-256 (1994))
presents a detailed study of methylmercury excretion kinetics based on measurement of i.v.
administered methylmercury (1.7-7.4 jug) in blood, urine and feces of 7 male volunteers. The
authors claim that data from this study are superior to those from previous studies in accounting
for the portion of the label which metabolized to inorganic mercury. Based on the linear
extrapolation of the plot of blood concentration of methylmercury versus time, they calculate that
approximately 7.5% of the methylmercury is in the blood following rapid equilibration among
tissue compartments. Based on fitting the experimental data to a five compartment
pharmacokinetic model, they calculate that 7.7% (geometric mean) of the methylmercury is found
in the blood. Given that previous studies did not correct for that fraction of the dose which
during the time course of the study was no longer present as methylmercury, these data may
reflect a more accurate estimate of this parameter. It should be noted however, that the values
for this parameter among the 7 subjects ranged from 6.5-9.5%.
pg.5-19, ^ 3 The only significant route for the elimination of methylmercury is via conversion
D-118
-------
A. H. STERN
to inorganic mercury by gut flora and elimination in the feces. Thus, more than other parameters
inputs in this model, the determination of this parameter value is likely to reflect significant
population variability resulting from differences in nutritional and health status which influence the
activity of these specific gut flora.
The value of 0.014 reported in Cox et al. (1989) is based on analysis of Hg in hair. The
median half-life is reported as 48 days. This corresponds to a value for b of 0.0144 day1. Based
on figure 4 of that paper, the range of half-lives was about 18-73 day'1. The mean value is not
reported, but based on a Monte Carlo simulation of the data in figure 4 (estimating values from
the y axis) I estimate that the mean value is about 47 days. The most frequently reported value
(mode), however, is 55 days which corresponds to value for b of 0.013 day'1. The 1990 WHO
Criteria document points out that although half-lives calculated from hair closely follow those
calculated from blood, they have a larger variability. WHO suggests that this may be due to
analytical variability in hair analysis. However, I believe that the uncertainty in estimating hair
growth rate may also be a significant source of uncertainty. In either case, the value of 0.014
day1 reported from the Cox et al. paper does not appear to have a unique claim on "accuracy".
Furthermore, even though this model is being used to derive a dose-response relationship for
the Iraqi maternal-child cohort, all of the other inputs to the calculation of d are derived from
various and different human populations. Therefore, the value of the elimination rate constant
derived from this study should not be given particular weight merely because it is cohort-specific.
Bakir et al. (Science 181:210-241 (1973)), reporting on the general population affected in the Iraqi
epidemic (16 hospital admissions), measured a mean half-life in blood of 65 days and a range
of 40-105 days. These observations may be biased toward somewhat larger values because they
represent individuals whose toxicity was severe enough to result in a hospital admission. Severe
toxicity may select for those with longer half-lives (and thus, slower excretion) for methylmercury.
Smith et al. (reference above) taking conversion to inorganic mercury into account, report an
overall estimate (geometric mean) of the half-life in blood (methylmercury-specific as per
discussion in previous paragraph) of 45 days (0.015 days"1). The relatively small sample size in
this study (n=7) should be noted. Nonetheless, even for this small population, the half-lives
ranged from 35 days (0.020 days'1) to 53 days (0.013 days*1). WHO (1990) references studies
by Kershaw et al. (1980) and Sherlock et al. (1984). Both studies were based on controlled fish
D-119
-------
A. H. STERN
consumption. The former from a single meal, and the latter from controlled chronic intake. The
respective means and ranges are 52 (39-67) days and 50 (42-70) days. Since both of these
studies followed blood rather than hair Hg levels, the variability is likely to largely reflect true inter-
individual variability rather than analytical or methodological uncertainty. Given the ranges seen
in all the studies, one cannot properly speak about the choice of an appropriate input value in
terms of accuracy. The choice of which value to use for the elimination rate constant will largely
reflect policy considerations rather than scientific considerations. The choice is between a
"typical" (average) value and a reasonably "inclusive" (upper percentile) value. The choice is
dictated by the degree of conservatism which the model output is intended to reflect and, to my
knowledge, this has not been articulated in the RfD guidelines. If a typical value is desired, the
half-life of 53 days from Kershaw et al. (corresponding to a b value of 0.013 day"1)lseems
reasonable. This choice gives some weight to the higher value from Bakir et al. and to the most
frequently reported value from Cox et al. On the other hand, if an inclusive and reasonably
protective value is intended, then I believe that the value of 70 days (b = 0.01 day1) used by
WHO (1976) for both the adult (paraesthesia) and developmental endpoints, is clearly
appropriate. A case can be made for using a reasonably conservative value from the observation
that the half-life may be the largest source of inter-individual variability in this model and that
given this variability, no is value is truly typical.
In this context, it is interesting to note that Smith et al. state that the risk associated with
dietary methylmercury may have been overestimated in previous assessments because larger
values of the half-life (smaller values of the elimination rate constant) than that derived in their
study were employed in calculating intakes corresponding to adverse effects. This, by itself,
would result in a larger estimate of the dose (o) corresponding to the blood concentration (Q
associated with observed adverse effects. However, their estimate of the fraction of the dose
which is contained in the blood (7.7%) is also different from the value which was used in previous
assessments (5%) (see comment on pg. 19 ^ 2). If both of the values derived by Smith et al. are
employed in this calculation, the relationship between otend Cis essentially unchanged from that
employed in previous assessments.
pg. 5-19, 1 5 Ideally, the calculation of c/from C would be specific to the population which
D-120
-------
A. H. STERN
generated the dose-response relationship (the Iraqi maternal cohort). In this case, however, the
parameter with the largest uncertainty, the hair/blood ratio, is completely generic and is not
based in any significant way on the study population. Thus, this calculation is, of necessity,
generic and other input values specific to this population should have no more weight than those
from other populations. Therefore, even if body weight data for Iraqi women were available, 60
kg would still be a reasonable estimate for this variable.
pg. 5-19, 5 6 - 5-21, t 1 Since this section largely reiterates (albeit in more detail) material
already presented on pg. 5-16 to 5-17, it is somewhat confusing in terms of the overall structure
of the document. It appears that the reiteration of this material partly, but not completely follows
the structure of the IRIS document. That however, is not necessarily a model of form. I believe
that the material on pages 5-19 to 5-21 should be integrated with the similar material on pages
5-16 to 5-17.
pg. 5-21, table 5-2 This table is not self explanatory and should have more informative labels.
pg. 5-21, ? 2 While I agree in principle with the overall modifying factor of 10,1 believe that one
of the most important justifications for this value is the possible overestimation of the effective
maternal dose in the Marsh 1987 paper. Briefly, maternal dose was defined as the peak hair
concentration. However, assuming that some periods of in utero neurologic development are
more sensitive to the effects of methylmercury than others, there is no reason to assume that
these periods coincided with the time of peak hair concentration. Therefore, use of peak hair
levels may overestimate the effective dose. Furthermore, different members of the cohort were
at different points in gestation at the time that exposure began. Thus, some may already have
passed critical periods of development prior to the onset of exposure and some may not have
reached these critical periods until the peak period of exposure had passed and maternal
concentrations were in decline. Therefore, there may be significant misclassification of exposure
in the cohort, with some of the cohort effectively unexposed. This would also result in
overestimation of the effective maternal dose.
The last justification given for the value of the overall uncertainty factor, lack of data for
D-121
-------
A. H. STERN
possible manifestations of adult paraesthesia observed during gestation, is unclear. Does this
refer to some influence of maternal paraesthesia on the developmental endpoint, or is it a
reference to an undetermined risk of adult paraesthesia at the benchmark dose?
pg. 5-21, ] 4 ff (the equation on pg. 5-22) This is quite confusing. The previous paragraph
states that an overall uncertainty factor of 10 was used with a modifying factor of 1. Here we see
an overall uncertainty factor of 30. Furthermore, I don't understand the origin of the benchmark
dose of 3.4 /^/kg/day.
As per my preceding comments on input values to the pharmacokinetic model for the
calculation of d, if b is taken to be 0.013 day"1 (53 days) and f is taken to be 7.7% (0.077), with
all other inputs as per the values in the report (i.e. C= 44 /jg/l as per the report, rather than 47
/yg/l as per the IRIS document), then c/is calculated to be 0.65 /jg/kg/day (RfD = 0.07 /jg/kg/day).
If b is kept at the value in the report of 0.014 day1, then d calculated to be 0.70 pg/kg/day. This
is a small, but significant difference relative to the value of 1.0 /jg/kg/day derived in the report.
As discussed previously, I believe that this change is justified on the basis of the new and more
specific data on the fraction of the methylmercury dose in the blood (/) reported in Smith et al.
pg. 5-22, ^ 2 I believe that there is considerable uncertainty inherent in the Marsh et al. (1987)
data base and in the model-based derivation of a specific benchmark dose from those data.
Nonetheless, viewed in the context of other RfD derivations, there are many things about this
derivation which place it among the least uncertain of those done to date. These include data
from a human population with a good exposure metric and the knowledge that the chemical in
question can cause the critical effect(s) in human populations at some dose level. I believe that
these observations justify the medium confidence levels within the context of IRIS. On the other
hand, there is recently completed as well as nearly completed studies which may cast new light
on this RfD derivation. Because it is not clear that these studies will, or should alter this
derivation and because it may be some time before these studies are available, interpreted and
placed in perspective relative to each other and to the pre-existing literature, the issuance of a
revised RfD for developmental effects at this time is warranted. However, in light of the
uncertainty in the current assessment and in light of the evolving nature of the evidence, I believe
D-122
-------
A. H. STERN
that this RfD should be classified as an interim RfD whose review will continue.
pg. 5-24, 1 5 Regarding developmental effects of elemental Hg, see comment on pg. 5-8, f
1-2.
pg. 5-25, 1 3 The dosage series is given here as mg Hg/kg/day (i.e. already converted from
mercuric acetateto Hg). The NOAEL is reported as 5 mg/kg mercuric chloride, or 2 mg/kg Hg.
This dose (as Hg) is not among those in the series. Furthermore, there seems to be confusion
as to the compound tested.
pg. 5-27, 1 2 The report is vague as to the daily fish consumption level used to derive the
hypothetical FDA RfD. This assumed value should be presented explicitly. Based on the FDA
standard and the hypothetical RfD as given in the text, I calculate the assumed daily fish
consumption level referred to in the text to be 28 g fish/day. The provenance of this value is not
clear to me.
pg. 6-1.11 ff. (developing fetuses) I have been unable to locate justification in the 1990 WHO
document for the statement in the text that estimates of fetal (developmental) toxicity are 2-3 fold
lower than for adults. In fact the discussion in section 10 (Evaluation of Human Health Risks) of
the WHO document seems to imply fairly clearly that it estimates a 4-5 fold greater toxicity to the
fetus. This can also be seen from the WHO estimates of threshold blood concentrations (50 jjg/\
maternal concentration for fetal toxicity, versus 200 //g/l for adult toxicity). Other references also
support this range (e.g. Clarkson Env. Health Perspec. 100:31-38 (1992)).
A third possible reason for the increased sensitivity of the fetus is the lack of
methylmercury excretion in the fetus or neonate (Grandjean et al. Env. Health Perspec.102:74-77
(1994)).
pg. 6-2, 11 I do not have the Rice 1989b paper referred to here. However, since
methylmercury absorption in adults is close to 100%, I do not see how absorption could be much
higher in infants.
D-123
-------
A. H. STERN
pg. 6-2, ^ 3 The recent paper of Igata (Env. Research 63:157-169 (1993)) should also be cited
in support of the aging process unmasking previously sub-clinical effects. Igata discusses the
occurrence of "late onset Minimata disease" as a function of aging.
pg. 7-1,^5 I do not have the Satoh et al. paper, but I have read the Nishikido et al. paper on
the effect of selenium on the developmental effects of methylmercury in some detail. This paper
does not clearly support the statement that selenium protects against the developmental effects
of methylmercury. Although maternal selenium exposure did protect against the fetotoxicity of
methylmercury, even the highest maternal dose of selenium in this study, failed to prevent the
decrease in fetal growth caused by maternal methylmercury exposure. Furthermore, the adverse
developmental effects examined in this study, lethality and growth retardation, are not those of
concern for the Reference Dose. In fact, the partial protection against the methylmercury
mediated decrease in fetal liver glutathione peroxidase activity given by maternal selenium
administration suggests that the adverse fetal effects examined in this study result from oxidative
damage which is similar or identical to that which may mediate adult methylmercury toxicity, but
may not be directly responsible for the developmental effects on brain architecture. Magos
(Advances in Mercury Toxicology. T. Suzuki et al. eds. Plenum Press. 1991), in a review of the
available literature on the protective effects of selenium on methylmercury toxicity (including the
Nishikido et al. paper), concludes that maternal selenium exposure provides no protection for the
fetus. His review, however, is also based on consideration of gross developmental abnormalities.
More recently, Fredriksson et al. (Pharm. Toxicol. 72:377-382 (1993)) examined the effect of
maternal selenium exposure (1.3 ppm in diet) and methylmercury (2, or 6 mg/kg on days 6-9 of
gestation) on neurologic developmental motor endpoints in rats. No gross signs of
developmental toxicity were observed. Selenium exposure resulted in significantly elevated Hg
levels in blood of the offspring, but had no effect on brain levels. Maternal selenium antagonized
the effect of methylmercury on some, but not all the developmental motor endpoints when
compared to the rats exposed maternally to selenium, but not to methylmercury. However, the
interpretation of these results is complicated by the observation that selenium alone, results in
decrements in these endpoints. In total, I do not believe that any conclusions can be drawn at
the present time regarding the influence of selenium exposure on the developmental toxicity of
methylmercury.
D-124
-------
A. H. STERN
IRIS Documents
NOTE: Much of my evaluation of the IRIS documents has already been included in my specific
and general comments on the report proper. Rather than duplicate those comments, I will refer
to them where appropriate.
Elemental Mercury - RfC
Overall, I believe that the database supporting this RfC is unusually strong and consistent.
However, I think that the assessment suffers from the lack of any discussion of the uncertainty
in the use of the conversion factors and whether this uncertainty is adequately addressed in the
uncertainty adjustments.
The four studies used as the primary sources are impressive in their close agreement on
a LOAEL concentration. Of some concern, however, is the fact that the Pikivi and Toulonen and
Pikivi and Hanninen studies addressed conditions of mixed exposure (i.e. a chlor-alkali plant).
In Fawer et al. and Pikivi study, the nature of the exposure is not stated. It is important to
consider to what extent exposure to other toxicants may have been responsible for some of the
adverse responses.
The conversion from occupational exposure to general population exposure is fairly
standard and generally reasonable. However, there should be some discussion of the
uncertainty associated with this conversion. In particular the assumption of 20 m3 as a 24 hr
inhalation volume is fairly conservative. This should be acknowledged.
The conversion factor (blood concentration to air concentration) is critical for the
derivation of the RfC. However, there is no discussion of the confidence and reliability
associated with this factor. Some discussion of the uncertainty and interindividual variability
associated with this factor seems necessary.
Given the lack of discussion of the uncertainty associated with the blood concentration
D-125
-------
A. H. STERN
to air concentration conversion factor, the appropriateness of the overall uncertainty factor (UF)
adjustment is unclear. I agree with the use of a UF of 10 to account for sensitive populations
and the use of a UF of 3 to address the inadequate database for developmental effects.
However, I do not know whether a UF is needed to address uncertainty in the conversion.
As discussed in my comment for pg. 5-8,1 1-2,1 believe that insufficient consideration
has been given to the developmental study of Fredriksson et al.
Elemental Mercury - Carcinoqenicitv Assessment
I agree with the weight of evidence assignment of elemental mercury to group D
(insufficient data). The data on human carcinogenicity are little more than mildly suggestive.
Inorganic Mercury (Mercuric Chloride) - RfD
Overall, I believe that too much of the underlying rationale for this RfD is unstated or
implied for this document to be evaluated as a self-contained justification for this RfD.
It is not clear why it was necessary to back-calculate from the Drinking Water Equivalent
Level (DWEL). If the major studies are evaluated in the IRIS document anyway, why is it not
appropriate to derive the LOAEL in the IRIS document. The DWEL can be cited as a
corroborating assessment. Merely back-calculating from another assessment obscures some
of the scientific reasoning.
Since the Druet et al. study exposed rats via subcutaneous injection, the derivation of a
LOAEL for oral exposure must have involved a conversion factor. This should be described.
The first sentence of the summary of the Bernausin et al. study should be re-written.
D-126
-------
A. H. STERN
Since the only studies discussed in the IRIS document which clearly demonstrate
induction of autoimmune glomerulonephritis are those using Brown Norway rats, it is not possible
to evaluate whether autoimmune glomerulonephritis occurs with inorganic Hg exposure (albeit
at a higher dose level of exposure) in species/strains which are not particularly sensitive to this
effect.
The use of Brown Norway rats (apparently a particularly sensitive strain for autoimmune
glomerulonephritis) needs discussion and justification.
The use of an overall UF of 1000 is not discussed. It seems likely, however, that this
represents three consecutive UFs of 10 for LOAEL to NOAEL conversion; animal to human
conversion; and sensitive populations. Given that the Brown Norway rat is already a sensitive
species, it is not clear that the UF of 10 to account for sensitive individuals is appropriate. This
requires discussion.
