Acid Rain Program; Nitrogen Oxides Emission Reduction Program
Federal RegisterDec 19, 1996
Ask Donna
What actually matters in this document.
Text
SUMMARY: This action promulgates standards for the second phase of the
Nitrogen Oxides Reduction Program under Title IV of the Clean Air Act
(``CAA'' or ``the Act'') by establishing nitrogen oxides (NOX)
emission limitations for certain coal-fired electric utility units and
revising NOX emission limitations for others as specified in
section 407(b)(2) of the Act. The emission limitations will reduce the
serious adverse effects of NOX emissions on human health,
visibility, ecosystems, and materials.
EFFECTIVE DATE: December 19, 1996.
ADDRESSES: Docket. Docket No. A-95-28, containing information
considered during development of the promulgated standards, is
available for public inspection and copying between 8:30 a.m. and 3:30
p.m., Monday through Friday, at EPA's Air Docket Section (LE-131),
Waterside Mall, Room M1500, 1st Floor, 401 M Street, SW, Washington, DC
20460. A reasonable fee may be charged for copying.
Background information document. The background information
document containing responses to public comments on the proposed
standards may be obtained from the docket. Please refer to ``Phase II
Nitrogen Oxides Emission Reduction Program--Response to Comments
Document''.
FOR FURTHER INFORMATION CONTACT: Peter Tsirigotis, Source Assessment
Branch, Acid Rain Division (6204J), U.S. Environmental Protection
Agency, 401 M Street S.W., Washington, DC 20460 (202-233-9620).
SUPPLEMENTARY INFORMATION:
Regulated Entities
Entities regulated by this action are electric service providers
that run or operate coal-fired electric utility boilers including dry
bottom wall-fired and tangentially fired boilers (Group 1) and certain
other boiler types including boilers applying cell-burner technology,
cyclone boilers, wet bottom boilers, and other types of coal-fired
boilers (Group 2). Regulated entities and boilers include:
------------------------------------------------------------------------
Regulated Entities Regulated Boilers
------------------------------------------------------------------------
Electric Service Providers................ Dry bottom wall-fired.
Tangentially fired.
Cell Burners.
Cyclones (larger than 155
MWe).
Vertically fired.
Wet bottoms (larger than 65
MWe).
------------------------------------------------------------------------
This table is not intended to represent a definitive enumeration of
all existing and future entities regulated by this action. Rather, its
intent is to provide a general guide for readers and to list entities
that EPA is now aware will be regulated by this action. Other types of
entities not listed in the table could also be regulated. To determine
whether your (facility, company, business, organization, etc.) is
regulated by this action, you should carefully examine the
applicability criteria in Secs. 72.6 and 76.1 of title 40 of the Code
of Federal Regulations. If you have questions regarding the
applicability of this action to a particular entity, consult the person
named in the preceding ``For Further Information Contact'' section.
The information in this preamble is organized as follows:
I. Rule Background
A. Purpose of Acid Rain NOX Emission Reduction Program
B. Summary of Final Rule
1. NOX Standards Promulgated by this Rule
2. Rationale for Revising Group 1 NOX Emission Limits and
Environmental Impact of Group 2 NOX Emission Limits
II. Public Participation
III. Summary of Major Comments and Responses
A. Phase II, Group 1 Boiler NOX Emission Limits
1. Boiler Population Used to Assess NOX Emission Limits
2. Time Period/Averaging Basis Used to Evaluate Performance of
Low NOX Burner Technology
3. Analysis Method Used to Establish Reasonably Achievable
Emission Limitations for Phase II, Group 1 Boilers
4. Percentile Used to Define Achievability
B. Group 2 Boiler NOX Emission Limits
1. Cost Comparability and Its Basis
2. Cost Comparison Methodology
3. Retrofit Nature of Group 2 Controls
4. Group 2 Boiler Size Exemption
5. Cyclone Boiler NOX Controls
6. Wet Bottom Boiler NOX Controls
7. Vertically Fired Boiler NOX Controls
8. Cell Burner Boiler NOX Controls
9. Revision of Proposed Group 2 Boiler NOX Emission Limits
C. Compliance Issues
D. Title IV NOX Program's Relationship to Title I and
NOX Trading Issues
IV. Administrative Requirements
A. Docket
B. Executive Order 12866
C. Unfunded Mandates Act
D. Paperwork Reduction Act
E. Regulatory Flexibility Act
F. Submission to Congress and the General Accounting Office
G. Miscellaneous
I. Rule Background
A. Purpose of Acid Rain NOX Emission Reduction Program
The primary purpose of the Acid Rain NOX Emission Reduction
Program is to reduce the multiple adverse effects of the oxides of
nitrogen, a family of highly reactive gaseous compounds that contribute
to air and water pollution, by substantially reducing annual emissions
from coal-fired power plants. Since the 1970 passage of the Clean Air
Act, NOX has increased about 7%; it is the only conventional air
pollutant to show an increase nationwide.
Electric utilities are a major contributor to NOX emissions
nationwide: in 1980, they accounted for 30 percent of total NOX
emissions and, from 1980 to 1990, their contribution rose to 32 percent
of total NOX emissions. In 1994, electric utility emissions
represented about 33 percent of the total annual NOX emissions.
Approximately 90 percent of estimated electric utility NOX
emissions were attributed to coal combustion (see docket item IV-A-8
(USEPA, National Air Pollution Emission Trends, 1900-1994 (EPA-454/R-
95-011) at 2-2, October 1995)).
The NOX emissions discharged into the atmosphere from the
burning of fossil fuels consists primarily of nitric oxide (NO). Much
of the NO, however, reacts with organic radicals in the air to form
nitrogen dioxide (NO2) and, over longer periods of time, reacts
with and forms other pollutants, including ozone (O3), nitric acid
(HNO3) and fine particles. These pollutants are harmful to public
health and the environment.
NO2 and airborne nitrate also degrade visibility, and when
they return to the earth through rain, snow, or fog (``wet
deposition'') or as gases (``dry deposition''), they contribute to
acidification of lakes and streams and to excessive nitrogen loadings
to estuaries and coastal water systems such as in the Chesapeake Bay
(``eutrophication'').
NO2 has been documented to cause eye irritation, either by
itself or when oxidized photochemically into peroxyacetyl nitrate
(PAN). Ozone, the most abundant of the photochemical oxidants, is a
highly reactive chemical compound which can have serious adverse
effects on human health, plants, animals, and materials. Fine particles
at current ambient levels contribute adversely to morbidity and
mortality.
[[Page 67113]]
B. Summary of Final Rule
1. NOX Standards Promulgated by This Rule
EPA today is promulgating new emission limitations to be
implemented for nitrogen oxides (NOX) emissions for wall-fired and
tangentially fired boilers (Group 1 boilers) and establishing emission
limitations for certain other boilers (Group 2 boilers). The final rule
implements section 407 (b)(2) of the Act, which applies to NOX
emission limitations for Group 1 and Group 2 boilers during Phase II of
the Acid Rain Program (January 1, 2000 and beyond). Under section
407(b)(2) the Administrator ``may revise'' the applicable NOX
emission limitations for Group 1 boilers in Phase II if the
Administrator determines that ``more effective low NOX burner
technology is available,'' i.e., that data on the effectiveness of low
NOX burner technology (LNB) installed after passage of the Clean
Air Act Amendments of 1990 supports emission limitations more stringent
than the limitations established for Group 1 boilers during Phase I of
the Acid Rain Program pursuant to section 407(b)(1) of the Act. 42
U.S.C. 7651f(b)(2). Also under section 407(b)(2) of the Act, the
Administrator must establish NOX emission limitations (on a lb/
mmBtu annual average basis) for Group 2 boilers, which include wet
bottom boilers, cyclone boilers, cell burner boilers, and all other
types of utility boilers not classified as dry bottom wall-fired and
tangentially fired boilers, and must meet certain requirements in
establishing these limitations. In setting the final emission
limitations for Group 1 and Group 2 boilers, as summarized below, the
Administrator has met the requirements in section 407(b)(2) of the Act.
i. Revision of NOX Emission Limits for Phase II, Group 1 Boilers
The Agency has developed a computerized database containing
detailed information on the characteristics and emission rates of all
coal-fired units with Group 1 boilers on which low NOX burners
(LNBs) have been installed without any other NOX controls, and for
which EPA has both quality assured long-term post-retrofit hourly
NOX emission rate data, measured by continuous emission monitoring
systems (CEMS), certified pursuant to 40 CFR part 75 (Acid Rain
Continuous Emission Monitoring Rule), and quality assured short-term
CEM or test data measurements of uncontrolled emission rates. This
database, called the ``LNB Application Database,'' consists of 39 dry
bottom wall-fired boilers and 14 tangentially fired boilers and forms
the technical basis for EPA's evaluation of the effectiveness (percent
NOX removal) of LNBs applied to Group 1 boilers.
For the final rule, EPA has adopted a methodology that employs
``load-weighted annual average NOX emission rates'' over the full
``post-optimization period'' for evaluating the effectiveness of LNBs.
The post-optimization period includes all available data beginning with
the first hour of the low NOX period,1 when the LNBs were
operating under optimized NOX removal conditions, and extending to
the end of the entire data set, i.e., through June 30, 1996, the end of
the latest available reporting period from the Acid Rain Emissions
Tracking System (ETS). The post-optimization period contains quality
assured CEM data spanning at least 4 calendar months for every boiler
and at least 11 calendar months for most boilers (83%). In addition,
EPA applied a NOX/load weighting scheme, using hourly load data
reported for 1995, to develop ``load-weighted'' annual average NOX
emission rates from the data set (see discussion in section III.A.2.iii
of this preamble). Two advantages of using load-weighted annual average
NOX emission rates over the post-optimization period are that the
criteria used to define the ``post-optimization period'' take into
account the site-specific nature of the LNB equipment optimization and
operator training processes while the use of ``load weighting''
accounts for any potential impact of annual load dispatch patterns on
NOX emissions.
---------------------------------------------------------------------------
\1\ The ``low NOX period'' EPA used for assessing
performance of LNBs applied to Group 1 boilers was defined by
identifying the lowest average NOX emission rate each boiler
has sustained for at least 52 days, i.e., over a period of 1,248
hours when the boiler was operating and valid CEM data, measured by
CEMS certified pursuant to 40 CFR part 75, were available. (Data for
30 calendar days following estimated date boiler began operating
after shutdown for LNB retrofit are not used when making this
determination. See Table 1, DQO #4D).
---------------------------------------------------------------------------
Following the identification of appropriate LNB applications and
time period for analysis, EPA developed a two-part model to estimate:
(1) Annual average emission rates that can be sustained by LNBs
installed on Phase II units with Group 1 boilers and (2) percentile
distributions of Phase II units that can comply with various
performance standards. The first part of the model calculates the
percent reduction achievable by LNBs as a function of uncontrolled
emission rate, and the second part applies the estimated percent
reduction to boiler-specific uncontrolled emission rates for the
population of units that will be subject to any revised NOX
emission limitations in Phase II. EPA used the percentile distributions
to select reasonably achievable emission limits for the two types of
Group 1 boilers, where ``reasonably achievable'' is defined as the
controlled emission rate 85 to 90 percent of the affected population of
units can meet or exceed on an annual average basis.
EPA concludes that more effective low NOX burner technology
is available for dry bottom wall-fired and tangentially fired
boilers. Further, EPA concludes that for dry bottom wall-fired
boilers, 0.46 lb/mmBtu is a reasonable emission limitation that is
achievable using such technology. EPA estimates that 85 to 90% of
the Phase II dry bottom wall-fired boilers can achieve this emission
rate. The implementation of this standard, will result in an
additional NOX emissions reduction of approximately 90,000 tons
per year, beginning in 2000, below the emission levels anticipated
under the Phase I Acid Rain NOX Emission Reduction Rule (60 FR
18751, April 13, 1995).
Finally, EPA concludes that for tangentially fired boilers, 0.40
lb/mmBtu is a reasonable emission limitation that is achievable
using such technology. EPA estimates that 85 to 90% of the Phase II
tangentially fired boilers can achieve this emission rate. The
implementation of this standard will result in an additional
NOX emissions reduction of approximately 30,000 tons per year,
beginning in 2000, below the emission levels anticipated under the
Phase I Acid Rain NOX Emission Reduction Rule. As discussed
below, EPA exercises its discretion under section 407(b)(1) to adopt
these revised Group 1 NOX emission limitations because the
resulting additional reductions are a reasonable step toward
achieving necessary, significant NOX reductions and are
consistent with the guideline in section 401(b) concerning the level
of NOX reductions to be achieved.
ii. Establishment of Group 2 Emission Limitations
In order to meet the requirements of section 407(b)(2), EPA is
using the following methodology for establishing Group 2 emission
limitations:
First, EPA determines what NOX control technologies are the
best systems of continuous emission reduction available for each
category of Group 2 boilers. Further, EPA considers only technologies
for which there is reliable cost information on which to base a
determination of whether they are of comparable cost to LNBs, applied
to Group 1 boilers.
Second, EPA evaluates each such NOX control technology and
estimates the dollar cost per ton of NOX removed using the control
technology on each boiler in the Group 2 population that is in the
appropriate Group 2 boiler category. EPA then compares the dollar cost
per ton of NOX removed for each
[[Page 67114]]
NOX control technology applied to the Group 2 boiler category to
the dollar cost per ton of NOX removed for low NOX burners
applied to dry bottom wall-fired and tangentially fired boilers. Based
on this comparison, EPA determines whether the NOX control
technology applied to the Group 2 boiler category has a cost-
effectiveness comparable to that of LNBs applied to Group 1 boilers.
Third, EPA estimates the percent change in electricity rates for
consumers resulting from costs (in mills per kilowatt-hour) associated
with the application of emission limitations on Group 2 boilers. This
value is then compared to the percent change in nationwide electricity
rates due to the establishment of emission limitations for LNBs on
Group 1 boilers. EPA also estimates the emission reductions that are
likely to be achieved and considers any other environmental impacts
likely to result from application of each NOX control technology.
Fourth, EPA assesses the performance (percent NOX reduction)
of each cost-comparable Group 2 control technology and applies that
reduction percentage to data on the uncontrolled emissions of each
boiler that is in the particular category of Group 2 boilers and that
will be subject to the Group 2 emission limitation. The emission
limitation that will be achievable by 85 to 90% of the boiler
population is generally selected, after taking account of energy and
environmental impacts, as the emission limitation for that category of
Group 2 boiler.
EPA concludes that for cell-burner fired boilers, 0.68 lb/mmBtu is
a reasonable emission limitation that meets the requirements of section
407(b)(2). For cell burner boilers, plug-in retrofits and non-plug in
retrofits are the best continuous control systems that are available
and meet the cost comparability requirement. EPA bases the emission
limitation on the use of these control technologies and estimates that
80% of the cell burner population can achieve the limitation. The
energy impact, i.e., impact of mills/kWh cost on electricity consumers,
of using these technologies to meet the emission limitation is small
and similar in magnitude to the energy impact of using LNBs on Group 1
boilers. The emission limitation will result in a total NOX
emissions reduction of approximately 420,000 tons per year, beginning
in 2000, without significant increases in other air pollutants or solid
waste. As discussed below, the resulting NOX reductions are a
reasonable step toward achieving necessary, significant NOX
reductions and are consistent with section 401(b).
EPA concludes that for cyclone fired boilers larger than 155 MWe,
0.86 lb/mmBtu is a reasonable emission limitation that meets the
requirements of section 407(b)(2). For cyclone fired boilers, gas
reburning, and SCR are the best continuous control systems that are
available and meet the cost comparability criteria. The energy impact,
i.e., impact of mills/kWh cost on electricity consumers, of using these
technologies to meet the emission limitation is small and similar in
magnitude to the energy impact of using LNBs on Group 1 boilers. EPA
bases the emission limitation on the use of these technologies and
estimates that 85 to 90% of the cyclone fired boiler population can
achieve the emission limitation. The emission limit will result in a
total NOX emissions reduction of approximately 225,000 tons per
year, beginning in 2000, without significant increases in other air
pollutants or solid waste. As discussed below, the resulting NOX
reductions are a reasonable step toward achieving necessary,
significant NOX reductions and are consistent with section 401(b).
