Order Code RL32744
CRS Report for Congress
Received through the CRS Web
Mercury Emissions from Electric Generating
Units: A Review of EPA Analysis and
MACT Determination
January 21, 2005
Dana A. Shea, Larry Parker, James E. McCarthy,
and Thomas Chapman
Resources, Science, and Industry Division
Congressional Research Service ˜ The Library of Congress

Mercury Emissions from Electric Generating Units: A
Review of EPA Analysis and MACT Determination
Summary
The Environmental Protection Agency (EPA) has identified mercury as the
hazardous air pollutant emitted from electric generating units that is of greatest public
health concern. As a result, EPA proposed regulations limiting electric generating
unit (EGU) mercury emissions. In December 2000, EPA committed to promulgating
Maximum Achievable Control Technology (MACT) emissions limits under Section
112 of the Clean Air Act.
Section 112 sets specific requirements for MACT standards. For new facilities,
the MACT standard must be at least as stringent as the degree of emissions control
achieved at the best controlled similar source. For existing facilities, the MACT
standard must generally achieve limits equal to the average performance of the best
12% of comparable sources. Determination of this performance is complicated by
the lack of installed commercial technology specifically for capture of EGU mercury
emissions, thereby necessitating data collection on other existing technologies and
extensive analysis of potential control levels. To determine the level of allowable
mercury emissions, EPA collected data regarding coal composition and mercury
emissions from an 80 EGU sample. Analysis of these data led EPA to subcategorize
EGUs and propose MACT standards for each subcategory in January 2004.
The proposed MACT standards have been criticized by a wide range of
stakeholders on several criteria, including EPA methodology to determine the
allowable emissions threshold. The EPA methodology incorporates two statistical
treatments, the use of, first, a 97.5% upper confidence limit to account for the
variability in input coal, and, second, another 97.5% upper confidence interval to
account for variance in plant operation. The result is proposed standards that are
substantially less stringent than the average emissions rate of the top 12% of the 80
unit sample. While EPA justifies these statistical treatments, others assert they
unnecessarily weaken the proposed regulation. Indeed, the proposed MACT standard
allows a greater amount of mercury emission than most stakeholders, from all
viewpoints, had recommended prior to the EPA proposal. The proposed MACT
standards (pounds of mercury per trillion Btu) are shown, along with the average
mercury emission and the results of the first statistical treatment, in the table below.
Coal
Average Measured
Average 97.5% Upper
Proposed MACT
Subcategory
Mercury Emission
Confidence Limit
Level
Bituminous
0.1
1.1
2.0
Subbituminous
0.8
3.1
5.8
Lignite
5.0
7.8
9.2
IGCC
5
6
19
Waste Coal
0.09
0.14
0.38
The EPA has agreed to promulgate final mercury emission standards by March
15, 2005 for EGUs. Whether these standards should follow the proposed regulation
in its current form, an alternate form, or through an alternate mechanism is a topic
of likely congressional interest. This report will not be updated.

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Maximum Achievable Control Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
EPA Methodology for Generating MACT Levels . . . . . . . . . . . . . . . . . . . . . . . . . 5
Alternate Interpretations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Sampling Concerns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Subcategories and MACT Floor Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Variability and MACT Floor Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Appendix A: Effects of Statistical Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
List of Figures
Figure 1. Effect of Statistical Treatment on Proposed MACT Standard
for Bituminous- and Subbituminous-fired Facilities . . . . . . . . . . . . . . . . . . . 8
Figure 2. Proposed Mercury MACT Relative to Projected Mercury Emissions
for Facilities Used to Set MACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
List of Tables
Table 1. Proposed MACT Emission Limits for Coal-Fired Electric Generating
Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Table 2. Existing Coal-Fired Electric Generating Units Compared to MACT
Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Table 3. Comparison of Measured Mercury Emissions, 97.5% Upper
Confidence Limits, and Proposed MACT Standard . . . . . . . . . . . . . . . . . . . 7
Table 4: Pre-proposal Stakeholder Group Recommendations . . . . . . . . . . . . . . . 11
Note: Thomas Chapman is on detail with the Congressional Research Service from
the National Science Foundation.

Mercury Emissions from Electric Generating
Units: A Review of EPA Analysis and
MACT Determination
Introduction
Mercury is a heavy metal and potent neurotoxin that bioaccumulates and
concentrates as it passes through the food chain. The Environmental Protection
Agency (EPA) considers it “both a public health concern and a concern in the
environment.”1 These concerns have grown in recent years as research has indicated
mercury’s presence at significant levels in numerous species of fish. Analyses of
human dietary intake and resulting levels of mercury in blood have pointed to
potential health risks from mercury ingestion, particularly for women of child-bearing
age.2
To limit mercury exposure, 45 states have issued fish consumption advisories.3
At the same time, EPA and the states have promulgated a series of regulations to
reduce the amount of mercury used in products4 and limit emissions for most sources
of mercury emitted to the atmosphere.5
U.S. emissions of mercury to the air come from eight principal sources. Of
these, the largest source, and the only major source for which emission standards
have not yet been promulgated, is coal-fired electric generating units. Coal-fired
1 U.S. EPA, “Regulatory Finding on the Emissions of Hazardous Air Pollutants from
Electric Utility Steam Generating Units,” 65 Federal Register 79830, December 20, 2000.
2 For specific information see U.S. EPA, “Office of Children’s Health Protection, America’s
Children and the Environment: Measures of Contaminants, Body Burdens, and Illnesses
,
2nd edition, February 2003. See also Kathryn R. Mahaffey, “Methylmercury: Epidemiology
Update,” presented at U.S. EPA National Forum on Contaminants in Fish, San Diego, CA,
January 26, 2004, p. 5, available online at
[http://www.epa.gov/waterscience/fish/forum/2004/presentations/monday/mahaffey.pdf].
3 Twenty-one states (primarily in the Midwest and Northeast) have issued statewide
advisories for mercury in all their freshwater lakes and/or rivers. In all, these advisories
cover more than 13 million acres of lakes and roughly 767,000 river miles. Eleven states,
primarily in the South, have statewide advisories for mercury in their coastal waters. In
addition, Hawaii has a statewide advisory for mercury in marine fish. For more information,
see U.S. EPA, Office of Water, “National Listing of Fish Advisories,” Fact Sheet, August
2004, p. 4, available at [http://www.epa.gov/waterscience/fish/advisories/factsheet.pdf].
4 For more information, see Mercury in Products and Waste: Legislative and Regulatory
Activities to Control Mercury
, CRS Report RL31908 by James E. McCarthy.
5 For more information, see Mercury Emissions to the Air: Regulatory and Legislative
Proposals
, CRS Report RL31881, pp. 2-4 (federal regulations) and p. 7 (state regulations).

