United States Sugar Corp v. EPA , 830 F.3d 579 ( 2016 )

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued December 3, 2015 Decided July 29, 2016
No. 11-1108
UNITED STATES SUGAR CORPORATION,
PETITIONER
v.
ENVIRONMENTAL PROTECTION AGENCY,
RESPONDENT
AMERICAN CHEMISTRY COUNCIL, ET AL.,
INTERVENORS
Consolidated with 11-1124, 11-1134, 11-1142, 11-1145,
11-1159, 11-1165, 11-1172, 11-1174, 11-1181, 13-1086,
13-1087, 13-1091, 13-1092, 13-1096, 13-1097, 13-1098,
13-1099, 13-1100, 13-1103
On Petitions for Review of Final Action of the
United States Environmental Protection Agency
William L. Wehrum Jr., David M. Friedland, and
Douglas A. McWilliams argued the causes for Industry
Petitioners. With them on the briefs were Allen A. Kacenjar,
Katy M. Franz, Lisa Marie Jaeger, Sandra Y. Snyder, Peter
H. Wyckoff, Claudia M. O=Brien, Stacey L. VanBelleghem, Eli
Hopson, Jane C. Luxton, Lauren E. Freeman, Elizabeth L.
Horner, William F. Lane, Alan H. McConnell, Timothy S.
Bishop, Kevin G. Desharnais, Chad M. Clamage, Ronald A.
Shipley, Quentin Riegel, Linda E. Kelly, and Jeffrey A.
Knight. Rachel Brand, Leslie A. Hulse, Harry M. Ng, Scott J.
Stone, John P. Wagner, and Lee B. Zeugin entered
appearances.
James S. Pew and Sanjay Narayan were on the briefs for
Environmental Petitioners. Neil Gormley entered an
appearance.
Perry M. Rosen and Norman L. Rave, Jr., Attorneys, U.S.
Department of Justice, argued the causes for respondent.
With them on the brief was John C. Cruden, Assistant
Attorney General, and Norman L. Rave, Jr., Attorney. Sam
Hirsch and Madeline P. Fleisher, Attorneys, entered
appearances.
James S. Pew and Neil Gormley argued the causes for
Environmental Respondent-Intervenors. With them on the
briefs was Sanjay Narayan.
William L. Wehrum, Quentin Riegel, Linda E. Kelly,
Patrick Forrest, Douglas A. McWilliams, Peter H. Wyckoff,
Jeffrey A. Knight, Claudia M. O=Brien, Stacey L.
VanBelleghem, Lisa Marie Jaeger, Sandra Y. Snyder, David
M. Friedland, William F. Lane, Alan H. McConnell, Ronald
A. Shipley, Carol F. McCabe, Suzanne Ilene Schiller, Michael
Dillon, Charles Howland Knauss, Shannon S. Broome,
Timothy S. Bishop, Kevin G. Desharnais, Chad M. Clamage,
Lauren E. Freeman, Elizabeth L. Horner, Larry B. Alexander,
and Leslie A. Hulse were on the brief for Industry Intervenor-
Respondents. Allen A. Kacenjar Jr., Rachel L. Brand, Harry
M. Ng, Scott J. Stone, John P. Wagner, and Lee B. Zeugin
entered appearances.
No. 11-1125
AMERICAN FOREST & PAPER ASSOCIATION, ET AL.,
PETITIONERS
v.
ENVIRONMENTAL PROTECTION AGENCY,
RESPONDENT
AMERICAN CHEMISTRY COUNCIL, ET AL.,
INTERVENORS
Consolidated with 11-1140, 11-1144, 11-1154, 11-1155,
11-1161, 11-1171, 11-1173, 11-1180, 11-1183, 11-1188,
13-1111, 13-1113, 13-1114, 13-1116, 13-1118, 13-1119,
13-1120, 13-1121, 13-1123, 13-1124, 13-1127
On Petitions for Review of a Final Action of the
United States Environmental Protection Agency
William L. Wehrum, Douglas A. McWilliams, and Jason
T. Morgan argued the causes for Industry Petitioners. On the
briefs were Richard G. Stoll, Leslie A. Hulse, Lisa Marie
Jaeger, Sandra Y. Snyder, Peter H. Wyckoff, Jeffrey A.
Knight, David M. Friedland, Jessalee Landfried, Michael B.
Wigmore, Ronald A. Shipley, Chet M. Thompson, Linda E.
Kelly, Quentin Riegel, William F. Lane, Alan H. McConnell,
Carol F. McCabe, Suzanne Ilene Schiller, and Michael
Dillon. David Y. Chung, Rachel L. Brand, Julia L. German,
Jeffrey W. Leppo, and Jane C. Luxton entered appearances.
Neil Gormley and James S. Pew argued the causes and
filed the briefs for Environmental Petitioners.
Perry M. Rosen and Norman L. Rave, Jr., Attorneys, U.S.
Department of Justice, argued the causes for respondent.
With them on the brief was John C. Cruden, Assistant
Attorney General. Madeline P. Fleisher, Attorney, entered an
appearance.
James S. Pew and Neil Gormley argued the causes and
filed the briefs for Environmental Respondent-Intervenors.
David M. Friedland and William L. Wehrum argued the
causes for Industry Intervenor-Respondents. With them on
the briefs were Jessalee Landfried, Leslie A. Hulse, Richard
G. Stoll, Ronald A. Shipley, William F. Lane, Alan H.
McConnell, James T. Morgan, Lisa Marie Jaeger, Sandra Y.
Snyder, Jeffrey A. Knight, Shannon S. Broome, Carol
McCabe, Suzanne Ilene Schiller, Michael Dillon, Linda E.
Kelly, Quentin Riegel, and Charles H. Knauss. Scott J. Stone,
Lori A. Rubin, and Jeffrey W. Leppo entered appearances.
No. 11-1141
AMERICAN CHEMISTRY COUNCIL,
PETITIONER
v.
ENVIRONMENTAL PROTECTION AGENCY,
RESPONDENT
AMERICAN FOREST & PAPER ASSOCIATION, ET AL.,
INTERVENORS
Consolidated with 11-1182, 11-1207, 11-1208, 13-1105,
13-1107
On Petitions for Review of a Final Action of the
United States Environmental Protection Agency
William L. Wehrum and David M. Friedland argued the
causes for Industry Petitioners. On the briefs were Lisa Marie
Jaeger, Sandra Y. Snyder, Jeffrey A. Knight, Quentin Riegel,
and Leslie A. Hulse. Harry M. Ng, Scott J. Stone, and John P.
Wagner entered appearances.
Neil Gormley argued the cause for Environmental
Petitioners. With him on the briefs was James S. Pew.
Perry M. Rosen and Norman L. Rave, Jr., Attorneys, U.S.
Department of Justice, argued the causes for respondent.
With them on the brief was John C. Cruden, Assistant
Attorney General. Madeline P. Fleisher, Attorney, entered an
appearance.
David M. Friedland and William L. Wehrum argued the
causes for Industry Intervenor-Respondents. With them on
the briefs were Lisa Marie Jaeger, Sandra Y. Snyder, Jeffrey
A. Knight, William F. Pedersen, Pamela A. Lacey, William F.
Lane, Linda E. Kelly, Quentin Riegel, James W. Conrad, Jr.,
and Leslie A. Hulse. Harry M. Ng, Scott J. Stone, John P.
Wagner, and Nidhi J. Thakar entered appearances.
James S. Pew and Neil Gormley were on the brief for
Environmental Respondent-Intervenors.
Before: HENDERSON, BROWN, and GRIFFITH, Circuit
Judges.
PER CURIAM: In these consolidated petitions for review,
we address approximately thirty challenges to three
regulations promulgated by the United States Environmental
Protection Agency (EPA or Agency): (1) the “Major Boilers
Rule,”1 (2) the “Area Boilers Rule,”2 and (3) the
“Commercial/Industrial Solid Waste Incinerators (CISWI)
Rule.”3 Collectively, these rules—all promulgated under the
Clean Air Act (CAA or Act), 42 U.S.C. §§ 7401 et seq.—set
emissions limits on certain combustion machinery known to
release hazardous air pollutants (HAPs). Roughly one-half of
1
National Emission Standards for Hazardous Air Pollutants
for Major Sources: Industrial, Commercial, and Institutional Boilers
and Process Heaters (2011 Major Boilers Rule), 76 Fed. Reg.
15,608 (Mar. 21, 2011), as amended, National Emission Standards
for Hazardous Air Pollutants for Major Sources: Industrial,
Commercial, and Institutional Boilers and Process Heaters (2013
Major Boilers Rule), 78 Fed. Reg. 7,138 (Jan. 31, 2013).
2
National Emission Standards for Hazardous Air Pollutants
for Area Sources: Industrial, Commercial, and Institutional Boilers
(2011 Area Boilers Rule), 76 Fed. Reg. 15,554 (Mar. 21, 2011), as
amended, National Emission Standards for Hazardous Air
Pollutants for Area Sources: Industrial, Commercial, and
Institutional Boilers (2013 Area Boilers Rule), 78 Fed. Reg. 7,488(Feb. 1, 2013).
3
Standards of Performance for New Stationary Sources and
Emission Guidelines for Existing Sources: Commercial and
Industrial Solid Waste Incinerator Units (2011 CISWI Rule), 76
Fed. Reg. 15,704 (Mar. 21, 2011), as amended, Commercial and
Industrial Solid Waste Incineration Units: Reconsideration and
Final Amendments; Non-Hazardous Secondary Materials that Are
Solid Waste (2013 CISWI Rule), 78 Fed. Reg. 9,112 (Feb. 7,
2013).
2
the challenges are advanced by a group of municipal-electric
organizations, industrial-trade associations, oil-and-gas
industry representatives, and other entities that own and
operate boilers, process heaters, and incinerators (Industry
Petitioners). The other one-half are pressed by organizations
interested in safeguarding the environment (Environmental
Petitioners).
I. BACKGROUND
The three rules at issue address a common phenomenon:
when combustion occurs, emissions result. The emissions
include numerous materials, some of which pose risks to the
environment in general and to human health in particular.
Because combustion is an inevitable occurrence in the
machinery that helps to power modern society, the Congress
has authorized the EPA to provide for a regulatory framework
that minimizes the deleterious effects of the incineration
industry while simultaneously allowing it to operate.
In 2013, the EPA finalized its efforts to do so for discrete
types of combustion machinery: boilers, process heaters, and
incinerators. Two of the three rules at issue—the Major
Boilers Rule and the Area Boilers Rule—govern boilers and
process heaters. The former are enclosed devices that use a
controlled flame to heat water and convert it into steam or hot
water. 40 C.F.R. § 63.11237. The latter are also enclosed
devices that use a controlled flame but, instead of generating
steam, they indirectly heat a “process material,” whether
liquid, gas, or solid, or a “heat transfer material” like glycol or
a mixture of glycol and water. Id. For simplicity, our use of
“boilers” covers both machinery types.
The two boiler-specific rules further divide the machinery
into three categories: industrial, commercial, and
3
institutional. See 2011 Area Boilers Rule, 76 Fed. Reg. at
15,557. Industrial boilers are used for manufacturing,
processing, mining, refining, and other similar operations.
See id. Commercial boilers are used by shopping malls,
laundromats, apartment complexes, restaurants, and hotels.
See id. And institutional boilers include those used by, e.g.,
medical centers, schools, churches, prisons, and courthouses.
See id. Collectively, over 200,000 boilers at over 100,000
separate facilities must comply with the standards set out in
the Major Boilers Rule or the Area Boilers Rule.
The third rule that we address—the CISWI Rule—
governs combustion machinery known as “solid waste
incineration unit[s].” 42 U.S.C. § 7429. The Act defines an
incinerator as a “distinct operating unit of any facility” that
burns solid waste from either commercial establishments,
industrial establishments, or the general public. Id.§ 7429(g)(1). An incinerator subjects “waste material” to
“high temperatures until it is reduced to ash.” Incinerator,
NEW OXFORD AMERICAN DICTIONARY 853 (2d ed. 2005).
Incinerators fall into different subcategories and, in the past,
the EPA has issued rules governing many of them, including,
e.g., municipal solid-waste incinerators, medical-waste
incinerators, and sewage-sludge incinerators.4 At issue in the
CISWI Rule are incinerators located in commercial or
industrial facilities that combust solid waste as defined in the
Resource Conservation Recovery Act (RCRA), 42 U.S.C.
§§ 6901 et seq. See 2011 CISWI Rule, 76 Fed. Reg. at
15,706.
4
See, e.g., 40 C.F.R. pt. 60, subpts. Cd, Ce, Eb, AAAA,
BBBB, EEEE, FFFF, LLLL, MMMM.
4
A. THE CLEAN AIR ACT, 42 U.S.C. §§ 7401 ET SEQ.
Enacted “to 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,” 42
U.S.C. § 7401(b)(1), the Act has been amended several times
since the Congress first attempted to control air pollution via
legislation in 1963. In 1970, the Congress required the EPA
to identify and publish a list of HAPs, which the CAA defined
as substances that increase “mortality,” “serious irreversible”
illness, or “incapacitating reversible” illness. Clean Air
Amendments of 1970, Pub. L. No. 91-604, § 4(a), 84 Stat.
1676, 1685 (1970). The EPA had to set emission limits for
every HAP based on the risk it posed to human health. See
Sierra Club v. EPA (Sierra Club I), 353 F.3d 976, 979 (D.C.
Cir. 2004). In other words, the EPA was to “consider[] levels
of HAPs at which health effects are observed, factor[] in an
ample margin of safety to protect the public health, and set
emission restrictions accordingly.” Id. (quotation marks
omitted).
The risk-focused approach to capping HAP emissions left
something to be desired. “In light of unrealistic time frames
and scientific uncertain[t]y over which substances posed a
threat to public health,” the EPA “only listed eight pollutants
as hazardous between 1970 and 1990,” Nat. Res. Def. Council
v. EPA (NRDC II), 529 F.3d 1077, 1079 (D.C. Cir. 2008), and
set “emission standards for [only] seven of them,” Sierra Club
I, 353 F.3d at 979; see also S. REP. NO. 101-228, at 3 (1989)
(“Very little has been done since the passage of the 1970 Act
to identify and control hazardous air pollutants.”). After
twenty years of the risk-based approach, the Congress went
back to the drawing board and, via the 1990 CAA
Amendments, Pub. L. No. 101-549, 104 Stat. 2399 (1990),
5
established the technology-based approach that governs
today. See Sierra Club I, 353 F.3d at 979.
1. 42 U.S.C. § 7412—“Hazardous Air Pollutants”
The 1990 CAA Amendments overhauled the Act’s
“Hazardous Air Pollutants” provision, codified at 42 U.S.C.
§ 7412. Although earlier iterations of the Act had assigned
HAPs-identification responsibility to the EPA, the slow pace
at which the EPA discharged its duty prompted the Congress
to create a list of pollutants itself.5 See Sierra Club I, 353
F.3d at 979-80 (citing 42 U.S.C. § 7412(b)). After identifying
nearly two hundred HAPs that warranted emissions
restrictions, see 42 U.S.C. § 7412(b)(1), the Congress directed
the EPA, first, to identify the sources of each HAP, see id.§ 7412(c). The Agency then was to set emissions limits for
each source that result in HAPs reduction to the greatest
extent achievable by current technology. See generally Nat’l
Ass’n for Surface Finishing v. EPA, 795 F.3d 1, 4 (D.C. Cir.
2015) (citing 42 U.S.C. § 7412(b)(1), (c), (d)).
a. Identifying and Categorizing HAP Sources
The EPA’s first task is to create HAP-source categories
and subcategories. See 42 U.S.C. § 7412(c). The Act
distinguishes “major” from “area” sources, defining the
former as “any stationary source or group of stationary
sources” that neighbor each other, share common control, and
emit (or have the potential to emit) either ten tons per year or
more of any single HAP or twenty-five tons per year or more
5
The EPA must keep the HAPs list current. See 42 U.S.C.
§ 7412(b)(2), (3).
6
of any HAP combination.6 Id. § 7412(a)(1). The latter are
sources that do not emit enough HAPs to qualify as “major.”
Id. § 7412(a)(2). Although the EPA must set stringent
restrictions on major sources, it has discretion to set more
lenient emissions caps on area sources. See id. § 7412(d)(5).
Apart from the statutory distinction between major and
area sources, the EPA has discretion to differentiate “among
classes, types, and sizes of sources within a category or
subcategory.” Id. § 7412(d)(1). Once the EPA finalizes
HAPs-source categories and subcategories, the CAA
mandates that it draw one final dividing line—between “new”
sources and “existing” sources. See id. § 7412(d)(3). “New”
sources are those “on which construction begins after EPA
publishes emission standards,” Cement Kiln Recycling Coal.
v. EPA, 255 F.3d 855, 858 (D.C. Cir. 2001); most of the
others are “existing” sources, see 42 U.S.C. § 7412(a)(10).
But if an existing source experiences either a physical change
or a change in operation method and the change increases
HAP emissions by more than a de minimis amount, the Act
mandates that the source meet the standards set for new
sources. See id. § 7412(a)(5), (g).
b. Setting Emission Standards for Major Sources—the
“MACT” Standard
After the EPA identifies HAP-source categories and
subcategories, it then sets emissions limits for each. See id.
§ 7412(d)(2). “[W]henever . . . feasible,” the caps must use
numeric HAPs limits. Id. § 7412(h)(4). The size of the
6
The CAA defines “stationary source” as “any building,
structure, facility, or installation which emits or may emit any air
pollutant.” 42 U.S.C. § 7411(a)(3).
7
source—either “major” or “area”—dictates whether the EPA
must set the numeric limit at the most stringent level that
current technology allows or at the level set by “generally
available control technologies.” Id. § 7412(d)(5). For major
sources, the CAA directs the EPA to establish emissions caps
that result in the “the maximum degree of reduction in
emissions” that the EPA determines is “achievable.” Id.
§ 7412(d)(2). We refer to an emissions cap that reflects the
current “maximum achievable control technology” as a
“MACT” standard. See NRDC II, 529 F.3d at 1079. Setting a
MACT standard is a two-step process.
First, the EPA establishes a “MACT floor” for each
category or subcategory. Sierra Club I, 353 F.3d at 980. The
MACT floor ensures that all HAPs sources “at least clean up
their emissions to the level that their best performing peers
have shown can be achieved.” Id. For new sources—those
built after promulgation of a HAPs limit, see 42 U.S.C.
§ 7412(a)(4)—the MACT floor cannot be less stringent than
the emissions levels achieved by the best performing similar
source. Id. § 7412(d)(3). For existing sources in categories or
subcategories that have thirty or more sources, the MACT
floor cannot be less stringent than the average emissions
limits achieved by the best performing 12 per cent of existing
sources in that category or subcategory. Id. § 7412(d)(3)(A).
And for existing sources in categories or subcategories with
fewer than thirty sources, the MACT floor cannot be less
stringent than the average emissions achieved by the best
performing five sources. Id. § 7412(d)(3)(B). When setting
the MACT floor, the EPA considers only the performance of
the cleanest sources in a category or subcategory; it does not
take into account other factors, including the cost of putting a
source in line with its better-performing counterparts. See
Cement Kiln, 255 F.3d at 857-58 (citing Nat’l Lime Ass’n v.
8
EPA, 233 F.3d 625, 629 (D.C. Cir. 2000), as amended on
denial of reh’g, No. 99-1325 (D.C. Cir. Feb. 14, 2001)).
Second, the EPA must determine whether current
technology makes it possible for a source to perform even
better than the best performing similar source or sources. In
other words, the CAA directs the EPA to consider whether it
should set a “beyond-the-floor” MACT standard. Nat’l Lime
Ass’n, 233 F.3d at 629. In determining whether a beyond-the-
floor standard is “achievable,” the Agency must consider
additional factors like “the cost of achieving such emission
reduction,” “any non-air quality health and environmental
impacts” and “energy requirements.” 42 U.S.C. § 7412(d)(2).
It has broad discretion in its determination. See id.; cf. Nat’l
Ass’n of Clean Water Agencies v. EPA (NACWA), 734 F.3d
1115, 1157 (D.C. Cir. 2013) (noting, in section 7429 case,
that “Congress gave EPA broad discretion in considering
whether to go beyond-the-floor”).
c. Setting Emission Standards for Area Sources—the
“GACT” Standard
Although the EPA must cap HAP emissions from major
sources at the “maximum degree of reduction,” see 42 U.S.C.
§ 7412(d)(2), it has discretion to set less stringent caps on
emissions from area sources. Indeed, the EPA need not list
categories of area sources at all unless: (A) it finds that the
sources in that category or subcategory “present[] a threat of
adverse effects” to the environment or human health, see id.§ 7412(c)(1), (3); or (B) control of a particular area source
category or subcategory is necessary to ensure that sources
accounting for at least 90 per cent of the aggregate emissions
of the thirty HAPs the EPA believes “present the greatest
threat to public health in the largest number of urban areas”
9
are subject to CAA control, id. § 7412(c)(3), (k)(3)(B). If it
finds that controlling emissions from a particular area source
subcategory is necessary to achieve a 90 per cent reduction in
the aggregate emissions of any of seven CAA-enumerated
HAPs, section 7412(c)(6) requires the Agency to impose
MACT caps on that subcategory. See id. § 7412(c)(6).
With the exception of section 7412(c)(6)’s MACT-
standard requirement, the EPA need not cap emissions from
area sources at the MACT level. Instead, it may set more
lenient emissions limits based on “generally available control
technologies.” Id. § 7412(d)(5). We refer to these caps as
GACT standards. The Act provides no guidance for setting
GACT standards but the legislative history of the 1990 CAA
Amendments describes GACT “as methods, practices and
techniques [that] are commercially available and appropriate
for application by the sources in the category considering
economic impacts and the technical capabilities of the firms
to operate and maintain the emissions control systems.” S.
REP. NO. 101-228, at 171 (1989). According to the EPA, it
can and will consider the following in setting a GACT
standard:
 “costs and economic impacts . . . , which
[are] particularly important when developing
regulations for source categories that may
have many small businesses . . . ”;
 “the control technologies and management
practices that are generally available to the
area sources in the source category”;
 “the standards applicable to major sources in
the analogous source category to determine if
10
the control technologies and management
practices are transferable and generally
available to area sources”; and
 “technologies and practices at area and major
sources in similar categories to determine
whether such technologies and practices
could be considered generally available for
the area source categories at issue.”
2011 Area Boilers Rule, 76 Fed. Reg. at 15,556. And, unlike
the EPA’s duty to consider a beyond-the-floor MACT
standard, it need not consider a more stringent GACT
standard.
d. Work-Practice and Management-Practice Standards
Although the CAA requires numeric emission standards
where possible, the EPA can “promulgate a design,
equipment, work practice, or operational standard, or
combination thereof” if it determines that a numeric limit is
“not feasible.” 42 U.S.C. § 7412(h)(1). In other words, the
EPA can require that all sources in a given category or
subcategory take a certain action (e.g., conduct a periodic
tune-up) or install certain emissions-control technology (e.g.,
install a fabric filter). Although the EPA has discretion to
impose a work-practice standard, the Act limits it by defining
the operative phrase “not feasible” narrowly to mean:
(A) a hazardous air pollutant or pollutants cannot
be emitted through a conveyance designed
and constructed to emit or capture such
pollutant, or that any requirement for, or use
of, such a conveyance would be inconsistent
with any Federal, State or local law, or
11
(B) the application of measurement methodology
to a particular class of sources is not
practicable due to technological and
economic limitations.
Id. § 7412(h)(2).
Similarly, for area sources, the EPA can impose a
“management-practice standard” in lieu of a numeric GACT
standard. See id. § 7412(d)(5). A management-practice
GACT standard is like a work-practice MACT standard in all
ways but one—the EPA need not consider feasibility when
setting management-practice standards. Compare id.
§ 7412(d)(2), with id. § 7412(d)(5).
2. 42 U.S.C. § 7429—“Solid Waste Combustion”
In addition to amending the Act’s “Hazardous Air
Pollutants” provision, see id. § 7412, the 1990 CAA
Amendments added to the U.S. Code section 7429, titled
“Solid Waste Combustion.” Section 7429 regulates “solid
waste incineration units” generally, see id. § 7429(a)(1)(A),
and CISWI specifically, see id. § 7429(a)(1)(D). Although
section 7412 requires the EPA to control emissions of nearly
two hundred HAPs, see id. § 7412(d)(1), section 7429
mandates that the EPA control emissions from only nine
specific pollutants (as well as opacity, where appropriate),
none of which the Congress included on its initial section
7412 list, see id. § 7429(a)(4); see also Nat. Res. Def. Council
v. EPA (NRDC I), 489 F.3d 1250, 1256 (D.C. Cir. 2007). We
have held that this difference “makes
promulgating . . . standards under [section 7412] and [section
7429] mutually exclusive.” NACWA, 734 F.3d at 1119. In
other words, if a source (or facility) is considered a CISWI
12
and, therefore, regulated under section 7429, it cannot be
regulated under section 7412. See id.
Whether a source falls under section 7412 or section
7429, “the statutory directive on setting MACT standards is
virtually identical.” Id.; see also Nat’l Lime Ass’n, 233 F.3d
at 631. That said, regulation under one section instead of the
other “has practical consequences.” NACWA, 734 F.3d at
1120. For example, section 7412 allows the EPA to impose a
GACT standard for area sources only but section 7429
requires the EPA to impose MACT standards for all covered
units, regardless of their size. Compare 42 U.S.C.
§ 7412(d)(1), (5), with id. § 7429(a)(1)(A); see also NRDC I,
489 F.3d at 1256. Moreover, section 7412 mandates that the
EPA control HAP emissions from “major source[s],” which
the Act defines broadly to include “group[s] of stationary
sources located within a contiguous area and under common
control.” 42 U.S.C. § 7412(a)(1) (emphasis added). Section
7429, in contrast, mandates that the EPA control emissions
from “solid waste incineration unit[s],” which the Act defines
more narrowly as “a distinct operating unit of any facility
which combusts any solid waste material,” 42 U.S.C.
§ 7429(g)(1) (emphases added). And finally, section 7429
does not provide for work-practice standards.
3. 42 U.S.C. §§ 7661 et seq.—“Title V Permits”
Finally, the 1990 CAA Amendments added a provision to
Title V of the Act that requires all owners and operators of
HAP sources to obtain operating permits. See id. § 7661a.
Title V does no more than consolidate “existing air pollution
requirements into a single document, the Title V permit, to
facilitate compliance monitoring” without imposing any new
substantive requirements. Sierra Club v. Leavitt, 368 F.3d
131300, 1302 (11th Cir. 2004). The legislative history of the
1990 CAA Amendments indicates that the Congress required
the “Title V permits” so that the public might “better
determine the requirements to which the source is subject, and
whether the source is meeting those requirements.” S. REP.
NO. 101-228, at 347. Although owners and operators of all
major HAP sources must obtain Title V permits, see generally
42 U.S.C. § 7661a(a), the EPA has discretion to exempt
certain area source categories if it “finds that compliance with
such requirements is impracticable, infeasible, or
unnecessarily burdensome,” id.
B. THE MAJOR BOILERS, AREA BOILERS,
AND CISWI RULES
On March 21, 2011, the EPA issued the first iteration of
all three rules under review. That same day, however, the
EPA announced that it intended to reconsider certain aspects
of each rule. Not long after, multiple parties filed the
petitions for review that we now address. Earlier, the EPA
had concluded its reconsideration and issued the most recent
iteration of the three rules. Because of this procedural quirk,
each “rule” we address is in fact two separate rules—the
EPA’s “final” 2011 version and its “final” 2013 version. The
EPA’s analyses remained mostly consistent from 2011 to
2013 and we indicate, where necessary, the instances in which
the EPA changed course in a significant way.
1. The Major Boilers Rule
The Major Boilers Rule sets HAPs emission caps for all
industrial, commercial, and institutional boilers that emit a
large volume of HAPs. See 2011 Major Boilers Rule, 76 Fed.
Reg. at 15,611. The EPA further divided the major boiler
categories into subcategories based on the primary fuel
14
combusted by the boilers in the subcategory (e.g., coal,
biomass, gas, etc.) and, for some subcategories, based on the
method used to “feed” the fuel into the boiler. See 2013
Major Boilers Rule, 78 Fed. Reg. at 7,144. For most of the
subcategories, the EPA set a numeric MACT standard for four
different HAPs: particulate matter (PM); hydrogen chloride
(HCl); mercury (Hg); and carbon monoxide (CO). See id. at
7,142 tbl.3; No. 11-1108 EPA Br. 9. The EPA used some of
these HAPs—particularly CO—as a surrogate (or proxy) to
set emissions limits for others on the section 7412(b) HAPs
list. See 2013 Major Boilers Rule, 78 Fed. Reg. at 7,144-45.
For the other major boiler subcategories, the EPA set a work-
practice standard (specifically, a tune-up requirement) in lieu
of numeric MACT standards. See 2011 Major Boilers Rule,
76 Fed. Reg. at 15,613.7 The EPA also established a tune-up
work-practice standard to control for dioxin/furan emissions
across all major boiler subcategories. 2013 Major Boilers
Rule, 78 Fed. Reg. at 7,138.
In addition to these emission standards, the Major Boilers
Rule includes several other provisions relevant to the current
petitions for review.
a. The “Upper Prediction Limit”
Several factors complicate the process of setting MACT
floors. The first is the CAA itself, which mandates that all
MACT floors (1) must be achievable, see 42 U.S.C.
7
The four major boiler subcategories for which the EPA
established work-practice standards include “[n]ew and existing
units that have a designed heat input capacity of less than 10
MMBtu/hr, and new and existing units in the Gas 1 (natural
gas/refinery gas) subcategory and in the metal process furnaces
subcategory.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613.
15
§ 7412(d)(2); (2) must ensure continuous regulation of the
covered sources, see id. § 7602(k); and (3) must be no less
stringent than the emissions levels being achieved by the best-
controlled sources, see id. § 7412(d)(3). The second is that no
source emits any HAP at a constant level; rather, HAP
emissions fluctuate over time and for many reasons,
including, e.g., “operation of control technologies, variation in
combustion materials and combustion conditions, variation in
operation of the unit itself, and variation associated with the
emission measurement techniques.” Memorandum from
Stephen D. Page, EPA Director of Air Quality Planning and
Standards, EPA’s Response to Remand of the Record for
Commercial and Industrial Solid Waste Incineration Units
(Page Mem.) (July 14, 2014), at 3 (No. 11-1125 J.A. 1316).
Finally, most sources do not measure their HAP emissions at
all times and under all conditions.8 Id. at 6. Instead, data are
usually gathered when a source conducts a “three-run stack
test.” Id. This test provides three “snapshots” of a source’s
emissions in a limited set of conditions and, accordingly, it
fails to demonstrate accurately a source’s emissions during all
times and under all conditions. Id.
To compensate for the lack of adequate emissions data,
the EPA uses a statistical tool known as the “upper prediction
limit” (UPL) to account for the expected variability in
emissions levels. See 2011 Major Boilers Rule, 76 Fed. Reg.
at 15,630. The UPL, in turn, allows the Agency to set a
MACT floor that is continuously achievable. Id. We discuss
the UPL mechanics at greater length below, see infra § IV.C,
but, in short, the EPA: (1) ranks all sources in a given
8
As discussed below, however, the EPA does allow sources
to demonstrate MACT compliance by use of “continuous
monitors.” See infra § IV.I.