It may well be that, given the extensive discussion underlying this derivation, a "high" level
of confidence is appropriate for the database supporting this RfD. However, given questions
regarding the qualitative appropriateness of this endpoint in Brown Norway rats for less sensitive
species, I cannot evaluate the appropriateness of this confidence assignment. Likewise, given
questions regarding the need for a UF of 10 to account for sensitive individuals with a LOAEL
derived form a sensitive strain, I cannot evaluate the appropriateness of the "high" confidence
level assigned to the overall RfD.
Inorganic Mercury (Mercuric Chloride) - Carcinogenicity
Overall, I agree with the weight of evidence classification of C - possible human
carcinogen. The carcinogenicity database is limited by the occurrence of nephrotoxicity which
obscures the determination of carcinogenicity independent of necrosis. The observation of
thyroid tumors, however, deserves more discussion. This database is somewhat unusual in my
experience in presenting germ cell and somatic cell mutagenicity data of more than passing
D-127
-------
A. H. STERN
interest. Particularly interesting is the observation of dominant lethality in rats resulting from male
exposure (Zasukhina et al.). I believe that the mutagenicity data argue for more mechanistic
research.
I agree that the NTP (1993) study provides only equivocal evidence of carcinogenicity.
There is clear evidence of extensive non-carcinogenic toxicity which makes it impossible to
identify carcinogenicity which is independent of chronic high dose tissue damage. However, I
am confused by the rationale cited from NTP which is used to dismiss the significance of the
thyroid carcinomas in the rats. My understanding is that the occurrence of hyperplasia may point
to chronic high dose tissue damage (necrosis) and would, therefore, seem to reduce the possible
significance of thyroid tumors observed in conjunction with the hyperplasia. On the other hand,
thyroid tumors without hyperplasia would seem to point toward a mechanism for carcinogenicity
which operates independently of necrosis and one which is more likely to have significance at
lower doses. The explanation of hypersecretion of thyroid stimulating hormone by the pituitary
is intriguing, but it is not clear how this explanation eliminates the significance of the observation.
Is the implication that pituitary stimulation results from some generalized high dose response?
This is not to argue for the interpretation of this study as adequate for the determination of
carcinogenicity. However, I think that a more thorough explanation for the dismissal of the
thyroid tumor observations is warranted.
Methvlmercury - RfD
NOTE: My comments on this RfD document, unless otherwise stated, refer to the IRIS document
in Appendix B rather than to the report text.
Overall, I believe that this RfD represents a reasonable synthesis of the available data.
The database has some clear limitations which, in general, have been adequately discussed in
the text. Additionally, I think that it should be mentioned that the derivation of the benchmark
dose form the Marsh et al. 1987 study is somewhat sensitive to the grouping of the exposure
groups in the statistical analysis. Based on my limited experience, I believe that, despite the
D-128
-------
A. H. STERN
limitations inherent in this RfD, if the key study and the overall database supporting this RfD are
compared to those of other RfDs which have undergone far less intensive reviews, this RfD will
be found to have a considerably stronger basis than most. A useful perspective is that if there
were no appropriate human study,,the use of one or more of the animal studies for the derivation
of an RfD would have been completely consistent with RfD practice. It is likely that such an
approach would lead to an RfD value quite similar to that derived from the human data.
Nonetheless, in light of continuing research, I believe that this RfD should be specifically
designated as interim. My primary criticism of this RfD as it currently stands, is that the values
selected for calculation of the pharmacokinetic model do not seem to reflect a coherent policy
regarding the choice of input values when interindividual variability results in a wide range of
possible values. As detailed below, my reading of the pharmacokinetic literature, including at
least one study apparently not available to the authors, suggests input values which yield a
slightly smaller RfD (0.07 /^/kg/day) based on the same benchmark maternal hair concentration.
The value of C, the concentration of methylmercury in blood corresponding to the
benchmark dose, is given in the IRIS document as 47 /vg/l. I believe that the correct value is 44
/jQ/\ as given in the text of the report.
The IRIS document presents two RfDs, one for developmental effects (specific to women
of childbearing age) and one for the general population based on paraesthesia. The text of the
report presents only the RfD for developmental effects. I am given to understand that the correct
version is that of the report. While I understand the desire to keep the IRIS database as general
as possible, I disagree with the decision as reflected in the report to have only the single RfD
based on developmental effects. There is clearly a population for whom the developmental
endpoint is irrelevant. For those environmental exposures over which the public can exercise
little or no control, it makes sense to issue one RfD which is protective of the most sensitive
group even if it is overprotective of the remainder of the population. In the case of
methylmercury, however, exposure is essentially only from the consumption offish and seafood.
This is a highly voluntary exposure which is controllable on an individual level. This RfD will serve
as the basis for consumption advisories. Many of these advisories (such as that in New Jersey)
are based on consumption frequency guidance, such that fish with lower mercury concentration
D-129
-------
A. H. STERN
can be consumed more frequently than those with higher concentration. These advisories call
for conscious decisions on the part of consumers. Further discriminating on the basis of risk
category (i.e. women of childbearing age and all others) will not further confuse or alienate the
public. Failure to discriminate on the basis of risk categories will lead to consumption guidance
based only on the high risk group, which will unnecessarily discourage a significant portion of
the population from taking advantage of a desirable and affordable source of quality nutrition.
This is particularly the case for tuna, which for the non-high risk population will have few if any
consumption restrictions under most advisory schemes if based on the non-developmental
endpoint, but which will not be able to be used to provide a major source of regular nutrition if
advisories are based on the developmental endpoint (most advisory schemes would produce a
recommendation of about one can of tuna a week based on the developmental RfD).
Some mention should be made regarding the possible implications of the use of the
pharmacokinetic model under non-steady state conditions. The WHO 1990 document contains
data from less-than-chronic human exposure studies which suggest that even when steady-state
conditions are not attained, this model still gives a reasonable estimate of exposure.
The observation in the RfD document that the occurrence of paraesthesia and other
sensory disturbances in the mothers in the Iraqi cohort was not necessarily correlated with
developmental effects in their children is important in demonstrating the weak degree of linkage
between maternal toxicity and developmental toxicity. However, I do not see any mention of this
in the report text.
I believe that my comment on pg. 5-18 ^ 1 provides a stronger rationale for the value
of 250 for the hair/blood ratio. In addition, see that comment regarding sources of variability in
this ratio.
See my comment on pg. 5-19, 5 2 regarding additional confirmation of 95% absorption
of methylmercury.
I believe that the recent study of Smith et al. (Pharm. 128:251 -256 (1994)) provides a more
D-130
-------
A. H. STERN
representative value for the fraction of the absorbed dose found in blood (see my comment on
pg. 5-19, 5 2). Based on this study, I suggest a value of 7.7% rather than 5%.
The discussion of the appropriate value for the elimination constant is incomplete and
requires consideration of the appropriate basis for choosing a value when interindividual
variability results in a wide range of possible values (see my comment on pg. 5-19, 1 3). I
suggest a value of 0.013 day1 (53 days) as being more representative of the upper range of
possible values.
The IRIS document cites a body weight of 58 kg while the report text cites a value of 60
kg. I believe that the 60 kg value is the more appropriate (see my comment on pg. 5-19,1 5).
As per my preceding comments on input values to the pharmacokinetic model for the
calculation of d, if b is taken to be 0.013 day"1 (52 days) and f is taken to be 7.7% (0.077), with
all other inputs as per the values in the report (i.e. C = 44 jjgf\ as per the report, rather than 47
jjgll as per the IRIS appendix), then d'\s calculated to be 0.65 ^g/kg/day (RfD = 0.07 jug/kg/day).
If b is kept at the value in the report of 0.014 day'1, then d calculated to be 0.70 //g/kg/day. This
is a small, but significant difference relative to the value of 1.0 pg/kg/day derived in the report.
As discussed previously, I believe that this change is justified on the basis of the new and more
specific data on the fraction of the methylmercury dose in the blood (/) reported in Smith et al.
The IRIS document gives a value of 35 ppm for the 10% risk level. This risk level is given
(correctly, I believe) in the report text as 11 ppm.
The uncertainty factor (UF) given in the IRIS document (total of 30) is, I believe, incorrect
and should be 10 as given in the report text. While I agree with the UF value of 10,1 have some
problems with the rationale used to generate that value (see my comment on pg. 5-21, f 2).
See my comments on pg. 4-59, \ 5 ff. regarding the summary of the McKeown-Eyssen
paper.
D-131
-------
A. H. STERN
See my comments on pg. 4-60, 1 1 and 4-61, 1 1 regarding the interpretation of the
Kjellstrom et al. studies. In particular, I do not agree with statements summarizing the limitations
of these studies.
Because of its consistency with the other supporting studies the recent paper of Delgard
et al. (Env. Health Perspec. 102:548-550 (1994)) should be included as a supporting study
despite its limitations (see my comment on pg. 4-61 *? 2).
Two additional monkey studies by Burbacher et al. should be included among the animal
studies (see my comment on pg. 4-63, 1 2).
The rat study of Musch et al. (Arch. Toxicol. 40:103-108 (1978)) which is closely related
to the Bornhausen et al. study, should be included.
The too brief discussion of the half-time (elimination rate constant) in blood in the
"kinetics" section is redundant of the discussion in the section on "calculation of dietary intake,"
which itself is incomplete.
See my comment on the influence of selenium on pg. 7-1, ^ 5.
I agree with the assignment of medium confidence to all the categories. However, in light
of ongoing research, I believe that this RfD should specifically be given interim status (see my
comment on pg. 5-22, ^ 2).
Methvlmercurv - Carcinogenicitv
I agree with the weight of evidence classification of C - possible human carcinogen.
However, I think that even for this classification, the data in support of a possible carcinogenic
activity are weak.
I agree that there is no significant evidence for the carcinogenicity of methylmercury from
D-132
-------
A. H. STERN
the cited human studies.
I believe that the evidence for the carcinogenicity of methylmercury from the animal
studies is equivocal at best. Where an increased incidence of tumors were observed, it appears
to have been associated with significant non-cancer toxicity particularly renal necrosis. I find the
mutagenicity data intriguing, and there appears to be a general tendency for mutagenicity tests
to indicate a range of chromosomal and DNA aberrations. Nonetheless, the chronic
carcinogenicity studies do not clearly reflect this mutagenic activity in terms of tumor production.
Comments on Appendix C
I find the uncertainty modeling presented in appendix C to be an interesting exercise in
attempting to understand the nature of the risk presented by the Marsh et al. data. However,
because it is a highly subjective treatment, I see it as no more than an exercise, merely one of
many such possible treatments and I therefore do not believe that it should be applied directly
to the formal characterization of the RfD. The subjectivity in this exercise is evident from the
choice of the form of the input distributions (log-normal, beta, correlations), to the weighting of
these distributions (the means, standard deviations and limits of the input distributions). While
I agree that some of these choices are plausible, I believe that others are not. Even for those
which are plausible though, other choices would be equally plausible. The modeling of true
uncertainly (as opposed to inter-individual variability) is, of necessity a subjective process
because, to put it simply, if we know enough about something to model it objectively, than it is
not uncertain. To some extent, this subjectivity can be addressed by iterative processes which
attempt to weight possible choices by some quasi-objective procedure (e.g. expert panels) and
then generates a weighted distribution reflecting the relative likelihood of each possible choice.
This is sometimes referred to as a two dimensional Monte Carlo analysis. It is a huge
undertaking and, I believe, of limited value and utility. The analysis in appendix C reflects only
one possible facet of such an analysis. While Monte Carlo analysis is a very powerful tool, it is
limited by the availability of data. There are times, such as this, where, I believe, we have to
admit that we do not have enough information and leave it at that.
D-133
-------
-------
Anthony Verity
D-135
-------
M.A. Verity, M.D.
Review: ''Mercury Study Report to Congress: Volume IV Health Effects of Mercury and
Mercury Compounds"
Introduction:
Volume IV on the Mercury Study Report to Congress entitled Health Effects of Mercury
and Mercury Compounds is an extensively well documented, current statement of the
present status of clinical, bioepidemiological and risk assessment data of elemental
mercury, mercuric mercury and methyl mercury toxicity in animals and humans. Volume
IV is subdivided by sections which individually describe pharmacogenetics, specific forms
of organ system toxicity as apply to the individual mercury species, and attempt at hazard
identification with dose-response assessment including an evaluation of human
carcinogenicity and a summary of ongoing research and proposed future research needs
pertaining to the health effects of organic and inorganic mercury compounds.
The presentation is detailed, coherent, quite comprehensive and provides a broad view of
the differences in organ toxicity induced by the mercury compounds with special reference
to toxicity in the kidney, immune, reproductive and nervous systems.
D-136
-------
M.A. Verity, M.D.
Specific Critique:
3-2. Further details are now available on the hepato-enteric mechanisms involved in
handling of mercury. In particular, the role of mercurial binding to GSH allowing for
transport of the mercurial from liver to bile is now well recognized.
3-5. A description of our present knowledge concerning the mechanism for renal handling
of mercury, especially methyl mercury may be relevant The importance of GSH, S-
conjugates, gamma-GTP and peptidases has been well documented in the cyclic renal
handling of methyl mercury (Hirayama and colleagues, Yasutake and colleagues).
»
4-68. In this review of the carcinogenicity of organic mercury in animals, evidence for
oncogenesis is provided. This is in marked contrast to the virtual absence of any such
correlation in human organic mercurial neurotoxicity or human inorganic mercury
exposure. Careful analysis of these individual studies reveal species variation and even
some dose exposure variation. Even so, the relative ease of oncogenesis in mice
compared to rats is identified. Certainly, this data should provide the stimulus for a future
research program.
4-76. "Proposed mechanism of action for neurotoxicity." I have taken time to rewrite
this section in the hope of providing more balance and emphasis on the likely mechanistic
pathways leading to neural injury, (see below)
D-137
-------
M.A. Verity, M.D.
The molecular basis for methyl mercury neurotoxicity is likely complex and multifactorial.
The broad affinity of mercury for -SH groups leads to membrane, enzyme and cytoplasmic
organelle interaction. Major mechanistic pathways have been proposed to include: 1.
inhibition of macromolecular metabolism, especially that of protein translation and nucleic
acid biogenesis. 2. oxidative injury. 3. disturbance in Ca2+ hemostasis. 4. aberrant
protein phosphorylation. Even so, the mechanisms underlying inhibition of protein and
RNA synthesis are multiple. Depending upon the systems used with in vitro, in vivo or
neuronal cell suspensions, evide'nce for inhibition of translation associated with a change in
ATP/ADP concentration has been found. On the other hand, direct inhibition of
elongation was documented secondary to the selective inhibition of certain aminoacyl-
tRNA synthetases (Cheung and Verity, 1985) Syversen, (1977) investigated the effects of
methyl mercury on protein synthesis in rats using techniques which allow analysis of
different cell populations from the central nervous system. Results of this study indicated
selective irreversible damage to granule cells of the cerebellum, whereas damage to the
other neurons, such as Purkinje cells was reversible. Such selectivity of toxicity is a
feature of the neuronal loss seen in human and experimental disease. Methyl mercury has
also been suggested to cause neuronal degeneration by promoting the formation of
reactive oxygen species (Ali et al, 1992; Le Bel et al, .1990, 1992; Verity and Sarafian,
1991). While contributory, such oxidative injury does not appear primary to the site of
toxicity as appropriate protective measures blocking oxidative stress and lipoperoxide
formation are only minimally cytoprotective.
D-138
-------
M.A. Verity, M.D.
A recent review by Atchison and Hard (1994) discusses several proposed mechanisms of
action of methyl mercury on Ca2+ hemostasis and ion channel function. Individual studies
have demonstrated that the neuromuscular actions of methyl mercury occur predominantly
at the presynaptic site (Atchison et al, 1984). Methyl mercury may interfere with
acetylcholine neurotransmitter release and subsequently synaptic transmission (Atchison et
al, 1986; Barrett et al, 1974; Schafer et al, 1990; Schafer and Atchison, 1989, 1991).
Finally, Sarafian (1993) demonstrated that the methyl mercury induced stimulation of
protein phosphorylation in cerebellar granule cell culture is coupled to Ca2"1" uptake,
changed intracellular Ca2"1" hemostasis and inositol phosphate metabolism. These latter
observations invoke the activation of the PKC pathway.
-Lt
5-14. "Dose-response assessment for mercury" and RfD for inorganic mercury of 3 x10 -
-4 mg/gl/day proposed. Response measure is kidney toxicity due to auto-immune disease
(Hapten-mercury complex in the glomerulus). Selection of this target system appears well
documented and use of the Brown-Norway rat is an appropriate model. The Loael values
were obtained from three individual studies. No statement as to the magnitude of the
--**-
uncertainty factor (UF). Organic mercury, RfD proposed as 3 x 10 -4 mg/kg/day
calculated using an uncertainty factor of 10. The target system was to include ataxia
paresthesias in populations of humans. Also obtained from single study.
D-139
-------
M.A. Verity, M.D.
The need for an RfD based upon developmental toxicity has been suggested. If the
proposed model were embryotoxic or fetotoxic than what animal species would be
appropriate? I see no evidence at this time to mandate the need for such further study.
The similarity in RfD for both inorganic mercury using the autoimmune kidney model and
methyl mercury, using human derived data are remarkably similar. However, the
magnitude of uncertainty factor is quite variable and perhaps difficult to defend at a level
of 10 in the human study where general variables and confounders are expected to be
much greater than in the more tightly controlled animal model.
8-5. Research needs. We fully agree with the existence of significant data gaps in the
carcinogenicity assessments for each of the mercury compounds. Mercury has not been
considered carcinogenic, but the documented experiments highlight a particular form of
renal neoplasia associated with the organomercurial. In each of the positive studies,
significant nephropathy occurred simultaneous with the onset of neoplasia and appropriate
future studies must of necessity utilize longer term, lower dose exposure paradigms.