CRS-2
electric generating units account for between one-third and one-half of total U.S.
mercury emissions. In January 2004, EPA proposed regulations to limit emissions
from these units, positing two possible control mechanisms. One is to establish
Maximum Achievable Control Technology (MACT) standards for emissions of
mercury from coal-fired electric power plants; the other is to develop a cap-and-trade
mechanism for those same emissions. This report focuses on the proposed EPA
MACT standards, including critiques thereof, and does not treat the proposed cap-
and-trade alternative model.
A number of controversies have arisen concerning the proposed MACT
standards. In general, critics argue that the proposed standards are not sufficiently
stringent to meet the statutory requirements for MACT. The EPA responds that
technology to meet more stringent standards is not currently available, and that the
standards reflect the best performance achieved by currently installed control
equipment.6
This report examines this controversy, i.e., whether the proposed MACT
standards do reflect the best emission control achieved at existing coal-fired electric
generating units as required by statute. This report begins by describing the statutory
requirements and the standards proposed by EPA. It then proceeds to describe the
sampling methods used by EPA to generate data on existing mercury control levels.
The methodology used by EPA to determine the proposed MACT standard, based on
the sampling data, is described. Finally, some of the comments received by EPA on
the proposed standards are reviewed.
Maximum Achievable Control Technology
On January 30, 2004, under Section 112(d) of the Clean Air Act, the EPA
proposed the establishment of Maximum Achievable Control Technology (MACT)
standards for emissions of mercury from coal-fired electric power plants.7 The
standards rely on the statutory requirements in Section 112(d)(3) of the Clean Air
Act, which states that for new sources, “the degree of reduction ... shall be not less
stringent than the emission control that is achieved in practice by the best controlled
similar source.” For existing sources, the degree of reduction shall be (with some
qualifications) “not less stringent, and may be more stringent than ... the average
emission limitation achieved by the best performing 12 percent of the existing
sources (for which the Administrator has emissions information)...,” or, if there are
fewer than 30 sources in a category or subcategory, it shall be not less stringent than
“the average emission limitation achieved by the best performing 5 sources (for
6 U.S. EPA, Proposed National Emission Standards for Hazardous Air Pollutants; and, in
the Alternative, Proposed Standards of Performance for New and Existing Stationary
Sources: Electric Utility Steam Generating Units
, Preamble, Section IV.D.2, 69 Federal
Register 4698, January 30, 2004.
7 Standards were also proposed for nickel emissions from oil-fired power plants, but these
standards have not proven as controversial, and are not discussed in this report.

CRS-3
which the [EPA] Administrator has or could reasonably obtain emissions
information)....”8
The statute establishes a minimum requirement (a “MACT floor”), but it
explicitly permits the Administrator to impose more stringent requirements. It is
unclear how often EPA has used this authority to go “beyond the floor,” although two
examples of the Agency’s doing so involve mercury controls imposed on municipal
and medical waste incinerators in the mid-1990s.9 This report focuses on EPA’s
calculation of the current MACT floor for mercury emissions from coal-fired
powerplants, not whether the Agency should have gone beyond it.10 Thus, this report
will not discuss at any length beyond-the-floor technologies or standards.
No U.S. power plant has, as yet, installed emission control technologies
specifically designed to capture mercury emissions. Therefore, the currently
proposed MACT floor for these sources relies on calculations of mercury removal
rates (“co-benefits”) achieved by other technology that was installed to control sulfur
dioxide, particulates, or other regulated pollutants. The EPA’s objective was to
identify the technologies designed for other pollution control objectives that have the
maximum level of these co-benefits for mercury emissions.
The proposed MACT standards for mercury emissions from coal-fired
powerplants are shown in Table 1. Using the authorities provided in Section 112(c)
and (d), EPA divided coal-fired plants into five subcategories — four based on coal
type (bituminous, subbituminous, lignite, and coal refuse) and the fifth based on
combustion technology (integrated gasification combined cycle, also known as
IGCC). Each type of coal has differing properties and mercury contents.
Additionally, the number of facilities burning each type of coal varies, with the
majority of facilities burning bituminous or subbituminous coal. The EPA set
standards that vary by as much as a factor of 30, depending on the coal type. The
most stringent standards apply to coal-refuse-fired and bituminous-fired facilities,
while the other categories have less stringent standards. As shown in Table 2, the
size of the five subcategories also varies widely: 95% of the existing units fall into
8 42 U.S.C. 7412
9 At the time these regulations were promulgated, these technologies were considered
“beyond the floor,” although now they are standard controls on incinerators. The most
common of the technologies is activated carbon injection (ACI), which, when used with a
fabric filter (or “baghouse”), captures more than 90% of mercury emissions from
incinerators. Similar technology has been the subject of successful full-scale field tests at
at least four coal-fired power plants, but has not yet been commercially installed.
10 Considerable controversy has arisen over whether EPA is correct in deciding not to go
“beyond the floor.” This determination is generally supported by industry groups, who
agree that mercury-specific controls are not commercially available. Others, including
environmental groups and some state and local agencies, contend that because technologies
such as ACI exist, EPA’s MACT standards for mercury from electric utilities should have
been more stringent by imposing standards that would be achieved by the use of ACI or
some other promising technology. For an example, see comments from State and Territorial
Air Pollution Program Administrators/Association of Local Air Pollution Control Officials,
Docket ID No. OAR-2002-0056 (June 29, 2004), p. 3.