16
category based on their three-run stack-test data;
(2) determines the HAP emissions level of the “best
controlled similar source” to establish standards for new
sources, 42 U.S.C. § 7412(d)(3), and determines the average
HAP emissions levels of the best performing 12 per cent of
sources to establish standards for existing sources, id.§ 7412(d)(3)(A); and then (3) applies the UPL methodology
to provide the cushion necessary to account for the expected
peaks and valleys in HAP emissions not reflected in the three-
run stack-test “snapshots.” See Page Mem. 4, 6.
b. The “Pollutant-By-Pollutant” Approach
In identifying the best performing sources in a given
category, often the EPA could not identify a single source that
controlled all HAPs better than all other sources. Instead, the
EPA found that one source effectively controlled emissions
from one HAP but was nonetheless one of the worst-
performing sources at controlling emissions from a different
HAP. For this reason, the EPA adopted a “pollutant-by-
pollutant” approach in setting MACT floors for major boiler
subcategories. See 2011 Major Boilers Rule, 76 Fed. Reg. at
15,622-23. That is, instead of identifying the one source that,
on balance, best controlled all HAPs in the aggregate, the
EPA used one source to set the MACT floor for, e.g., PM, and
used a different source to set the MACT floor for, e.g., HCl.
For at least two subcategories of major boilers—new heavy
oil-fired units and existing stoker coal-fired units—the EPA’s
pollutant-by-pollutant approach resulted in MACT floors that
no source had achieved in toto.
c. Startups, Shutdowns, and Malfunctions
The EPA found it difficult to account for HAP emissions
when sources start up, shut down, and malfunction. All three
17
occurrences alter HAP emissions and, historically, the EPA
exempted sources from normal numeric MACT-standard
compliance when these events occurred. See, e.g., Standards
of Performance for New Stationary Sources, 42 Fed. Reg.
57,125 (Nov. 1, 1977). Nevertheless, concluding that the Act
“require[s] that there must be continuous section [7412]-
compliant standards” and observing that the exemption meant
that “no section [7412] standard governs these events,” in
2008 we vacated the exemption for startups, shutdowns, and
malfunctions when the issue arose in a case challenging a
different rule. Sierra Club v. EPA (Sierra Club III), 551 F.3d
1019, 1027-28 (D.C. Cir. 2008) (emphasis added).
In response to the Sierra Club III vacatur, the EPA
established a work-practice standard in lieu of a numeric
MACT standard during startup and shutdown periods (but not
during malfunctions) when it promulgated the Major Boilers
Rule. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613.9
It did so after determining that the “physical limitations and
the short duration of startup and shutdown periods” made it
technologically infeasible to conduct the requisite testing for
numeric emissions limits. Id. A work-practice standard
sufficed, in the EPA’s view, because “[p]eriods of startup,
normal operations, and shutdown are all predictable and
routine aspects of a source’s operations.” Id.
9
Specifically, the startup and shutdown work-practice
standard requires a source to follow “the manufacturer’s
recommended procedures for minimizing periods of startup and
shutdown.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613. “If
manufacturer’s recommended procedures are not available,” the
Major Boilers Rule provided that “sources must follow
recommended procedures for a unit of similar design for which
manufacturer’s recommended procedures are available.” Id. at
15,642.
18
But because a malfunction is “sudden, infrequent, and not
reasonably preventable,” id. (quoting 40 C.F.R. § 63.2), the
EPA declined to treat a malfunction as a “distinct operating
mode,” id. As a result, the EPA did not account for
malfunctions when it set the MACT floors and it required
sources to comply with all MACT floors even during periods
of malfunction. Id. At the same time and recognizing that
even the best equipment can fail and that such failure can
spike emissions, the EPA added to the Major Boilers Rule “an
affirmative defense to civil penalties for exceedances of
numerical emission limits that are caused by malfunctions.”
Id. In reviewing a challenge to a different EPA rule, however,
we vacated a materially identical affirmative-defense
provision and held that the EPA has no power under the CAA
to create a defense to civil liability. See Natural Res. Def.
Council v. EPA (NRDC III), 749 F.3d 1055, 1062-64 (D.C.
Cir. 2014). Here, the EPA defends its decision not to address
malfunctions by asserting that it will use its enforcement
discretion regarding malfunctions on a case-by-case basis.
d. The One-Time Energy Assessment
The EPA also promulgated a “beyond-the-floor”
requirement for all facilities with existing major boilers. See
2011 Major Boilers Rule, 76 Fed. Reg. at 15,613.
Specifically, the Major Boilers Rule mandates a “a one-time
energy assessment . . . on the affected boilers and facility to
identify any cost-effective energy conservation measures,”
id., which assessment includes, inter alia, a review of fuel
usage, energy management practices, and conservation
measures, see 2013 Major Boilers Rule, 78 Fed. Reg. at
7,198-99. In some respects, the energy assessment is limited:
it (1) need occur only one time, see 40 C.F.R pt. 63, subpt.
DDDDD tbl.3; (2) is “based on energy use by discrete
19
segments of a facility and not by a total aggregation of all
individual energy using elements of a facility,” 2013 Major
Boilers Rule, 78 Fed. Reg. at 7,146; and (3) does not require
an owner or operator to implement any of the energy-saving
findings the assessment makes. In one respect, however, it is
expansive—it requires owners and operators to assess not
only the boilers themselves but also other components
“located on the site of the affected boiler that use energy
provided by the boiler,” including “compressed air systems”
as well as “facility heating, ventilation, and air conditioning
systems.” 40 C.F.R. § 63.11237.
e. The Health-Based Emissions Limits for HCl
Although the EPA set numeric MACT standards to
control HCl emissions, see 2013 Major Boilers Rule, 78 Fed.
Reg. at 7,193-98 tbls.1 & 2, in an earlier iteration of the Major
Boilers Rule, the EPA did not set MACT standards for HCl.
See National Emission Standards for Hazardous Air
Pollutants for Industrial, Commercial, and Institutional
Boilers and Process Heaters (2004 Boilers Rule), 69 Fed. Reg.
55,218, 55,227 (Sept. 13, 2004). Instead, the Agency opted
for a less stringent health-based emissions limit under section
7412(d)(4). See id. The EPA changed course after
concluding that HCl emissions posed health concerns the
Agency had not previously considered—in particular, the
EPA feared the “potential cumulative public health and
environmental effects” of HCl emissions, 2011 Major Boilers
Rule, 76 Fed. Reg. at 15,643-44 (emphasis added)—and after
recognizing that it did not have the requisite data to weigh
adequately the newly identified health risks.
20
2. The Area Boilers Rule
In the Area Boilers Rule, the EPA set emissions limits for
the same three boiler categories it controlled in the Major
Boilers Rule, see supra § I.B.1: industrial, commercial, and
institutional boilers. See 2013 Area Boilers Rule, 78 Fed. Reg.
at 7,488. It further split the categories into seven
subcategories, see id., and set emissions limits for three of
them, see id. at 7,517-18 tbls.1 & 2.10 These include:
(1) coal-fired boilers (i.e., “any boiler that burns any solid
fossil fuel and no more than 15 percent biomass,” 40 C.F.R.
§ 63.11237); (2) oil-fired boilers (i.e., “any boiler that burns
any liquid fuel and is not in either the biomass or coal
subcategories,” id.); and (3) biomass-fired boilers (i.e., “any
boiler that burns any” “biomass-based solid fuel that is not a
solid waste” and “is not in the coal subcategory,” id.). See
2013 Area Boilers Rule, 78 Fed. Reg. at 7,517-18 tbls.1 & 2.
For these subcategories, the EPA set emissions limits for
three HAPs: Hg, PM, and CO, with PM functioning as a
surrogate for non-Hg urban metals and CO functioning as a
surrogate for polycyclic organic matter (POM). See 2011
Area Boilers Rule, 76 Fed. Reg. at 15,586. Because Hg and
POM are both listed in section 7412(c)(6), the EPA had to set
MACT standards for Hg and for CO (as surrogate for POM)
for any area source category that, in the EPA’s view, required
10
As noted above, see supra § I.A.1.a, the EPA has some
discretion in promulgating emissions limits for area HAP sources.
Exercising its discretion, the EPA had previously determined that
natural gas-fired area boilers did not emit HAPs at a level
necessitating regulation. See National Emission Standards for
Hazardous Air Pollutants for Area Sources: Industrial, Commercial,
and Institutional Boilers (2010 Proposed Area Boilers Rule), 75
Fed. Reg. 31,896, 31,900 (June 4, 2010).
21
MACT control to assure a 90 per cent reduction in the
aggregate emissions of these two HAPs. See 42 U.S.C.
§ 7412(c)(6). The Agency complied, setting numeric MACT
standards for Hg and CO emissions from large coal-fired
boilers and a MACT work-practice standard (specifically, a
tune-up requirement) for emissions from small coal-fired
boilers. See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488,
7,517-18.11 It did not, however, set MACT standards for Hg
and POM emissions from biomass or oil-fired boilers, finding
it unnecessary to assure a 90 per cent reduction in aggregate
emissions of those two HAPs. See 2011 Area Boilers Rule,
76 Fed. Reg. at 15,566.
Thus, with the exception of Hg and CO emissions from
coal-fired boilers, the EPA had discretion to promulgate
GACT standards for all other HAPs in all other source
subcategories. See 42 U.S.C. § 7412(d)(5). Exercising this
discretion resulted in the following standards:
11
As used in the Area Boilers Rule, the difference between
“large” and “small” units depends on the heat-input capacity of the
unit. See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488. It is not
the same as the difference between “major” and “area” sources,
which is based on the volume of HAPs a source emits. See 42
U.S.C. § 7412(a).
22
Boiler Hg Hg
Size Age
Subcategory Limit Type
New Numeric MACT
Large
Existing Numeric MACT
New Tune- MACT
Coal
Up
Small
Existing Tune- MACT
Up
New --- ---
Large
Existing --- ---
Biomass
New --- ---
Small
Existing --- ---
New --- ---
Large
Existing --- ---
Oil
New --- ---
Small
Existing --- ---
Boiler CO CO
Size Age
Subcategory Limit Type
New Numeric MACT
Large
Existing Numeric MACT
Coal
New Tune-up MACT
Small
Existing Tune-up MACT
New Tune-up GACT
Large
Existing Tune-up GACT
Biomass
New Tune-up GACT
Small
Existing Tune-up GACT
New Tune-up GACT
Large
Existing Tune-up GACT
Oil
New Tune-up GACT
Small
Existing Tune-up GACT
23
Boiler PM PM
Size Age
Subcategory Limit Type
New Numeric GACT
Large
Existing Tune-up GACT
Coal
New Tune-up GACT
Small
Existing Tune-up GACT
New Numeric GACT
Large
Existing Tune-up GACT
Biomass
New Tune-up GACT
Small
Existing Tune-up GACT
New Numeric GACT
Large
Existing Tune-up GACT
Oil
New Tune-up GACT
Small
Existing Tune-up GACT
2013 Area Boilers Rule, 78 Fed. Reg. at 7,488-89, 7,517-19.
The Area Boilers Rule shares many of the same features
as the Major Boilers Rule; for example, the Area Boilers Rule
treats startups, shutdowns, and malfunctions in the same
fashion as the Major Boilers Rule, see supra § I.B.1.c—i.e.,
the Area Boilers Rule creates work-practice (or management-
practice) standards for startup and shutdown periods but does
not account for malfunctions at all, save for the Agency’s
commitment to consider malfunctions on a case-by-case basis.
See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,496; 2011 Area
Boilers Rule, 76 Fed. Reg. at 15,560-61. Additionally, the
Area Boilers Rule imposes the same one-time energy-
assessment requirement for existing large area boilers that the
Major Boilers Rule imposes for existing major boilers. See
supra § I.B.1.d; see also 2013 Area Boilers Rule, 78 Fed.
Reg. at 7,500; 2011 Area Boilers Rule, 76 Fed. Reg. at
24
15,560, 15,567-68. There are, however, two unique features
of the Area Boilers Rule that warrant brief discussion.
a. Exclusion of “Temporary Boilers”
After the EPA promulgated the 2011 Area Boilers Rule
but before it promulgated the 2013 version, it proposed an
amendment to 40 C.F.R. § 63.11195 that added temporary
boilers to the list of those boilers not regulated by
section 7412. See National Emission Standards for
Hazardous Air Pollutants for Area Sources: Industrial,
Commercial, and Institutional Boilers (2011 Proposed Area
Boilers Rule on Reconsideration), 76 Fed. Reg. 80,532,
80,535 (Dec. 23, 2011). The EPA created the exclusion
because, in its view, temporary boilers are “insignificant
sources[] and were not included in the EPA’s analysis of the
source category.” Id. The Agency eventually defined
“temporary boiler” as “any gaseous or liquid fuel boiler that is
designed to, and is capable of, being carried or moved from
one location to another by means of, for example, wheels,
skids, carrying handles, dollies, trailers, or platforms.” See
2013 Area Boilers Rule, 78 Fed. Reg. at 7,491 (quoting 40
C.F.R. § 63.11237).
b. Title V Permit Exemption for Synthetic Area Sources
As noted, see supra § I.A.3, Title V of the CAA imposes
a permit requirement on all owners and operators of major
and area HAP sources. See 42 U.S.C. § 7661a. The EPA,
however, can exempt an area source subcategory if it finds
“that compliance with such requirements is impracticable,
infeasible, or unnecessarily burdensome on such categories.”
Id. § 7661a(a). When it proposed the Area Boilers Rule in
2010, the EPA considered exempting some area sources
because, in its view, the existing restrictions on those sources
25
made Title V duplicative. See 2010 Proposed Area Boilers
Rule, 75 Fed. Reg. at 31,910-13. At the time, the EPA
announced that it did not intend to exempt “synthetic” area
sources (i.e., area sources that, but for existing air-pollution
controls, would be considered major sources). Id. at 31,913.
In so doing, the EPA reasoned that synthetic area sources:
(1) more closely resemble major sources than area sources,
(2) are often located in populous areas, and (3) have high
HAP emissions potential when uncontrolled. Id.
But in the 2011 Area Boilers Rule, the EPA changed
course and exempted synthetic area sources from the Title V
permitting requirement. See 76 Fed. Reg. at 15,578. It
reasoned that the “observations and data . . . relied upon in
other rulemakings for distinguishing between sources that
became synthetic area sources due to controls and other
synthetic and natural area sources did not necessarily apply to
this source category.” Id. In its view, it no longer had
“sufficient information” to distinguish synthetic area sources
from the others it exempted and, accordingly, “the rationale
for exempting most area sources subject to this rule . . . is also
now relevant for” synthetic area sources. Id.; see also 2013
Area Boilers Rule, 78 Fed. Reg. at 7,497.
3. The CISWI Rule
In the CISWI Rule, the EPA created four CISWI
subcategories: (1) incinerators (i.e., “units designed to burn
[solid] waste materials for the purpose of disposal”);
(2) small, remote incinerators (“SRIs”) (i.e., units that burn
small waste batches); (3) energy recovery units (“ERUs”)
(i.e., units that would be classified as boilers but for the fact
they combust solid waste); and (4) waste-burning kilns (i.e.,
units that would be classified as cement kilns if they did not
26
burn solid waste). 2013 CISWI Rule, 78 Fed. Reg. at 9,118.
Initially, the EPA proposed a fifth subcategory—burn-off
ovens—but eliminated burn-off ovens after comments
revealed that it had greatly underestimated the number of
units in that subcategory (36 versus 15,000) and that it lacked
the requisite data to set limits for the units. See 2011 CISWI
Rule, 76 Fed. Reg. at 15,734. Of the four CISWI
subcategories, the EPA further divided the ERU subcategory
(for CO emissions only) into coal-fired, biomass-fired and
oil/gas-fired ERUs and it further divided the waste-burning
kiln subcategory (again, for CO emissions only) into long and
preheater/precalcinator kilns. See 2013 CISWI Rule, 78 Fed.
Reg. at 9,118 tbl.2.
The EPA then set numeric MACT limits for the section
7429(a)(4) pollutants.12 See 2011 CISWI Rule, 76 Fed. Reg.
at 15,709-10 tbl.1. Unlike the Major Boilers Rule and the
Area Boilers Rule, the CISWI Rule contains no beyond-the-
floor MACT standards. The EPA also declined to promulgate
work-practice standards, concluding that it had no authority to
do so because section 7429 includes no work-practice
standard provision similar to that in section 7412. See id. at
15,721.
The CISWI Rule shares several features with the Major
Boilers Rule. In the CISWI Rule, for instance, the EPA also
used the UPL, see id. at 15,722-27, as well as the pollutant-
by-pollutant approach, see id. at 15,719-21, in setting MACT
floors. Based in part on the differences between section 7412
12
These pollutants are (1) PM, (2) sulfur dioxide (SO2),
(3) HCl, (4) nitrogen oxide (NOx), (5) CO, (6) lead (Pb),
(7) cadmium (Cd), (8) Hg, (9) dioxins and dibenzofurans, and (10)
opacity (where appropriate). 42 U.S.C. § 7429(a)(4).
27
and section 7429, the CISWI rule has four unique
characteristics we briefly describe.
a. Startups, Shutdowns, and Malfunctions
As discussed, see supra § II.B.1.c, the EPA imposed a
work-practice standard for major and area source boilers
during periods of startup and shutdown but declined to make
any regulatory modification for malfunctions. See 2011
Major Boilers Rule, 76 Fed. Reg. at 15,613; 2011 Area
Boilers Rule, 76 Fed. Reg. at 15,560-61. The CISWI Rule,
however, makes no modification for any of these periods,
mandating instead that the numeric MACT standards “apply
at all times,” even when CISWI units are starting up or
shutting down. 2011 CISWI Rule, 76 Fed. Reg. at 15,711,
15,737-38. The Agency concluded that it had no legal
authority under section 7429 to impose anything but a
numeric MACT standard on CISWI units. See id. at 15,709
tbl.1; see also id. at 15,737-38.
b. The Record-Keeping Requirement
Whether the EPA considers a combustion unit to be a
boiler (and thus subject to section 7412) or a CISWI (and thus
subject to section 7429) turns entirely on whether the unit
combusts “solid waste.” See id. at 15,709. The term “solid
waste” is defined in RCRA, 42 U.S.C. §§ 6901 et seq., and
clarified by EPA regulation, see Identification of Non-
Hazardous Secondary Materials that Are Solid Waste (NHSM
Rule), 76 Fed. Reg. 15,456, 15,457 (Mar. 21, 2011). See also
2011 CISWI Rule, 76 Fed. Reg. at 15,709. If the unit
combusts solid waste, it is a CISWI. Id.
The source owner or operator initially decides whether
the material its combustion unit burns meets the definition of
28
solid waste. See id. at 15,740. For this reason, the CISWI
rule requires that the owner or operator of a combustion unit
that burns materials “not clearly listed as traditional fuels”
keep records explaining how the materials meet the regulatory
definition of “non-solid waste.” Id.; see also 40 C.F.R.
§ 60.2175(v). Failure to do so means, for the purposes of the
EPA, that “the operating unit is a CISWI unit.” 40 C.F.R.
§ 60.2265; see also 2013 CISWI Rule, 78 Fed. Reg. at 9,188.
c. Emissions Averaging
During the notice-and-comment period, certain industry
entities urged the EPA to allow a facility containing more
than one CISWI unit to demonstrate compliance with the
CISWI MACT standards by averaging the HAP emissions of
all units in the facility. See Commercial and Industrial Solid
Waste Incineration Units: Reconsideration and Proposed
Amendments; Non-Hazardous Secondary Materials that Are
Solid Waste (2011 Proposed CISWI Rule on
Reconsideration), 76 Fed. Reg. 80,452, 80,463 (Dec. 23,
2011). Although it allowed facility-wide averaging in the
Major Boilers Rule, the Agency declined to allow it for
facilities with CISWI units. See id. The EPA explained, first,
that “[t]he applicability of CISWI is such that each unit is an
affected facility.” Id. In response to further comments, the
EPA subsequently explained that it did “not believe [it had]
the legal authority to allow emissions averaging in CISWI or
under section [7429] generally because each individual unit is
an affected facility.” Summary of Public Comments and
Responses for Commercial and Industrial Solid Waste
Incineration Units (CISWI Rule—Responses to Comments),
EPA-HQ-OAR-2003-0119-2638-A2 (Dec. 2012), at 195.
29
d. Treatment of Units that Begin Combusting Solid Waste
Finally, in the preamble to the 2011 CISWI Rule, the
EPA stated broadly that “[u]nits that begin combusting solid
waste are considered existing sources under CISWI.” 76 Fed.
Reg. at 15,714 (emphasis added). This categorical
pronouncement drew objections from commentators who
insisted that, if such units experienced an increase in HAP
emissions, the units would meet the statutory definition of
“modified solid waste incineration unit[s],” see 42 U.S.C.
§ 7429(g)(3), and would, accordingly, be subject to the
MACT standards for new units, see id. § 7429(g)(2). In the
subsequent proposed CISWI Rule, the EPA clarified that
“[a]n existing source will not be considered a new source
solely due to a combustion material switch. Assuming new
source applicability is not triggered, existing sources that
change fuels or materials are considered existing
sources . . . .” 2011 Proposed CISWI Rule on
Reconsideration, 76 Fed. Reg. at 80,459.
II. STANDARD OF REVIEW
For each issue, the Petitioners argue that the EPA either
misinterpreted the CAA, acted arbitrarily and capriciously, or
both. We review the EPA’s construction of the statute under
the two-part framework established in Chevron, U.S.A., Inc. v.
Nat. Res. Def. Council, Inc., 467 U.S. 837 (1984). At
Chevron step 1, we ask whether the Congress “has directly
spoken to the precise question at issue”; if it has, we “must
give effect to [its] unambiguously expressed intent.” Id. at
842-43. In so doing, we examine the CAA’s text, structure,
purpose, and legislative history to determine if the Congress
has expressed its intent unambiguously. See Bell Atl. Tel. Co.
v. FCC, 131 F.3d 1044, 1047 (D.C. Cir. 1997). If the statute
30
is “silent or ambiguous with respect to the specific issue,” we
proceed to Chevron step 2 and defer to the EPA’s
interpretation so long as it is “based on a permissible
construction of the statute.” Chevron, 467 U.S. at 842-43.
The CAA authorizes the Court to “reverse any [EPA]
action found to be . . . arbitrary, capricious, an abuse of
discretion, or otherwise not in accordance with law.” 42
U.S.C. § 7607(d)(9)(A). Our review is “narrow” and we will
“not . . . substitute [our] judgment for that of the
agency.” Motor Veh. Mfrs. Ass’n v. State Farm Mut. Auto.
Ins. Co. (State Farm), 463 U.S. 29, 43 (1983). We “must
uphold an agency’s action where [the agency] ‘has considered
the relevant factors and articulated a rational connection
between the facts found and the choice made,’ and has not
‘relied on [improper] factors.’” Nat’l Ass’n of Clean Air
Agencies v. EPA (NACAA), 489 F.3d 1221, 1228 (D.C. Cir.
2007) (citations omitted) (quoting Allied Local & Reg’l Mfrs.
Caucus v. EPA, 215 F.3d 61, 68 (D.C. Cir. 2000), and State
Farm, 463 U.S. at 43). A rule is arbitrary and capricious if
the agency: (1) “has relied on factors which Congress has not
intended it to consider,” (2) “entirely failed to consider an
important aspect of the problem,” (3) “offered an explanation
for its decision that runs counter to the evidence before the
agency,” or (4) “is so implausible that it could not be ascribed
to a difference in view or the product of agency expertise.”
State Farm, 463 U.S. at 43.
We review the EPA’s factual determinations for
substantial evidence. 5 U.S.C. § 706(2)(E). We also “owe[]
particular deference to EPA when its rulemakings rest upon
matters of scientific and statistical judgment within [its]
sphere of special competence and statutory jurisdiction.” Am.
Coke & Coal Chems. Inst. v. EPA, 452 F.3d 930, 941 (D.C.
31
Cir. 2006). But “[w]e are hesitant to rubber-stamp EPA’s
invocation of statistics without some explanation of the
underlying principles or reasons why its formulas would
produce an accurate result.” NACWA, 734 F.3d at 1145.
III. INDUSTRY PETITIONERS’ CHALLENGES
A. STARTUPS, SHUTDOWNS, AND MALFUNCTIONS
Industry Petitioners raise two sets of challenges to
startup, shutdown, and malfunction periods: (1) a challenge to
the EPA’s failure to take malfunctions into account in the
Major Boilers and Area Boilers Rules and (2) a challenge to
EPA’s failure to take into account periods of startup,
shutdown, and malfunction in the CISWI Rule. For the
reasons that follow, we reject all of the Industry Petitioners’
claims related to startups, shutdowns, and malfunctions.
1. Periods of Malfunction in the Major Boilers and Area
Boilers Rules
First, Industry Petitioners challenge the Major Boilers
and Area Boilers Rules’ failure to take malfunctions into
account in setting MACT floors. See 2011 Major Boilers
Rule, 76 Fed. Reg. at 15,613; 2011 Area Boilers Rule, 76 Fed.
Reg. at 15,560-61. The EPA defends its refusal to account for
malfunctions on the basis of (1) the impracticability of
accounting for events that are necessarily unpredictable, and
(2) the EPA’s assertion that it will use its prosecutorial
discretion to determine on a case-by-case basis whether an
exceedance of emission standards is attributable to an
excusable malfunction or whether applicable regulatory
penalties should be imposed instead. See No. 11-1108 EPA
Br. 38; No. 11-1141 EPA Br. 29.
32
Both sides agree that malfunctions are inevitable in the
operation of area and major boilers. According to the EPA,
“even equipment that is properly designed and maintained can
sometimes fail and . . . such failure can sometimes cause an
exceedance of the relevant emission standard.” 2011 Major
Boilers Rule, 76 Fed. Reg. at 15,613; 2011 Area Boilers Rule,
76 Fed. Reg. at 15,561. Thus, the EPA defined a malfunction
as a “sudden, infrequent, and not reasonably preventable
failure of air pollution control and monitoring equipment,
process equipment or a process to operate in a normal or usual
manner.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613
(citing 40 C.F.R. § 63.2); 2011 Area Boilers Rule, 76 Fed.
Reg. at 15,560 (same). In attempting to write rules to account
for emissions, however, the EPA faced an intractable
problem: how to account for a malfunction which is, by
definition, unpredictable in terms of timing, duration,
magnitude, and effect. While the existence of malfunctions is
entirely predictable, the nature of those malfunctions is not,
and it is the malfunction’s nature that affects emissions and
thus is relevant to the application of emission limits.
At first glance, the EPA’s chosen approach to
malfunctions may seem counterintuitive, as the Agency
appears to have several reasonable alternatives: it could
exempt periods of malfunction entirely from the application
of the emission standards; or it could apply the standards to
malfunctions while giving boiler owners the opportunity to
defend against a penalty by demonstrating they were not at
fault for the malfunction. But the EPA has previously been
stymied in its attempts to implement either of these solutions,
as this court has concluded neither approach is consistent with
the Agency’s enabling statutes. For instance, in Sierra Club
III, the EPA attempted to exempt major sources from
complying with emission standards during start up, shut
33
down, and malfunction. See 551 F.3d at 1027-28. This court
rejected that approach because the Congress “required that
there must be continuous section 112-compliant standards”
and so the EPA lacked discretion to exempt certain periods
from compliance, regardless of their unpredictability. Id. at
1027. In NRDC III, this court considered a challenge to the
affirmative defense provision the EPA adopted for persons
defending against civil suits under 42 U.S.C. § 7604(a), which
allows “any person” to “commence a civil action on his own
behalf” against any entity alleged to be in violation of an
emission standard or limitation. The affirmative defense
provision was meant to shield alleged violators from liability
for certain emissions violations caused by “unavoidable”
malfunctions; under the provision, therefore, “the district
court [could] assess penalties only if violators fail[ed] to meet
[their] burden of proving all of the requirements in the
affirmative defense.” NRDC III, 749 F.3d at 1062 (internal
quotation omitted). The court rejected this provision as an
impermissible intrusion on the judiciary’s role. See id. at
1063 (“[U]nder this statute, deciding whether penalties are
‘appropriate’ in a given private civil suit is a job for the
courts, not for EPA.”).
Faced with an obvious dilemma, the EPA arrived at the
approach it defends today. Malfunctions receive no special
treatment and the EPA instead exercises “its enforcement
discretion to address exceedances of emission limits that may
be caused by such uncertain, unpredictable events, on a case-
by-case basis.” No. 11-1108 EPA Br. 38; see also No. 11-
1141 EPA Br. 29. The EPA’s current treatment of
malfunctions thus differs from its invalid affirmative defense
provision because the Agency is exercising its own regulatory
enforcement power on an ad hoc basis outside the context of
citizen suits. When an exceedance occurs during a
34
malfunction, the EPA determines what enforcement action—
if any—it should take by considering “the good faith efforts
of the source to minimize emissions during malfunction
periods, including preventative and corrective actions, as well
as root cause analyses to ascertain and rectify excess
emissions.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613;
see also 2011 Area Boilers Rule, 76 Fed. Reg. at 15,561
(same). The EPA also considers whether the exceedance was
in fact “not reasonably preventable” or whether it was
“caused in part by poor maintenance or careless operation.”
2011 Major Boilers Rule, 76 Fed. Reg. at 15,613 (citing 40
C.F.R. § 63.2); see also 2011 Area Boilers Rule, 76 Fed. Reg.
at 15,561 (same).
For our purposes, we need not (indeed, must not)
evaluate the policy implications of the EPA’s regulatory
choice because our review is confined to determining whether
the EPA’s regulation reflects a permissible reading of the
applicable statute under Chevron. Here, we conclude that it
does. The relevant statute requires only that the EPA set
“achievable” standards, 42 U.S.C. § 7412(d)(2), and it defines
achievability to be no less “than the emission control that is
achieved in practice by the best controlled similar source,” 42
U.S.C. § 7412(d)(3). The “best controlled similar source,”
however, is unlikely to be a malfunctioning source, and the
EPA is bound to enact a standard in keeping with emission
limits achieved by that “best controlled similar source.” If
anything, then, the statutory language on its face prevents the
EPA from taking into account the effect of potential
malfunctions when setting MACT emission standards. At the
very least, the language permits the EPA to ignore
malfunctions in its standard-setting and account for them
instead through its regulatory discretion. Our Sierra Club III
decision confirms this. See 551 F.3d at 1027-28. Because the
35
EPA had no option to exclude these unpredictable periods, its
approach is reasonable. We therefore reject Industry
Petitioners’ argument that the EPA either misinterpreted the
CAA or acted arbitrarily and capriciously in failing to account
for malfunctions when setting MACT floors in the Major and
Area Boilers Rules.
Nor do we agree with the Industry Petitioners’ secondary
argument that the EPA acted arbitrarily and capriciously by
failing to set a work-practice or a GACT management-
practice standard for malfunction periods. First, the statute
makes clear that these kinds of standards are to be set at the
discretion of the EPA, so it would be difficult to interpret the
statute consistently with its text while holding that the text’s
permissive language in fact sets out a requirement that the
Agency set work-practice or GACT management-practice
standards. As to work-practice standards, “[t]he
Administrator may, in lieu [of a numeric standard],
promulgate a design, equipment, work practice, or operational
standard, or combination thereof,” and any such standard set
must “in the Administrator’s judgment [be] consistent with
the provisions of subsection (d).” 42 U.S.C. § 7412(h)(1). As
to GACT management-practice standards, “the Administrator
may . . . elect to promulgate” such standards with respect to
certain “categories and subcategories of area sources.” Id.§ 7412(d)(5). It should go without saying that “may means
may.” McCreary v. Offner, 172 F.3d 76, 83 (D.C. Cir. 1999)
(internal quotations omitted).
Second, the Petitioners have not demonstrated and the
EPA does not concede that setting work-practice or GACT
management-practice standards would even be feasible for
periods of malfunction. As for work-practice standards, the
EPA would have to conceive of a standard that could apply
36
equally to the wide range of possible boiler malfunctions,
ranging from an explosion to minor mechanical defects. Any
possible standard is likely to be hopelessly generic to govern
such a wide array of circumstances. Similar problems exist
for setting GACT management practices. These management
practices would also need to apply to the wide range of
possible malfunctions, and the EPA would need to determine
that the standard would “reduce emissions of hazardous air
pollutants,” an evidence-based standard that is difficult
(perhaps impossible) to apply to the unpredictable
circumstances of malfunctions. 42 U.S.C.
