Clean Air Act: Emerging Mercury Control Technologies Have Shown  
Promising Results, but Data on Long-Term Performance Are Limited 
(31-MAY-05, GAO-05-612).					 
                                                                 
In March 2005, the Environmental Protection Agency (EPA) issued a
rule that will limit emissions of mercury--a toxic element that  
causes neurological problems--from coal-fired power plants, the  
nation's largest industrial source of mercury emissions. Under	 
the rule, mercury emissions are to be reduced from a baseline of 
48 tons per year to 38 tons in 2010 and to 15 tons in 2018. In	 
the rule, EPA set the emissions target for 2010 based on the	 
level of reductions achievable with technologies for controlling 
other pollutants--which also capture some mercury--because it	 
believed emerging mercury controls had not been adequately	 
demonstrated. EPA and the Department of Energy (DOE) coordinate  
research on mercury controls. In this context, GAO was asked to  
(1) describe the use, availability, and effectiveness of	 
technologies to reduce mercury emissions at power plants; and (2)
identify the factors that influence the cost of these		 
technologies and report on available cost estimates. In 	 
completing our review, GAO did not independently test mercury	 
controls. GAO provided the draft report to DOE and EPA for	 
comment. DOE said that it generally agreed with our findings. EPA
provided technical comments, which we incorporated as		 
appropriate.							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-612 					        
    ACCNO:   A25457						        
  TITLE:     Clean Air Act: Emerging Mercury Control Technologies Have
Shown Promising Results, but Data on Long-Term Performance Are	 
Limited 							 
     DATE:   05/31/2005 
  SUBJECT:   Air pollution control				 
	     Cost analysis					 
	     Electric powerplants				 
	     Environmental law					 
	     Environmental monitoring				 
	     Environmental policies				 
	     Environmental research				 
	     Hazardous substances				 
	     Health hazards					 
	     Operational testing				 
	     Pollution control					 
	     Pollution monitoring				 
	     Research and development				 
	     Coal						 
	     Environmental protection				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-05-612

                 United States Government Accountability Office

                     GAO Report to Congressional Requesters

May 2005

CLEAN AIR ACT

Emerging Mercury Control Technologies Have Shown Promising Results, but Data on
                       Long-Term Performance Are Limited

                                       a

GAO-05-612

[IMG]

May 2005

CLEAN AIR ACT

Emerging Mercury Control Technologies Have Shown Promising Results, but Data on
Long-Term Performance Are Limited

                                 What GAO Found

Mercury controls have not been permanently installed at power plants
because, prior to the March 2005 mercury rule, federal law had not
required this industry to control mercury emissions; however, some
technologies are available for purchase and have shown promising results
in field tests. Overall, the most extensive tests have been conducted on
technologies using sorbents-substances that bind to mercury when injected
into a plant's exhaust. Tests of sorbents lasting from several hours to
several months have yielded average mercury emission reductions of 30-95
percent, with results varying depending on the type of coal used and other
factors, according to DOE and other stakeholders we surveyed. Further, the
most recent tests have shown that the effectiveness of sorbents in
removing mercury has improved over time. Nonetheless, long-term test data
are limited because most tests at power plants during normal operations
have lasted less than 3 months.

The cost of mercury controls largely depends on several site-specific
factors, such as the ability of existing air pollution controls to remove
mercury. As a result, the available cost estimates vary widely. Based on
modeling and data from a limited number of field tests, EPA and DOE have
developed preliminary cost estimates for mercury control technologies,
focusing on sorbents. For example, DOE estimated that using sorbent
injection to achieve a 70-percent reduction in mercury emissions would
cost a mediumsized power plant $984,000 in capital costs and $3.4 million
in annual operating and maintenance costs. If this plant did not have an
existing fabric filter and chose to install one-an option a plant might
pursue to increase the efficiency of mercury removal and reduce related
costs-capital costs would increase to about $28.3 million, while annual
operating and maintenance costs would decrease to about $2.6 million. Most
stakeholders generally expect costs to decrease as a market develops for
the control technologies and as plants gain more experience using them.
Furthermore, EPA officials said that recent tests of chemically enhanced
sorbents lead the agency to believe that its earlier cost estimates likely
overstated the actual cost power plants would incur.

     Coal-Fired Power Plant United States Government Accountability Office

Contents

  Letter

Results in Brief
Background
Mercury Controls Have Not Been Permanently Installed at Power

Plants but Are Available for Purchase and Have Shown Promising Results in
Field Tests Mercury Control Costs Depend on a Variety of Factors, and
Current

Estimates Vary Widely Concluding Observations Agency Comments

1 3 4

10

18 28 29

Appendixes

Appendix I:

Appendix II:

Appendix III:

Appendix IV:

Appendix V:

Appendix VI: Appendix VII: Objectives, Scope, and Methodology

Availability and Costs of Mercury Monitoring Technology

Summary of Field-Scale Tests of Mercury Controls

Summary of Stakeholder Perceptions about Availability of Mercury Controls

Stakeholder Confidence in Ability of Technologies to Achieve Mercury
Reductions under Three Scenarios

Sorbent Injection Cost Estimates from EPA and DOE

GAO Contact and Staff Acknowledgments

32 42 45

50

54 58 63

Tables	Table 1: Table 2:

Table 3:

Table 4:

Table 5:

Table 6:

Table 7:

Summary of Mercury Control Field Test Data 45
Stakeholder Perceptions on Availability of Sorbent
Technologies 51
Stakeholder Perceptions on Availability of Non-Sorbent
Mercury Controls 52
Stakeholder Confidence in Reducing Nationwide Mercury
Emissions 50 Percent by 2008 55
Stakeholder Confidence in Achieving Mercury Reductions
of 50 Percent at Nearly Every Plant by 2008 55
Stakeholder Confidence in Reducing Nationwide Mercury
Emissions 70 Percent by 2008 56
Stakeholder Confidence in Achieving Mercury Reductions
of 70 Percent at Nearly Every Plant by 2008 56

Contents

Table 8:	Stakeholder Confidence in Reducing Nationwide Mercury Emissions
90 Percent by 2008 57

Table 9:	Stakeholder Confidence in Achieving Mercury Reductions of 90
Percent at Nearly Every Plant by 2008 57

Table 10: Select EPA Cost Estimates of Sorbent Injection for a
100-Megawatt Coal-Fired Power Plant, 2003 59

Table 11: Select EPA Cost Estimates of Sorbent Injection for a
975-Megawatt Coal-Fired Power Plant, 2003 59

Table 12: DOE's Cost Estimates for Sorbent Injection Installed on a
500-Megawatt Coal Power Plant, 2003 61

                               Overview of a Coal-Fired Power Plant Sample    
Figures Figure 1: Figure 2: Layout of Mercury Controls at a Coal-Fired   6
                                                  Power                    
                                                  Plant                     9 
                     Figure 3: Stakeholder Perceptions about Availability  
                                               of Mercury                  
                                                Controls                   12 

Contents

Abbreviations

ACI Activated Carbon Injection
CEMS Continuous Emissions Monitoring Systems
DOE Department of Energy
EPA Environmental Protection Agency
ESP Electrostatic Precipitator
FDA Food and Drug Administration
FF Fabric Filter
FGD Flue Gas Desulfurization
MACT Maximum Achievable Control Technology
MW Megawatt
NESCAUM Northeast States for Coordinated Air Use Management
NETL National Energy Technology Laboratory

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

A

United States Government Accountability Office Washington, D.C. 20548

May 31, 2005

Congressional Requesters

Mercury, a toxic element that poses human health threats, enters the
environment through natural and human activities, such as volcanic
eruptions and fuel combustion. Coal-fired power plants release mercury
into the air when burning coal to generate electricity and were, prior to
March 2005, the largest unregulated industrial source of mercury emissions
in the United States.1 The Environmental Protection Agency (EPA) estimated
that in 1999, the most recent year for which data were available,
coal-fired power plants within the United States emitted 48 tons of
mercury into the air, or about 42 percent of the total man-made emissions
nationwide.2 The Clean Air Act Amendments of 1990 required EPA to study
the environmental and health effects of hazardous air pollutants from
coalfired power plants and determine whether it was "appropriate and
necessary" to regulate emissions of these pollutants.

In 2000, the agency determined that it was appropriate and necessary to
regulate emissions of mercury, a hazardous air pollutant, from coal-fired
power plants by requiring these plants to meet specific emissions
standards reflecting the application of control technology (the
"technology-based" approach).3 In January 2004, EPA issued a proposed rule
with two options for controlling mercury from power plants-the
technology-based approach and an alternative approach that would set a
national cap on mercury emissions and allow power plants flexibility
either to achieve reductions or to purchase allowances from plants that
achieved excess reductions (the "cap-and-trade" option).4

1In this report, "power plants" refers to coal-fired electricity
generating units larger than 25 megawatts in size that produce electricity
for sale.

2The 48 ton emissions level reflects reductions in mercury emissions
achieved by existing controls for other pollutants. In this report, we use
the amount of mercury in coal that is burned by power plants (75 tons) as
a baseline when discussing the effectiveness of mercury controls.

3The technology-based approach is commonly known as the Maximum Achievable
Control Technology (MACT) approach.

4For information about EPA's economic analysis of the mercury control
options, see our related report, GAO, Clean Air Act: Observations on EPA's
Cost-Benefit Analysis of Its Mercury Control Options, GAO-05-252
(Washington, D.C.: Feb. 28, 2005).

In March 2005, EPA revised its finding that it was appropriate and
necessary to regulate mercury emissions from power plants under the
technology-based approach and issued a final rule based on the cap and
trade option that established a mercury cap of 38 tons for 2010 and a
second phase cap of 15 tons for 2018.5 Although power plants were not
previously required to control mercury emissions, some already captured
mercury as a side benefit of using controls designed to reduce other
pollutants such as sulfur dioxide. In developing the rule, EPA determined
that technologies specifically intended to capture mercury were not
adequately demonstrated and therefore were not "commercially available."
As a result, the agency decided that it could not reasonably impose
requirements to use these technologies in the near-term and set emissions
targets for 2010 based on the level of mercury control it expects to
result as a side benefit of another rule it issued in March 2005-the Clean
Air Interstate Rule (the interstate rule)-that calls for further
reductions in emissions of nitrogen oxides and sulfur dioxide.

Controlling mercury from power plants poses unique challenges because it
is emitted in low concentrations, making removal difficult, and in several
different forms, some of which are harder to capture than others. In
addition, the relative ease of removal varies from plant to plant
depending upon such site-specific factors as the type of coal burned.6 EPA
and the Department of Energy (DOE) coordinate research and development of
mercury controls, with EPA conducting small-scale research on new
technologies, while DOE partners with the power industry and other
stakeholders to conduct field tests of mercury control technologies at
power plants.

The DOE field tests have focused on (1) mercury controls known as sorbent
injection technologies, in which powdered substances (known as sorbents)
that bind to mercury are injected into a plant's exhaust; (2)

5EPA has estimated that power plants will achieve emissions reductions
beyond the 38 ton cap in 2010 and then use the resulting emissions
allowances to comply with the more stringent cap for 2018, resulting in
actual mercury emissions of about 31 tons in 2010 and about 26 tons in
2018. Relative to the estimated 75 tons of mercury in coal, this equals a
59 percent reduction in 2010 and a 65 percent reduction in 2018.

6The main types of coal burned, in decreasing order of rank, are
bituminous, subbituminous, and lignite. Rank is the coal classification
system based on factors such as the heating value of the coal. High-rank
coal generally has relatively high heating values (i.e., heat per unit of
mass when burned) compared with low rank coals, which have relatively low
heating values.

enhancements to existing controls for other pollutants to increase mercury
removal; (3) multipollutant controls, which simultaneously capture mercury
and other pollutants; and (4) oxidation technologies, which convert
mercury to a chemical form that is easier to remove. As of February 2005,
13 of DOE's field tests were completed and 26 were planned or not yet
completed.

In this context, you asked us to (1) describe information on the use,
availability, and effectiveness of technologies to reduce mercury
emissions at power plants; and (2) identify the factors that influence the
cost of these technologies and report on available cost estimates. To
respond to these objectives, we reviewed data about technologies
specifically designed to reduce mercury, including modifications to
pollution controls already in use that would target and improve mercury
capture.7 We included test data on mercury controls used in field-scale
tests but did not include test data on controls that were at earlier
stages of development. We surveyed 59 key stakeholders-including mercury
control vendors, representatives of the coal-fired power industry,
technology researchers, and government officials-and received 40
responses. In addition, we reviewed technical documents addressing the
performance of mercury controls and discussed technology research and
development with 14 key stakeholders who view mercury reduction from a
policy perspective. We did not independently test mercury control
technologies. Finally, we interviewed vendors and researchers of mercury
emissions monitoring technology to obtain and analyze information on the
availability and reliability of mercurymonitoring devices; this
information is presented in appendix II. (See app. I for a more detailed
description of the scope and methodology of our review.) We performed our
work between May 2004 and May 2005 in accordance with generally accepted
government auditing standards.

Results in Brief	Mercury controls have not been permanently installed at
power plants because, prior to the March 2005 mercury rule, federal law
had not required this industry to control mercury emissions; however, some
technologies are available for purchase and have shown promising results
in field tests. Overall, tests of varying duration of the most developed
mercury control, sorbent injection, have achieved average mercury
reductions of 30 to 95 percent, with results depending on the rank of coal
burned and other

7We did not assess the effectiveness of controls for other pollutants in
capturing mercury as a side benefit because EPA had already conducted an
extensive analysis of that topic.

factors, according to DOE and other stakeholders we surveyed. More recent
DOE-funded monthlong tests, particularly those for chemically enhanced
sorbents, have shown average removal rates of over 90 percent. However,
data on the long-term performance of mercury controls or the effect that
they have on the overall reliability and efficiency of power plants are
limited, especially for plants using low-rank coals, because most field
tests have lasted less than 3 months. Ongoing tests may better inform
stakeholders within the next year about the longer-term capabilities of
mercury controls for these coals.

The cost to install and operate mercury controls depends on a number of
factors, including the extent to which controls already in place to reduce
other pollutants also reduce mercury emissions. As a result, cost
estimates vary widely. Available EPA and DOE cost estimates for mercury
controls have focused primarily on sorbent injection and were based on
modeling and data from a limited number of field tests, making them
preliminary and uncertain. Nonetheless, DOE estimated that using sorbent
injection to achieve a 70 percent reduction in mercury emissions would
cost a mediumsized power plant-one that has the capacity to generate 500
megawatts of electricity and operates for about 80 percent of the time
over the course of a year-$984,000 in capital costs and $3.4 million in
annual operating and maintenance costs. If this same plant were to install
a supplemental fabric filter-an option a plant might pursue to increase
the efficiency of mercury removal and reduce related costs-capital costs
would increase to about $28.3 million, while annual operating and
maintenance costs would decrease to about $2.6 million. Regardless of the
exact magnitude of costs, most stakeholders we contacted generally expect
mercury control technologies to cost less over time as a market develops
for the controls and as plants gain more experience using them.
Furthermore, EPA officials said that recent tests of chemically enhanced
sorbents lead the agency to believe that its earlier cost estimates likely
overstated the actual costs power plants would incur.