CRS-4
the bituminous or subbituminous groups while the other three subcategories have
relatively few units.
The proposed standards would apply on a unit-by-unit basis for each unit with
a capacity greater than 25 megawatts electric (MW); but the proposal allows facilities
three forms of flexibility in achieving compliance. First, emissions can be averaged
among all of the units at a facility.11 A large power plant commonly has several
units, so this provision might allow facilities to comply by installing stringent
controls on one or more of their units, while leaving other units uncontrolled.
Second, compliance will be determined not by whether a facility is within the
standard continuously, but by whether the rolling 12-month average of emissions is
within the standard. This would allow a facility to exceed the standard for days or
weeks, provided that it lowers emissions sufficiently at other times during the 12
months to offset peak emissions. The EPA states that use of an averaging period is
appropriate because mercury emissions from power plants are not an acute health
hazard.
Table 1. Proposed MACT Emission Limits for Coal-Fired
Electric Generating Units
Emission Limit for Existing Sources
Emission Limit for
Input Basis
or
Output Basis
New Sources
Unit Type
(lb Hg/TBtu)a
(10-6 lb Hg/MWh)b
(10-6 lb Hg/MWh)b
Bituminous-fired
2.0
21
6.0
Subbituminous-fired
5.8
61
20
Lignite-fired
9.2
98
62
IGCC
19
200
20
Coal-refuse-fired
0.38
4.1
1.1
Source: 69 Federal Register 4662, January 30, 2004. Limits for all sources are based on a 12-month
rolling average.
a Pounds of mercury per trillion British thermal units.
b Millionths of a pound of mercury per megawatt-hour.
Third, the standards for existing facilities are expressed both in input-based
(lbs/trillion Btu) and output-based (lbs/megawatt-hour) form. Output-based
standards provide an incentive for more efficient generation: a plant that can exceed
the efficiency factor used by EPA to correlate the input-based and output-based
standards would be able to burn more coal before reaching the mercury emissions
limit.
11 A facility is defined as “a contiguous plant site where one or more electric utility steam
generating units are located.”

CRS-5
The proposed standards do not include a percentage emission reduction
requirement. The EPA estimates that the standards would result in emissions of 34
tons of mercury from the affected sources, a 29% reduction as compared to the 1999
emission level of 48 tons.12
Sampling
The MACT standard is based on data obtained during an information collection
request (ICR) initiated by the EPA in April, 1998 to gather data on mercury
emissions from electric generating units.13 The industry was canvassed and those
data were used to determine the amount of mercury emitted with respect to the coal
burned. The collection of comprehensive data is key to understanding the range of
possible plant configurations, coal compositions, and subsequent mercury emissions.
EPA recognized that competing considerations, the cost of the data collection
effort versus the comprehensiveness of the data obtained, were important factors in
determining the final number of sites sampled.14 The EPA decided that, due to the
factors of cost and effectiveness, 3 units would be selected for each combination of
mechanical sulfur dioxide control, coal type, and particulate matter control.15 This
led to a maximum of 135 units considered, assuming that at least three units
populated each combination.16 In practice, no site existed for several of the
combinations. Data was subsequently obtained from 80 sites and considered to
establish the MACT floors.17 Table 2 provides a comparison of the number of
existing coal-fired power plants to the number of plants used to determine the
proposed MACT standard.
EPA Methodology for Generating MACT Levels
EPA chose to subcategorize the mercury-emissions data by unit type because of
the different mercury collection characteristics following combustion. A different
MACT floor was determined for each of these categories by examining the mercury-
emission rates. In the case of coal-refuse-fired, lignite-fired, and IGCC units, fewer
than 30 units exist. For these subcategories, the MACT floor was to be established
by the average emission limitation achieved by the best performing 5 sources. In the
case of subbituminous-fired and bituminous-fired units, more than 30 units exist, and
12 69 Federal Register 4661, January 30, 2004.
13 63 Federal Register, 17406, April 9, 1998.
14 63 Federal Register, 17408, April 9, 1998.
15 The EPA referred to the last category as electrostatic precipitator control, though it
included other forms of particulate control, such as fabric filters.
16 Office of Air Quality Planning and Standards, Standard Form 83-I Supporting Statement
for OMB Review of EPA ICR No. 1858.01: Information Collection Request for Electric
Utility Steam Generating Unit Mercury Emissions Information Collection Request
, U.S.
EPA, Research Triangle Park, NC, p. 13, November 16, 1998.
17 69 Federal Register 4661, January 30, 2004.