§ 7412(d)(5). Thus, we reject the Industry Petitioners’
argument that the EPA was required to set a work-practice or
GACT management-practice standard for malfunction
periods.
In doing so, we are mindful that the EPA is not the only
entity able to bring enforcement actions under the CAA, but
that private citizens are also empowered to enforce emission
standards by filing suit in district court. 42 U.S.C.
§ 7604(a). Assurances that the EPA will use its prosecutorial
discretion to account for malfunctions would mean little if
private citizens could seek strict enforcement of those same
standards. But as we stated in NRDC III, “the Judiciary, not
any executive agency, determines ‘the scope’—including the
available remedies—‘of judicial power vested by’ statutes
establishing private rights of action.” 749 F.3d at 1063
(quoting City of Arlington v. FCC, 133 S. Ct. 1863, 1871
(2013)). Accordingly, in citizen suits under the CAA, “the
courts determine, on a case-by-case basis, whether civil
penalties are ‘appropriate.’” Id. Boiler operators can argue
that penalties should not be assessed because of an
unavoidable malfunction, and they can support that argument
with other relevant facts, “such as the defendant’s ‘full
37
compliance history and good faith efforts to comply.’” Id.(quoting 42 U.S.C. § 7413(e)(1)). The EPA can also provide
supporting argumentation as intervenor or amicus. Id. Courts
should not hesitate to exercise their judicial authority to craft
appropriate civil remedies in the case of emissions
exceedances caused by unavoidable malfunctions.
2. Periods of Startup, Shutdown, and Malfunction in the
CISWI Rule
In the CISWI Rule, the EPA made no modification for
periods of startup, shutdown, or malfunction. The Industry
Petitioners argue that failing to account for these periods
violated the EPA’s statutory instruction to set “achievable”
standards. Additionally, the Industry Petitioners claim it was
arbitrary and capricious for the EPA to set work-practice
standards for startup and shutdown periods under the Major
Boilers Rule but not under the CISWI Rule. Both arguments
are without merit.
First, the EPA’s emission standards for small incinerators
do take into account periods of shutdown and startup. The
EPA based its standards for these machines on “short term
stack tests for pollutants,” in which incinerators are monitored
during the course of normal operation, which includes daily
startup and shutdown periods. See 2011 CISWI Rule, 76 Fed.
Reg. at 15,738. Thus, startup and shutdown times are already
incorporated into the standards the EPA set, and what is more,
nearly all pollutants are present in smaller numbers during
startup and shutdown anyway, when incinerators are burning
fuels alone rather than fuels and solid waste. See Standards of
Performance for New Stationary Sources and Emission
Guidelines for Existing Sources: Commercial and Industrial
Solid Waste Incineration Units (2010 Proposed CISWI Rule),
38
75 Fed. Reg. 31,938, 31,964 (June 4, 2010). Given this
reality, the CISWI Rule satisfies the statutory standard of
“achievability” and is not arbitrary and capricious.
Second, as to periods of malfunctions, the same analysis
applies to the CISWI Rule as applies to the Boilers
Rules. The EPA adopted a reasonable interpretation of the
CAA when it excluded periods of malfunction from its
calculations of achievability given that malfunction periods
are by their very nature unpredictable in terms of their effect
on emissions. The EPA’s decision to account for
malfunctions in its discretion is likewise a reasonable
interpretation of 42 U.S.C. § 7412(d)(2) and (3).
For these reasons, we reject the Industry Petitioners’
challenges to the EPA’s regulatory choices with regard to
periods of startup, shutdown, and malfunction.
B. THE POLLUTANT-BY-POLLUTANT APPROACH
The EPA must look to the performance of the best major
boilers and CISWI incinerators when setting MACT floors for
a pollutant. As described above, for new units, the EPA must
set floors at the level achieved by the best similar unit in each
subcategory. For existing units, the Agency must set floors at
the level achieved by the best 12 per cent of similar units in
each subcategory. 42 U.S.C. §§ 7412(d)(3)(A), 7429(a)(2).
As a result, the EPA had to identify the best performing units
in each subcategory when setting the MACT floors for the
Major Boilers and CISWI Rules. But the EPA often could not
identify a single unit or set of units that controlled all HAPs
better than the other units in the subcategory. Instead, the
EPA sometimes found that a unit might rank among the best
39
in its subcategory at controlling emissions of one HAP, but
among the worst at controlling emissions of a different HAP.
To address this problem, the EPA adopted a “pollutant-
by-pollutant” approach in setting the MACT floors: instead of
identifying the unit or units that best controlled all HAPs in
the aggregate, the EPA used one unit or set of units to set the
MACT floor for, e.g., PM, and used a different unit or set of
units to set the MACT floor for, e.g., HCl. See 2011 Major
Boilers Rule, 76 Fed. Reg. at 15,621-23; 2011 CISWI Rule,
76 Fed. Reg. at 15,720-21. For at least two subcategories of
major boilers—new heavy oil-fired units and existing stoker
coal-fired units—the EPA’s pollutant-by-pollutant approach
resulted in MACT floors that no unit in the subcategory had
achieved in toto. Similarly, for small, remote incinerators
(SRIs), the approach resulted in standards for existing units
that only two of the 28 SRI units had met in toto, and
standards for new units that no existing SRI had met in toto.
The Industry Petitioners challenge the EPA’s use of the
pollutant-by-pollutant approach. According to the Industry
Petitioners, the CAA’s plain language requires the Agency to
identify the best overall unit or set of units—not the best unit
or set of units for a particular pollutant—in each subcategory
when setting MACT floors. They further claim the EPA’s
pollutant-by-pollutant approach was unreasonable with regard
to SRIs because it resulted in a set of emission standards that
no single unit in the subcategory had achieved in practice.
We disagree, and conclude that the EPA’s pollutant-by-
pollutant approach is a reasonable interpretation and
application of the statute.
For the purposes of this challenge, the MACT floor
provisions for major boilers and CISWI units are identical.
40
Under both provisions, the EPA must set emission standards
for new units based on “the emissions control that is achieved
in practice by the best controlled similar unit, as determined
by the Administrator.” 42 U.S.C. § 7429(a)(2) (CISWI); see
also id. § 7412(d)(3) (major boilers). For existing units, the
MACT floor is based on “the average emissions limitation
achieved by the best performing 12 percent of units in the
category.” Id. § 7429(a)(2) (CISWI); see also id.
§ 7412(d)(3)(A) (major boilers).
The Industry Petitioners claim this language
unambiguously forecloses the EPA’s pollutant-by-pollutant
approach. For new units, they assert, the statute requires the
EPA to find the single unit that performs best overall and use
this unit—and only this unit—to set standards for all
regulated pollutants. For example, if Incinerator 3 were
deemed the best overall performer in a subcategory, then the
EPA would use Incinerator 3’s emissions levels to set
standards for PM, CO, and each of the other regulated
pollutants. This would be true even if Incinerator 1 in the
same subcategory had lower CO emissions and Incinerator 2
had lower PM emissions. The Industry Petitioners also make
this argument for existing sources. For these units, under
their interpretation, the mandate to identify the “best
performing 12 percent of units” required the EPA to use data
from the 12 per cent of sources with the lowest overall
emissions in the subcategory. In short, the Industry
Petitioners argue that the best “unit” referred to by the
provision cannot be a “hypothetical composite” of multiple
units that result in standards for new units that no actual unit
has met in practice with regard to every pollutant, or
standards for existing units that 12 per cent of actual units
have not met with regard to every pollutant.
41
The Industry Petitioners read too much into the statutory
language. It is true that the statute requires the EPA to base
MACT standards on what is “achieved” by the best “unit” or
“12 percent of units.” But, as the EPA argues, the statute says
nothing about how the Agency should determine which units
are the best. Cf. Sierra Club v. EPA, 167 F.3d 658, 661 (D.C.
Cir. 1999) (noting that section 7429(a) “on its own says
nothing about how the performance of the best units is to be
calculated”). Both the industry-favored method of choosing
the best overall unit and the EPA’s method of choosing the
best unit as to each particular pollutant facially comport with
the statute’s mandate to determine which units are best.
Because the statute is ambiguous as to how the EPA should
identify those units, we must defer to the Agency’s choice so
long as it is reasonable. See Sierra Club I, 353 F.3d at 990.
Here, the EPA’s choice is reasonable. The statute
provides that emission standards shall reflect “the maximum
degree of reduction in emissions of [regulated pollutants] that
the Administrator . . . determines is achievable for new or
existing units in each category.” 42 U.S.C. § 7429(a)(2); see
also id. § 7412(d)(2). It then provides that the “degree of
reduction in emissions that is deemed achievable for new
units in a category shall not be less stringent than the
emissions control that is achieved in practice by the best
controlled similar unit, as determined by the Administrator.”
Id. § 7429(a)(2); see also id. § 7412(d)(3). Reading these
provisions together, they support a pollutant-by-pollutant
approach. The “best controlled similar unit” language does
not exist in a vacuum; rather, it exists to measure the “degree
of reduction in emissions that is deemed achievable.” Id.
§ 7429(a)(2); see also id. § 7412(d)(3). That “reduction in
emissions” is the reduction in emissions of each pollutant
listed in sections 7429(a)(4) and 7412(b)(1). The EPA’s
42
approach to setting standards on a pollutant-by-pollutant basis
thus comfortably fits within this statutory scheme.
Moreover, the Industry Petitioners have not explained
how their preferred approach would better comport with the
statute. Were the EPA required to determine which units
perform best “overall,” we see at least two possibilities for
how it could do so: First, the EPA could calculate a unit’s
average emissions for each pollutant in consistent units of
measurement, add these emissions together, and then choose
the unit with the smallest overall sum in each subcategory.
But this approach could produce arbitrary results, because the
“best performing” overall unit might emit unusually low
quantities of some pollutants and unusually high quantities of
others. This would mean the emission standards for some
pollutants would be lenient while others would be stringent,
with no principled reason for the difference. Alternatively,
the Agency could identify which source is best overall based
on which emits the lowest level of the riskiest pollutants. But
this approach would require the Agency to rank pollutants’
relative risks without any congressional guidance on how to
do so. This approach would also contravene our previous
understanding of the congressional intent behind the MACT
floor provisions. As we have explained, the MACT floors
“are to be based not on an assessment of the risks posed by
[pollutants], but instead on the maximum achievable control
technology (MACT) for sources in each category.” Sierra
Club I, 353 F.3d at 980.
The Industry Petitioners nevertheless argue that the
CAA’s legislative history supports their preferred approach.
In particular, they point to the floor comments of Senator
Durenberger discussing the potential impact on MACT floors
of mutually incompatible control technologies. 136 Cong.
43
Rec. S17,238 (daily ed. Oct. 26, 1990) (statement of Senator
Durenberger). Mutually incompatible control technologies
cannot be used at the same time and therefore present
regulators with a dilemma. For example, say Technology 1
and Technology 2 cannot be used together. If Technology 1
is better at reducing PM than Technology 2, and Technology
2 is better at reducing CO than Technology 1, the EPA would
have to choose which of the two technologies to factor into
emission standards. In such situations, Senator Durenberger
anticipated that the “EPA should judge MACT to be the
technology which best benefits human health and the
environment on the whole.” Id. The Industry Petitioners
argue this statement demonstrates that Congress intended the
EPA to make an overall determination of which units are the
best performing “on the whole.”
Senator Durenberger’s statement does not support this
broad principle. The statement merely explains that, where
two technologies cannot be used together, the EPA should
base MACT standards on the technology it considers best
overall. Here, the Industry Petitioners do not identify any
relevant control technologies that are mutually incompatible.
Indeed, the EPA found in the CISWI Rule that “there is no
technical reason why [the] air pollution control systems
cannot be combined.” 2011 CISWI Rule, 76 Fed. Reg. at
15,721; see also 2011 Major Boilers Rule, 76 Fed. Reg. at
15,623 (“All available data for boilers and process heaters
indicate that there is no technical problem achieving the floor
levels contained in this final rule for each HAP
simultaneously, using the MACT floor technology.”). There
is thus no reason to believe that the EPA’s current MACT
floor standards cannot be achieved. Instead, the Industry
Petitioners merely insist that no units currently meet the
EPA’s new unit standards with regard to every regulated
44
pollutant in certain subcategories, and only a few sources
meet all of the standards for existing units in the same
subcategories. But, if the statute permits the EPA to
determine which units are best on a pollutant-by-pollutant
basis—and it does—then the EPA’s choice to adopt that
approach does not become unlawful merely because few or no
units have achieved those standards for all pollutants.
Finally, the Industry Petitioners argue that even if the
pollutant-by-pollutant approach is reasonable in some
circumstances, it is arbitrary and capricious as applied to
certain SRIs because it exacerbates certain problems posed by
the “batch” nature of SRIs. As explained at infra § III.E,
SRIs burn waste in small batches. According to the
Petitioners, this means that the SRIs that the EPA identified as
best performing were, in reality, burning cleaner waste at the
time emissions testing was done; they were not actually better
than other units at removing or destroying waste. The
pollutant-by-pollutant approach, the Industry Petitioners
argue, “simply captures the results from units that happened
to be burning wastes with low levels of that particular
pollutant during testing,” and this reality makes it harder for
SRI units to meet emission standards for all pollutants at the
same time. No. 11-1125 Indus. Pet’rs’ Reply Br. 8 (emphasis
omitted).
This argument fails because the Industry Petitioners have
not demonstrated that the Agency considered impermissible
factors, failed “to consider important aspect[s] of the
problem,” or offered an unreasonable explanation for its
decision when setting the MACT floors for SRIs. See State
Farm, 463 U.S. at 43. Rather, their argument is a back-door
attempt to challenge the Agency’s alleged failure to consider
waste inputs, which we reject below at infra § III.E.
45
Petitioners have also not shown that it is infeasible for the SRI
units to meet the MACT floor standards or that any individual
pollutant standard was not achieved in practice by an existing
SRI unit. They merely assert, without evidence, that no
existing unit burning high sulfur garbage can match the SO2
performance achieved by the unit the EPA used to set SO2
standards because that latter unit was burning low sulfur
waste at the time of the emissions testing. But MACT floors
are not unreasonable simply because they are difficult to
achieve in practice. As such, we find the EPA’s pollutant-by-
pollutant approach to be a reasonable interpretation and
application of the statute, and deny the Industry Petitioners’
challenge to the EPA’s use of this approach.
C. THE ENERGY-ASSESSMENT REQUIREMENT
The Major Boilers Rule and the Area Boilers Rule
generally require sources with existing boilers to perform a
one-time energy assessment. In the assessment, facilities
must “identify energy conservation measures”—such as
“process changes or other modifications to the facility”—
“that can be implemented to reduce the facility energy
demand,” thereby “reduc[ing] fuel use.” 2011 Area Boilers
Rule, 76 Fed. Reg. at 15,573; see also 2011 Major Boilers
Rule, 76 Fed. Reg. at 15,632. While facilities must conduct
the assessment, they need not implement its conclusions. See
2011 Area Boilers Rule, 76 Fed. Reg. at 15,573; 2011 Major
Boilers Rule, 76 Fed. Reg. at 15,632.
The logic behind the assessment is straightforward.
Boilers produce HAP emissions when fuel is combusted.
Less combustion means fewer emissions. The EPA primarily
justified the assessment as a beyond-the-floor MACT
requirement under section 7412(d)(2). See 2011 Area Boilers
46
Rule, 76 Fed. Reg. at 15,573; 2011 Major Boilers Rule, 76
Fed. Reg. at 15,632. With respect to certain biomass and oil-
fired boilers located at area sources, the assessment was
justified as a GACT management practice under
section 7412(d)(5). See 2011 Area Boilers Rule, 76 Fed. Reg.
at 15,567.
Industry Petitioners raise three principal challenges to the
energy-assessment requirement, none of which have purchase.
The first challenge claims that the energy assessment
regulates aspects of facilities that are off limits to the EPA—
namely, the energy needs supplied by regulated boilers.
Petitioners point to the language of the CAA, which requires
the EPA to “list . . . categories and subcategories of major
sources and area sources” of enumerated air pollutants. 42
U.S.C. § 7412(c)(1). “For the categories and subcategories
the Administrator lists, the Administrator” must set
“emissions standards under” section 7412(d). Id.§ 7412(c)(2). As relevant here, the EPA defined the source
categories to include “industrial boilers and commercial and
institutional boilers.” 2011 Area Boilers Rule, 76 Fed. Reg. at
15,557; 2011 Major Boilers Rule, 76 Fed. Reg. at 15,608. To
the extent the assessment concerns parts of the facility other
than the boiler itself, the Industry Petitioners claim it exceeds
the EPA’s authority.
The Industry Petitioners misapprehend both the scope of
the assessment and the CAA. The assessment requires
facilities to evaluate energy systems “located on the site of the
affected boiler,” including “[p]rocess heating[,] compressed
air systems[,] . . . facility heating, ventilation, and air
conditioning systems,” and “[o]ther systems that use steam,
hot water, process heat, or electricity, provided by the affected
boiler.” 40 C.F.R. § 63.11237; see id. § 63.7575. Based on
47
that evaluation, facilities must compile a “comprehensive
report detailing the ways to improve efficiency, the cost of
specific improvements, [anticipated] benefits, and the time
frame for recouping those investments.” 40 C.F.R. pt. 63,
subpt. JJJJJJ tbl.2; id. pt. 63, subpt. DDDDD tbl.3.
Contrary to the Industry Petitioners’ argument, the EPA
has not “regulate[d] virtually every piece of equipment at all
affected facilities.” No. 11-1141 Indus. Pet’rs’ Br. 19. Only
“energy use systems” that “us[e] energy clearly produced by
affected boilers” must be evaluated; facilities need not review
the “total aggregation of all individual energy using segments
of a facility.” 2013 Area Boilers Rule, 78 Fed. Reg. at 7,493
(emphasis added); see also 2013 Major Boilers Rule, 78 Fed.
Reg. at 7,188. The assessment focuses on “discrete segments
of a facility,” such as “production area[s] or building[s]”
associated with a particular boiler. 2013 Area Boilers Rule,
78 Fed. Reg. at 7,493; see 2013 Major Boilers Rule, 78 Fed.
Reg. at 7,188. Energy requirements satisfied by other
sources—not by a HAP-emitting boiler—fall outside of that
mandate. See 2011 Area Boilers Rule, 76 Fed. Reg. at 15,573
(limiting the assessment to “specific portions of the source
that directly affect emissions from the affected boiler”). And
regulated facilities are under no obligation to implement the
results they reach. In essence, rather than setting inflexible
and generally applicable beyond-the-floor numeric limits, the
EPA required facilities to take stock of the actual energy
demands placed on their boilers. By reducing energy
demands and associated fuel consumption, facilities could
reduce HAP emissions. That requirement is more measured
than the Industry Petitioners contend.
And that measured requirement falls within the EPA’s
statutory authority. The CAA authorizes the EPA to regulate
48
“major sources and area sources” of HAPs, and to subdivide
those sources into categories and subcategories. 42 U.S.C.
§ 7412(c)(1), (c)(2). To Industry Petitioners, the authority to
subdivide sources means the EPA may only regulate the
narrowest applicable categorization—in this instance,
commercial and industrial boilers. But the statute does not
require so rigid a reading. While the EPA is permitted to
subdivide sources, each subdivision remains a component of
either a major or area “source.” Dividing sources into
categories and subcategories does not make them any less of a
“source” subject to the EPA’s regulation.
For that reason, the EPA explained that the Rules reach,
respectively, “[a]ny area source facility using a boiler,” 2011
Area Boilers Rule, 76 Fed. Reg. at 15,555 (emphasis added),
and “major source facilities having affected boilers or process
heaters,” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613
(emphasis added). Likewise, the regulations implementing
the energy assessment requirement apply to those who “own
or operate an existing affected boiler,” not merely to the
boiler itself. 40 C.F.R. § 63.11214(c); see id. § 63.7485.
Going further, the relevant part of the CFR applies, by its own
terms, to the “owner or operator of any stationary source.” Id.
§ 63.1(b)(1).
The Congress’s definition of the terms major and area
source supports this reading. At bottom, both terms refer to a
“stationary source.” See 42 U.S.C. § 7412(a)(1), (a)(2).
Stationary source, in turn, means “any building, structure,
facility, or installation which emits or may emit any air
pollutant.” Id. § 7411(a)(3). Against that backdrop, the Rules
apply to any “building, structure, facility, or installation” that
contains a boiler emitting the specified HAPs. The EPA’s
49
regulatory authority reaches the relevant stationary source, of
which the boiler is part.
That the EPA may regulate stationary sources does not
mean it may regulate every nook and cranny of those sources.
The CAA directs its authority to the establishment of
emission standards; it does not provide some general power to
superintend the business processes of plants and
manufacturing facilities. In this case, however, we have no
occasion to parse the precise parameters of the EPA’s
authority to regulate aspects of area sources. It is enough to
conclude that the challenged energy assessment—which
applies only to systems that “us[e] energy clearly produced by
affected boilers”—falls within the EPA’s authority under the
CAA. 2013 Area Boilers Rule, 78 Fed. Reg. at 7,493; 2013
Major Boilers Rule, 78 Fed. Reg. at 7,188.
In the remaining two challenges, the Industry Petitioners
take issue with the EPA’s justification of the energy
assessment as a beyond-the-floor MACT standard and a
GACT management-practice standard. We reject both
challenges.
The assessment represents a valid beyond-the-floor
MACT standard.13 As discussed, after the Agency sets the
MACT floor, it must determine “whether stricter standards
are ‘achievable,’” Nat’l Lime Ass’n, 233 F.3d at 629 (quoting
42 U.S.C. § 7412(d)(2)), considering costs, “any non-air
13
In addition to challenging the assessment as a beyond-the-
floor measure, the Industry Petitioners claim the assessment
represents an invalid work-practice standard. But “[t]he energy
assessment is not . . . a work practice standard, and EPA makes no
claim that it is.” No. 11-1141 EPA Br. 47 n.9. Therefore, we
decline to address that contention.
50
quality health and environmental impacts and energy
requirements,” 42 US.C. § 7412(d)(2). These “measures,
processes, methods, systems or techniques includ[e], but [are]
not limited to, measures which—
(A) reduce the volume of, or eliminate emissions
of, such pollutants through process changes,
substitution of materials or other
modifications, . . .
(D) are design, equipment, work practice, or
operational standards . . . or
(E) are a combination of the above.
Id. The EPA primarily justified the energy assessment as a
beyond-the-floor measure designed to identify “process
changes or other modifications to the facility” that would
reduce fuel use and thereby reduce hazardous emissions.
2011 Area Boilers Rule, 76 Fed. Reg. at 15,573; 2011 Major
Boilers Rule, 76 Fed. Reg. at 15,632.
The Industry Petitioners argue that the EPA skipped a
step, imposing the energy assessment as a beyond-the-floor
measure without first setting a relevant MACT floor. That is
incorrect. The EPA first set a numeric MACT emissions limit
for the categories and subcategories of sources subject to the
energy assessment. See 40 C.F.R. pt. 63, subpt. JJJJJ tbl.1; id.
pt. 63, subpt. DDDDD tbl.2. The energy assessment
represents a step beyond that—a measure designed to
discover energy efficiencies that, once implemented, could
decrease emissions below the floor level.
Before setting a beyond-the-floor measure, the EPA must
consider whether it is “achievable” based on a number of
51
factors, among them cost, “non-air quality health and
environmental impacts and energy requirements.” 42 US.C.
§ 7412(d)(2). The EPA did so here. To begin, the EPA
adequately considered costs. In the EPA’s estimation, “[t]he
one-time cost of an energy assessment ranges from $2500 to
$55,000 depending on the size of the facility.” 2010 Proposed
Area Boilers Rule, 75 Fed. Reg. at 31,907; National Emission
Standards for Hazardous Air Pollutants for Major Sources:
Industrial, Commercial, and Institutional Boilers and Process
Heaters (2010 Proposed Major Boilers Rule), 75 Fed. Reg.
32,006, 32,026 (June 4, 2010). Because saving fuel saves
money, common sense suggested that sources would often
find the energy assessment “cost-effective” to implement.
2011 Area Boilers Rule, 76 Fed. Reg. at 15,568 (“By
definition, any emission reduction [achieved as a result of the
energy assessment] would be cost effective or else it would
not be implemented.”); see also 2011 Major Boilers Rule, 76
Fed. Reg. at 15,633.
In addition to costs, the EPA considered non-air quality
health and environmental impacts in general terms,
concluding that “improving energy efficiency reduces
negative impacts on the environment.” 2010 Proposed Area
Boilers Rule, 75 Fed. Reg. at 31,907; 2010 Proposed Major
Boilers Rule, 75 Fed. Reg. at 32,026. Given the nature of the
assessment, the EPA’s somewhat terse analysis of health and
environmental impacts suffices. Performing the assessment
involves rudimentary tasks—examining the boiler and
associated energy systems and drafting a report—that do not
impose meaningful health or environmental impacts. The
same holds for the EPA’s consideration of energy use
requirements. Facilities would expend very little energy in
conducting the one-time assessment, and could conserve
52
energy by implementing the results. The assessment therefore
represents a lawful beyond-the-floor measure.
We also find that the assessment is a valid GACT
management practice. With respect to area sources, the EPA
has discretion to require the use of “generally available
control technologies or management practices . . . to reduce
emissions of hazardous air pollutants.” 42 U.S.C.
§ 7412(d)(5). The EPA justified the energy assessment as a
GACT management practice for oil- and biomass-fired
boilers. See 2011 Area Boilers Rule, 76 Fed. Reg. at 15,567.
The Industry Petitioners challenge that justification,
claiming the energy assessment—which does not require
implementation—cannot “reduce emissions of hazardous air
pollutants.” 42 U.S.C. § 7412(d)(5). We disagree. The EPA
did not need to make implementation mandatory to make the
assessment lawful. Under the CAA, the EPA may sometimes
act with a soft touch, rather than a firm hand. Here, the EPA
selected a soft touch, requiring an assessment but not
implementation. It was not unreasonable for the EPA to
conclude, “after considering the structure of the requirement,
the incentives it presents, and the likely behavior of
sources, . . . that sources will find it cost-effective to
implement the conservation measures identified in the energy
assessment.” 2011 Area Boilers Rule, 76 Fed. Reg. at 15,573.
If the results were implemented, HAP emissions would be
reduced. For present purposes, that is enough.
For those reasons, we reject the Industry Petitioners’
challenges to the energy-assessment requirement.
53
D. RECORDKEEPING REQUIREMENT FOR CISWI UNITS
Section 7429 regulates combustion units that burn solid
waste; units that do not burn solid waste will generally be
regulated under section 7412. RCRA defines the term “solid
waste” to mean (in part) “discarded material . . . resulting
from industrial [or] commercial . . . operations.” 42 U.S.C.
§ 6903(27); see id. § 7429(g)(6) (directing that “solid waste”
carry “the meanings established by the Administrator
pursuant to” RCRA). On the same day the EPA issued a rule
setting emission standards for CISWI, it issued a separate rule
fleshing out the meaning of solid waste in the context of
combustion units. See NHSM Rule, 76 Fed. Reg. at 15,456.
The NHSM Rule generally provides that “non-hazardous
secondary materials that are combusted are solid wastes,”14
subject to several exceptions and exemptions. 40 C.F.R.
§ 241.3(a). Among the exceptions, non-hazardous secondary
materials that meet certain “legitimacy” criteria do not qualify
as solid waste. See id. § 241.3(b), (d). Source owners and
operators may also seek a finding from the EPA that
particular materials do not constitute solid waste when
combusted by a third party. Id. § 241.3(c). And the rule
exempts altogether a variety of materials from the definition
of solid waste, including “traditional fuels.” Id. § 241.2.
The NHSM Rule is self-implementing: each source
owner or operator must determine whether combusted
materials meet the definition of solid waste. See 2011 CISWI
Rule, 76 Fed. Reg. at 15,740. To ensure that owners and
14
The NHSM Rule defines non-hazardous secondary material
to “mean[] a secondary material that, when discarded, would not be
identified as a hazardous waste.” 40 C.F.R. § 241.2.
54
operators “review and apply” the NHSM Rule and its
exceptions, the EPA issued strict recordkeeping requirements.
Id. Owners and operators who determine the secondary
materials they combust are not solid waste must “keep a
record” justifying that decision. 40 C.F.R. § 60.2175(v).
Failing to file records carries consequences. For units
combusting discarded material other than traditional fuels, the
failure to “keep and produce records” results in the
determination that “the operating unit is a CISWI unit.” Id.§§ 60.2265, 60.2875 (containing an identical provision).
Industry Petitioners challenge this last provision of the
CISWI Rule.15 They argue that the EPA cannot automatically
treat units that fail to keep certain paperwork as CISWI units.
Section 7429 permits regulation of “solid waste incineration
units”—not units whose owners fail to file paperwork. As a
result, the Industry Petitioners ask this court to invalidate the
regulatory provision as exceeding the EPA’s statutory
authority.16
We decline the invitation. At Chevron’s first step, we
find that “Congress did not speak directly, let alone clearly, to
15
In their reply brief, the Industry Petitioners clarify that they
do not challenge the EPA’s authority to require sources to keep
records.
16
The Industry Petitioners also argue the EPA arbitrarily
failed to provide sufficient notice of the recordkeeping
presumption. We disagree. The Industry Petitioners had sufficient
notice of the CISWI Rule, which was promulgated after notice and
comment and “give[s] fair warning of the conduct it prohibits.”
Gen. Elec. Co. v. EPA, 53 F.3d 1324, 1328 (D.C. Cir. 1995)
(quoting Gates & Fox Co. v. OSHRC, 790 F.2d 154, 156 (D.C. Cir.
1986)).
55
this issue.” Am. Chem. Council v. EPA, 337 F.3d 1060, 1064
(D.C. Cir. 2003). Section 7429 regulates “solid waste
incineration units,” a phrase that Congress defined “plainly
and broadly to include ‘a distinct operating unit of any facility
which combusts any solid waste material from commercial or
industrial establishments or the general public.’” NRDC I,
489 F.3d at 1257 (emphasis omitted) (quoting 42 U.S.C.
§ 7429(g)(1)). In NRDC I, we vacated an earlier iteration of
the CISWI Rule that narrowed the scope of that definition
beyond what its language would bear. See id. at 1257-58.
When the Congress commanded the EPA to regulate units
that burn “any” solid waste, the Congress meant what it said.
See id.In this case, the EPA included within the revised CISWI
Rule a presumption designed to enforce the Congress’s
command. Section 7429 nowhere addresses whether the EPA
may establish presumptions to ensure its regulations reach all
sources burning solid waste. At the same time, the Congress
plainly intended the EPA to regulate sources burning “any”
solid waste, a goal presumably advanced by the
recordkeeping presumption. See id. Against that backdrop,
we cannot conclude that the presumption offends the text or
purpose of section 7429.
Moving to Chevron’s second step, we conclude the
recordkeeping presumption is reasonable. In American
Chemistry Council, we upheld a regulation issued under
RCRA defining hazardous waste to include any mixture or
derivative of hazardous substances. See 337 F.3d at 1064-65.