EPA has decided not to set a NOX emission limitation for cyclone
boilers of 155 MWe or less.
EPA concludes that for wet bottom boilers larger than 65 MWe, 0.84
lb/mmBtu is a reasonable emission limitation that meets the
requirements of section 407(b)(2). For wet bottom boilers, gas
reburning, and SCR are the best continuous control systems that are
available and meet the cost comparability requirement. EPA bases the
emission limitation on the use of these technologies and estimates that
85 to 90% of the wet bottom boiler population can achieve the emission
limitation. The energy impact, i.e., impact of mills/kWh cost on
electricity consumers, of using these technologies to meet the emission
limitation is small and similar in magnitude to the energy impact of
using LNBs on Group 1 boilers. The emission limitation will result in a
total NOX emissions reduction of approximately 80,000 tons per
year, beginning in 2000, without significant increases in other air
pollutants or solid waste. As discussed below, the resulting NOX
reductions are a reasonable step toward achieving necessary,
significant NOX reductions and are consistent with section 401(b).
EPA has decided not to set a NOX emission limitation for wet
bottom boilers of 65 MWe or less.
EPA concludes that for vertically fired boilers 0.80 lb/mmBtu is a
reasonable emission limitation that meets the requirements of section
407(b)(2). For vertically fired boilers, combustion controls are the
best continuous control system available and meet the cost
comparability requirement. EPA bases the emission limitation on the use
of these technologies and estimates that 85 to 90% of the vertically
fired boiler population can achieve this emission limitation. The
energy impact, i.e., impact of mills/kWh cost on electricity consumers,
of using these technologies to meet the emission limitation is small
and similar in magnitude to the energy impact of using LNBs on Group 1
boilers. The emission limitation will result in a total NOX
emissions reduction of approximately 45,000 tons per year, beginning in
2000, without significant increases in other air pollutants or solid
waste. As discussed below, the resulting NOX reductions are a
reasonable step toward achieving necessary, significant NOX
reductions and are consistent with section 401(b). EPA has decided not
to set a NOX emission limitation for arch-fired boilers, a subset
of the vertically fired boiler category.
Finally, EPA has decided not to set a NOX emission limitation
for FBC boilers. Because these units are already low NOX emitters
by design, the NOX emissions reduction achieved by installing any
additional control technology, would not meet the cost-comparability
requirement of section 407(b)(2). Moreover, setting an emission
limitation that can be achieved by every existing FBC boiler without
installing any additional control technology would have an adverse
environmental impact. Some existing boilers emit at rates considerably
below the highest annual rate observed among FBC boilers and these
boilers could offset the emission reductions otherwise required of
other affected boilers through emissions averaging under Sec. 76.10.
EPA has also decided not to set a NOX emission limitation for
stoker boilers. EPA has not found any continuous control technology for
stoker boilers that meets the cost-comparability requirement.
2. Rationale for Revising Group 1 NOX Emission Limits and
Environmental Impact of Group 2 NOX Emission Limits
EPA is exercising its discretion to revise the Phase II, Group 1
NOX emission limitations because: (1) NOX emissions have
significant adverse effects on human health and the environment; (2)
significant, additional regional NOX reductions from current
levels are likely to be necessary; (3) without additional actions
NOX emissions are projected to increase
[[Page 67115]]
nationwide starting in 2002; (4) the revision of Phase II, Group 1
emission limitations is one of the most cost-effective means of
achieving additional NOX reductions; and (5) the additional
reductions from the revision represent a reasonable step toward
achieving necessary NOX reductions. In addition, the resulting
NOX reductions are consistent with section 401(b). The adverse
health and environmental effects of NOX emissions are discussed in
the proposed rule on Phase II NOX emission limitations. 61 FR
1442, 1453-55, January 19, 1996. EPA reaffirms that discussion, which
summarizes the adverse impact of NOX emissions through: The
formation of ozone, particulate matter, and nitrogen oxides; and
atmospheric deposition resulting in eutrophication of water bodies and
acidification of lakes and streams. For the same reasons, EPA also
concludes that the adoption of the Group 2 emission limitations set
forth in today's rule is supported by the environmental impact of the
emission reductions that will result.
The contribution of nitrogen oxides to the formation of ozone, acid
deposition and eutrophication of water bodies is substantial.
Consequently, in order to address these problems, significant NOX
emission reductions are likely to be needed on a regional scale,
particularly in the eastern half of the U.S. This is the portion of the
nation in which most of the boilers subject to NOX emission
limitations under the Acid Rain Program are located; 87% of Phase II,
Group 1 boilers and 89% of Group 2 boilers covered by today's final
rule are in the eastern U.S.
i. Ozone
With regard to ozone, additional regional NOX reductions of at
least 50% from current levels are likely to be needed over large
portions of the nation to attain and maintain the national ambient air
quality standard for ozone. Modeling results using EPA's Regional
Oxidant Model (ROM) estimated that NOX reductions of about 75%
will be needed over large portions of the nation to reduce ozone
concentrations to levels at or below the NAAQS (see docket item IV-J-8
(EXISTMOD.TXT, OTAG Modeling and Assessment Subgroup Files on EPA's TTN
Bulletin Board, February 7, 1996)). The ROM modeling results were among
the reasons for the formation of the Ozone Transport Assessment Group
(OTAG), comprised of the 37 eastern-most States and tasked with
developing a consensus approach for reducing regional NOX
emissions. OTAG recently completed atmospheric modeling simulations
using SAI's Urban Airshed Model (UAM-V) (see docket item IV-J-21 (OTAG
Air Quality Analysis Workgroup, 1996)). The results indicate that:
broad NOX emission reductions will decrease regional ozone, high
ozone, and ozone in non-attainment areas; and NOX emission
reductions in each OTAG sub-region will be needed to both lower ozone
in that same sub-region, as well as other sub-regions.
Further, necessary NOX reductions to achieve or maintain the
ozone standard have been estimated for several other areas of the
country: 50-75% from 1990 levels throughout the Northeast Ozone
Transport Region (OTR) (60 FR 4712, 4722, January 24, 1995); up to 90%
reductions in the Southeast (see docket item II-I-98 (State of the
Southern Oxidants Study, 1995)); and a combination of 75% reductions
for NOX and 25% for VOCs regionally, combined with 25% for
NOX and 75% for VOCs locally in the New York region (60 FR 4721);
and significant NOX reductions in the Lake Michigan area, not yet
quantified. The results of a study analyzing ozone non-attainment in
the eastern U.S. found that nationwide NOX emission reductions of
about 50% from 1990 levels will be needed to approach achievement of
the necessary ozone standards (see docket item IV-J-9 (Rao, S.T.,
et.al., Dealing with the Ozone Non-Attainment Problem in the Eastern
United States, AWMA journal, January 1996)).
ii. Acid Deposition
Similarly, additional, regional NOX reductions of at least 40%
are likely to be necessary in order to mitigate the effects of acid
deposition. In particular, it is estimated that between 40-50%
reductions of NOX in the Eastern U.S. beyond those already
required in the Clean Air Act may be necessary simply to keep the
number of acidified lakes in the Adirondacks in New York at 1984
levels. (See docket item IV-A-6 (Acid Deposition Standard Feasibility
Study (EPA 430-R-95-001a) at xvi).) Without additional reductions, the
number of acidic lakes in the Adirondacks are projected to increase by
almost 40% by 2040. Id. at 47. Significant, additional reductions may
also be necessary with regard to the Mid-Appalachian region (see docket
item IV-A-6 (Acid Deposition Standard Feasibility Study at xvi)).
iii. Eutrophication
NOX emissions also contribute significantly to eutrophication,
i.e., an overabundance of nitrogen to water bodies that leads to
problems of nutrient enrichment. Regional NOX emission reductions
of up to 40% are likely to be needed. The signatories to the Chesapeake
Bay Agreement, (Maryland, Pennsylvania, Virginia, the District of
Columbia, the Chesapeake Bay Commission, and the federal government)
have agreed on a goal of a 40% reduction in nitrogen loadings to the
Bay by 2000 (relative to a 1985 baseline), representing a reduction of
34 million kilograms of nitrogen (see docket item IV-J-11 (Hicks et
al., 1995:6)). In addition, they agreed to maintain, after 2000, a cap
on nitrogen loadings at 60% of baseline loadings. Present estimates are
that approximately 27% of total nitrogen loading to the Bay system
comes from atmospheric sources in the form of NOX emissions (see
docket items IV-J-26 (Linker et al., 1993) and IV-J-19 (Valigura et
al., 1995)). Since reducing nitrogen loading through the control of
NOX emissions can be as cost-effective as controlling non-
atmospheric sources of nitrogen loading (e.g., point sources such as
waste water treatment and non-point sources such as farms), up to a 40%
reduction of the contribution in NOX emissions to the Bay in areas
contributing to the eutrophication of the Bay is likely to be
necessary.
Although the watershed of the Chesapeake Bay encompasses
approximately 64,000 square miles, the Chesapeake Bay ``airshed,''
which is the contiguous area providing 70% of the atmospheric
deposition loads to the watershed (see docket item IV-J-18 (Dennis,
1996)), covers up to 600,000 square miles in area (see docket item IV-
J-3 (Valigura et al., 1996:23)). The airshed extends upwind of, as well
as bordering the water body itself: south to South Carolina, north to
Ontario, Canada, and westward up to 500 miles (see docket item IV-J-11
(Hicks et al., 1995:6)). NOX emissions from outside this area not
only contribute to eutrophication in the Bay but also to the entire
coastline, such as from the Carolinas to New York (see docket item IV-
J-3 (Valigura et al., 1996:23)).
iv. Utility Contribution to Atmospheric NOX Emissions
Electric utilities contributed approximately 33% of total
atmospheric NOX emissions in 1994, thus substantially contributing
to ozone formation, acid deposition, and eutrophication.
Table 1 summarizes the reductions in atmospheric NOX emissions
likely needed and the additional reductions provided by today's final
rule. Although the additional reductions from coal-fired utility
boilers under the final rule are substantial, they represent only
[[Page 67116]]
about 5% of all atmospheric NOX emissions from all sources of
NOX emissions. The additional reductions under the final rule
represent about a 15% reduction in total utility emissions. Since
utilities presently contribute about 33% of total NOX emissions,
the final rule provides reductions of about 5% of total NOX
emissions. This reduction level is significantly less than the
reduction level likely to be needed to mitigate ozone, acid deposition,
and eutrophication (see docket item IV-A-8 (EPA, ``National Air
Pollution Emission Trends, 1900-1994'' at 2-2, October, 1995, EPA-454/
R-95-011)).
Table 1.--Estimated Regional Reductions Necessary to Mitigate Various Environmental Effects
----------------------------------------------------------------------------------------------------------------
Environmental effect
-----------------------------------------------------------------------------------------------------------------
Ozone Acid deposition Eutrophication
----------------------------------------------------------------------------------------------------------------
Regional NOX Reductions Necessary.... More than 50%.......... More than 40%.......... Up to 40%
NOX Reductions Achieved from the 5%..................... 5%..................... 5%
Final Rule as Percentage of Total
NOX Emissions.
----------------------------------------------------------------------------------------------------------------
v. NOX Reductions Not Sustained
Although national NOX emissions are expected to decrease up to
the year 2000, (see docket item IV-A-8 (EPA, ``National Air Pollution
Emission Trends, 1900-1994'' at 5-5, October, 1995, EPA-454/R-95-011)),
emissions are projected to begin increasing after 2000 (id. at 5-2 and
6-8 2). The existing NOX control programs under the Clean Air
Act (including the Mobile Source Program under title II and the Acid
Rain NOX Program under title IV) limit NOX emission rates
(e.g., the pounds of NOX emissions per amount of fuel consumed
(under title IV)) for emission sources. The programs do not cap the
total tonnage of nationwide emissions. As the number of emission
sources and the use of emission sources increases, reductions due to
emission rate limitations are offset to an increasing extent. For this
reason, after 2002, when implementation of these NOX control
programs is largely completed and growth in sources and source use
continues, NOX emissions will gradually increase for the
foreseeable future (id. at 5-5). Section 401(b) of the Act suggested,
as a guideline, that NOX emissions should be reduced nationwide by
2 million tons from the 1980 level. By about 2006, total NOX
emissions will surpass that guideline unless additional efforts are
made (e.g., under title IV) to reduce NOX emissions (See figure 1,
below). The projected increase in total NOX emissions is well
within the time frame considered by Congress in title IV. EPA notes
that the nationwide annual cap for SO2 emissions, also established
under section 402, begins to apply in the year 2010. Until 2010, total
annual allocated SO2 allowances will exceed the cap, because of
additional allowances allocated under section 409 for repowered units
and bonus allowances under section 405. Additional NOX reductions,
such as these under today's final rule, are necessary both in light of
the likely need to reduce NOX to address ozone, acid deposition,
and eutrophication, and in light of the NOX reduction guideline in
section 401(b) of the Act. In short, new initiatives are needed to
reduce NOX emissions on a regional scale in order to improve
environmental quality and health beyond 2000.
\2\ Report's projections take into account requirements for
Reasonably Available Control Technologies (RACT) under title I,
enhanced programs for inspection and maintenance of mobile sources
under title I, and title IV Group 1 emission limits promulgated
April 13, 1995 (id. at 6-8, (assuming, for analytical purposes, that
title IV emission limits are set at RACT)).
---------------------------------------------------------------------------
BILLING CODE 6560-50-P
[[Page 67117]]
[GRAPHIC] [TIFF OMITTED] TR19DE96.000
BILLING CODE 6560-50-C
vi. Cost-Effectiveness
The revision of Phase II, Group 1 emission limitations and
establishment of Group 2 emission limitations is a cost-effective means
of achieving the likely necessary, additional regional NOX
reductions. The control technologies on which the revised Group 1
limits and the Group 2 limits are based are more cost-effective (i.e.,
have a lower cost per ton of NOX removed) when applied to the
respective Group 1 and Group 2 boiler types than most other control
technologies applied to these boiler types or to non-utility sources.
As shown below, the dollar cost per ton of NOX removed for
reductions under the final rule is less than, or at the lower end of,
the range of dollar cost per ton of NOX removed for most
alternative reductions. In short, the NOX reductions achievable
under this final rule are among the less expensive that can be made.
[[Page 67118]]
Utility Sources: For coal-fired utility boilers using higher level
control technologies, (e.g., SCR with higher NOX reduction
capability) than the technologies on which the title IV limits are
based, the average cost-effectiveness for typical wall-fired boilers
ranges from $1,226/ton to $1,670/ton with percent reductions ranging
from 60-90%. For typical tangentially fired boilers, the cost-
effectiveness ranges from $1,439/ton to $1,935/ton with percent
reductions ranging from 60-90%. For typical cyclone boilers, the cost-
effectiveness ranges from $440/ton to $880/ton with percent reductions
ranging from 60-90%. For typical cell-burner boilers, the cost-
effectiveness ranges from $624/ton to $801/ton with percent reductions
ranging from 60-80%. For typical wet bottom boilers, the cost-
effectiveness ranges from $572/ton to $733/ton with percent reductions
ranging from 60-90%. For typical roof-fired (vertically-fired) boilers,
the cost-effectiveness ranges from $750/ton to $907/ton with percent
reductions ranging from 60 to 90%. For typical oil and gas utility
boilers, the average cost-effectiveness for wall-fired dual-fired
boilers under various NOX reduction technologies ranges from $748/
ton to $2,263/ton with percent reductions ranging from 40-90%. For
typical tangentially fired dual-fired boilers, the cost-effectiveness
ranges from $507/ton to $1,573/ton with percent reductions ranging from
30-90% (see docket item IV-J-4 (Ozone Transport Assessment Group,
Control Technologies and Options Workgroup, Final Report, April 11,
1996)).