We provided a draft of this report to DOE and EPA for review and comment.
DOE said that it generally agreed with our findings. EPA's Office of Air
and Radiation and Office of Research and Development provided technical
comments, which we incorporated as appropriate.

Background	Mercury enters the environment through natural and man-made
sources, including volcanoes, chemical manufacturing, and coal combustion,
and poses ecological threats when it enters water bodies, where small
aquatic

organisms convert it into its highly toxic form-methylmercury. This form
of mercury may then migrate up the food chain as predator species consume
the smaller organisms. Through a process known as bioaccumulation,
predator species may consume and store more mercury than they can
metabolize or excrete.

Fish contaminated with methylmercury may pose health threats to people
that rely on fish as part of their diet. Mercury harms fetuses and can
cause neurological disorders in children, including poor performance on
behavioral tests, such as those measuring attention, motor and language
skills, and visual-spatial abilities (such as drawing). The Food and Drug
Administration (FDA) and EPA recommend that expectant or nursing mothers
and young children avoid eating swordfish, king mackerel, shark, and
tilefish and limit consumption of other potentially contaminated fish.
These agencies also recommend checking local advisories about
recreationally caught freshwater and saltwater fish. According to EPA, 45
states issued mercury advisories in 2003 (the most recent data available).

According to the United Nations Environment Program, global mercury
emissions are uncertain but fall within an estimated range of 4,850 to
8,267 tons per year. Of this total, EPA estimates that man-made sources in
the United States emit about 115 tons per year, with about 48 tons emitted
by power plants. Because mercury can circulate for long periods of time
and be transported thousands of miles before it gets deposited, it is
difficult to link mercury accumulation in the food chain with individual
emission sources.

The United States has 491 power plants that rely in whole or in part on
coal for electricity generation, and these plants produced 52 percent of
all electricity generated in 2004, according to DOE's most recent data.
These plants generally operate by burning coal in a boiler to convert
water into steam, which in turn drives turbines that generate electricity.
Figure 1 provides a general overview of a power plant's layout.

Coal combustion releases other harmful air pollutants in addition to
mercury, including sulfur dioxide and nitrogen oxides.8 EPA has regulated
these pollutants since 1995 and 1996, respectively, through its program
intended to control acid rain. In addition, the March 2005 interstate rule
will require further cuts in these pollutants beginning in 2009.9 To
comply with these and other regulations, the coal-fired power industry has
installed a variety of technologies that, while intended to control
nitrogen oxides, particulate matter, or sulfur dioxide, may also affect or
enhance mercury capture. Examples of such technologies include selective
catalytic reduction (SCR) for nitrogen oxides, electrostatic precipitators
(used by about 80 percent of all facilities) and fabric filters (used by
the remaining 20 percent) to control particulate matter and wet or dry
scrubbers to remove sulfur dioxide.

EPA estimates that power plants capture about 27 tons of mercury each
year, primarily through the use of controls for other pollutants. In
general, the exhaust from coal combustion (called flue gas) exits the
boiler and may flow through a device intended to control nitrogen oxides
before entering the particle control device and then through a scrubber
prior to release from the smokestack. The combination of these devices in
use at power plants differs greatly among facilities and is likely to
change as a result of the interstate rule, which, according to EPA, will
result in additional installations of equipment to control nitrogen oxides
and sulfur dioxide. EPA believes that the steps power plants will take to
control nitrogen oxides and sulfur dioxide under the interstate rule will
enable them to meet the first phase mercury cap of 38 tons beginning in
2010.10 As noted above, EPA determined that mercury control technologies
were not commercially available and that the agency could not reasonably
impose requirements to use them in the near-term.

8Nitrogen oxides and sulfur dioxide contribute to acid rain and the
formation of fine particles that have been linked to aggravated asthma,
chronic bronchitis, and premature death. Nitrogen oxides also contribute
to the formation of ozone, a regulated pollutant, when they react with
volatile organic compounds in the presence of heat and sunlight.

9The interstate rule requires further reductions in nitrogen oxide and
sulfur dioxide emissions in 2009 and 2010, respectively.

10According to EPA, a large share of the mercury captured under the two
rules will be its forms that are of greatest concern with respect to
deposition in the United States and eventual uptake by freshwater aquatic
organisms.

Nonetheless, a number of mercury control technologies have been developed
over the past several years as a result of public and private investments
in research and development, and these technologies generally fall into
the following categories:

o 	Sorbent (carbon-based, chemically enhanced carbon-based, and non-carbon
based). This technology involves injecting a powdered substance (sorbent)
into the flue gas that binds to mercury prior to collection in a particle
control device. Regardless of the chemical composition of the sorbent,
this technology involves adding a silo or other structure containing the
sorbent and a system that injects the sorbent into ducts that carry the
flue gas.

o 	Enhancements to existing controls for other pollutants to increase
mercury capture. This class of technologies focuses on retrofitting
existing controls for other pollutants to improve their ability to capture
mercury. Examples of enhancements include adding sorbents to wet scrubbers
used for sulfur dioxide removal or modifying selective catalytic reduction
devices used to reduce nitrogen oxides.

o 	Multipollutant controls. This class of technologies is designed from
the outset to simultaneously control or enhance the removal of multiple
pollutants, such as mercury, nitrogen oxides, or sulfur dioxide.11 These
technologies may also use sorbents.

o 	Oxidation technologies. This class includes methods, chemicals, or
equipment designed to oxidize mercury into a form that is more readily
captured.

o 	Other technologies. This category includes other technologies that
capture mercury using approaches such as removing mercury from coal prior
to combustion and fixed adsorption devices that rely on precious metals
such as gold to separate mercury from flue gas.

The intended location of these technologies in a power plant's overall
layout may vary. As shown in figure 2, some may be located between the
boiler and the particulate matter collection device, while others may be
located further downstream in a plant's process. This figure also shows
that

11Multipollutant controls do not include those that are intended to
capture other pollutants that may also remove some mercury.

some plants can either install sorbent injection upstream of the existing
particulate matter removal device or downstream of the device using a
supplemental filter to collect the spent sorbent, keeping it separate from
the fly ash collected in the particulate matter collection device. The
latter configuration may be relevant for those facilities that sell their
fly ash as a raw material for use in other applications, such as cement
manufacturing, because carbon-based sorbent can render fly ash unsuitable
for some of these applications. According to EPA, power plants sell about
35 percent of their fly ash for use in other applications, with 15 percent
going to uses, such as cement manufacturing, where carbon contamination
could pose a problem.

Figure 2: Sample Layout of Mercury Controls at a Coal-Fired Power Plant

The Department of Energy's (DOE) National Energy Technology Laboratory
partners with the private sector to evaluate the use of mercury control
technologies at power plants in tests lasting up to 5 months. The testing
program focuses on mercury controls, such as sorbent injection, and ways
to better and more consistently capture mercury with

technologies for other pollutants. Participants in DOE's program evaluate
concepts in laboratories and develop promising technologies in
progressively larger-scale applications, including actual power plants.12
The duration of the tests that have been completed has varied from several
hours to 5 months, with most of the completed DOE-funded tests lasting
between 1 week and several months.13 The most recent phase of DOE testing
has focused on the longer-term performance of mercury control
technologies. Appendix III provides more information on the DOE tests
completed, ongoing, or planned as of February 2005.

Mercury Controls Have Not Been Permanently Installed at Power Plants but
Are Available for Purchase and Have Shown Promising Results in Field Tests

Power plants in the United States do not currently use mercury controls,
but some technologies are available for purchase and have shown promising
results in full-scale tests in power plants. These tests have shown that
mercury controls known as sorbent technologies-which involve injection of
a powdered material that binds to mercury in the plant's exhaust-have
shown the greatest effectiveness in removing mercury during tests at power
plants. However, long-term test data are limited because most of these
tests have lasted less than 3 months.

Mercury Controls Are Not Currently Used by Power Plants, but Some
Technologies Are Available for Purchase

According to all 40 survey respondents, coal-fired power plants were not,
as of November 2004, using mercury controls, although several plants have
subsequently announced plans to install them. The coal-fired power
industry has not used mercury controls because, prior to EPA's March 2005
rule, federal law had not required mercury emissions reductions at power

12As stated in appendix I, we focused our data collection on tests at
actual power plants. The tests at power plants were conducted on varying
scales, with some controls applied to a diverted fraction of the flue gas
and other controls-primarily the sorbents-applied to the entire stream of
flue gas, e.g., full-scale tests.

13The longest continuously operating test lasted for 5 months as part of a
yearlong project at a plant in Wilsonville, Alabama.

plants.14 In fact, most of the power industry survey respondents (13 of
14) cited uncertainty about future regulations as one of the top three
reasons for not installing mercury controls. Thus, in the absence of
federal requirements to reduce mercury emissions, limited demand existed
for mercury controls.

We found that although some mercury controls, such as activated carbon
injection, are currently available for purchase from vendors, perceptions
about their availability vary widely among stakeholders, primarily because
stakeholders do not consistently define "availability." That is, some
stakeholders believe that mercury controls become available when they have
been demonstrated in long-term tests under normal commercial operations,
rather than when they are available for purchase. Thus, some stakeholders'
views on availability reflect more of a judgment about the proven
effectiveness of a control technology than their availability for
purchase.15 In this context, we found that views regarding the
availability of mercury controls generally varied by stakeholder group and
by the type of control. A greater portion of the vendors described mercury
controls as available than either of the other two groups we surveyed,
with the power industry group citing these controls as available least
frequently. As shown in figure 3, the stakeholders were overall most
optimistic about the availability of activated carbon injection
technologies, followed by multipollutant controls and enhancements to
existing controls for other pollutants.

14Thirteen of the 14 power industry respondents also identified inadequate
performance guarantees and the belief that technologies are unproven as
reasons for not installing mercury controls.

15In our survey, we asked respondents separate questions about mercury
controls addressing their availability for purchase, their effectiveness,
and the need for further testing.

    Figure 3: Stakeholder Perceptions about Availability of Mercury Controls

Mercury control

                                    Sorbent: Activated carbon injection (ACI)
                                                              Multipollutants
                                                                 Enhancements
                                             Sorbent: Chemically-enhanced ACI
                                                            Mercury oxidation
                                                Sorbent: Non-carbon injection

                                      6 10

26 24 23

                                       31

0 5 1015202530

Number of stakeholders viewing mercury control as available

Source: GAO analysis of survey responses.

Note: This figure is based on responses from the stakeholders that
participated in either our surveys (40) or structured interviews (14). In
asking survey respondents and interview participants about their views on
the availability of all mercury controls, we categorized sorbent injection
technologies as activated carbon, chemically enhanced carbon, and
non-carbon injection in order to reflect the research and development of
various sorbent materials.

Appendix IV provides more detailed information on stakeholder perceptions
of the availability of mercury controls.

In evaluating the availability of mercury controls prior to finalizing the
March 2005 mercury rule, EPA found that mercury controls were available
for purchase but concluded that they had not been sufficiently
demonstrated in long-term tests, and therefore were not available for
permanent installation at power plants before 2010. As a result, EPA set
the 2010 mercury reduction targets at a level that power plants could
achieve as a side benefit of using technologies for other pollutants that
the agency expects many plants will install to comply with the interstate
rule, and set more stringent limits for 2018. Thus, power plants will not
need to install mercury-specific controls until well after 2010. According
to an EPA white paper assessing test results as of February 2005, the
agency expects that mercury control technologies will be available for
commercial application on most, if not all, key combinations of coal type
and control technology to

provide mercury removal levels between 60 and 90 percent after 2010 and
between 90 and 95 percent in the 2010-2015 time frame.16

Some Mercury Controls Have Shown Promising Results in Short-Term Field
Tests, but Data on Long-Term Performance Are Limited

Because mercury controls have not been permanently installed at power
plants, the data on the performance of these technologies come from field
tests. We obtained data from 29 completed field tests, including 13 which
were part of DOE's mercury control research and development program, and
16 other tests identified by survey respondents.17 Most of the available
test data (21 of 29 tests) related to the effectiveness of sorbents.
According to DOE and EPA, the tests have shown promising results, although
the extent of mercury removal varies at each plant.

Tests of varying duration have identified sorbent technologies as the most
developed mercury controls, which show promising results in achieving high
mercury reductions. For example, tests of activated carbon and chemically
enhanced carbon-based sorbents at power plants using a variety of air
pollution controls have shown average reductions of 30 to 95 percent
overall, providing the following average mercury reductions for each coal
type:18

o  70-95 percent average removal on bituminous coals;

o  30-90 percent average removal on subbituminous coals;

o  63-70 percent average removal on lignite coals; and19

o  94 percent removal on blends of bituminous/subbituminous coals.

16EPA, Office of Research and Development, Control of Mercury Emissions
from Coal Fired Electric Utility Boilers: An Update (Research Triangle
Park, N.C., Feb. 18, 2005).

17We obtained data about 55 field tests, 39 of which are part of DOE's
mercury control research and development program. As of February 2005,
long-term testing was either planned or had not been completed at 26 of
the 39 DOE-funded field tests. Sixteen of the 55 field tests we reviewed
were identified by survey respondents and did not correspond to DOE-funded
tests.

18These data consider the amount of mercury in coal-75 tons-as the
baseline for estimating the percent mercury reduction.

19One test on lignite coal also used a sorbent enhancement, i.e.
additional chemicals to improve mercury capture.

As the scale and duration of testing has increased, researchers have
gained a better understanding of site-specific variables that affect
results, and more recent full-scale, monthlong tests, particularly those
using chemically enhanced carbon-based sorbents, have shown sustained high
removal rates. For example, a monthlong test conducted in 2004 showed that
a chemically enhanced sorbent reduced mercury emissions from a primarily
subbituminous blend of coal by 94 percent, and a monthlong test of another
chemically enhanced sorbent at a different plant burning subbituminous
coal achieved a 93 percent reduction.

A number of the stakeholders we surveyed pointed out that the results of a
particular test cannot be generalized or extrapolated to estimate
potential reductions at other power plants because the reductions achieved
during a test may have resulted in part from factors unique to that
facility, such as its size, the type of boiler used, the temperature of
its flue gas, or the combination of controls for other pollutants. For
example, available data show that the extent of mercury reduction achieved
by sorbent injection at facilities using electrostatic precipitators
depends largely on the location of these devices at the plant. The
location of an electrostatic precipitator in turn affects the temperatures
of the flue gas entering the device, with more mercury captured at cooler
temperatures. Thus, the results achieved at a particular plant may not
serve as a reliable indicator of the performance of that control at all
plants.