“[B]ecause many mixtures of and derivatives from hazardous
wastes are themselves hazardous, it [was] reasonable for the
EPA to assume that all such mixtures and derivatives are
hazardous until shown otherwise.” Id. at 1065. In that
56
context, it made good sense for the EPA to “[p]lac[e] the
burden upon the regulated entity” to show that a given
substance lacked “hazardous characteristic[s].” Id.Similar reasoning applies here. The EPA crafted the
presumption to reach sources likely to be burning solid waste,
namely, those burning discarded materials other than
traditional fuels. See 42 U.S.C. § 6903(27) (defining “solid
waste” to include, among other things, “discarded material”);
40 C.F.R. § 241.2 (exempting traditional fuels, defined as
“materials that are produced as fuels . . . that have not been
discarded,” from the definition of solid waste). Such sources
are subject to strict recordkeeping requirements. See 40
C.F.R. § 60.2175(v). Within those confines, placing the
burden on unit operators who have the mandatory obligation
and the information to establish their non-regulable status is
reasonable. Cf. Am. Chem. Council, 337 F.3d at 1065.
There is, however, a difference between the presumption
in this case and the one we upheld in American Chemistry
Council. The CISWI recordkeeping presumption appears to
turn on the failure to file paperwork, rather than the presence
of a regulated substance. However broadly the Congress
defined “solid waste incineration unit” in section 7429, the
Congress did not allow for the regulation of non-waste
burning sources—even when those sources fail to file
paperwork. Indeed, had the EPA attempted to regulate
sources based purely on a failure to file paperwork, we may
well have reached a different conclusion.
But the CISWI presumption does not stretch so far. As
explained, the presumption depends on factors beyond the
mere failure to keep records. Sources subject to the
presumption burn materials likely to qualify as solid waste,
57
and must satisfy demanding recordkeeping requirements. The
EPA acted reasonably when it presumed such sources were
burning solid waste.
Despite the provision’s narrow reach, the Industry
Petitioners fear it will sweep up sources not burning solid
waste. To the extent that possibility exists, sources
wrongfully regulated as CISWI have multiple forms of
recourse. Most obviously, sources can prepare and file the
records they were already required to make under 40 C.F.R.
§ 60.2175(v). They can also avail themselves of procedures
designed to identify non-waste materials in 40 C.F.R. § 241.3.
The existence of these safety valves calms concerns that the
presumption will regulate non-waste burning sources.
We therefore reject the Industry Petitioners’ challenges
to the recordkeeping presumption.17
E. WASTE-STREAM VARIANCE FOR SRI UNITS
The EPA regulated SRIs as a subcategory in the CISWI
Rule. See Memorandum from Eastern Research Group, Inc.,
to Toni Jones, U.S. Environmental Protection Agency, CISWI
Emission Limit Calculations for Existing and New Sources
for the Reconsideration Final Rule (Jones Mem.) (Nov. 16,
2012) (No. 11-1125 J.A. 1159, 1162). There are 28 SRI units,
17
The Industry Petitioners also contend that the CISWI Rule
functions as a form of injunctive relief in violation of 42 U.S.C.
§ 7413(a). That is incorrect. The provision is neither styled nor
operated as a form of injunctive relief. Cf. 42 U.S.C. § 7413(a)
(permitting the Administrator to issue, among other forms of relief,
“an administrative penalty order” or “an order requiring [a person
in violation of EPA regulations] to comply with such requirement
or prohibition”).
58
all of which are located in Alaska, and the EPA had emissions
data for nine of them. Id. As explained supra § I.B.3, the
EPA used the pollutant-by-pollutant approach to establish
MACT emission standards for these units. For new-unit
standards, the EPA determined which of the nine units had the
lowest emissions for a particular pollutant and set the MACT
floor for that pollutant at the level achieved by the identified
unit. See 42 U.S.C. § 7429(a)(2) (explaining that MACT
floors for new units must be set at “the emissions control that
is achieved in practice by the best controlled similar unit”).
When setting MACT floors for existing units, the EPA had to
calculate the average level of emissions achieved by the best
performing 12 per cent of units. See id. It therefore
determined which four sources had the lowest emissions for a
given pollutant and set the emissions standard for that
pollutant at the average level achieved by those four units.
The Industry Petitioners argue that the EPA’s approach
was unlawful because it failed to account for the unique role
that waste inputs play in emissions from SRIs. Unlike larger
incinerators, SRIs burn small batches of waste at a time. Some
batches include cleaner waste, such as wood and cardboard,
while others include waste, such as sewage, that generates
large quantities of SO2 and other pollutants. Moreover,
existing SRIs cannot use certain “end-of-stack” control
technologies like wet scrubbers due to the Alaskan climate.
The Industry Petitioners thus contend that emissions from
SRIs are more closely tied to waste input than are emissions
from other types of incinerators. This difference, they assert,
required the EPA to take into account, when determining
which SRI units were best performing for MACT floor
purposes, the kind of waste an SRI unit was burning at the
time of testing. Because the Agency did not do so, the
59
Industry Petitioners contend the MACT standards for SRIs are
arbitrary and capricious. We disagree.
To support their challenge, the Industry Petitioners
advance two arguments, neither of which has merit.
Petitioners first point to section 7429(a)(3), which directs the
EPA to base emission standards on “methods and
technologies for removal or destruction of pollutants before,
during, or after combustion.” 42 U.S.C. § 7429(a)(3).
According to the Industry Petitioners, this language requires
the EPA to identify best performing units for MACT purposes
by considering which units are best at removing or destroying
pollutants. The Industry Petitioners assert that the Agency did
not do this. Instead, they contend, the EPA set standards
without regard to whether that unit happened to be burning
cleaner waste. And, according to the Industry Petitioners,
remote incinerators in Alaska cannot control their waste
inputs because the core purpose of SRIs is to burn waste that
is impracticably far from municipal landfills. The fact that
emissions levels varied dramatically during test runs for the
SRI units, they claim, is thus the result of random variance in
the type of waste the unit was combusting, rather than any
“method” or “technology” aimed at “removing” or
“destroying” pollutants.
The EPA responds that the approach it adopted for SRIs
complies with section 7429(a)(3) because “waste
segregation”—that is, diverting dirtier waste to landfills and
burning only cleaner waste—is a “method . . . for removal . . .
of pollutants before . . . combustion.” See 42 U.S.C.
§ 7429(a)(3). In fact, during notice and comment, the EPA
estimated that many SRIs would choose to comply with the
MACT standards by segregating their waste instead of by
installing expensive control technologies. See Jones Mem.
60
The Agency also determined that waste segregation was
possible for SRIs because their waste often contained
materials that could be recycled. Id. Finally, the Agency
factored in any additional variance in emissions from these
units by calculating the MACT floors according to the UPL
formula described at supra §§ I.B.1.a, IV.C. For these
reasons, the Agency contends, it did not need to consider
further any variation in emissions that might be caused by
differences in waste inputs for SRIs.
The EPA has the better argument, based on both text and
precedent. Textually, waste segregation plainly can be a
“method[]” for “removal” of pollutants “before” combustion.
See 42 U.S.C. § 7429(a)(3). Accordingly, the EPA, when
setting MACT floors, could not have looked solely to
technologies used to reduce emissions during combustion.
Accord Sierra Club v. EPA (Sierra Club II), 479 F.3d 875,
883 (D.C. Cir. 2007) (per curiam). Instead, the plain
language of section 7429(a)(3) requires the Agency to
consider whether emission reductions can be achieved by
non-combustion-related controls such as using cleaner fuels
or waste inputs. Accord id. The statute supports the approach
that the Agency took here.
Our holding in Sierra Club II confirms that our
conclusion is correct. In that case, the EPA had
acknowledged that kilns emitted lower levels of pollutants
when burning cleaner clay but nevertheless based MACT
standards only on the emission reductions achieved by control
technology during the combustion process. Id. at 882. The
Agency explained that clean clay existed only in certain areas
and that transportation of the clay over long distances was
impractical. Id. The EPA therefore considered only those
emission reductions that were attributable to “deliberate steps
61
kiln operators [took] to reduce emissions rather than to the
‘happenstance’ of being located near cleaner clay.” Id. at
883. But we rejected that approach, finding that “the Clean
Air Act requires neither an intentional action nor a deliberate
strategy to reduce emissions.” Id. Instead, where “non-
technology factors” affect emission levels, we held the EPA
must consider those effects when setting MACT floors. Id.Applying that same reasoning, the EPA acted reasonably
when it decided to consider the emissions reduction that could
be achieved by waste segregation in SRI units before
combustion. This is true even if an element of
“happenstance” plays into an SRI unit’s ability to segregate
its waste. And, had the EPA instead determined that the best
performers were those SRI units that most effectively reduced
pollutants only during combustion, as the Industry Petitioners
suggest, the resulting MACT standards may have run afoul of
our holding in Sierra Club II. We cannot, as a result, find the
Agency’s choice to avoid that outcome unreasonable.
The Industry Petitioners’ second argument also comes up
short. According to Petitioners, the EPA selected the best
performers for SRIs merely because those units happened to
be burning batches of cleaner waste at the time of the
emissions test. They claim this happenstance resulted in test
data that did not reasonably estimate the typical performance
of the units, and thus misidentified the best performers. See
Cement Kiln, 255 F.3d at 862 (finding that although the EPA
has authority to estimate which units perform best, its
methodology must “provide[] an accurate picture of the
relevant sources’ actual performance”). Petitioners further
argue that the Agency’s use of the UPL method to account for
variability did not fix this problem because the EPA applies
that method only after identifying the best performers.
62
If the record supported this argument, it might well be
persuasive; in NACWA, we accepted a similar contention that
the EPA’s dataset for determining MACT floors must fairly
represent a unit’s typical performance. See 734 F.3d at 1146.
But the record here does not support the Industry Petitioners’
position. None of the evidence on which Petitioners rely can
bear the weight they would have us place on it.
First, Petitioners cite evidence indicating that XTO
Energy, which operates the incinerator that the EPA deemed
the best performer for SO2, was burning low-sulfur “waste
wood, cardboard, and oily waste” during the relevant test
runs. See ConocoPhillips Co., Comment on EPA’s Proposed
National Emission Standards for Hazardous Air Pollutants,
EPA-HQ-OAR-2003-0119 (Feb. 12, 2012) (No. 11-1125 J.A.
1036). But the record does not show that the resulting test
data were unrepresentative of XTO’s typical performance
because the record says nothing about what XTO typically
burns. Id.
Second, Petitioners note that Drift River, the unit the
EPA deemed the worst performer for SO2, had emissions
results similar to XTO Energy’s when burning low-sulfur
waste, but results over 1,000 times higher when burning high-
sulfur waste. See id. (No. 11-1125 J.A. 1032-33). But again,
the record does not say anything about the type of waste Drift
River typically burns or its sulfur content; it merely
demonstrates that the unit’s test results varied greatly from
one run to the next. See id.
Third, Petitioners point to additional test data they
provided for the Kuparuk unit, a source that met the EPA’s
MACT standards for NOx. See id. (No. 11-1125 J.A. 1017,
1027-28). They claim this data shows that the Kuparuk unit
63
“consistently” emits NOx levels exceeding that standard when
burning sewage sludge. Id. This claim is both factually
untrue—as the data reveals exceedances on only one day—
and says nothing about whether the test data that the EPA
used was representative of Kuparuk’s typical performance.
Id.
Instead, the record supports the EPA’s assertion that it
gave Petitioners “multiple opportunities” to present data on
the variability of waste streams for SRIs, but Petitioners never
provided a reasonable empirical basis upon which the Agency
could adjust the MACT standards due to this variability. The
Industry Petitioners have thus not met their burden to show
that the EPA’s test data was unrepresentative of SRI units’
actual or typical performance.
In sum, no record evidence suggests that the current SRI
emission standards are not achievable. The Industry
Petitioners instead offer only general statements about the
“small batch” nature of SRIs and the difficulty of using waste
segregation or other controls in remote locations. These
factors alone do not call into question the EPA’s assertion that
controls such as waste segregation and technology upgrades
are a feasible means of achieving compliance with the MACT
floors that it established. See 2011 CISWI Rule, 76 Fed. Reg.
at 15,730 (explaining that the MACT floors will require SRIs
to employ “the best demonstrated technologies that are
technologically feasible at these facilities,” such as
afterburners and waste segregation, and noting that such
controls “are sufficient to meet the MACT floor limits”). As
a result, the EPA’s action here was reasonable; the Agency
did not need to account further for waste stream variance in
setting MACT floor standards for these SRI units.
64
F. CARBON MONOXIDE AS A SURROGATE
In setting MACT standards for major boilers, the EPA
used carbon monoxide (CO) as a surrogate for several of the
HAPs that the Agency was required to regulate. A surrogate
is another chemical that stands in as a proxy for the regulated
HAP when the EPA sets numeric emission standards. The
EPA regulates the surrogate in order to regulate the HAP,
sometimes because the HAP itself is too difficult to measure.
We have previously approved the use of surrogates where
the EPA’s choice of a surrogate for the HAP is “reasonable.”
See, e.g., Nat’l Lime Ass’n, 233 F.3d at 637. Here, the
Industry Petitioners claim the EPA’s use of CO as a surrogate
was not reasonable for a particular type of emissions—
organic HAP emissions from coal-fired boilers—for two
reasons. First, the EPA based the MACT standards on
datasets that contained numerous “non-detects” for these
organic HAPs. Second, the Agency failed to explain why it
used CO as a surrogate for major boilers, but used work-
practice standards to regulate similar emissions from other
types of boilers in another rule. We find no merit in either
argument and, accordingly, deny this challenge.
The Industry Petitioners base their first argument on a
deficiency in the EPA’s dataset for coal-fired boilers’
emissions—i.e., the dataset contained numerous “non-detects”
for organic HAP emissions. A test result is considered a
“non-detect” when emissions testing returns a value below
that which the test methods are capable of detecting.
According to the Industry Petitioners, multiple non-detects in
a dataset demonstrate that it is “not feasible” to set a numeric
emission standard for the affected HAP. As a result, they
argue, the EPA should have set work-practice standards for
65
these HAPs under section 7412(h)(2), which permits the EPA
to set such standards when it is “not feasible” to set a numeric
emission standard. See 42 U.S.C. § 7412(h)(2).
This argument fails because Petitioners have not
explained how the non-detects here made setting numeric
emissions “not feasible,” as that term is defined in the CAA.
The CAA expresses a clear preference for MACT emission
standards and limits the EPA’s ability to fashion more flexible
work-practice standards. Compare id. § 7412(d)(3)
(providing that emission standards “shall not be less
stringent” than the MACT floor), with id. § 7412(h)(1)
(permitting work-practice standards only if MACT standards
are “not feasible”). To set a work-practice standard for these
emissions, in fact, the EPA would need to find that it is
infeasible to set a numeric standard for a particular HAP. Id.
§ 7412(h)(1). And, as relevant here, the statute defines setting
a numeric standard as “not feasible” where “the application of
measurement methodology to a particular class of sources is
not practicable due to technological and economic
limitations.” Id. § 7412(h)(2)(B).
This is a high bar and Petitioners have not demonstrated
that the non-detects they have identified meet it. During
notice and comment, the Agency reasonably explained that
non-detects are present in many of its datasets because they
are inherent to the imprecision associated with measuring
boiler emissions. See, e.g., 2011 Major Boilers Rule, 76 Fed.
Reg. at 15,623. The EPA’s scientific conclusion that its data
was nevertheless sufficient to set numeric standards receives
an “extreme degree of deference.” Kennecott Greens Creek
Mining Co. v. Mine Safety & Health Admin., 476 F.3d 946,
954-55 (D.C. Cir. 2007) (quotations omitted). And the
Industry Petitioners never explain here why the particular
66
level of non-detects found in this dataset nevertheless made a
numeric standard infeasible. Although the Industry
Petitioners point to several comments asserting that no coal-
fired boiler could meet the current numeric standards in all
HAP categories, these general comments say nothing about
the relevant question under the statute: whether it was feasible
to establish numeric standards for organic HAP emissions in
light of the non-detects in the coal-fired boiler datasets.
We also reject the Industry Petitioners’ second argument
that the EPA needed to explain why it established work-
practice standards for other types of boilers in the unrelated
“Utility MATS” rule. We take an “every tub on its own
bottom” approach to the EPA’s setting of emission standards
pursuant to the CAA. Sierra Club I, 353 F.3d at 986. The
adequacy of the underlying justification offered by the
Agency is what matters in an arbitrary-and-capricious
review—not what the Agency did on a different record
concerning a different industry. Id. As a result, we cannot
find that it was unreasonable for the EPA to use CO as a
surrogate in setting numeric standards for coal-fired boilers on
this basis. Nor can we find that the EPA was required on
reconsideration to explain the discrepancy between its
approach to organic HAP emissions in these two rules, as
Petitioners assert. See id. at 987 (“EPA could have noted
where the bases for its decision in this case differed from
those with respect to other decisions in other cases, as was
done in the EPA’s brief to this court . . . but such explanations
are not required given the different contexts of the various
rulemakings.”).
67
G. HEALTH-BASED EMISSIONS LIMITATION FOR HCL
In the Major Boiler Rule, the EPA chose not to exercise
its discretion to create more lenient emission standards for
hydrogen chloride (HCl) based on health. The Industry
Petitioners challenge this decision as arbitrary and capricious
because, they claim, the Agency considered impermissible
factors in reaching the decision and departed from its previous
position without adequate justification. We disagree and hold
the EPA reasonably chose not to establish a health-based
emissions limitation for HCl.
The EPA generally must establish emission standards for
all listed pollutants emitted from a source category based on
what the best performing similar sources have achieved, i.e.,
the MACT floor. The Agency, however, may consider
adopting alternative health-based emission standards—which
are more lenient—for pollutants with an established health
threshold. See 42 U.S.C. § 7412(d)(4). The statutory
language permitting these alternative standards is
discretionary, providing that “[w]ith respect to pollutants for
which a health threshold has been established, the
Administrator may consider such threshold level, with an
ample margin of safety, when establishing emission standards
under this subsection.” Id. (emphasis added). But, even if the
EPA considers, in its discretion, a health-based emission
standard, the statutory text nowhere requires that the EPA
adopt a more lenient standard than the MACT floor. This
provision thus allows, but does not require, the EPA to adopt
a standard more lenient than the MACT floor, subject to two
critical restrictions: the Agency must determine (1) that there
is an established health threshold, and (2) that the established
threshold would provide “an ample margin of safety.”
68
Using this authority, the EPA considered and adopted
health-based emission standards for HCl in an earlier
rulemaking for major boilers. See 2004 Boilers Rule, 69 Fed.
Reg. at 55,240-41. At the time, the Agency based its decision
on three key findings: a health threshold was established for
HCl, adverse health effects were unlikely at emissions below
that level, and low HCl emissions from major source boilers
made HCl a “particularly well-suited” candidate for more
lenient standards. Id. at 55,241. The EPA also said, however,
that it was not embracing a general policy for HCl, but would
instead “undertake in each individual rule to determine
whether it is appropriate to exercise [the Agency’s]
discretion” to adopt such standards. Id. We later vacated that
rule without considering the merits of the EPA’s HCl
decision. See NRDC I, 489 F.3d 1250.
The EPA again chose to consider a health-based standard
for HCl in the current rulemaking, but this time declined to set
such a standard. 2010 Major Boilers Rule, 75 Fed. Reg. at
32,030. The EPA explained that it continued to interpret its
authority under section 7412(d)(4) to require that it find a
health threshold exists, with an ample margin of safety, before
using its discretion to depart from an established MACT floor.
Id. The Agency reasoned further that, even if it made a
finding that a health threshold exists, the discretionary nature
of the authority allowed it to weigh additional factors when
choosing whether to adopt the more lenient health-based
standard. Id. Those factors included: the potential for
cumulative adverse health effects due to concurrent exposure
to other HAPs or emissions from other nearby sources;
potential impacts of increased emissions on ecosystems; and
reductions in emissions of other pollutants, also known as
“co-benefits,” achieved through enforcement of the HCl
MACT floor. Id. at 32,030-31.
69
Applying this interpretation, the EPA suggested in its
proposed rule that a health-based standard for HCl might not
be appropriate because these additional health and
environmental considerations cautioned against a more lenient
emission standard. Id. at 32,031. The Agency acknowledged,
in particular, that its decision in the 2004 rule was based on
data that considered only the chronic respiratory effects of
HCl exposure. Id. While affirming the validity of those
findings, the EPA explained that those chronic impact studies
did not consider the additional variables it had now identified,
nor did it consider the potential acute or carcinogenic effects
that might be caused by HCl exposure. Id. And, because of
these potential (though unproven) risks, the Agency resolved
that it currently lacked sufficient information to establish an
HCl emission standard that would protect health with an
ample margin of safety. Id. It thus requested additional data
from stakeholders and the regulated community to help
address its concerns, including information regarding the
potential cumulative effects of HCl emissions from boilers
and other nearby sources. Id.
After receiving numerous comments on the issue, the
EPA declined to set a health-based standard in the final rule
for two primary reasons: (1) the comments had not provided
sufficient data on potential cumulative health and
environmental effects caused by HCl emissions from boilers
and other nearby sources; and (2) the comments affirmed the
potential co-benefits that limiting HCl emissions might have
in lowering emissions of other HAP and non-HAP pollutants.
2011 Major Boilers Rule, 76 Fed. Reg. at 15,643-44.
According to the EPA, its consideration of these co-benefits
was not a regulation of other pollutants; rather, it was simply
choosing not to ignore the purpose of the CAA—to reduce the
negative health and environmental effects of HAP
70
emissions—when exercising its discretionary authority under
the Act. Id. at 15,644.
The Industry Petitioners contend that the EPA’s
consideration of the broad potential health and environmental
impacts of HCl rendered the Agency’s decision arbitrary and
capricious. In particular, they argue that the Agency based its
decision on two impermissible factors that were not supported
by the record: (1) the potential cumulative effects of
emissions from boilers and other nearby sources, and (2) the
co-benefits of setting a more stringent MACT floor standard
for HCl. We disagree on both counts.
The statutory text and purpose of section 7412(d)(4)
amply support the Agency’s decision to consider potential
cumulative risks associated with emissions from boilers and
other nearby sources. Although other CAA provisions require
the EPA to set emission standards based on the emissions
from a particular source, section 7412(d)(4)’s plain language
is not focused on emissions from any particular source.
Compare 42 U.S.C. § 7412(d)(3) (instructing the EPA to set
emission standards for sources at the level achieved in
practice by the best controlled similar source), with id.§ 7412(d)(4) (containing no mention of emissions from a
particular source). The EPA’s consideration of the
cumulative impacts from these emissions is also relevant to
the Agency’s statutory mandate to ensure that a health
threshold would protect health with an “ample margin of
safety.” As such, the Agency had discretion to consider the
potential risks associated with the cumulative emissions of
boilers and other nearby sources under this provision.
The EPA was likewise free to consider potential co-
benefits that might be achieved from enforcing the HCl
71
MACT floor. Section 7412(d)(4)’s text does not foreclose the
Agency from considering co-benefits and doing so is
consistent with the CAA’s purpose—to reduce the health and
environmental impacts of hazardous air pollutants. The
Agency was under no obligation to ignore the CAA’s purpose
in making a final decision on whether to exercise a
discretionary authority.
The Industry Petitioners attempt to refute this
straightforward conclusion by pointing to “restrictions” in
another provision, section 7412(d)(2). No. 11-1108 Indus.
Pet’rs’ Br. 55-56. This provision requires the EPA to
consider costs, non-air quality health and environmental
impacts, and energy requirements in setting maximum
achievable emission standards. Petitioners contend that these
same “restrictions” must be read into section 7412(d)(4). But,
even if we assume Petitioners are correct that these factors
restrict the Agency’s ability to consider other factors under
section 7412(d)(2), that provision furthers the statute’s
command to set the strictest possible emission standards
above what has already been achieved (i.e., the MACT
floors). Section 7412(d)(4), by contrast, is a permissive
authority for the EPA to abandon already achieved emission
standards. We do not read limits on the EPA’s authority to
set more stringent standards into a provision laying out the
EPA’s authority to set more lenient standards. If anything,
the difference between the provisions cuts the other way.
Section 7412(d)(4) does not specify the factors that
Petitioners argue for, while section 7412(d)(2) does. This
difference shows that Congress knew how to provide such
limits where it found them necessary. We thus find no basis
to conclude that the EPA could not consider potential
cumulative effects or co-benefits in rejecting a more lenient
health-based HCl standard.
72
Finally, the Industry Petitioners claim that the EPA’s
decision was arbitrary because the Agency failed to support
its reversal from the 2004 rule, in which it set health-based
emission standards for HCl. Because the EPA changed its
position, the Petitioners contend that the Agency had to
present factual support for its decision to disregard the facts
and circumstances that underlay its prior adoption of a health-
based HCl standard. See FCC v. Fox Television Stations,
Inc., 556 U.S. 502, 516 (2009). The Agency failed to do this,
Petitioners say, because it relied on a data gap regarding the
potential cumulative effects of HCl exposure. But this
argument fares no better than Petitioners’ first.
At the outset, Petitioners misstate the EPA’s burden to
justify its change in policy. Although an agency does not
generally need to provide a more substantial explanation or
reason for a policy change than for any other action, it must
do so where “its new policy rests upon factual findings that
contradict those which underlay its prior policy.” Id. at 515.
In that circumstance, “it is not that further justification is
demanded by the mere fact of policy change; but that a
reasoned explanation is needed for disregarding facts and
circumstances that underlay or were engendered by the prior
policy.” Id. at 515-16. The EPA, therefore, was not required
to refute the factual underpinnings of its prior policy with new
factual data. The Agency only needed to provide a reasoned
explanation for discounting the importance of the facts that it
had previously relied upon. Id.The EPA did so here by explaining that its prior decision
focused too narrowly on the chronic respiratory effects of HCl
emissions without considering the broader implications of
such emissions on health and environmental conditions. See
2010 Proposed Major Boilers Rule, 75 Fed. Reg. at 32,030-
73
31. In so doing, the EPA neither contradicted nor abandoned
the factual findings it made in its earlier rulemaking. It
instead acknowledged that those findings were more limited
than what it now considered necessary to justify the exercise
of its discretion to set a health-based standard. Id. For
example, the Agency noted that: (1) little research had been
done on HCl’s carcinogenicity or on the toxicity of mixtures
of HCl and other respiratory irritants emitted from boilers;
and (2) the Agency had no data about peak short-term
emissions of HCl from major boilers that might create risks of
acute exposure. Id.
These enumerated concerns were sufficient to support the
Agency’s decision not to adopt a health-based standard.
Section 7412(d)(4) does not require that the EPA present
affirmative factual data to reject a health-based standard. The
provision requires just the opposite: in order to impose a
health-based standard, the Agency must find that a health
threshold can be set that provides an ample margin of safety.
The EPA here determined that it could not do so, in part
because it lacked relevant data like that discussed above.
2011 Major Boilers Rule, 76 Fed. Reg. at 15,643-44. In other
words, the EPA could not determine that any health threshold
would provide an ample margin of safety to protect health.
Without such a finding, the EPA could not invoke its
discretionary authority under the statute. Id. There was thus
nothing impermissible in the EPA’s reliance on a lack of data
in rejecting a more lenient health-based standard. The EPA’s
decision not to adopt health-based emission standards for HCl
was not arbitrary and capricious.
74
H. EMISSIONS AVERAGING OF MULTIPLE CISWI UNITS
IN ONE FACILITY
Certain industry entities urged the EPA to allow facilities
with more than one CISWI unit to demonstrate MACT
compliance by showing that the average HAP emissions
across all units at that location fell under the relevant cap.
They pointed to the EPA’s allowance of emissions averaging
in the Major Boilers Rule but the Agency defended its
disparate treatment because, in its view, “[t]he applicability of
CISWI is such that each unit is an affected facility.” See 2011
Proposed CISWI Rule on Reconsideration, 76 Fed. Reg. at
80,463. It subsequently elaborated that it did “not believe [it
has] the legal authority to allow emissions averaging in
CISWI or under section [7429] generally because each
individual unit is an affected facility.” CISWI Rule—
Responses to Comments, at 195-96. The Industry Petitioners
challenge the disallowance of facility-wide averaging for
CISWIs, arguing that “unit” cannot mean “facility” because
section 7429(g)(1) defines “solid waste incineration unit” as
“a distinct operating unit of any facility” and therefore the
EPA’s rule fails Chevron step 1. They also argue the EPA’s
conflation of “unit” and “facility” is unreasonable, and thus
violates Chevron step 2, because the EPA has allowed
emissions averaging in a different section 7429 rule and in a
number of section 7412 rules.
Although the Industry Petitioners’ point is well taken—
the plain terms of the CAA foreclose the EPA’s conflation of
a CISWI “unit” and “affected facility,” see 42 U.S.C.
§ 7429(g)(1) (“facility” is comprised of “units”)—we agree
that the EPA has no statutory authority to allow emissions
75
averaging under section 7429.18 Section 7429 requires the
EPA to regulate emissions from all “solid waste incineration
units,” 42 U.S.C. § 7429(a)(2); see also id. § 7429(a)(4), and
the CAA defines a “solid waste incinerator unit” as “a distinct
operating unit” of a “facility,” id. § 7429(g)(1) (emphasis
added). In other words, because the CAA mandates that the
EPA regulate each “distinct” CISWI unit in a “facility,” the
EPA cannot allow emissions averaging of all CISWI units in
a facility. See id.
For this reason, the Industry Petitioners’ Chevron
challenge fails, notwithstanding the EPA’s minimal
explanation set forth in its proposed CISWI Rule. It is
axiomatic that an agency must “articulate[] an adequate
explanation for its action,” Int’l Fabricare Inst. v. EPA, 972
F.2d 384, 389 (D.C. Cir. 1992) (emphasis added); see also
State Farm, 463 U.S. at 48, but the EPA’s failure to do so
here cannot create statutory authority that does not exist. And
because the EPA has no authority under section 7429 to allow
emissions averaging of multiple CISWI units in one facility,
the Petitioners’ Chevron argument does not carry the day.19
18
The EPA does have statutory authority under section 7412
to allow facility-wide emissions averaging in the Major Boilers
Rule. See 42 U.S.C. § 7412(a)(1) (“major source[s]” defined as
“any stationary source or group of stationary sources located within
a contiguous area and under common control” (emphasis added));
see also id. § 7411(a)(3) (“stationary source” defined as “any
building, structure, facility, or installation which emits or may emit
any air pollutant”).
19
The EPA concedes that it once allowed, in a different rule,
emissions averaging for units subject to section 7429 but has since
concluded that it does not have the statutory authority to do so.
Although the Industry Petitioners argue that the Agency arbitrarily
76
IV. ENVIRONMENTAL PETITIONERS’
CHALLENGES
A. CARBON MONOXIDE AS A SURROGATE
As explained at supra §§ I.B.1 and III.F, the EPA used
carbon monoxide (CO) as a surrogate for several non-
dioxin/furan organic HAPs when the Agency set the MACT
floors for major boilers. In support of this approach, the EPA
found that both CO and these HAPs were the products of
“incomplete combustion.” 2010 Proposed Major Boilers
Rule, 75 Fed. Reg. at 32,018. The Agency concluded as a
result that CO was a reasonable surrogate because:
(1) minimizing CO emissions would minimize these HAPs;
(2) methods used for the control of these HAP emissions
would be the same methods used to control CO emissions
(i.e., good combustion or using an oxidation catalyst);
(3) standards limiting CO emissions would result in decreases
in these HAP emissions; and (4) establishing emission limits
for individual organic HAPs would be impractical and costly.
Id. Although several commenters challenged aspects of this
reasoning, the EPA ultimately stuck with its decision to use
CO as a surrogate for non-dioxin/furan organic HAP
emissions, without further explanation, in the final Major
Boilers Rule. See 2013 Major Boilers Rule, 78 Fed. Reg. at
7,145 (explaining the EPA was denying Sierra Club’s petition
to reconsider the suitability of CO as a surrogate for non-
organic HAPs based on the reasoning provided by the Agency
in the 2010 proposed rule).
changed its position, the fact that the EPA may have acted outside
its authority in a rule is not at issue here. “[P]revious statutory
violations,” of course, “cannot excuse” new ones. New Jersey v.