As compared to the cost-effectiveness ranges for higher level
control technologies applied to typical utility boilers, the average
cost-effectiveness for meeting the Group 1 and Group 2 emission limits
under today's final rule, using the control technologies on which the
limits are based, is approximately $229/ton of NOX removed.
Non-Utility Point Sources: Non-utility point sources NOX
reductions are less cost effective, on average, than NOX
reductions under today's final rule. For example, the average cost-
effectiveness for process heaters ranges from $290-50,000/ton at an
average reduction of 5-90%. For cement manufacturing, the average cost-
effectiveness ranges from $470-4,870/ton at an average reduction of 20-
90%. For wood manufacturing, the average cost-effectiveness ranges from
$1,000 to over $10,000/ton at an average reduction of 0-60% (see docket
item IV-J-4 (Ozone Transport Assessment Group, Control Technologies and
Options Workgroup, Final Report, April 11, 1996)).
Mobile Sources: For mobile sources, the cost-effectiveness under
various NOX control options is also high, on average, as compared
to reductions under today's final rule. For example, the average cost-
effectiveness for light-duty on highway vehicles ranges from $1,100-
$260,000/ton, with percent reductions ranging from 0.2-21%. For heavy-
duty on highway vehicles, the average cost-effectiveness ranges from
$1,000/ton to $40,000/ton, with percent reductions ranging from 0.02-
5.6%. For non-road sources, the average cost-effectiveness ranges from
$119/ton to $23,000/ton, with percent reductions ranging from 0.4-3.4%
(see docket item IV-J-6 (Mobile Sources Assessment: NOX and VOC
Reduction Technologies for Application by the Ozone Transport
Assessment Group, Final Report, March 4, 1996)).
Table 2 summarizes the cost-effectiveness ranges of NOX
controls for the three major NOX emitting sources, as compared to
the cost-effectiveness of reductions under the revised Group 1 limits
and Group 2 limits.
Other: The reductions from applying control technologies to coal-
fired power plants under today's final rule can be as cost-effective to
achieve as reductions from other point sources (e.g., wastewater
plants) and area sources (e.g., farms, animal pastures). Studies
concerning eutrophication in the Chesapeake Bay estimate the following
average cost-effectiveness of control technologies applied to non-
utility sources: chemical addition or biological removal of nitrogen
from wastewater processing, $4,000 to over $20,000/ton of nitrogen
removed; and management practices to reduce nitrogen from fertilizers,
animal waste, and other non-point sources, $1,000 to over $100,000/ton
of nitrogen removed (see docket items IV-J-25 (Camacho, 1993:97-98) and
IV-J-27 (Shulyer, 1995:6)).
Table 2.--Average Cost-Effective of NOX Controls by Source
[Utility, other point source, mobile]
------------------------------------------------------------------------
Range in typical
cost- Percent
effectiveness ($/ reduction
ton)
------------------------------------------------------------------------
Utility sources (Coal w/advanced NOX
controls):
Wall-fired........................... $1,226-1,670 60-90
Tangentially-fired................... 1,439-1,935 60-90
Cyclones............................. 440-880 60-90
Cell burners......................... 624-801 60-80
Wet bottoms.......................... 572-733 60-90
Roof (vertically-fired).............. 750-907 60-90
Utility sources (Oil and Gas):
Wall dual-fired...................... 748-2,263 40-90
Tangential dual-fired................ 507-1,573 30-90
------------------------------------------------------------------------
Source: Ozone Transport Assessment Group, Control Technologies and
Options Workgroup, Final Report, April 11, 1996.
------------------------------------------------------------------------
Average cost-
effectiveness Percent
Title IV phase II NOX rule of Sec. reduction
407(b)(2) ($/ under Sec.
ton) 407(b)(2)
------------------------------------------------------------------------
Group 1 and group 2..................... $229 20
------------------------------------------------------------------------
See section IV.B (Table 17) of this preamble.
[[Page 67119]]
------------------------------------------------------------------------
Range in typical
cost- Percent
Non-utility point sources effectiveness ($/ reduction
ton)
------------------------------------------------------------------------
Non-utility boilers...................... $490-19,600 5-90
Process heaters.......................... 290-50,000 20-90
I.C. engines............................. 180-13,400 5-98
Gas turbines............................. 130-2,760 60-90
Residential fuel combustion.............. 1,600-62,500 50-100
Cement manufacturing..................... 470-4,870 20-90
Metals processing........................ 120-11,600 12-96
Wood manufacturing....................... 1,000-10,000+ 0-60
Agriculture chemical manufacturing....... 76-715 44-99
Incineration............................. 800-10,000 10-77
------------------------------------------------------------------------
Source: Ozone Transport Assessment Group, Control Technologies and
Options Workgroup, Final Report, April 11, 1996.
------------------------------------------------------------------------
Range in typical
cost- Percent
Mobile sources effectiveness ($/ reduction
ton)
------------------------------------------------------------------------
Light-duty (on highway).................. $1,100-260,000 0.2-21
Heavy-duty (on highway).................. 1,000-40,000 0.02-5.6
Non-road................................. 119-23,000 0.4-3.4
------------------------------------------------------------------------
Source: Mobile Sources Assessment: NOX and VOC Reduction Technologies
for Application by the Ozone Transport Assessment Group, Final Report,
March 4, 1996.
------------------------------------------------------------------------
Average cost-
effectiveness Percent
Title IV phase II NOX rule of Sec. reduction
407(b)(2) ($/ under Sec.
ton) 407(b)(2)
------------------------------------------------------------------------
Group 1 and Group 2..................... $229 20
------------------------------------------------------------------------
vii. Need to Revise Group 1 Limits and Establish Group 2 Limits
As discussed above, in order to mitigate adverse effects on health
and the environment due to NOX emissions, significant, additional
reductions in regional atmospheric NOX emissions from current
levels are likely to be necessary. Further, the contribution of the
final rule toward the overall NOX reduction goal is approximately
5%. The NOX reductions under the rule represent only a portion of
the much larger NOX reductions likely to be needed and are among
the most cost-effective reductions available. EPA concludes that the
reductions under the final rule represent a reasonable step toward
achieving necessary NOX reductions.
Some commenters suggested that, because the authority to revise the
Phase II, Group 1 emission limitations and to issue Group 2 emission
limitations arises under title IV of the Clean Air Act, EPA must
consider only the acidification impacts of NOX emissions in
deciding whether to revise or issue limitations. Allegedly, all other
impacts must be addressed only under other provisions of the Act. EPA
rejects this crabbed view of its authority under section 407(b)(2) as
having no basis in statutory language or logic. In granting EPA the
authority to decide to revise the Phase II, Group 1 emission
limitations, section 407(b)(2) only requires a determination of the
availability of more effective LNB technology and does not bar
consideration of non-acidic deposition impacts. Similarly, in requiring
EPA to issue Group 2 emission limitations, section 407(b)(2) sets forth
several criteria for setting the limitations but none of the criteria
bars consideration of non-acidic deposition impacts. On the contrary,
section 407(b)(2) has a general requirement that EPA take account of
``environmental impacts'' in setting Group 2 emission limitations. 42
U.S.C. 7651f(b)(2).
In the absence of a statutory bar on considering all environmental
impacts of NOX emissions and in light of the general purpose of
the Clean Air Act to, inter alia, ``protect and enhance the quality of
the Nation's air resources so as to promote the public health and
welfare and the productive capacity of its population'', it would be
illogical for EPA to focus exclusively on acid deposition.3 42
U.S.C. 7401(b)(1). The latter approach would require EPA to regulate on
a piecemeal basis and to blindly ignore a major part of the harmful
effects of NOX emissions when setting nationwide NOX emission
limits under title IV. In any event, EPA maintains that, even if the
Agency were confined to considering only the acidic deposition effects,
referred to above, of NOX emissions, it would still conclude that
additional NOX reductions are necessary and that the emission
limitations set forth in today's rule should be adopted.
---------------------------------------------------------------------------
\3\ Although, as discussed below, section 401(b) states that the
general purpose of title IV is ``to reduce the adverse effects of
acid deposition'', this provision should not be interpreted as
barring consideration of other environmental impacts for purposes of
setting emission limitations under section 407. 42 U.S.C. 7651(b).
EPA's interpretation--which harmonizes sections 101(b)(1) (stating
the general purposes of the Clean Air Act) and 401(b) (stating the
general purposes of title IV)--is that, while the primary focus in
promulgating regulations under title IV is reduction of acidic
deposition, other environmental impacts may also be considered.
---------------------------------------------------------------------------
Some commenters also noted that section 401(b) states that the
purpose of title IV is to reduce acidic deposition through reduction of
annual SO2 emissions of ten million tons from 1980 levels ``and,
in combination with other provisions of this Act, of nitrogen oxides
emissions of approximately two million tons from 1980 emission levels,
in the forty-eight contiguous States and the District of Columbia.'' 42
U.S.C. 7651(b). According to such commenters, because this goal is
already met by the existing Phase II, Group 1 emission limitations (as
well as by regulations under other parts of the Clean Air Act), there
is no basis for revising the limitations. However, section 401(b)
provides only general guidance concerning implementation of title IV
and, in light of the imprecision of its language, does not--and was not
intended to--impose an absolute limit on the amount of NOX
reductions that can be required under emission limitations promulgated
under section 407.
In contrast to the SO2 provisions of title IV, which set a
nationwide cap on total tonnage of SO2 emissions (i.e., 8.95
million tons starting in 2010), the NOX provisions of title IV
provide only for limits on the NOX emitted per mmBtu of fuel
burned. Even if the NOX emission limitations are met, increased
use of existing coal-fired and other
[[Page 67120]]
utility boilers in the future in response to growth in demand for
electricity can result in increased tonnage of NOX emissions. The
NOX emissions reductions projected to be achieved through adoption
of any given set of NOX emission limitations under title IV are
therefore not permanent. For this reason, when EPA estimates NOX
reductions resulting from title IV emission limitations, the estimates
are tied to a specific year, in this case the year 2000. Regulatory
Impact Analysis of NOX Regulations at 1-7 and 1-8, December 8,
1995. Moreover, as discussed above, total NOX emissions are
projected to decline through 2000, increase thereafter, and exceed the
two million guideline by around 2006. In short, the commenters' claim
that a two-million-ton emission reduction ``goal'' is ``satisfied'' by
the existing Group 1 emission limitations is inaccurate because a two-
million-ton level of reductions from 1980 achieved for a given year
(e.g., for 2000) through these limitations is unlikely to be
maintained, in the near future without further reductions.
Although EPA maintains that the 2 million ton guideline in Section
401(b) aims at total NOX emissions of 2 million tons below the
1980 levels, EPA notes that the final rule will result in total Group 1
and Group 2 boiler NOX emissions around 2 million tons less than
what they otherwise would have been in 2000. The annual NOX
reductions anticipated from the existing Group 1 emission limitations
under the April 13, 1995 rule and additional annual reductions
anticipated from the Phase II, Group 1 and Group 2 emission limitations
under today's final rule are about 1,170,000 tons and 890,000 tons
respectively for the year 2000, for a total of about 2,060,000 tons.
EPA's current estimate of reductions from the April 13, 1995 rule is
lower than the reductions originally estimated (i.e., about 1,890,000
tons for the year 2000) for that rule. 59 FR 13538, 13562-63 (March 22,
1994); see also 59 FR 18760 (adopting for April 13, 1995 rule the
Regulatory Impact Analysis originally promulgated for the March 22,
1994 rule).
In making the original estimates of reductions, EPA used emissions
factors (i.e., estimated uncontrolled emission rates based on coal type
and boiler type) to determine the uncontrolled emissions of boilers to
which the existing Group 1 emission limitations were to be applied. In
response to comment in today's rulemaking concerning the inaccuracy of
emission factors, EPA has minimized its use of emission factors and
instead relied almost exclusively on actual, short-term, uncontrolled
emissions data from continuous emissions monitoring obtained during
annual monitor certification testing (i.e., CREV data) or submissions
of CEM, EPA reference method, or other test data by utilities. This
data was not generally available to EPA when the April 13, 1995 rule
was published.4 As a result of using more accurate uncontrolled
emissions data, EPA's estimates of anticipated reductions under the
existing Group 1 emission limitations are now more accurate and are
lower. Even if section 401(b) were viewed as imposing a ``ceiling'' of
``approximately two million tons'' of NOX reductions under section
407, the reductions anticipated under the emission limitations adopted
in the April 13, 1995 rule and today's final rule are consistent with
that ``ceiling.''
---------------------------------------------------------------------------
\4\ For the January 19, 1996 proposal in the instant rulemaking,
EPA replaced many, but not all, of the emissions factors with actual
data, which resulted in estimated annual reductions under the
current Group 1 emission limitations of about 1,540,000 million
tons. See Regulatory Impact Analysis for the proposed rule (docket
item II-F-2).
---------------------------------------------------------------------------
For the reasons discussed above, EPA concludes that it should
exercise its discretion under section 407(b)(2) to revise the Phase II,
Group 1 emission limitations. The revised Group 1 limits represent a
reasonable step toward achieving the significant NOX reductions
that are likely to be necessary, and are consistent with the 2 million
ton guideline for NOX reductions. The revision of the Group 1
emission limitations will result in about 120,000 tons of additional
annual NOX reductions. Actions to achieve NOX reductions
beyond those realized under title IV are being considered, or will be
considered in the future, under other titles of the Clean Air Act.
Unlike the Group 1 limitation revisions, which are discretionary
under section 407(b)(2), the issuance of Group 2 emission limitations
is mandatory under that section so long as the requirements of the
section (e.g., cost comparability) are met. However, as noted above,
EPA is required, when setting Group 2 emission limitations under
section 407(b)(2), to consider environmental impacts. EPA's application
of the section 407(b)(2) requirements for setting Group 2 emission
limitations--including the consideration of environmental impacts--is
set forth in detail below in section III.B of this preamble. EPA
concludes that, like the Group 1 revisions, the Group 2 emission
limitations supported and adopted in that section of the preamble
represent a reasonable step toward achievement of necessary,
significant NOX reductions and are consistent with the 2 million
ton guideline for NOX reductions.
II. Public Participation
Regulations were proposed in the Federal Register on January 19,
1996 (61 FR 1442). The notice invited public comments and copies of the
proposed rule were made available to interested parties.
EPA held a public hearing to provide interested parties the
opportunity for oral presentation of data, views, or arguments
concerning the proposed regulations. The hearing was held on February
8, 1996 in Washington, DC. Four persons testified at the hearing
concerning issues related to the proposed regulations. The hearing was
open to the public, and each attendee was given an opportunity to
comment on the proposed regulations. (See docket items IV-F-1, IV-F-2
and IV-F-3.) The initial public comment period (January 19, 1996 to
March 4, 1996) was extended by two weeks to March 19, 1996 to allow
additional time for inspection of interagency review materials which
EPA added to the docket on January 26, 1996. (See docket item III-A-2.)
III. Summary of Major Comments and Responses
EPA received approximately 100 comment letters regarding the
proposed regulations, presenting more than 200 issues. Commenters
included public and municipal utilities, utility associations, state/
local agencies and Attorneys General, environmental organizations,
vendors, general industry, research/trade groups, and private citizens.
A copy of each comment letter received is included in the rulemaking
docket. A list of commenters, their affiliations, and the EPA docket
item number assigned to their correspondence is included in the
background information document.
All of the comments have been carefully considered, and where
determined to be appropriate by the Administrator, changes have been
made in the final regulations. The background information document
includes a summary of all the comments and EPA's response on each of
the relevant issues. The following sections of the preamble provide a
summary of the major comments received and the Agency's response to
those major comments.