DOE's research and development program has funded tests of mercury
controls on each coal type in light of its and EPA's conclusions that the
form of mercury emitted-which varies by coal type-and other chemical
variations among coal types, such as chlorine content, can have an impact
on a control's removal effectiveness. For example, lower removal rates in
activated carbon injection tests have occurred primarily at plants burning
low rank coal or at plants with existing controls that are less conducive
to mercury removal. One university-based researcher attributes the
challenge of mercury reductions on lignite-a low rank coal-to its chemical
composition, but believes that chemically enhanced sorbents and special
additives can improve the ability of the sorbent to bind to this form of
mercury, thereby addressing this problem. The more recent mercury removal
results we reviewed tended to support this view as monthlong tests using
chemically enhanced carbon-based sorbents achieved average reductions of
70 percent or greater on low-rank coals, including lignites, suggesting
that this technology may achieve high-level mercury reductions from
low-rank coals (See app. III for more information on these results).

Since most of the field tests have focused on sorbent injection, fewer
data are available on the performance of non-sorbent mercury controls,
such as multipollutant controls, enhancements to existing controls, and
mercury oxidation technologies. Results from 11 of the 19 tests of such
controls were not yet available (9 of the tests were not planned to begin
until after February 2005). The few available results show that average
mercury removal achieved by multipollutant controls and enhancements has
ranged from about 50 percent to 90 percent. The field tests of mercury
oxidation technologies, multipollutant controls, enhancements and other
nonsorbent technologies, lasting several days to several months, have
included all coal types, but most (7 of 10) to date have focused on
bituminous coal. In addition, a future DOE project will fund a test of a
multipollutant control on a plant burning subbituminous coal and three
tests of mercury controls, including mercury oxidation and enhancements,
on plants burning lignite coal.20

Stakeholders Generally Agree That Sorbent Injection Is the Most Promising
Control and That Some Additional Tests Are Needed

As noted above, EPA determined as part of its March 2005 mercury rule that
it could not reasonably impose requirements that would force the use of
mercury-specific controls before 2010. Specifically, EPA believes that
chemically enhanced carbon-based sorbents could reduce mercury emissions
at a broad spectrum of plants but regards long-term testing as necessary
in order to evaluate (1) the mercury removal performance of technologies
when operated continuously for more than several months at a time; and (2)
the impact that these controls have on a plant's overall efficiency and
operations. Furthermore, DOE officials have said that while sorbent
injection holds much promise, it is unwise to depend solely on one
approach for mercury control in part because the site-specific variables
at each power plant affects the performance of mercury controls. DOE has
concluded that it will be necessary to build a broad portfolio of mercury
control options.

Likewise, technical papers and presentations about the field tests by
research and development participants express a high degree of confidence
in the capability of sorbents, particularly chemically enhanced
carbonbased sorbents, but also suggest the need for additional evaluation
of the impact of these controls, if any, on the efficiency and reliability
of power plants. For example, a paper written by a sorbent vendor
conducting DOE

20DOE has required most projects in this round of testing to last at least
for 1 month. The exact duration of these tests has not yet been
determined.

funded tests concluded that recent monthlong tests of chemically enhanced
carbon-based sorbent injection have shown high mercury removal at plants
that burn subbituminous coals, but also discussed concerns about the
long-term use of this control on a power plant's operations. This vendor
concluded that although these tests did not show any adverse effects
resulting from the chemically enhanced carbon-based sorbent, concerns and
issues surrounding the contamination of fly ash that can render it
unsuitable for sale for certain applications have not yet been resolved.
With regard to potential adverse impacts at plants, no serious adverse
effects have been associated with sorbent injection tests lasting up to 1
month in duration, according to EPA.

To provide additional perspective on the expected long-term performance of
mercury controls, we asked survey respondents to indicate whether they
believed power plants could use mercury controls to achieve industrywide
mercury reductions of 50, 70, or 90 percent by 2008.21 We also asked the
respondents whether their perceptions would differ if the reductions were
averaged across the industry (as in an emissions trading program) or if
they were required at each plant. We found that many survey respondents
(22 of the 38 answering this question) were confident in the ability of
power plants to achieve a 50 percent reduction by 2008 regardless of
whether the reductions were achieved at each plant or averaged across the
industry.22 EPA set the mercury emissions cap for 2010 based on a 50
percent reduction from the 75 tons in coal.

The stakeholders were progressively less confident in the ability of
plants to achieve 70 and 90 percent reductions by 2008. For the 70 percent
reduction scenario, stakeholders were more confident in the ability of
plants to achieve this reduction averaged across the industry rather than
at each plant; 16 stakeholders described themselves as confident or very
confident in the ability of plants to achieve this level of reduction
nationwide, while 21 described themselves as less confident or not at all
confident. For the 90 percent scenario, the vast majority of the survey
respondents (33 of 38 that answered this question) described themselves as
not at all confident or less confident in the ability of plants to achieve
this

21We asked respondents to consider the amount of mercury in coal-75
tons-as the baseline when considering each mercury reduction.

22This would result in nationwide emissions of 37.5 tons per year, given
the baseline of 75 tons of mercury in coal.

level of reduction nationwide by 2008. Appendix V summarizes the survey
responses for each of the three scenarios.

Furthermore, we asked the 40 survey respondents to identify additional
testing needed to assess the ability of mercury control technologies to
effectively and reliably reduce mercury emissions by 70 percent. Most of
the survey responses (40 of 45)23 showed that stakeholders believe that
some additional testing is needed for at least one technology. For
example, the 14 power industry respondents said that additional testing is
needed for sorbent injection. In addition, 3 of the 4 carbon-based sorbent
vendors answering this question as well as 9 of the 12 researchers and
government officials believed that some additional testing is needed to
show that carbon-based sorbent injection would reliably and effectively
achieve mercury reductions of 70 percent.

Three policy stakeholders representing the power industry believed that
more tests are needed to evaluate factors such as the performance of
controls on low-rank coals, the impact on small power plants, and the
ability of plants to use mercury controls without compromising electricity
generation. Several of the power industry respondents expressed concern
about the potential for mercury controls to interfere with a plant's
overall efficiency or cause malfunctions, and a power industry
representative pointed out that such disruptions are a concern because
power plants cannot store electricity for use as a backup when they
experience technical problems. Information from ongoing and planned
long-term tests will provide important information on both the long-term
performance of mercury controls and the effect, if any, that these
controls have on the efficiency or reliability of power plants.

In addition, several plants have recently announced plans to install
mercury controls to comply with either state permit requirements or the
terms of legal settlements. For example, a power plant in New Mexico
announced in March 2005 that it would install sorbent injection within the
next 2 years to reduce mercury emissions as part of a settlement agreement
with two environmental groups. A plant representative stated that while he
believes sorbent technology "is not that advanced ... it is advanced
enough to use it to reduce mercury emissions" at the power plant. Another
power

23The number of survey responses exceeds the number of survey participants
because technology vendors were given the option of submitting a survey
for each technology they produce. Five of the 14 technology vendors
submitted two surveys.

plant currently under construction in Iowa has a state air pollution
permit requiring the company to control mercury emissions and is
installing sorbent injection technology. The company expects to reduce
mercury emissions from subbituminous coal by 83 percent. Finally, under an
agreement with the state of Wisconsin, a Michigan power plant owned by a
Wisconsin-based company has begun to install a multipollutant control that
will use sorbent injection to reduce mercury and other pollutants.

Mercury Control Costs Depend on a Variety of Factors, and Current
Estimates Vary Widely

The estimated costs to install and operate mercury controls vary greatly
and depend on a number of site-specific factors, including the amount of
sorbent used (if any), the ability of existing air pollution controls to
remove mercury, and the type of coal burned. EPA and DOE have developed
the most comprehensive estimates available for mercury controls based on
modeling and data from a limited number of field tests, making them both
preliminary and uncertain.24 These estimates, as well as other available
estimates, focus on sorbent injection, the most developed mercury control
technology. Estimated costs for sorbent injection vary greatly depending
on whether facilities achieve mercury reduction targets by using this
technology in combination with their existing air pollution control
devices or instead add fabric filters to collect the spent sorbent.
Regardless of the exact costs of the controls, most of the stakeholders we
contacted generally expect the costs to decrease over time.

Cost Estimates Depend on Several Site-Specific Factors

The available cost estimates are projections based on a limited number of
tests, primarily of activated carbon injection. The cost estimates we
reviewed show that the total costs of installing and operating mercury
controls vary depending on factors such as sorbent consumption, the
ability of existing air pollution controls to remove mercury, and the type
of coal burned. We discuss each of these factors in more detail below:

24Environmental Protection Agency, Office of Research and Development,
National Risk Management Research Laboratory, Performance and Cost of
Mercury and Multipollutant Emission Control Technology Applications on
Electric Utility Boilers (Research Triangle Park, N.C., 2003).

Jeff Hoffmann and Jay Ratafia-Brown, Science Applications International
Corporation,

Preliminary Cost Estimate of Activated Carbon Injection for Controlling
Mercury Emissions from an Un-Scrubbed 500 MW Coal-Fired Power Plant, a
report prepared for the Department of Energy, National Energy Technology
Laboratory, November 2003.

o 	Sorbent consumption: The amount of sorbent that a facility needs to use
greatly influences control cost estimates. According to DOE, sorbent
consumption levels for activated carbon injection technology directly
relate to the desired level of mercury control. Further, while increasing
the amount of carbon injected increases mercury removal, the performance
of the carbon eventually levels off, requiring increasingly greater
amounts of carbon to achieve an incremental mercury reduction. For
example, test data from a plant burning subbituminous coal show that more
than twice as much sorbent would be needed to remove 60 percent of the
mercury from the plant's flue gas than to remove 50 percent. Therefore,
the cost of the activated carbon can increase dramatically, depending on
the desired level of mercury removal and the type of coal burned.

o 	Other air pollution controls already installed: The air pollution
controls already installed at a facility-especially fabric filters and
electrostatic precipitators used for controlling particulate matter-can
have a major effect on the cost of controlling mercury because some of
these devices already remove varying amounts of mercury. For example,
DOE's tests have shown that fabric filters generally remove more mercury
than electrostatic precipitators. Thus, facilities with fabric filters may
already remove enough mercury to achieve a desired or required level of
reduction. However, plants that do not have an existing fabric filter and
choose to install one may incur significant costs due to their high
capital expense. Additionally, EPA believes that controls for other
pollutants some plants will install to comply with the interstate
rule-such as selective catalytic reduction to control nitrogen oxides and
wet scrubbers to control sulfur dioxide-will result in further mercury
capture. Therefore, the combination of other air pollution controls may
reduce or in some cases eliminate the need for a plant to install
mercury-specific controls to reduce its mercury emissions. As noted above,
EPA based its mercury reduction goals for 2010 to 2018 on the level of
control it expects plants will achieve with controls for these other
pollutants.

o 	Type of coal burned: According to EPA, the amount of mercury captured
by a given control technology is generally higher for plants burning
bituminous coals than for those burning subbituminous coals. This
difference arises because the flue gas from bituminous coal contains
higher levels of substances that facilitate mercury capture. Along these
lines, DOE's cost estimates assume that an electrostatic precipitator will
capture 36 percent of mercury from plants that burn

bituminous coal, but none of the mercury from plants that burn
subbituminous coal. Thus, DOE estimated that mercury removal costs are
higher for subbituminous-fired plants than bituminous-fired plants.

Available Mercury Control Cost Estimates Are Preliminary and Vary Greatly

Most of the available cost estimates for mercury control focus on sorbent
injection, the most developed technology. DOE and EPA have developed
comprehensive cost estimates; however, they are preliminary and, in EPA's
case, based on model plants rather than actual power plants. Further,
while DOE developed its estimates from tests in power plants, the agency
indicated that its mercury control costs may be off by as much as 30
percent in either direction because (1) the estimates were developed from
a limited data set of relatively short-term tests and thus are highly
uncertain, and (2) they are based on a number of assumptions that, if
changed, would result in significantly different estimates. According to
DOE, further testing of sorbent injection for a variety of coals is needed
to accurately assess the costs of implementing the technology throughout
the United States. In addition, EPA's and DOE's cost estimates were
published in October and November 2003, respectively, and do not reflect
the more recent test data. For example, more recent field tests with
chemically enhanced sorbents have shown that these sorbents may be more
efficient at removing mercury than the sorbents used in earlier tests.
Thus, chemically enhanced sorbents may achieve a high level of mercury
removal using less sorbent and without the high capital cost of installing
a fabric filter. DOE expects to issue revised cost estimates which will
reflect lower costs based on recent testing. As a result, the available
cost estimates may not accurately reflect the costs that power plants
would incur if they chose to install mercury controls.

In addition, the two agencies' cost estimates relied on different
assumptions and are not directly comparable. Most notably, the two
agencies based their cost estimates on plants of different size and made
varying assumptions about the percentage of time that an average plant
operates (called capacity factor). For example, EPA conducted its modeling
for 100- and 975-megawatt plants, while DOE based its estimates

on a 500-megawatt plant.25 As a result, EPA provided a wider range of cost
estimates. Furthermore, EPA assumed a plant capacity factor of 65 percent,
while DOE assumed an 80 percent capacity factor, which resulted in higher
operating costs in the DOE estimates.26 Additionally, based on available
data for plants with an existing electrostatic precipitator that burn
bituminous coal, EPA's modeling predicted the existing control equipment
would achieve a 50 percent mercury removal without sorbent injection,
while DOE assumed that this configuration would remove no more than 36
percent of mercury and that sorbent injection was needed even for
achieving 50 percent mercury removal.27

Although the DOE and EPA estimates reflect different assumptions as
discussed above, we are providing the two agencies' cost estimates for
achieving a 70 percent mercury reduction at a bituminous-fired coal power
plant under two scenarios (using an existing electrostatic precipitator
and installing a supplemental fabric filter) to provide a perspective on
the costs power plants could incur to install sorbent injection
technologies.

o 	For a 100-megawatt plant using an existing electrostatic precipitator,
EPA estimated that capital costs would total $527,100 ($5.27 per kilowatt,
2003 dollars), and the operating and maintenance costs would total
$531,820 annually for a plant operating at 65 percent capacity ($0.93 per
megawatt-hour).28 Alternatively, if this plant were to install a
supplemental fabric filter, the capital costs would increase to about $5.8
million ($57.73 per kilowatt) and the operating and maintenance costs
would decrease to $171,959 annually ($0.30 per megawatt-hour).

o 	For a 500-megawatt plant using an existing electrostatic precipitator,
DOE estimated the capital costs would total $984,000 ($1.97 per kilowatt),
and the annual operating and maintenance costs would total

25A megawatt is a unit of power equal to one million watts, or enough
electricity to power about 750 homes at any given time.

26According to a DOE official, the varying assumptions regarding the plant
capacity factor reflect different assumptions about which coal-fired power
plants will use sorbent technologies.

27According to EPA, while 36 percent is an average removal rate for
bituminous coals, the 50 percent rate they used in this case was based on
specific assumptions about a particular type of bituminous coal in the
scenario they analyzed.