EPA, 517 F.3d 574, 583 (D.C. Cir. 2008).
77
The Environmental Petitioners challenge this decision,
arguing that the EPA has not adequately explained how
setting emission standards for CO will accomplish what the
statute plainly requires: that the EPA set emission standards
for organic HAPs at the average level achieved by the best
performers with regard to those HAPs. We agree and remand
to the EPA to adequately explain how CO acts as a reasonable
surrogate for non-dioxin/furan organic HAPs. We do not,
however, vacate the current emission standards because we
conclude that the Agency will likely be able to adequately
explain its decision on remand and that vacatur would prove
substantially disruptive.
The EPA may use a surrogate to regulate HAPs under
section 7412 where “reasonable.” See, e.g., Nat’l Lime Ass’n,
233 F.3d at 637. To be reasonable, the emission standard set
for the surrogate must reflect what the best source or best 12
per cent of sources in the relevant subcategory achieved with
regard to the HAP. See Sierra Club I, 353 F.3d at 984. This
requires the surrogate’s emissions to share a close relationship
with the emissions of the HAP. Id. One crucial factor we
have identified for determining whether that close relationship
exists is the availability of alternative control technologies.
See id. at 985. These technologies regulate the HAP without
impacting a surrogate’s emissions, or regulate the surrogate
without impacting the HAP. Id. As we have explained, the
importance of this factor to our reasonableness analysis “is
clear: if EPA looks only to [the surrogate], but HAPs are
reduced [in another] way that does not reduce [the surrogate],
the best achieving sources, and what they can achieve with
respect to HAPs, might not be properly identified.” Id.In the Major Boilers Rule, the EPA proposed using CO as
a surrogate because, as relevant here: (1) the lowest possible
78
CO emissions resulted in the lowest possible HAP emissions,
and (2) the same combustion and oxidation control methods
reduce both types of emissions. See 2010 Proposed Major
Boilers Rule, 75 Fed. Reg. at 32,018. But, during notice and
comment, the EPA failed to directly consider and respond to
several comments that introduced evidence suggesting that
other control technologies and methods could be effectively
used to reduce HAP emissions without also impacting CO
emissions, or vice versa. See, e.g., Inst. of Clean Air Cos.,
Comments on National Emission Standards for Hazardous Air
Pollutants for Major Sources, EPA-HQ-OAR-2002-0058
(Aug. 23, 2010), at 20-21 (No. 11-1108 J.A. 822-23);
Responses to Public Comments on EPA’s National Emission
Standards for Hazardous Air Pollutants for Major Sources,
vol. 2, EPA-HQ-OAR-2002-0058 (Feb. 2011) (No. 11-1108
J.A. 1033, 1035-36, 1049-52). The EPA ultimately decided to
use CO as a surrogate for all non-dioxin/furan organic HAPs
in its final rule without ever addressing whether such
alternative control technologies and methods might be used to
lower organic HAP emissions further. See 2011 Major
Boilers Rule, 76 Fed. Reg. at 15,654; 2013 Major Boilers
Rule, 78 Fed. Reg. at 7,138. Instead, the Agency responded
by doubling down on its assertion that both CO and organic
HAP emissions were the product of poor combustion and, as a
result, optimal combustion would minimize the emissions of
both CO and non-dioxin/furan organic HAPs. 2013 Major
Boilers Rule, 78 Fed. Reg. at 7,145. But this response was no
response at all to the substantial concerns raised in the
comments that other variables might also affect emissions.
Although we afford an agency’s scientific decision “an
extreme degree of deference,” see Kennecott Greens, 476
F.3d at 954-55 (quoting Hüls Am., Inc. v. Browner, 83 F.3d
445, 452 (D.C. Cir. 1996)), we cannot uphold an agency
79
decision that does not consider all relevant factors or fails to
establish a reasonable connection to the facts in the record.
Cf. Ethyl Corp. v. EPA, 51 F.3d 1053, 1064 (D.C. Cir. 1995).
The EPA could not conclude that CO acts as a reasonable
surrogate in this statutory context without at least considering
a key factor: whether the best performing boilers might be
using alternative control technologies and methods that
reduce organic HAP emissions beyond what they achieve by
regulating CO alone. See Sierra Club I, 353 F.3d at 985. We
therefore reject the EPA’s contention that its reason for using
CO as a surrogate—that good combustion would minimize
both CO and non-dioxin/furan organic HAP emissions—was
alone sufficient to support its decision.
We recognize that there might be a context where a
surrogate’s use is reasonable despite the presence of
alternative control methods or technologies, but the Agency
does not explain why it did not need to even consider whether
such methods might further reduce HAPs here. For example,
if the EPA used a surrogate that was closely correlated to the
HAP and set surrogacy emission standards at a level that
would eliminate HAP emissions altogether, the Agency might
not need to account for alternative control technologies in its
final rule. In that case, the use of the surrogate would not call
into question whether the Agency had regulated the HAP as
required by the statute because, after all, nothing is better than
eliminating HAP emissions entirely. But the Agency offers
us no analogous explanation or supportive data here.
Although it is possible that all of the challenged CO emission
standards are in fact set at such a level, the Agency has not
defended the rule on such reasoning. Indeed, the Agency
failed to consider or even comment directly on this issue,
including whether certain post-combustion processes might
increase organic HAP emissions without a corresponding
80
increase in CO emissions. We cannot ignore such an
oversight in this context.
We reject, however, the Environmental Petitioners’ other
argument that combustion-related issues preclude the EPA
from using CO as a surrogate for non-dioxin/furan organic
HAPs. The Petitioners contend that the EPA’s decision to use
CO was arbitrary because record evidence demonstrated a
breakdown in the correlation between CO and organic HAP
emissions at CO emission levels below 130 parts per million
(ppm). But the EPA explained that this apparent breakdown
was most likely caused by the difficulty of measuring the
regulated HAP at such extremely low emission levels, rather
than by a flaw in the correlation between CO and organic
HAPs. 2013 Major Boilers Rule, 78 Fed. Reg. at 7,144-45;
Memorandum from Eastern Research Group, Inc. to Jim
Eddinger, EPA, Revised MACT Floor Analysis for the
Industrial, Commerical, and Institutional Boilers and Process
Heaters National Emission Standards for Hazardous Air
Pollutants—Major Source (2012 MACT Floor Memorandum)
(Aug. 2012), at 11-12 (No. 11-1108 J.A. 1462-63). This is
precisely the sort of scientific judgment to which we must
defer and accordingly, we do so on this point. See Kennecott
Greens, 476 F.3d at 954-55. The Environmental Petitioners
fail to provide any reason to believe that organic HAP
emissions can, in fact, be accurately measured at such low
levels. And the Agency’s explanation also addresses why the
EPA discounted record evidence regarding extremely high
burn temperatures that demonstrated a potential breakdown in
the CO and organic HAP relationship as HAP emissions
approached zero.
Still, the EPA’s failure to address substantial record
evidence on the potential availability of alternative control
81
technologies or methods rendered the Agency’s use of CO as
a surrogate for certain organic HAPs arbitrary and capricious.
We thus remand the portion of the Major Boilers Rule
providing for CO’s use as a surrogate for non-dioxin/furan
organic HAPs to the Agency for further consideration. We do
not, however, vacate the current emission standards based on
CO’s use as a surrogate. We may remand without vacatur
where there is a likelihood of (1) cure on remand, and (2) a
substantial disruptive effect that would result from vacatur.
See Heartland Reg’l Med. Ctr. v. Sebelius, 566 F.3d 193, 197-
98 (D.C. Cir. 2009). Here, vacatur would cause substantial
disruptive effects by removing emission limits for the
regulated HAPs. And it is likely that the EPA will be able to
adequately explain its use of CO on remand after properly
considering the matter. As a result, we decline to vacate the
current standards in the interim.
B. EXCLUSION OF CERTAIN UNITS FROM MACT ANALYSIS
In the Major Boilers Rule, the EPA created subcategories
based primarily on the fuel combusted. See 2013 Major
Boilers Rule, 78 Fed. Reg. at 7,142. To qualify for certain
subcategories, the EPA required that a source burn a fuel
mixture comprised of only 10 per cent of the subcategory-
defining fuel. See, e.g., id. at 7,193 (“Unit designed to burn
solid fuel subcategory means any boiler . . . that burns . . . at
least 10 percent solid fuel . . . in combination with liquid fuels
or gaseous fuels.” (emphasis added)). Notwithstanding the
low bar for inclusion, we conclude, and discuss at greater
length below, see infra § IV.J, that the EPA reasonably
exercised its discretion when it subcategorized boilers this
way.
82
We cannot say the same about the EPA’s exclusion of
certain high-performing units from its MACT-floor
calculation. Although the EPA allowed sources that combust
only 10 per cent of a subcategory-defining fuel to join that
subcategory, it declined to consider emissions from any
source that burned less than 90 per cent of the subcategory-
defining fuel when determining the average emissions level of
the best performing sources in setting MACT floors for
existing sources. And when it set a subcategory’s MACT
floors for new sources, the Agency declined to consider the
emissions levels from any source that did not burn 100
per cent of the fuel. This disparate treatment makes a
difference; several sources excluded from the MACT-floor
determination were among the best performing sources (or, in
some cases, the single best performing source) in that fuel-
based subcategory.
The CAA, however, demands that source subcategories
take the bitter with the sweet. Section 7412 mandates,
without ambiguity, that the EPA set the MACT floor at the
level achieved by the best performing source, or the average
of the best performing sources, in a subcategory. See 42
U.S.C. § 7412(d)(3)(A), (B). It thus follows that if the EPA
includes a source in a subcategory, it must take into account
that source’s emissions levels in setting the MACT floor.
The Agency, however, claims discretion to omit from
MACT-floor computation sources it considers dissimilar. In
support, it cites section 7412(d)(3), which provides that
MACT standards must be no less stringent than “the best
controlled similar source, as determined by the [EPA].” Id.§ 7412(d)(3) (emphases added). Our decision in Sierra Club
II, 479 F.3d 875, however, forecloses this argument. In
Sierra Club II, the EPA set MACT standards for brick and
83
ceramic kilns. Id. at 879. For some subcategories, the EPA
based its MACT-floor determination on “the pollution control
devices used by the second-best performers,” not the best
performers. Id. (emphasis added). Although the EPA argued
that it “reasonably construe[d] the term ‘best performing’ . . .
to allow it to consider whether retrofitting kilns with a
particular pollution control technology is technically
feasible,” id. at 880 (alterations in original), we held that the
EPA could not circumvent the requirement that it base the
MACT floor “on the emission level actually achieved by the
best performers (those with the lowest emission levels).” Id.at 880-81 (citing Cement Kiln, 255 F.3d at 861) (emphasis in
original). We reach the same conclusion here.
The EPA tries to distinguish Sierra Club II, arguing that
the issue in that case “was whether [the] EPA could exclude
all units using the most-effective emission control technique
because it might not be applicable to all existing units”;
however, “[h]ere, [the] EPA is excluding a test result that is
unrepresentative of typical operations of units in the
subcategory, and thus is inappropriate to use in establishing
the MACT floor.” No. 11-1108 EPA Br. 81. But it makes no
difference whether the EPA exempts from consideration units
with certain highly effective technology or units with
impressive test results driven by the fuel combination it
combusts. Either approach contravenes our holding in Sierra
Club II that the EPA cannot ignore “the emission level
actually achieved by the best performers (those with the
lowest emission levels)” in the subcategory. 479 F.3d at 880(emphasis omitted). In any event, the EPA has not simply
excluded aberrant test results; it has excluded an entire class
of units—those burning less than 90 per cent of the
subcategory’s fuel—even though every one of those units fits
84
the subcategory’s parameters. This is no different from what
we rejected in Sierra Club II.
The EPA insists that if a source is “unrepresentative of
typical operations of units in the subcategory,” it is
“inappropriate to use [it] in establishing the MACT floor.”
No. 11-1108 EPA Br. 81. Not so. “The idea is to set limits
that, as an initial matter, require all sources in a category to at
least clean up their emissions to the level that their best
performing peers have shown can be achieved.” Sierra Club
I, 353 F.3d at 980 (citing 42 U.S.C. § 7412(d)(3)). For this
reason, an unusually high-performing source should be
considered; indeed its performance suggests that a more
stringent MACT standard is appropriate. Accordingly, we
vacate the MACT standards for all major boiler subcategories
that would have been affected had the EPA considered all
sources included in the subcategories.20
C. UPPER PREDICTION LIMIT
Sections 7412 and 7429 create MACT-floor criteria that,
for our purpose, are materially the same. Compare 42 U.S.C.
20
In its brief, the EPA argued that the Environmental
Petitioners’ challenge was moot either because the challenged
MACT standards had been remanded for other reasons or because
inclusion of the allegedly dissimilar sources would not have
affected the MACT standard. During oral argument, however, it
conceded that it misunderstood the scope of the Petitioners’
argument, which argument challenges unremanded MACT
standards that have in fact been affected by the EPA’s decision to
omit certain high-performing sources from its MACT-floor
analysis. See Oral Arg. Recording pt. B at 48:28-49:22. We
believe that the Environmental Petitioners’ challenge is not moot
and has not been waived.
85
§ 7412(d), with id. § 7429(a)(2). In both provisions, the CAA
mandates that MACT floors have maximum stringency but
also be continuously achievable. See id. § 7412(d)(2), (k); id.
§ 7429(a)(2); id. § 7602(k). Satisfying the statutory criteria is
no easy task, especially because no source emits any HAPs at
a constant level. See Page Mem. 6. Rather, emissions levels
fluctuate over time and for many reasons. See id. at 3.21 We
have held, see Mossville Envtl. Action Now v. EPA, 370 F.3d
1232, 1242 (D.C. Cir. 2004), and recently reaffirmed, see
NACWA, 734 F.3d at 1133-34, that the EPA can consider this
variability when setting MACT floors.
Further complicating the task is the way in which sources
typically measure emissions. Virtually all of the data the EPA
collects to set MACT floors come from the three-run stack
test. Page Mem. 6. The three-run stack test, as the name
suggests, involves three measurements of the source’s
emissions taken over a short time period (i.e., no more than a
few days) with each of the three test “runs” lasting from one
hour to four hours. Id. at 3. Because the tests provide three
“snapshots” of a source’s emissions performance, they cannot
accurately represent the source’s full range of emissions over
all times and under all conditions. Id. at 3-4. Because stack
testing typically involves “three separate runs,” however, it
“will in most cases show some of a particular source’s
21
See also Page Mem. 2-3 (“This variability occurs due to a
number of factors, including measurement variability (both
sampling and analysis) and short term fluctuations in the emission
levels that result from short-term changes in fuels, processes,
combustion conditions, and controls.”).
86
variability over the short period of time during which testing
was conducted.” Id. at 6 (emphasis added).22
1. NACWA, 734 F.3d 1115 (D.C. Cir. 2013)
Based on the limitations inherent in stack testing, the
EPA concluded that it could not set MACT floors based on
that testing alone. It began using the UPL to account for the
HAPs-emissions variety that stack-testing data do not reflect.
See NACWA, 734 F.3d at 1122. The Agency did so in several
rules promulgated in 2011, including not only the Major
Boilers Rule and the CISWI Rule but also the Sewage Sludge
Incinerator Rule addressed in NACWA. See id. In that case,
the petitioners challenged the EPA’s UPL use, arguing that
the Agency failed to establish that the UPL fairly represented
the “average emissions limitation achieved” by the best
performing sources to set the Sewage Sludge Incinerator
MACT floors and, accordingly, was “unlawful and arbitrary.”
Id. at 1130. We agreed in part. See id. at 1119.
Specifically, we struggled to pin down the EPA’s precise
interpretation of the phrase “average emissions limitation
achieved by the best performing 12 percent of units.” Id. at
1142-43 (quoting 42 U.S.C. § 7429(a)(2)).23 As best we
could tell, the EPA defended its use of the UPL as follows:
“[b]ecause the [UPL] represents the value which [the EPA]
22
See also Page Mem. 5 (“[E]ven single three run tests, which
are performed over a short period of time, typically show different
emissions levels during each individual test run.”).
23
See also NACWA, 734 F.3d at 1142 (“[I]t seems EPA has
adopted yet another interpretation of the phrase ‘average emissions
limitation achieved by the best performing 12 percent of units.’”
(emphasis added)).
87
can expect the mean (i.e., average) of three future
observations (3-run average) to fall below, based upon the
results of the independent sample size from the same
population, the [UPL] reflects average emissions.” Id. at
1142 (quoting Standards of Performance for New Stationary
Sources and Emission Guidelines for Existing Sources:
Sewage Sludge Incineration Units, 76 Fed. Reg. 15,372,
15,389 (Mar. 21, 2011)) (emphasis added) (some alteration in
original). In our view, however, “the word ‘average’ . . .
seems to mean the average emissions limitation that the
existing population of the best-performing 12 percent of
incinerators has achieved.” Id. (emphases added).
Despite these doubts, we reasoned that the EPA could
have “plausibl[y]” concluded that the UPL represents the
“average emissions limitation achieved” by the best
performing sources. Id. at 1143. That said, we were not
willing to assume the EPA’s responsibility of “supply[ing] a
reasoned basis” for its UPL use. Id. (quoting Bowman
Transp., Inc. v. Ark.-Best Freight Sys., Inc., 419 U.S. 281(1974)). For that reason, we remanded—but did not vacate,
see id. at 1161—the UPL portion of the Sewage Sludge
Incinerator Rule and ordered the EPA to “clarify how the
[UPL] represents the average emissions limitation achieved
by the best performing 12 percent.” Id. at 1143 (internal
quotation marks omitted).24
24
See also NACWA, 734 F.3d at 1151 (“[W]hile we determine
that [the] EPA’s use of the [UPL] may be lawful, we are remanding
this portion of its rulemaking for further explanation on the issue[]
of how the upper prediction limit represents the average emissions
limitation achieved . . . .” (internal quotation marks omitted)).
88
Because the EPA also used the UPL in the Major Boilers
Rule and the CISWI Rule, the Agency moved for a limited
remand of the current petitions so that it could include its
revised UPL explanation in the administrative records of these
two regulations.25 See Page Mem. 2. On July 14, 2014, the
EPA published a fifteen-page memorandum authored by
Stephen D. Page, the EPA Director of Air Quality Planning
and Standards (Page Memorandum), in response to NACWA.
See id. at 1. The EPA’s current explication of the UPL is now
before us.26
25
In NACWA, we had other problems with the EPA’s use of
the UPL. Specifically, the EPA had explained that “a smaller
dataset may have greater variability, and thus a higher [UPL].”
NACWA, 734 F.3d at 1144. We instructed the EPA not only to
explain its use of the UPL in general but also to “explain why the
[UPL] could still be considered accurate given a small dataset” in
particular. Id. at 1144-45 (emphasis added). In its remand motion,
the EPA represented that it could “adequately explain why [its] use
of the UPL in general is consistent with Clean Air Act requirements
through a remand of the record for a limited time” but that “the
question of whether the UPL is an appropriate statistical method for
small data sets requires more analysis . . . [along with] additional
notice and comment rulemaking.” No. 11-1108 Mot. for Remand
9, 13 (Feb. 28, 2014). We agreed and, for this reason, the only
issue we decide today is whether the EPA carried its burden of
establishing, as a general matter, that the UPL reasonably estimates
the average emissions level achieved by the best performing source
or sources to set MACT floors.
26
The Environmental Petitioners urge us to ignore the Page
Memorandum, insisting that it “provide[s] a series of new
interpretations and assertions that, rather than ‘explaining’ the prior
record, instead contradict and revise the agency’s earlier position,”
in contravention of NACWA and the scope of the remand the
Agency requested regarding the Major Boilers Rule and the CISWI
89
2. The Page Memorandum
The Page Memorandum recognized our “concern about
the interpretation [we] believed [the] EPA was taking” of the
word “average.” Page Mem. 3. It clarified that the Agency
“does not interpret the term ‘average’” to mean “the average
of a future 3-run compliance test.” Id. (emphasis added)
(quoting NACWA, 734 F.3d at 1143). Rather, it explained that
the “EPA interprets the average to mean the average
emissions over time,” based not only on the “average of all
emissions test data from the best performing source or
sources” but also on “information regarding the variability of
emissions.” Id. (emphasis added).
In the EPA’s judgment, “variability is a key factor in
establishing” MACT standards because “[e]ach MACT
standard is based on limited data from sources whose
emissions are expected to vary over their long term
Rule. No. 11-1108 Envtl. Pet’rs’ Br. 41. But our NACWA decision
did not, as the Petitioners would have it, require the EPA to adopt
our belief that the Agency construed “average” to mean “the
average of a future 3-run compliance test.” See NACWA, 734 F.3d
at 1143. Rather, we asked the EPA to clarify how, in its view, the
UPL “represents the ‘average emissions limitation achieved by the
best performing 12 percent.’” Id. (emphasis added). Nor do we
think that the EPA altered its initial basis for using the UPL, which
the EPA has consistently held out as “a statistical formula designed
to estimate a MACT floor level that is equivalent to the average of
the best performing sources based on future compliance tests.”
2011 Major Boilers Rule, 76 Fed. Reg. at 15,630 (emphasis added).
What the EPA failed to do before NACWA was to explain how the
UPL functions and why it is a reasonable way to calculate
“average” emissions levels. The Page Memorandum does precisely
that.
90
performance.” Id. Specifically, “[t]he available emissions
data are generally in the form of short term, three-run stack
tests, with each test run lasting for between 1 and 4 hours.”
Id. For this reason, the EPA concluded that it did not have
information “encompass[ing] the emissions performance of a
source over time.” Id. (emphasis added). And because the
“EPA interprets ‘emissions performance’ . . . to mean the
emissions of a source over the long term, rather than just
during a short-term stack test,” the EPA found it necessary to
“appl[y] a methodology that predicts the actual emissions
levels the source is achieving at times other than when stack
testing was conducted.” Id. at 3-4 (emphases added).
The UPL is the methodology the EPA selected to account
for these limitations. Id. at 4. “[A] value derived from widely
accepted and commonly used statistical principles,” the UPL
“represents the upper end of a prediction interval.” Id. In
layman’s terms, the UPL uses an equation that considers
(1) the average of the best performing source or sources’
stack-test results (i.e., the mean); (2) the pattern the stack-test
results create (i.e., the distribution); (3) the variability in the
best performing source or sources’ stack-test results (i.e., the
variance); and (4) the total number of stack tests conducted
for the best performing source or sources (i.e., the sample
size). Id. at 4-5.
The UPL, however, cannot demonstrate with absolute
certainty the average emissions levels achieved by the best
performing sources at all times (indeed, certainty is
impossible without continuous monitoring). See id. Instead,
the UPL equation produces a range of values that is expected,
given the variance in the relevant stack-test data, to
encompass the average emissions levels achieved by the best
performing sources a specified percentage of the time. Id. at
91
4. To establish the MACT floor, the EPA calibrated the UPL
equation to produce a range in which the average emissions
levels of the best performing source or sources would be
expected to fall 99 per cent of the time, which is referred to as
a 99 per cent confidence interval. Id. Once the EPA had this
range, it set the MACT floor at the top level of that range—
hence, the “upper” in “upper prediction limit”—to arrive at a
figure that, 99 out of 100 times, it expected the average
emissions levels of the best performing sources to “achieve.”
Id. Or, in the EPA’s words, “the 99 percent UPL is the level
of emissions that” the EPA is “99 percent confident is
achieved by the average source represented in a dataset over a
long-term period based on its previous, measured
performance history as reflected in short term stack-test data.”
Id.
One of the equations the EPA used to calculate the UPL
is as follows:27
27
The EPA used “one of several equations” to calculate the
UPL depending on “certain characteristics of [the] dataset,”
including the distribution of data within the dataset. Page Mem. 4.
Here, we set out the equation the EPA used for a dataset with a
“normal distribution.” Id. at 10. For our review, we need not
recount the other, somewhat more complicated equations the EPA
used in determining the UPL for datasets with, e.g., a “lognormal
distribution.” See id. (“Even though they differ due to separate
mathematical properties associated with each distribution, the UPL
equations share a common format . . . .”); see generally id. at 11
(describing lognormal distribution equation).
92
NACWA, 734 F.3d at 1139. In this equation:
 “x̄ ” is the mean;
 “t(0.99, n-1)” is a value called the “t-statistic,” the
statistical tool used to set the confidence interval
(here, 99 per cent);
 “n” is the sample size;
 “m” is the number of stack tests that were run to
calculate the mean (“x̄ ”); because most stack tests
involve 3 “runs,” m usually equals 3;
 “s” represents the “standard deviation.”
See id.; see also Page Mem. 10-11.
3. Instant Challenges to UPL
After the EPA issued the Page Memorandum, the
Environmental Petitioners renewed their argument that the
UPL represents neither (1) the “average” emissions limit of
the best performing source or sources in a subcategory, nor
(2) the emissions levels “achieved” by the best performing
sources in a subcategory. We believe that the EPA has
carried its burden of demonstrating that the UPL “reflect[s] a
reasonable estimate of the emissions achieved in practice by
93
the best performing sources.” Cement Kiln, 255 F.3d at 871-
72 (internal quotation marks omitted); see also NACWA, 734
F.3d at 1148 (“[H]aving decided to account for variability,
and having decided to estimate that variability, EPA bears the
burden of demonstrating with substantial evidence that its
estimate is reasonable.”).
Our conclusion is driven, in large part, by the deference
we owe the EPA when it determines how best to meet the
technical challenges in its area of expertise. Indeed, the EPA
“typically has wide latitude in determining the extent of data-
gathering necessary to solve a problem” and, for that reason,
we have “accorded Chevron deference to [its] interpretation
of [the CAA] as allowing it to estimate MACT floors.”
NACWA, 734 F.3d at 1131. Moreover, “the requirement that
the existing unit floors not be less stringent than the average
emissions limitation achieved by the best performing 12
percent of units does not, on its own, dictate how the
performance of the best units is to be calculated,” id. (internal
quotation marks omitted)—“[f]loors need not be perfect
mirrors of the best-performers’ emissions,” Cement Kiln, 255
F.3d at 871. So long as the EPA “demonstrate[s] with
substantial evidence—not mere assertions”—that the UPL
“allows a reasonable inference as to the performance of the
top 12 percent of units,” NACWA, 734 F.3d at 1131
(quotations omitted) (emphasis added), the EPA has
conducted reasoned decision making.
The Agency has done so here. The Page Memorandum
explains the limitations of stack-test data—i.e., the
“snapshots” cannot reflect the best performing source’s or
sources’ average emissions levels at all times and under all
operating conditions. Page Mem. 6. The Page Memorandum
also explains that the Agency chose the UPL as a tool
94
“derived from widely accepted and commonly used statistical
principles,” id. at 4, that “reasonably account[s] for variability
in the emissions of . . . sources,” id. at 2. Finally, the Page
Memorandum plugs the analytical gap we identified in
NACWA—it thoroughly explains how and why the UPL
accounts for the variance and therefore how and why it
reasonably represents the emissions level “achieved by the
average source” or sources. Id. at 3-5. In so doing, the EPA
has “clarif[ied],” to our satisfaction, “how the upper
prediction limit represents the average emissions limitation
achieved.” NACWA, 734 F.3d at 1143 (internal quotation
marks omitted).
The Environmental Petitioners’ arguments to the contrary
are unavailing. Their primary objection is that the UPL
cannot reasonably estimate the “average” emissions level
achieved by the best performing source or sources because the
UPL represents “a level [the] EPA expects any future
compliance test by any [source] in the top 12 percent to fall
below.” No. 11-1108 Envtl. Pet’rs’ Br. 35 (emphases added)
(internal quotation marks omitted).28 But the Page
Memorandum counters the Environmental Petitioners’
mistaken understanding of what the UPL represents.29
According to the EPA, “the UPL does not represent the worst
emissions performance of the best performing units at any
28
See also No. 11-1108 Envtl. Pet’rs’ Reply Br. 15 (“It is, as
its name indicates, an upper limit—the emissions limitation that
every member of the best-performing 12 percent will fall
below . . . .” (emphasis in original) (quotation marks omitted)).
29
The Environmental Petitioners’ argument rests, at least in
part, on their contention that we should not consider the Page
Memorandum at all. We decline their invitation to ignore the
explanation we ordered the EPA to provide.
95
time.” Page Mem. 4 (emphasis in original).30 It is instead
“the average level expected to have been achieved over time”
by the best performing source or sources. Id. (emphasis in
original). “In other words, the 99 percent UPL is the level of
emissions that [the EPA is] 99 percent confident is achieved
by the average source . . . over a long-term period based on
its previous, measured performance history as reflected in
short term stack test data.” Id. (emphasis added).
Next, the Environmental Petitioners criticize the Page
Memorandum’s explanation that the UPL represents the long-
term average emissions levels achieved because “the first
element of the UPL equation is the average of the short-term
emissions test data from the best-performing sources.” Id. In
their view, the UPL is no different from “saying that, over
time, the average of 1, 2, and 3 = 2 + 500 because the first
element in the equation (2) is the average of 1, 2, and 3.” No.
11-1108 Envtl. Pet’rs’ Br. 48. But the UPL does not simply
tack an arbitrary increase on top of the stack-test average of
the best performing sources. Rather, the UPL “allows [the]
EPA to use emissions test data and the data characteristics,”
which include “the distribution and sample size, along with
the intrinsic variability associated with those data,” to
estimate “an emissions limit based on a specified level of
confidence such that an average best performing existing
30
See also Page Mem. 5 (It is “generally . . . reasonable to
establish a [MACT floor] standard that all the best performing 12
percent of existing sources can meet without any modification
because the statute requires the Agency to establish the standard at
the average level of performance of the best 12 percent of sources.”
(emphasis in original)); id. at 14 (“[T]he MACT floor represents the
average emission level achieved by the best performing sources,
not the worst emission level achieved by those sources.” (emphases
in original)).
96
source would not be expected to exceed the limit a specified
number of times.” Page Mem. 6 (emphases added). In other
words, the UPL does not simply add an arbitrarily chosen
value but instead turns entirely on the features inherent in the
stack-test data and how those features reflect the natural
variance in emissions experienced by the best performing
sources over time. See id. at 4 (“[T]he MACT floor
calculation takes into account the inherent variability in
emissions performance to more accurately reflect the range of
the best performing sources’ emissions over time.” (emphasis
added)).31 Thus, as the Page Memorandum amply
demonstrates, see id., the EPA’s use of the UPL is not
arbitrary.
The Environmental Petitioners also attack the results
produced by the UPL. They provide a series of charts that, in
their view, demonstrate that the UPL sets MACT floors far
too high to comport with the CAA’s mandate that floors
represent “the maximum degree of reduction in emissions.”
See 42 U.S.C. § 7412(d)(2). True, some of the charts show
that the EPA has set a MACT floor above the highest
emissions level recorded by the best performing sources’
stack testing. See No. 11-1108 Envtl. Pet’rs’ Br. 14-15; No.
11-1108 Envtl. Pet’rs’ Reply Br. 23. But this does not mean
that the UPL is an arbitrary “average” proxy—for at least two
reasons.
31
See also Page Mem. 6-7 (“[T]he UPL equation that is used
to account for variability and [to] calculate the MACT floor
standard depends on the distribution of the data.”); id. at 11 (“The
UPL . . . is directly related to the confidence level and to the
variance, meaning that as either of these values go up or down, so
does the UPL value.”).