CRS-6
so the MACT floor was to be determined by the best performing 12% of existing
units.
Table 2. Existing Coal-Fired Electric Generating Units
Compared to MACT Sample
Unit Type
Number of Existing Units
Number of Units in Sample
Bituminous-fired
701
32
Subbituminous-fired
236
32
Lignite-fired
24
12
IGCC
2
2
Coal-refuse-fired
17
2
Source: 69 Federal Register 4662, January 30, 2004. Memorandum from Clark Allen, RTI
International to Jeffrey Cole, RTI International, December 2003.
The EPA found it necessary to statistically treat the data received from the
information collection request. Because the mercury-emission measurements were
limited in time, often described as a snapshot, EPA determined that it was necessary
to project the probable mercury emissions from coal burned at the facility over the
course of a year. Coal mercury-content and heating-value data were combined with
estimates of control efficiencies, based on the observed data, of various air-pollution
control equipment, in some cases taking into account the effect of coal chlorine level,
to estimate the distribution of mercury emitted.18
Rather than use the directly measured mercury emissions, EPA chose instead
to use the range of coal compositions reported by each plant to estimate an emissions
value which would be exceeded no more than 2.5% of the operating time, the 97.5%
upper confidence limit. The EPA considers this value to represent the operation of
the unit under conditions reasonably expected to occur at each unit.19
Table 3 presents the average measured mercury emissions for the facilities
included in the MACT determination, the results of the first statistical treatment, and
the proposed MACT level for each coal subcategory.
18 In cases where a sufficient correlation was established between the coal chlorine content
and the emitted mercury, the coal chlorine content was used to predict mercury emissions
from coal data. In cases where this correlation was insufficient, the estimated control
efficiencies of the existing air-pollution control equipment was used to predict mercury
emissions from coal data.
19 69 Federal Register 4673, January 30, 2004.

CRS-7
Table 3. Comparison of Measured Mercury Emissions, 97.5%
Upper Confidence Limits, and Proposed MACT Standard
(in pounds of mercury per trillion Btu coal)
Coal
Average Measured
Average 97.5% Upper
Proposed
Subcategory
Mercury Emissiona
Confidence Limit
MACT Level
Bituminous
0.1
1.1
2.0
Subbituminous
0.8
3.1
5.8
Lignite
5.0
7.8
9.2
IGCC
5
6
19
Waste Coal
0.09
0.14
0.38
Source: CRS analysis from EPA data.
a. The average measured mercury emission is for those facilities directly included in the MACT
determination, not for all facilities within the subcategory.
For bituminous and subbituminous-fired units, there are more than 30
established units; thus the MACT floor is established by the best performing 12% of
units. The EPA determined that the sample obtained from the information collection
request was representative of existing facilities, and therefore the best performing
12% of the sample would suffice to establish the MACT floor. In both the
bituminous- and subbituminous-fired cases, EPA took the four units which had the
lowest measured mercury emissions and combined the observed collection
efficiencies with coal-composition values reported by those units over the course of
a year to calculate the expected mercury emission. From these estimated emission
rates, EPA determined, for each unit, the projected mercury-emission level that
would not be exceeded more than 2.5% of the time. These four values were
averaged, and their average was then statistically adjusted to a level which would not
be exceeded more than 2.5% of the operating time. This second statistical
adjustment is intended to reflect the possible variability in operation of different units
burning the same fuel type.20 This final value was assigned as the MACT standard.
For a graphical representation of the effects of the statistical treatments on the MACT
standard, see Figure 1.
20 Ibid.

CRS-8
Figure 1. Effect of Statistical Treatment on Proposed MACT
Standard for Bituminous- and Subbituminous-fired Facilities
0
1
2
3
4
5
6
0
1
2
3
4
5
6
Source: Figure developed by CRS from EPA data regarding coal composition for electric generating
units chosen by EPA for determination of proposed MACT.
Note: The above figures display the EPA projected mercury emissions based on coal analyses of those
facilities included in the MACT determination. The horizontal axis is displayed in units of pounds of
emitted mercury per trillion Btu coal. The vertical axis is displayed in relative frequency, the number
of times a projected mercury emission was predicted for a given value divided by all predictions for
that facility. The data has been normalized by CRS so as to appropriately weight uneven reporting
between facilities. The average 97.5% cutoff is the average of the upper-97.5% confidence limit
determined from the calculated distributions for each set of facilities. The measurement average refers
to the average mercury emissions measured for the facilities included in the MACT standard
determination, not the average mercury emissions measured for all facilities of that subcategory. The
incorporation of each statistical treatment serves to make the proposed MACT standard less stringent.
Similarly, for lignite-fired units, the observed mercury emissions from the five
best performing units were adjusted to reflect the coal variability, being adjusted to
a value which would be exceeded no more than 2.5% of the operating time, and then
averaged. This average value was then adjusted so that it would not be exceeded
more than 2.5% of the operating time. This final value was assigned as the MACT
floor.
For IGCC and coal-refuse-fired units, EPA had mercury-emissions data from
only two units, even though the MACT floor is to be set from the five best
performing units.21 The EPA used the same method as was used to determine the
MACT floor for lignite-fired units, except that they only used the data for the two
units.
21 While there are only 2 existing IGCC units, there are 17 existing coal-refuse-fired units.
The information collection request only gathered data from 2 coal-refuse-fired plants.