28Costs expressed in dollars per megawatt-hour and mills per kilowatt-hour
are numerically equal.

about $3.4 million ($0.97 per megawatt-hour) for a plant operating at 80
percent capacity (2003 dollars). Alternatively, if this plant were to
install a supplemental fabric filter, the capital costs would increase to
about $28.3 million ($56.53 per kilowatt), and the operating and
maintenance costs would decrease to about $2.6 million annually ($0.74 per
megawatt-hour).

o 	For a 975-megawatt plant using an electrostatic precipitator, EPA
estimated that capital costs would total about $2.4 million ($2.47 per
kilowatt), and the operating and maintenance costs would be about $5.1
million annually for a plant operating at 65 percent capacity ($0.92 per
megawatt-hour). Alternatively, if this plant were to install a
supplemental fabric filter, the capital costs would increase to about
$35.4 million ($36.32 per kilowatt), and the operating and maintenance
costs would decrease to about $1.6 million annually ($0.30 per
megawatt-hour).

These data show that DOE estimated lower capital costs per unit of power
generating capacity than EPA, while EPA estimated slightly lower operating
and maintenance costs than DOE. This may result from the fact that EPA
assumed higher rates of mercury removal with existing controls than DOE,
as well as DOE's use of a higher plant capacity factor than EPA. Appendix
VI provides additional information on EPA's and DOE's cost estimates for
sorbent injection control technologies.

According to EPA, the costs of sorbent injection technologies to control
mercury emissions are very small compared to other air pollution control
equipment when other retrofits, such as the addition of fabric filters,
are not required. EPA also reports that the fixed operating costs for
these systems are also relatively low, stemming from the simplicity of the
equipment. In EPA's rulemaking documents, the agency said that in light of
the more recent tests of chemically enhanced sorbents, their earlier
estimates likely overstated the actual costs power plants would incur. DOE
officials said they shared this view.

EPA also estimated costs for multipollutant controls, including advanced
dry scrubbers. Although these controls cost substantially more than
sorbent injection, they would provide additional benefits by controlling
other types of pollutants such as nitrogen oxides and sulfur dioxide.29
EPA regarded cost information for multipollutant controls as preliminary,
because there had been limited commercial experience with these
technologies in the United States. In part because the agency estimated a
range of capital and operating costs for each scenario, EPA's estimates of
the cost of these technologies varied widely.30 For example, for advanced
dry scrubbers, EPA estimated the capital costs as $115.46 to $243.08 per
kilowatt, with costs per kilowatt generally higher for smaller plants.31
For 100-megawatt and 975-megwatt plants, capital costs could be as low as
$16.2 million and as high as $168.7 million respectively. EPA estimated
operating and maintenance costs for a 100-megawatt plant to be between
$1.1 million and $1.3 million per year, assuming a plant capacity factor
of 65 percent (or between $1.93 and $2.35 per megawatt-hour). For a
975megawatt plant, operating and maintenance costs were estimated to be
between $9.3 million and $37.5 million per year, assuming a plant capacity
factor of 65 percent (or between $1.68 to $6.76 per megawatt-hour).

In addition to the cost estimates from EPA and DOE, we surveyed technology
vendors, representatives of coal-fired power plants, and researchers about
the cost of these technologies. Seventeen of these stakeholders provided
sorbent injection cost information, but these estimates were incomplete
and not always comparable due to site-specific variations and differing
assumptions. The vendors generally provided lower cost estimates than
those provided by the power industry, while estimates provided by
researchers had the broadest range.

EPA and DOE officials and other stakeholders identified relevant cost
estimates compiled by other nongovernmental entities:

29When combined with existing equipment, advanced dry scrubbers were
estimated to achieve mercury removal rates between 96 and 99 percent in
EPA's models.

30In calculating these estimates, EPA assumed that the unit capital cost
could vary by as much as 20 percent, while operating and maintenance costs
were calculated assuming a range of reagent costs that varied by as much
as plus or minus $20 per ton. Due to these variations, cost ranges
presented in unit costs, such as dollars per kilowatt, do not always match
the calculated cost ranges in total dollars for a plant of a given size.

31In estimating costs for advanced dry scrubbers, EPA only presented costs
for plants burning bituminous coal.

o 	Charles River Associates, an economics and business consulting firm,
provided cost estimates for activated carbon sorbent injection in
combination with an existing or supplemental fabric filter.32 Rather than
presenting estimates of costs for particular plant sizes and mercury
removal percentages, Charles River Associates provided formulas with
variables for mercury removal and plant size.33 Using these formulas and a
plant size of 500 megawatts, Charles River Associates' analysis would
generate estimates of total capital costs of about $749,278 for using
sorbent injection with an existing fabric filter and about $20.6 million
for sorbent injection and a supplemental fabric filter (1999 dollars).
Operating and maintenance costs comprise a fixed cost based on plant size
and a variable component that could be calculated for a range of mercury
removal percentages. For example, a 90 percent mercury reduction using
sorbent injection with an existing fabric filter for a bituminous
coal-fired 500-megawatt plant operating at 80 percent capacity over the
course of a year (7,008 hours) would cost $999,473 per year, or about
$0.29 per megawatt-hour. A 90 percent reduction at the same size plant
burning subbituminous coal would cost $1.3 million per year or about $0.38
per megawatt-hour. Annual operating and maintenance costs were about
$75,000 higher for the configuration where a supplemental fabric filter
was installed.

In its modeling, Charles River Associates considered only sorbent
injection technology with an existing or retrofitted fabric filter because
the firm expects that this combination would have a lower cost per pound
of mercury removed than sorbent injection alone. Charles River Associates'
operating and maintenance cost estimates for activated carbon injection
alone are lower than the EPA and DOE estimates; however, the Charles River
estimates reflect the assumption that plants already had a fabric filter,
while EPA and DOE assumed plants already had an electrostatic
precipitator.

32Anne Smith et al., Charles River Associates , and John H. Wile, E&MC
Group, Projected Mercury Emissions and Costs of EPA's Proposed Rules for
Controlling Utility Sector Mercury Emissions (Washington, D.C., 2004).

33These formulas allow capital and fixed operating and maintenance costs
to vary by the size of the plant and allow variable operating and
maintenance costs to vary depending on the desired level of mercury
reduction.

o 	MJ Bradley & Associates, an engineering and environmental consulting
firm, summarized costs for other multipollutant controls that have
undergone full-scale testing.34 One technology, which uses ozone to
oxidize nitrogen oxide and mercury, has been estimated to remove over 90
percent of nitrogen oxide and mercury from a plant's flue gas; it also
controls sulfur dioxide.35 This technology is estimated to cost between
$90 and $120 per kilowatt in capital costs and $1.7 to $2.37 per
megawatt-hour in annual operating and maintenance costs. For a 500megawatt
plant operating at 80 percent capacity, this would equate to $45 million
to $60 million in capital costs and $6.0 million to $8.3 million in annual
operating and maintenance costs.36 MJ Bradley also estimated the costs of
a system that removes sulfur dioxide and mercury and decomposes nitrogen
oxide through a multi-stage oxidation, chemical, and filter process. The
target mercury removal rate for this process is 85 to 98 percent, which MJ
Bradley reports the manufacturer guarantees. The estimated capital cost of
this process is between $110 and $140 per kilowatt, or $55 million to $70
million for a 500-megawatt plant. A downstream fabric filter is associated
with this process to remove particulate matter, which could add an
additional cost.

In considering the cost estimates, it is important to note that plants may
identify and choose the most cost-effective option for complying with
EPA's mercury rule. The cost-effectiveness of a given mercury control will
vary by facility, depending on site-specific factors, including the type
and configuration of controls already installed. Furthermore, the desired
level of mercury control at a plant will affect its control costs and some
plants may meet their mercury reduction goals by modifying existing air
pollution control equipment, thereby negating the need for additional
mercury controls. In cases where plants decide to install mercury
controls, the desired control level will affect the cost-effectiveness of
the various technologies. For example, sorbent injection with a downstream
fabric filter may prove cost effective for facilities seeking a high level
of reduction, but less cost effective for plants seeking lower level
reductions

34MJ Bradley also presented cost estimates for sorbent injection, but
presented the same cost information reported by DOE.

35M.J. Bradley & Associates, Status of Development of Mercury Control
Technologies (Concord, Mass., Aug. 5, 2004).

36Calculated annual operating and maintenance costs assume a 500-megawatt
plant operating at 80 percent capacity, i.e. 7008 hours per year.

because of the relatively high capital costs. In the example given above
for a 70 percent mercury reduction at plants burning bituminous coal,
based on annualized costs, EPA's estimates suggest it is more
cost-effective for both the 100- and 975-megawatt plants to achieve that
reduction without installing a supplemental fabric filter; however, DOE's
estimates suggest it is more cost-effective for the 500-megawatt plant to
install the supplemental filter when accounting for the loss of revenue
and increased disposal costs plants could incur from not being able to
sell their fly ash.37

Fly ash disposal plays a role in determining the most cost effective
compliance option because the plants that sell their fly ash and choose to
use carbon-based sorbents may lose revenue and face increased disposal
costs if they can no longer sell their fly ash.38 According to EPA, power
plants sell about 35 percent of their fly ash for use in other
applications, with 15 percent going to uses, such as cement manufacturing,
where carbon contamination could pose a problem. The presence of
carbon-based sorbent in fly ash may render it unusable for such purposes,
particularly as a cement substitute in making concrete. Therefore, in some
cases, plants using carbon-based sorbent may not be able to sell their fly
ash and instead have to pay for its disposal. Plants may mitigate this
problem by installing sorbent injection downstream of the electrostatic
precipitator. This would, however, require the plants to install a fabric
filter to collect the spent sorbent. DOE estimated that this configuration
may be a cost-effective method to achieve mercury reductions for plants
that wish to continue selling their fly ash, but the high capital costs of
installing a fabric filter may

37EPA's estimates suggest that the installation of the fabric filter is
more cost-effective than carbon injection alone to achieve an 80 percent
mercury reduction at a 975-megawatt plant and a 90 percent mercury
reduction at both the 100-and 975-megawatt plants.

38DOE's estimates indicate that for a plant that sells its fly ash, loss
of fly ash sales and related disposal costs could increase the cost of
mercury removal by between $31,232 and $213,133 per pound of mercury
removed for a plant using activated carbon injection with an existing
electrostatic precipitator. Costs vary depending on the type of coal
burned and the desired level of mercury reduction. For example, the cost
per pound of mercury removed for a 50 percent mercury reduction at a
bituminous coal-fired plant increases from $32,598 to $245,731 when
accounting for the potential impact in lost fly ash sales. EPA estimated
that using current technology, the marginal cost of mercury control will
be $23,200; $30,100; and $39,000 per pound of mercury removed in 2010,
2015, and 2020 respectively (in 1999 dollars). EPA also conducted a
sensitivity analysis-assuming that mercury controls will improve over time
and therefore cost less-that showed this marginal cost falling to $11,800;
$15,300; and $19,900 respectively in 2010, 2015, and 2020. These mercury
removal analyses were conducted by EPA using the Integrated Planning
Model, and are therefore based on different assumptions and modeling
efforts than those that went into the 2003 mercury control cost report.

render this choice uneconomic for some facilities. However, based on more
recent tests, EPA believes that chemically enhanced sorbents can be more
efficient at achieving a high level of mercury removal and may not render
fly ash unusable for other purposes. Therefore, the use of these sorbents
might prevent a plant from having to install a fabric filter and allow
them to continue selling fly ash.

Most Stakeholders Expect the Costs to Decrease over Time

Regardless of the exact magnitude of costs, 22 of the 40 survey
respondents, all of the 14 policy stakeholders we interviewed, EPA, and
DOE expect mercury control costs to decrease over time. Stakeholders cited
a number of reasons for this belief, including the presence of a mercury
rule, the expected development of a market that would lead to competition
and increased demand for technologies, and anticipated improvements in
technology performance as a result of innovation and experience. According
to EPA and DOE officials, the most recent test results of injected sorbent
technologies suggest that the cost of using these technologies will be
less than these agencies estimated in 2003, stemming from advances in the
sorbents. Likewise, EPA's economic impact analysis of the mercury rule
reports that the actual cost of mercury control may be lower than
currently projected, since the rule may lead to further development and
innovation of these technologies, which would likely lower their cost over
time.

In addition to the views of these stakeholders, experience with pollution
control requirements under other air quality regulations also suggests
that costs may decrease over time. While factors affecting the cost of
mercury control technology may or may not be analogous to that of
technologies to control other regulated pollutants, an examination of the
cost trends for other air pollution controls shows that costs have
declined over time. For example, according to EPA, the acid rain sulfur
dioxide trading program was shown in recent estimates to cost as much as
83 percent less than originally projected.39 Furthermore, studies
conducted by other researchers demonstrate that costs of air pollution
control technologies have declined. For example, research conducted by
Carnegie Mellon University found that the capital cost of sulfur dioxide
control technology

39Part of the fall in acid rain costs is due to lower costs of
transportation, since the deregulation of rail made it cheaper to ship
low-sulfur coal greater distances.

for a coal-fired power plant decreased from approximately $250 to $130 per
kilowatt of electricity generating capacity between 1976 and 1995 (1997
dollars). Similarly, case studies analyzed by the Northeast States for
Coordinated Air Use Management (NESCAUM) found the total operating and
maintenance costs of sulfur dioxide controls decreased about 80 percent
between 1982 and 1997.40 NESCAUM also found a reduction in the capital
cost of nitrogen oxide controls, which it attributed to improvements in
operational efficiency.

Concluding Observations

Because data on the performance of mercury controls stem from a limited
number of tests rather than permanent installations at power plants, data
on the long-term performance of these technologies are limited.
Furthermore, while the available data show promising results, forecasting
when power plants could rely on these technologies to achieve significant
mercury reductions-such as by 2008 or later-involves professional
judgment. The judgment of the stakeholders we contacted varied
substantially, with control vendors and some researchers expressing
optimism about the potential for sorbent technologies to achieve
substantial mercury reductions in the near term, while power industry
stakeholders, DOE, and EPA highlighted the need for more long-term tests.
Current and future DOE tests will enhance knowledge about these controls,
especially on their effectiveness in removing mercury and the potential
impacts they may have on plant operations. In addition, information from
the power plants that plan to install mercury controls as part of
settlement agreements or to meet state-level requirements could shed
additional light on these issues.

A number of factors complicate efforts to estimate the costs of installing
mercury controls. For example, available data suggest that site-specific
variables will dictate the level of expense that power plant owners and
operators will incur should they install one of the available mercury
control technologies. While even the current cost estimates for the most
advanced of the technologies-sorbent injection-are highly uncertain for
individual plants, many of the stakeholders we contacted expect these
costs to decline. Further, past experience with other air pollution
control regulations suggests that the costs of pollution controls decline
over time

40Based on studies by the Electric Power Research Institute and the
Massachusetts Institute of Technology that showed operating and
maintenance costs decline from $17.3 per megawatt-hour to $3.34 per
megawatt-hour in 1999 dollars.

due to technological improvements, the development of a market, and
increased experience using the controls.