97
First, the charts selectively included are generated from
data sets with considerable variance between the highest
recorded stack test and the lowest. Unsurprisingly, if a
handful of “snapshots” in a data set demonstrate that
emissions levels experience high spikes and low plummets at
discrete times, it is more likely that the average emissions
level achieved by the best performing sources at all times
might be high. This is because a data set with high variability
will produce a higher UPL than a data set with low variability,
even if the two sets share the same average. In other words,
the UPL takes large variance into account and therefore
naturally goes higher to arrive at the 99 per cent certainty the
EPA thinks is appropriate.32 Second, where the UPL
suggested a MACT floor higher than the results of the stack
tests, it often did so by insubstantial amounts. Indeed, for at
least one chart, “the limit is a mere 4 millionths of a pound per
million Btu above the emissions test results of best
performers, an unalarming amount given that the
methodology is supposed to account for variable results.”
No. 11-1108 Indus. Intervenors’ Br. 10 (emphases in
original). For these reasons, the Environmental Petitioners
have not convinced us that the EPA failed to satisfy the
32
The EPA “selected the 99 percent level in order to provide
reasonable assurance that the limit can be met at all times by a
source with emissions at the average level achieved by the best
performing source or sources.” Page Mem. 10. The Environmental
Petitioners have not challenged the EPA’s choice of a 99 per cent
confidence level, as opposed to a lower level of certainty, and we
express no opinion on that choice. And we reiterate that the more
specific concerns we had with the UPL when we decided
NACWA—in particular, the UPL’s accuracy “given a small
dataset”—are not before us. 734 F.3d at 1144-45.
98
“minimal standard[] of rationality” that we require. Ethyl
Corp. v. EPA, 541 F.2d 1, 36 (D.C. Cir. 1976) (en banc).
Finally, the Environmental Petitioners insist that “[t]he
UPL predicts a level that hypothetical future tests will fall
below, rather than estimating what boilers actually achieved,”
in contravention of the requirement that MACT floors “reflect
what the best-performing sources achieved.” No. 11-1108
Envtl. Pet’rs’ Reply Br. 24 (internal quotation marks omitted).
But the Environmental Petitioners ignore the Page
Memorandum’s explanation that, because the UPL is not
time-dependent, it “not only is a prediction of the emissions
performance of those sources in tests conducted in the future,
but is also an indication of the range of current average
emissions performance of those units.” Page Mem. 3;33 see
also No. 11-1108 Indus. Intervenors’ Br. 9 (“Because this
statistical method is not time-dependent, it is equally valid for
predicting past performance (i.e., the range of emissions
levels expected to have been experienced in the past by the
best performers during periods when actual emissions testing
was not underway) and future performance.”).
We believe that the UPL “reflect[s] a reasonable estimate
of the emissions achieved in practice by the best-performing
sources,” Cement Kiln, 255 F.3d at 871-72 (internal quotation
marks omitted), and, accordingly, we reject the
Environmental Petitioners’ challenge to it.
33
See also Page Mem. 4 (“[T]he 99 percent UPL is the
emissions level that the source would be predicted to be below 99
out of 100 performance tests, including emissions tests conducted
in the past, present, and future.”); id. at 10 (“The confidence level,
in this case 99 percent, is the percentage of measurements (past,
present, and future) that are predicted to fall at or below the UPL
value.”).
99
D. BEYOND-THE-FLOOR STANDARDS FOR CISWI UNITS
The EPA declined to set beyond-the-floor standards for
CISWI units. The Environmental Petitioners challenge that
determination in three primary respects, each of which we
reject.34
Section 7429 of the CAA directs the EPA to set MACT
standards in two steps. It first sets a floor level based on the
best performing sources. See 42 U.S.C. § 7429(a)(2). Next, it
determines “whether a more restrictive standard is
‘achievable,’” NRDC III, 749 F.3d at 1057, “taking into
consideration the cost of achieving such emission reduction,
and any non-air quality health and environmental impacts and
34
Although the EPA does not argue that the court lacks
jurisdiction to consider this argument, Environmental Petitioners
raise the issue defensively, contending that they satisfied the CAA’s
administrative exhaustion provision. We agree. During the
rulemaking process, the Petitioners comprehensively critiqued the
EPA’s proposed rationale for rejecting beyond-the-floor standards.
See, e.g., Natural Resources Defense Council, Comments on
Standards of Performance for New Stationary Sources and
Emission Guidelines for Existing Sources, EPA-HQ-OAR-2003-
0119 (Aug. 23, 2010), at 11-16 (No. 11-1125 J.A. 668-73). Many
of those comments challenged the EPA’s consideration of costs and
other factors—the same types of issues Petitioners now ask the
Panel to resolve. Because the Environmental Petitioners raised the
relevant issues “with reasonable specificity” during the period for
public comment, our jurisdiction is not in question. 42 U.S.C.
§ 7607(d)(7)(B); see Portland Cement Ass’n v. EPA, 665 F.3d 177,
186 (D.C. Cir. 2011) (“While we certainly require some degree of
foresight on the part of commenters, we do not require telepathy.
We should be especially reluctant to require advocates for affected
industries and groups to anticipate every contingency.”).
100
energy requirements,”35 42 U.S.C. § 7429(a)(2). The “EPA
calls these stricter requirements ‘beyond-the-floor’
standards.” NRDC III, 749 F.3d at 1057.
In section 7429, the “Congress gave EPA broad
discretion in considering whether to go beyond-the-floor.”
NACWA, 734 F.3d at 1157. The Congress required the EPA
to consider a variety of factors without telling the EPA how to
weigh them. That calculus belongs to the EPA’s discretion.
See 42 U.S.C. § 7429(a)(2) (delegating to the EPA
Administrator the responsibility to “tak[e] into consideration”
the statutory factors). Against that backdrop, challenges to
the EPA’s beyond-the-floor determinations “must clear a high
bar, as we are at our most deferential when an agency is
‘making predictions, within its area of special expertise, at the
frontiers of science.’” NACWA, 734 F.3d at 1156 (quoting
Husqvarna, 254 F.3d at 199).
When establishing MACT standards for CISWI, the EPA
declined to establish beyond-the-floor standards in the
proposed rule, see 2010 Proposed CISWI Rule, 75 Fed. Reg.
at 31,956-59, and the final rule, see 2011 CISWI Rule, 76
Fed. Reg. at 15,729-32. The EPA also declined requests to
reconsider that decision. See Memorandum from Eastern
Research Group, Inc., to Amy Hambrick, U.S. Environmental
Protection Agency, Revised Draft CISWI Reconsideration
Issues (Dec. 20, 2012), at 22-23 (No. 11-1125 J.A. 1219-20).
35
EPA interprets the statutory factor of “cost” to permit
consideration of cost-effectiveness, NRDC III, 749 F.3d at 1060-61,
which is often calculated “on [a] per ton of emissions removed
basis,” Husqvarna AB v. EPA, 254 F.3d 195, 200 (D.C. Cir. 2001).
We have previously upheld that interpretation. See, e.g., NRDC III,
749 F.3d at 1060-61.
101
The first challenge targets several instances in which the
EPA refused to require sources to adopt, as a beyond-the-floor
measure, controls that most sources would employ to meet the
MACT floor standard. In each instance, the EPA determined
that the relative costs outweighed the expected emissions
gains. In the first such case, the EPA decided not to require
liquid-fired energy recovery units to install dry sorbent
injection and fabric filters as a beyond-the-floor measure,
despite the fact that “four of the six” units would need to
install those systems to meet the floor standard. 2011 CISWI
Rule, 76 Fed. Reg. at 15,731. That decision satisfied the
statute. Had the EPA mandated the control measures, the
remaining two units would have needed to expend “$1.1
million per year” to achieve only a small emissions reduction,
“which translates into an incremental cost-effectiveness of
about $230,000 per ton” of emission. Id. Nothing in section
7429(a)(2) requires the Agency to impose a cost so
disproportionate to the expected emissions gains.
The Environmental Petitioners take issue with two other
decisions along these lines. In the first, the EPA declined to
set beyond-the-floor mercury control measures for kilns,
citing a cost-effectiveness of roughly $351 million per ton.
See Memorandum from Eastern Research Group, Inc., to Toni
Jones, U.S. Environmental Protection Agency, Final
Reconsideration Beyond-the-Floor Analysis for CISWI Units
(Reconsideration Mem.) (Dec. 20, 2012), at ¶ 3.4.2 (No. 11-
1125 J.A. 1232). In the second, a $26,000 per-ton
implementation cost led the Agency not to establish stricter
carbon monoxide control measures for calciner kilns. See id.
¶ 3.4.3. Energy use—a factor mandated in section
7429(a)(2)—also entered the equation. With respect to
calciner kilns, the technology used to reduce carbon
102
monoxide would also increase energy requirements, and
therefore increase energy costs. See id. In each of these
decisions, the EPA reasonably applied the statutory factors.
That Petitioners would have weighed the costs differently
provides no grounds to displace the EPA’s otherwise
reasonable determination.
In the second challenge to the decision not to set beyond-
the-floor standards, the Environmental Petitioners contend the
Agency arbitrarily failed to set emission levels lower than the
MACT floor for categories likely to adopt technology capable
of meeting those lower levels. Specifically, according to the
Environmental Petitioners, the EPA knew waste-burning kilns
and energy recovery units would adopt fabric filters that
“achieve particulate matter emissions levels dramatically
lower than the floor, but refused to set the standard at that
lower level.” See No. 11-1125 Envtl. Pet’rs’ Br. 50.
That is incorrect. The Environmental Petitioners spin this
yarn based on a line in the proposed rulemaking. There, the
Agency speculated that kilns and energy recovery units would
adopt fabric filters to comply with the MACT floor limit, and
would “likely achieve a level of performance” below the
floor. 2010 Proposed CISWI Rule, 75 Fed. Reg. at 31,958.
That statement represented a preliminary prediction, which
was subject to change during the notice-and-comment
process. And change it did. In the final rulemaking, the EPA
further subcategorized the energy recovery unit subcategory
and revised the MACT floor for waste-burning kilns. See
2013 CISWI Rule, 78 Fed. Reg. at 9,122 (explaining the
changes). New information received during the rulemaking
inspired those changes, which the EPA made after
considering the statutory factors. See id.; Reconsideration
Mem., ¶ 2.3-3.4.5. The evidence does not suggest that the
103
EPA refused to set beyond-the-floor emission levels it knew
were reasonably achievable.36
In the third challenge, the Environmental Petitioners take
issue with three determinations that rested on factors other
than cost. First, the EPA declined to require coal-fired energy
recovery units to adopt linkageless boiler management
systems as a beyond-the-floor measure for carbon monoxide.
See Reconsideration Mem., ¶ 2.3.1.1. While acknowledging
that linkageless systems were available at “fairly low-cost,”
the EPA concluded it had insufficient data to determine the
“actual reductions this control option would achieve” relative
to an alternative control system. Id.
The EPA acted reasonably. The record suggests the EPA
had scant evidence on the efficacy of linkageless control
measures applied to coal-fired energy recovery units. See id.
Had the Agency imposed a stricter standard based on controls
for which it had precious little (if any) evidence, a reviewing
court may well have concluded the decision lacked “a rational
connection between the facts found and the choice made.”
State Farm, 463 U.S. at 43 (internal quotation marks omitted).
Second, the EPA rejected regenerative thermal oxidizers
as a beyond-the-floor control for carbon monoxide in solid
waste energy recovery units. See 2011 CISWI Rule, 76 Fed.
Reg. at 15,732. Thermal oxidizers could do the job “but
likely at a far greater energy requirement (specifically natural
36
This argument suffers from an additional flaw: the
Environmental Petitioners appear to treat as interchangeable
proposed emissions rules for new units with the final rules
applicable to existing ones. That apples-and-oranges comparison
underscores the weakness of the argument.
104
gas) [and] with comparable control efficiency” as carbon
monoxide catalysts, which “some units will need to install to
meet the MACT floor . . . limits.” Id. In other words, even
though oxidizers work as well as carbon catalysts, oxidizers
would be unsuitable because they use more energy. See id.
(concluding that beyond-the-floor controls “would be
unreasonable for this subcategory due to additional cost and
energy impacts”).
The Environmental Petitioners contend that the EPA
failed to “suggest that these natural gas requirements are high
in an absolute sense or relevant to achievability.” No. 11-
1125 Envtl. Pet’rs’ Br. 51. We agree that the EPA’s analysis
is less than fully satisfying. Among other reasons, nowhere
did the Agency estimate the per-ton cost of mandating
thermal oxidizers, or compare the energy costs relative to
other control measures.
Despite these imperfections, we reject the challenge. See
Dist. Hosp. Partners, L.P. v. Burwell, 786 F.3d 46, 61 (D.C.
Cir. 2015) (“[I]mperfection alone does not amount to arbitrary
decision-making.”). The EPA’s somewhat sparse analysis on
this issue reflects a somewhat sparse record. At bottom, the
Agency rejected thermal oxidizers because it lacked sufficient
evidence to support their utility, at least compared with
control measures whose efficacy and costs were better known.
The Agency’s determination should be read in context.
Elsewhere in the final rule, the EPA expanded on the energy
and environmental impacts of thermal oxidizers, concluding
that “[t]he combustion of fuel needed to generate additional
electricity and to operate [thermal oxidizer] controls would
yield slight increases in emissions, including NOX, CO, PM,
and SO2 and an increase in CO2 emissions.” 2011 CISWI
105
Rule, 76 Fed. Reg. at 15,744. The EPA addressed another
statutory factor—cost—by reasonable implication. Energy—
natural gas, in this case—is not free. A technology that
demands “far greater energy requirement[s]” naturally comes
at a cost. See id. at 15,732.
Though courts are powerless to “supply a reasoned basis
for the agency’s action that the agency itself has not given,”
“[w]e will . . . uphold a decision of less than ideal clarity if
the agency’s path may reasonably be discerned.” State Farm,
463 U.S. at 43 (internal quotation marks omitted). Here, the
Agency’s path may reasonably be discerned: mandating
thermal oxidizers was not achievable due to increased energy
demands and a corollary increase in cost, see 2011 CISWI
Rule, 76 Fed. Reg. at 15,732 (declining to set a beyond-the-
floor limit “due to additional cost and energy impacts”). For
those reasons, EPA did not act unreasonably.
Third, and finally, the Environmental Petitioners
challenge the rejection of dry sorbent injection and wet
scrubbers as beyond-the-floor measures for waste-burning
kilns. The EPA determined those measures would be cost-
effective (at only $5,000 per ton) but declined to require them
due to “uncertainty” surrounding “the appropriate control
system that some existing kilns would need to employ to
meet” a stricter standard, “especially kilns that use ingredients
with a high sulfur content.” See Reconsideration Mem.,
¶ 3.4.5. Adding to that uncertainty, the EPA could not
“account for potential costs at existing sources for additional
scrubber water and spent sorbent.” Id. As before, the EPA
reached a reasonable conclusion in the face of imperfect
information. Had the EPA set a beyond-the-floor standard
based on sorbent injection and wet scrubbers, the Agency
106
would have been flying blind. In avoiding that course, we
conclude the EPA acted reasonably.
E. REGULATION OF CERTAIN CISWI UNITS
The final CISWI Rule did not contain emission standards
for burn-off ovens, cyclonic burn barrels, foundry sand
reclamation units, soil treatment units, and space heaters. The
Environmental Petitioners claim that the EPA unlawfully
exempted these units from regulation by creating
subcategories that capture only a subset of the units that the
Agency is required to regulate as CISWI. The EPA, however,
protests that it did not exempt these five types of units from
regulation. Rather, the Agency determined that it lacked
sufficient data to regulate the units at this time, and, with
respect to some, it received comments suggesting the units
were not CISWI.37
37
The EPA asserts that it has not made a final decision with
regard to the regulation of the five units at issue here—a claim that
calls into question our jurisdiction, which under the CAA is limited
to “final” actions. See Portland Cement, 665 F.3d at 193 (citing 42
U.S.C. § 7607(b)). We disagree with the Agency. Because the
statutory deadline for the EPA to establish emission standards for
all CISWI has passed, see 42 U.S.C. § 7429(a)(1)(D), “the
promulgated regulations must be deemed the [A]gency’s complete
response in compliance with the statutory requirement[].” Hercules
Inc. v. EPA, 938 F.2d 276, 282 (D.C. Cir. 1991) (internal quotation
marks omitted). Accordingly, “even if [the Agency] promulgates
additional . . . rules sometime in the future, petitioners’ claim that
the existing final regulations are unlawful remains reviewable by
this court.” Id. (internal quotation marks omitted). Moreover, the
EPA did not signal in the administrative record that it was
“continu[ing] the rulemaking process” as to these five units.
Portland Cement, 665 F.3d at 194 (holding that the EPA’s action
107
We agree with the Environmental Petitioners that the
Agency has violated its nondiscretionary statutory duty (1) to
promulgate standards with respect to cyclonic burn barrels,
and (2) to determine whether the remaining four types of units
fall within the statutory definition of CISWI. The CAA
requires the EPA to “establish performance standards . . . for
each category of solid waste incineration units” no later than
November 15, 1994. See 42 U.S.C. § 7429(a)(1)(A), (D).
The statute then defines “solid waste incineration unit” as a
“distinct operating unit of any facility which combusts any
solid waste material from commercial or industrial
establishments or the general public.” Id. § 7429(g)(1)
(emphasis added). That provision unambiguously requires the
EPA to set emission standards for “any facility that combusts
any commercial or industrial solid waste material at all,”
subject only to the listed statutory exceptions. NRDC I, 489
F.3d at 1257-58. Because the statutory deadline to regulate
these units has long passed, the EPA has “breached a non-
discretionary duty” if it has failed to promulgate standards for
any facilities combusting solid waste from commercial or
industrial establishments that do not fit into the listed
exceptions. Sierra Club v. EPA, 992 F.2d 337, 346 (D.C. Cir.
1993); cf. id. (explaining that the “plain language” of a similar
provision in RCRA “obligates the Agency to issue, by the
deadline, revisions for all facilities” covered by the statute
and therefore “does not contemplate partial compliance”).
The Agency makes no effort to claim that cyclonic burn
barrels fall outside the statutory definition for CISWI units.
was not “final” under 42 U.S.C. § 7607(b) because the Agency
expressly stated in its final rule that the rulemaking process
remained underway). We therefore need not consider whether our
conclusion regarding finality would change had it done so.
108
Nor could it—both the administrative record and the EPA’s
brief make clear that cyclonic burn barrels “combust” solid
waste. See 2011 Proposed CISWI Rule on Reconsideration,
76 Fed. Reg. at 80,460 (describing a cyclonic burn barrel as
“a combustion device for waste materials”); No. 11-1125
EPA Br. 68 (same); see also 42 U.S.C. § 7429(g)(1) (defining
“solid waste incineration unit” as a “distinct operating unit of
any facility which combusts any solid waste material from
commercial or industrial establishments or the general
public”). Because they combust solid waste, cyclonic burn
barrels clearly fall within the statutory definition of “solid
waste incineration unit” and, as established above, the EPA
had a nondiscretionary statutory duty to establish emission
standards for all these units by 1994. We therefore conclude
that the Agency violated that duty by failing to promulgate
emission standards for cyclonic burn barrels.
The EPA protests that it reasonably chose not to regulate
cyclonic burn barrels at this time, given how little information
it had on them. According to the EPA, comments revealed
there were many more cyclonic burn barrels in use than
originally thought, the Agency lacked data on these units, and
it was “difficult, if not impossible, to test such units for the
section 7429 pollutants.” No. 11-1125 EPA Br. 69. But this
argument misses the point: in light of the unambiguous
statutory command to promulgate numeric standards for all
solid waste incineration units, the EPA had no discretion to
avoid regulating any such units—even if its choice to avoid
regulating these units would have been otherwise reasonable.
The Agency was obligated to collect the data it needed, and
Congress gave it the authority to do so. See 42 U.S.C.
§ 7414(a) (explaining that for the purpose of regulating solid
waste combustion under section 7429, the EPA may, for
example, require owners and operators of those units to
109
sample emissions, keep records, and offer other information
that the Agency needs). Moreover, the Agency provides no
evidence that it would be infeasible to set emission standards
for these units. Instead, the EPA merely states that it
“received information” that measuring emissions is difficult,
“if not impossible,” but points to no comments or evidence
supporting this assertion. 2011 Proposed CISWI Rule on
Reconsideration, 76 Fed. Reg. at 80,460.
The EPA also had a duty to determine whether the other
challenged sources—burn-off ovens (including foundry sand
reclamation units), soil treatment units, and space heaters—
were units that “combust” solid waste. Several commenters
told the Agency that these units fell within the statutory
definition of CISWI, and the EPA itself initially viewed some
of these units as combusting waste. See, e.g., CISWI Rule—
Responses to Comments, at 74-76; 2010 Proposed CISWI
Rule, 75 Fed. Reg. at 31,941. Under these circumstances, the
Agency was obligated to determine whether the units in fact
combust solid waste. Yet the EPA concedes it never made
that determination. As we have explained, the EPA had a
nondiscretionary duty to promulgate standards for all solid
waste combustion units. This obligation includes the
subsidiary duty to determine whether the units identified by
the commenters in fact combust solid waste. Any other
conclusion would allow the Agency to ignore its statutory
mandate altogether by not taking the initial step of identifying
such units.
The CAA unambiguously requires that the Agency
establish standards for all CISWI units. As a result, we grant
the Environmental Petitioners’ petition for review on this
issue and remand to the Agency to set emission standards for
cyclonic burn barrels. The EPA must also determine whether
110
the remaining four types of units are CISWI units and, if it
finds that they are, it must set standards for them as well.
F. DELISTING UNDER 42 U.S.C. § 7412(C)
In contrast to major source subcategories (all of which
the EPA must control), the CAA does not require the EPA to
control emissions in every area source subcategory. See 42
U.S.C. § 7412(c)(1), (3). The Act does, however, mandate
that the EPA control area source emissions if the area source
subcategory meets certain criteria. Section 7412(c)(1), for
instance, requires the EPA to control any area source
subcategory upon the Agency’s finding that emissions from
the sources in the subcategory jeopardize either the
environment or human health. See id. § 7412(c)(3). If so, the
EPA can establish either a MACT or a GACT standard. See
id. § 7412(d)(5). Similarly, if the EPA finds that capping
emissions from an area source subcategory is necessary to
achieve a 90 per cent reduction in the aggregate emissions of
one of seven CAA-enumerated HAPs, section 7412(c)(6)
requires the Agency to impose caps in that subcategory as
well. See id. § 7412(c)(6). Upon that finding, however, the
EPA must impose a MACT standard. Id.
In addition to prescribing requirements for inclusion of
area source subcategories, the CAA provides a mechanism for
removal of area source subcategories that, in the EPA’s view,
no longer need to be controlled. Specifically, the EPA can
“delete” any subcategory if it finds that no source or group of
sources in it (1) emits cancer-causing HAPs at a volume
sufficient to increase the lifetime risk of cancer in the
population by more than one in one million and (2) emits non-
cancer-causing HAPs at a level in excess of that which is
adequate “to protect public health with an ample margin of
111
safety” and to prevent against environmental harm. Id.
§ 7412(c)(9)(B). The section 7412(c)(9) process is known as
“delisting.”
In 1998, the EPA identified several area source boiler
subcategories—including oil-fired, industrial wood,
commercial oil-fired and commercial wood-combustion
boilers—as contributors to the “90 per centum of the
aggregate emissions” of Hg and POM under section
7412(c)(6). See Source Category Listing for Section
112(d)(2) Rulemaking Pursuant to Section 112(c)(6)
Requirements, 63 Fed. Reg. 17,838, 17,839 (Apr. 10, 1998).
When it decided to “list” these sources, however, the EPA
included a caveat. It explained that it used the best emissions
information it had at the time to conclude that these boiler
subcategories produced enough Hg and POM emissions to
justify section 7412(c)(6) control but it also admitted that it
could not “assure that this calculation of the 90 percent will
remain constant.” Id. at 17,840.
The caveat proved prescient. When the EPA issued the
2010 Proposed Area Boilers Rule, it decided it needed to
regulate only coal-fired boilers at the MACT level to control
90 per cent of Hg emissions. See 75 Fed. Reg. at 31,898.
And when it finalized the 2011 Area Boilers Rule, the Agency
similarly decided that it needed to regulate only coal-fired
boilers at the MACT level to control 90 per cent of POM
emissions. See 76 Fed. Reg. at 15,566.
For this reason, the EPA established GACT, rather than
MACT, standards for the oil-fired and biomass-fired area
source subcategories regarding these two pollutants. See id.
It did not, however, make any of the “delisting” findings
required by section 7412(c)(9) when it removed these area
112
source subcategories from section 7412(c)(6)’s purview. See
2011 Area Boilers Rule, 76 Fed. Reg. at 15,566 (“[W]e have
not removed or ‘delisted’ oil-fired and biomass-fired area
source boilers by this action. We are not promulgating
MACT-based regulations at this time because they are
unnecessary to meet the requirements of CAA section
112(c)(6).”). The Environmental Petitioners challenge the
EPA’s imposition of GACT standards, arguing that, because
once the EPA “listed” these sources under section
7412(c)(6)’s MACT requirement, the CAA mandates that the
EPA “delist” them under section 7412(c)(9) before putting
them under the more lenient GACT standards. In their view,
the EPA’s contrary approach fails at Chevron step 1. The
EPA responds that section 7412(c)(9) applies only if it
decides to “delist” a subcategory entirely from section 7412
regulation, resulting in neither MACT nor GACT restrictions.
Because section 7412(c)(9) does not unambiguously
apply to section 7412(c)(6) and because the EPA’s
interpretation of section 7412(c)(9)’s delisting requirement is
reasonable, we uphold the EPA’s decision as permissible
under Chevron step 2. Section 7412(c)(9) provides that the
EPA “may delete any source category from the list under this
subsection” on its finding that the source category is not a
threat to human health or the environment. 42 U.S.C.
§ 7412(c)(9)(B). The inclusion of a singular “list” to govern
“this subsection” seems, most naturally, to refer to the list
contemplated by section 7412(c)(1), which states that the
EPA “shall publish, and shall from time to time . . . revise . . .
a list of all categories and subcategories of . . . area sources
(listed under paragraph (3)).” Id. § 7412(c)(1) (emphasis
added). In other words, it appears that section 7412(c)(1)
directs the EPA to create one “list” of source categories and
subcategories to subject to emission controls and section
113
7412(c)(9) instructs how to remove source categories from
that list. This conclusion finds support in section 7412(c)(1)’s
cross reference to “paragraph (3)” of section 7412(c), which
lays out the circumstances under which the EPA “shall list”
area source categories for emissions control. Id. § 7412(c)(3).
In the Environmental Petitioners’ view, section
7412(c)(9) also applies to a second, subsidiary list—that
contemplated by section 7412(c)(6), requiring imposition of
the MACT standard. Granted, section 7412(c)(6) mandates
that the EPA “shall . . . list” source categories and
subcategories if doing so is necessary to control 90 per cent of
the aggregate emissions from seven enumerated pollutants.
Id. § 7412(c)(6) (emphasis added). But the use of the verb
“list” in section 7412(c)(6) does not unambiguously establish
that 7412(c)(9), titled “[d]eletions from the list,” applies.
Because section 7412(c)(9) is ambiguous, we defer to the
EPA so long as its interpretation is “based on a permissible
construction.” Chevron, 467 U.S. at 842-43. And the EPA’s
reading of section 7412(c)(9)—that it applies only if the EPA
wants to remove a category from all section 7412
regulation—is reasonable.
First, the EPA’s approach harmonizes sections
7412(c)(1), 7412(c)(3), and 7412(c)(9). Because the EPA
must find that an area source “presents a threat of adverse
effects to human health or the environment” before it
regulates the source category at all, id. § 7412(c)(3), it makes
sense to require the EPA to find that “no source in the
category or subcategory . . . exceed[s] a level which is
adequate to protect public health . . . and no adverse
environmental effect will result from emissions from any
source” before it completely deregulates that category, id.
§ 7412(c)(9). It makes less sense to require the EPA to make
114
the same findings before it opts for GACT instead of MACT
standards, which occurs when the EPA removes a source from
section 7412(c)(6)’s purview but continues to regulate it
under section 7412(c)(1).
Second, the EPA’s approach is consistent with our
decision in New Jersey, 517 F.3d 574. There, we held that
“the only way EPA could remove [a source category] from
the section [7412(c)(1)] list was by satisfying section
[7412(c)(9)’s] requirements.” Id. at 582. In other words, New
Jersey held that the EPA cannot remove a source category
from all section 7412 regulation without delisting it; it said
nothing about the process by which the EPA moves source
categories from section 7412(c)(6).
Finally, the Petitioners’ argument would freeze the
EPA’s decision as to which sources need to be controlled to
reach the requisite 90 per cent emissions reduction for the
section 7412(c)(6) pollutants until it determines that “no
source in the category or subcategory . . . exceed[s] a level
which is adequate to protect public health . . . and no adverse
environmental effect will result from emissions from any
source.” 42 U.S.C. § 7412(c)(9). This, in turn, would hamper
the EPA’s ability to respond to updated data, thereby
substantially complicating its attempts to control the
pollutants. Nothing in the CAA suggests that the Congress
intended to so hamstring the Agency.
G. TITLE V PERMIT EXEMPTION FOR SYNTHETIC BOILERS
The EPA has discretion to exempt one or more area
source categories from Title V permitting requirements upon
a finding “that compliance with such requirements is
impracticable, infeasible, or unnecessarily burdensome on
such categories.” 42 U.S.C. § 7661a(a). The EPA originally
115
proposed exempting some area source categories because
existing “testing, monitoring, notification, and recordkeeping
requirements” rendered Title V permitting cumulative. 2010
Proposed Area Boilers Rule, 75 Fed. Reg. at 31,910. At the
time, however, the EPA elected not to exempt synthetic area
sources as one of those categories. Id. at 31,913. Synthetic
area sources are boilers that “naturally” emit pollutants at a
major source level but which qualify as area sources due to
the voluntary adoption of air pollution control technologies.
Id. Despite its initial stance, the EPA ultimately decided to
exempt all area sources—including synthetic area sources—
from Title V’s permitting requirements. See 2011 Area
Boilers Rule, 76 Fed. Reg. at 15,578.
Environmental Petitioners argue the EPA’s decision to
exclude synthetic boilers from Title V licensing requirements
is arbitrary and capricious for two reasons. First, they say, the
EPA arbitrarily concluded synthetic area sources would bear
the same level of burden as other area sources in complying
with Title V permitting requirements, rather than a lesser one.
See No. 11-1141 Envtl. Pet’rs’ Br. 39-43. And second, they
contend the EPA arbitrarily dismissed the additional
compliance benefits of Title V licensing for these synthetic
sources. See id. at 43-47. Under State Farm, “an agency rule
[is] arbitrary and capricious if the agency . . . offered an
explanation of its decision that runs counter to the evidence
before the agency.” 463 U.S. at 43. A court may not accept
an agency’s “post hoc rationalizations” for its
decisionmaking. Id. at 49.
The EPA has authority under the CAA to exempt sources
from Title V permitting requirements if those requirements
would be “impracticable, infeasible, or unnecessarily
burdensome” on the area source. 42 U.S.C. § 7661a(a). The
116
EPA previously developed a four-factor balancing test to
determine whether Title V’s requirements are “unnecessarily
burdensome.” See Exemption of Certain Area Sources from
Title V Operating Permit Programs, 70 Fed. Reg. 75,320,
75,323 (Dec. 19, 2005). Under this test, the EPA considers
whether: (1) Title V permitting would result in significant
improvements in compliance with emission standards;
(2) whether Title V permitting would impose significant
burdens on the area source category; (3) whether the costs are
justified, taking into account potential gains; and (4) whether
there are existing enforcement programs in place sufficient to
ensure compliance. See id. at 75,323-26. The EPA also must
consider, consistent with the legislative history of the CAA,
whether exemption would “adversely affect public health,
welfare, or the environment.” Id. at 75,333-34. These factors
are considered in combination and not every factor must point
in favor of exemption for the EPA to choose that course. See
id. at 75,323.