Moreover, the known sex and species difference in renal handling of methyl mercury
among rodents must be assessed in any such long term study.
Specific areas of concern:
Is the information provided on pharmacogenetics sufficient?
D-140
-------
M.A. Verity, M.D.
Such information is variable between the three considered species of mercury. Certainly,
elemental and inorganic mercury may be considered to be handled in a similar manner once
incorporated within the biomass and target system. However, methyl mercury will
undergo variable transformations and demethylation providing change in ratio of
organic/inorganic mercury concentrations confounding in the analysis of
compartmentation. However, the data as presented is sufficient to allow for first principle
derivations from which estimates of disposition may be made allowing for evaluation of
human health effects.
"Categorization for carcinogenicity, developmental toxicity and germ cell mutagenicity."
A priori, "weight of evidence" data used in defining a toxic effect must be considered
unsatisfactory. If the data is derived from specific epidemiological studies with
appropriate statistical analysis, then the term "weight of evidence" would be inappropriate.
In the case detailed here for carcinogenicity, then group C (possible human carcinogen) is
appropriate given the definition of limited evidence of carcinogenicity in animals in the
absence of human data. This classification appears stronger for the methyl mercury
derivative than the inorganic mercury based upon available animal evidence. However,
insufficient evidence is really applicable to both these mercurial species suggesting a
mandatory need for reevaluation of animal data in the future. In contrast, developmental
toxicity is well documented for the methyl mercury derivative with suggestive but
insufficient data pertaining to the inorganic compound. This is similar to the germ cell
D-141
-------
M.A. Verity, M.D.
mutagenicity where strong evidence is derived from both human and animal data
pertaining to the organic mercurial with more equivocal evidence associated with
inorganic mercury.
Dose-response assessment for carcinogenicity. No scientific data is presented to
document human carcinogenesis associated with mercurial exposure. Although animal
models do identify some oncogenic potential (see above), this reviewer would agree that
any meaningful quantitative dose-response assessment extrapolated to human exposure
would therefore be meaningless. The arguments appear cogent and are adequately
supported by information.
Calculation of reference dose (RfD) and reference concentration (RfC). This field is
beyond my area of competence. Examination of formulation appears appropriate, but the
somewhat variable and capricious inclusion of the uncertainty factor (UF) may need
further clarification. Moreover, the inclusion of the term "reference" appears excessive
when defining a minimum dose providing measurable effect at a certain probability level.
Reference to what?
D-142
-------
Volume V
An Ecological Assessment for Anthropogenic Mercury Emissions in the United States
D-143
-------
-------
Thomas Atkeson
D-145
-------
Atkeson and Parks
Volume V
MERCURY STUDY REPORT TO CONGRESS VOLUME V:
AN ECOLOGICAL ASSESSMENT FOR ANTHROPOGENIC MERCURY EMISSIONS
IN THE UNITED STATES
Tom Atkeson, Paul Parks
Division of Technical Services
Florida Department of Environmental Protection
COMMENTS ON THE DRAFT OF December 13,1994.
Introduction
This volume presents three results that might be used to make decisions about the
control of anthropogenic mercury:
1. Maps indicating regions of concern where high rates of mercury deposition
occur along with wildlife species of concern;
2. A numerical wildlife criterion, which is a concentration of mercury in water
that, if not exceeded, protects avian and mammalian wildlife populations
from adverse effects resulting from ingestion of surface waters and from
ingestion of aquatic life taken from these surface waters;
3. An outline of research needs for an assessment of the ecological impacts of
anthropogenic mercury emissions.
The following comments consider the utility of these three results for decision making.
Regions of Concern
The reasons for mapping regions of concern are, first, to present a general picture of
the nationwide extent of the co-occurance of high mercury deposition and wildlife
species of concern and, secondly, to help locate the areas of greatest concern where
efforts should be focused.
D-146
-------
Atkeson and Parks
Volume V
The general picture is presented in the Base Case Map (Figure 3-1), which is a map of
the United States showing three ranges of modeled deposition rates. This map would
be more informative if the areas with the highest rates were darkest, the areas with
intermediate rates were intermediate in darkness, and the areas with the lowest rates
were white (as is done in Figure 3-4). The time scale of the deposition rate has been
omitted. The map itself is useful and informative. In Section 3.2.1, which refers to the
Base Case Map, some comment or forward reference to the likely significance to
wildlife of the chosen ranges of deposition rates might be helpful.
Figures 3-5 through 3-11 show the distribution of species of concern in relation to
deposition rates. These figures eliminate from the "regions of concern" those areas
where the pH of the water is greater than 5.5, "based on the generally accepted
observation that methylmercury concentrations in fish flesh have beeri positively
correlated with low pH" (Section 3.2.4). While methylmercury concentrations in water
and fish are related to pH, this does not mean that there are no problems at higher pH.
For example, this cut-off point eliminates the Florida Everglades, where surface water
pH is circumneutral, from consideration even though it has some of the highest
recorded mercury values for fish over an area exceeding 1 million acres in extent.
Acidity is but one of many factors that affect the rate of uptake of mercury by fish. The
evidence is not sufficient to support the exclusion of large areas with high mercury
deposition rates because their waters have pH values greater than 5.5.
The legends for Figures 3-5 through 3-11 are unclear, as are the maps. What do the
percentages represent? These maps might be made clearer by leaving the range
background white and the non-range gray. Deposition rates could be eliminated from
the non-range areas indicated within the range using patterns that are different from
the non-range gray and leave the white background essentially clear.
D-147
-------
Atkeson and Parks
Volume V
The Wildlife Criterion
The wildlife criterion is of fundamental importance because it is the basis for any
mercury management action that may be taken.
The No Observed Adverse Effect Level Basis for the Wildlife Criterion
Feeding study toxicity data for ring-necked pheasant, mallard duck, river otter, mink,
and house cat were available. The wildlife criterion derived in this work is based on the
NOAEL, the no observed adverse effect level, obtained by feeding mallard ducks
methylmercury dicyandiamide. The NOAEL value used here applied only to the first
generation fed. Second and third generations did show effects at the NOAEL value for
the first generation. In addition, there was no NOAEL for behavioral effects in the
ducklings. Thus, the NOAEL used is limited to this species and represents a no
observed adverse effect level in a very limited sense. A species sensitivity factor of 0.5
(an arbitrary value) multiplies the mallard NOAEL in the wildlife criterion calculation.
In Section 2.3.2.3., the sublethal effects of mercury on birds are discussed and the
statement is made: "Reproductive effects, however, are the primary concern for avian
mercury poisoning and can occur at dietary concentrations well below those which
cause overt toxicity." In Section 2.3.3.1, there is a further discussion of mercury effects
on reproduction. Also, elevated levels of mercury were associated with abnormally high
infestations of parasites in a die-off of 2,500 loons in the Gulf of Mexico.
While it is probably true that fisheating birds and mammals are most at risk from
atmospheric mercury, some further consideration should be given to the effects of
mercury on fish and reptiles. These components of the ecosystem are important perse
and are part of the food supply for fisheating birds and mammals. Also, alligators and
snakes are long-lived and would be expected to accumulate mercury; alligators from
the Florida Everglades have been found to contain up to 7 - 8 mg/kg mercury in
muscle.
D-148
-------
Atkeson and Parks
Volume V
The wildlife toxicology studies upon which the NOAEL is based are regrettably limited;
a shortcoming which can only be rectified by further work in this area. Thus, while they
are the best available, the mallard feeding data provide a very tenuous basis for
computing a wildlife criterion.
The Bioaccumulation Factor Basis for the Wildlife Criterion
Calculation of the bioaccumulation factor - The wildlife criterion value is inversely
proportional to a weighted sum of the bioaccumulation factors (BAF) for trophic levels 3
and 4. The BAF is the ratio of methvlmercurv concentration in fish to total mercury
concentration in filtered water. The reason for choosing to define the BAF in this way is
not addressed directly, but is treated in Sections 4.2.1, 4.3.9, 5.1 and 5.3. This
decision appears to have been a combination of precedent (considerable agency effort
having gone into development of various EPA guidance documents and the Great
Lakes Water Quality Initiative), the rapid changes in analytical methods for total and
methylmercury that have taken place in recent years, and the paucity of methylmercury
data from natural water bodies. Some of the arguments justifying this choice, however,
are specious. In Section 4.2.1, Para. 4, the very real concern about "...lack of scientific
consensus about which forms of methylmercury are available for uptake by aquatic
organisms..." applies equally to total mercury.
While the authors of the report are certainly cognizant of the dependence of this
approach upon a large number of assumptions, it nonetheless represents the greatest
conceptual weakness in the report. To base the BAF on total mercury is to interpose a
clearly nonlinear, multivariate 'process box' between the water and the biota,
multiplying the assumptions greatly and increasing uncertainty in proportion. Logically
the best method for determining the BAF is to base it on methylmercury in fish :
methylmercury in water, and with present measurement techniques this has become
feasible. Although data are yet sparse, it is clear that the state of the art has
progressed to the point that the BAF should be based on methylmercury and the
biogeochemical processes that relate total mercury to methylmercury in water should
be treated separately.
D-149
-------
Atkeson and Parks
Volume V
New values for field-derived fish methylmercury : water methylmercury BAFs are being
reported from a number of areas in North America, a number of which estimate BAF4 to
be on the order of 1 x 107. For example, largemouth bass from the Florida Everglades
average - 2 mg/kg mercury, whereas determinations of methylmercury in surface
waters of the region by a number of investigators are typically < 200 pg/L; a BAF
approximating 1 x 107. As such values become available from a representative
sampling of water bodies it will be desirable to revisit the estimation of the BAF used in
this risk assessment.
The relation between water total mercury concentration and fish mercury concentration
is so complex and poorly understood that, if in this report the authors must persist with
the present method of estimation of the BAF, the presentation of any value for a wildlife
criterion should be accompanied by a more explicit discussion of all known limitations of
the value. It would be helpful if the limitations of the wildlife criterion were summarized
in a table. Such a table would also be helpful for defining research needs (see below).
Bioaccumulation factor variability - The BAF is assumed to be a constant at a given
trophic level. In fact, there is not much reason for believing that the concentration ratio
defined above is a constant and there is evidence to the contrary. The concentration of
methylmercury in fish does not appear to be a simple function of the total mercury
concentration in water. In samples taken along Everglades canals Stober, Jones and
Scheldt (1994, In Press) found that the gradient in Gambusia total mercury
concentration was opposite to the gradient in water total mercury concentration, but
was parallel to the gradient of methylmercury in canal bottom sediments. The
Gambusia gradient was the same as the gradient found in largemouth bass in another
study. In this case, the bioaccumulation factor is not a constant defined by the ratio of
fish methylmercury to water total mercury concentrations.
When the effects of pH, hardness, and degree of oligotrophy on the concentration of
methylmercury in fish are considered, it likewise appears that the use of total mercury
D-150
-------
Atkeson and Parks
Volume V
concentration in water as the denominator for the bioaccumulation factor is an
oversimplification.
The Wildlife Criterion: Conclusion
For a report of broad, national applicability the effects of uncertainty in the wildlife
criterion depend on what use is made of this value. The conceptual and quantitative
uncertainties are large, and it will be these uncertainties surrounding the estimation of
this criterion that will limit its utility in the formulation of policy for the control of the
widespread problem of excessive mercury in fish. Problems stem principally from
selection of appropriate NOAEL or LOAEL and from the uncertainty introduced by the
selection of total mercury as the basis for calculation of the BAF. These uncertainties
are sufficiently large that it really is not known if the value selected here is too high or
too low for "the protection of avian and mammalian wildlife populations from adverse
effects resulting from ingestion of surface waters and from ingestion of aquatic life
taken from these surface waters."
Research Needs
The research topics proposed in Section 5 adequately cover the questions that are the
major sources of uncertainty or variability pertinent to the present effort to develop a
wildlife criterion for mercury.
D-151
-------
-------
Steven Bartell
D-153
-------
S.M. Bartell
Volume V. An Ecological Assessment for Anthropogenic Mercury Emissions in the
United States
The following comments are offered in review of the Mercury Study Report to Congress.
The comments have been organized according to the specific topic areas requested.
Additionally, general and specific comments that might facilitate the review have been
included.
Methods for Generating BAF3 and BAF4 Values
The BAFs were calculated as the ratio of methylmercury in whole fish to total mercury in
filtered lake water. This relies on the assumption that the ratio of methylmercury to total
mercury in filtered lake water remains essentially constant (e.g., p. A-5). This assumption
remains difficult to justify. Several environmental factors influence the chemical
speciation of mercury in natural waters. Water characterized by high concentrations of
particulate or dissolved organic matter might result in markedly different bioavailability of
mercury. Also, on p. A-22, a value of 5% was used as the percent of total mercury that is
methylated, yet in the other calculations, a value of 17% was used (p. A-5).
Instead of accepting the large variability with respect to mercury concentrations in fish (p.
A-3), efforts should have been made to at least explore regional variability in water
chemistry, food chain (better, food web) structures, spatial-temporal scales of deposition,
and spatial-temporal scales of the receiving organisms. The regional bias (recognized on
p. A-24) towards lakes of the northern Midwest for sources of data questions the utility of
the overall assessment.
It was disconcerting that the PPF2 and PPF3 values were based on a single study (i.e.,
Watras and Bloom 1992). Also, the food chain multipliers do not appear to consistently
increase (as assumed) and it was recognized that this limitation exists (p. A-17). It was
further recognized that this assumption could have a "large impact on subsequent
D-154
-------
S.M. Bartell
calculation of the BAFs..." (p. A-17).
Provide a reference that supports the statement that the beta distribution is commonly
used to represent ratios (p. A-12).
One of the more critical assumptions (e.g., p. 3-19, 4-16) concerns the fish trophic levels
fed upon by the various species of interest. Particularly, it remains difficult to believe that
eagles would only feed on trophic level 4 fish, while ospreys and kingfishers consumed
only level 3 fish! Similar concerns are expressed for the fish dietary assumptions for
otters and mink. This assumption should have been tested as part of the overall
sensitivity analysis in estimating the dose and resulting WC values for the wildlife
populations of concern.
Methods for Generating an Uncertainty Analysis
In describing the Monte Carlo analyses (p. A-2), it was mentioned that the parameters are
assumed to vary independently. This is not the case and the Crystal Ball software permits
the specification of correlations among parameters. In fact, the influence of such
correlations were examined in the reported sensitivity analysis. Also, the number of
simulations (i.e., p. A-2) depends on the nature of the model and the particular output of
interest. Ideally, the number of simulations should be determined by defining the
relationship between the number of simulations and the variability about the model
output.
The sensitivity analysis described in section A.3.6 focused mainly on three model
assumptions (p. A-25). However, it was not a sensitivity analysis that addressed the
implications of uncertainty associated with the input model parameters. This was
puzzling, given the effort put forth in documenting and developing distributions for the
model parameters. The results of such traditional analysis of parameter sensitivity would
have produced a rank ordering of parameters to be examined in future research to reduce
D-155
-------
S.M. Bartell
overall uncertainties ( as performed for the methylmercury RfD analysis reported in
Volume IV-Appendix C).
The investigators admit that the potentially greatest source of uncertainty is model
uncertainty (e.g., failure of the simple linear BAF model to account for more complex real
world processes, p. A-28). However, this line of discussion was not followed up and the
reviewer is left alone to understand the meaning and utility of the overall assessment in
light of this potential source of uncertainty.
Endpoints and Studies Selected for Wildlife RfDs
The toxicity of mercury, particularly methylmercury, seems to have been fairly well
characterized in the Report.
The studies used to develop the different components of the wildlife RfDs seem to
represent a fairly comprehensive summary.
Assumptions Used in Developing Wildlife Water Criteria
The framework for ecological risk assessment is inherently probabilistic. Therefore, it is
questionable that in developing the water criteria (WC) that only the mean value of the
corresponding BAF was used to provide a single estimate of the WC for the different
endpoints of interest (p. 4-6). Distributions of WC values should have been developed for
comparisons with expected distributions (or at least ranges) of reported or modeled
mercury concentrations. The uncertainties associated with the estimates of the BAFs
should have been propagated through the estimates of the WC - it would have been
possible to set this up in the Crystal Ball - Excel environment.
The precision of the WC values described on p. 4-17 may be somewhat an artifact of
D-156
-------
S.M. Bartell
using the same NOAEL values for each of the avian or wildlife species in combination
with allometrically derived parameters for mercury intake. These assumptions limit the
degree of dispersion possible for the WC estimates. Thus, it's difficult to assess the
meaning and value of this precision.
The NOAELs were based on relatively short term studies and it remains difficult to
determine the relevance of these toxicity data to longer term assessments for a nation
wide assessment.
The species sensitivity factors can range from 0.01 to 1 for this study. The use of these
factors was recognized as a limitation in the assessment, however it was not apparent
that uncertainties associated with values assigned to these factors were included
explicitly in the calculations of WC values.
Migratory behavior and life history characteristics of the avian and wildlife species were
not included in assessing spatial-temporal patterns of exposure to mercury (p. 4-16).
Possible incompatibilities in scales between exposure and effects assessments should
be examined to evaluate the meaning of the overlays of modeled mercury deposition and
species distributions for the national assessment.