CRS-9
The outcome of the EPA efforts to account for variations with fuel composition
within the performance of a single unit and between different units burning the same
types of coal is to significantly increase the MACT floor. The resulting MACT floor
is higher than the average of the measured emissions and higher than the emissions
average corrected for coal variability in all cases. For a graphical example of the
effect of the statistical treatments on the MACT analysis for other MACT
subcategories, see Appendix A.
Alternate Interpretations
EPA received over 680,000 public comments on its 2004 proposed mercury
rule, including almost 5,000 unique comments.22 The comments represent another
phase in a regulatory process that began in earnest with EPA’s 1998 identification of
mercury as the hazardous air pollutant emitted by electric utilities that was of the
“greatest potential concern.”23 Many of these comments build on recommendations
provided by forums and stakeholder groups that met before release of the 2004
proposal.
During the formulation of the proposed mercury rule process, EPA convened
a stakeholder forum called the Utility MACT Working Group. Consisting of
industry, environmental groups, and state and local agencies, the working group
attempted to develop recommendations to assist EPA in setting a utility mercury
MACT. Unable to develop consensus, the working group’s efforts resulted in
separate sets of recommendations from four groups of stakeholders: (1) an
environmental group;24 (2) a state and local agencies group;25 (3) a major industry
group;26 and (4) a minority industry group (the Clean Energy Group, CEG).27 Many
22 Environmental Protection Agency, Proposed National Emission Standards for Hazardous
Air Pollutants
; Notice of Data Availability, p. 10.
23 U.S. EPA, Study of Hazardous Air Pollutant Emissions from Electric Utility Steam
Generating Units — Final Report to Congress
, 1998.
24 Environmental Stakeholders included Natural Resources Defense Council, National
Wildlife Federation, National Environmental Trust, Clean Air Task Force, and
Environmental Defense.
25 State and Local Agencies included The Northeast States for Coordinated Air Use
Management (NESCAUM), State and Territorial Air Pollution Program Administrators /
Association of Local Air Pollution Control Officials (STAPPA/ALAPCO), State of New
Jersey, and the Regional Air Pollution Control Agency (Dayton, Ohio).
26 Majority Industry Group included Cinergy, National Mining Association, West
Associates, Latham & Watkins, Southern Company Generation, the United Mine Workers
of America, the Utility Air Regulatory Group, the Class of 85 Regulatory Response Group,
American Public Power Association, and the National Rural Electric Cooperative
Association.
27 CEG members included PG&E National Energy Group, Consolidated Edison, Public
Service Enterprise Group, Conectiv, Exelon Corporation, KeySpan, Northeast Utility, and
Sempra Energy.

CRS-10
of the comments subsequently received by EPA referred to recommendations
developed during this process.
As analyzed and summarized by the Northeast States for Coordinated Air Use
Management (NESCAUM) in Table 4, the four stakeholder groups developed
recommendations considerably different from each other and from EPA’s proposal.28
The following observations are among those that can be made:
! Unlike EPA’s proposal, all the stakeholder groups determined that
fluidized-bed combustion (FBC) units could achieve a more
stringent mercury MACT than other facilities.
! Like EPA’s proposal, the industry groups used some form of
subcategorization, generally based on coal type, while the
environmental group and state and local agencies group did not.
! EPA’s variability analysis resulted in less stringent MACT floors
than those discussed or recommended by any of the four stakeholder
groups.
The rationale behind these recommendations intersect in interesting ways with
the rationale of EPA’s proposals. These interactions are evident in the comments
various members of these groups submitted in response to the EPA proposal. The
following Table 4 examines these intersections in three categories: the recommended
mercury emission standard, the relevant annual mercury baseline (in tons), and the
post-MACT annual mercury emissions (in tons).
28 See NESCAUM, Mercury MACT Under the Clean Air Act: An Assessment of the
Mercury Emissions Outcomes of Stakeholder Group Recommendations
(May 8,
2003). The National Mining Association (NMA) characterized the MACT floor attributed
to the majority industry group as “illustrative,” not definitive recommendations. See NMA,
Docket ID No. OAR-2002-0056 (May 14, 2004), p. 9.

CRS-11
Table 4: Pre-proposal Stakeholder Group Recommendations
Stakeholder
Relevant Annual
Post-MACT Annual
Recommended Mercury Emission Standard
Group
Baseline (tons)
Emissions (tons)
Environmental
C 0.19 lbs/TBtu for FBC facilities
44.6
1.9
Stakeholders
C 0.21 lbs/TBtu for all other facilities for all coal types
State and
C 0.4-0.6 lbs/TBtu for bituminous and subbituminous coal
41.5
6.3-6.7
Local Agencies
Clean Energy
C 0.320 lbs/TBtu for bituminous or subbituminous FBC
44.6
13.1
Group (CEG)
facilities
C 11.984 lbs/TBtu for lignite FBC facilities
C 1.223 lbs/TBtu for all other bituminous or subbituminous
boilers
C 9.091 lbs/TBtu for all other lignite boiler types
Majority
C 3.7 lbs/TBtu for hot stack bituminous facilities
44.6
25.0-30.0
Industry
C 2.2 lbs/TBtu for saturated stack facilities burning
Groupa
bituminous coal
C 3.2 lbs/TBtu for wet stack facilities burning bituminous coal
C 4.2 lbs/TBtu for subbituminous facilities
C 6.5 lbs/TBtu for facilities burning lignite
C 2.0 lbs/TBtu for FBC facilities
Environmental
C 2.0 lbs/TBtu for bituminous coal facilities
48
34
Protection
C 5.8 lbs/TBtu for subbituminous coal facilities
Agency
C 9.2 lbs/TBtu for lignite coal facilities
Proposed
C 19 lbs/TBtu for IGCC facilities
MACT
C 0.38 lbs/TBtu for waste coal facilities
Source: NESCAUM, Mercury MACT Under the Clean Air Act: An Assessment of the Mercury Emissions Outcomes of Stakeholder Group
Recommendations
(May 8, 2003), Appendix A, and 69 Federal Register 4662, January 30, 2004.
Note: NESCAUM’s analysis generally uses a 44.6 ton annual mercury baseline (instead of EPA’s 48 ton estimate) because a small subset of units
in EPA’s Utility Air Toxic Study database lack either the coal supply or mercury emissions data used to calculate projected mercury reductions.
NESCAUM further reduced the baseline for the State and Local Agency Stakeholder group as that group’s recommendations did not cover the 11
lignite units (3.1 tons mercury) contained in the database.
a. The National Mining Association (NMA) characterized the MACT floor attributed to the majority industry group as “illustrative,” not definitive
recommendations. See NMA, Docket ID No. OAR-2002-0056 (May 14, 2004), p. 9.