In its 2010 Proposed Area Boilers Rule, the EPA applied
this balancing test and excluded almost all area source boilers
except synthetic boilers that achieved “area” status via
installation of a control technology (although it exempted
those that achieved “area” status through operational
changes). The EPA provided an extensive rationale for its
decision to exclude these “natural” area sources from Title
V’s permitting requirements. See 2010 Proposed Area
Boilers Rule, 75 Fed. Reg. at 31,910-13. With respect to
factor one, the EPA found its proposed rule already required
“direct monitoring of emissions,” both continuously and
periodically, recordkeeping that would allow for additional
monitoring, and “semi-annual reporting to assure
compliance.” Id. at 31,911. Moreover, under the proposed
rule, “records are required to be maintained in a form suitable
117
and readily available for expeditious review” for up to five
years. Id. The EPA acknowledged Title V permitting could
provide some additional compliance benefits; specifically,
that Title V has an every-six-month monitoring and reporting
requirement. See id. But the EPA ultimately concluded the
monitoring, recordkeeping, and reporting requirements of its
proposed rule were sufficient to assure compliance: “Given
the nature of the operations at most area sources and the types
of requirements in this rule, Title V would not significantly
improve those compliance requirements.” Id.
As to the second factor, the EPA noted that subjecting
most area sources to Title V would “impose[] certain burdens
and costs that do not exist outside of the [t]itle V program.”
Id. at 31,912. One of the EPA’s major concerns was that
“requiring permits for the large number of area sources could,
at least in the first few years of implementation, potentially
adversely affect public health, welfare, or the environment by
shifting [s]tate agencies[’] resources away from assuring
compliance for major sources with existing permits to issuing
new permits for these area sources, potentially reducing
overall air program effectiveness.” Id. at 31,913. For the
third factor, the EPA concluded the costs of compliance
would “impose a significant burden on many of the
approximately 137,000 facilities affected by this proposed
rule” with only “low” potential gains in compliance. Id. at
31,912. Finally, for the fourth factor, the EPA determined
that “[s]tate delegated programs are sufficient to assure
compliance with this [rule],” and noted that the Agency
retains authority to enforce this rule “anytime.” Id. The EPA
therefore proposed exempting these area sources from the
permitting requirements. See id. Environmental Petitioners
are not currently challenging the exemption for non-synthetic
area boilers.
118
However, in this 2010 rulemaking, the EPA also
explained precisely why it declined to exempt synthetic area
sources that installed air pollution controls from Title V
requirements. First, the EPA noted these synthetic area
sources “represent less than one percent of the total number of
sources that will be subject to the final rule.” Id. at 31,913.
The EPA also characterized these sources as “much more like
the major sources” that are not exempt from Title V
permitting requirements. Id. Further distinctions included
that “many of these sources are located in cities, and often in
close proximity to residential and commercial centers where
large numbers of people live and work,” that they “have
significantly higher emissions potential when uncontrolled”
(even compared to synthetic boilers that adopted operational
limits to attain area source status), and that many of these
sources “are large facilities with comprehensive compliance
programs in place” as opposed to small facilities, like schools
or hospitals. Id. Given these distinctions, the EPA concluded
additional public involvement and compliance oversight
through Title V was “important to ensure that these sources
are maintaining their emissions at the area source level.” Id.
But the EPA shifted its position in the 2011 Area Boilers
Rule by deciding to exempt all area sources, including
synthetic sources. See 76 Fed. Reg. at 15,578. The EPA
provided only a cursory explanation for this shift, noting how
a further review of the record led it to conclude “observations
and data we have relied upon in other rulemakings for
distinguishing between sources that became synthetic area
sources due to controls and other synthetic and natural area
sources [do] not necessarily apply to this source category.”
Id. (emphasis added). Because the EPA asserted it no longer
had “sufficient information” to identify control-technology-
dependent synthetic sources, it decided to apply the same
119
rationale used to exempt “natural” sources to these synthetic
sources. Id. (“[T]he rationale for exempting most area
sources subject to this rule . . . is also now relevant for sources
which we proposed to permit [under Title V].”). But—even if
the EPA truly cannot distinguish between synthetic sources
relying on control technologies and other sources—it does not
invariably follow that the justifications the Agency relied on
for exempting “natural” sources under the four-factor
balancing test can be transposed onto these synthetic sources.
Cf. Sierra Club II, 479 F.3d at 884 (“We agree with the Sierra
Club that EPA’s use of work practice standards instead of
emission floors violates section 7412(h). That provision
allows EPA to substitute work practice standards for emission
floors only if measuring emissions levels is technologically or
economically impracticable. Here, EPA never determined
that measuring emissions from ceramic kilns was
impracticable; it determined only that it lacked emissions data
from ceramics kilns. EPA thus had no basis under section
7412(h) for using work practice standards.”).
In its next iteration of the rule, the EPA endeavored to
further explain its exemption of synthetic sources. The EPA
again stated it “lacked sufficient information” to distinguish
these synthetic sources from other area sources. See 2011
Proposed Area Boilers Rule on Reconsideration, 76 Fed. Reg.
at 80,538. The Sierra Club challenged this exemption in a
comment, and the EPA responded with “additional analysis”
of the synthetic exemption. Id. In this analysis, the EPA first
reiterated the difference in number between the two types:
estimating there to be at least 48 control-technology-
dependent synthetic sources versus 137,000 other area
sources, most of which are located at small facilities like
schools, hospitals, and churches. See id. The EPA then
provided a new rationale for the exemption: that these
120
synthetic facilities “may already have a Title V permit for
other reasons.” Id. The EPA also found that “synthetic area
sources would likely be subject to more stringent permit and
monitoring requirements than natural area sources” because
they have a “legal duty to use the control equipment” to keep
them at an “area” level. Id. (emphasis added). Finally, the
EPA made several assertions about the similarities between
synthetic and natural sources. Specifically, that synthetic
sources are “similar in size and sophistication to those that are
natural area sources,” that their “uncontrolled emissions are
generally on the same order of magnitude as the emissions of
natural sources,” and that “the facilities and owners are
comparable in size.” Id. The EPA provided no data or
examples in support of these assertions, which appear to
directly contradict the distinctions the EPA listed in its earlier
version of the rule. Compare id., with 2010 Proposed Area
Boilers Rule, 75 Fed. Reg. at 31,913. In its final rule, the
EPA declined to make any changes to its Title V
exemptions—exempting all area sources including synthetic
sources using a control technology. See 2013 Area Boilers
Rule, 78 Fed. Reg. at 7,497.
Based on this record, the EPA’s reasoning has several
fatal flaws that render its exemption decision arbitrary. The
EPA put forward two primary justifications for exempting
synthetic sources: (1) that it could not necessarily rely on
existing data for distinguishing the different type of sources,
and (2) that these facilities are “similar in size and
sophistication” to natural area sources. See 2011 Proposed
Area Boilers Rule on Reconsideration, 76 Fed. Reg. at
80,538. The second justification flatly contradicts the EPA’s
earlier, extensive discussion about how these synthetic
sources have higher emissions potential and are often located
on large sites with existing compliance programs, in addition
121
to being uniquely few in number and generally found near
cities. 2010 Proposed Area Boilers Rule, 75 Fed. Reg. at
31,912-13. These factors all undercut the EPA’s assertion
that synthetic sources are “similar”—in size, sophistication, or
otherwise—to natural sources. With respect to the lack of
data for distinguishing, the EPA was able to estimate in its
proposed rule that 48 synthetic sources would have been
affected by this rule—which suggests the EPA possesses
some mechanism for distinguishing the types. See 2011
Proposed Area Boilers Rule on Reconsideration, 76 Fed. Reg.
at 80,538. Moreover, the EPA does not explain why the data
it used in prior rulemakings to distinguish these source types
is not accurate in this context. Environmental Petitioners also
point out that, “to qualify for area-source status, synthetic area
sources must notify the EPA or the state permitting authority
of the limits on their emissions,” such that the EPA “need
only ask these authorities to identify the sources operating in
their states.” No. 11-1141, Envtl. Pet’rs’ Br. 39-40. The EPA
never endeavors to explain why that mechanism (or any other
existing mechanism) is insufficient for identifying synthetic
area sources.
Because its justifications for the final rule contradict
earlier findings, the EPA must provide some reasoning to
explain why its final decision “runs counter to the evidence
before the agency.” State Farm, 463 U.S. at 43. The EPA’s
proffered explanation fails. This court has “often declined to
affirm an agency decision if there are unexplained
inconsistencies in the final rule.” See Dist. Hosp. Partners,
786 F.3d at 59; see also Gulf Power Co. v. FERC, 983 F.2d
1095, 1101 (D.C. Cir. 1993) (“[W]hen an agency takes
inconsistent positions . . . it must explain its reasoning.”);
Gen. Chem. Corp. v. United States, 817 F.2d 844, 846 (D.C.
Cir. 1987) (holding agency action to be arbitrary because its
122
analysis was “internally inconsistent and inadequately
explained”). The EPA had a duty here to examine and justify
the “key assumptions” underlying its decision, and it failed to
do so. See Appalachian Power Co. v. EPA, 135 F.3d 791, 818
(D.C. Cir. 1998) (“EPA retains a duty to examine key
assumptions as part of its affirmative burden of promulgating
and explaining a nonarbitrary, non-capricious rule.” (internal
quotation marks omitted)).
The EPA’s major oversight was its failure to explain why
the rationale it used to exempt natural area sources from Title
V could be identically applied to synthetic area sources. One
of the Agency’s main justifications for exempting natural area
sources was that their prolific numbers might overwhelm state
and local regulatory agencies, diverting resources from other
important environmental programs, thereby harming public
health and welfare. The EPA never explained why requiring
48 synthetic area sources to comply with Title V would strain
government resources to a comparable degree as would
requiring the 137,000 natural area sources to comply. As
discussed above, the EPA also did not explain how it
suddenly determined these synthetic area sources were
“similar in size and sophistication” to natural sources, when it
had previously articulated several key differences. It is
particularly unclear how these synthetic sources could have
“uncontrolled emissions . . . generally on the same order of
magnitude as the emissions of natural area sources.” 2011
Proposed Area Boilers Rule on Reconsideration, 76 Fed. Reg.
at 80,538. Given that synthetic sources are defined as
“major” sources that have artificially reduced their emissions
to an “area” level, it is difficult to understand how the
uncontrolled emissions of these sources would be similar to
natural area sources. Additionally, the EPA asserted that
synthetic source “facilities and owners are comparable in
123
size” to natural sources. Id. This contradicts earlier findings
that synthetic sources tend to be large, located on sites with
existing compliance plans, and near population-dense areas.
The EPA provides no data or explanation to support this shift.
The EPA relies on another problematic premise when it
claims the potential benefits of subjecting synthetic area
sources to Title V requirements are low. Both the EPA and
Industry Intervenors argue that the added benefits of Title V
would be minimal for these synthetic sources, relying solely
on the rationale given for natural sources. But the EPA
originally asserted “additional public involvement and
compliance assurance requirements through title V [are]
important to ensure that these sources are maintaining their
emissions at the area source level.” 2010 Proposed Area
Boilers Rule, 75 Fed. Reg. at 31,913 (emphasis added). The
EPA never explains why these additional benefits were
considered “important” before but are now “not important”
simply because it allegedly determined that synthetic sources
may be hard to distinguish from natural sources. The
difficulty in identifying synthetic sources says nothing about
the benefits that may be gained by requiring Title V permits,
assuming the sources can be identified. Synthetic sources
retain the attributes which first motivated the EPA to subject
them to Title V permitting: they tend to be near cities,
specifically near large residential populations, and they have
greater emission potential if their control technology is
removed, turned off, or not kept up to standards. The EPA
arguably finds Title V’s additional compliance benefits
unnecessary because synthetic sources have “a legal duty to
use the control equipment” and that use is “not optional.”
2011 Proposed Area Boilers Rule on Reconsideration, 76 Fed.
Reg. at 80,538. But that observation does not speak to the
need for public oversight; just because facilities are obligated
124
to use the control technology does not mean they will always
do so. Title V’s process requires facilities to submit
compliance documentation every six months—far more
frequently than under the EPA’s current rule—which expands
the opportunity for public oversight and compliance. Perhaps
this “legal duty” provides a stronger incentive for compliance
than public oversight but, if so, the EPA still fails to explain
how.
Similarly, for factor three’s balancing of costs and
benefits, the EPA never justifies applying to natural sources—
which tend to be small sites like schools, hospitals, and
churches—the same rationale it applies to these larger
synthetic sources, which tend to be located at refineries,
chemical plants, and factories. Given these distinctions, it is
at least possible this balancing would lead to a different
outcome for synthetic sources. Taken as a whole, the EPA’s
analysis fails to explain why several of the facts and
characteristics it relied on for its initial assessment are no
longer relevant—creating several glaring inconsistencies in
the rulemaking record. The EPA offers no plausible reason
for applying the results of the four-factor test for natural
sources wholesale to these control-technology-dependent
synthetic sources. We do not hold, however, that the EPA can
never remove synthetic area sources from the ambit of Title V
compliance. The outcome the EPA ultimately reached may
be reasonable; however, “[n]ot only must an agency’s decreed
result be within the scope of its lawful authority, but the
process by which it reaches that result must be logical and
rational.” Allentown Mack Sales & Serv., Inc. v. NLRB, 522
U.S. 359, 374 (1998). The EPA should have applied its four-
factor balancing test directly to synthetic sources or, at a
minimum, provided an explanation for adopting the natural
125
source balancing test that is not premised on inconsistencies
in the record.
With respect to remedy, there is a strong possibility that
the Agency can properly explain its decision to exclude
synthetic boilers from the Title V permitting requirement;
moreover, vacating the decision would be unnecessarily
disruptive for synthetic boiler operators who, in the interim,
would not know whether they needed to begin the expensive,
time-consuming process of obtaining a Title V permit. See
Allied-Signal, Inc. v. U.S. Nuclear Regulatory Comm’n, 988
F.2d 146, 150-51 (D.C. Cir. 1993). We therefore remand this
issue (without vacating) for further explanation by the EPA.
H. GACT STANDARD DETERMINATIONS
With few exceptions, the EPA has broad discretion to
choose how to control area source emissions. For instance,
the EPA has discretion to choose between GACT and MACT
standards in the majority of cases. See 42 U.S.C.
§ 7412(d)(5). Even if the EPA chooses a MACT standard, it
has discretion—although somewhat circumscribed—to set a
work-practice standard instead of a numeric standard. Id.§ 7412(h)(1). And the EPA has discretion when choosing
among different GACT-standard options. See id.
§ 7412(d)(5).
Accordingly, we must uphold the EPA’s GACT-standard
determinations so long as it “has considered the relevant
factors and articulated a rational connection between the facts
found and the choice made, and has not relied on [improper]
factors.” Nat’l Ass’n of Clean Air Agencies, 489 F.3d at 1228
(citations and internal quotation marks omitted). But for all
of the discretion the EPA enjoys, it must nonetheless
demonstrate that it exercised its judgment in a reasoned way.
126
The cases establishing this principle are legion. See, e.g.,
Transactive Corp. v. United States, 91 F.3d 232, 236 (D.C.
Cir. 1996) (agency must “identif[y] and explain[] the
reasoned basis for its decision”); Int’l Fabricare Inst., 972
F.2d at 389 (agency must “examine[] the relevant data and
. . . articulate[] an adequate explanation for its action”). The
EPA need not go to great lengths to meet its burden; indeed,
we “uphold a decision of less than ideal clarity” so long as
“the agency’s path may reasonably be discerned.” State
Farm, 463 U.S. at 43 (quoting Ark.-Best Freight Sys., Inc.,
419 U.S. at 286).
With these principles in mind, we address the
Environmental Petitioners’ two challenges to the EPA’s
discretionary decisions regarding the Area Boilers Rule.
1. EPA’s Selection of GACT Standards
for Non-Hg Metals
The Environmental Petitioners argue that the EPA failed
to support its decision-making when it established MACT
standards for Hg and POM emissions from some coal-fired
boilers but declined to regulate non-Hg emissions under the
MACT standard from the same boilers. We agree. Although
the EPA thoroughly explained why it chose to impose one
GACT standard instead of another, nothing in the record
explains why the EPA decided to impose GACT standards
instead of MACT standards in the first place. Despite the
Agency’s broad discretion, we cannot sustain its action in the
absence of some explanation for why GACT standards are
more appropriate than MACT standards for these sources and
types of pollutants. See Transactive Corp., 91 F.3d at 236.
For this reason, we remand (but do not vacate) the EPA’s
choice of GACT standards for non-Hg emissions from coal-
127
fired boilers. See Sierra Club, 167 F.3d at 664; Nat’l Lime
Ass’n, 233 F.3d at 634-35.
2. EPA’s Selection of Certain GACT Standards
The Environmental Petitioners also challenge several of
the EPA’s choices among different GACT standards. As
noted, see supra § I.A.1.c, the CAA provides virtually no
instruction regarding GACT standards but the standards
generally take the form of “methods, practices and techniques
which are commercially available and appropriate for
application by the sources in the category considering
economic impacts and the technical capabilities of the firms
to operate and maintain the emissions control systems.” S.
REP. NO. 101-228, at 171 (1989). Because the EPA has
ample discretion to choose the appropriate GACT standard,
we will affirm its choices so long as we can discern reasoned
decision-making from the record. State Farm, 463 U.S. at 43.
For the reasons set forth below, we can do so here and,
accordingly, we reject the Environmental Petitioners’ GACT-
focused challenges.
First, the Environmental Petitioners challenge the data set
the EPA used to arrive at the numeric GACT standards for
non-Hg-metal emissions from coal-fired boilers. Specifically,
they contend that the EPA set the GACT limit based on
boilers with no control technology, which resulted in a
numeric standard of 0.42 lb/mmBtu. They insist that the EPA
should instead have examined boilers outfitted with fabric
filters, which would have resulted in a numeric standard of
0.03 lb/mmBtu. The EPA, however, thoroughly explained
why it considered the uncontrolled boiler data set.
Specifically, the controlled data set derives from the EPA’s
“New Source Performance Standards” (NSPS) data, which, in
128
the Agency’s view, could be used to set the non-Hg-metal
GACT standard for boilers with a heat input capacity of 30
mmBtu/hr or greater but did not suffice for boilers with a
lower heat input capacity. For this reason, the EPA examined
its original data set, found that none of the coal-fired boilers
in that set used control technology and, accordingly, set the
GACT numeric standard at the emissions level achieved by
the best performing uncontrolled source in that data set (i.e.,
0.42 lb/mmBtu). We are satisfied that the EPA exercised its
discretion in a reasoned manner and, accordingly, we do not
disturb it. See Transactive Corp., 91 F.3d at 236.
Next, the Environmental Petitioners challenge the EPA’s
decision to establish a tune-up requirement as a GACT
management-practice standard for Hg and POM emissions
from large biomass-fired and oil-fired boilers. In their view,
other, more restrictive control technologies, including
multiclones,38 are “generally available” and their availability
mandates that the EPA set numeric standards based on boilers
that use those controls. But the EPA explained its approach:
A boiler tune-up requirement would potentially
result in the same non-mercury metallic HAP
reduction as a PM emission limit based on
performance of multiclones but would also
reduce emissions of organic HAP. In addition
the cost of a boiler tune-up appears minimal
compared to the cost for testing and
38
A multiclone is a PM “mechanical separator[].” See 2010
Proposed Area Boilers Rule, 75 Fed. Reg. at 31,908. It diverts
particles from the exhaust stream by creating a circular air flow.
See id.
129
monitoring to demonstrate compliance with an
emission limit.
See 2010 Proposed Area Boilers Rule, 75 Fed. Reg. at 31,908.
The EPA also explained that multiclones were “minimally
effective” for controlling non-Hg metals, ineffective for POM
and Hg, and expensive. Id. Because the EPA’s decision to
impose a tune-up requirement fits within its “technical
expertise,” we owe the Agency an “extreme degree of
deference” so long as its explanation is rational. Catawba
Cty., N.C. v. EPA, 571 F.3d 20, 41 (D.C. Cir. 2009) (quoting
City of Waukesha v. EPA, 320 F.3d 228, 247 (D.C. Cir.
2003)). And because its explanation was rational, we reject
the Petitioners’ challenge thereto.
Finally, the Petitioners challenge the EPA’s decision to
set a tune-up requirement as a management-practice standard
for small biomass-fired and oil-fired area boilers. The EPA
adopted this approach because measuring PM emissions for
smaller boilers is “not feasible.” 2010 Proposed Area Boilers
Rule, 75 Fed. Reg. at 31,906. When the EPA explained its
decision regarding small biomass-fired and oil-fired area
boilers, it provided the same reasons it gave for its use of a
tune-up requirement for small coal-fired area boilers, which
we address (and uphold), infra, § IV.M. For those reasons,
we reject the challenge to the EPA’s tune-up requirement for
small biomass-fired and oil-fired area boilers.
I. 30-DAY ROLLING AVERAGE
As discussed, see supra § I.B.1.a, when the EPA sets a
MACT floor, it begins by examining data generated by stack
testing. Once the MACT standard is established, however, a
source may (and in some cases, must) demonstrate
compliance by implementing “continuous monitoring” instead
130
of conducting additional stack tests.39 See 2011 Proposed
CISWI Rule on Reconsideration, 76 Fed. Reg. at 80,464-65.
For a source using a continuous monitor, the EPA determines
MACT-standard compliance based on the source’s thirty-day
“rolling average.” Id. at 80,465.
The calculation of a thirty-day rolling average is
straightforward: the average of a source’s daily emissions for
the immediately preceding thirty days. Each day produces a
new rolling average and each “average is a separate
compliance determination.” No. 11-1125 EPA Br. 88 n.17.
In the EPA’s view, this “allow[s] operators sufficient
flexibility for operational and control device adjustments
should they be needed for short term fuel or waste
characteristics variability.” 2011 Proposed CISWI Rule on
Reconsideration, 76 Fed. Reg. at 80,465. The EPA also
concluded that thirty-day rolling average violations will occur
almost as frequently as violations of shorter rolling-average
periods. Id.
The CAA vests the EPA with authority to “prescribe
procedures and methods for determining compliance and for
monitoring and analysis of pollutants.” 42 U.S.C. § 7661c(b).
We have emphasized that the EPA has “broad discretion in
selecting a monitoring regime that ensures compliance, and as
long as it reasonably articulate[s] the basis for its decision,
[we] will defer to the informed discretion of the Agency,
39
As the name suggests, a continuous monitoring system
measures the source’s emissions at all times and generally takes one
of two forms: (1) a continuous parameter monitor, which measures,
e.g., a source’s temperature, pressure or oxygen content; or (2) a
continuous emissions monitor, which measures the pollutant
concentration in the source’s emissions.
131
recognizing that analysis of this issue requires a high level of
expertise.” White Stallion Energy Ctr., LLC v. EPA, 748 F.3d
1222, 1255 (D.C. Cir. 2014) (per curiam) (citation and
internal quotation marks omitted), rev’d on other grounds by
Michigan v. EPA, 135 S. Ct. 2699 (2015). Notwithstanding
this deference, the Environmental Petitioners argue that
allowing a source to demonstrate compliance by way of a
thirty-day rolling average not only fails Chevron review but is
also arbitrary. We disagree.
First, they argue that the thirty-day rolling average fails at
Chevron step 1 because it allows sources to emit HAPs
continuously at the UPL-established MACT floor. Because
they do not believe that the UPL represents the average
emissions level achieved by the best performing sources, they
argue that, ipso facto, allowing sources to continuously emit
HAPs at the UPL level means that sources are permitted to
emit at levels higher than the average levels achieved by the
best performing sources. Because we have already concluded
that the UPL is in fact a reasonable proxy for the average
emissions level achieved by the best performing sources, see
supra § IV.C, the Environmental Petitioners’ premise is
inaccurate. And because the “total emissions from a unit
complying with a rolling average must still be below the total
emissions from a unit emitting continuously at the level of the
standard,” No. 11-1125 EPA Br. 90, the Environmental
Petitioners’ Chevron step 1 argument fails.
The Environmental Petitioners’ Chevron step 2 argument
fares no better. The EPA explained that (1) it expects to catch
violations using a thirty-day rolling average “almost as much
as for a shorter term average” and (2) it believes the longer
average to be more effective in addressing “[c]oncerns of
variability outside the operators[’] control such as fuel
132
content, seasonal factors, load cycling, and infrequent hours
of needed operation.” 2011 Proposed CISWI Rule on
Reconsideration, 76 Fed. Reg. at 80,465. Because the EPA
“reasonably articulate[d] the basis for its decision,” we uphold
it. White Stallion, 748 F.3d at 1255 (internal quotation marks
omitted).
Finally, we conclude that the EPA’s allowance of thirty-
day rolling averaging does not reflect an arbitrary change in
position. Although the Petitioners cite other rules that, in
their view, manifest that the EPA once believed that longer
averaging periods resulted in less stringent enforcement, most
of the rules they cite have nothing to do with MACT-setting
or MACT compliance40 and none evidences an unexplained or
unjustified deviation. Similarly, the Petitioners point to the
EPA’s explanation in the 2011 CISWI Rule that “24-hour
block averages . . . would be inconsistent with the sampling
time for the stack test data” to indicate an arbitrary change in
position. See 76 Fed. Reg. at 15,728. But the EPA made this
statement while discussing why stack-test data and
continuous-monitoring data could not be used in tandem to set
a MACT level, which says nothing about allowing emissions
40
The only exception is the EPA’s 1996 Medical-Waste
Incinerators Rule, which provides that “[t]he period of time over
which emissions are measured and then averaged to determine
compliance with the regulation . . . must correspond to the period of
time over which emission levels were measured and averaged in
determining the emission limits included in the regulation.”
Standards of Performance for New Stationary Sources and
Emission Guidelines for Existing Sources: Medical Waste
Incinerators, 61 Fed. Reg. 31,736, 31,748 (June 20, 1996). This
twenty-year-old statement, however, does not detract from the
EPA’s well-reasoned defense of the thirty-day rolling average in
the CISWI Rule.
133
averaging—long-term or otherwise—to gauge MACT-floor
compliance. See id. And even if the Environmental
Petitioners had directed us to a real about-face, the EPA’s
justification for allowing the thirty-day rolling average
convinces us that any change was not arbitrary.
J. FUEL-COMBUSTION-BASED SUBCATEGORIES
Section 7412 provides that the EPA may distinguish
among “classes, types, and sizes” of sources when
establishing emission standards. 42 U.S.C. § 7412(c), (d)(1).
Under this authority, the EPA created subcategories of major
source boilers based on the fuel the boiler was designed to
burn. 2013 Major Boilers Rule, 78 Fed. Reg. at 7,144. The
Environmental Petitioners challenge this decision on three
grounds: First, they claim that the text of the statute
forecloses the EPA from creating such subcategories. Next,
they argue that the EPA’s subcategories are arbitrary because
they permit a boiler to switch subcategories from year to year.
Finally, they contend that the EPA’s action was arbitrary
because the Agency failed to demonstrate with substantial
evidence that burning a different fuel alters the boiler’s class,
type, or size.
These arguments fail. Section 7412(d) gives the EPA
discretion to create subcategories based on boiler type, and
nothing in the statute forecloses the Agency from doing so
based on the type of fuel a boiler was designed to burn. Nor
was the EPA’s decision to create such subcategories arbitrary
and capricious. The Agency considered the relevant factors in
coming to a reasoned decision that the type of fuel a boiler is
designed to burn impacts the feasibility of emission standards.
And, finally, the EPA based its technical judgment on
sufficient record evidence. As a result, we deny the
134
Environmental Petitioners’ challenge to the EPA’s
subcategorization of major source boilers based on the type of
fuel the boiler is designed to burn.
The Environmental Petitioners first claim that the text of
the CAA forecloses the EPA from creating subcategories of
“types” of boilers based on the fuel a boiler burns because a
single boiler may use different fuels over the course of its
lifetime. This may be true, but the Petitioners never explain
what it is about the word “type” that bars the EPA from
regulating a boiler that burns “x” differently from a boiler that
burns “y.” According to its ordinary meaning, “type” is easily
broad enough to accommodate changes in boiler
characteristics from year to year. See OXFORD ENGLISH
DICTIONARY (2013) (defining “type” as a “general form,
structure, or character distinguishing a particular kind, group,
or class of beings or objects”). There is no textual reason then
to assume that a boiler’s type must be written in stone.
Nor does our understanding of “type” write it out of the
statute, as the Petitioners contend. The EPA has done what
the term plainly encompasses: it has distinguished among
boilers based on the kind of fuel the boiler burned over the
last year. It is thus not surprising that we have interpreted a
similar provision to permit distinctions based on fuel inputs.
See Sierra Club v. Costle, 657 F.2d 298, 318-19 (D.C. Cir.
1981) (holding that the text of 42 U.S.C. § 7411, which
allows the EPA to “distinguish among classes, types and
sizes,” permits distinctions based on variations in the sulfur
content of coal used by utility plants). Likewise, we conclude
that section 7412’s undefined and unrestricted use of class,
type, or size does not foreclose the EPA’s interpretation.
135
This court will, as a result, defer to the EPA’s
interpretation so long as it is reasonable. See, e.g., Sierra
Club I, 353 F.3d at 990. And here, it is. The Agency
explained that boilers vary in their designs depending on the
type of fuel they burn. 2010 Proposed Major Boilers Rule, 75
Fed. Reg. at 32,016-17. These differences, according to the
Agency, affect boiler emissions and the feasibility of emission
controls. Id. And, because design constraints also restrict a
boiler’s ability to switch fuels, the Agency concluded that it
could determine a boiler’s type by looking at the fuel it had
burned over the previous 12-month period. Id. at 32,014.
The Environmental Petitioners point to nothing in the record
that calls into question either of these technical judgments,
which receive great deference. See NRDC I, 489 F.3d at
1375. Nor do the Petitioners offer any additional reasons in
support of their argument that the Agency has ventured
beyond its authority under the statute. In fact, the EPA’s
reasoning from the emissions data is consistent with the very
existence of a subcategorization authority because the grant of
this authority implicitly acknowledges that the EPA may need
to set different emission standards within a category of major
sources based on what is achievable for a subset of those
sources. Because the statutory text readily encompasses the
EPA’s interpretation for the reasons explained above, and
because the Environmental Petitioners offer no additional
argument as to why the EPA’s interpretation was
unreasonable, we reject the Petitioners’ Chevron challenge to
the EPA’s interpretation of its subcategorization authority.
The Environmental Petitioners nevertheless claim that the
EPA’s subcategories are arbitrary because a boiler is not of a
different type when it can be a boiler “designed to burn coal”
one year, and a boiler “designed to burn biomass” the next.
But this argument fails for the same reasons as the Chevron
136
argument we just rejected. The fact that boilers may switch
from one type to another over time does not, alone, render a
subcategorization arbitrary. With no discernable basis to find
the EPA’s choice here questionable, much less arbitrary, we
reject this argument too.
Finally, the Environmental Petitioners contend that the
Agency failed to demonstrate with sufficient evidence, rather
than mere assertions, that burning a different fuel makes the
boiler a different class, type, or size. The Petitioners largely
fail to develop this argument and, regardless, the EPA easily
met its burden. The EPA based its decision on documented
emissions data, several reports provided by the National
Energy Technology Laboratory on boiler operations, and
operating manuals provided by boiler manufacturers. See,
e.g., Summary of Public Comments and Responses for
National Emission Standards for Hazardous Air Pollutants for
Major Sources: Industrial, Commercial, and Institutional
Boilers and Process Heaters, EPA-HQ-OAR-2002-0058-
3511-A1 (Dec. 2012), at 558-63. These sources support the
EPA’s decision to distinguish boilers based on the type of fuel
they are designed to burn and the Agency’s conclusion that
boilers designed for one fuel type are unlikely to use another
fuel type. Id. The Petitioners present no contrary evidence,
nor do they attack the validity or accuracy of the data that the
EPA relied upon. We thus find no merit in the Petitioners’
various challenges to the EPA’s decision to subcategorize
major boilers based on the fuel the boiler is designed to burn.