Recent data already show a similar trend with respect to mercury controls.
For example, EPA and DOE have stated that advanced sorbent technologies
have the potential to achieve greater mercury removal at lower cost than
previously estimated. Also, the emissions trading program established
under EPA's mercury rule gives industry flexibility in determining how it
will comply with the control targets, enabling plants to choose the most
cost-effective compliance option, such as installing controls, switching
fuels, or purchasing emissions allowances. Finally, because the power
industry must also further reduce its emissions of nitrogen oxide and
sulfur dioxide to comply with the interstate rule, the power industry has
the opportunity to cost-effectively address emissions of all three
pollutants simultaneously.

Agency Comments	We provided a draft of this report to DOE and EPA for
review and comment. DOE reviewed the report and said that it generally
agreed with our findings. EPA's Office of Air and Radiation and Office of
Research and Development provided technical comments, which we
incorporated as appropriate.

As agreed with your offices, unless you publicly announce the contents of
this letter earlier, we plan no further distribution until 15 days from
the report date. At that time, we will send copies of the report to the
EPA Administrator, DOE Secretary, and other interested parties. We will
also make copies available to others upon request. In addition, the report
will be available at no charge on the GAO Web site at http://www.gao.gov.

If you have any questions about this report, please contact me at (202)
5123841 or [email protected]. Contact points for our Offices of
Congressional Relations and Public Affairs may be found on the last page

of this report. GAO staff who made major contributions to this report are
listed in appendix VII.

John B. Stephenson Director, Natural Resources and Environment

List of Requesters

The Honorable Olympia J. Snowe
Chair, Committee on Small Business and Entrepreneurship
United States Senate

The Honorable James M. Jeffords
Ranking Minority Member
Committee on Environment and Public Works
United States Senate

The Honorable Joseph I. Lieberman
Ranking Minority Member
Committee on Homeland Security and Governmental Affairs
United States Senate

The Honorable Patrick J. Leahy
Ranking Minority Member
Committee on the Judiciary
United States Senate

The Honorable Thomas R. Carper
Ranking Minority Member
Subcommittee on Clean Air, Climate Change, and Nuclear Safety
Committee on Environment and Public Works
United States Senate

The Honorable Barbara Boxer
Ranking Minority Member
Subcommittee on Superfund and Waste Management
Committee on Environment and Public Works
United States Senate

The Honorable Hillary Rodham Clinton
United States Senate

The Honorable Mark Dayton
United States Senate

The Honorable Frank Lautenberg
United States Senate

Appendix I

                       Objectives, Scope, and Methodology

Congressional requesters asked us to (1) describe the use, availability,
and effectiveness of technologies to reduce mercury emissions at power
plants; and (2) identify the factors that influence the cost of these
technologies and report on available cost estimates. To respond to these
objectives, we surveyed a nonprobability sample of 59 key stakeholders in
three groups, including 22 mercury control technology vendors, 21
representatives of the coal-fired power industry, and 16 individual
researchers and/or government officials.1 We supplemented and
corroborated, to the extent possible, the survey information through
structured interviews with 14 stakeholders who view the reduction of
mercury emissions from a policy perspective, including senior staff at
EPA's Office of Policy Analysis and Review and DOE's Office of Fossil
Energy. Finally, we interviewed vendors and researchers of mercury
emissions monitoring technology to obtain and analyze information on the
availability and reliability of mercury monitoring devices.

Our work dealt with (1) technologies or measures that are specifically
intended to control mercury emissions and (2) modifications to existing
controls for other pollutants (e.g., nitrogen oxides, particulate matter,
or sulfur dioxide) that are specifically intended to enhance mercury
removal. We did not assess the availability, use, cost, or effectiveness
of controls for other pollutants that capture mercury as a side-benefit
because EPA had already conducted an extensive analysis of that topic as
part of the rule development process. As a result, our work addressed only
technologies specifically intended to control mercury. We did not
independently test these technologies. Lastly, we focused on technologies
that had advanced to the field-test stage rather than on technologies in
earlier stages of testing. Most of the test data we reviewed were from
full-scale tests, but the field tests of less developed controls, such as
some multipollutant controls, were not full-scale. In these cases, the
data were obtained from slipstream tests at power plants, where segments,
rather than the entire stream, of the flue gas were diverted for testing.

We relied primarily on surveys to obtain current data and professional
judgment on the status of mercury controls. We developed three different
surveys, one for each stakeholder group, which requested information about
the availability, use, effectiveness, and cost of mercury control

1Results from nonprobability samples cannot be used to make inferences
about a population because in a nonprobability sample some elements of the
population being studied have no chance or an unknown chance of being
selected as part of the sample.

Appendix I
Objectives, Scope, and Methodology

technologies. The scope and nature of some questions varied between the
three surveys in order to reflect the varying expertise of each
stakeholder group. To the extent possible, we structured the questions to
facilitate comparisons between the responses of each stakeholder group. We
used this format because we expected researchers, government officials,
and power industry respondents to possess broad knowledge about a
portfolio of mercury controls while technology vendors would have
extensive information about a limited number of controls, or those that
they produce, develop or sell. The most significant difference between the
three surveys was that we asked technology vendors to answer questions
only about the control produced, developed, or sold by each vendor,
whereas the questions for researchers, government officials, and power
industry respondents were not limited to one mercury control.

We developed the three surveys with survey specialists between July 2004
and October 2004. We took steps in the design, data collection, and
analysis phases of the work to minimize nonsampling and data processing
errors. We conducted pretests of the surveys, and staff involved in the
evaluation and development of mercury control technologies within EPA's
Office of Research and Development and DOE's Office of Fossil Energy also
reviewed and commented on the three surveys. We made changes to the
content and format of the final surveys based on the pretests, comments of
EPA and DOE officials, and comments of our internal reviewer. We followed
up with those that did not respond promptly to our surveys. We also
independently verified the entry of all survey responses entered into an
analysis database as well as all formulas used in the analyses.

We mailed paper copies of the surveys to 59 stakeholders and received 45
surveys from 40 stakeholders (68 percent response rate), which included 14
representatives of coal-fired power plants, 12 researchers and government
officials, and 14 technology vendors. Because we asked technology vendors
to complete one survey for each mercury control technology that they
develop, produce, or sell, the number of surveys exceeded the number of
respondents-five of the 14 vendors responding to our survey submitted more
than one survey. Upon receiving the surveys and reviewing the questions,
four stakeholders (1 power industry representative, 1 vendor, and 2
researchers/government officials) informed us that they were unable to
participate. Finally, we contacted each stakeholder who did not return a
survey by the deadline several times, either via email, phone, or both.

Appendix I
Objectives, Scope, and Methodology

We developed separate nonprobability samples for each of the three groups
we surveyed, identifying stakeholders based on the extent of their
expertise and involvement with the research, development, and
demonstration of mercury control technologies.

o 	To compile a list of mercury control technology vendors, we spoke with
DOE staff overseeing the mercury technology demonstration program to
identify companies that either manufacture a mercury control technology
for coal-fired power plants or research these technologies to develop them
commercially. Although we excluded from the technology vendors group any
company or organization that conducts research solely for evaluative or
academic reasons and lacks a significant financial interest in the
performance of the technology, we did include these stakeholders in the
researcher and government official group. Next, we spoke with DOE and
mercury technology vendors and reviewed available documents to identify
the stage of testing of each company's product(s), and we included on our
list the companies whose product(s) have undergone commercial
demonstrations, full-scale field tests, pilot-scale tests, or slipstream
tests. We then corroborated the list of mercury control technology vendors
with the Institute of Clean Air Companies, the national trade organization
for air pollution control vendors, to ensure the completeness of the list
of mercury control vendors. Our survey of mercury control technology
vendors included a representative from each of the 22 companies we
identified as meeting these criteria.

o 	We identified an initial list of 21 representatives from the coal-fired
power industry to participate in our survey based primarily on a list
generated from Platts' POWERdat database of the power generators who
burned the most coal in calendar year 2002, which is the most recent year
of available data. We determined that this database was sufficiently
reliable for this purpose. We based our selection of stakeholders on the
quantity of coal burned because it correlated more closely with mercury
emissions than any other available variable. We included a representative
from each of the 20 generators that burned the most coal in calendar year
2002, accounting for 60 percent of the coal burned for power generation in
that year in the United States. One company from this list declined to
participate in our survey. Therefore, we added the next-largest company on
the list. This final group of 20 generators accounted for 59 percent of
the coal burned for power generation in that year. Additionally, we added
one company to our group of generators-resulting in a total of 21
generators surveyed-

Appendix I
Objectives, Scope, and Methodology

because it had begun a commercial demonstration of a mercury control
technology. Next, we corroborated our list of generators by asking
representatives of the following organizations to identify contacts within
the coal-fired power industry who would be knowledgeable of mercury
control technologies: (1) three power companies that have actively
participated in mercury control technology demonstrations; (2) the Edison
Electric Institute, the trade association for electric utilities; and (3)
the National Rural Electric Cooperative Association, which represents
utilities serving rural communities. The power industry stakeholders
identified by these three organizations all corresponded with those we had
placed in the group of 21 generators.

o 	For the survey targeting researchers and government officials, we
included senior agency staff involved in the evaluation and development of
mercury control technologies within EPA's Office of Research and
Development and DOE's National Energy Technology Laboratory, state
government officials in states that initiated action to limit mercury
emissions from power plants, and experts from companies and nonprofit
organizations that do research on mercury control technologies. We
coordinated with the State and Territorial Air Pollution Program
Administrators/Association of Local Air Pollution Control Officials, the
national association of state and local air pollution control agencies, to
identify nine states that had initiated actions to reduce mercury
emissions from power plants and the state officials that had been involved
with research and development of mercury control technologies. After
speaking with representatives from these states, we eliminated one of the
states because the legislation did not specifically target mercury
emissions. We spoke to representatives of the following eight states:
Connecticut, Illinois, Iowa, Massachusetts, New Hampshire, New Jersey,
North Carolina, and Wisconsin.

We recognized that the technology vendors and power industry respondents
might have had concerns about disclosing sensitive or proprietary
information. Therefore, although we have included a list of the survey
respondents below, this report does not link individual survey responses
to any particular technology vendor or representative of the coal-fired
power industry. We mailed the survey to stakeholders on October

Appendix I
Objectives, Scope, and Methodology

22, 2004, and asked to receive responses by November 8, 2004. Of the 59
stakeholders we contacted, the following 41 responded to our survey:2

o  ADA Environmental Solutions

o  ADA Technologies Incorporated

o  AES Corporation

o  Alstom Power

o  American Electric Power Company, Incorporated

o  Andover Technologies

o  Apogee Scientific, Incorporated

o  Babcock Power Incorporated

o  Basin Electric Power Cooperative

o  CarboChem

o  Cormetech, Incorporated

o  Dominion Resources, Incorporated

o  Electric Power Research Institute

o  Enerfab Clean Air Technologies (CR Clean Air Technologies)

o  FirstEnergy Corporation

o  EnviroScrub Technologies Corporation

o  Hamon Research Cottrell

2We received responses from 41 stakeholders, but 2 of these respondents
completed one survey together in order to describe a product produced by
both companies. Because the 2 stakeholders completed one survey for one
mercury control, we counted this as one response as part of our survey
analysis.

Appendix I
Objectives, Scope, and Methodology

o  Illinois Environmental Protection Agency, Bureau of Air  o KFx

o  Mobotec USA

o  NORIT-Americas, Incorporated

o  New Hampshire Department of Environmental Sciences

o  New Jersey Department of Environmental Protection

o  North Carolina Division of Air Quality

o  Powerspan

o  PPL Corporation

o  Progress Energy, Incorporated

o  Reaction Engineering

o  Reliant Energy Incorporated

o  Scottish Power Plc (Known as Pacificorps in the U.S.)

o  Sorbent Technologies Corporation

o  Southern Company

o  Southern Research Institute

o  TXU Corporation

o  Tennessee Valley Authority

o  United Technologies

o  U.S. Department of Energy, National Energy Technology Laboratory

o 	U.S. EPA, Office of Research and Development, Air Pollution Prevention
and Control Division

Appendix I
Objectives, Scope, and Methodology

o  We Energies

o 	Wisconsin Department of Natural Resources, Bureau of Air Management

o  Xcel Energy, Incorporated

We supplemented and corroborated, to the extent possible, the survey
information with testimonial evidence. This included structured interviews
with 14 policy stakeholders familiar with the policy implications of
mercury control technology research, including senior staff at EPA's
Office of Policy Analysis and Review and DOE's Office of Fossil Energy,
state and local regulatory organizations, electric utility associations,
and environmental organizations.3 We developed a nonprobability sample for
the group of policy stakeholders. We worked with a survey expert to
develop a set of structured interview questions about the availability,
use, effectiveness, and cost of mercury control technologies. In order to
minimize nonsampling error, we took steps to ensure that the questions
were unambiguous, balanced, and clearly understandable. The interview
questions were similar to the survey questions, but tailored to reflect
the policy expertise of the interview participants. For example, rather
than asking interview participants to provide data on mercury technology
demonstrations, we sought their views on the implications of mercury
technology demonstrations for mercury policies. We conducted pretests of
the structured interview, including one with an EPA official in the Office
of Policy Analysis and Review. We made changes to the content and format
of the final interview questions based on the pretests.

We conducted the 14 structured interviews between November 2004 and
December 2004 with stakeholders from the following 13 organizations:4

o  American Public Power Association

o  Clean Air Task Force

3The policy stakeholders we interviewed did not participate in the three
surveys we conducted.

4We conducted 14 interviews with stakeholders representing these 13
organizations. In order to include the perspective of several senior air
policy staff at EPA, we conducted two interviews with the agency.

Appendix I
Objectives, Scope, and Methodology

o  Edison Electric Institute

o  Institute of Clean Air Companies

o MJ Bradley

o  National Rural Electric Cooperative Association

o  National Wildlife Federation

o  Northeast States for Coordinated Air Use Management

o  Regional Air Pollution Control Agency

o 	State and Territorial Air Pollution Program Administrators/Association
of Local Air Pollution Control Officers

o  U.S. Department of Energy, Office of Fossil Energy

o 	U.S. Environmental Protection Agency, Office of Air and Radiation,
Office of Policy Analysis and Review

o 	U.S. Environmental Protection Agency, Office of Air and Radiation,
Office of Air Quality Planning and Standards

Finally, because of the important role monitoring data play in the
regulation of air pollutants, we gathered and analyzed information on the
availability and reliability of two kinds of mercury monitoring devices-
sorbent traps and continuous emissions monitors-by conducting seven
structured interviews with the technology vendors and researchers in the
government and private sectors. We developed the list by consulting with
EPA's lead expert on mercury monitoring technology and then comparing it
to the list of presenters at DOE's Mercury Measurements Workshop, which
was conducted in July 2004. Because this list of monitoring technology
vendors primarily represented one of the two advanced mercury monitors, we
included an organization regarded as a major developer of the other
mercury monitoring device. Finally, we also included researchers and
government stakeholders with broad knowledge of the mercury monitoring
industry.