Other Species of Concern That Should be Included
Piscivorous wading birds, as well as birds and waterfowl that consume large amounts of
potentially contaminated plankton and benthic invertebrates should have been addressed
in the study. Indeed, other species of potential interest were recognized (p. 4-9). The
herring gull, raccoon, and snapping turtle probably have more relevance to a national
assessment than does the Florida panther.
The assessment addressed direct impacts on higher trophic level piscivores. There
should have been additional focus on the risks posed by mercury to primary and
secondary production at the lower trophic levels (e.g., phytoplankton, zooplankton,
benthic invertebrates). Indirect impacts on higher trophic levels through reduced energy
D-157
-------
S.M. Bartell
flow throughout the food web should have been explored. Effects on other individuals,
populations, communities, and ecosystems were discussed (pp. 2-13 to 2-20), but
endpoints were not developed for these levels of organization.
Given that the results of the RELMAP predictions of high mercury deposition coincided
with the location of endangered plant populations in central and southern Florida, the
northeastern coast, and throughout the Midwest, why weren't some of these species
selected as endpoints for the risk assessment?
Other Geographic Areas of Concern
The assessment was heavily influenced by data and analyses for northern oligotrophic
lakes. While this was recognized (p. A-28) by the investigators, it remains difficult to
determine the implications of this bias on the overall nationwide assessment.
Given that the Florida panther was selected as one of the species of concern, there
should have been additional effort placed on more site-specific (at least regional)
assessment of differences in mercury speciation, transport, and bioaccumulation.
Important Additional Data or Analyses
The investigators seem to have performed a comprehensive review of the available
information regarding mercury fate and effects.
Presentation of Arguments and Conclusions
The conceptual models for mercury accumulation and biomagnification presented in
Volume V appear valid on the basis of what is known about the chemistry and biology of
mercury and its compounds. Perhaps the most questionable logic takes the form of the
assumed simple linear food chains leading ultimately to the higher level piscivores,
including humans. This steady-state, linearized view of nature appears overly simplistic:
D-158
-------
S.M. Bartell
the result might be overestimating mercury risks for some regions and species and
underestimating risks for others. Regional differences in food webs, seasonality in
patterns and rates of food consumption, and patterns of mercury deposition should ideally
be factored into an assessment that is truly national in scope.
The BAF, PPF, and FCM values are more likely species-specific than trophic-level
specific (p. 2-13).
The assessment endpoints for communities (Table 2-5) seem unrelated to ecology. Also,
the use of biomarkers for individual measurement endpoints focuses more on exposure
than risk.
There needs to be better justification for using annual average values of modeled
mercury deposition (e.g., RELMAP). In highly seasonal environments like the northern
Midwestern oligotrophic lakes, annual averages might not accurately reflect exposure to
aquatic biota. Also, mercury released in combustion processes will likely exhibit strong
temporal patterns, again questioning the validity of annual average deposition values. It
might have been more useful to estimate the maximum deposition values.
\
The interrelations among the different models (i.e., RELMAP, COMPMERC, and IEM2)
were not sufficiently described. If the COMPMERC addresses deposition from a local
facility, what are the source terms to the RELMAP? Are source terms double-counted in
using the combination of RELMAP and COMPMERC? This was not clear. Also, what data
substantiate the use of a 30-y period for the typical lifespan of these different facilities
(e.g., such as listed in Table 3-2)?
Research Needs to Address Uncertainties
The investigators identified research needs regarding (1) process-based models and
understanding of the environmental chemistry and biological effects of mercury, (2)
additional wildlife toxicity data for mercury, (3) improved analytical methods for mercury
and methylmercury in environmental samples, (4) more realistic representation of food
D-159
-------
S.M. Bartell
webs, (5) research on mercury accumulation in lower trophic levels, (6) additional filed
residue data, and (7) incorporation of life history data into future assessments.
These specified research needs follow logically from the discussion of the assumptions
and limitations stated throughout the assessment. However, the research needs were not
identified through rigorous sensitivity analysis of the equations adopted for estimating the
WC values for the species examined. Had this been the strategic approach, it would be
possible to determine the relative value of new information in reducing uncertainties
associated with the WC values.
General Comments
A major concern lies with the overall assessment being based on estimates of the WC
derived from NOAELs for the selected species. If actual exposures exceed the WC, it is
not straightforward how the ecological impacts of such exposures can be determined in
the context of ecological risk. The assessment might be more powerful if endpoints were
developed in terms of population dynamics: what is the probability of local extinction of
endangered species? what is the probability of unacceptable (perhaps defined regionally)
decreases in the population sizes of raptors exposed to mercury? what is the probability
of regional declines in primary and secondary production as a function of mercury
exposure? These and other similar ecological endpoints should have been the focus of
the nationwide assessment, at in addition to the NOAEL-based endpoints.
Another concern is that most, if not all, of the limitations pointed out in this review were
also mentioned in this volume of the report. However, few if any attempts were made to
quantify the impacts of the recognized assumptions and limitations on the resulting
assessment. Given the scope and importance of the assessment, efforts should have
been directed at quantifying the impacts of the often well described uncertainties on the
results of the assessment.
D-160
-------
Volume VI
Characterization of Human Health and Wildlife Risks From
Anthropogenic Mercury Emissions in the United States
D-161
-------
-------
Steven Bartell
D-163
-------
S.M. Bartell
Volume VI. Characterization of Human Health and Wildlife Risks
The following comments are offered in review of the Mercury Study Report to Congress.
The comments have been organized according to the specific topic areas requested.
Additionally, general comments that might facilitate the review have been included.
Sufficiency of Risk Assessment Summaries for Scientific Critique
Volume VI is long on detailed description and recapitulation of the other volumes in the
Report. The complexities, assumptions, and limitations of the assessment are
comprehensively laid out in detailed discussion. However, Volume VI should focus more
attention on the actual risk estimation, which currently distills to two tables (i.e., Table 4-3
and 4-4). If this volume is to serve as a stand-alone document, the repetition of material
from early volumes can be understood. If the emphasis is on risk characterization,
Volume VI would benefit from serious modification.
A 10,000-fold span between the 5th and 95th percentile for the developmental effects
threshold (p. 2-10) really calls into question the utility of such a distribution for purposes
or risk assessment.
There are two seemingly important deficiencies in the risk characterization. One, risk is
inherently probabilistic in its conceptualization. The mercury concentrations in fish tissues
corresponding to the NOAELs or LOAELs for humans and wildlife (i.e., Tables 4-3 and 4-
4) should be developed as distributions, not single values. The uncertainties in the
assessment should have been carried through the entire process of calculation using
Monte Carlo methods to produce these distributions of tissues concentrations. Two, the
concentrations in fish tissues corresponding to the NOAELs or LOAELs need to be
compared converted to distributions of WC values and then compared with regional (or
site-specific) distributions of water column mercury concentrations. The risk assessment
should estimate the probability that the distribution of mercury in the water column (ideally
D-164
-------
S.M. Bartell
in "methylmercury equivalents") significantly overlaps the distribution of mercury
concentration that translates to fish tissue values corresponding to the NOAEL or LOAEL.
That is, the characterization of mercury impacts should be put into the statistical context
of risk assessment.
Uncertainties, Defaults, and Assumptions Not Discussed
The uncertainties were comprehensively addressed, but primarily in a qualitative manner.
Except for the analysis in Appendix C (Volume IV), little was presented concerning the
impacts of these uncertainties on the overall assessment of risk.
One major potential shortcoming results from using the results of short term toxicity
studies (e.g., for wildlife) and analysis of comparatively short term human exposures
(e.g., the Iraqi women) to assess long term risks on a national scale. This may have been
mentioned, however the issue was not resolved in the context of risk and uncertainty nor
were the impacts of this assumption on the resulting risk characterization addressed in
any quantitative manner.
Critique of the Uncertainty Analysis
Very little of the uncertainty analysis is presented in Volume VI. Rather, the reader is
referred to Appendix C of Volume IV for the formulas and parameter distributions. The
description provided in the Appendix C is fairly comprehensive in the context of the
overall assessment. That is, the distributions for the parameters were well described. The
Monte Carlo analysis was fairly standard (as determined by the capabilities of the
selected software, Crystal Ball and Excel). The sensitivity analysis was also standard in
the sense of Crystal Ball.
Methods and Results of the Comparative Risk Discussion
D-165
-------
S.M. Bartell
It is not clear why the large variances in estimation of the BAFs justify the selection of the
mean values for use in exposure assessments (p. 3-6). A more useful approach would
have been to examine the sources of these large variances through rigorous sensitivity
analysis of the BAF calculations. The results of such analyses may well have identified
the key contributions to the large variances, as well as possible bias in the estimates of
the mean values.
Otherwise, the comments regarding the discussion are similar to those developed for the
previous volumes. For example, the review comments for Volume V apply to the effects
discussion in Volume VI - most of the information is repetitive.
Important Additional Data or Analyses
No comment.
Presentation of Arguments and Conclusions
An important premise in the development of the methylmercury RfD was that the 81
pregnant Iraqi women (i.e., Marsh et al. 1987) were somehow representative of the
general range of susceptibilities of the US fish consuming population (p. 2-8). However,
this premise is not discussed further. Also, if the distribution of the threshold dose was
defined arbitrarily, in a way not supported by statistical analysis, what is the utility of such
a distribution for risk assessment?
The fate and transport discussion (Section 4.1) indicates that the remote northeastern
lake location had higher mercury deposition than the remote midwestern location.
However, much of the data for assessing ecological risks derives from the midwestern
lakes. Is there substantial reason to believe that the toxicity data from the midwestern
location applies to the remote northeastern location?
A serious limitation in the risk characterization lies in forcing the comparison of exposures
D-166
-------
S.M. Bartell
(i.e., mercury concentration in fish tissues) based on NOAELs developed for different
human and wildlife endpoints (Table 4-3). It is difficult to determine the value of making
the comparison. If fish tissue concentrations exceed the values in Tables 4-3 or 4-4, what
are the implications for human health or wildlife in the context of risk? This shortcoming in
terms of interpretation results mainly from the selection of endpoints. The endpoints
seem justifiable in terms of our understanding of the toxicology of mercury, yet the
methods and approach for using this information to develop endpoints results in the
inability to realistically address risk.
Research Needs to Address Uncertainties
No research needs were identified in Volume VI.
D-167
-------
-------
Leonard Levin
D-169
-------
Leonard Levin
External Review Comments Concerning
U.S. Environmental Protection Agency
MERCURY STUDY REPORT TO CONGRESS,
January 12, 1995
LEONARD LEVIN, Ph.D.
Electric Power Research Institute
3412 Hillview Ave.
Palo Alto, California 94303
INTRODUCTORY COMMENT
The limited time allotted for reviewing this extremely important document is an unfortunate
consequence of doing science under a schedule. The EPA report suffers further from lacking
critical synthesizing and summarizing sections, particularly for Volume VI on risks (to which
I was assigned) and the Executive Summary Volume I for the entire report. The latter
Executive Summary can, with enough time and commitment to make needed changes and re-
analyses in the other parts of the report, serve as the true Synthesis Document for EPA's
"screening" national assessment of mercury. The disparate pieces contained here form the
nucleus of an analysis that needs to be done in a preliminary way, to guide research for the
next 5-10 years. Instead, EPA has been compelled to attempt a more detailed set of
calculations for which confirming data are lacking or contradict the findings of the EPA
model runs. These disparities can only be corrected through a deliberate program of research,
conducted under a guiding Integrating Framework, to assure the results will not only withstand
technical examination, but guide both researchers and decisionmakers clearly. This report is
not yet that Framework.
UNDERSTANDING OF CHARGE TO REVIEWERS
Under requirements of the 1990 Amendments to the Clean Air Act ("CAAA"), the EPA is to
carry out a study of mercury in the environment from all sources, and an evaluation of its
risks to human health and to ecosystem viability. Reviewers have been requested to focus on
individual volumes assigned, but are free to comment on other volumes relevant to the
technical topics they are covering. Reviewers are requested to assess the data and analyses
used and provide information on additional input that might contribute to the report; to
assess clarity of arguments and conclusions; and to evaluate the listed Research Needs. The
Jan. ,2, 1995
-------
Leonard Levin
assignment for the Volume VI risk characterization is to assess whether sufficient
information on human and wildlife risk is presented to do a technical critique; to evaluate if
major areas supporting the conclusions are lacking; and to critique both the uncertainty
analyses and the comparative risk discussion.
SUMMARY OF ASSIGNED VOLUME VI ("Characterization of Human Health and
Wildlife Risks from Anthropogenic Mercury Emissions in the United States") AND
OF OTHER RELEVANT REPORT SECTIONS
The EPA report on mercury distributed for external review covers emissions and
"downstream" issues regarding mercury emissions from some (but not all) anthropogenic
sources, and excludes background "natural" sources (biogeochemical-origin mercury, or
"paleoanthropogenic" mercury from earlier human emissions sequestered in biogeochemical
reservoirs). The analysis evaluates mercury emissions by source category aggregate, and also
poses scenarios of particular individual sources from within these categories for evaluation of
local-scale mercury plume dispersion, transformation, deposition, and ecosystem fate using
settings scenarios applied to each of the source scenarios. Additionally, source inventories
are used as input to regional-scale dispersion and deposition modelling.
Deposition to surveyed sensitive ecosystems and to modeled local hydrologic systems is
modeled, as is bioaccumulation through aquatic trophic levels to evaluate concentrations in
fish used as food by pisciverous species and by humans. Human exposure is evaluated based
on consumption rates. Health data are used as inputs to Monte Carlo models to arrive at a
new recommended reference dose (RfD) that is one-third the currently listed federal value.
The risk characterization in Volume VI begins by summarizing human cancer and noncancer
health endpoints for mercury species, and the earlier uncertainty analysis on these endpoints,
to arrive at a recommended RfD for methylmercury of 1 x 10'4 mg/kg/day. The wildlife
health effects assessment summarizes results in Volumes III and V. This is followed by a
summary of the exposure conclusions from Volume III, and then by the risk characterization
for wildlife and for humans. Conclusions concerning wildlife risk are set forth in relatively
quantitative terms. Human health risk is apparently dismissed in one sentence, apparently
intended to warn consumers away from consuming fish with higher levels of methylmercury
("As methylmercury concentrations in fish tissue increase, the risk to consumers of these
fish also increases."), although even this sentence is ambiguous, since it might also refer to
pisciverous animals.
COMMENTS ON REPORT
General
Jan. 12, 1995 D-171
-------
Leonard Levin
The key conclusion of the report, summarized in Volume III, is the purported demonstrated
link between the modeled anthropogenic combustion sources of mercury and "significant
incremental exposures, above background," to humans and wildlife via consumption of
contaminated fish. This conclusion rests on a set of assumptions and approximations, used as
inputs to a set of models that has not been peer-reviewed, that result in a series of
conclusions not supported by nor even compared with observation. The most general set of
approximations that drives the final conclusions, and that is by the report's own admission
unsubstantiated by any observation, is that of the wet deposition rate for total mercury.
Stating that these numbers "seem to agree with actual measurements within a factor of 2 or
3" is damning with faint praise, since this "agreement" appears to be a systematic high bias.
In the regional modelling section, this results in deposition rates for total mercury of 20 to
60 ng/m^/yr over much of the eastern U.S., while the cited observations fail to support
numbers beyond 30 ng/m^/yr. (Although there are isolated instances of measured deposition
above 60 u,g/m2/yr, these were short term measurements in an urban area, not applicable to
yearly-average values over an entire 40 x 40 grid square as in the EPA analysis.) These
"citations" in themselves are not referenced, so that one is unable to judge their applicability
in this case. (Discussion below analyzes possible reasons for these high estimates of
deposition rate.)
The consequences of these high deposition rates (and the presumed systematic bias they
indicate for such rates throughout the modelling) are that all subsequent estimates throughout
the report are skewed high, particularly the local-scale COMPMERC modelling which (based
on the footnote to Table 4-20 of Volume III [or "Table 4-20 (III)"]) results in instances of
modeled local deposition rates of more than 500 (ig/m^/yr. These deposition rates, which
appear to be systematically high by about a factor of 2 or so for long-range transport, and
about a factor of 10 for local deposition, result in the listed peak concentrations in fish flesh
for the case study sources exceeding federal advisory levels. These values contrast with those
cited in the report by Brookhaven National Laboratory (1994), where no observed value
exceeds 35 |ig/m2/yr (and that not in the U.S.). The deposition rates locally and regionally
result in the cited mass deposition of 77% of U.S. anthropogenic-source divalent mercury
depositing within the (lower 48) United States. This total mass deposition would be
substantially reduced with reduction in the wet deposition rate.
This anomaly in wet deposition, combined with anomalously high in-cloud conversion rates
of Hg(0) to Hg(II) (when, in fact, net transformation appears more likely to be Hg(II) to
Hg(0); Constantinou, Wu, & Seigneur, 1995), results in relatively high flux of ionized
Jan. 12, 1995
D-172
-------
Leonard Levin
mercury into aquatic systems where it is subject to methylation and bioaccumulation. The
final results, by receptor location and source scenario, are biased high by these anomalies.
A basic premise of the report — that data are inadequate for assessing "natural"
(biogeochemical + paleoanthropogenic) mercury emissions, but are (barely adequate for
evaluating U.S. anthropogenic emissions — results in an incomplete assessment of the
mercury issue. For example, emissions from one of the anthropogenic source categories
(U.S. utility boilers) is roughly 0.5-1.0% of global anthropogenic sources of mercury. Current
thinking is that natural terrestrial sources of mercury globally are about equal to
anthropogenic sources. Coincidentally, U.S. land area is roughly 1.6% of global terrestrial
surface area. Thus, in a global context, and assuming natural emissions are proportional to
land area, these natural emissions from the land surface of the U.S. might be as great as utility
emissions alone. Yet this major source term is unqualified, leaving EPA to postulate a
"plausible link" between anthropogenic point source emissions alone enough to account for
fish methylmercury levels above advisory levels.