[[Page 67121]]
A. Phase II, Group 1 Boiler NOX Emission Limits
1. Boiler Population Used To Assess NOX Emission Limits
Background. For the proposed rule, EPA developed a computerized
boiler database containing detailed information on the characteristics
and pre-retrofit and post-retrofit emission rates of coal-fired units
with Group 1 boilers on which low NOX burners (LNBs) had been
installed without any other NOX controls (``the LNB Application
Database''). This database contained all known applications of LNBs to
Group 1 boilers that were installed subsequent to 11/15/90 (the date of
enactment of the 1990 amendments to the CAA) and for which EPA had at
least 52 days of quality assured post-retrofit data measured by
continuous emission monitors (CEMs) certified according to 40 CFR part
75. The 24 wall-fired boilers and 9 tangentially fired boilers in this
database formed the empirical basis for EPA's assessment of the
effectiveness of low NOX burner technology and the revised annual
NOX emission limitations provisions for Group 1 boilers in the
proposed rule.
Comment/Analyses: EPA received approximately 25 comment letters
(from 19 utilities, 3 utility associations, 2 states, and an
environmental organization) on the appropriateness of including or
excluding certain boilers and the selection criteria used to define
eligibility for the LNB Application Database.
Several commenters suggested that EPA include specific boilers to
increase the size and improve the representativeness of the
tangentially fired subset in the LNB Application Database: Riverbend 7
and 8, Allen 1 and 3, J.H. Campbell 3, Gallatin 4, and Lansing Smith 2
(see, for example, docket items IV-D-22, p. 1; IV-D-21, pp. 2-3; IV-D-
20, pp. 7-9, and IV-D-65, p. 22). The commenters acknowledged that many
of these retrofit cases did not satisfy the quality assurance criteria
that EPA had established for inclusion in the LNB Application Database.
They believed, however, that the general benefits of broadening the
experiential basis for tangentially fired boilers outweighed specific
data quality concerns. As one commenter said, ``Although not [based on]
CEM data, Gallatin Unit 4's performance test result of 0.47 lb/10
6 Btu is reliable, relevant evidence * * * and should be
considered by EPA.'' (See docket item IV-D-20, p. 9.)
Commenters also suggested that EPA include specific boilers to
improve the representativeness of the wall-fired subset in the LNB
Application Database, particularly with respect to boilers with high
uncontrolled emission rates: Hammond 4, Watson 4 and 5, Valley 1 and 2
(see, for example, docket items IV-D-65, p.22). Several commenters
cited additional wall-fired retrofit cases within the context of the
related issue of the dependence of NOX emissions on boiler load:
Conesville 3, Picway 9, Amos 1 and 2, Big Sandy 2, Glen Lyn 6, Colbert
5, Valley 1-4; Presque Isle 5 and 6 (see docket items IV-D-73, p.1; IV-
D-20, p.5; IV-D-26, p.2).
On the other hand, several commenters fully endorsed the quality
assurance criteria EPA has used to determine eligibility for the LNB
Application Database (see, for example, docket items IV-D-063, p.12;
IV-D-046, p.3-4). They said that EPA properly excluded older LNB
installations (such as Gallatin 4, Lansing Smith 2, and Hammond 4) for
which quality assured long-term post-retrofit CEM data did not exist.
(EPA notes that this criterion generally excludes experimental or
otherwise short-lived LNB installations such as those used for
technology demonstrations, and the Allen units.5) These commenters
also recommended that EPA should attach greater significance to (or
rely exclusively on) LNB applications in the 13-state Northeast Ozone
Transport Region (OTR) for the evaluation of LNB technology
effectiveness because these applications have been required to meet a
NOX emission limit beginning May 31, 1995, whereas most other
applications have not had to comply with a recently established
NOX standard.
---------------------------------------------------------------------------
\5\ The Allen plant is located in Gaston County, NC, which,
until July 1995, was considered in non-attainment for ozone. The
utility installed LNBs on two Allen boilers, the vendor is reported
to have optimized in mid 1995. In July 1995, Gaston County was
redesignated to ozone attainment and low NOX operation was
discontinued on Allen 1 and 3 on September 1, 1995 (see docket item
IV-D-22, p. 1). As a result, Allen units 1 and 3 each have less than
52 days of emissions data after optimization of their respective
LNBs.
---------------------------------------------------------------------------
Some commenters correctly noted that one wall-fired boiler in the
LNB Application Database used for the proposed rule analysis, North
Valmy 1, should be excluded because this boiler had pre-existing
NOX controls (i.e., Babcock and Wilcox (B&W) DRB version LNBs) so
its baseline measurement does not represent an uncontrolled emission
rate. EPA notes that this NSPS boiler, when retrofitted with modern
LNBs (i.e., B&W XCL version), has sustained an average post-retrofit
controlled emission rate of 0.264 for calendar year 1995 (see docket
item II-A-9). ``NSPS boilers'' are new coal-fired utility units on
which construction commenced after August 17, 1971, which are subject
to New Source Performance Standards (NSPS) (40 CFR part 60, subparts D
or Da). Some NSPS boilers had early versions of LNBs and/or some other
type of NOX combustion control installed as original equipment.
EPA has excluded these ``controlled NSPS boilers'' from the LNB
Application Database and regression models because their measured
baseline emission rates do not generally represent uncontrolled
emissions. EPA has included all NSPS boilers, both controlled and those
without built-in NOX combustion control equipment, in the Phase
II, Group 1 boiler set to which the models are applied since NSPS
boilers represent approximately one third of the units affected by this
rulemaking.
One commenter recommended that EPA exclude two boilers, Coleman C1
and Pulliam 7, because, according to this commenter, these boilers have
low NOX combustion controls beyond the LNB definition in 40 CFR
76.2. EPA disagrees with this commenter's opinion that these two
retrofits include auxiliary combustion air outside the waterwall hole
which are `` `staging' combustion on active burners analogous to
overfire air'' (see docket item IV-D-51, p. 9). EPA also notes that
another commenter, who represents 67 utilities, included both units in
their regression analyses on the performance of LNBs applied to wall-
fired Group 1 boilers (see docket item IV-D-65, p. 58 and Enclosure 8,
Table 4-1). DOE included Coleman C1 in its regression analyses, but
excluded Pulliam 8 (probably because, as EPA learned after the rule
proposal, the utility switched to Powder River Basin coal for both
Pulliam 7 and 8) (see docket item II-D-62).
Some commenters recommended that EPA include Group 1 boilers that
installed both LNB and overfire air (OFA) in the LNB Application
Database, primarily because they believe units with high uncontrolled
emission rates were under-represented in the proposed rule analysis
(see, for example, docket item IV-D-58, p. 4). These commenters
provided supporting data for certain boilers, including: Eastlake 1, 3,
and 4; and Ashtabula 7 (see docket item IV-D-23, p. 5). As discussed
later in this section of the preamble, EPA disagrees with this
recommendation. First, OFA cannot be considered in determining whether
to revise the Group 1 limits and the assessment of the achievable
performance of LNBs alone is problematic when LNBs are used in
combination with other technologies. Further, the addition of 20 units
to the LNB Application Database has
[[Page 67122]]
significantly improved the robustness of EPA's regression models for
units with high uncontrolled emission rates.
Several commenters agreed with EPA's decision to exclude boilers
using Powder River Basin or other subbituminous coal from the LNB
Application Database (see, for example, docket items IV-D-15, p. 3; IV-
D-65, p. 20). For such boilers, measured post-retrofit NOX
emission reductions reflect the combined effects of switching to a coal
with inherently lower NOX emissions plus the application of LNBs.
Response: In light of the comments requesting the inclusion and/or
exclusion of specific boilers from the LNB Application Database, EPA
has formalized and expanded the data quality assurance criteria used in
the rule proposal into Data Quality Objectives (DQOs). The DQOs are
rigorous and precisely defined rule tables which were used to screen
all candidate boiler retrofit cases and hourly CEM data observations.
The DQOs are designed to ensure that the LNB Application Database
satisfies objective and consistent data quality assurance standards.
Table 3 presents EPA's DQOs for evaluating candidate boiler retrofit
cases (DQOs Applied to Boilers) and for quality assuring hourly post-
retrofit CEM data (DQOs Applied to Data).
Table 3.--Data Quality Objectives Applied to Boilers and Data to Screen Boilers for Inclusion in the LNB
Application Database
----------------------------------------------------------------------------------------------------------------
DQO# DQOs applied to boilers Rationale
----------------------------------------------------------------------------------------------------------------
1B Only dry bottom wall-fired and tangentially NOX emission rates for Group 1 boilers affect
fired boilers will be included in the dry bottom wall-fired and tangentially fired
database. boilers only.
2B Boilers must have an installed LNB control Consistent with Alabama Power v. EPA, 40 F.3d
technology only. Boilers with LNB plus 450 (D.C. Cir. 1994), EPA cannot consider
overfire air (OFA) or other controls will LNB+OFA installations when setting Group 1
not be included in the database. This limits.
determination is made by either (1)
information in EPA's Program Tracking System
Database or (2) direct contact with
individual utilities.
3B Any boiler with an LNB installation date Revised Group 1 limits are to be based on
prior to November 15, 1990 will not be improved performance of LNBs installed after
included in the database. LNB installation passage of 1990 Clean Air Act Amendments
dates are determined from (1) EPA's Program (CAAA).
Tracking System Database, (2) estimation of
the dates from visual interpretation of
hourly emissions plots, or (3) direct
contact with the utilities.
4B Only boilers with at least 52 days of post- 52 days is generally accepted as the minimum
retrofit data, following an equipment time period for assessing long-term
``break-in'' period of 30 calendar days, performance of NOX combustion control
will be included in the database. technology (see preamble section
III.A.2.ii). Vendors and utilities
acknowledge existence of ``break-in''
period, lasting about 30 calendar days,
during which boiler operations are often
highly irregular.
5B Boilers for which LNB design, installation Boilers with serious and persistent LNB
and/or operations are known to be seriously design, installation, and operational flaws
flawed will be excluded from the database. do not reflect the true NOX emission
This determination will be made on the basis reduction associated with LNB retrofit.
of published utility papers or information (This DQO is a logical extension of a
submitted to EPA for a rulemaking docket. pertinent statutory concept. Section 407(d)
(This DQO, however, was never used as the requires selection of appropriate control
sole basis for rejecting any candidate equipment ``designed to meet the applicable
boiler retrofit cases from current emission rate'' as well as proper
database.). installation and operation of such equipment
for determining eligibility, and an
appropriate emission rate, for an
alternative emission limitation).
6B Boilers must have a pre-retrofit uncontrolled Quality assured short-term uncontrolled
emission rate based on quality assured short- emission rate data are needed to perform
term CEM or test data that is verifiable in consistent analysis and projections using
the CREV database, the Acid Rain Cost Form first and second parts of model (see
for NOX Control Costs, or another source preamble, section III.A.3.ii.).
available to EPA.
7B Quarterly report submissions for boilers must Quarterly report submissions that do not
pass the quality assurance (QA) criteria in satisfy the CEM and other QA criteria in 40
40 CFR part 75. CFR part 75 contain insufficient information
to verify the accuracy of reported NOX
emission rate data.
8B NSPS boilers are excluded from the database.. Pre-NSPS boilers differ from NSPS boilers
with regard to furnace volume and heat
release rates and, as a result, NSPS units
can more easily meet a NOX reduction target
by retrofitting LNBs. This makes NSPS units
unrepresentative for establishing overall
LNB NOX reduction efficiency.
9B Only boilers not using Powder River Basin Powder River Basin coal has been identified
coal will be included in the database. by utilities as a subbituminous coal which
produces very low NOX emission rates. Its
performance cannot necessarily be reproduced
by any other type of coal for LNB
applications.
----------------------------------------------------------------------------------------------------------------
DQO# DQOs applied to data Rationale
----------------------------------------------------------------------------------------------------------------
1D Data generated using EPA's missing data The missing data routines include a penalty
substitution procedures will not be used (40 for not properly maintaining CEM equipment.
CFR part 75). In order to assess actual LNB performance,
only measured NOX emission rate data will be
used.
2D Hourly emission rate data will be adjusted Using bias adjusted NOX emission rates will
using the appropriate bias adjustment factor ensure compatibility of CEM NOX emission
for the boiler. rate measurements obtained from different
monitors.
[[Page 67123]]
3D NOX emission rates greater than 10 lb/mmBtu Such reported data values are clearly
and less than or equal to 0 lb/mmBtu will be erroneous (i.e., physically impossible) and,
discarded. thus, should not be included when estimating
achievable emission rates.
4D Hourly emission rate data for ``break-in'' Vendors and utilities acknowledge existence
period, defined as the 30 calendar days of ``break-in'' period, lasting about 30
following estimated date the boiler began calendar days, during which boiler
operating after shutdown for LNB retrofit operations are atypical due to vendor
(denoted on tables as ``LNB retrofit performance guarantee testing. Discarding
date''), will be discarded. hourly emissions data for ``break-in''
period also allows for any uncertainty
associated with exact date of beginning of
post-retrofit period.
----------------------------------------------------------------------------------------------------------------
EPA applied these DQOs to candidate boilers: those used in the
Phase II proposed rule analysis (Tables 2 and 3, 61 FR 1442, 1446-1447,
January 19, 1996); those that commenters requested EPA to consider
(many of which are named above); and additional LNB boiler applications
which EPA identified using 1995 and first and second quarter, 1996 CEM
data submitted pursuant to 40 CFR part 75 and other program
information. A detailed presentation of the results of EPA's
comprehensive data evaluation appears in docket item IV-A-6. The
resulting LNB Application Database, presented in Tables 4 and 5,
consists of 39 wall-fired boilers and 14 tangentially fired boilers and
contains over 477,800 hours of quality assured post-retrofit CEM data
on LNB performance.
Table 4.--Wall-fired Boilers in the LNB Application Database
----------------------------------------------------------------------------------------------------------------
Load weighted
Uncontrolled post-
Obs. No. ORISPL Unit name/unit ID Phase NoX rate (ln/ optimization Percent NoX
mmBtu) NoX rate (ln/ removal
mmBtu)
----------------------------------------------------------------------------------------------------------------
1. 26 Gaston unit 1..... 1 0.900 0.384 57.3
2. 26 Gaston unit 2..... 1 0.780 0.384 50.8
3. 26 Gaston unit 3..... 1 0.800 0.413 48.4
4. 26 Gaston unit 4..... 1 0.800 0.413 48.4
5. 47 Colbert unit 1.... 1 0.800 0.421 47.4
6. 47 Colbert unit 2.... 1 0.670 0.421 37.2
7. 47 Colbert unit 3.... 1 0.830 0.421 49.3
8. 47 Colbert unit 4.... 1 0.860 0.421 51.0
9. 47 Colbert unit 5.... 1 0.780 0.434 44.4
10. 641 Crist unit 6...... 1 1.040 0.492 52.7
11. 641 Crist unit 7...... 1 1.160 0.517 55.4
12. 856 Edwards unit 2.... 2 1.000 0.514 48.6
13. 1043 Ratts unit 1SG1... 1 1.080 0.508 53.0
14. 1043 Ratts unit 2SG1... 1 1.090 0.468 57.1
15. 1295 Quindaro unit 2... 1 0.635 0.405 36.2
16. 1355 Brown unit 1...... 1 1.000 0.495 50.5
17. 1357 Green River unit 5 1 0.836 0.400 52.2
18. 1381 Coleman unit 1.... 1 1.410 0.489 65.3
19. 1381 Coleman unit 2.... 1 1.290 0.466 63.9
20. 1384 Cooper unit 1..... 1 0.900 0.419 53.4
21. 1384 Cooper unit 2..... 1 0.900 0.419 53.4
22. 2049 Watson unit 4..... 1 1.100 0.413 62.5
23. 2049 Watson unit 5..... 1 1.220 0.431 64.7
24. 2629 Lovett unit 4..... 2 0.570 0.349 38.8
25. 2629 Lovett unit 5..... 2 0.585 0.329 43.8
26. 2840 Conesville unit 3. 1 0.852 0.412 51.6
27. 2843 Picway unit 9..... 1 0.866 0.415 52.1
28. 3131 Shawville unit 1.. 1 0.990 0.486 50.9
29. 3131 Shawville unit 2.. 1 1.020 0.483 52.6
30. 3159 Cromby unit 1..... 2 0.600 0.378 37.0
31. 3178 Armstrong unit 2.. 1 1.042 0.420 59.7
32. 3948 Mitchell unit 1... 1 0.999 0.500 50.0
33. 3948 Mitchell unit 2... 1 0.999 0.500 50.0
34. 4042 Valley unit 1..... 1 1.100 0.477 56.6
35. 4042 Valley unit 2..... 1 1.100 0.477 56.6
36. 4042 Valley unit 3..... 1 1.050 0.473 55.0
37. 4042 Valley unit 4..... 1 0.925 0.473 48.9
38. 6041 Spurlock unit 1... 1 0.900 0.414 54.0
39. 6085 RM Schahfer unit 2 0.420 0.228 45.7
15.