We could not interview all 18 stakeholders we identified for the sorbent
trap and continuous emissions monitors because of time constraints.

Appendix I
Objectives, Scope, and Methodology

Therefore, we decided to (1) interview four researchers and government
officials, (2) interview the major producer of sorbent traps, and (3)
interview a random sample of the multiple vendors involved with the eight
kinds of continuous emissions monitors. Within this last group, we
compiled a list of 13 mercury monitoring vendors, which was then
randomized by a senior GAO methodologist. We interviewed the first 3
stakeholders on the randomized list of 13 mercury monitoring vendors in
order to include their knowledge and perspectives on the industry. We were
not able to reach the sorbent trap producer for an interview.

We based the questions for the monitoring interviews on those posed in the
mercury control technology surveys, including the same concepts and
emphasizing the availability and level of demonstration of monitoring
technologies, and again took steps to minimize nonsampling errors. We
conducted two pretests of the monitoring interviews. Finally, we
corroborated the numerical values used in questions about the accuracy and
reliability of mercury monitors with EPA's mercury monitoring expert in
the Office of Research and Development. We made changes to the content and
format of the final interview questions based on the pretests and the EPA
official's comments.

Lastly, we identified and reviewed governmental and nongovernmental
reports estimating the cost of mercury control technologies. We identified
two government cost reports-one from EPA and one from DOE-and four
nongovernmental cost reports. We excluded two of the nongovernmental
reports from our analysis because these reports addressed cost issues that
were either too limited in scope or were not germane to our research
objectives. We then reviewed the results of both government reports and
two remaining nongovernmental reports as part of our technology cost
analysis. We took several steps to assess the validity and reliability of
computer data underlying the cost estimates in the EPA, DOE, and
nongovernmental reports which were discussed in our findings, including
reviewing the documentation and assumptions underlying EPA's economic
model and assessing the agency's process for ensuring that the model data
are sufficient, competent, and relevant. We determined that these four
reports are sufficiently reliable for the purposes of this report.

As part of our effort to consider data on mercury control demonstrations
and costs, we assessed compliance with internal controls related to the
availability of timely, relevant, and reliable information. We also
obtained data on mercury emissions. Because the emissions data are used
for background purposes only, we did not assess their reliability.

Appendix I
Objectives, Scope, and Methodology

We performed our work between May 2004 and May 2005 in accordance with
generally accepted government auditing standards.

Appendix II

Availability and Costs of Mercury Monitoring Technology

This appendix provides information on technologies that facilities may use
to monitor mercury emissions, including background information on
monitoring technologies and requirements under EPA's mercury rule, as well
as on the availability and cost of different monitoring technologies.

Background	In addition to technologies that control emissions, those that
monitor the amount of a pollutant emitted can play an equally important
role in the success of an air quality rule's implementation. For example,
effective emissions monitoring assists facilities and regulators in
assuring compliance with regulations. In some cases, monitoring data can
also help facilities better understand the efficiency of their processes
and identify ways to optimize their operations.

Accurate emissions monitoring is particularly important for trading
programs, such as that established by the mercury rule. According to EPA,
the most widespread existing requirements for using advanced monitoring
technologies stem from EPA's Acid Rain program. Under the program, power
plants have been allowed to buy and sell emissions allowances, but each
facility must hold an allowance for each ton of sulfur dioxide it emitted
in a given year; furthermore, facilities must continuously monitor their
emissions.1 According to EPA, monitoring ensures that each allowance
actually represents the appropriate amount of emissions, and that
allowances generated by various sources are equivalent, instilling
confidence in the program. Conversely, a study by the National Academy of
Public Administration found that the lack of monitoring in other trading
programs led to difficulty in ensuring the certainty of emissions
reductions.

EPA's mercury rule requires mercury emissions monitoring and quarterly
reporting of mercury emissions data. For plants that emit at least 29
pounds of mercury annually, EPA requires continuous emissions monitoring,
while sources that emit less than this amount may instead conduct periodic
testing-testing their emissions once or twice a year depending on their
emissions level. According to EPA, the mercury emissions from sources
exempt from continuous monitoring comprise approximately 5 percent of
nationwide emissions. EPA estimates that the

1The Clean Air Interstate Rule revised these provisions of the Acid Rain
Program to require additional allowances beginning in the year 2010.

      Appendix II Availability and Costs of Mercury Monitoring Technology

annual impact in monitoring costs for the entire industry will total $76.4
million.2

EPA Expects That Monitoring Technologies Will Be Available Prior to the
Compliance Deadlines

EPA expects that two technologies will be available to monitor mercury
emissions continuously prior to the rule's deadline and requires
continuous emissions monitoring for most facilities either by a Continuous
Emissions Monitoring System (CEMS) or a sorbent trap monitoring system,
while facilities that emit low levels of mercury can conduct periodic
monitoring using a testing protocol known as the Ontario-Hydro Method:

o 	CEMS continuously measures pollutants released by a source, such as a
coal-fired power plant. Some CEMSs extract a gas sample from a facility's
exhaust and transport it to a separate analyzer while others allow
effluent gas to enter a measurement cell inserted into a stack or duct.
This allows for continuous, real-time emissions monitoring. EPA estimates
that a unit's annual CEMS operating, testing, and maintenance cost would
be about $87,000, while a unit's capital cost would be about $70,000.

o 	Sorbent trap monitoring systems collect a mercury sample by passing
flue gas through a mercury trapping medium, such as an activated carbon
tube. This sample is periodically removed and sent to a lab for analysis.
The rule requires that the average measurement of two separate sorbent
trap readings be reported. Sorbent trap monitoring allows for continuous
monitoring, but is not considered a real-time method. EPA estimates that a
unit's annual sorbent trap operating and testing costs would be about
$113,000 per year, while a unit's capital cost would be about $20,000.

o 	The Ontario-Hydro Method, a periodic testing method, involves manually
extracting a sample of flue gas from a coal-fired plant's stack or duct,
usually over a period of a few hours, which is then analyzed in a
laboratory. EPA estimates this method would cost about $12,500 a year for
two tests and about $7,000 for one test.

2Based on the annualized capital and operating costs of the technologies
units are expected to use and the number of units expected to use each
technology.

      Appendix II Availability and Costs of Mercury Monitoring Technology

Stakeholders Believe That Mercury Monitoring Technology Is Available,
Reliable, and Will be Able to Meet Quality Control and Assurance Standards
by 2008

All of the stakeholders we asked about the availability of CEMS or sorbent
trap systems said that the technologies were available for purchase.
Furthermore, an EPA monitoring technology expert and the vendors we
interviewed agreed that there were no technical or manufacturing
challenges that would prevent vendors from supplying monitors to coalfired
power plants by 2008. However, some researchers identified factors that
could affect vendors' ability to supply monitors by that date, including
whether vendors had sufficient production capacity to meet the industry's
demand for the equipment. All three vendors we interviewed were aware of
power plants in other countries that had installed mercury monitoring
equipment (including Germany, Japan, and the United Kingdom). Two
respondents were aware of power plants in the United States that had
permanently installed mercury monitoring equipment.

Most researchers considered CEMS and sorbent trap technologies to be
accurate and reliable, and the CEMS vendors also characterized their
technologies as accurate and reliable. Researchers cited the need for
additional testing of certain subcomponents of the continuous monitoring
systems. Stakeholders were generally confident that these technologies
would be able to meet proposed quality control and assurance standards by
2008, although two researchers expressed concerns that EPA's proposed
standards might be too strict for CEMS to meet.

According to EPA, recent field tests have demonstrated that sorbent trap
systems can be as accurate as CEMS. The rule requires the implementation
of quality assurance procedures for sorbent trap monitoring systems, which
EPA says are based on field research and input from parties that commented
on the agency's mercury rule during the public comment period. EPA
acknowledges that there may be problems with the technology, such as the
possibility of the traps becoming compromised, lost, or broken during
transit or analysis, which could result in missing data; however, EPA also
believes steps can be taken to minimize these possibilities.

Appendix III

Summary of Field-Scale Tests of Mercury Controls

The table below summarizes data about mercury control tests, including the
power plant location, duration of continuous testing, coal type, and
average mercury removal. We obtained data from DOE's National Energy
Technology Laboratory and from the 40 survey respondents about field
tests. The tests that have been partially funded by DOE's National Energy
Technology Laboratory are identified in the table below by an asterisk
symbol.

              Table 1: Summary of Mercury Control Field Test Data

Mercury                                     Test                 Average   
control                                                          mercury   
category  Technology    Location   Duration year   Coal type   reductiona  
            Activated                                               Various   
Sorbent  carbon*      Wilsonville,          2001  Bituminous      test     
                              AL       9 days                       results   
                                                                  reported to 
                                                                  GAO: 78-90  
                                                                    percent   
            Activated    Pleasant     Three    2001 Subbituminous    46-73    
            carbon*      Prairie,     5-day                         percent   
                              WI       tests                      
            Activated                                               Various   
            carbon*                            2002  Bituminous      test     
                         Somerset, MA 10 days                       results   
                                                                  reported to 
                                                                   GAO: 85 to 
                                                                           90 
                                                                   percent.   
            Activated     Salem, MA    4 days  2002  Bituminous      85-95    
            carbon*                                                percentb   
            Activated    Underwood,    5 days  2003    Lignite    70 percent  
            carbon       ND                                       
            Activated     Denver, CO  6 daysc  2004 Subbituminous 64 percent  
            carbon                                                
            Activated     Denver, CO  3 hours  2004 Subbituminous 86 percent  
            carbon                                                
            Activated    Undisclosed   1 day   2004 Subbituminous 30 percent  
            carbon                                                
            Activated    Undisclosed   2 days  2004 Subbituminous 55 percent  
            carbon                                                
            Activated                 1 month  2004                           
            carbon       Stanton, ND                   Lignite    63 percent
            and sorbent                                           
            enhancement*                                          

                                  Appendix III
                    Summary of Field-Scale Tests of Mercury
                                    Controls

                         (Continued From Previous Page)

              Activated carbon* Newnan, GA 1 month 2004 Bituminous

According to preliminary analysis, removal varied by measurement point
within the process:

ESPd+ACI, removal ranged from a minimum of 50 to a maximum of 91 percent
(majority data 60-85 percent);

ESP+ACI+scrubber, removal ranged from a minimum of 50 to a maximum of 97
percent (majority data 70-94 percent)

       Activated                                                          Not 
         carbon*    Newnan, GA         Not 2004-      Bituminous   available: 
                                                                      testing 
                                available:  2005                     ongoinge 
                                   testing                       
                                  ongoinge                       
       Activated                                                          Not 
          carbon                  2 months  2005                   available: 
                    Beulah, ND                           Lignite      testing 
     and sorbent                                                     ongoinge 
    enhancement*                                                 
       Activated                            2005          Blend:      Not yet 
        carbon*f    Monroe, MI     Not yet        Subbituminous/      testede 
                                   testede            Bituminous 
       Activated Conesville,                2005      Bituminous      Not yet 
        carbon*f OH                Not yet                            testede 
                                   testede                       
                                                                      Average 
                                   Several  2003      Bituminous      varied; 
      Chemically Cliffside, NC                                        mercury 
        enhanced                                                      removal 
          carbon               multi-hour                         ranged from 
                                                                            a 
                                tests over                         minimum of 
                                        1-                         20 percent 
                                                                         to a 
                               week period                         maximum of 
                                                                   90 percent 
      Chemically    Athens, OH     Several  2003      Bituminous   70 percent 
        enhanced               multi-hour                        
          carbon                                                 
                                tests over                       
                                        2-                       
                               week period                       
      Chemically St. Louis, MO     30 days  2004   Subbituminous   90 percent 
        enhanced                                                 
         carbon*                                                 
      Chemically   Near Garden     30 days  2004   Subbituminous   90 percent 
        enhanced      City, KS                                   
         carbon*                                                 
                    East China              2004     Blend:                   
      Chemically                   30 days         Bituminous      94 percent
        enhanced  Township, MI                    /Subbituminous 
         carbon*                                                 
      Chemically   Undisclosed     Greater  2004         Lignite   70 percent 
        enhanced                   than 10                       
          carbon                                                 
                                      days                       
      Chemically   Undisclosed       1 day  2004   Subbituminous   60 percent 
        enhanced                                                 
          carbon                                                 

                                  Appendix III
                    Summary of Field-Scale Tests of Mercury
                                    Controls

                         (Continued From Previous Page)

Chemically    Stanton, ND   24 days  2004    Lignite        70 percent     
    enhanced                                               
     carbon*                                               
Chemically    Stanton, ND   1 month  2004    Lignite    Not yet availablee 
    enhanced                                               
     carbon*                                               
Chemically    Spencer, NC   3 months 2005  Bituminous   Not yet availablee 
    enhanced                                               
     carbon*                                               
Chemically    Stanton, ND     TBDg   TBDg    Lignite     Not yet testede   
    enhanced                                               
     carbon*                                               
Chemically   Portland, PA     TBDg   TBDg  Bituminous    Not yet testede   
    enhanced                                               
     carbon*                                               
Chemically   Located near     TBDg   TBDg  Bituminous    Not yet testede   
    enhanced   Goldsboro, NC                               
     carbon*                                               
Chemically  Romeoville, IL    TBDg   TBDg Subbituminous  Not yet testede   
    enhanced     (tentative                                
     carbon*                                               
                  location)                                
Chemically   Glenrock, WY     TBDg   TBDg Subbituminous  Not yet testede   
    enhanced                                               
     carbon*                                               
Chemically    Chicago, IL     TBDg   TBDg Subbituminous  Not yet testede   
    enhanced                                               
     carbon*                                               
Chemically   Muscatine, IA    TBDg   TBDg Subbituminous  Not yet testede   
    enhanced                                               
     carbon*                                               
Chemically  Council Bluffs,   TBDg   TBDg Subbituminous  Not yet testede   
    enhanced         IA                                    
     carbon*                                               
Non-Carbon    Denver, CO    6 hours  2004 Subbituminous     28 percent     

Non-Carbon Denver, CO 6-7 daysh 2004 Subbituminous	Various test results
reported to GAO: 51 percent reported for 7-day test; 57-68 percent
reported for 6-day test

Non-Carbon* North Bend, OH 1 month 2005 Bituminous Not yet availablee

Multipollutant Activated      Wilsonville,    5    2003 Bituminous      86 
                  carbon         AL           months                  percent 
                   and enhanced                                       
                  particulate                                         
                  collection*i                                        