CRS-12
Sampling Concerns
While the EPA is confident the data sample is representative of the industry,29
others have raised concerns whether the method used to select both the number of
sites sampled and the specific sites included in the dataset provides sample data
representative of those units already employed in industry settings. For example, the
National Mining Association (NMA) states:
The ICR emissions data represent, at best, a limited ‘snap shot’ of emissions
from a few units, taken over a very short period of time, with a limited number
of fuels, and cannot account for the wide variability of coals and process
conditions encompassed by the full fleet of U.S. utility boilers. In addition, the
units chosen by EPA for mercury emissions sampling in the ICR program is [sic]
unrepresentative of U.S. coal-fueled power plants.30
Based on an analysis conducted for NMA by AEMS, LLC (a consulting firm), that
found inadequacies and biases in the EPA 80 unit sample, NMA states:
As a result of these inadequacies the ICR Part III data cannot be used to identify
and evaluate the emissions performance of specific units that would rank in the
top 12 percent of their subcategory (or among the 5 best) on an annual basis.31
Indeed, NMA argues that the data and related analysis are so flawed that EPA should
abandon a MACT-based form of mercury regulation, stating that if EPA proceeds:
“NMA and its members reserve all potential objections to such standards arising
under the Data Quality Act and other applicable provisions of Federal law.”32
The State of Missouri, Department of Natural Resources, contending that the
MACT standard, at least with respect to subbituminous coal, is too lenient, states:
The Missouri source calculations indicate that the proposed MACT emission
limitation for existing subbituminous coal-fired units is too lenient. This
leniency was likely caused by a sampling error resulting from basing the
determinations on a sample population too small to be representative of the norm.
In this case, the sample size of 32 units used to determine the floor greatly
increased the likelihood of sampling errors.33
The EPA determined that an appropriate experimental design for the ICR was
to vary three parameters, coal type, sulfur dioxide control technology, and particulate
29 69 Federal Register 4670, January 30, 2004.
30 National Mining Association, Docket ID No. OAR-2002-0056 (May 14, 2004), p. 9.
31 NMA, Docket Number OAR-2002-0056 (May 14, 2004) pp. 16-17.
32 NMA, Docket Number OAR-2002-0056 (May 14, 2004) p. 43.
33 Department of Natural Resources, State of Missouri, Docket ID No. OAR-2002-0056, p.
9-10.

CRS-13
matter control technology, and include one facility for each combination.34 Some
groups have questioned whether an experimental strategy including a more
proportionate distribution of facilities or including technologies that are in
development, but not yet implemented, would have provided for a more stringent
mercury standard. For example, NESCAUM asserts that EPA’s attempts to treat
variability in the ICR data is flawed, in part, due to the lack of any data from mercury
specific controls, such as activated carbon injection.35 The EPA has stated that
mercury-specific control technologies are not yet commercially available for electric
generating units, and will likely not be commercially available prior to 2010, and
therefore consideration of these technologies in determining the MACT floor would
be incorrect.
Subcategories and MACT Floor Levels
A primary source of disagreement about EPA’s proposal is the use of
subcategories. While Section 112(d) provides the EPA Administrator with the
authority to subcategorize MACT standards, the criteria for determining these
subcategories are not defined. The EPA used a combination of coal types
(bituminous, subbituminous, lignite) and firing configurations (IGCC, boilers) to
develop subcategories within the utility mercury MACT. This decision has proven
highly controversial, both with respect to the subcategories chosen, and the MACT
floor determination for each subcategory.
Environmental groups have attacked the proposed subcategories as “without
rational basis, arbitrary and capricious.”36 The groups note that many sources,
including some in EPA’s data sample, use a blend of different coal types and that
combustion technologies vary little with respect to coal type burned. In addition,
these groups state that many boilers have some flexibility to adapt to different coal
feedstocks. Also criticizing the use of subcategories, the State of New Jersey
suggests in its comments that because the proposed standard for subbituminous coal
is nearly three times that for bituminous coal, utilities may be able to comply with the
proposed MACT standards by simply switching coal supplies, emitting more mercury
than before.37 That the weaker subbituminous proposal could encourage western
subbituminous coal production at the expense of eastern bituminous production has
been expressed by eastern coal interests.38
34 Office of Air Quality Planning and Standards, Standard Form 83-I Supporting Statement
for OMB Review of EPA ICR No. 1858.01: Information Collection Request for Electric
Utility Steam Generating Unit Mercury Emissions Information Collection Request
, U.S.
EPA, Research Triangle Park, NC, p. 28, November 16, 1998.
35 NESCAUM, Docket Number OAR-2002-0056 (June 29, 2004) p. 5.
36 Comments of Eleven Environmental and Public Health Organizations, Docket ID No.
OAR-2002-0056 (June 29, 2004) p. II-7.
37 Department of Environmental Protection, State of New Jersey, EPA’s Proposed MACT
Floor Standards for Mercury Emissions from Coal-fired Utility Units: A Statistical and
Analytical Assessment
, (July, 2004) pp. 11-12.
38 For example, see “EPA’s Leavitt Promises Mercury Rule Will Treat Coal Types Equally,”
(continued...)