----------------------------------------------------------------------------------------------------------------
[[Page 67124]]
Table 5.--Tangentially Fired Boilers in the LNB Application Database
--------------------------------------------------------------------------------------------------------------------------------------------------------
Uncontrolled Load weighted
NOX rate post-
---------------- optimization Percent NOX
Obs. No. ORISPL Unit name/unit ID Phase NOX rate removal
(ln/mmBtu) ----------------
(ln/mmBtu)
--------------------------------------------------------------------------------------------------------------------------------------------------------
1. 710 McDonough unit 1......................... 1 0.657 0.388 40.9
2. 710 McDonough unit 2......................... 1 0.600 0.388 35.3
3. 728 Yates unit Y4BR.......................... 1 0.561 0.421 25.0
4. 728 Yates unit Y5BR.......................... 1 0.650 0.421 35.2
5. 1374 Elmer Smith unit 2....................... 1 0.859 0.419 51.2
6. 1710 Campbell unit 1.......................... 1 0.690 0.456 33.9
7. 2554 Dunkirk unit 1........................... 2 0.478 0.343 28.2
8. 2554 Dunkirk unit 2........................... 2 0.478 0.331 30.8
9. 2642 Rochester 7 unit 4....................... 2 0.587 0.365 37.8
10. 2732 Riverbend unit 7......................... 2 0.580 0.421 27.4
11. 2732 Riverbend unit 8......................... 2 0.640 0.383 40.2
12. 2732 Riverbend unit 10........................ 2 0.772 0.357 53.8
13. 4041 S. Oak Creek unit 7...................... 1 0.661 0.377 43.0
14. 4041 S. Oak Creek unit 8...................... 1 0.665 0.377 43.3
--------------------------------------------------------------------------------------------------------------------------------------------------------
The Agency believes that the addition of 20 units to the LNB
Application Database increases the overall representativeness of the
database for use in analyzing the achievable emission rates for Group 1
boilers and addresses commenters'' concerns that the original database
may not adequately represent units with high uncontrolled emission
rates. The current database contains 22 units with uncontrolled
emission rates above the rates classified by one utility commenter as
``high'' (i.e., for wall-fired boilers, above 0.90 lb/mmBtu and for
tangentially fired boilers, above 0.68 lb/mmBtu, see docket item IV-G-
16, p. 7). For several reasons, the Agency believes these additions to
the database are more appropriate than adding boilers with LNB and
overfire air (OFA) as suggested by some commenters. First, under the
ruling in Alabama Power v. EPA, 40 F.3d 450 (D.C. Cir. 1994), EPA
cannot consider LNB with OFA installations in the LNB Application
Database for setting Group 1 limits. Second, isolating the true
NOX reduction performance of the LNB portion of LNB+OFA systems is
problematic because the controls are designed to reduce NOX as an
integrated system and site-specific factors influence the relative
contribution that each component (LNB vs. OFA) is designed to achieve.
Further, there is no basis for assuming that the performance of the LNB
portion, even if this could be measured accurately, is representative
of the performance that could be achieved by LNBs without the addition
of OFA.
2. Time Period/Averaging Basis Used To Evaluate Performance of Low
NOX Burner Technology
i. Background
Because the Acid Rain Phase I NOX Emission Reduction Program
did not go into effect until January 1, 1996, EPA did not have, at the
time the proposed rule was issued, CEM data on the performance of LNBs
applied to Group 1 boilers during a period when affected boilers were
required to meet the annual Phase I NOX emission limitations.
Further, for the reasons discussed below, it could not be assumed that
all the CEM data available, some of which had been recorded as early as
January 1, 1994, reflected LNB performance during optimized NOX
removal conditions.
As discussed in the Regulatory Impact Analysis (RIA) for the
proposed rule (see docket item II-F-2), plants incur both fixed and
variable operation and maintenance (O & M) costs when operating LNBs to
reduce NOX emissions to the lowest practicable level consistent
with prudent boiler operations to comply with regulatory emission
limitations. Therefore, even though LNB controls are installed,
utilities have a financial incentive not to operate units throughout an
extended period of pre-compliance to sustain the emission reductions
the controls were designed to achieve, since this would increase O & M
costs when the NOX emission reductions are not yet required. Thus,
the average NOX emission rate measured over an extended pre-
compliance period may not be a good predictor of LNB performance under
actual compliance conditions. On the other hand, it is reasonable to
expect that utilities operated their newly installed NOX controls
for some period of time following optimization of the equipment to
simulate compliance conditions, perhaps as a dry run or for training
purposes.
EPA's objective, then, was to identify the time period in the
stream of post-retrofit hourly CEM data that corresponds to operation
under optimized NOX removal conditions. EPA believed this time
period should contain 52 days of valid CEM data since, in publications
and in past rulemakings, the Department of Energy (DOE) and the utility
industry have stated that acceptable results of long-term performance
require data sets of at least 51 days with each day containing at least
18 valid hourly averages (see docket items II-I-99, Advanced
Tangentially-Fired Combustion Techniques for the Reduction of Nitrogen
Oxide (NOX) Emissions from Coal-Fired Boilers, and II-I-100,
Demonstration of Advanced Wall-Fired Combustion Modifications for the
Reduction of Nitrogen Oxide (NOX) Emissions from Coal-Fired
Boilers). EPA defined a 52-day ``low NOX period'' for the purposes
of assessing performance of LNBs applied to Group 1 boilers in the
proposed rule. The ``low NOX period'' was determined by
identifying the lowest average NOX emission rate each boiler has
sustained for at least 52 days, i.e., over a period of 1,248 hours when
the boiler was operating and valid CEM data (measured by CEMS certified
pursuant to 40 CFR part 75) were available. The low NOX period for
most boilers is considerably longer than 52 calendar days since hours
during which the boiler did not operate or hours for which valid CEM
data were not recorded are ignored and do not count
[[Page 67125]]
towards the required total of 1,248 hours.
Even prior to the proposed rule, utility commenters and DOE had
expressed the concern that by not using essentially all the recorded by
post-retrofit CEM data, EPA was not accurately assessing the long-term
performance capabilities of LNBs (61 FR 1442).\6\ Further, these
commenters believed that using a fixed-length shakedown period of 30 to
90 days, applied universally to all installations, to allow for
optimizing LNBs and operator training was more objective than using the
variable-length and site-specific shakedown periods implicit in EPA's
low NOX period methodology. Accordingly, for the proposed rule,
EPA also developed estimates of post-retrofit average NOX emission
rates for another time period beginning 30 calendar days after the
estimated date the boiler began operating after shutdown for LNB
installation and continuing to the end of the CEM data set. This period
is referred to as the ``overall post-retrofit period'' in the proposed
rule (61 FR 1447 (Tables 4 and 5); also see docket item II-A-9, Table 2
) and as the ``post-retrofit minus 30 days period'' (abbreviated as
``30-day post-retrofit period'' in tabular column headings) in the
technical support document for the final rule (see docket item IV-A-6).
---------------------------------------------------------------------------
\6\ EPA notes that the tangentially fired boilers in the LNB
Application Database used for the proposed rule had little more than
the requisite 52 days of quality assured post-retrofit CEM data.
Only CEM data reported through June 30, 1995, the end of the second
quarter reporting period, were available for analysis and the LNB
retrofit dates for tangentially fired boilers occurred in late 1994
or early 1995.
---------------------------------------------------------------------------
For the proposed rule, EPA developed estimates of post-retrofit
average NOX emission rates for a third period which, like the
overall post-retrofit period, uses most of the recorded post-retrofit
CEM data and, like the low NOX period, allows for a variable-
length shakedown period to accommodate the site-specific nature of LNB
equipment optimization and operator training processes. This time
period begins with the first hour of the low NOX period and
continues to the end of the CEM data set. It is referred to as the
``post-optimization period'' in both the proposed rule and final rule
analyses. As mentioned previously in section B of this preamble, the
post-optimization period forms the basis for EPA's final assessment of
the effectiveness of LNBs applied to Group 1 boilers.
Another concern, which was raised prior to the proposed rule by
utility commenters and DOE, is that limited time periods such as the
low NOX period may not adequately capture annual dispatch patterns
and seasonal variations in demand for electrical power generation.
Accordingly, for the proposed rule, EPA also investigated the
representativeness of load dispatch during the low NOX period by
comparing it to the load dispatch during calendar year 1994 for each
boiler or common stack in the LNB Application Database. EPA developed
two histograms using ``load bins'' for the horizontal axis: (1) Average
hourly NOX emission rate as a function of load during the low
NOX period; and (2) frequency of various boiler operating loads
throughout 1994 (for which EPA had actual performance data from the CEM
data set ). Then, EPA used these histograms to estimate ``load-weighted
annual average NOX emission rates'' based on weighted averages of
the average emission rate during the low NOX period for each load
bin times the number of hours the boiler operated in that load bin
during 1994 (61 FR 1448 (Tables 6 and 7)). To test the
representativeness of boiler operations during the low NOX period,
EPA also created bar charts comparing the percentage of time a boiler
operated in each load bin during the low NOX period to the
percentage of time it operated in that load bin during calendar year
1994 (see docket item II-A-9, Appendix B). Using these graphical
analyses, EPA concluded that most boilers in the LNB Application
Database had a load dispatch pattern during their low NOX period
similar to their annual dispatch pattern in 1994.
When analyzing long-term post-retrofit CEM data for the proposed
rule, EPA found no strong correlation between boiler operating loads
and hourly average NOX emission rates for either wall-fired
boilers or tangentially fired boilers in the LNB Application Database.
While earlier technical analyses performed for EPA in support of other
utility NOX emission rulemakings had generally adopted the
industry accepted presumption of a NOX vs. boiler load
relationship for many uncontrolled Group 1 boilers, they also showed
the direction, magnitude, and form of this correlation to be both
highly boiler-specific and difficult to predict (see, for example,
docket item IV-J-20).
Nevertheless, EPA recognized that a predictable systematic
correlation between hourly average NOX emission rates and boiler
load for all or some boilers could have significant ramifications for
proper application of a 52-day low NOX period methodology.
Accordingly, EPA developed the ``load-weighted annual average NOX
emission rates,'' defined above, to account for the potential existence
of a NOX vs. boiler load relationship. Because the load-weighted
annual average NOX emission rates were essentially the same as or
lower than the average NOX emission rates for the low NOX
period for these boilers (see 61 FR 1446 (Tables 5 and 6)) EPA selected
the simpler form, a straight average over the low NOX period, as
the basis for the proposed rule.
The Agency received many detailed comments and supporting data
about the appropriateness of using a limited low NOX period for
assessing LNB performance, the merits of site-specific variable-length
vs. universal fixed-length shakedown periods to reflect LNB equipment
optimization and operator training, the advantages and disadvantages of
the alternative time periods EPA had considered for the proposed rule
analysis, and the technical issue of the existence of a NOX vs.
load relationship and its relevance for assessing LNB performance
applied to Group 1 boilers. The first three issues are discussed in the
next section within the context of the low NOX period methodology
whereas the last issue, for which EPA received approximately 25 site-
specific data submissions from utility boiler owners or operators, is
treated separately in the subsequent section.
ii. Use of 52-Day Low NOX Period
Comment/Analyses: EPA received approximately 29 comment letters
(from 22 utilities, 2 utility associations, 3 states, a gas industry
representative, and an environmental association) on the
appropriateness of using a 52-day low NOX period for assessing LNB
performance when, for some boilers, considerably more post-retrofit
data was available.
Some commenters fully endorsed EPA's 52-day methodology and
implicit assumption that utilities not under a compliance obligation
are unlikely to operate the controls for maximum emission reductions
following LNB optimization and a low NOX test period. They
believed EPA had demonstrated that the 52-day methodology and ``load-
weighted annual average NOX emission rates'' adequately addressed
annual dispatch and load patterns in most cases. A utility that owns
and operates coal-fired units which have become subject to state-
mandated NOX Reasonably Available Control Technology (RACT)
requirements in 1995 said EPA should go even further and ``use NOX
data only from units that have had to comply with a recent NOX
standard (such as NOX RACT)'' for
[[Page 67126]]
evaluating the effectiveness of LNB technology (see docket item IV-G-
14, p. 1). EPA notes that 6 wall-fired boilers and 3 tangentially fired
boilers in the LNB Application Database are located in the Northeast
Ozone Transport Region and are subject to NOX RACT requirements.
The mean load-weighted annual average NOX emission rates over the
post-optimization period for these boilers are: 0.403 lb/mmBtu (wall-
fired) and 0.344 lb/mmBtu (tangentially fired).
One commenter noted that utilities had an explicit disincentive for
operating their LNBs to achieve the maximum practicable emission
reductions during 1994 and 1995, since section 407(b)(2) allows EPA to
promulgate revisions to Group 1 emission standards if measured average
post-retrofit NOX emission rates during this time frame indicate
``more effective low NOX burner technology is available'' (see
docket item IV-D-63, p.14). Another commenter endorsed the conclusion
that observations during the 52-day low NOX period may understate
the actual reduction capability of LNBs (see docket items IV-D-047, p.
2 and IV-D-063, p. 12-14).
Other commenters disagreed with the assumption that utilities did
not have any incentive to operate the installed LNBs to achieve maximum
emission reductions consistent with prudent boiler operations. One
utility stated that plant personnel ``operated [their] NOX control
systems in a compliance mode even though its units were technically not
yet subject to the Phase I NOX standard. [The utility] established
performance goals based on operating NOX reductions systems to
meet the standard and management bonuses were geared to meeting these
goals'' (see docket item IV-D-020, p. 6). EPA notes that all of this
utility's wall-fired units sustained average NOX emission rates
below 0.44 lb/mmBtu throughout their ``post-optimization'' periods
(i.e., the post-retrofit period excluding a shakedown period based on
actual boiler experience). The post-optimization periods for these
units varied in length from 12 to 18 months. Another utility stated
that boilers were operated in a manner to optimize NOX emission
reduction; to do otherwise would be ``counterproductive to the design
of the burners and would defeat the training of the operating staff''
(see docket item lV-D-023, p. 4). EPA notes that the units owned and
operated by both of these utility commenters are located outside
designated ozone nonattainment areas and are not subject to NOX
RACT or any other state-mandated NOX control requirements. Their
decision to operate in a low-NOX mode, therefore, was voluntary
and not made on the basis of whether a compliance obligation existed.
Several commenters indicated that the best approach for estimating
annual average NOX emission rates is to use a full year of post-
retrofit monitoring data (see, for example, docket item IV-D-38, p. 3).
Commenters reiterated the concern raised prior to the proposal rule,
that by not using essentially all the recorded post-retrofit CEM data,
EPA is not accurately assessing the long-term performance capabilities
of LNBs (see, for example, docket items IV-D-35, p. 3; IV-G-15, pp. 2-
3). They said EPA's 52-day low NOX period methodology fails to
take into account all of the operating variables that affect LNB
performance and biases the LNB performance assessment toward emission
reduction levels that may not be achievable over the long term.
Further, commenters who participated in DOE Clean Coal Technology
Demonstrations where the 52-day methodology was used, said the ``52-day
rule'' defines ``the minimum number of continuous days of data needed
before a data set can be considered `long-term' data. It is not a rule
that justifies selective editing of data, when more data are
available'' (see docket item II-D-65, p. 29).