Responses to "Charge to Reviewers"
• Are additional data or analyses available that -would have a major impact on the
conclusions presented in any volume of the report?(all reviewers)
A number of new reports have emerged in the last few months that provide additional
information for consideration in any report on mercury in the environment. These recent
references are cited in the attached bibliography. Key references are: Levin and Torrens
(1994) (provided to EPA OAQPS in November 1994), where data on mercury emissions
from oil-fired utility plants differ substantially from values listed in the EPA report; Clewell
et al. (1994) (provided to EPA in October 1994), which provides analyses of mercury
exposure and health response data that indicate the new RfD cited in Volume IV of the EPA
report is poorly-founded; and Constantinou, Wu, and Seigneur (1994) indicating the key in-
cloud conversion reaction of mercury is Hg(II) to Hg(0), rather than the Hg(0) to Hg(II)
reaction employed by EPA.
• Are arguments and conclusions presented clearly and in a logical manner?(all reviewers)
Although the goal of the report is laudatory — a comprehensive review of mercury sources,
transport & deposition, effects, and control potential — execution of the report is
fragmented and of highly varying quality. The major failing is either a lack of any
comparison with observation (admittedly these observations are still fragmentary,
conflicting, and preliminary) and the calculated results. Since the entire structure of the
report rests on extensive modelling, such comparisons are essential. In addition, there are
Jan. 12, 1995 3
-------
Leonard Levin
significant discrepancies among sections of the report, and indeed within the same volume.
For example, Table 3-2 (IV) differs substantially from the emissions in Volume II, the
emissions inventory volume. Below are some examples:
SOURCE CATEGORY Value in Table 5-1 (II) Value in Table 3-2 (III)
(tons/yr) (tons/yr)
Utility boilers 52.8 54.0
Municipal Waste Combustion 63.5 52.3
Medical Waste Incinerators 64.7 DO NOT APPEAR
Non-ferrous Smelting 9.7 14.8
Other sections are at variance with the facts. Oil-burning power plant emissions of mercury
appear to be systematically overestimated by a factor of 10 or so, despite data on oil utility
emissions having been provided to EPA in early 1994 by EPRI (Levin & Torrens, 1994, for
these values).
• Do the Research Needs chapters... present... research projects that will address
uncertainties? (all reviewers)
Although many of the Research Needs chapters present recommendations that parallel those
called for in other technical forums (see, e.g., Expert Panel on Mercury Atmospheric
Processes, 1994), other clear needs are dismissed as irrelevant or unimportant on the basis of
the model- and assumption-specific analyses in the report, even though the report itself
notes that the conclusions reached have such specificity. For example, page 4-80 of Volume
III states that, based on the sensitivity analysis, research resources need not be put into
mercury wet deposition, but focussed instead on dry deposition and other method-specific
parameters. Such facile conclusions with regard to additional research require a re-
examination of the entire issue.
• Are the summaries of human and wildlife risk assessment sufficient for a scientific
critique? (Volume VI)
Since Volume VI itself states that "U.S. EPA has not done an estimate of actual risk from
exposure to levels above the threshold for any adverse health effect" (1st sentence, section
2.3, volume VI), the answer is no. However, this caveat (again inserted inconspicuously in
the midst of the volume) is in fact well-supported by our current state of knowledge of all
aspects of mercury. With the single exception of mercury in utility fuels and utility boiler
emissions, none of the data needed for assessing mercury health risk or management methods
D-174
-------
Leonard Levin
is currently well-founded in data. Thus, EPA's decision to forego a risk assessment on
mercury, based on their own analysis, makes sense.
• Are there major areas of uncertainty, defaults, or assumptions that were not discussed?
(Volume VI)
Although the report covers many of the shortcomings, caveats, and model-specificities of
the approach used, these are rarely quantified, nor are major alternative values or approaches
used. Indeed, there is throughout an uncritical inclusion of inappropriate values in the
derivation of parameters that form the core of later modelling steps in the report. For
instance, fish consumption rates listed in Table 4-8 (III) lists the 90th percentile fish
consumption rate as 170 g/day. Figures in BNL (1994) show a 98.5 percentile figure of 85
g/day - half as much. This is a further example of figures throughout the report that seem to
depart from any "reality checks," yet are determiners of subsequent values in the analysis.
• Critique of Uncertainty Analysis
As noted above, Section 2.3 of Volume VI states that EPA has not done an estimate of risk
from exposure above threshold. Nonetheless, the acute exposure data from the Iraqi tainted-
grain incident are used to extrapolate a revised RfD for the U.S. population. The proposed
new RfD of 1 x 10'4 (one-third the currently listed value) is less than the 5th percentile of
the calculated response, and thus exceeds the level recommended by the NAS report on risk
for a level at which to select simulation results for risk analyses.
The uncertainty analysis for the exposure modeling omits several variables which appear to
be determining of subsequent concentrations, in particular, stack height. Particularly for wet
deposition and for complex terrain effects, stack height may make a significant difference in
resulting deposition. Additionally, the values for the BAF appear skewed toward the high end
by the inclusion without critical examination of the Clear Lake, California, results to
"anchor" the probabilistic analysis. Section 3.2.3 of Volume VI notes that available measured
data near municipal waste combustors (the source category with the highest near-source
deposition rate) "do not consistently indicate local deposition of mercury around a point
source." This is another example of two parts of the report apparently not communicating
with each other, since Volume III consistently reports extremely high deposition rates in the
modelling results; one wonders why these model results are employed if they are so
inconsistent with observation.
• Critique of methods and results in comparative discussion of risk.
Section 3.2.2. presents a useful summary of the "gaps in information [that] result in
uncertainties in the risk characterization." Since this section lists essentially every element
Jan. 12, 1995
D-175
-------
Leonard Levin
that makes up the modelling analysis, it appears to be an admission that detailed quantitative
results are not attainable based on current information.
Specific Comments, by Volume
Primary Volume: Volume VI ("Characterization of ...Risks...")
• There is some evidence that the hair/blood ratio for mercury due to chronic exposure is
somewhat less than that for such acute exposures as the Iraqi poisoning episode. The
result of this would be that equivalent health (behavioral) responses would occur at higher
blood Hg levels under chronic circumstances than for acute episodes.
• Blanket statements that use of activated carbon in "mass burn combustors" captures 90%
of the mercury are still unsubstantiated, as noted below.
• Use of a bioaccumulation factor seems inappropriate, particularly where site-specific case
studies are undertaken. The BAF is an amalgam of derived values from a number of
waterways, some of which receive mercury primarily from current or formerly-active
waste discharges on their periphery. With a series of case study sites, water chemistry
details can easily be obtained for input to more accurate, and field-qualified,
biogeochemical models, such as the EPRI Mercury Cycling Model (delivered to EPA
more than a year ago). Use of this model in other studies (see, e.g., Levin & Torrens,
1994) indicates a better specification of bioaccumulation for a particular aquatic system,
more than use of a generic adjustment constant like the BAF does.
Volume II ("Inventory of Anthropogenic Mercury Emissions...")
• Table 3-5: The list of states with crematoria excludes California and Arkansas, among
other states; are these emissions significant?
• Table 4-2: Estimates of utility boiler emissions of mercury are reasonable (though high)
for coal units, but extreme over-estimates for oil and gas boilers, as shown in the
following comparison table (EPRI numbers drawn from Levin & Torrens, 1994):
FUEL EPA Figure (tons/yr) EPRI Figure (tons/yr)
Coal 48.7 43
Oil 4.0 0.2
Gas 0.2 0.001
TOTAL 52.9 43.2
• Although the aggregate total mercury emissions for oil and gas plants are insignificant
compared to coal emissions (when the correct emissions numbers are used), that is
Jan. 12, 1995
D-176
-------
Leonard Levin
irrelevant here, since EPA employs single-source scenarios. Thus, overestimating oil
emissions by a factor of, say, 20 from a given plant may give wildly high estimates of
deposition and subsequent impact under some case study scenarios used, and result in some
of the apparently high concentrations of mercury in fish that are reported. The high
numbers for oil reported here are apparently the old numbers reported in the 1993 report
by EPA, "Locating and Estimating Air Emissions from Sources of Mercury and Mercury
Compounds," rather than the more recent figures provided to EPA by EPRI about a year
ago.
• The figures for utility emissions in Table 4-4, and associated text, of 62.36 tons/year, is
in conflict with the value of 52.9 tons per year in Tables ES-5, 4-2, 5-1, and A-l. One
wonders at this point whether these are typos (i.e., last-minute revisions in the
calculations that were not revised in the report on those calculations), or actual
inconsistencies in analyses carried out over time (so that different runs reproduced in the
report actually used different numbers).
• Table A-2: The column labelled "Distillate Fuel" should probably be labelled "Residual or
Distillate Fuel;" most utility oil boilers bum residual oil
Volume III ("...Exposure from Anthropogenic Mercury Emissions ...")
• General: The lack of source modelling for near-boundary sources in Canada and Mexico
is a serious drawback to the comprehensiveness of the analysis. The approach should
include at some point a comparison of results with observations, which would require the
non-domestic major sources to be included.
• Page ES-4: Text indicates 144 metric tons/year of divalent mercury depositing to the
surface domestically from domestic emissions, while Item 2b above that text indicates
that 137.2 metric tons/year of divalent mercury are actually emitted. This appears to
violate conservation of mass.
• Page ES-5: The statement listing geographic areas with modeled deposition of greater
than 60 (Xg/m2/yr is an unreasonably alarmist linkage of "real-world" geographic
(especially particular urban) areas with purely modeled numbers that are never tied to
observed deposition rates. Also, since these are fairly large grid cells being modelled, it is
only coincidental if one grid cell is entirely urban. In particular, since no coal-fired (and
few oil-fired) power plants operate in California, the large number cited for deposition
over Los Angeles requires closer examination against observation.
• Page ES-6: The items dealing with proximity of lakes to anthropogenic combustion
sources of mercury (both beginning with "In situations where major anthropogenic ...")
Jan. 12, 1995
D-177
-------
Leonard Levin
once more are unwarranted assertions based on the specifics of the limited modelling runs
reported here, and not in any way linked to the "real world" of remote lakes near or far
from anthropogenic sources of mercury. Since biogenic sources, for example, are not
modeled, nor are waste sites, these assertions should be clearly labelled as limited
conclusions from the case study modeling done, not in any way related to conditions that
occur in actual settings (pending much more extensive data).
• Page ES-6: Fish concentrations summarized in Item 6 are higher than nearly all fish
concentrations ever measured in the U.S., even from waterways directly impacted by
mercury discharges.
• The modelled lakes appear to be extremely shallow, and would tend to have higher
mercury concentrations for a given deposition because of the lower dilution volume.
Volume IV ("Health Effects of Mercury and Mercury Compounds")
• The acute nature of the Iraqi exposure makes it less than appropriate for a re-analysis of
reference dose for use with a U.S. population.
Volume VII ("... Mercury Control Technologies ...")
• Page B-14: It is stated that chloride levels are assumed high enough that all flue gas
mercury occurs as HgCl2. Yet page 2-21 of Volume VII states that ionic mercury ranges
from 12 to 99 percent of total mercury, with an average of 79 percent. Thus, how can
100% be assumed ionic, since that exceeds not only the mean but the extreme of the
observations? This is another example of the report's consistent selection of high-end,
rather than central-estimate, values in calculations that significantly determine later
values of exposure and potential risk.
• Page B-14: Despite EPRI measurement data for oil-fired boilers showing a geometric
mean mercury concentration of 0.5 ng/dscm, the EPA report "assumes" a value of 2
|ig/dscm. In essence, the control measures modeled are capturing more mercury than is
actually present!
• Page B-15: The report "assumed" that carbon injection removes 90 percent of mercury
in coal plant flue gas, and 50 percent in oil flue gas. There are in fact no data to support
any capture efficiency for carbon injection in oil plants, so that the 50 percent figure
appears to be without foundation.
CONCLUSIONS
Although EPA's goal was commendable — a comprehensive analysis of mercury emissions,
transport, deposition, bioconcentration, and effect on ecosystems and health — and although
Jan. 12, 1995
D-178
-------
Leonard Levin
the report drew heavily on the considerable technical expertise of the EPA staff scientists,
the result is seriously flawed in light of its implications for national discussion of
environmental policy. Generally, the report has erred by choosing not only highly
conservative, but unrealistically conservative, parameter values for the modelling effort
presented.
The key aspect of the analysis is that it is based solely on the use of untested, unreviewed
models developed by EPA for this purpose, with no validation conducted against field data to
test the models. Although the report properly notes many of the caveats that should be
incorporated in an assessment of the mercury issue, one is left with the feeling that the
analysis carried out has little substantiating observation to back up its component parts, and
that the values reported here have such great uncertainty that any decisionmaking based on
them is foolhardy. Given that, more involvement of outside reviewers and advisors at an
earlier stage, with direct input to researchers designing the analysis, would have been
extremely helpful.
It would be unfortunate if this product is forwarded to Congress, or other national bodies,
without substantial alteration. The report takes the proper cautionary stance throughout
with regard to remaining uncertainties and the inability to extrapolate results very far, but
nonetheless puts so many figures and "factoids" in the public arena that nonspecialists are
likely to misapprehend the limitations placed on the analysis by our current state of
knowledge of mercury. That current uncertainty (although much less now in some areas,
such as utility emissions, than it was 5 years ago) can only be rectified by additional research,
not by issuing reports with the caveats found in the middle of sections, but the much weaker
numerical conclusions featured at the front.
BIBLIOGRAPHY
• Brookhaven National Laboratory, 1994. DOE/FDA/EPA Workshop on Methylmercury
and Human Health. Biomedical and Environmental Assessment Group, Brookhaven
National Laboratory, Upton, N.Y. Report Conf-9403156.
• E. Constantinou, X.A. Wu, and C. Seigneur, 1995. "Development and application of a
reactive plume model for mercury emissions." Water, Air, and Soil Pollution (to appear).
• K. Crump, J. Viren, A. Silvers, H. Clewell, J. Gearhart, and A. Shipp, 1995. "Re-analysis
of dose-response data from the Iraqi methylmercury poisoning episode." Risk Analysis (to
appear)
Jan. 12, 1995 D-179
-------
Leonard Levin
Expert Panel on Mercury Atmospheric Processes, 1994; Mercury Atmospheric
Processes: A Synthesis Report. Electric Power Research Institute, Palo Alto, California.
EPRI Report TR-104214.
• J.M. Gearhart, H.J. Clewell III, K. Crump, A. Shipp, and A. Silvers, 1995.
"Pharmacokinetic dose estimates of mercury in children and dose-response curves of
performance test in a large epidemiological study." Water, Air, and Soil Pollution (to
appear).
• L. Levin, I. Torrens, et al., 1994; Electric Utility Trace Substances Synthesis Report.
Electric Power Research Institute, Palo Alto, California. EPRI Report TR-104614 (4
volumes)
• U.S. EPA, 1993. Locating and Estimating Air Emissions from Sources of Mercury and
Mercury Compounds. U.S. Environmental Protection Agency, Research Triangle Park,
' N.C. EPA/454/R-93-023.
Jan. 12, ,995
-------
Pamela Shubat
D-181
-------
Pamela Shubat
January 10,1995
Review of the Mercury Study Report to Congress. Volume VI: Characterization of Human
Health and Wildlife Risks from Anthropogenic Mercury Emissions in the United States.
Comments specific to the questions/instructions in "Charge to reviewers"
1. Are the summaries for wildlife and humans sufficient for a scientific critique ?
The various BMDs and reference doses were difficult to keep straight. More descriptive references
are necessary. The in-depth appendix on the BMD and RfD (Appendix C Volume IV) confused this
issue because it appears that the RfD is NOT used for the risk characterization step. It is not clear
what the RfD is to be used for. It is also confusing to refer to the NOAEL for human health in
section 4.4 if it is actually the modelled BMD that was used. Finally, the confusion is exacerbated by
the use of different values in different sections of the report. A benchmark dose of 1 ug/kg/day is
used in Volume VI, section 2.1.1. A benchmark "dose" of 11 ppm Hg in hair is used in Volume IV,
page 5-17, and is not actually a dose but a hair concentration. A dose of 1.1 ug/kg/d is calculated
(page 5-18). A developmental marker threshold of 12 ppm in hair is used in the uncertainty analysis
of the pharmacokinetic model shown in Appendix c, Volume IV, page C-9.
The following simplified description, used in section 2.1, would be helpful:
A BMD for the hair concentration corresponding to a 10% adverse response was derived
from a dose-response curve (show the curve and where the work is described). The resulting
value of 11 ppm mercury in hair is equivalent to a level of 44 ug/1 in blood (Volume IV,
page 5-17). This calculated threshold for adverse effect was used in a pharmacokinetic
model to determine the chronic intake level that would be necessary to sustain this blood or
hair level (Volume IV, page 5-18). The resulting intake estimate of 1 ug/kg/d is used as a
surrogate for a NOAEL in the risk characterization step (Volume VI, table 4-1, page 4-4 of
Volume VI). The estimated NOAEL of 1 ug/kg/d is also used to derive a Reference dose of
IxlO"4 mg/kg/d (Volume IV, page 5-22«or was it supposed to be a NOAEL of 3.4 ug/kg/d?