CRS-14
These groups believe that the MACT floor should not be stratified by coal type,
but should be a simple average of the best performing facilities.39 By eliminating
coal type and EPA’s variability adjustment, the environmental groups’ joint
comments developed a MACT floor of 1.0 (10-6lbs/MWh output basis) — higher
than the environmental stakeholders recommended levels in Table 4, but generally
more stringent than EPA’s proposal. In a similar vein, NESCAUM found that if the
average observed mercury emissions from the top-performing 12% of EPA’s 80 unit
data sample, regardless of coal type, was used as the emissions standard, an
emissions reduction of 91.1% would be achieved, in comparison with the predicted
29% reduction in mercury emissions expected under EPA’s proposed MACT
standard.40
Most industry groups agreed with EPA’s determination that subcategories were
appropriate, but disagreed on their specification. In particular, several commenters
argued that North Dakota and Gulf Coast lignite should be regulated separately, not
combined. Stating that mercury content of Gulf Coast lignite was at least double that
of North Dakota lignite, the Edison Electric Institute (EEI) argued that the proposed
lignite standard would be unachievable for units using Gulf Coast lignite.41 Other
concerns were expressed about the absence of any units burning Powder River Basin
subbituminous coal in EPA sample data.42
Variability and MACT Floor Levels
A major source of disagreement surrounding the proposed mercury rule is
EPA’s variability analysis. To determine the proposed standard, EPA used a
methodology to identify and quantify expected variability in mercury emissions
because of the variability of mercury in coal. The result is proposed standards that
are substantially less stringent than the simple average emissions rate of the top 12%
of its 80 unit sample.
This variability analysis has come under particular scrutiny by all parties.
Environmental groups label the analysis as “statistical gimmicks” that result in
MACT floors that represent “virtually the worst short-term emissions from the worst
performing of the best units, and assumes that these pollution levels will persist
through the year.”43 One of the more detailed analyses of EPA’s methodology has
been done by the State of New Jersey. In a white paper by its Department of
38 (...continued)
Energy Washington Week (December 8, 2004).
39 The environmental groups would allow MACT levels stratified by firing configuration.
As noted earlier, both groups believe that “beyond the floor” regulation should be
promulgated.
40 NESCAUM, Docket Number OAR-2002-0056 (June 29, 2004) p. 3.
41 EEI, Docket Number OAR-2002-0056 (June 29, 2004) pp. 54-55.
42 NMA, Docket Number OAR-2002-0056 (May 14, 2004) p. 14.
43 Comments of Eleven Environmental and Public Health Organizations, Docket ID No.
OAR-2002-0056 (June 29, 2004) p. II-25.

CRS-15
Environmental Protection, New Jersey attacks EPA’s approach that uses variability
in short-term peak emissions to set a standard that would be based on a 12-month
rolling average,44 stating:
If it were essential to identify and quantify variability in emissions, the relevant
measure of variability would be what was experienced over 12 months, not what
was experienced in any single stack testing event. A transient peak in emissions
during the 12-month period will not cause noncompliance with the standard,
because lower emissions during the rest of the period will average out the effect
of the peak. For that reason, using the short-term worst-case scenario to set an
annual standard immediately makes the standard unnecessarily lax. If compliance
is to be based on a long-term average, EPA should be estimating long-term
averages, not worst case emissions, to determine the MACT floor.45
New Jersey proceeds to argue that EPA’s proposed standards are indeed less stringent
as a result of their variability analysis. Eliminating EPA’s variability analysis,
including subcategorization, would reduce the MACT floor to 0.18 lbs/TBtu.
Alternately, if the data were only stratified by coal type, the average of the observed
mercury emissions would be 0.12 lbs/TBtu for bituminous units, in contrast with the
proposed MACT standard of 2.0 lbs/TBtu.
In contrast, while industry generally supported EPA’s efforts, some industry
analysis found that EPA insufficiently dealt with the mercury variability issue. In
arguing for MACT floors substantially higher than EPA’s proposal, NMA asserted
that a combination of data quality and variability concerns require an increase in
MACT floor estimates for new sources to 5.1 lbs/TBtu for bituminous units and 7.4
lbs/TBtu for subbituminous units.46 According to analysis by AEMS, LLC, the EPA
MACT standards for new sources, when converted from an output basis to an input
basis, are equivalent to 0.63 lbs/TBtu for bituminous units and 2.1 lbs/TBtu for
subbituminous units.47 As stated by NMA when comparing the converted standards
developed by AEMS, LLC with those standards suggested by NMA: “These
estimates of appropriate emission limits for new units are an order of magnitude
higher than those estimated by EPA, reflecting AEMS’ more comprehensive
44 Department of Environmental Protection, EPA’s Proposed MACT Floor Standards for
Mercury Emissions from Coal-fired Utility Units: A Statistical and Analytical Assessment
,
(July, 2004).
45 Department of Environmental Protection, EPA’s Proposed MACT Floor Standards for
Mercury Emissions from Coal-fired Utility Units: A Statistical and Analytical Assessment
,
(July, 2004), pp. 3-4.
46 NMA also argued in its comments that “the combination of uncertainty in the emissions
analysis and repeat test variability means that the MACT floors for new units must admit
the possibility of measured emission levels in excess of the mercury content of the coals
(negative removals).” See NMA, Docket Number OAR-2002-0056 (May 14, 2004) p. 34.
47 In converting from an output basis to an input basis, AEMS LLC assumed a heat rate of
9,500 Btu/kwh. AEMS, LLC, Review and Critique of Data and Methodologies Used in EPA
Proposed Utility Mercury MACT Rulemaking,
(April 2004) p. 79.