Some of these commenters suggested using all CEM data recorded
after a fixed-length shakedown period whereas others believed a
variable-length shakedown period is more appropriate given the site-
specific nature of the LNB equipment optimization and operator training
processes. EPA notes that one utility commenter reported that burner
optimization for each of their five tangentially fired retrofits was
completed within 120 days of startup (see docket item IV-D-23, p.4),
which is considerably longer than the fixed 30-day shakedown period
recommended by DOE and others. Another utility commenter reported that
one of their wall-fired boilers, E.D. Edwards 2, was still being
optimized more than a year after the retrofit date (see docket item IV-
D-73, p. 3).
Several commenters indicated support for the post-optimization
period approach, which EPA had presented in the proposed rule together
with the 52-day low NOX period methodology and load-weighted
annual average NOX emission rates. As one utility said, `the post-
optimization period' emission results are the best data set
characterizing long-term low-NOX mode boiler operation. This
database maximizes the amount of low-NOX mode data (i.e., sample
size) collected following a period of demonstrated minimum NOX
operation.'' (See docket item IV-D-051, p. 8.)
Some commenters indicated a 52-day low NOX period methodology
would be credible for assessing the long-term performance of LNB
technology if NOX emission rates following LNB optimization do not
vary significantly with boiler load (see, for example, docket item IV-
D-72, p. 4). While these commenters generally believe NOX emission
rates are a function of load for many boilers (see discussion below
under NOX vs. Boiler Load Relationship), they do endorse the
concept of using less than essentially all the recorded post-retrofit
CEM data for assessing LNB performance.
Response: EPA believes that the 52-day low NOX period
methodology is technically justified for evaluating the achievable
NOX reduction capability of LNBs. This time period is sufficiently
long, in most instances, to reflect long-term operation as evidenced by
the generally similar load dispatch patterns observed during the low
NOX period and for calendar year 1994 for most boilers in the LNB
Application Database. However, assuring proper selection of a low
NOX period that is representative of long-term boiler operating
conditions in all instances can be difficult. An example of this is
E.D. Edwards 2 where, according to the utility, the 52-day low NOX
period EPA had selected for the proposed rule analysis was atypical
because it represents ``a period of testing in a low NOX mode when
the boiler was not optimized.'' Shortly thereafter, the utility re-
tuned the boiler for improved efficiency, to reduce loss on ignition
(LOI), and to maintain full compliance with particulate and opacity
emissions standards. (See docket item IV-D-073, pp. 3-4.) Another
commenter suggested possible adverse plant impacts may have occurred
during the low NOX period for a few other boilers in the LNB
Application Database (see docket item IV-D-65, Enclosures 7 and 14);
EPA's analysis of the specific impacts and remedial actions cited
indicates that these possible issues are adequately addressed by
extending the low NOX period into the longer post-optimization
period. Therefore, to maximize the likelihood that the performance
evaluation period is representative and to assure observations over the
broadest possible range of boiler operating variables and electric
power generation demand scenarios, EPA is using the longer post-
optimization period as the basis for assessing the performance of LNBs
applied to Group 1 boilers for the final rule.
[[Page 67127]]
EPA's decision to use the post-optimization period is also based,
in part, on the comments utilities have submitted regarding their
actions to operate installed LNBs in a compliance mode during 1995,
prior to the effective date of the Acid Rain Phase I NOX Emission
Reduction Program. EPA believes that there were reasons for utilities
to operate installed LNBs as if the emission standards were in effect,
even though such operation could increase utility O & M costs. EPA has
rejected the concept of using a ``post-retrofit minus 30 (or 60 or 90)
days period'' approach because utilities submitted significant evidence
documenting that the time required for LNB optimization is highly
variable and can be much longer than any of the fixed shakedown periods
under consideration (see, for example, docket items IV-D-023, IV-D-073,
and IV-G-04). Nonetheless, for comparison purposes, EPA has computed
average NOX emission rates based on the post-retrofit minus 30
days period for boilers in the LNB Application Database (see docket
item IV-A-6, Table 3-1).
The addition of four more quarters of CEM data to the LNB
Application Database substantially lengthens the post-optimization
period for most boilers.7 The post-optimization period also
includes six months of 1996 compliance data for each Phase I boiler in
the database. Table 6 presents summary statistics on the amount of
hourly CEM data and calendar months encompassed by the post-
optimization periods.
---------------------------------------------------------------------------
\7\ A notable exception is the post-optimization time period for
E.D. Edwards 2, which has been lengthened by a lesser amount. In
response to the utility's comments, EPA has selected another low
NOX period, beginning after October 1, 1995, the date on which
EPA believes corrections for adverse opacity and particulate
emissions were substantially complete.
Table 6.--LNB Application Database: Hours of CEM Data and Calendar
Months in Post-Optimization Periods
------------------------------------------------------------------------
Hours of CEM Calendar
Boiler types data months
------------------------------------------------------------------------
Wall-fired boilers: 85% have at least 11
months of CEM data in post-optimization
period:
Range.................................... 3,877-15,829 6-30
Average.................................. 9,547 16
Total.................................... 372,324 610
Tangentially fired boilers: 79% have at least
11 months of CEM data in post-optimization
period:
Range.................................... 1,280-12,327 4-18
Average.................................. 7,537 14
Total.................................. 105,523 190
------------------------------------------------------------------------
iii. NOX vs. Boiler Load Relationship
Comment/Analyses: EPA received approximately 23 comment letters
(from 21 utilities and 2 utility associations) criticizing EPA's
decision in the proposed rule to base revised Group 1 emission
limitations on a time period and averaging method which do not
explicitly recognize the existence of a NOX vs. load relationship.
As mentioned previously under section III.A.2.i. of this preamble, EPA
found no strong correlation between boiler operating loads and hourly
average NOX emission rates for either wall-fired boilers or
tangentially fired boilers in the LNB Application Database when
analyzing long-term post-retrofit CEM data for the proposed rule.
Nevertheless, to test the potential impact of a NOX/load
relationship, in the analysis accompanying the proposed rule EPA
developed a methodology that assumed the existence of a functional
relationship between NOX and boiler load. EPA then used this
methodology to estimate ``load-weighted annual average NOX
emission rates'' for each boiler or common stack in the LNB Application
Database (see docket item II-A-9, pp. 9-10).
The load-weighting methodology produced a weighted average based on
the frequency of various operating load intervals (or ``bins'') during
calendar year 1994 as reported in the CEM data set and the mean hourly
NOX emission rates for each load bin observed during the low
NOX period. (The computational procedures EPA used to estimate
load-weighted annual average NOX emission rates for the proposed
rule are described under preamble section III.A.2.i.) Finding that the
load-weighted annual average NOX emission rates for these boilers
were essentially the same as or lower than the average NOX
emission rates for the low NOX period without the assumption of a
NOX/load relationship (see 61 FR 1446 (Tables 5 and 6)), EPA
believed it was not necessary to investigate the NOX vs. load
relationship further and selected the more conservative (i.e., higher)
of the two sets of estimates for modeling annual average emission rates
that could be sustained by LNBs installed on Phase II, Group 1 boilers.
The commenters who criticized EPA's treatment of the NOX/load
relationship raised the following main issues:
Lack of statistical measures to quantify the extent of the
NOX/load relationship: Several commenters indicated that a
critical missing link in EPA's analysis of this issue for the proposed
rule was the failure to develop any statistical measures describing the
strength of the association, if any, between NOX and boiler load.
As one utility said, EPA concluded ``through observance of the data''
that the relationship between NOX and load is not strong for wall-
fired boilers (see docket item IV-D-023, p. 5)
Inconsistency with earlier EPA studies: Some commenters claimed
that earlier EPA studies and utility emission rulemakings supported the
existence of the NOX/load relationship.
Examples to show presence of a NOX/load relationship: Many of
the commenters on this issue included site-specific data intended to
document the presence of a well-correlated NOX/load relationship.
On the other hand, some commenters who supported EPA's use of the
low NOX period for evaluating the performance of LNBs also said
EPA's comparison of load-weighted annual average NOX emission
rates vs. average NOX emission rates without the assumption of a
NOX/load relationship satisfactorily addresses this issue (see,
for example, docket items IV-D-46, p. 5 and IV-D-56, p. 1). According
to a state agency, the ``52-day time frame is representative of a wide
range of operations in a facility'' because the load variations over a
seven-day week are likely to be more significant than seasonal
variations. This agency said that, for most load-following units, load
changes are likely to be more significant between weekends and weekdays
than between seasons. Only the highest base-loaded units do not exhibit
this load cycle and such units are ``likely not affected by seasonal
changes'' (see docket item IV-D-27, p. 9).
Response: After further extensive boiler-by-boiler analysis of
NOX and boiler load, using both data provided by commenters and
reported independently under 40 CFR part 75 requirements, EPA has
determined that the installation of LNBs dampens any NOX/load
correlation that may have
[[Page 67128]]
existed at uncontrolled boilers and, in many instances, virtually
eliminates any long-term relationship. A NOX vs. load relationship
appears to have persisted for none of the tangentially fired boilers
and for only a few of the wall-fired boilers (Colbert 5, E.D. Edwards
2, Quindaro 2, and Jack Watson 5) in the LNB Application Database (see
docket item IV-A-6, pp. 4-2 through 4-7). However, despite these
findings, in response to commenters' insistence that a definite
functional relationship exists between NOX and boiler load, EPA
has employed a NOX/load weighting scheme in establishing NOX
emission limits in this final rule. This load-weighting method
incorporates at least two distinct improvements over the method used
for the proposed rule analysis. First, following commenters'
recommendation, the load weighting method employs ten load bins
consistent with the convention specified in 40 CFR part 75, rather than
the 25-MW increments used in the proposal. Second, the method uses
post-retrofit CEM data over the longer post-optimization period, rather
than the 52-day low NOX period, to estimate mean hourly NOX
emission rates for each load bin, thus making it unnecessary to combine
load bins due to sparse data. (Commenters had also said the combining
of load bins with little or no data tended to mask the NOX/load
relationship. See docket item, IV-D-65, p. 35.) The load weighting
method uses hourly boiler or common stack load as reported in the CEM
data set for 1995 to establish the frequency of operation in different
load bins over a year. EPA has rigorously investigated the relationship
of individual load patterns of boilers sharing a common stack to the
combined load patterns over a year and, thus, to the annual average
NOX emissions for the common stack (see discussion of common stack
issues in section III.A.3.v of this preamble). Finally, EPA has
compared, where data are available, boiler or common stack load
patterns for 1994 and 1995 to assess inter-year variations in dispatch
and demand for electrical power generation (see docket item IV-A-6).
This improved load weighting scheme accounts for any potential
impact that annual load dispatch patterns may have on NOX
emissions. Its use should allay concerns raised by commenters on how
the presence of a NOX/load relationship might impede accurate
assessment of long-term LNB performance. In addition, EPA's specific
responses to the main NOX/load issues are presented below:
Lack of statistical measures to quantify the extent of the
NOX/load relationship: Even among those commenters who most
strongly assert the presence of a NOX/load correlation, there is
little consistency from boiler to boiler in either the functional form
or the direction of the NOX/load relationship. For example, of the
three commenters submitting regression equations as evidence of a
NOX/load relationship, one was based on a cubic model (see docket
item IV-D-20, Figure 3), another was based on a logarithmic model (see
docket item IV-G-14, p. 3), and a third was based on a quadratic model
(see docket item IV-G-16). A fourth commenter, represented the
NOX/load relationship from one-third to full load for eight
boilers as straight line plots with slopes varying from approximately
15 deg. to 45 deg. (see docket item IV-D-72, Attachment 1). Although no
supporting documentation was provided explaining how these plots were
derived, they would imply a linear model was appropriate. The situation
is further complicated when a NOX/load relationship is discernible
over only a portion of the load range. This is particularly an issue
for wall-fired boilers retrofit with LNBs. EPA's plots of data from
post-retrofit wall-fired boilers show that if a NOX/load
relationship is discernible at all, it occurs almost entirely in the
upper 10-20% of the boiler load range.
The absence of a consistent functional form for the NOX/load
relationship and a failure to persist across the full load range makes
application of a statistical measure to quantify the extent of the
NOX/load correlation difficult. Nonetheless, assuming a linear
relationship between NOX and boiler load, EPA estimated the
strength of correlation as indexed by R2 during post-retrofit
period for 30 wall-fired and 11 tangentially fired boilers or common
stacks in the LNB Application Database and, during the pre-retrofit
period, for 13 wall-fired and 6 tangentially fired boilers or common
stacks (see docket item IV-A-6, Cadmus Group 1 technical report, Table
4-1). The R2 statistic measures the fraction of the variability in
the dependent variable, hourly average NOX emission rate,
explained by the model. EPA chose an R2 of 40% as a threshold for
detection of the possible existence of a predictable correlation. For
the post-retrofit hourly average NOX emission rate measurements,
only 13% of the wall-fired and none of the tangentially fired boilers
or common stacks had an R2 of 40% or higher (suggesting no
predictable correlation). EPA compared the load dispatch pattern during
the post-optimization period for each boiler or common stack crossing
the R2 threshold to its annual dispatch pattern in 1995 and
concluded the patterns were similar enough that the improved load-
weighting methodology would mitigate the effects of any NOX/load
correlation on estimated controlled annual average emission rates.
Inconsistency with earlier EPA studies: Earlier technical analyses
performed for EPA in conjunction with other utility NOX emission
rulemakings generally adopted the industry accepted presumption of a
NOX vs. boiler load relationship. However, this was almost
exclusively for uncontrolled Group 1 boilers, not boilers retrofit with
LNBs. Prior studies also showed the direction, magnitude, and form of
this correlation to be both highly boiler-specific and difficult to
predict. (See, for example, docket item IV-J-20). Thus, for example, in
these earlier studies, some uncontrolled tangentially fired boilers
exhibit increasing NOX emission rates with decreasing boiler
loads, others show precisely the reverse correlation, and still others
have U-shaped curves. Uncontrolled wall-fired boilers typically exhibit
increasing NOX emission rates with increasing boiler loads.
However, this relationship was not found to be universally valid
either, and the strength of the correlation, when present, varies
considerably from one boiler to another.
For this final rule, EPA's analysis is more exhaustive than these
earlier studies. It encompassed more boilers, longer data streams, and
better quality data. Separate graphs were generated for every boiler or
common stack in the LNB Application Database, plotting NOX hourly
emission rates as a function of hourly load, using long-term quality
assured CEM data. To allow comparison of uncontrolled and controlled
emissions, wherever available, pre- and post-retrofit hourly data were
plotted on the same graph, differentiated by distinct symbols.
A comparison of the pre-retrofit and post-retrofit plots shows
that, with one exception, for both wall-fired boilers and tangentially
fired boilers, if any NOX vs. load relationship existed for
uncontrolled emissions, the installation of LNBs both reduced the
magnitude and shortened the effective range of that relationship.
As discussed above, EPA also developed a statistical measure
(R2) of the strength of the correlation between NOX and
boiler load, assuming a linear relationship. This statistical analysis
corroborates the visual assessment of the data plots. For the post-
retrofit hourly average NOX emission rate measurements, only 13%
of the wall-
[[Page 67129]]
fired and none of the tangentially fired boilers or common stacks had
an R2 of 40% or higher, suggesting possible presence of a
predictable correlation. Even though this analysis confirms that the
occurrence of a NOX-load relationship is generally slight and for
only some boilers, to eliminate all concerns in this regard, EPA has
based the final rule on load-weighted annual average NOX emission
rates (instead of a straight average emission rates) observed over the
post-optimization period (instead of the 52-day low NOX period).
Examples to show presence of a NOX/load relationship: A number
of commenters provided data intended to demonstrate the presence of a
NOX/load relationship. The submissions either had drawbacks which
rendered their conclusions questionable or corroborate EPA's finding
that the installation and operation of LNBs generally dampen any pre-
retrofit correlation of NOX and boiler load and, in many
instances, virtually eliminate any long-term relationship. The salient
aspects of each submission and EPA's responses are summarized below:
Docket item IV-D-020, Figure 20: Using CEM data for the period 06/
30/95 through 07/18/95, this submission included a regression analysis
for a 550 MW wall-fired boiler retrofit with LNBs. The regression model
fit NOX emissions to boiler load during the period analyzed. The
R2 statistic, which captures the explanatory power of the
regression model, was 77.3%, indicative of a good fit with the data.