Something seems to be wrong). The distributions for the parameters used in the
pharmacokinetic model, including the BMD and the uncertainty factors used to derive a
reference dose, are subjected to Monte Carlo simulations in an uncertainty and sensitivity
analysis (Volume IV, Appendix C).
D-182
-------
Pamela Shubat
January 10, 1995
It was not clear when the value in question was the LOAEL or RED developed in 1994 by the EPA's
RfD work group and when the value was the BMD developed for this report (1 ug/kg/d). This is
even more confused because two EPA NOAELs were discussed in Volume IV, 1.1 ug/kg/d and 3.4
ug/kg/d. From reading this draft, it is not clear if there are simple editing mistakes where the two are
confused, or if there is a deliberate intention of using both-one from a early and one from a later
EPA evaluation. This problem shows up in sections 2.1.2 (page 2-4) and 2.2 (page 2-7) of Volume
VI and on page 5-22 of Volume IV.
There seems to be very little documentation, either in this volume or in Volume IV, of the adult BMD
and RfD based on paresthesia. In one description of the data (pg 5-21, Volume IV), it is stated that
quantitative estimates of exposure were based on blood concentrations in the affected population, but
the population was not described. In another section (appendix c, page C-2, Volume IV), thresholds
were determined on the data shown in table c-1 (pregnant females), but the number of individuals
involved was not clearly stated. In Volume VI, page 2-7, it says 35 female subjects were used hi the
BMD modelling. On page 2-8, there is a long discussion of the adult data from the Bakir data,
implying the general population, while in the next paragraph the 81 pregnant mothers are discussed.
I cannot tell from this discussion just who was used for the adult BMD modelling and how they
compare either to the Iraqi population or any other group.
Finally, there is no overall discussion of how critical endpoints should be chosen from amongst the
various endpoints discussed in the animal data on neurological effects of methylmercury.
2. Are there major areas that were not discussed?
Population risks and discussion of the extent and magnitude of potential hazardous exposures. See
number 4, below.
3. Critique the uncertainty analysis.
The quantitative uncertainty analysis (Appendix C, Volume IV) involves the pharmacokinetic model
used to derive a Reference Dose. But, the RfD is not used in the final step of the risk assessment,
instead the BMD-derived surrogate for a NOAEL is used. Much of the uncertainty is attributed to
the use of uncertainty factors, such as uncertainty for latent effects. It is not clear if this work is
D-183
-------
Pamela Shubat
January 10,1995
therefore applicable to the use of the BMD in devising the fish tissue tables that comprise the risk
characterization. The uncertainty analysis is not used to develop a spread of numbers for use in the
risk characterization that follows. It is not used to justify the use of a LOAEL or NOAEL in the risk
characterization. It is not clear how this uncertainty analysis should be used in conjunction with the
final risk characterization that is performed. On the other hand, it seems to be an excellent analysis
of the sources of variability and uncertainty for a reference dose.
The reference dose is derived from Iraqi data using a blood volume for a pregnant woman. It may be
useful to derive the safe intake level for a nonpregnant woman or male adult (a value of 0.91
ug/kg/day). This number can be used in the uncertainty/sensitivity analysis.
4. Critique the methods and results of the comparative discussion of risk.
The risk assessment could characterize the extent and magnitude of current, ambient mercury
exposure exceeding the NOAEL or LOAEL and then discuss the implications of that risk. For
example:
How many children and adults are in the classifications that have been devised
Where are the vulnerable populations (humans and wildlife) located relative to distributions
of high fish tissue levels? Describe the geographical and demographic (socio-economic)
distribution of the high exposure humans.
How high are exposures? Are anticipated exposures high enough that further research
would document health effects?
If these type of questions are not answerable, they should appear in the further research section. If
they are site specific, that needs to be spelled out as the missing information that would complete the
assessment of risk at the local level. A risk assessment should end up with population and unit risk
characterizations even if it is for selected subpopulations. This is made difficult when the toxicant is
not a carcinogen because we are not used to dealing with such numbers. But incidence of adverse
effect for the highest exposure groups can still be characterized using the dose-response curves
available (i.e. a possible 10% of the population consuming fish contaminated with X ppm mercury
D-184
-------
Pamela Shubat
January 10,1995
at a rate of one meal per week would experience transient paresthesia). These potential health
concerns can be compared to incidence of these adverse effects in the general population.
Page 1-2 of Volume VI mentions the elements that are still missing: presentation of the risk estimate
and communication of the risk analysis. I fear that this volume will not convey to the public the
findings of the EPA because definitive statements characterizing the extent and magnitude of
underlying and incremental risk of mercury toxicity are missing. The obvious has not been stated
The concentrations of mercury in trophic level 3 and 4 fish shown in table 4-3 are already present in
fish sampled throughout the country.
There was no discussion of using rat and monkey data in the wildlife health effects assessment—
either as endpoints or as a way to describe the comparative doses that produce different levels of
health effects.
The latent period should also be discussed further. Page 4-73 of volume four reports a latent period
of 16-38 days in Iraq (for neurological effects) and a latent period of up to several years in
Minamata.
General comments:
Somewhere near the end of this volume it should be sated that the values for harmful levels in fish
tissue that are derived to characterize risk are not the same as the values that states and other entities
may use to issue fish consumption advisories or to perform other risk assessments. Although much
discussion has been devoted to derivation of a reference dose, that dose, which includes factors to
account for uncertainty in the data, was not used for the purposes of characterizing risk. Discuss why
the reference dose was not used, what was thereby left out (uncertainty factors), what the strengths of
using the LOAEL and NOAEL are, and how the uncertainty around the LOAEL and NOAEL (or
BMD) should be characterized.
It would also be helpful, and assist in describing the context for this risk assessment, to compare the
RfD values in this report with values that are currently in use -- such as the old RfD in IRIS and the
numbers used by FDA in their recent analysis (alluded to but not well-described). Although this type
D-185
-------
Pamela Shubat
January 10, 1995
of discussion is clearly risk management, the risk assessment should address these factors since they
are what the risk manager will want to consider in determining the impact of this risk assessment on
programs now in existence. This is covered to some extent in section 5.4, Volume IV, but it does not
offer a critique and comparison of FDA and older EPA methods.
In general, the conclusions lack detail and quantitive information. This volume would be improved
by providing more detailed directions on where in the other volumes the reader can find the detailed
information. The final risk characterization does not have a description of ambient conditions and
risks or a projection of anticipated conditions and risks. Statements are made that modelling mercury
accumulation in fish results in tissue levels that exceed the LOAELs and that mercury in fish may
pose health risks to these species. The risk characterization for the hypothetical situations used in
the tables is adequate to describe levels offish contamination that would be a concern. But the
reader is directed back into previous volumes in order to compare the resulting values with measured
values in fish. Some of the conclusions from Volume III could be refined and repeated here. Most
important are quantitative descriptions of how many are being exposed to how much with what
attendant health effects and uncertainty. Perhaps some of that is being discussed in the missing
executive summary.
D-186
-------
Pamela Shubat
January 10,1995
Comments on specific sections of the volume.
Section 2.1 Human Health Effects Assessment
The second paragraph is a very good, succinct description of the different targets of mercury toxicity.
However, the statement that "methylmercury is the form to which humans are most exposed"
dismisses the previous suggestion of immune-mediated toxicity of inorganic mercury without
sufficient explanation. At the very least, this paragraph should say that all of the health effects
relevant to airborne and environmental exposures are discussed in section 2.1.3. Paragraph 2.1
should close with a description of what will be discussed in the succeeding sections (methylmercury,
cancer and other). Then, section 2.1.3 should be amended to include a discussion of each of the
endpoints discussed in 2.1. The most helpful characterization of these various health effects will be
estimates/ranges of toxic doses or reference doses when available. Then some short discussion
pointing out the obvious-the most sensitive health endpoint is associated with methylmercury.
Section 2.1.1
I strongly suggest that the description of the benchmark dose(BMD) also include the modifier of
"95% lower confidence limit on a 10% effect level" (Volume IV, page 5-20). I did not find the
modifier that would fully describe the chosen BMD anywhere n the section. A graph of the data
would clarify the dose-response analysis that produced the BMD. BMDs are not widely used, but are
of great interest to people. Since there is no general agreement on the percent response level
appropriate for a BMD, this should be a part of the description of the chosen number. I also suggest
adding the following item from Volume IV, page 5-20: "the 10% level for the benchmark dose
roughly correlates with a NOAEL for developmental toxicity data." Combined with the above, my
suggestion is to add a sentence to the first paragraph of section 2.1.1 as follows:
The RfD value is IxlO"4 mg/kg body weight/day based on a benchmark dose of 1 ug/kg body
weight /day where the critical effect is developmental neurologic abnormalities in human
infants; paresthesia in the mothers also occurred at slightly higher benchmark doses. The
benchmark dose estimate is the 95% lower confidence limit on a 10% incidence level of six
adverse neurologic symptoms and correlates with a NOAEL for developmental toxicity data.
D-187
-------
Pamela Shubat
January 10, 1995
In paragraph two, I assume the statement "Fetal effects of methylmercury exposure were based on
hair mercury analysis" was intended to mean "Fetal exposure to methylmercury was based on hair
mercury analysis..." Otherwise, the statement "Fetal effects of methylmercury exposure were based
on studies of 81 Iraqi infants" would also be correct.
In paragraph three (page 2-1), it is stated that the RfDs are based on benchmark doses for human
populations. This is confusing since the RfD is not used in the risk assessment (section 4.4) and in
fact it is a BMD model of the data that is a surrogate for the chronic NOAEL.
Paragraph four (page 2-2) mentions the appropriateness of extrapolating a short-term exposure to a
chronic exposure. Due to methylmercury's relatively short half-life, it might be good to describe the
difference between short-term and long-term in terms relative to half-life. For example, is the RfD
for an adult (but not fetal protection) meant to protect someone who has reached an equilibrium of
Intake and elimination? In that case, it is not a lifetime exposure but a year-long exposure. This has
important implications because (1) paresthesia appears to be reversible after acute or subchronic
exposures, (2) there are no lifetime human exposure data, and (3) there may also be latent effects that
occur late in life but are not due to a lifetime of exposure. The benchmark dose for fetal development
is likewise applicable to women a year before and during pregnancy. This last concept has not been
articulated but is very important is assessing who is at risk (i.e. not just pregnant women).
Paragraph six (page 2-2) mentions the "other serious disease processes present" in the Iraqi
population. This is also mentioned in Appendix C of Volume FV (page C-l) where the text suggests
that the Iraqi population are sensitive due to their health and nutrition, and in Volume VI in section
2.2 (page 2-7) "competing health problems and section 2.3 (page 2-8) "general nutritional and health
status." I could not find a description of their health and nutrition status or any comparison with U.S.
populations. This is troublesome because there is some widespread understanding that the
population was malnourished and ate the treated grain because they were hungry. However, at the
conference on methylmercury held in Bethesda, Dr. Myers made clear that the population was not
starving and had sources of protein in their diet. These statements require further explanation. Why
are the New Zealand and Canadian studies more reflective of those in the United States? I have not
seen a description of the diets of the New Zealand or Cree populations.
D-188
-------
Pamela Shubat
January 10,1995
Paragraph seven (page 2-2) discusses data needs. Is it appropriate to do so here? It may be more
appropriate to present a compendium of research needs at the end of the volume, as was done in at
least one other volume. In this brief paragraph, concerns are highlighted without an explanation of
the state of the knowledge about these concerns or the potential impact on the risk assessment.
Please direct the reader to the volume that contains a fuller explanation of these factors that may
affect the evaluation of the toxicity of methylmercury. There is an inadequate discussion in volume
IV of the mechanistic explanation for protein to alter the toxicokinetics or toxicity of methylmercury.
I am also concerned about the way selenium is discussed (pages 2-2 and 2-7). The data on selenium
are equivocal, incomplete, and potentially very important (both in an assessment of toxicity and in
discussions of mitigation). Ocean and freshwater fish may have very different levels of selenium. I
am not aware of any data that shows that freshwater fish have the high levels of selenium that may be
present in ocean fish. In addition, selenium toxicity has been a concern for wildlife in the some
Western states and might need some explanation.
Section 2.1.2
This section was well presented. The opening paragraph contains information on where (which
volume) to find the original discussion and this technique should be used in section 2.1.1.
The BMDs that are discussed on page 2-4, paragraph four, appear to be the EPA doses under
discussion by the RfD Work group, not the BMD derived for this report.
The tables that are used here and in Volume IV are very useful references.
Section 2.1.3
This section was somewhat confusing in its organization. The first paragraph would be more helpful
if it contained a description of what would be covered in the section (i.e. development, mutagenicity).
It would also be helpful to add a paragraph supporting paragraph two of section 2.1. with
information on the inhalation reference dose and kidney damage from ingested inorganic mercury.
The table which does provide information on inorganic mercury toxicity, Table 2-2, page 2-5,
appears without explanation. Finally, data on the developmental effects of methylmercury are
summarized while these same data were described in great detail as neurological effects in section
D-189
-------
Pamela Shubat
January 10,1995
2.1.1. It should be make very clear, in the first line of the paragraph (paragraph 3 of section 2.1.3)
what the relationship between neurological and developmental effects are and if there are important
differences between the two.
This section lacks an overall qualitative or quantitative summary. While each paragraph ends with a
description of the suitability of the data for risk assessment purposes, there is no overall discussion
of how critical endpoints should be chosen and how these various endpoints compare to the previous
discussion of neurological effects of methylmercury.
Section 2.2 Sensitivity of the Subpopulation
Paragraph one discusses the BMD for adults. I could not find the analysis of the data on the 35
women in Volume IV. A clear reference to that step would be helpful. I also have not found a
discussion of how these women compared to the general population in terms of sensitivity to
methylmercury. Was the dose-response relationship that they exhibited typical of the rest of the
population?
Paragraph two discusses the uncertainty over the latency period and the manner in which this
uncertainty is factored into the EPA Reference Dose. The result is a reference dose that is lower in
value than the NOAEL The risk assessment used in this volume relies instead on a BMD approach
with a resulting value that is similar to the NOAEL. The assumptions used in the BMD (a 95%
lower confidence interval and a 10% response level) do not address the uncertainty in applying the
BMD to exposure scenarios that are different than the data derived from the infant-mother pairs.
This seems to be worth discussing here, as a sensitive subpopulation might be the aged who were
exposed early in life. This is also worth discussing in a quantitative way in the uncertainty analysis.
Deborah Rice's monkey data provides some quantitative guidelines for latency .
There is a typographical error in a sentence in paragraph four that should read "This may explain the
longer latent period in Japan where people were exposed to methylmercury in fish." My earlier
comment is pertinent here in reading the next sentence. Some data should be available in one of the
volumes to substantiate the statement that a diet offish, freshwater and ocean fish, is enriched in
selenium.
D-190
-------
Pamela Shubat
January 10,1995
Section 2.3 Uncertainty Analysis - Methylmercury RfD
I am confused about the intended use of this section. It appears as though the RfD is not used in the
final step of the risk assessment, instead the BMD-derived surrogate for a NOAEL is used. The
uncertainty analysis revolved around the pharmacokinetic model used to derive a Reference Dose. If
such emphasis is placed on understanding the Reference Dose, why it is not used hi the risk
characterization?
It is not clear, from the discussion in paragraph two, whether the "susceptible subgroup" that we are
assuming is covered in the analysis is the group of fetuses or the group of adult women. It is very
possible that the most susceptible fetuses and the most susceptible adults will not be represented in
one set of infant-mother pairs. This may be explained biologically by considering the fetus as a
compartment to which mercury distributes. The more mercury that distributes to this compartment,
the less that is available to the mother. A fetus that accumulates a large amount of mercury would
potentially protect its mother from mercury toxicity. It would be useful to describe the estimated
doses or body burdens associated with paresthesia in the other populations as well as in the infant-
mother pairs.
Paragraph five discusses the "model" for estimating ingested dose levels. The term pharmacokinetic
model would help distinguish this model from the BMD model used to derive a threshold. I could
not find the calculations for the value used for the threshold parameter in Appendix C of Volume IV.
The uncertainty analysis shown in Appendix C of Volume IV for the Reference dose seems to be
quite complete, although it is not always clear when data for the distributions were available and
when some professional judgement was used (e.g. A, fraction of mercury hi the diet that is absorbed).
hi addition, the input variables hi table C-5 do not always correspond to the text (e.g. 12 ppm for the
"threshold" for hair). Finally, the uncertainty factors used in the analysis (30 fold) are not the same
uncertainty factor (10-fold) described on page 5-21 of Volume IV.
The table is difficult to interpret and the reader should not have to search the text to find an
explanation for a table heading The text does a better job describing the results of the analysis than
the table does.
D-191
-------
Pamela Shubat
January 10, 1995
Since the Reference dose is not used in the risk characterization in this volume, this uncertainty
analysis appears to be wasted. It is not used to develop a spread of numbers for use in the risk
characterization that follows. It is not used to justify the use of a LOAEL or NOAEL in the risk
characterization. It is not clear how this uncertainty analysis should be used in conjunction with the
final risk characterization that is performed.
On the other hand, it seems to be a useful analysis of the sources of variability and uncertainty for a
reference dose.
Section 2.4 Wildlife Health Effects Assessment
One of my concerns about this section is that recent information on mercury and fish reproduction is
not included (James Wiener, personal communication). In addition, there are data on loons and other
wildlife generated by the Minnesota Pollution Control Agency that have not been considered
(Mercury and lead in Minnesota common loon, MPCA 1992; Contaminants in Minnesota Wildlife
1989-1991, MPCA 1993).