CRS-16
approach to variability analyses, as well as the use of available data on mercury
content of U.S. coal.”48
Conclusion
The proposed regulations to limit mercury emissions from electric generating
units have been criticized by a wide range of stakeholders based on several criteria,
including the method used by EPA to determine the allowed emissions threshold.
The proposed emissions threshold incorporates two statistical treatments. This serves
to make the proposed MACT standard less stringent than the average of the actual
emission measurements, and less stringent than the recommendations of
stakeholders, including industry and environmental groups, prior to the MACT
proposal. While EPA justifies these statistical treatments as necessary to account for
the variability in input coal and plant operation, others assert that they represent a
significant, unnecessary weakness of the proposed regulation. Whether mercury
emissions from electric generating units should be controlled via the proposed
regulation in its current form, an alternate form, or through an alternate mechanism
is a topic of likely congressional interest.
48 NMA, Docket Number OAR-2002-0056 (May 14, 2004) p. 35.

CRS-17
Appendix A: Effects of Statistical Treatment
The effect of each statistical treatment performed by EPA serves to make the
proposed MACT standard less stringent than the average of the measured mercury-
emissions data. The EPA found it necessary to statistically treat the data received
from the information collection request for several reasons. First, the
mercury-emission measurements were limited in time, often described as a snapshot,
and may not represent the actual emissions performance of the facility. As a
consequence, EPA projected the probable mercury emissions from coal burned at the
facility over the course of a year.
The EPA developed these projections through one of two methods. In the case
of some equipment configurations, it was determined that a correlation between the
mercury emissions and the coal chlorine-content existed. In these situations, the coal
chlorine content was used to predict the mercury emissions. In other cases, a suitable
correlation did not exist. In these situations, the mercury emissions for that facility
were determined by using an empirical estimate of the control efficiency for mercury
of existing emissions-control equipment. By obtaining these projected mercury
emissions, EPA was able to determine the expected mercury emissions level over a
twelve-month period.
The EPA took the projected mercury emissions, developed from the coal
compositions reported by each plant, and determined the emissions value which
would be exceeded no more than 2.5% of the operating time, the upper 97.5%
confidence limit.49 The EPA considers this value to represent the operation of the
unit under conditions reasonably expected to occur at each unit.
The proposed MACT standard was then determined from the average of these
upper 97.5% confidence limits. Once the average of the upper 97.5% confidence
limits was determined, a second statistical treatment was performed to determine the
variation between the units. From this variation, a calculation was made to determine
what the upper 97.5% confidence limit would be for these averaged units.50 This
value was then set as the MACT standard.
The number of facilities averaged depended on the subcategory considered. In
the case of bituminous- and subbituminous-fired facilities, 4 facilities were included
in the MACT standard determination. For lignite-fired facilities, 5 facilities were
included. For waste-coal-fired facilities and integrated gasification combined cycle
(IGCC) facilities, 2 facilities were included.
49 The data at the extremes of the distribution, i.e. the highest and lowest measurements, play
a predominant role in determining the confidence interval. Therefore, the quality of these
measurements is crucial, since any inaccuracies in this data will be magnified in the
calculation of the confidence interval.
50 The number of facilities averaged influences the width of this confidence interval. The
greater the number of facilities averaged together, the lower the uncertainty regarding the
distribution of these values. This lower uncertainty translates to a narrower confidence
interval.

CRS-18
The effects of EPA’s statistical treatment of the data is shown graphically in
Figure 2.
Figure 2. Proposed Mercury MACT Relative to Projected Mercury
Emissions for Facilities Used to Set MACT
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.25
0.2
0.15
0.1
0.05
0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0
0.1
0.2
0.3
0.4
Source: Figure developed by CRS from EPA data regarding coal composition for electric generating
units chosen by EPA for determination of proposed MACT.
Note: The above figures display the EPA projected mercury emissions based on coal analyses of those
facilities included in the MACT determination. The horizontal axis is displayed in units of pounds of
emitted mercury per trillion Btu coal. The vertical axis is displayed in relative frequency, the number
of times a projected mercury emission was predicted for a given value divided by all predictions for
that facility. The data has been normalized by CRS so as to appropriately weight uneven reporting
between facilities. The average 97.5% cutoff is the average of the upper-97.5% confidence limit
determined from the calculated distributions for each set of facilities. The measurement average refers
to the average mercury emissions measured for the facilities included in the MACT standard
determination, not the average mercury emissions measured for all facilities of that subcategory.