    Activated carbon  Cheshire, OH   TBDg  TBDg  Bituminous   Not yet testede 
      and enhanced                                           
      particulate                                            
      collection*i                                           

    Activated carbon  Newark, AR  TBDg  TBDg  Subbituminous   Not yet testede 
      and enhanced                                           
      particulate                                            
      collection*i                                           

Activated carbon Near           TBDg TBDg      Lignite or          Not yet 
                    Fairfield,                                        testede 
     and enhanced         TX                 Lignite/Subbituminous 
     particulate                                     blend         
     collection*i                                                  

                                  Appendix III
                    Summary of Field-Scale Tests of Mercury
                                    Controls

                         (Continued From Previous Page)

             Sorbent and    Moncure, NC   14 days  2002   Bituminous   80 percent 
                 high                                                  
             velocity air                                              
               Wet ESP     Shippingport,    Not    2001- Subbituminous            
                           PA                                          78 percent
                                         specified 2003                
                                                                Blend:            
            Corona                                 2004     Bituminous 
            Dischargej     Shadyside, OH  6 days                   and 75 percent
                                                         subbituminous 
  Mercury   Chlorine-based Located near  2 months  2005                 Not yet   
 oxidation                                                  Lignite     testede   
              additives*    Center, ND   expected                      
            Chlorine-based Mt. Pleasant,  1 month  2005                 Not yet   
                           TX                               Lignite     testede   
              additives*                 expectede                     
Enhancement    Wet FGDk     Moscow, OH    2 weeks  2001   Bituminous   52 percent 
              Additive*                                                
            Wet FGD         Litchfield,  4 months  2001   Bituminous   79 percent 
            Additive*           MI                                     
               Wet FGD       Mt. Storm             2004   Bituminous              
               Additive        Lake,      3 days                       71 percent
                           northeastern                                
                                WV                                     
               Wet FGD       Mt. Storm             2004   Bituminous    Over 90   
               Additive        Lake,      7 days                        percent   
                           northeastern                                
                                WV                                     
            Wet FGD         Newnan, GA     TBDg    TBDg   Bituminous    Not yet   
            Additive*                                                   testede   
            Wet FGD         Conesville,    TBDg    TBDg   Bituminous    Not yet   
            Additive*           OH                                      testede   
            Wet FGD        Mt. Pleasant,   TBDg    TBDg     Lignite     Not yet   
            Additive*      TX                                           testede   
                                                                              Not 
Other    Fixed sorbent   Stanton, ND  6 months  2004- Lignite, then available: 
                                                         switched         testing 
              structure*                 expected  2005       to        ongoinge  
                                                         subbituminous 
                                                            during     
                                                            testing    
            Fixed sorbent                5 months  2005   Bituminous    Not yet   
                            Newnan, GA                                 availablee 
              structure*                 expected                      
              Combustion    Rogersvile,    TBDg    TBDg   Bituminous    Not yet   
                                TN                                      testede   
            modification*                                              

Source: DOE National Energy Technology Laboratory and GAO analysis of
survey responses.

*Field tests partially funded by DOE's National Energy Technology
Laboratory.

aAverage mercury removal reflects the total mercury removal achieved by
the entire system of pollution controls, not just the mercury control,
installed at the power plant.

bMeasurements obtained over a four-day test showed overall mercury capture
of 85 to 95 percent.

cThe test was conducted for 3 to 8 hours per day.

dESP is the abbreviation for electrostatic precipitator.

eThis is based on DOE's information as of February 2005.

fThe research team has not yet finalized the selection of sorbent for this
test. The research team is testing activated carbon, but will also
consider using chemically enhanced carbon injection at this site.

gTBD means to be determined. DOE's National Energy Technology Laboratory
had awarded funding for this project but a specific testing timeframe had
not been identified yet as of February 2005.

hThe test was conducted for 3 to 8 hours per day. Survey respondents
reported test durations of 6 days and 7 days.

Appendix III
Summary of Field-Scale Tests of Mercury
Controls

iThis combination of pollution controls includes an enhanced, compact
fabric filter designed to capture mercury and particulates at plants
already using an electrostatic precipitator.

jEPA describes corona discharge technology as the "generation of an
intense corona discharge (ionization of air by a high voltage electrical
discharge)" in the flue gas (page 7-43). The corona discharge triggers a
series of chemical reactions that are intended to improve the capture of
mercury and particulate matter. US EPA, National Risk Management Research
Laboratory, Control of Mercury Emissions from Coal-Fired Electric Utility
Boilers: Interim Report Including Errata Dated March 21, 2002 (Research
Triangle Park, NC, 2002).

kFGD is the abbreviation for flue gas desulfurization.

Appendix IV

Summary of Stakeholder Perceptions about Availability of Mercury Controls

This appendix provides more detailed information on stakeholders' views
regarding the availability of the different mercury controls. Please refer
back to appendix I for details about our survey methodology.

Of the stakeholders that either responded to our survey (40) or
participated in an interview (14), a majority (40) believed that at least
one technology was currently available for purchase. As shown in table 2,
many of the researchers and government officials said that activated
carbon injection (8 of 12) and chemically enhanced carbon (7 of 12) are
currently available, while less than half of the power industry officials
also believe activated carbon injection technology is available (6 of 14).
All of the vendors associated with carbon-based sorbent injection,
including activated carbon (4) and chemically enhanced carbon (2),
described their technology as available. In addition, 13 of the 14 policy
stakeholders we interviewed- those who do not participate in technology
research but are involved in the development of mercury control policy,
including representatives of EPA, DOE, regional and local air pollution
agencies, environmental advocacy groups, and the electric utility
industry-believe that sorbent technology is currently available for
purchase.

                                  Appendix IV
                    Summary of Stakeholder Perceptions about
                        Availability of Mercury Controls

Table 2: Stakeholder Perceptions on Availability of Sorbent Technologiesa

    Technology     Stakeholder    Available Not         Do not Did not  Total 
                      group                 available   know    answer  
     Activated                                                          
      carbon     Coal-fired power                                       
     injection       industry             6           3      3        2 
       (ACI)                                                            
                 Researchers and                                        
                    government            8           1      1        2 
                    officials                                           
                    Technology            4           0      0        0 
                     vendorsb                                           
                      Policy             13           1      0        0 
                   stakeholders                                         
       Total                             31           5      4        4   44b 
     responses                                                          
    Chemically                                                          
     enhanced    Coal-fired power                                       
        ACI          industry             3           5      4        2 
                 Researchers and                                        
                    government            7           1      1        3 
                    officials                                           
                    Technology            2           0      0        0 
                     vendorsb                                           
                      Policy             11           1      2        0 
                   stakeholders                                         
       Total                             23           7      7        5   42b 
     responses                                                          
    Non-carbon                                                          
      sorbent    Coal-fired power                                       
                     industry             0           8      4        2 
                 Researchers and                                        
                    government            1           2      5        4 
                    officials                                           
                    Technology            1           1      0        0 
                     vendorsb                                           
                      Policy              4           4      6        0 
                   stakeholders                                         
       Total                              6          15     15        6   42b 
     responses                                                          

Source: GAO.

aGiven the uncertainty about federal mercury reduction goals that existed
prior to the March 2005 mercury rule and the fact that field testing of
mercury controls is ongoing, some of the stakeholders were reluctant to
make conclusions about the availability of all mercury controls when we
asked them in November and December 2004. Therefore, some participants did
not answer this question, and the number of responses for each mercury
control reflects in part the extent of field testing.

bThe number of responses for the question on availability does not
correspond to the overall number of survey responses because the
availability question differed slightly for technology vendors. We did not
seek the technology vendors' perceptions of all mercury controls, an
option we gave the other stakeholders, but asked the vendors whether the
mercury control they produce, develop, and/or sell is available for
purchase without regard to technology effectiveness.

The survey responses regarding the availability of other mercury controls
were more limited and less optimistic than those for sorbent injection.
While 40 of the 54 stakeholders answered questions about the availability
of activated carbon injection, far fewer respondents answered the

                                  Appendix IV
                    Summary of Stakeholder Perceptions about
                        Availability of Mercury Controls

questions about the availability of other controls.1 As shown in table 3,
the stakeholders who responded to questions about nonsorbent control
technologies, such as multipollutant controls, mercury oxidation
technologies, and enhancements to existing controls for other pollutants,
were more mixed in their views about the availability of these
technologies. For example, researchers and government officials expressed
a range of views about mercury oxidation technologies-4 believe they are
available, 3 do not think they are available, 2 did not know, and 3 chose
not to answer this question.

    Table 3: Stakeholder Perceptions on Availability of Non-Sorbent Mercury
                                   Controlsa

                         Stakeholder             Not       Do   Did not       
       Technology           group      Available available not  answer  Total
                                                           know         
                         Coal-fired                                     
    Mercury oxidation       power                                       
      technologies        industry             0         8    4       2 
                       Researchers and                                  
                         government            4         3    2       3 
                          officials                                     
                         Technology            1         1    0       0 
                          vendorsb                                      
                           Policy              5         6    3       0 
                        stakeholders                                    
     Total responses                          10        18    9       5   42b 
     Multipollutant      Coal-fired                                     
        controls            power                                       
                          industryc            4         3    0       9 
                       Researchers and                                  
                         government            6         2    0       6 
                         officialsc                                     
                         Technology            4         4    0       0 
                         vendorsb,c                                     
                           Policy             12         4    2       3 
                        stakeholdersc                                   
     Total responses                          26        13    2      18 59b,c 
     Enhancements to     Coal-fired                                     
                            power                                       
    existing controls     industryd            0         2    1      12 
                       Researchers and                                  
                         government            5         4    0       6 
                         officialsd                                     
                         Technology            1         0    0       0 
                         vendorsb,d                                     
                           Policy             18         1    0       5    24 
                        stakeholdersd                                   
     Total responses                          24         7    1      23 55b,d 

Source: GAO.

aGiven the uncertainty about federal mercury reduction goals that existed
prior to the March 2005 mercury rule and the fact that field testing of
mercury controls is ongoing, some of the stakeholders

1Ten of the 14 vendors were not asked to provide views on the availability
of activated carbon because these vendors do not produce, develop, or sell
this technology.

Appendix IV
Summary of Stakeholder Perceptions about
Availability of Mercury Controls

were reluctant to make conclusions about the availability of all mercury
controls when we asked them in November and December 2004. Therefore, some
participants did not answer this question, and the number of responses for
each mercury control reflects in part the extent of field testing.

bThe number of responses for the question on availability does not
correspond to the overall number of survey responses because the
availability question differed slightly for technology vendors. We did not
seek the technology vendors' perceptions of all mercury controls, an
option we gave the other stakeholders, but asked the vendors whether the
mercury control they produce, develop, and/or sell is available for
purchase without regard to technology effectiveness.

cThe number of responses for the question on availability for
multipollutants controls does not correspond to the overall number of
survey responses because some stakeholders identified more than one
multipollutant control and provided different responses about the
availability of those controls.

dThe number of responses for the question on availability for enhancements
to existing controls does not correspond to the overall number of survey
responses because some stakeholders identified more than one enhancement
and provided different responses about the availability of those
enhancements.

Finally, the 14 policy stakeholders we interviewed also expressed mixed
views on the availability of mercury controls. Nine described various
multipollutant controls as available, 5 viewed mercury oxidation as
available, and 8 regarded various enhancements to existing technologies as
available.

Appendix V

Stakeholder Confidence in Ability of Technologies to Achieve Mercury
Reductions under Three Scenarios

This appendix summarizes the perceptions of survey respondents in the
ability of mercury controls to reduce emissions under three scenarios.
(Appendix I provides details about our survey methodology.)

We asked survey respondents to assess their confidence in the ability of
power plants to achieve mercury reductions of 50, 70, and 90 percent by
the year 2008 under two different scenarios. The first scenario resembled
the cap-and-trade approach recently finalized by EPA in that it asked
stakeholders to consider whether the industry could use available
technologies to achieve industrywide reductions of 50, 70 or 90 percent by
2008. The second scenario was similar to an alternative approach
considered by EPA that would have required each plant to reduce emissions;
for this scenario we asked respondents whether each individual plant could
use available technologies to achieve the percentage reductions by 2008.1

As shown in tables 4 through 9, the confidence levels depended on the
level of reduction required and by stakeholder group. Overall, the
technology vendors answering this question expressed the greatest
confidence, while the power industry respondents were the least confident.
Within each stakeholder group, respondents expressed the greatest
confidence overall in achieving a 50 percent reduction by 2008-a reduction
that EPA requires under its 2010 cap-and progressively less confidence in
the 70 and 90 percent scenarios. For both possible control scenarios-the
national limit and facility-specific reductions-a majority of the 38
respondents2 expressed confidence in achieving the 50 percent reductions
(see tables 4 and 5), but many lacked confidence in the feasibility of 90
percent mercury reductions by 2008 (see tables 8 and 9). Respondents
expressed mixed opinions about the feasibility of 70 percent reductions by
2008, as shown in tables 6 and 7.

1GAO instructed respondents to consider whether such reductions were
feasible at most, but not all, power plants. This allowed survey
respondents to report confidence in mercury reduction at nearly all power
plants without considering highly unusual situations that might arise at
certain plants.

2This number differs from the number of responses because two of the 40
respondents did not answer these questions.