Another concern is the discrepancy between the severity of effects that are considered for humans vs
wildlife. The effects observed in humans are subtle. The effects that are of concern in wildlife are
gross effects such as death. Rat and monkey data provides both endpoints-sensitive neuromotor
and/or cognitive effects as well as frank toxicity. An obvious question is why aren't some of these
animal data being used? If not used as endpoints, then perhaps used to understand the comparative
doses that produce different levels of health effects. These could then be applied to wildlife for
which only frank toxicity data are available.
The section (2.4.2) on co-location of selected wildlife species and high mercury levels would be made
more interesting and understandable with some maps.
D-192
-------
Pamela Shubat
January 10, 1995
Section 3 Characterization of Mercury Exposure of Selected Human and Wildlife Populations
Section 3.1 Introduction
A caveat that could be added to the discussion of other sources of mercury (i.e. dental amalgams and
occupational exposures) is that the risk assessment may underestimate the risk from environmental
sources of mercury, by ignoring what might be considered background exposures, such as dental
amalgams.
Section 3.2.1
It is unclear what constitutes a background level of mercury and what the range of values for
background could be. Although it has been stated that this report is not concerned with background
levels, there are obvious risk management responsibilities that make it important to consider them.
Section 3.2.2
Although there were insufficient data for assessment purposes, were there sufficient data to validate
the models? A description of how such reality checks were used would be reassuring.
Section 3.2.4.
The definitions for trophic levels 3 and 4 would be appropriate here. Bioaccumulation depends on
the duration and magnitude of exposure a fish has to methylmercury in the water and its diet. The
size of a fish is a useful surrogate for both age of the fish and its ability to catch and eat other fish.
Some researchers have developed data showing sharp increases of mercury by age classes as fish (for
example perch) switch from one prey item to another.
Section 3.2.5
Do the fish consumption rates include commercially-available fish or are they based on estimates of
recreationally-caught fish? The values that are presented are likely to be contentious, but they are
consistent with other estimates that have been used in risk assessments and are defensible. If the
values include commercially-available fish, a portion of the exposure may be estimated from the data
available on these fish. Tuna would be important to estimate since it is a major source of fish in the
U.S. and there are recent data on mercury in canned tuna.
D-193
-------
Pamela Shubat
January 10, 1995
I had not heard that mercury concentrations are high in migratory fish. What species offish are
representative of migratory fish?
Section 4 Characterization of Risk
Section 4.2 Exposure Assessment
This was a very nice, concise summary of the decision points that had been made. The assumptions
are clearly stated. The explanations for these 'assumptions are assumed to be found in the previous
volumes of the report. This is presented as an oversimplified, hypothetical situation. It would be
useful to include an uncertainty analysis that discusses propagated uncertainties and also compare the
estimates against measured values.
The uncertainties are carefully outlined, in earlier sections. However, once the final step is arrived at,
these uncertainties might be revisited.
Section 4.3 Health Assessment
In the latter part of the first paragraph on page 4-4, it is stated that the human health endpoint of
concern is fetal development. However, the data presented earlier suggest that the adult paresthesia
data are comparable to the fetal development data.
It is also stated that the severity of the wildlife and human health endpoints are not comparable.
From rat and monkey data, for which there exists both mortality data, reproductive data, and the
more sensitive developmental, cognitive, and neuromuscular effects data, an estimation of the
difference in magnitude between these gross and subtle effects might be possible. This could enter
into this discussion about the disparity between the animal and human data.
Table 4-1 is an instance of the confusion surrounding the various calculated BMDs, RfDs, and
NOAELS. The data presented as human health NOAEL appears to be the BMD model for 10%
effect level. Or is it a NOAEL? This has not been clarified in the previous sections. It is not clear
where the LOAEL for human health, 3 ug/kg/d, came from. These values should be identified in the
preceding sections and in Volume IV as values that will be used for the risk characterization.
D-194
-------
Pamela Shubat
January 10,1995
Table 4-3 is not entirely clear. Trophic levels three and four are not explained in the footnotes and
are not self-explanatory. In fact, without explanation, they appear counter-intuitive.
The statement (page 4-6) that the recreational anglers are least at risk for exceeding the NOAEL is a
little difficult to understand. Is the intention to say that recreational anglers face the smallest risk of
overexposure to methylmercury since they can "safely" eat the most contaminated fish?
Section 4.5 Conclusions
The tables that summarize the data (4-3 and 4-4) represent fish advisories for humans and wildlife.
these are useful indicators of concern and similar to the fish tissue levels of concern many states
already use in their fish advisory programs.
The concern that I have about this section is that it does not summarize the risks to the population.
1) Are these fish tissue levels of concern to be coupled with models of deposition and
bioaccumulation to anticipate geographic areas or plant types that are a concern?
2) Can these fish tissue levels of concern be coupled to fish anywhere-commercial, any
lake or river—to estimate populations at risk?
3) Are there data that describe the size of the at-risk populations (number of eagles, otter,
people who are exposed to unsafe levels of mercury)?
4) Is it impossible to discuss the quantitative results of the modelling offish tissue levels
from anthropogenic sources?
This section should be characterizing the extent and magnitude of the concern. If these fish tissue
levels of concern are useful and perhaps valid, it would be simple to compare them to data that are
collected for pollution monitoring (such as 305b reports) to describe the geographic area that already
exceeds these levels.
The obvious has not been stated The concentrations of mercury in trophic level 3 and 4 fish shown
in table 4-3 are already present in every state that has monitored for mercury in fish.
D-195
-------
-------
Volume VH
An Evaluation of Mercury Control Technologies,
Costs, and Regulatory Issues
D-197
-------
-------
Tim Eder
D-199
-------
MERCURY STUDY REPORT TO CONGRESS
VOLUME VH
AN EVALUATION OF MERCURY CONTROL TECHNOLOGIES,
COSTS AND REGULATORY ISSUES
(December 11, 1994 DRAFT)
Review Comments of Tim Eder and Wayne Sckmidl
National Wildlife Federation
Overall Comments;
• The report is a good summary of information on the costs and effectiveness of control
technologies for reducing mercury releases from coal-fired utilities, municipal and medical waste
incinerators, and chlor-alkali plants.
• The report does an inadequate job of investigating and evaluating the potential for
pollution prevention and market-based solutions to reducing mercury pollution. Pollution prevention
techniques that have potential include source separation at municipal and medical waste handling
facilities, product reformulation, bans and phaseouts, product labeling requirements, reclamation and
recycling (such as reclaiming mercury from fluorescent tubes and thermostats), and deposit-refunds.
Market-based solutions include a tax on mercury use, a government buy-back of recycled mercury to
retire surplus stocks (including the Department of Defense stockpile) thereby decreasing supply and
raising costs. The report should be revised to explore the feasibility of these solutions more thoroughly
and to propose means of gathering additional information on their feasibility.
• Our evaluation of the report is severely limited by EPA's failure to include
recommendations. The final report must include EPA's conclusions and recommendations to address
mercury contamination, a serious nation-wide problem. The recommendations we would propose to
address mercury contamination include a mix of strategies, with pollution prevention to reduce use and
disposal of mercury-containing products as the top priority. Pollution prevention alone will not be
sufficient. Additional end-of-pipe controls to reduce, to the maximum extent possible, releases of
mercury from utilities, incinerators and other sources will be necessary. Based on the information in this
report, it appears to us that control technologies will be highly effective in reducing mercury releases.
It also appears that the costs of these technologies are affordable and are balanced by the benefits that
will be derived from reducing mercury contamination in the environment.
Executive Summary
The tables included in the Executive Summary (pg. ES 2-8), especially table ES-1, are useful.
However, nowhere in the report is this information analyzed and evaluated. EPA needs to revise the
D-200
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
report to formulate conclusions and recommendations based on this information. As we discuss in more
detail below, the Summary needs to present information on the cost-effectiveness of various control
technologies in conjunction with:
• information on the relative importance of each source category in contributing to total
atmospheric mercury emissions, and
• a table or other graphical presentation of information on the range of benefits of
reducing mercury contamination.
The Executive Summary should answer the following questions: Based on the information in
this report, what solutions do EPA propose to address the problem of mercury contamination? Which
sources or industries will provide the most cost effective results in terms of mercury releases reduced?
Within sources or industries, which technologies or mix of technologies will be most cost effective?
L Introduction
Risk Management Principles;
The framework of the study is flawed in that pollution prevention technologies are given less
emphasis and evaluation in comparison to "end-of-pipe" (EOF) controls. The report does a good job of
evaluating the effectiveness of various control technologies. However, because of a lack of empirical data
on the costs, benefits and efficacy of materials separation, product reformulation, bans and other
pollution prevention approaches, there is little discussion of these types of solutions. It is likely that
pollution prevention solutions would prove to be more cost effective than EOF controls and that
pollution prevention will enhance the effectiveness of EOF controls. Pollution prevention will
undoubtedly be more politically palatable to the regulated community than more command and control,
EOF solutions.
The study performs fairly rigorous financial analyses to evaluate the cost effectiveness of
mercury control technologies, hi addition to evaluating the cost effectiveness of trying to capture mercury
from the waste stream with EOF technology, the study should have conducted equally rigorous
evaluations of the costs and effectiveness of pollution prevention solutions.
Despite the paucity of existing empirical data on materials separation and other pollution
prevention solutions, EPA should devote more attention in the final document to investigating and
exploring pollution prevention. What would it take to make pollution prevention work? What sort of
training would medical and municipal trash handlers need? What sort of systems would be needed to
encourage consumers to recycle batteries, fluorescent lamps and thermostats, switches and other
D-201
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
household products containing mercury? What sorts of incentives should government provide large
commercial establishments, such as factories and office buildings to recycle fluorescent tubes? What
would be the costs of phasing out uses of mercury in thermometers, switches and other instruments for
which alternatives and substitutes for mercury are available?
As an example, Honeywell Corporation operates a program in Minnesota to take back mercury-
bearing thermostats. How much does this cost the company and how effective is it? Does this solution
have potential application for other products?
All of these and other pollution prevention solutions need to be investigated EPA should collect
available information on pollution prevention solutions and include it in the final report. The final report
should include EPA's plans for conducting additional studies to find the answers to critical questions
necessary to make pollution prevention work.
The report framework is limited to several of the largest sources of mercury. An additional
source category that may be a large source of atmospheric emissions of mercury is refineries. Information
on mercury releases from refineries and control technologies should be added to the final report
"L_ Mercury Controls
As we stated above, the report directs little attention to materials separation and product
reformulation. The report includes no consideration of bans and phase-outs, or product labeling as a
means to reduce mercury content in products and the waste stream. The report includes no discussion
of market-based solutions, such as a tax on mercury use, or manipulation of the supply through
reclaiming and retiring mercury stocks, including the Department of Defense stockpile.
The draft overlooks several potential pollution prevention solutions, including:
• Product content bans.
• Input taxes on the use of mercury in products.
• Labeling of products to indicate to consumers which products contain mercury and
which are mercury-free. This would be especially helpful in the area of switches and devices that most
consumers would not expect to contain mercury1.
These solutions are important for a number of reasons, including the fact that currently there is
little or no incentive to prevent or avoid new uses of mercury in products. For example, L.A. Gear was
probably unaware of the environmental risks and the negative publicity that would be generated by their
use of mercury in tennis shoes. It has been reported that several automotive companies are planning to
D-202
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
switch to a new, high intensity headlight that contains mercury. What incentives or mechanisms would
prevent these new, unnecessary uses? Input taxes, bans and product labels would help.
Generally, the discussion on the effectiveness of various end-of-pipe control technologies
appears adequate to us, although this is an area where our expertise is limited.
The report overlooks the fact that most EOF control technologies merely transfer the mercury
from one form of waste to another. This media shifting is either overlooked or glossed over in the report.
This is an important advantage of pollution prevention over end-of-pipe control. All of the add-on
control technologies (carbon filter beds, wet scrubbing, activated carbon injection, fabric filters) result
in a contaminated medium that must be disposed of either by burning or land filling, often in a hazardous
waste landfill. It is not clear that these costs have been factored into the annual operation costs. These
added costs might be avoided or reduced by pollution prevention solutions, such as product bans,
reformulation or materials recycling if those programs sufficiently reduced the amount of mercury in the
waste stream.
Energy conservation. The report directs virtually no attention to the benefits of reducing energy
consumption through demand-side management and other conservation programs. Decreasing energy
production and use will decrease mercury released and provide ancillary benefits of reducing emissions
of S 02 and other pollutants.
EPA has missed an opportunity to investigate some creative market-based solutions for reducing
mercury emissions from power plants. Perhaps credits or vouchers could be given to utilities for mercury
reduction goals. Utilities would have the option of meeting their reduction targets in the manner they
deemed most efficient, either through HOP technology, or conservation, or other means.
Coal Washing The report overlooks mercury reduction that occurs as a result of coal washing.
Coal washing is apparently done to reduce sulfur content and has the added benefit of reducing mercury
content. According to Volume n of the mercury study (page 4-3)coal washing can reduce mercury
content by as much as 64 percent (average of 21 percent), but is highly dependant on the type of coal.
Why wasn't this technology given consideration in Volume VII.
3. Cost, Impact and Benefit Considerations
Again, the report directs little attention or effort at evaluating the effectiveness of pollution
prevention solutions. "Reliable cost data on battery separation programs as mercury control options were
not available, so maximum price increases arising from these programs could not be estimated." While
this statement is true, EPA should have investigated the costs of establishing a battery collection and
D-203
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
recycling program. Some communities, such as Ann Arbor, Michigan, provide curbside pickup of
batteries. Many communities and private companies offer household hazardous waste collection
programs. Some research on EPA's part could have yielded important information on the costs and
effectiveness of these programs, and how they might be adapted to be more effective in recycling
mercury.
The report correctly recognizes that materials separation should be much more effective at
hospitals. Therefore, mercury reduction should be more effective in hospitals where staff can be more
easily trained to separate from the waste stream the batteries and instruments that likely contain mercury.
There has been much discussion of this in the Great Lakes, and, in fact, EPA has established a task force
that is working with hospitals and health care professionals to develop educational materials and
programs to reduce mercury contaminated wastes in hospitals.
The costs of using activated carbon as a control technology in medical waste incinerators
(MWIs) is presented in an inconsistent and misleading manner on page 3-11. The costs of adding
activated carbon to small MWIs is averaged across the cost of hospital services, yielding an increase of
0.02 percent. For large MWIs the increased costs from adding activated carbon injection are averaged
across only the revenues of commercial MWIs, or the cost of waste disposal, as opposed to the costs of
all hospital services.
The report should recognize that the costs of installing EOP controls to capture mercury
emissions from incinerators and power plants will have added benefits by reducing other pollutants as
well. Thus, the costs should not all be attributed to a mercury reduction program.
3.2 Social Costs
We are pleased that this section is included. It is important to recognize that contamination of
the environment from mercury and other pollutants does result in social, economic and ecological costs.
Chapter 3 of Volume VII could be improved by the addition of a summary of the costs of mercury
contamination. Even though most of the costs cannot be precisely quantified or isolated to mercury
contamination alone, they can be described in qualitative terms and summarized in a table. NWF's
report, "Our Priceless Great Lakes" includes a table that lists in summary fashion all of the benefits of
clean Great Lakes. A similar format might be used in this chapter and would help summarize the costs.
This is especially important since it is much easier to quantify the costs of any given control technologies
or measures. Presentation of the list of benefits in a table or other summary would provide a balance to
the cost of controls.
D-204
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
The report mentions that in the State of Michigan there is a general advisory on all inland lakes
because of mercury contamination. Latest data indicate that 66 percent of the inland lakes and reservoirs
tested have at least some fish exceeding 0.5 ppm and 33 percent have at least one species with an
average mercury concentration at or above 0.5 ppm2. This information underscores two important themes
of this section of EPA's mercury report: mercury is a widespread, ubiquitous environmental contaminant
that is causing problems at current levels in the environment, and the problems caused by environmental
mercury contamination are resulting in very real economic costs. The State of Michigan issues an
advisory to all of this state's more than one million anglers that when fishing on any inland lake or
reservoirs they should restrict their consumption because of mercury contamination. This means that the
value of this resource is being decreased by mercury contamination. When confronted with this
information, it is likely that some anglers will decide to fish less often or not at all.
The report describes many of these and other costs of mercury in a qualitative or general manner.
It is disappointing that EPA does not have more data to include. The area of the social costs of a
contaminated fishery and the benefits of a clean fishery is a very important justification for many
environmental clean-up programs. EPA should commission additional research to quantify the economic
costs associated with contaminated fisheries and fish consumption advisories.
The report correctly notes that cultural values of fishing are an important consideration when
placing a value on a fishery that is free from contamination. We are pleased that the report recognizes
that Native Americans, some Asians and others have diets and cultures closely tied to fish and fishing.
As the report notes, "..the potential for such reductions in fishing values is clear..." However, the a
paucity of data prevents estimation of the economic and socials costs associated with mercury
contamination. This is another dimension of a serious problem which should be addressed through
additional research. As the report notes, this problem is not limited to lost values of fisheries from
mercury contamination, the same problem exists with other contaminants.
Prior to finalizing the report, EPA should contact the States of Michigan and New York and the
Province of Ontario to inquire as to whether there are any data on the lost revenues that resulted when
sport fishing in their jurisdictions was shut down due to contamination. In the early 1970's the Lake St.