K. “UNITS THAT BEGIN COMBUSTING SOLID WASTE”
AS “EXISTING” SOURCES
Section 7429(a)(2) distinguishes between “existing” and
“new” CISWI units. The former must comply with floors set
137
at the “average emissions limitation achieved by the best
performing 12 percent of units” and the latter must comply
with stricter floors set at the level achieved by “the best
controlled similar unit.” 42 U.S.C. § 7429(a)(2). “Modified”
units, defined as units “at which modifications have occurred”
that either experience changes that cost more than 50 per cent
of the original construction price or result in increased
emissions, see id. § 7429(g)(3), must be treated as “new,” see
id. § 7429(g)(2).
The preamble to the 2011 CISWI Rule states, “[u]nits
that begin combusting solid waste are considered existing
sources.” 76 Fed. Reg. at 15,714 (emphasis added).
Commentators objected that this blanket statement
contravened the Act’s plain terms, which mandate that the
EPA treat such sources as “new,” not “existing,” if they meet
the section 7429(g)(3) requirements. In its subsequent 2011
Proposed CISWI Rule on Reconsideration, the EPA refined
its position: “An existing source will not be considered a new
source solely due to a combustion material switch. Assuming
new source applicability is not triggered, existing sources that
change fuels or materials are considered existing
sources . . . .” 76 Fed. Reg. at 80,459.
The Environmental Petitioners argue that the EPA’s
broad statement in the 2011 CISWI Rule indicates that it
impermissibly changed its treatment of “modified sources” in
contravention of the CAA. The EPA, however, agrees that
any CISWI unit fitting the statutory criteria for a modified
source must comply with new-unit MACT levels, not
existing-unit MACT levels. See 42 U.S.C § 7429(g)(2). It
also recognizes that its categorical statement in the 2011
CISWI Rule “may have been imprecise” and, in any event, it
138
argues that the Environmental Petitioners have taken its
statement out of context. See No. 11-1125 EPA Br. 73.
We agree with the Agency. The EPA’s later statement
made clear that it intended to treat “sources that change fuels
or materials” as “existing sources” unless “new source
applicability,” as mandated by the Act, is “triggered.” See
2011 Proposed CISWI Rule on Reconsideration, 76 Fed. Reg.
at 80,459. Moreover, the Agency provided its more precise
statement while discussing specifically what constitutes a
“modification” for the CISWI Rule. See id. (“An existing
source will not be considered a new source solely due to a
combustion material switch.”). It made its earlier,
“imprecise” comment, in contrast, while describing when a
fuel change could mean the difference between regulation
under section 7412 or section 7429. See 2011 CISWI Rule,
76 Fed. Reg. at 15,714. Convinced that the EPA has not
impermissibly changed the statutory definition of “modified”
CISWI, we reject the Petitioners’ challenge.
L. EXCLUSION OF “TEMPORARY” BOILERS FROM AREA
BOILERS RULE
In the final 2013 Area Boilers Rule, the EPA excluded
“temporary boilers” from regulation under section 7412. See
78 Fed. Reg. at 7,491. The Rule defined “temporary boilers”
as “any gaseous or liquid fuel boiler that is designed to, and is
capable of, being carried or moved from one location to
another by means of, for example, wheels, skids, carrying
handles, dollies, trailers, or platforms.” Id. Moreover, a
boiler is not a temporary boiler if any of the following apply:
139
(1) The equipment is attached to a foundation.
(2) The boiler or a replacement remains at a
location within the facility and performs the
same or similar function for more than 12
consecutive months, unless the regulatory
agency approves an extension. An extension
may be granted by the regulatory agency
upon petition by the owner or operator of a
unit specifying the basis for such a request.
Any temporary boiler that replaces a
temporary boiler at a location within the
facility and performs the same or similar
function will be included in calculating the
consecutive time period unless there is a gap
in operation of 12 months or more.
(3) The equipment is located at a seasonal
facility and operates during the full annual
operating period of the seasonal facility,
remains at the facility for at least 2 years, and
operates at that facility for at least 3 months
of each year.
(4) The equipment is moved from one location
to another within the facility but continues to
perform the same or similar function and
serve the same electricity, steam, and/or hot
water system in an attempt to circumvent the
residence time requirements of this
definition.
2013 Area Boilers Rule, 78 Fed. Reg. at 7,491-92.
140
Environmental Petitioners challenge this exclusion as a
violation of the EPA’s obligations under the CAA to regulate
all boilers listed under section 7412. By its own terms, the
2011 Area Boilers Rule “applies to all existing and new
industrial boilers, institutional boilers, and commercial boilers
located at area sources. Boiler means an enclosed combustion
device having the primary purpose of recovering thermal
energy in the form of steam or hot water.” 76 Fed. Reg. at
15,557; see also 76 Fed. Reg. at 15,554 (“EPA is
promulgating national emission standards for control of
hazardous air pollutants from two area source categories:
Industrial boilers and commercial and institutional boilers.”).
Environmental Petitioners claim the general term “boiler”
necessarily encompasses temporary boilers: “[T]he category
of ‘boilers’ plainly includes temporary boilers, just as the
category of ‘courts’ includes federal courts, or the category of
‘dogs’ includes brown dogs.” No. 11-1141, Envtl. Pet’rs’
Reply Br. 7. According to Petitioners, then, sections 7412(c)
and 7412(d) of the CAA require the EPA to issue emission
standards for temporary boilers as well as “permanent”
boilers. See 42 U.S.C. § 7412(c)(2) (“For the categories and
subcategories the Administrator lists,
the Administrator shall establish emission standards . . . .”);
id. § 7412(d)(1) (“The Administrator shall promulgate
regulations establishing emission standards for each category
or subcategory of major sources and area sources of
hazardous air pollutants listed for regulation . . . .”).
To the extent Environmental Petitioners challenge as
unreasonable the EPA’s justifications for declining to set
emission standards for temporary boilers, they cannot prevail.
“Under arbitrary-and-capricious review, EPA’s
determinations are presumptively valid provided [they] meet[]
a minimum rationality standard.” Nat’l Ass’n for Surface
141
Finishing, 795 F.3d at 7 (internal quotation marks omitted).
So the question is whether the EPA offered a sufficiently
rational explanation for its exclusion of temporary boilers.
The EPA has done so here. First, contrary to Petitioners’
claims, temporary boilers were never considered an
inexorable part of the “industrial boiler” category section
7412 requires the EPA to regulate. While the EPA only
listed generic area source categories—“industrial boilers” and
“institutional/commercial boilers”—in its 1999 rulemaking, it
has since refined these broad categories pursuant to its
statutory authority. See National Air Toxics Program: The
Integrated Urban Strategy, 64 Fed. Reg. 38,706, 38,721 tbl.2
(July 19, 1999). In doing so, the EPA excluded several other
subgroups of boilers that might otherwise be read as falling
under one of the general boiler categories. See, e.g., 2013
Area Boilers Rule, 78 Fed. Reg. at 7,492 (excluding boilers
already regulated by other MACT standards); 2011 Proposed
Area Boilers Rule on Reconsideration, 76 Fed. Reg. at 80,539
(excluding electric and residential boilers as not part of either
source category). The EPA’s clarification that temporary
boilers were never considered part of the “industrial boilers”
category was simply another refinement, as contemplated by
the statute. See 42 U.S.C § 7412(e)(4) (precluding judicial
review until the EPA has issued its final emission standards
for a category or subcategory).
Second, as both the EPA and Industry Intervenors note,
the parallel rule for major source boilers has always explicitly
excluded temporary boilers from its “industrial boiler”
categorization. See 40 C.F.R. § 63.7491(j). The EPA thus
considered commenters’ requests to add a similar clarification
to the 2013 Area Boilers Rule and reasonably decided to do
so. See, e.g., American Forest & Paper Association,
Comments on Proposed Area Source Rule (AF&PA
142
Comments), EPA-HQ-OAR-2006-0790-1939 (Aug. 23,
2010), at 58 (No. 11-1141 J.A. 389). EPA explained this
choice in its proposed rule:
Owners and operators of regulated sources
have pointed out that temporary boilers are
small (less than 10 MMBtu/hr heat input) and
are generally owned and operated by
contractors, rather than the facility. As a
result, they are not included in the facility’s
operating permits because state and federal
CAA operating permit programs have
historically classified such units as
insignificant sources. The owners and
operators also noted that compliance with the
work practice requirements applicable to these
small boilers would be complicated because
they are typically located on site for less than a
year, but would be subject to biennial
management practice requirements. We agree
that the source category identified in subpart
JJJJJJ should specifically exclude these
temporary boilers because they have been
considered insignificant sources, and were not
included in the EPA’s analysis of the source
category.
2011 Proposed Area Boilers Rule on Reconsideration, 76 Fed.
Reg. at 80,535.
The unique nature of temporary boilers favors their
exclusion. These boilers tend to be rented for use on a
temporary basis and come in “shop-fabricated package
designs.” AF&PA Comments, at 58 (No. 11-1141 J.A. 389).
143
Temporary boilers also “typically only fire gas or liquid fossil
fuels (natural gas or distillate oil) which may be cleaner than
the boiler(s) they are temporarily replacing. In addition, these
units often do not have exhaust stacks that meet EPA Method
1 requirements for application of test methods.” Id.
Regardless, during the rulemaking, Environmental Petitioners
argued the EPA had “not explained why this is a distinction
that justifies differential treatment, let alone an exemption.”
See Area Boilers Rule—Responses to Comments, at 65. The
EPA responded by explaining that rather than having “created
a category or subcategory of ‘temporary boilers’ and then
exempted them from the standards,” the Agency never
“intend[ed] to regulate temporary boilers under the area
source standards” in the first place. See id. The EPA further
noted that, “[b]y their nature of being temporary, these boilers
operate in place of another non-temporary boiler while that
boiler is being constructed, replaced or repaired, in which
case we counted the non-temporary boiler as the one being
regulated.” Id. Finally, the Agency concluded regulation of
temporary boilers was not necessary to meet its statutory
emission requirements under sections 7412(c)(6) and
7412(c)(3) of the CAA. Id. In its final rule, the EPA
reiterated this explanation: “Similar to residential boilers, we
did not intend to regulate temporary boilers under the area
source standards because they are not part of either the
industrial boiler source category or the
commercial/institutional boiler source category.” 2013 Area
Boilers Rule, 78 Fed. Reg. at 7,491. The final regulation also
included a detailed explanation of how EPA decided on its
limited definition of “temporary boilers.” See id. at 7,499.
The evidence before the Agency supported its decision to
exclude temporary boilers. Indeed, the EPA “cogently
explain[ed]” why it exercised its discretion in this manner,
144
such that this court concludes the Agency’s choice “was the
product of reasoned decision making.” U.S. Telecom Ass’n v.
FCC, 227 F.3d 450, 460 (D.C. Cir. 2000). We therefore
uphold the EPA’s exclusion of temporary boilers from
regulation of area source boilers.
M. WORK-PRACTICE STANDARDS FOR COAL-FIRED
BOILERS
When setting emission limits for area sources, the EPA
enjoys greater discretion than when setting limits for major
sources. With respect to major sources, the EPA has to
promulgate MACT standards, see 42 U.S.C. § 7412(d)(2),
whereas for area sources the EPA can generally promulgate
more lenient GACT standards, see id. § 7412(d)(5). The
CAA, however, singles out seven particularly hazardous
pollutants that require stricter regulatory standards, even for
area sources.41 Under section 7412(c)(6), the EPA must “list
categories and subcategories of sources assuring that sources
accounting for not less than 90 per centum of the aggregate
emissions of each such pollutant” are regulated. The EPA
listed a variety of area sources under section 7412(c)(6) in its
1998 rulemaking based on their Hg and POM emissions. See
Proposed 2010 Area Boilers Rule, 75 Fed. Reg. at 31,898;
Source Category Listing for Section 112(d)(2) Rulemaking
Pursuant to Section 112(c)(6) Requirements, 63 Fed. Reg.
17,838, 17,849-50 (Apr. 10, 1998). But the Agency
subsequently refined that list and ultimately concluded only
41
These seven pollutants are: (i) alkylated lead compounds,
(ii) polycyclic organic matter (POM), (iii) hexachlorobenzene, (iv)
mercury (Hg), (v) polychlorinated biphenyls, (vi) 2,3,7,8-
tetrachlorodibenzofurans, and (vii) 2,3,7,8-tetrachlorodibenzo-p-
dioxin. See 42 U.S.C. § 7412(c)(6).
145
coal-fired area boilers needed to be listed to meet the statute’s
90 per cent emissions threshold. See 2010 Proposed Area
Boilers Rule, 75 Fed. Reg. at 31,898; see also No. 11-1141
EPA Br. 14 (“[T]he coal-fired subcategory is responsible for
over 82 percent of the mercury emissions from the [area
source] category in the inventory, even though it represents
only 2 percent of the boilers in the category.”).
Under section 7412(c)(6), the EPA was therefore
required to set either a MACT limit under section 7412(d)(2),
a health threshold under section 7412(d)(4), or a work-
practice standard under section 7412(h) for all coal-fired
boilers. The Agency chose to set MACT numerical emission
limits for Hg and CO42 at new and existing large coal-fired
boilers. 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488.
However, the EPA found it “technologically and
economically impracticable to apply [its] measurement
methodology to . . . small sources,” and so it chose to institute
a work-practice standard43 for all new and existing small coal-
fired boilers. Id. at 7,488-89. This work-practice standard
requires small coal-fired units to be periodically tuned up but
does not impose any numeric emission limit. See id. The
EPA similarly decided that, for large coal-fired boilers
undergoing a startup or a shutdown, a work-practice
standard—rather than a numeric emission standard—was
42
Because the EPA chose to regulate POM emissions
indirectly—by using CO emissions as a surrogate—the standards it
set under section 7412(c)(6) are for CO rather than POM. See 2013
Area Boilers Rule, 78 Fed. Reg. at 7,488, 7,503.
43
In their brief, Environmental Petitioners alternate between
the terms “operational standards” and “work-practice standards,”
both of which fall under section 7412(d)(2)(D). This opinion will
use “work-practice standards” for simplicity.
146
most practicable. See id. at 7,518 tbl.2 (requiring owners of
large “[e]xisting or new coal-fired” boilers to “[m]inimize the
boiler’s startup and shutdown periods and conduct startups
and shutdowns according to the manufacturer’s recommended
procedures”).
Environmental Petitioners challenge the EPA’s decision
to employ work-practice standards as a violation of
7412(d)(2)’s mandate to achieve the “maximum degree of
reduction in emissions.” We examine Petitioner’s statutory
argument step-by-step, as it hinges on the interplay between
several statutory provisions. First, section 7412(c)(6)—which
governs regulation of Hg and POM emissions—requires the
Administrator to regulate sources of these pollutants under
either section 7412(d)(2) or (d)(4). Section 7412(d)(4) allows
the Administrator to establish health-based emission
standards; it is not implicated here. Instead, the EPA decided
to regulate coal-fired boilers under section 7412(d)(2).
Section 7412(d)(2) instructs the Administrator to achieve “the
maximum degree of reduction in emissions of the hazardous
air pollutants . . . that the Administrator, taking into
consideration the cost of achieving such emissions reduction,
and any non-air quality health and environmental impacts and
energy requirements, determines is achievable for new or
existing sources.” The Administrator is authorized to use
several means to achieve this reduction including
implementing “design, equipment, work practice, or
operational standards . . . as provided in [section 7412(h)].”
42 U.S.C. § 7412(d)(2)(D). Section 7412(h)(1) states: “[I]f it
is not feasible in the judgment of the Administrator to
prescribe or enforce an emission standard for control of a
[pollutant], the Administrator may, in lieu thereof, promulgate
a . . . work-practice standard . . . , which in the
Administrator’s judgment is consistent with the provisions of
147
subsection (d) or (f) of this section.” Petitioners do not
dispute the EPA’s ability to set work-practice standards here;
they instead focus on section 7412(h)’s requirement that any
such standards be “consistent with” subsection (d)—which
requires the “maximum degree of reduction in emissions.”
According to Petitioners, the EPA’s decision to set these
particular work-practice standards fails at both Chevron steps.
With respect to Chevron step 1, Petitioners argue the
“EPA does not claim the operational standards [for coal-fired
boilers] are ‘consistent with the provisions of subsection (d)
or (f)’ of § 7412.” No. 11-1141 Envtl. Pet’rs’ Br. 33. In other
words, because these work-practice standards “do not even
purport” to be consistent with section 7412(d)’s mandate to
maximize reduction of emissions, “they are unlawful under
Chevron step one.” Id. Petitioners point to the EPA’s
specific findings to support this claim: for large coal-fired
boilers, the EPA found that mercury emissions could be
reduced by 75 to 82 per cent through the use of a fabric filter.
Id. But, according to Petitioners, the “EPA admits the tune-
up program [for small coal-fired boilers] will reduce
emissions by only one percent.” Id. And, with respect to
large coal-fired boilers undergoing startup or shutdown,
Petitioners argue the “EPA does not claim that ‘following the
manufacturer’s recommended procedures’ during startup and
shutdown will reduce emissions at all.” Id.
At the familiar Chevron step 1, the court must “first
examine the statute de novo, employing traditional tools of
statutory construction.” Nat’l Ass’n of Clean Air Agencies,
489 F.3d at 1228 (internal quotation marks omitted). If the
Congress’s intent is clear, then the Agency’s interpretation is
afforded no deference, and the court “must give effect to the
unambiguously expressed intent of Congress.” Id.
148
In this case, Environmental Petitioners place too much
emphasis on certain snippets of the statute without examining
the larger context. For one, Petitioners seem to argue that the
EPA must adopt work-practice standards that result in the
maximum possible reduction of emissions, without taking into
account any other considerations. But section 7412(d)(2)
itself belies this claim: it says the EPA must promulgate
standards that require “the maximum degree of reduction in
emissions . . . that the Administrator, taking into
consideration the cost of achieving such emission reduction,
and any non-air quality health and environmental impacts
and energy requirements, determines is achievable.” 42
U.S.C. § 7412(d)(2) (emphasis added). This portion of the
statute explicitly defers to the Administrator’s judgment
regarding a standard’s “achievability,” even though it directs
him to consider particular factors in making that assessment.
Section 7412(h) similarly requires the Administrator to adopt
a work-practice standard that in his judgment is consistent
with section 7412(d)(2)’s mandate. We therefore cannot
accept Petitioners’ suggestion that Congress unambiguously
required the EPA to adopt standards that result in the
maximum reduction of emissions that is technologically
feasible.
Environmental Petitioners’ challenge to these work-
practice standards as unreasonable under Chevron step 2 and
arbitrary under State Farm presents a closer call. With
respect to Chevron step 2, the court must “uphold an agency’s
interpretation if it is reasonable.” Ariz. Pub. Serv. Co. v. EPA,
211 F.3d 1280, 1287 (D.C. Cir. 2000). And, “even where
EPA’s construction satisfies Chevron, [the court] still must
ensure that its action is not otherwise arbitrary and capricious.
The arbitrary and capricious standard is ‘[h]ighly deferential,’
and it ‘presumes the validity of agency action.’” Nat’l Ass’n
149
of Clean Air Agencies, 489 F.3d at 1228 (citations omitted).
As long as an agency has “considered the relevant factors and
articulated ‘a rational connection between the facts found and
the choice made,’” then its decision must be upheld. Allied
Local & Reg’l Mfrs. Caucus, 215 F.3d at 68.
Petitioners mount both a “facial” and a substantive
challenge to the EPA’s rationale for adopting work-practice
standards. First, Petitioners claim the EPA’s decision is
arbitrary because it fails “to reconcile its approach with the
statutory requirement [of section 7412(d)(2)].” No. 11-1141
Envtl. Pet’rs’ Br. 34. Specifically, Petitioners insist the EPA
must explicitly state somewhere that these particular work-
practice standards are “consistent” with section 7412(d)(2).
See id. at 33-34. Otherwise, Petitioners contend, the Court
must simply “assume that the Agency heeded § 7412(h)’s
‘consistent with’ requirement, notwithstanding the EPA’s
failure to acknowledge and apply that requirement in the
record.” No. 11-1141 Envtl. Pet’rs’ Reply Br. 10.
The Agency responds that, “by identifying the tune-up
and startup/shutdown requirements as ‘work practices,’ [it] is
stating that those standards are issued under section 7412(h)
and are consistent with the requirements of section 7412(d)
(i.e., MACT).” No. 11-1141, EPA Br. 71. The EPA did
acknowledge it has an obligation to maximize emission
reductions under section 7412(d)(2) when promulgating
work-practice standards. See 2011 Area Boilers Rule, 76 Fed.
Reg. at 15,568 (“CAA section 112(h) authorizes the
Administrator to promulgate [a work-practice standard]
consistent with the provisions of CAA sections 112(d) or
(f) . . . .”). However, Petitioners are correct that the Agency
did not make a finding on the record that these work-practice
150
standards would achieve the highest emissions reduction
possible.
But the lack of an explicit statement does not
automatically condemn this portion of the rule. See Bowman,
419 U.S. at 286 (“[W]e will uphold a decision of less than
ideal clarity if the agency’s path may reasonably be
discerned.”). The Petitioners offer no support for their
contention that the EPA must make an express finding that its
standards are “consistent” with section 7412(d)(2). Nor does
our conclusion requires us to merely “assume” that the
Agency’s actions comport with section 7412(d)(2). Instead,
as we usually do when presented with such arguments, we
review the rulemaking record to determine whether the
justifications the EPA offered for adopting these work
practices standards were permissible.
1. Small Coal-Fired Boilers
First, with respect to small coal-fired boilers, the EPA
determined that a biennial tune-up requirement would best
comply with section 7412(h)’s requirements. As a starting
point, the EPA surveyed a sample of state regulations
mandating various work-practice standards for small coal-
fired boilers; the Agency found that ten states required tune-
ups, two required periodic inspections, one required operator
training, and one required operation in accordance with
manufacturer specifications. 2011 Area Boilers Rule, 76 Fed.
Reg. at 15,573-74. The EPA thus concluded that tune-ups
were the most typical work-practice standard employed for
this type of boiler. Id. The Agency also found that regular
tune-ups could lower HAP emissions by increasing the
efficiency of small coal-fired boilers. See id. at 15,575 (“A
tune-up performed to the manufacturer’s specifications would
151
ensure the highest energy efficiency and reduce fuel usage,
which will ultimately reduce HAP emissions.”); see also 2010
Proposed Area Boilers Rule, 75 Fed. Reg. at 31,908 (“A
boiler tune-up provides potential savings from energy
efficiency improvements and pollution prevention. . . . In
addition, the cost of a boiler tune-up appears minimal
compared to the cost for testing and monitoring to
demonstrate compliance with an emission limit.”).44
The EPA elected to implement a work-practice standard
because the typical method used to measure emissions of Hg
and CO could not be used to sample emissions from stacks
with small diameters (less than 12 inches). See 2011 Area
Boilers Rule, 76 Fed. Reg. at 15,568. Because many small
coal-fired boilers have stacks with diameters below 12
inches—and because many of these boilers “do not currently
have sampling ports or a platform for accessing the exhaust
stack”—the Agency concluded the cost of testing and
monitoring these small boilers would “present an excessive
burden for smaller sources.” Id. The Agency’s consideration
of cost effectiveness is particularly appropriate in this context
because the “vast majority” of area source boilers are
“generally owned and operated by small entities,” which
would be disproportionately burdened by a numeric emissions
limit. See Fact Sheet: Final Adjustments to the Air Toxics
Standards for Industrial, Commercial, and Institutional
Boilers at Area Source Facilities, 1, 2 (No. 11-1141 J.A. 684-
44
The EPA’s discussion of boiler tune-up advantages occurs
mainly in the context of its decision to select a GACT standard
rather than any numeric emission standards for certain boilers.
Environmental Petitioners challenge this decision on similar
grounds, see supra § IV.H. But the benefits of periodic tune-ups
also apply to the coal-fired boilers at issue here.
152
85); see also 2010 Proposed Area Boilers Rule, 75 Fed. Reg.
at 31,906 (“The results of the analysis indicate that total
compliance costs exceed 3 percent (and can reach as high as
19 percent) of the average firm revenues for 79 percent of the
facilities.”).
Environmental Petitioners counter that while tune-ups
may minimally reduce HAP emissions, they do not maximize
this reduction per section 7412(d)(2)’s mandate.45 But
Petitioners view section 7412(d) too myopically; under that
section, the Administrator is empowered to adopt standards
that “tak[e] into consideration the cost of achieving such
emission reduction.” 42 U.S.C. § 7412(d)(2). The EPA here
examined the costs of imposing a numeric emission standard
on small coal-fired boilers and found that option “not
feasible” due to high costs and monitoring difficulties,
considerations equally permissible under section 7412(d)(2).
Petitioners argue that requiring the use of a fabric filter would
have resulted in greater reductions, but they are unable to
point to any small coal-fired boiler that currently uses such a
filter. See 2010 Proposed Area Boilers Rule, 75 Fed. Reg. at
31,906 (“For existing [small] area source boilers, the only
work practice being used that potentially controls mercury
45
Petitioners also point to a comment they made in the record
arguing that a tune-up standard “would not achieve emission
reductions that are consistent with the definition of MACT,” and
urge that the EPA never addressed these concerns. See National
Association of Clean Air Agencies, Comments on EPA Proposals
for Regulation of Hazardous Air Pollutants (HAPS), EPA-HQ-
OAR-2006-0790, EPA-HQ-OAR-2002-0058, EPA-HQ-OAR-
2003-0119 (Aug. 23, 2010), at 21-22 (No. 11-1141 J.A. 417-18).
But this comment was specifically addressed to gas-fired boilers,
and it is inapposite to the EPA’s consideration of standards for
coal-fired boilers.
153
and POM emissions is a boiler tune-up.”). Evidence before
the agency in fact indicated that the best performing small
coal boilers for POM emission use no add-on controls. See
Memorandum from Amanda Singleton & Brandon Long,
Eastern Research Group, to Jim Eddinger, EPA
(MACT/GACT Mem.), App. E-2a (No. 11-1141 J.A. 540). It
was therefore reasonable for the EPA, when considering
costs, to conclude that biennial tune-ups would allow for the
maximum “achievable” reduction in emissions.
Petitioners’ most compelling argument involves the
EPA’s lack of data on small coal-fired boilers. As they point
out, the EPA’s summary of its 2008 combustion survey makes
no mention of any small coal-fired boilers. See
MACT/GACT Mem., App. D-3, tbl.1 (No. 11-1141 J.A. 523).
And the EPA never directly addressed whether control
technologies, such as fabric filters, were useable by small
boilers; “[t]he only claim EPA made in the record is that tune-
ups are the most effective option that [small] coal-fired
boilers . . . are currently using, not that tune-ups yield the
maximum reduction ‘achievable.’” No. 11-1141 Envtl.
Pet’rs’ Reply Br. 11. In Sierra Club II, this court agreed with
Sierra Club’s challenge to the EPA’s use of a work-practice
standard instead of an emission floor because the “EPA never
determined that measuring emissions from ceramics kilns was
impracticable; it determined only that it lacked emissions data
from ceramics kilns. EPA thus had no basis under section
7412(h) for using work practice standards.” 479 F.3d at 884.
That context is somewhat distinguishable, given that the
statute there explicitly required the EPA to make a
“feasibility” finding (as discussed above), but it could be
argued that the EPA here lacked the data to determine
whether tune-ups were “consistent with” section 7412(d)(2).
154
Ultimately, though, the high level of deference afforded
the EPA counsels in favor of upholding this work-practice
standard. Although the EPA did not explicitly state that tune-
ups were the best option to reduce emissions while still
“considering costs,” this finding can be inferred from the
record as a whole. For instance, the EPA found that “[n]one
of the States for which we have an inventory have an
applicable emissions limit” for small coal-fired boilers, except
New Jersey, which actually has tune-ups as its work-practice
standard. 2010 Proposed Area Boilers Rule, 75 Fed. Reg. at
31,909. Based on these findings, it can reasonably be inferred
that—given the prevalence of these small boilers—at least a
few would be using a control technology if that were
technologically or economically feasible. Because numeric
emissions cannot easily be measured from these smaller
sources and the costs of outfitting them with such technology
would be cost prohibitive, the EPA’s choice of tune-ups as the
work-practice standard is sufficiently reasonable to uphold
under both Chevron step 2 and State Farm. The Agency’s
choice is consistent with section 7412(h)’s “feasibility”
requirement and with section 7412(d)(2)’s instruction to
maximize emission reductions while also considering costs.
2. Large Coal-Fired Boilers Undergoing Startup or
Shutdown
The record for large coal-fired boilers undergoing startup
or shutdown is less extensive but again the EPA’s
determination is reasonable. While large coal-fired boilers
are required to meet numeric emission standards during
“normal” operations, the EPA adopted a work-practice
standard for the temporary periods of startup and shutdown.
See 40 C.F.R. § 63.11214 (“[M]inimize the boiler’s startup
and shutdown periods following the manufacturer’s
155
recommended procedures, if available. If manufacturer’s
recommended procedures are not available, you must follow
recommended procedures for a unit of similar design for
which manufacturer’s recommended procedures are
available.”). Environmental Petitioners contend that the EPA
never stated this practice would reduce emissions at all, and
therefore it has not met its burden under section 7412(d)(2).
But we have already explained that no express finding of
consistency with section 7412(d)(2) need be made. Here, the
record suggests that the work-practice standard the Agency
chose would reduce emissions, and we therefore can
“reasonably [] discern[]” the Agency’s path. Bowman, 419
U.S. at 286. Specifically, the EPA explained that requiring
boilers to operate in startup and shutdown mode for
“sufficient time to conduct the required test runs [to impose
numeric standards] could result in higher emissions than
would otherwise occur.” 2011 Major Boilers Rule, 76 Fed.
Reg. at 15,642. Industry stakeholders also pointed out that “it
is very common . . . for certain control devices to be out of
operation during periods of start-up due to the nature of the
equipment.” See American Chemistry Council, Comments
on EPA’s Proposed Rule for National Emission Standards for
Hazardous Air Pollutants for Area Sources, EPA-HQ-OAR-
2006-0790 (Aug. 23, 2010), at 31 (No. 11-1141 J.A. 386).
Because the control technology is temporarily offline, “it is
likely that emissions will exceed the standards proposed
[during that time period].” Id. A work-practice standard that
requires facilities to minimize the time their boilers spend in
startup or shutdown thus seems calculated to maximally
reduce emissions during those periods—and Petitioners fail to
provide any viable alternative. We therefore conclude the
EPA’s decision to adopt these work-practice standards for
large coal-fired boilers during startup and shutdown was
reasonable.
156
V. CONCLUSION
For the foregoing reasons, we grant the petitions in part
and deny them in part. Specifically, we vacate the MACT
standards for all major boiler subcategories that would have
been affected had the EPA considered all sources included in
the subcategories, as explained at supra § IV.B. We
also remand, without vacatur, to the EPA to: (1) adequately
explain how CO acts as a reasonable surrogate for non-
dioxin/furan organic HAPs; (2) set emission standards for
cyclonic burn barrels; (3) determine whether burn-off ovens,
soil treatment units, and space heaters are CISWI units and, if
so, to set standards for those types of units; (4) adequately
explain the exclusion of synthetic boilers from Title V’s
permitting requirements; and (5) adequately explain the
choice of GACT standards over MACT standards for non-Hg
metals.
So ordered.