Appendix V Stakeholder Confidence in Ability of Technologies to Achieve Mercury
                        Reductions under Three Scenarios

  Table 4: Stakeholder Confidence in Reducing Nationwide Mercury Emissions 50
                       Percent by 2008 Stakeholder group

     Scale of   Confidence     Power     Researchers/government         
     mercury                  industry                                  
    reduction      level    respondents               officials Vendors Total 
                   Very                                                 
    50 percent   confident                                              
    reduction       or                                                  
nationwidea   confident             2                      9   12    
                   Less                5                      1       1 
                 confident                                              
                Not at all                                              
                 confident             6                      0       0 
                Do not know            1                      1       0 
      Total                                                 11b   13b   
respondents                        14                                

Source: GAO.

aThe survey asked stakeholders how confident they were that power plants
could reduce mercury emissions 50 percent by 2008. In this case,
respondents were asked to consider reductions averaged across power plants
in the United States, which does not mean that each individual plant would
achieve the reductions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Table 5: Stakeholder Confidence in Achieving Mercury Reductions of 50 Percent at
                  Nearly Every Plant by 2008 Stakeholder group

     Scale of    Confidence     Power    Researchers/government         
      mercury                 industry                                  
     reduction      level    respondents              officials Vendors Total 
                    Very                                                
    50 percent    confident                                             
reduction at      or                                                 
    nearly each   confident            2                      9   11    
power planta                                                         
                    Less               5                      1    2    
                  confident                                             
                 Not at all                                             
                  confident            6                      0    0    
                 Do not know           1                      1    0    
       Total                                                11b   13b   
    respondents                       14                                

Source: GAO.

aThe survey asked stakeholders to consider the likelihood that a single
power plant could reduce mercury emissions 50 percent by 2008. In this
case, respondents were asked to consider whether most, but not necessarily
all, power plants in the United States would each be capable of achieving
a 50 percent reduction in mercury emissions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Appendix V Stakeholder Confidence in Ability of Technologies to Achieve Mercury
                        Reductions under Three Scenarios

  Table 6: Stakeholder Confidence in Reducing Nationwide Mercury Emissions 70
                       Percent by 2008 Stakeholder group

     Scale of   Confidence     Power     Researchers/government         
     mercury                  industry                                  
    reduction      level    respondents               officials Vendors Total 
                   Very                                                 
    70 percent   confident                                              
    reduction       or                                                  
nationwidea   confident             0                      6      10 
                   Less                1                      3       3 
                 confident                                              
                Not at all                                              
                 confident            13                      1       0 
                Do not know            0                      1       0 
      Total                                                             
respondents                        14                    11b     13b 

Source: GAO.

aGAO asked stakeholders how confident they were that power plants could
reduce mercury emissions 70 percent by 2008. In this case, respondents
were asked to consider reductions averaged across power plants in the
United States, which does not mean that each individual plant would
achieve the reductions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Table 7: Stakeholder Confidence in Achieving Mercury Reductions of 70 Percent at
                  Nearly Every Plant by 2008 Stakeholder group

     Scale of                   Power    Researchers/government         
      mercury                 industry                                  
     reduction   Confidence  respondents              officials Vendors Total 
                    level                                               
                    Very                                                
    70 percent    confident                                             
reduction at      or                                                 
    nearly each   confident            0                      5    7    
power planta                                                         
                    Less               1                      4    5    
                  confident                                             
                 Not at all                                             
                  confident           13                      1    1    
                 Do not know           0                      1    0    
       Total                                                11b   13b   
    respondents                       14                                

Source: GAO.

aThe survey asked stakeholders to consider the likelihood that a single
power plant could reduce mercury emissions 70 percent by 2008. In this
case, respondents were asked to consider whether most, but not necessarily
all, power plants in the United States would each be capable of achieving
a 70 percent reduction in mercury emissions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Appendix V Stakeholder Confidence in Ability of Technologies to Achieve Mercury
                        Reductions under Three Scenarios

  Table 8: Stakeholder Confidence in Reducing Nationwide Mercury Emissions 90
                       Percent by 2008 Stakeholder group

     Scale of   Confidence     Power     Researchers/government         
     mercury                  industry                                  
    reduction      level    respondents               officials Vendors Total 
                   Very                                                 
    90 percent   confident                                              
    reduction       or                                                  
nationwidea   confident             0                      2       2 
                   Less                1                      2       6 
                 confident                                              
                Not at all                                              
                 confident            13                      6       5 
                Do not know            0                      1       0 
      Total                                                 11b   13b   
respondents                        14                                

Source: GAO.

aGAO asked stakeholders how confident they were that power plants could
reduce mercury emissions 90 percent by 2008. In this case, respondents
were asked to consider reductions averaged across power plants in the
United States, which does not mean that each individual plant would
achieve the reductions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Table 9: Stakeholder Confidence in Achieving Mercury Reductions of 90 Percent at
                  Nearly Every Plant by 2008 Stakeholder group

     Scale of    Confidence     Power    Researchers/government         
      mercury                 industry                                  
     reduction      level    respondents              officials Vendors Total 
                    Very                                                
    90 percent    confident                                             
reduction at      or                                                 
    nearly each   confident            0                      2       2 
power planta                                                         
                    Less               1                      2       6 
                  confident                                             
                 Not at all                                             
                  confident           13                      6       4 
                 Do not know           0                      1       1 
       Total                                                11b   13b   
    respondents                       14                                

Source: GAO.

aThe survey asked stakeholders to consider the likelihood that a single
power plant could reduce mercury emissions 90 percent by 2008. In this
case, respondents were asked to consider whether most, but not necessarily
all, power plants in the United States would each be capable of achieving
a 90 percent reduction in mercury emissions.

bOne stakeholder in this group that responded to the survey did not answer
this question.

Appendix VI

Sorbent Injection Cost Estimates from EPA and DOE

This appendix summarizes estimates of the cost of activated carbon
injection reported by EPA and DOE in October and November 2003.1

Environmental Protection Agency. Using modeling data provided in EPA's
cost report, we selected control cost scenarios that are comparable to
those DOE presented in its cost study.2 These estimates include the cost
of fly ash disposal for plants that use sorbent injection without a fabric
filter, based on the assumption that the presence of sorbent in fly ash
makes it unsuitable for sale. EPA provided capital costs in dollars per
unit of generating capacity, and operating and maintenance costs in
dollars per unit of electricity generated (per hour) for 100-and
975-megawatt plants operating at 65 percent capacity over the course of a
year (5,694 hours). Tables 10 and 11 present the range of capital and
operating and maintenance costs for the selected EPA plant scenarios;
capital costs are in total dollars while operating and maintenance costs
are expressed in dollars per year.

1Environmental Protection Agency, Office of Research and Development,
National Risk Management Research Laboratory, Performance and Cost of
Mercury and Multipollutant Emission Control Technology Applications on
Electric Utility Boilers (Research Triangle Park, N.C., 2003).

Jeff Hoffmann and Jay Ratafia-Brown, Science Applications International
Corporation,

Preliminary Cost Estimate of Activated Carbon Injection for Controlling
Mercury Emissions from an Un-Scrubbed 500 MW Coal-Fired Power Plant, a
report prepared for the Department of Energy, National Energy Technology
Laboratory, November 2003.

2According to the EPA study, the agency identified a representative range
of plant configurations, coal types, and technologies. In developing the
range, EPA used 49 model plants. For the estimates presented here, we
selected 4 model plants, which were 100megawatt and 975-megawatt plants
with an existing electrostatic precipitator, burning lowsulfur bituminous
or subbituminous coals with and without a fabric filter installed, with
desired mercury removal levels between 50 and 90 percent, depending on
configuration and coal type. These model plants most closely align with
the assumptions presented in the DOE cost estimates discussed in this
report.

                                  Appendix VI
                   Sorbent Injection Cost Estimates from EPA
                                    and DOE

  Table 10: Select EPA Cost Estimates of Sorbent Injection for a 100-Megawatt
                          Coal-Fired Power Plant, 2003

                           Thousands of 2003 dollars

       Cost        Low         High    Low-end        High-end assumptions    
                 estimate  estimate  assumptions   
                   $16.5b             50 percent   90 percent mercury removal 
      Capital              $5,947.9    mercury                from            
                                     removal from    sorbent injection and    
                                     bituminous-         fabric filter        
                                      fired unit         retrofit, as well as 
                                    with existing          mercury monitoring 
                                      equipment    for a subbituminous-fired  
                                     only; costs              unit            
                                    include        
                                    mercury        
                                    monitoring     
      Annual         0.6b             50 percent   90 percent mercury removal 
     operating              1,342.6    mercury                from            
        and                          removal with   sorbent injection without 
maintenancea                        existing               a fabric filter 
                                    equipment                                 
                                    only; no       and mercury monitoring for
                                    sorbent        
                                      injection      bituminous-fired unit    
                                        needed     

Source: GAO analysis of EPA data.

aBased on a plant capacity factor of 65 percent, includes both variable
and fixed operating and maintenance costs.

bThis reduction is assumed to be met with existing equipment; therefore
costs are for mercury monitoring only, no sorbent injection.

  Table 11: Select EPA Cost Estimates of Sorbent Injection for a 975-Megawatt
                          Coal-Fired Power Plant, 2003

                           Thousands of 2003 dollars

       Cost        Low          High    Low-end       High-end assumptions    
                 estimate   estimate  assumptions  
                   $91.7b             50 percent   90 percent mercury removal 
      Capital              $36,210.5    mercury               from            
                                     removal from    sorbent injection and    
                                      bituminous-        fabric filter        
                                      fired unit         retrofit, as well as 
                                     with existing         mercury monitoring 
                                       equipment   for a subbituminous-fired  
                                      only; costs             unit            
                                     include       
                                     mercury       
                                     monitoring    
      Annual         5.6b             50 percent   90 percent mercury removal 
     operating              12,868.7    mercury               from            
        and                          removal with   sorbent injection without 
maintenancea                        existing               a fabric filter 
                                     equipment                                
                                     only; no      and mercury monitoring for
                                     sorbent       
                                       injection     bituminous-fired unit    
                                        needed     

Source: GAO Analysis of EPA data.

aBased on a plant capacity factor of 65 percent, includes both variable
and fixed operating and maintenance costs.

bThis reduction is assumed to be met with existing equipment; therefore
costs are for mercury monitoring only, no sorbent injection.

EPA estimated that the capital cost of sorbent injection for a
100-megawatt plant would range from $0.17 to $59.5 per kilowatt of
capacity, while operating and maintenance costs for the same plant would
range from $0.001 to $2.36 per megawatt-hour. For the 975-megawatt plant,
EPA

Appendix VI
Sorbent Injection Cost Estimates from EPA
and DOE

estimated that the capital cost would range from $0.09 to $37.1 per
kilowatt, while operating and maintenance costs would range from $0.001 to
$2.32 per megawatt-hour. EPA also estimated the total annualized cost of
these controls in 2003 dollars, which ranged from $0.005 to $2.64 per
megawatt-hour or between $2,847 and $1.5 million per year for a
100megawatt plant.3 For a 975-megawatt plant, annualized costs ranged from
$0.003 to $2.45 per megawatt-hour or between $16,655 and $13.6 million per
year.

Capital costs were much higher for scenarios where a fabric filter was
installed, while the highest operating and maintenance cost and annualized
cost were for achieving a 90 percent mercury reduction for a bituminous
coal-fired plant using sorbent injection without installing a fabric
filter, due to the amount of sorbent needed to achieve a high mercury
removal. At the low end of these costs, EPA assumed that existing
equipment is sufficient to achieve a 50 percent reduction in mercury for
plants that burn bituminous coal, therefore costs reflect only that of
monitoring mercury emissions and do not include actual sorbent injection
costs. While total capital and annual costs for the larger plant were
higher than for the smaller plant, the annualized cost in dollars per
megawatt-hour was actually lower, since costs were spread out over more
units of capacity and electricity generated.

Department of Energy. DOE's analysis of the cost of mercury control
technologies was based on field testing conducted by DOE's National Energy
Technology Laboratory. For its estimates, DOE used a hypothetical power
plant of 500 megawatts burning bituminous or subbituminous coal and
equipped with an electrostatic precipitator or a layout that consists of
sorbent injection and a fabric filter retrofitted downstream of an
existing electrostatic precipitator. Cost estimates were developed for
mercury removal requirements ranging from 50 to 90 percent as shown below
in table 12. DOE estimated capital costs between $1.97 and $57.44 per
kilowatt. The high end of the capital cost range represented cases where
facilities installed a supplemental fabric filter to achieve higher levels
of mercury reduction, while the high end of the operating and maintenance
costs represented achieving a 90 percent reduction in mercury emissions

3EPA's annualized cost reflects the capital cost annuitized using a
levelized carrying charge rate of 13.3 percent assuming a 30-year
operating period summed with operating and maintenance costs levelized
with a factor of 1.0.

                                  Appendix VI
                   Sorbent Injection Cost Estimates from EPA
                                    and DOE

for a plant burning bituminous coal using sorbent injection without a
fabric filter.

Table 12: DOE's Cost Estimates for Sorbent Injection Installed on a 500-Megawatt
Coal Power Plant, 2003 Thousands of 2003 dollars Cost Low estimate High estimate
                    Low-end assumptions High-end assumptions

Capital $984.0 $28,719.0 50 or 70 percent mercury 60 or 90 percent mercury 
                            removal from bituminous-   removal with sorbent   
                              fired unit, 50 or 60   injection and fabric     
                                                     filter                   
                            percent mercury removal     installation for a    
                            from subbituminous-fired subbituminous-fired unit 
                            unit with sorbent        
                            injection                
                             and existing equipment  
                               (no fabric filter)    

       First year      931.0 15,647.0 50 percent mercury  90 percent mercury  
      operating and                                       
       maintenance                       removal with     removal with        
                                            sorbent       sorbent             
                                         injection and    injection and       
                                           existing       existing            
                                         equipment (no    equipment (no       
                                            fabric        fabric              
                                      filter) from               filter) from 
                                      bituminous-                 bituminous- 
                                          fired unit          fired unit      

Source: GAO analysis of DOE data.

DOE also provided two sets of annualized cost estimates, one that included
a projected impact for the loss of fly ash sales and one that did not.
Without a by-product impact, DOE estimated annualized costs to range from
$0.37 to $5.72 per megawatt-hour, which equates to about $1.3 million to
$20.0 million per year. Estimates with the by-product impact ranged from
$1.82 to $8.14 per megawatt-hour, which equates to about $6.4 million to
$28.5 million per year. At the high end, these estimates represented the
cost of achieving a 90 percent mercury reduction at a bituminous-coal
fired plant with sorbent injection, an existing electrostatic
precipitator, and no fabric filter. The low-end cost without a by-product
impact represented a 50 percent mercury reduction at a bituminous-fired
plant using sorbent injection with an electrostatic precipitator, while
the low-end cost with a by-product impact was for the same configuration
and mercury reduction, but at a subbituminous-fired plant.

In addition, DOE's cost estimates suggest that plants may achieve a high
level of mercury control without a fabric filter. While achieving a higher
mercury removal rate without a fabric filter would require more sorbent,
plants can decide what air pollution control configuration is most cost
effective. Furthermore, according to EPA, test results suggest that

Appendix VI
Sorbent Injection Cost Estimates from EPA
and DOE

chemically enhanced sorbent may prove more efficient than activated carbon
in achieving high levels of mercury removal at relatively modest injection
rates, and thus less expensive to use. According to EPA, tests of these
sorbents have achieved mercury removal rates of 40 to 94 percent without a
fabric filter, with the highest removal rate achieved during a continuous
30-day test (the longest reported test of these sorbents). Therefore, some
facilities seeking to achieve high levels of mercury reduction may not
have to incur the substantial cost of adding a fabric filter.

Appendix VII

                     GAO Contact and Staff Acknowledgments

GAO Contact John B. Stephenson (202) 512-3841

Acknowledgments	In addition to the contact named above, Kate Cardamone,
Christine B. Fishkin, Tim Guinane, Michael Hix, Andrew Huddleston, Judy
Pagano, and Janice Poling made key contributions to this report. Nabajyoti
Barkakati, Cindy Gilbert, Jon Ludwigson, Stuart Kaufman, Cynthia Norris,
Katherine Raheb, Keith Rhodes, and Amy Webbink also made important
contributions.

GAO's Mission	The Government Accountability Office, the audit, evaluation
and investigative arm of Congress, exists to support Congress in meeting
its constitutional responsibilities and to help improve the performance
and accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone	The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone:	Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

Congressional	Gloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public Affairs	Paul Anderson, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548
*** End of document. ***