There were a number of drawbacks, however, with the analysis.
First, the period analyzed represents only 19 calendar days. This is
too short a period to adequately represent long-term performance or to
distinguish a strong, but transitory, NOX/load correlation from a
persistent NOX/load correlation.
Second, the data plot shows a wide range of NOX emission rate
points at zero load. These appear to be spurious measurements which
improperly dominated the regression results.
Docket item IV-G-14, Tables 1-4 and Figures 1 and 2: This
submission included ``before LNB'' and ``after LNB'' regression
analyses for a 80 MW tangentially fired boiler. The ``before LNB''
regression is based on five-and-a-half months of CEM data and the
``after LNB'' regression is based on eight months of CEM data. During
the ``after LNB'' period, this boiler had to comply with a state-
mandated NOX RACT limit of 0.42 lb/mmBtu on a 24-hr average basis.
The commenter rightly excludes NOX emission data points for
periods when load is zero, which is consistent with EPA's DQO 3D.\8\ In
both the ``before LNB'' and ``after LNB'' case, the highest NOX
emission rate is at minimum load. The R2 statistic in the ``before
LNB'' regression was 57.8%, indicating that the model had moderate
explanatory power, whereas the R2 value in the ``after LNB''
regression was only 29.1%, indicating poor explanatory power.\9\ EPA
believes that this ``before LNB'' and ``after LNB'' comparative
regression analysis illustrates how the installation and operation of
LNBs can dampen any NOX vs. load relationship which may be
observed at uncontrolled boilers.
---------------------------------------------------------------------------
\8\ The commenter applies a cutoff at 5 MW, to exclude periods
when a small positive heat input may be recorded, but boiler load is
actually zero.
\9\ The commenter concludes that the relationship between
NOX and boiler load is much less well-defined after LNB
retrofit, but maintains the relationship still exists based on an
analysis of variance which produces a correlation coefficient of
-0.54.
---------------------------------------------------------------------------
Docket items IV-D-65, Enclosure 8; IV-D-23, Attachment 1; IV-D-73,
Attachment A: Several commenters submitted line plots or histograms of
average and/or maximum NOX emission rates recorded for different
load intervals or ``bins''. There were several problems with these
submissions. First, although they criticize EPA in this regard, the
commenters themselves do not develop any statistical measures of the
association between NOX and load for the data they submit (perhaps
because it, too, fails to demonstrate the presumed relationship). Nor
do they suggest functional representations for their plots.
A second drawback of these submissions is that some of the plots
represent boilers retrofit with LNBs plus separated overfire air. As
noted previously, such applications cannot be considered in this
rulemaking.
Third, while the submitted graphs appear to support the commenters'
statements about the existence of a NOX vs. load relationship for
the boilers analyzed, the use of a single value (whether the average or
maximum) to represent all values in a load range bin is misleading. It
hides the variability within the bin, thereby avoiding the issue of
whether the range of values in one bin are distinguishable from those
in another bin.
To address this issue EPA generated NOX/load box-and-whisker
plots for each boiler or common stack in the LNB Application Database.
The box-and-whisker representation not only shows a mid-point value
(the median), but it also characterizes the range of values found in
each bin by displaying the minimum, maximum, and first and third
quartile values. Where sufficient data were available, separate graphs
were created for NOX/load correlation before and after LNB
retrofits. In response to the commenter's criticism that EPA's earlier
analysis for the proposed rule had used too few load bins, ten load
bins (in 10 percent increments from zero to maximum gross unit load)
were used in all the NOX/load analyses for the final rule.
The box-and-whisker plots reveal so much overlap in NOX values
from bin to bin that drawing conclusions about a NOX/load
relationship is technically inappropriate. This is particularly true
for the post-retrofit situation.
Docket items IV-D-73, p. 5 and Attachment A; IV-D-65, p. 32: One
utility reported that the NOX emission rate guarantee, in its
contract for LNBs on a 375 MW wall-fired boiler (E.D. Edwards 3),
``[is] designed specifically to achieve specific NOX rates at
specific loads.'' The annual NOX emission rate is guaranteed to
meet 0.50 lb/mmBtu based on a specified capacity. The NOX emission
rate guarantees for particular loads range from 0.28 lb/mmBtu at 40% of
MCR (150 MW) to 0.63 lb/mmBtu at 100% of MCR (375 MW). The commenter
also submitted graphs depicting the ``remarkable NOX vs. load
relationship'' for another wall-fired boiler (E.D. Edwards 2). The
graphs plotted the average, maximum, and minimum hourly NOX
emission rates recorded in each of ten load bins for the four quarters
of 1995 as well as the entire year.
EPA has analyzed all the post-retrofit CEM data for E.D. Edwards 2
to evaluate the extent of a discernible NOX/load relationship. The
analysis confirmed the existence of a well-defined NOX vs. load
relationship for this boiler, but only in the upper 20% of the load
range (see docket item IV-A-6, Appendix D).
Another commenter noted that Babcock & Wilcox (B&W), a primary
designer of wall-fired boilers and a major LNB vendor in the U.S.,
attests to the existence of a NOX/load correlation. This commenter
said EPA did not find a strong NOX vs. load relationship because
EPA did not examine closely the post-retrofit CEM data for wall-fired
boilers designed by B&W. B&W has stated, ``a definite correlation
[exists] between NOX emissions and boiler load'' (see docket item
IV-D-65, p. 32).
The LNB Application Database contains 18 wall-fired boilers
designed by B&W. Five of these boilers, EPA believes, have also been
retrofit with LNBs manufactured by B&W (Model DRB-XCL). Only one B&W
boiler and none of the B&W LNB retrofits appeared among the wall-fired
boilers or common
[[Page 67130]]
stacks that had an R\2\ of 40% or higher for the correlation of post-
retrofit hourly average NOX emission rate measurements with boiler
load.
3. Analysis Method Used to Establish Reasonably Achievable Emission
Limitations for Phase II, Group 1 Boilers
i. Background
For the proposed rule, EPA used a three-step analytical procedure
for establishing reasonably achievable annual emission limitations for
the populations of wall-fired boilers and tangentially-fired boilers,
retrofit with LNBs, that would be subject to any revised emission
limitations (i.e., those units subject to NOX emission limitations
only in Phase II). The first step (Model Building) consisted of
deriving linear regression equations, one for wall-fired boilers and
another for tangentially fired boilers, that captured the percent
reduction in post-retrofit load-weighted annual average NOX
emission rate as a function of the uncontrolled emission rate for
boilers in the LNB Application Database. The second step (Calculation
of Achievable Emission Rates) was to enter the uncontrolled emission
rates of the Phase II boilers into the regression equations in order to
derive the controlled NOX emission rate that each boiler could be
expected to achieve by LNB retrofit. Using the resulting set of
achievable emission rates, the third step was to identify the annual
emission limitation that a specified percentage (i.e., 85 to 90%) of
the Phase II boilers could achieve. Separate limits were identified for
wall-fired boilers and for tangentially fired boilers.
This three-step procedure afforded several advantages. First, by
using regression equations, the estimates of achievable emission rates
were not rough extrapolations from average Phase I post-retrofit
experience but were estimates specifically tailored to the pre-retrofit
NOX emission rates actually observed at the Phase II units. As
shown in Table 12 in the preamble to the proposed rule (61 FR 1452),
Phase II units typically operate at lower uncontrolled emission rates
than Phase I units (i.e., 23% lower for wall-fired boilers and 18%
lower for tangentially fired boilers) so a simple extrapolation of the
experience of the mostly Phase I units in the LNB Application Database
would significantly underestimate the number of boilers that would be
expected to achieve a given emission limitation.
Second, using regression models also allowed for quantitative,
statistical evaluation of the explanatory power implicit in the
resulting estimates and enabled objective comparison of different
analytical approaches. Incorporating load-weighted annual averaging
into Step 1 of the procedure meant that any NOX/load effects would
be factored into the model.
Furthermore, responding to comments criticizing the proposed rule
for basing the regression model on 52 days of low NOX post-
retrofit emission data, the final rule uses the much longer post-
optimization data stream to build the regression equations. Use of this
longer data stream increases confidence that the regression equations
model the long-term behavior of boilers in the LNB Application
Database.
The Agency received detailed comments from utilities and a utility
association on three data issues and related technical components of
EPA's analysis methods. First, commenters questioned EPA's use of
short-term data to characterize pre-retrofit uncontrolled emission
levels when, for some boilers, long-term data were available.
Uncontrolled emission rates are used in Step 1 (Model Building) and
Step 2 (Calculation of Achievable Emission Rates) of EPA's analytical
procedure for deriving the annual emission limitations. Second, in a
related data issue, commenters believed that the uncontrolled emission
rates used for the affected population of Phase II boilers were biased
low, due partly to a misperception about how controlled NSPS units were
treated in Step 2. (Controlled NSPS units have older LNBs or some other
early type of NOX combustion control installed as original
equipment, so their measured baseline emission rates do not represent
uncontrolled emissions.) Third, commenters disagreed with or raised
questions about certain technical assumptions built into the models--
namely, the methods used to estimate percent NOX reduction outside
the range of the observed model inputs and the form of the regression
model. Finally, commenters said that monitored emissions data from
boilers sharing a common stack should be used cautiously, if at all,
when evaluating LNB performance and offered suggestions on how to
properly assess such measurements.
Salient background points regarding EPA's treatment of certain data
issues for the proposed rule are summarized in the paragraphs below.
The subsequent sections of this preamble discuss the comments more
fully, EPA's response to the issues raised, and how these data and
technical components are treated in the analysis supporting the final
rule.
EPA is fully cognizant that ``long-term data collection is the
definitive method to determine actual NOX reduction
characteristics of a low NOX combustion system'' and that DOE
Clean Coal Technology Demonstrations routinely collect long-term CEM
data to measure the baseline uncontrolled emission rate (see docket
item II-I-99, p. 8). At the time of the proposed rule analysis,
however, EPA had quality assured pre-retrofit long-term CEM data for
only 21% of the boilers in the LNB Application Database. Such CEM data
were unavailable for most of the wall-fired boilers (21 of 24) and over
half of the tangentially fired boilers (5 of 9). Generally, CEM data on
uncontrolled emissions were unavailable because the LNB retrofit had
begun prior to certification of the CEM system in accordance with 40
CFR part 75. EPA decided that it was preferable to use consistent,
quality assured, short-term measurements of uncontrolled emission rates
based on EPA Reference Method, certified CEM, or other test data rather
than to limit the LNB Application Database to only those boilers for
which EPA had quality assured, pre-retrofit, long-term CEM data. EPA
also rejected the possible option of using short-term data for some
boilers and long-term data for other boilers for the reasons explained
in detail in the next section of this preamble.
To assure that consistent data of known high-quality was used for
the model projections, EPA identified specific sources of acceptable
short-term uncontrolled emission rate data. These sources, listed in
priority order, are: (1) Short-term CEM data reported in monitor
certification review (CREV) tests (see docket item II-A-9); (2)
utility-reported CEM or EPA Reference Method test data provided on the
Acid Rain Cost Form for NOX Control Costs; and (3) other short-
term CEM or test data provided by utilities, generally as a correction
or update to data previously submitted to EPA.
For the proposed rule analysis, EPA obtained acceptable short-term
uncontrolled emission rate data for all units in the LNB Application
Database and for 69% of the Phase II boiler population. For the
proposal, EPA used uncontrolled emission rates based on long-term CEM
data or, as a last resort, estimates in the National Utility Reference
File (NURF), which were developed using emission factors, for the other
boilers in the Phase II population. For the final rule, EPA has located
substantial additional quality assured short-term uncontrolled emission
rate data and has discontinued using both long-term CEM and NURF
[[Page 67131]]
estimates for the Step 2 (Calculation of Achievable Emission Rates)
projections.
ii. Short-term vs. Long-Term Uncontrolled Emission Rate Data
Comment/Analysis. EPA received approximately 7 comment letters
(from 6 utilities and 1 utility association) on the use of short-term
uncontrolled emission rate data for assessing the performance of LNBs
applied to Group 1 boilers. Concern was expressed that using short-term
uncontrolled emission rates to build the regression equation would
cause the model to overestimate or, at least wrongly estimate, the
achievable reductions, because short-term uncontrolled emissions would
tend to reflect full-load uncontrolled emissions whereas the
corresponding controlled emissions values, used to build the regression
model, would represent the ``average of 1248 points at different
loads'' (see docket item IV-D-65, p. 51). The comments raised two
issues:
(1) Misuse of Short-Term Data: EPA used short-term uncontrolled
emission rate data even when, for some boilers, quality assured long-
term CEM data were available for determining pre-retrofit uncontrolled
emission rates. (See docket items IV-D-38, p. 3 and IV-D-65, pp. 50-51.
No commenter suggested, however, that EPA restrict the analysis to only
those boilers for which pre-retrofit, long-term CEM data were
available. EPA notes that one commenter, who recommended using a full
year of pre-retrofit monitoring data, selected CREV emission rates as
the best available substitute for baseline measurements when long-term
CEM data were not available.\10\
---------------------------------------------------------------------------
\10\ This commenter combined annual CEM and CREV baseline
measures when assessing the effect of a fuel switch on NOX
emissions for four boilers the utility owns and operates. The
analysis used annual CEM data for the ``before'' measurement on two
boilers, CREV emission rates for the ``before'' measurement on two
other boilers, and annual CEM data for the ``after'' measurement on
all four boilers (see docket item IV-D-038, Attachment A).
---------------------------------------------------------------------------
(2) Load Cell 10 Approach: Several commenters said the use of
short-term measurements of pre-retrofit uncontrolled emission rates in
both the EPA and DOE studies led to high estimates of uncontrolled
emission rates which, in turn, exaggerated LNB reduction efficiencies
(see, for example, docket items IV-D-11, p. 3 and IV-D-72, p. 3).
(Traditionally, LNB percent reduction efficiency has been measured on a
consistent pre-retrofit/post-retrofit basis, normally short-term to
short-term though occasionally long-term to long-term.) The load cell
10 approach was suggested by one commenter as a solution: its argument
runs as follows. In building the regression model, since EPA used
short-term uncontrolled emission rate data, which tends to be obtained
at full load, for consistency the post-retrofit controlled emission
values should have been ``the average NOX data in load cell 10 or
the highest load cell experienced at the boiler,'' not the average of
controlled emission values at all load levels. (See docket item IV-D-
65, p. 52.)
Response. (1) Misuse of Short-Term Data: For analytical, practical,
and statistical reasons, EPA chose to use short-term uncontrolled
emission data rather than only long-term uncontrolled emission data or
a combination of short-term and long-term uncontrolled emission data.
For analytical consistency, it is desirable, if not essential, for all
uncontrolled emission data to be long-term or short-term but not a
mixture of both. Maintaining this consistency across both the LNB
Application Database and the Phase II, Group 1 boiler population
database provides the logical underpinnings for drawing inferences from
the regression model to the Phase II data set, insofar as the
uncontrolled emission rate represents the independent variable in the
regression model. From an analytical standpoint it is perfectly
acceptable for the regression model's dependent variable (controlled
emission rate) to be based on a different duration standard (e.g.,
long-term as opposed to short-term) than the independent variable.
Practical and statistical considerations favored the selection of
short-term data over long-term data. In particular, it was not possible
to obtain quality assured long-term uncontrolled data for many units
because CEM requirements for Phase I boilers were generally coincident
with LNB retrofits. Fewer data points would have reduced the
statistical confidence in the conclusions drawn from the data.
Some commenters were apparently unaware of certain practical data
limitations. One commenter said, ``utilities have been required to
provide EPA with CEM data since at least January 1, 1995 (pursuant to
40 CFR part 75 . . . (so) the CEM NOX data should be used in most
instances for uncontrolled emissions'' (see docket item IV-D-65, p.