Clair sport fishery was shut down due to mercury contamination from the Dow Chemical chlor-alkali
plant in Samia, Ontario, hi New York, an extensive fish consumption advisory (mostly on organic
contaminants — not mercury) in the 1980's closed down a significant portion of the Lake Ontario sport
D-205
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
fishery. Both of these states and the Province of Ontario may have some information on the lost revenues
associated with those extreme fish consumption advisories.
Of those few studies that have been conducted, the results are surprising. For example, a study
in Arkansas estimated a $5 million dollar decline in fishing related expenditures due to a mercury
contamination advisory. (And this is in a state that has only one advisory for mercury, according to figure
3-5.)
Costs of Human Health Effects
We agree that one of the benefits of decreasing mercury contamination in exposed populations,
especially children, could be higher IQS, fewer developmental problems, increased productivity and
associated benefits in later life. The report correctly cites NWF's report "Our Priceless Great Lakes,"
which was based on a 1991 study of the effects of lead poisoning in children that could be associated
with a lifetime loss in wages of 1.76 percent for each IQ point lost This section would be strengthened
by the inclusion of a reference to findings in the previous Volumes that mercury is a neuro-toxicant that
can be linked to decreased learning potential and loss of IQ in exposed populations.
4. Federal and State Authorities and Activities
The conclusions and recommendations EPA proposes in the final report should not be limited
to the authority available to EPA under the Clean Air Act. EPA should, in the final report, include
recommendations for controlling mercury by other means, including promoting pollution prevention
through non-regulatory incentives and the use of other statutes, such as the Clean Water Act, the
Resource Conservation and Recovery Act, Superfund (including SARA Title HI) and the Toxic
Substances Control Act
The report glosses over references to water quality standards for mercury and does not include
any reference to new standards to be promulgated under the Great Lakes Water Quality Initiative
(required under §118 of the Clean Water Act). Because mercury is highly bioaccumulative, the GLI will
impose extremely low standards for water discharges of mercury. These will be an important stimulus
to pollution prevention and source reduction of mercury within and upstream of point sources. In
addition, because water quality standards for mercury are already very low in some states like Michigan,
NPDES permits for municipal wastewater treatment plants currently include requirements for mercury
minimization and reduction. This is providing a strong inducement to municipalities to set up household
hazardous waste collection and battery separation programs.
D-206
-------
Tim Eder, Wayne Schmidt
National Wildlife Federation
As an example, NWF has been involved in contested case hearing negotiations with the City of
Detroit over the adequacy of the City's mercury minimization program. NWF and the City have invested
many resources in cooperative efforts to reduce and eliminate mercury releases from all sources within
the City.
An additional authority that EPA can use is to broaden the reporting requirements for mercury
use and release under the Toxics Release Inventory (SARA §313). We recommend that EPA lower the
threshold for mercury (and other bioaccumulative compounds) to releases of one pound or more. In
addition, to promote and monitor the effectiveness of pollution prevention, EPA should require reporting
on use of mercury, in addition to release.
5 Recommended Actions
This Volume lacks conclusions and recommended actions. The final report must include
recommendations to Congress and the Nation for reducing and controlling mercury contamination. The
report recognizes that mercury is causing degradation of the nation's waters and this degradation is
resulting in a variety of environmental and social costs, including impaired value of commercial and sport
fisheries, threats to human health and serious effects of panthers, eagles and other wildlife. Mercury
contamination is a serious nation-wide problem that demands a response from EPA. The report evaluates
the costs and effectiveness of a number of potential solutions: EPA would be doing a disservice to
Congress and the rest of the Nation if it does not take the report the final step and propose solutions.
EPA needs to synthesize and evaluate the information in the report and, based on this, develop
recommendations for controlling mercury. It is unfortunate that the summary and recommendations were
not completed for this review or for the January 25-26 meeting in Cincinnati. EPA's conclusions and
recommendations could have been refined and improved had they been presented in the report and
discussed by these reviewers.
Table 3-1 constitutes the closest thing in this Volume to conclusions. In addition to this
information on cost effectiveness of various technologies, the summary needs to include two additional
pieces of information:
• the relative importance of each of these sources in terms of total contribution of
atmospheric mercury, and
• a summary of the environmental costs of mercury contamination (benefits of mercury
reduction).
D-207
-------
TimEder, Wayne Schmidt
National Wildlife Federation
Some of this information is presented in the Executive Summary. It should be restated in the final chapter
to form the basis for conclusions and recommendations.
In our view the appropriate strategy for controlling mercury contamination will include a range
of actions, with pollution prevention as the priority. In addition to pollution prevention solutions, we
support and believe that the report provides justification for, national requirements for HOP controls on
mercury releases from utilities, incinerators, chlor-alkali plants and other sources. In part because of
trace contamination of mercury in many products, it is unlikely that pollution prevention and source
separation alone will be sufficient to eliminate mercury from the effluent of incinerators. Additional EOF
controls will be necessary.
It appears to us, based on our review of this Volume, that controls on releases of mercury from
power plants, incinerators, chlor-alkali plants and other sources are technologically feasible, and, in many
cases, appear to be highly effective. It also appears that the costs of these controls are not unreasonable.
When these costs are passed on to consumers, users of products, purchasers of electrical power and
people paying for disposal costs at incinerators, they will be very small. When balanced against the many
social, health and ecological costs of mercury contamination, we believe that the American public will
view the costs of mercury control as a bargain.
Enonofes
1. At a recent workshop sponsored by EPA's Great Lakes National Program Office on its Virtual
Elimination Project, product labeling found strong support from most participants (including many who
represented industry) as a'potentially effective solution for mercury-containing products.
2. Michigan Department of Public Health, 1994. Summary of Revisions to the Michigan Sport Fish
Consumption Advisory. Division of Health Risk Assessment.
D-208
-------
Edward Swain
D-209
-------
Vol. Vn review 1-10-95: Edward Swain
Review of "Mercury Study Report to Congress Volume VII: An Evaluation of
Mercury Control Technologies, Costs and Regulatory Issues"
Edward B. Swain, Minnesota Pollution Control Agency, 520 Lafayette Road, St.
Paul, MN 55155 612-296-7800
General Comments
This volume does a good job of identifying issues, and identifying control
technologies and their costs — but it doesn't do a good job of integrating issues
and costs. Perhaps what is missing will appear in section 5, "Recommended
Actions", which apparently hasn't been written yet. What needs to be integrated
is the cost of control and what people are willing to pay.
Although costs per pound of mercury removed are calculated, and there is
discussion that Americans are willing to pay over $100 a year to make their
waters fishable (page 3-22), these two approaches are not merged. Is $100 per
person per year sufficient to pay for the increased cost of electricity, lead, copper,
and other commodities due to mercury control? Or, conversely, how much
would mercury emissions go down if the $100 per person per year were spent on
the most cost-effective mercury controls? (Let us assume that regulations are
enlightened enough so that the most efficient removal methods are used first.)
Perhaps it is unfair to allocate all of the $100 to mercury control, when PCBs are
also reducing fishability. But the need for integration is still there: how much
would it cost per person in the U.S. to significantly reduce mercury emissions,
say by half, as was done for acid emissions? This document needs to answer that
question.
It should be stated somewhere that the benefit of some of the pollution control
devices extends beyond mercury control. Activated carbon injection, for
instance, captures a high proportion of dioxins from incinerators.
Another major point lost in this document is the sense that mercury control is
actually a relatively recent endeavor (compared to most other air pollution
control efforts), and that the specific dollar amounts per pound of mercury
removed may fall dramatically as research is conducted, discoveries are made,
D-210
-------
Vol. VII review 1-10-95: Edward Swain
and engineering technique fine-tuned. For instance, on page 2-26, brief mention
is made that "...iodide-impregnated carbon increased mercury removal to nearly
100 percent, an increase of 45 percent over results achieved with an equal amount
of nonimpregnated activated carbon..." Later on the page, it is noted that
"...impregnating activated carbon with chloride salts increases adsorptive
capacity of the activated carbon 300-fold..." Without being overly optimistic, it
would make sense to temper the discussion of removal costs in the Executive
Summary with a summary of emerging technologies, as discussed on page 2-26.
Other than the above, and specific comments below, I see no serious flaws with
the cost analysis presented in this volume. However, it is clear that considerably
more effort went into quantifying costs associated with flue gas treatment
technologies, in contrast to pollution prevention and materials separation. For
instance, the EPA's own documents on fluorescent lamp disposal options contain
information on recycling costs that is not contained in volume VII. Volume VII
puts too much reliance on battery collection, when the mercury content of
batteries is declining rapidly. The materials separation section should contain
information on three mercury-containing components of the waste stream:
batteries, fluorescent lamps (which will be the biggest source category soon), and
thermostats. Now that Honeywell is collecting thermostats from consumers and
HVAC contractors, Honeywell should be able to provide accurate costs on the
activity.
In general, the information in the volume gives the reader a lot to think about — I
just wish it were integrated better so that readers don't have to try to integrate it
on their own. Perhaps the limits to integration should be discussed.
Specific Questions and Comments
# page point
I p. 2-1, section 2.1 "Pollution Prevention and Other Management
Measures" The first pollution prevention measure
discussed is Materials Separation, which is not
generally regarded as pollution prevention. This
distinction is not merely semantics, because Materials
Separation is not anywhere close to 100% effective in
"preventing pollution".
D-211
-------
Vol. VII review 1-10-95: Edward Swain
2 vi
3 ES-1 (middle)
5 Table ES-2
I doubt that HVAC stands for "High volume air
compressor". Rather, it probably stands for "Heating,
Ventilating, and Air Conditioning", as used in
"HVAG dealers are required to properly manage or
recycle used mercury thermostats." (bottom of table
on page 4-9).
although a summary of a later point on page 1-2,
neither place in the volume is very clear on the
following: "e.g., fluorescent lamp breakage would
not be considered appropriate for a technology-based
standard under section 112 of the Clean Air Act". I
think what is meant here is that "fluorescent lamp
breakage as a mercury emission source category
would not be considered an appropriate category for
a technology-based standard...."
4 ES-2 (-line 30) typo: "control" is spelled "contorl'1
In the last column, what does the dash mean after 0.8
and 2.2?
6 ES-11
7 Table 2-1
last sentence: Batteries containing mercury are not
banned in Minnesota; the use of mercury in batteries
is restricted, but not banned. The addition of
mercury to alkaline batteries will be banned in 1996.
Under the source category "MWCs", a status
comment is made for carbon filter beds: "Currently
this technology is applied to five full-scale power
plants in Germany". Are these "waste to energy"
power plants? It is unclear why power plants are
discussed here.
D-212
-------
Vol. VU review 1-10-95: Edward Swain
8 Table 2-1
9 Table 2-1
Under MWCs, activated carbon injection is listed as
only being applied on a pilot scale on 3 MWCs. The
HERC facility in Minneapolis (1000 tons/day) has
been using activated carbon injection since February
1994, achieving more than 95% level of control.
Under utility boilers, "installation" is spelled
"instillation"
10 2-4 (top)
Hennepin County does not send alkaline and carbon-
zinc batteries to hazardous waste landfills, as stated.
They do not have any reason to, because the batteries
do not test hazardous.
11 2-4 (bottom)
12 2-4 (bottom)
"It projected that by 2000 fluorescent lights will
account for about 24 percent of the mercury in MSW."
This projection is undoubtedly a very low estimate,
because it assumes that the production of mercuric
oxide batteries will not decline, whereas it is because
of state actions and manufacturer phase-out.
"Currently, there are few locations in the United
States where the mercury from such lights can be
recovered." is a misleading statement. While it is
true that there are few recovery sites, the capacity for
recovery probably encompasses a majority of the
lamps used in the U.S.: the users simply have to ship
the lamps to the recovery sites, which have great
excess capacity for local needs. In addition, recycling
capacity is increasing rapidly across the country as
state regulations take affect. Entrepreneurs have
shown themselves to be responsive on short time
frames to opportunities created by regulations.
D-213
-------
Vol. VII review 1-10-95: Edward Swain
13 2-10 It is hard to believe that carbon filter beds are used to
remove acid gases. Perhaps the key phrase is in the
first paragraph, "Carbon filter beds have been
developed in Europe for use as a final cleaning
stage..." (italics added). It is not mentioned early
enough that acid gas control occurs before the carbon
bed.
14 2-16 "Once the lifetime of the filter mass has expired, the
HgSe mass is landfilled (it is not combustible)."
Perhaps it is not combustible, but is the Hg
recoverable through retorting, as it might be from
activated carbon? Mercury selenite may be "a very
stable compound, and the filter vendor indicated that
laboratory leach tests showed negligible leaching" but
the vendors definition of negligible may differ from
mine. On the other hand, the world needs a way to
immobilize mercury in the ground. Is this a valid
method? There is a lot behind these seemingly
innocuous sentences in the report.
15 2-18 (top) Under the Commercial Status and Performance of
Activated Carbon Injection, there is no mention that
the HERC facility in Minneapolis (1000 tons/day),
which has been using activated carbon injection since
February 1994, achieving more than 95% level of
control.
D-214
-------
Vol. VD review 1-10-95: Edward Swain . -.
16 Table 2-2, p 2-19 I'm concerned that tables such as this make oil-fired
utility boilers look clean in terms of atmospheric
mercury emissions, when in fact most of the mercury
in oil was already driven off during the refining
process, which is essentially unstudied (volume n,
page.4-59: "...no estimate of mercury emissions could
be made for this source category." I would like to see
a note to this effect in volume VII: 'It is not known if
the use of oil fired utility boilers releases less mercury
emissions than coal-fired boilers because the mercury
release during refining is essentially unstudied."
17 3-29 (bottom) FWS 1993b is not in the reference list.
18 3-30 Fig 3-6 FWS 1993a is not in the reference list.
19 3-32 "Another primary source of mercury is peat bogs,
which released 8.2 tons of methylmercury in 1991."
This statement is unattributed, but probably came
from Roelke et al. 1991. I wouldn't call peat bogs a
primary source, and I doubt that the specific figure of
8.2 tons of methylmercury is accepted by the
researchers now pursuing a detailed understanding of
mercury sources in Florida. Is it really
methylmercury?
20 4-2, Table 4-1, "Federal Mercury Controls":
The Surface Waters Ambient Water Quality Criterion
seems to be inaccurately listed as equal to 0.144 ug/L.
I believe that it is 0.012 ug/L.
21 4-9 In "white goods" for Minnesota, "The production and
distribution of mercury thermometers and
thermostats are limited" is inaccurate. It should read,
'The distribution of mercury-containing fever
thermometers is restricted."
D-215
-------
Vol. Vn review 1-10-95: Edward Swain
22 4-10 Batteries, Minnesota: "Mercury concentrations in
batteries must be < 0.25 mg by weight" is inaccurate
and should be deleted; it seems to be mixing two
accurate statements that should be added:
"Mercury concentrations in alkaline batters must be
less than 0.025% by weight until 1996, when mercury
additions are banned."
"Button batteries may not contain more than 25 mg of
mercury unless an exemption is granted."
23 Appendix B In general, Appendix B seems to use symbols and
acronyms that were not defined on page vi. For
instance, "gr" is used throughout, starting on the
middle of page B-l, apparently as an abbreviation for
grams, which is abbreviated earlier as "g". The
Acronyms in the tables of Appendix B are also not
defined on page vi, like PE for Purchased Equipment.
24 B-l 2nd line from bottom: "Hennepin Country" should
be "Hennepin County"
25 B-2 "Once household batteries have been collected they
must be disposed of at a hazardous waste facility or
sent to a metals recyder." This is not true. Only
batteries that test hazardous need be dealt with in this
manner. Alkaline, carbon-zinc, zinc-air, and lithium
all are OK to put into the regular waste stream.
26 B-2 2nd paragraph: The statement is made that the only
mercury-containing batteries that can be recycled are
mercury-zinc batteries. I believe that mercuric oxide
batteries can be recycled also.
D-216
-------
Vol. VII review 1-10-95: Edward Swain
27 B-2 "No facility recycles alkaline batteries or other
household batteries." This is not true; alkaline
recycling is coming, and certainly Ni-Cd, sealed lead
acid, and silver oxide batteries are now being
recycled.
28 B-2 footnote 1: The division symbols look like "+" on my
copy, which may be confusing to some readers.
29 B-6,B-28 Tables B-3 and B-18: The footnote symbols and the
footnotes themselves are overly tiny in these tables.
30 B-24 2nd paragraph: 'The cost effectiveness for the oil-
fired boiler is large because mercury concentration in
the oil is low compared to coal, and mercury removal
efficiency is also assumed to be lower than that for
coal." (italics added). There is a communication
problem here. Large is used here in the sense of poor,
in that the cost per pound of mercury removed is
large. Certainly the cost effectiveness is not high.
How about changing this to read, "The cost
effectiveness for the oil-fired boiler is poor because.../'
31 B-31 Last paragraph: Again, there is a communication
problem: "The cost effectiveness value, however,
decreases to $529/kg of mercury removed because of
the additional mercury collected annually." Some
readers may take the sentence to mean that the value
of mercury removal is less, whereas, the author
actually means that mercury removal is more cost
effective. I suggest changing the sentence to, "Cost
effectiveness improves to $529/hg of mercury
removed because...."
32 B-31 The last sentence on the page has the same problem as
noted above.
D-217
-------
Vol. VH review 1-10-95: Edward Swain
33 B-32 The units of the last sentence of the top paragraph
should be $550/lb and $614/lb rather than per ton.
D-218
------- |