51). However, while desirable, this approach was not a practical
option. Since utilities are not required to report even approximate
dates of LNB installations for Phase II units to EPA, as they did in
Phase I on the Acid Rain Cost Form for NOX Control Costs, it is
exceedingly difficult to accurately determine the control status of
each unit, the date and hour on which a specific unit is being taken
off-line for installation of LNBs, and the end (i.e., date and hour) of
the pre-retrofit monitoring period. In contrast, reliable information
on unit control status accompanies the short-term uncontrolled emission
data in the CREV database since utilities are required to report the
type of NOX controls, if any, on each unit to EPA with the annual
certification review test data.
Using the short-term CREV data for the final rulemaking, EPA was
able to amass uncontrolled NOX emission rates for 85% of the Phase
II, Group 1 boilers. This includes virtually every Phase II, Group 1
boiler whose uncontrolled emissions were not otherwise obscured by
complex ``mixed'' common stack arrangements, either with respect to
boiler type (e.g., wall and tangentially fired boilers sharing a common
stack) or control status (e.g., controlled and uncontrolled boilers
sharing a common stack). Quality assured short-term uncontrolled
emission data were obtained for an additional 13% of the Phase II,
Group 1 boilers from other acceptable sources. In all, about 98% of the
affected Phase II, Group 1 boilers were included in the Step 2 analysis
(Calculation of Achievable Emission Rates) for the final rule.
Notwithstanding commenters' concerns, the ability of the regression
model to estimate achievable NOX emission limits is not diminished
by using short-term uncontrolled emission values as the regression
model's independent variable. This is a consequence of the structure of
the model. In the model building stage (Step 1) of EPA's analytical
procedure a functional relationship is established between short-term
uncontrolled emissions and the post-optimization load-weighted
controlled average emission rate achieved by boilers in the LNB
Application Database. In Step 2, as long as the Phase II short-term
uncontrolled emission values that are fed into the regression equation
remain within the range for which the model was designed, the model's
ability to estimate the corresponding achievable post-optimization
annual emission rate should remain unimpaired.
To evaluate the effect of using short-term rather than long-term
data for uncontrolled emission rate on the annual emission limitations
derived from the 3-step analytical procedure, EPA was able to assemble
a database of 18 boilers containing long-term pre-
[[Page 67132]]
retrofit emission rate values.\11\ This database was used to perform
sensitivity tests on the effect of using long-term vs. short-term
measurements of uncontrolled emission rate on the projections of the
number of Phase II, Group 1 boilers that could comply with various
performance standards. For these tests, EPA used the long-term, instead
of short-term, measurements for uncontrolled emission rate in Step 1
(Model Building) wherever such pre-retrofit data were available (18 out
of 53 boilers). The R2 values for the resulting regression models
based on load-weighted annual average emission rates over the post-
optimization period were 65.3% (wall-fired boilers) and 78.9%
(tangentially fired boilers), indicating acceptable fit (see docket
item IV-A-6, Tables 4-6a and 4-6b). Applying these models to the Phase
II, Group 1 data set of uncontrolled emission rates produced the
results shown in docket item IV-A-6, Tables 4-7a and 4-7b. Within the
primary range of interest (i.e., from 80th to 90th percentiles), the
percentage of boilers estimated to achieve a specified emission limit
using the long-term data typically varies by less than 2% (and not more
than 5%) from the percentage derived using strictly short-term data.
Both positive and negative differences occur, depending on the exact
percentile and type of boiler, suggesting the emission limit could be
lowered in some instances and raised in others. EPA concludes that
using short-term measurements of uncontrolled emission rate has not
systematically nor significantly lowered the resulting estimates of
controlled emission rates achievable by Phase II, Group 1 boilers
retrofit with LNBs.
---------------------------------------------------------------------------
\11\ Long-term pre-retrofit emission rate values were defined
from the hourly CEM data as follows. The pre-retrofit period, which
is called ``pre-retrofit minus 30 days'' (abbreviated as ``30-day
pre-rate'' in tabular column headings), starts at the beginning of
the CEM data set. Because some uncertainty exists as to the exact
date of the LNB retrofit, EPA used only quality assured CEM data
recorded more than 30 calendar days before the primary boiler outage
for installation of LNBs. These days are excluded to assure that no
post-retrofit data are mixed with pre-retrofit data in the baseline
measurement. Consistent with the post-retrofit situation, EPA
included only boilers which had at least 1,248 hours (or 52 days) of
quality assured pre-retrofit CEM data.
---------------------------------------------------------------------------
(2) Load Cell 10: Use of load weighting in EPA's regression model
makes the load cell 10 restriction unnecessary. As noted in the
preceding paragraph, the regression model establishes a functional
relationship between the short-term uncontrolled emission rate and the
load-weighted annual average emission rate maintained over the post-
optimization period. If, as the commenter maintains, the load level can
be assumed to relatively constant for all the short-term uncontrolled
emission data (i.e., at full load), all the more reason exists for the
functional relationship captured in EPA's regression equation to remain
intact.
The load cell 10 approach would establish a functional relationship
between the short-term uncontrolled emission rate and the long-term
controlled emission rate achieved when the unit is operating at
essentially full load (i.e., in ``load cell 10,'' at 90-100% of total
unit operating load).\12\ The dependent variable in this regression
model would be the unit's average ``load cell 10'' (or full-load)
controlled emission rate. This approach would discard all post-retrofit
CEM hourly data recorded when the unit is operating in load cells 1
through 9 and thus, would not be representative of unit's average
emission rate over a calendar year. This would be inconsistent with the
purpose under section 407(b)(2) of analyzing LNB performance, which is
to determine whether the existing Group 1 emission limitation applied
on any annual average basis should be made more stringent.
---------------------------------------------------------------------------
\12\ The ``load cell 10 approach'' uses only data recorded for
the highest load cell experienced at the boiler, which is normally
load cell 10.
---------------------------------------------------------------------------
EPA notes that for boilers where NOX emission rate increases
with increasing load, the achievable full-load emission rate determined
using the load cell 10 approach would be higher than the average
emission rate observed over varying boiler loads throughout a year. At
least 25% of the wall-fired boilers in the LNB Application Database
operated at full load for less than 20% of total operating hours in
1995. Basing the annual performance standard on an achievable full-load
emission rate would inappropriately bias the emission limitation since
many boilers are typically operating at lower loads most of the time.
iii. Potential for Low Bias in Phase II Uncontrolled Emission Rate
Estimates/Treatment of NSPS Units
Comments/Analysis: EPA received approximately 5 comment letters
(from 3 utilities and 2 utility associations) saying that EPA's
estimates of uncontrolled emission rates for the Phase II boiler
population appeared too low. The commenters cited different reasons for
this outcome and some submitted unit-specific estimates of uncontrolled
emission rate (see, for example, docket item IV-D-39, p. 3). Several
commenters attributed the seemingly low rates to the inclusion of NSPS
units in the Phase II boiler population baseline of uncontrolled
emission rates. As one commenter stated, ``the NSPS units are by
original design low NOX emitters . . . and (if included), the
overall Phase II, Group 1 boiler baseline rate will be artificially
biased downward and will lead to conclusions that overstate the ability
of both non-NSPS and NSPS units to achieve the final emission limit for
this boiler group'' (See docket item IV-D-72, p. 3).
Response: These commenters correctly noted that the technical
support document for the proposed rule does not contain a separate
baseline for NSPS units nor any explicit discussion of the how these
units are treated in Step 1 (Model Building) and Step 2 (Calculation of
Achievable Emission Rates) of EPA's projection analyses. EPA developed
a table comparing the average uncontrolled emission rates, by boiler
category, for the Phase II, Group 1 boiler population with and without
NSPS Subpart D and Subpart Da units against the Phase I, Group 1 boiler
population (see docket item IV-A-10). This table shows that average
uncontrolled emission rate for the Phase II population excluding units
identified as ``NSPS-vintage units'' \13\ is definitely lower than the
average uncontrolled emission rate for the Phase I population: the
difference is estimated as 10% for wall-fired boilers and 9% for
tangentially fired boilers.
---------------------------------------------------------------------------
\13\ This classification of ``NSPS-vintage units'' was based on
boiler age as reported in the NURF data file.
---------------------------------------------------------------------------
Subsequent to the rule proposal, EPA obtained additional data to
refine both the classification of Phase II units subject to NSPS
NOX requirements, including both Subpart D and Subpart Da, and the
description of any pre-existing NOX combustion controls installed
on these units. EPA notes that since no percent reduction standard for
NOX applies to Subpart D boilers, Subpart D units frequently do
not have combustion controls installed as original equipment. Subpart
Da boilers are required to achieve a specified percent reduction for
NOX, so Subpart Da units generally had some early form of NOX
combustion controls installed prior to November 15, 1990.
As discussed previously in section III.A.1 of this preamble, EPA
has excluded controlled NSPS boilers from the model building regression
analyses because their measured baseline emission rates do not
represent uncontrolled emissions. However, EPA has included all NSPS
boilers, controlled and uncontrolled, in the
[[Page 67133]]
Phase II boiler data set on which the regression models are applied
because coal-fired NSPS boilers are subject to this rulemaking.
NSPS boilers are by original design inherently lower NOX
emitters and have larger furnace volumes per MW than most pre-NSPS
boilers which makes it easier for NSPS boilers, when retrofit with
current LNB technology, to achieve specified levels of controlled
NOX emission rates.\14\ The only NSPS boiler for which EPA has
long-term post-retrofit CEM data (North Valmy 1) corroborates the
assessment that NSPS boilers, when retrofit with current LNB
technology, can generally achieve lower NOX levels than most pre-
NSPS boilers. North Valmy 1 sustained an average controlled emission
rate of 0.264 for calendar year 1995 (see docket item IV-A-9). Although
several commenters discussed this particular LNB installation, none
provided any information which would suggest this boiler is not typical
of controlled NSPS boilers.
---------------------------------------------------------------------------
\14\ Reference: Smith, L. 1988. Evaluation of Radian/EPA
NOX Reduction Estimation Procedures. ETEC-88-20046. February.
---------------------------------------------------------------------------
iv. Technical Assumptions Used in Group 1 Regression Model
EPA received approximately 3 comment letters (from 2 utilities and
one utility association) on certain technical assumptions in the Group
1 regression model approach--namely, the methods used to estimate
percent NOX reduction outside the range of the observed model
inputs and the form of the regression model.
a. Estimation Method for Units with High Uncontrolled Emission Rates
Comment/Analysis: Commenters said that percent NOX reductions
and controlled emission rates that seemed to be predicted by EPA's
regression model at theoretically high values of uncontrolled emission
rates were ``curious'' and seemingly contrary to experience and common
sense.
Any regression model is statistically verifiable only for the range
of data used to construct the model. Not realizing that EPA had assumed
the percent NOX reduction for any Phase II boilers with
uncontrolled emission rates above the highest value in the LNB
Application Database was equal to the percent NOX reduction
estimated for the highest data point (see docket item II-F-2, p. 4-3),
some commenters said the model ``predicts NOX control scenarios
that lead to absurd results'' such that if one can only increase
uncontrolled NOX emissions to a sufficiently high level, one could
achieve 100% NOX removal!'' (see docket item IV-D-65, p. 43 and
IV-G-16, p. 6).
Response: EPA's failure in the proposal to explicitly state a
caveat that is routinely assumed in regression analysis led these
commenters to draw erroneous conclusions from the model. The required
caveat is that the statistically verifiable fit of a regression model
is only assured within the range of the data actually used to construct
the model. Thus, for the regression equations used in the proposed
rule, the statistically verifiable range (in uncontrolled emission
rates) for wall-fired boilers was from 0.51 lb/mmBtu to 1.34 lb/mmBtu
and for tangentially fired boilers was from 0.48 lb/mmBtu to 0.66 lb/
mmBtu. With the addition of 20 boilers to the LNB Application Database
in support of the final rulemaking, the current upper limits on the
ranges have increased to 1.41 lb/mmBtu for wall-fired, and to 0.86 lb/
mmBtu for tangentially fired boilers (see docket item IV-A-6, Tables 3-
1a and 3-1b).
Had the commenters been cognizant of the caveat described in the
previous paragraph, they probably would not have drawn the admittedly
``curious'' conclusions noted above. Further, had they assumed proper
application of the model instead of presuming improper application,
they would have noted that the model was not applied outside its
effective range.
Similarly, the commenters were also troubled by the seeming
implication that the mathematical form of the regression seemed to pre-
ordain that emissions could never exceed a certain maximum bound. As
the commenter in docket item IV-D-65 puts it: The model predicts ``. .
. that controlled emissions at wall-fired boilers will never exceed
0.454 lb/mmBtu.'' In fact, based on existing data, the model simply
shows a maximum predicted emission reduction over the model's
statistically verifiable range. For points outside the range of the
model, no specific bound is implied, and the maximum observed emission
reduction was not exceeded.
As in the proposal, when estimating the controlled emission rates
for Phase II, Group 1 boilers with uncontrolled emission rates higher
than the verifiable range of the model, EPA made the following
assumption \15\: the percent NOX emission reduction for such
boilers was assumed to be no greater than the reduction obtained by the
boiler with the highest uncontrolled emission rate in the LNB
Application Database. In effect, this assumption would lead to emission
limits that are less stringent than if it were assumed that the
emission reductions for such boilers could exceed those of boilers in
the LNB Application Database.
---------------------------------------------------------------------------
\15\ Only 1 wall-fired boiler and 3 tangentially fired boilers
in the Phase II boiler data set (representing less than 1% and less
than 2%, respectively, of the affected populations) have measured
uncontrolled emission rates higher than the range used to construct
the regression model and thus fall in this category.
---------------------------------------------------------------------------
b. Form of the Regression Model
In both the analysis for the proposed and final rules, EPA
considered two alternative forms of the regression models used to
predict the achievable controlled emission rates from uncontrolled
boiler NOX emission rates:
Model #1 (One-step approach): Direct linear fit, regressing
controlled emission rate on uncontrolled emission rate.
Model #2 (Two-step approach): Step 1--Direct linear fit, regressing
percent NOX reduction on uncontrolled emission rate. Step 2--
Controlled emission rate is computed from the percent reduction derived
in Step 1.
EPA chose Model #2 because the regression equations derived using
this model explain the data better than those derived using Model #1.
Statistically, this is expressed in the higher ``R \2\ value'' of Model
#2 (R \2\=73.1% for wall-fired boilers; R \2\=70.7% for tangentially-
fired boilers) as compared to Model #1 (R \2\=59.7% for wall-fired
boilers; R \2\=17.0% for tangentially-fired boilers) (see docket item
IV-A-6, Tables 4-9a and 4-9b).
Comment/Analysis: A commenter criticized EPA's choice of Model #2,
saying that it models the wrong parameter: ``. . . while the key issue
in this rulemaking is the level of controlled emissions at Phase II,
Group 1 boilers, . . . (EPA's) model is designed to predict NOX
removal efficiency--a related but secondary parameter'' (docket item
IV-D-65, p. 45). Consequently, the commenter questioned the
meaningfulness of a superior R \2\ value from a model that regresses
percent reduction on uncontrolled emissions, when the true parameter of
interest is not percent emission reduction but controlled emissions:
``. . . just because Model 2 predicts removal efficiency better than
Model 1 predicts controlled emissions does not mean that Model 2
predicts controlled emissions better than Model 1'' (docket item IV-D-
65, pp. 45-46).
Response: While on the surface this criticism appears plausible, on
further investigation it is incorrect because the two-step approach of
Model #2 is algebraically equivalent to a one-step second order linear
regression model that directly regresses controlled
[[Page 67134]]
emissions on uncontrolled emissions.16 Thus, although its two-step
formulation makes Model #2 appear not to regress controlled emissions
on uncontrolled emissions, in actuality, by simply restating Model #2
in its se
This text is long and has been trimmed here. Open the source document for the complete record.
This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.