United States Court of Appeals
         FOR THE DISTRICT OF COLUMBIA CIRCUIT




Argued November 9, 2017                 Decided July 6, 2018

                        No. 15-1487

SIERRA CLUB AND NATURAL RESOURCES DEFENSE COUNCIL,
                    PETITIONERS

                             v.

    ENVIRONMENTAL PROTECTION AGENCY AND E. SCOTT
                      PRUITT,
                   RESPONDENTS

           BRICK INDUSTRY ASSOCIATION, ET AL.,
                     INTERVENORS



   Consolidated with 15-1492, 15-1493, 15-1496, 16-1179



      On Petitions for Review of a Final Action of the
      United States Environmental Protection Agency



       James S. Pew argued the cause and filed the briefs for
Environmental petitioners.
                               2
         William L. Wehrum, Felicia H. Barnes, Todd E.
Palmer, Valerie L. Green, Cameron F. Field, Howard L.
Gilberg, Jean M. Flores, Jeffrey S. Longsworth, Charles M.
Denton, and Roger J. Marzulla were on the joint briefs for
Industry petitioners.

          Kate R. Bowers, Attorney, U.S. Department of
Justice, argued the cause for respondents. With her on the brief
were Sonya J. Shea, Attorney, and Sonja L. Rodman and Scott
J. Jordan, Attorneys, U.S. Environmental Protection Agency.

        Felicia H. Barnes argued the cause for Industry
respondent-intervenors. With her on the joint brief were
William L. Wehrum Jr., Todd E. Palmer, Valerie L. Green,
Cameron F. Field, Jeffrey S. Longsworth, and Charles M.
Denton.

        James S. Pew was on the brief for Environmental
respondent-intervenors.

       Before: ROGERS and MILLETT, Circuit Judges, and
SENTELLE, Senior Circuit Judge.

       Opinion for the Court filed by Senior Circuit Judge
SENTELLE.

          SENTELLE, Senior Circuit Judge: In this consolidated
proceeding, we consider petitions for review of an
Environmental Protection Agency (“EPA”) final rule entitled
“NESHAP for Brick and Structural Clay Products
Manufacturing; and NESHAP for Clay Ceramics
Manufacturing,” 80 Fed. Reg. 65,470 (Oct. 26, 2015)
(“Brick/Clay Rule”), and its partial denial of reconsideration of
that rule, 81 Fed. Reg. 31, 234 (May 18, 2016).
                                3
        In Case No. 15-1487, petitioners Sierra Club and
Natural      Resources   Defense      Council     (collectively,
“Environmental Petitioners”) contend that the EPA erred in its
use of health-based standards for acid gas emissions, failed to
properly explain its methodology in setting maximum
achievable control technology-based standards, and
improperly allowed brick plants to meet alternative emissions
floors. In Case Nos. 15-1492, 15-1493, 15-1496, and 16-1179,
the Brick Industry Association, the Tile Council of North
America, Inc., and the Kohler Company (collectively,
“Industry Petitioners”) contend that the EPA made multiple
errors in its methodology in the Brick/Clay Rule.

        Finally, the EPA moved to sever and hold in abeyance
the Industry Petitioners’ petition for review while it reconsiders
the Brick/Clay Rule. Industry Petitioners supported the
motion; the Environmental Petitioners opposed the motion to
hold the entire case in abeyance but not EPA’s motion to sever
and hold in abeyance the Industry Petitioners’ petition. See
Unopposed Mot. at 2, Doc. No. 1703072 (Filed Nov. 3, 2017).
We deferred consideration of the motion pending oral
argument. We now deny the motion and consider this case on
its merits. For the reasons stated below, we deny Industry
Petitioners’ petition for review and grant in part that of the
Environmental Petitioners and remand the Brick/Clay Rule to
the agency for further proceedings consistent with this opinion.

I.   Background

       Under the Clean Air Act (“CAA”), the EPA regulates
all major and area sources of hazardous air pollutants. 42
U.S.C. § 7412(d)(1). There are 189 hazardous air pollutants
subject to regulation under the CAA, including hydrogen
chloride, hydrogen fluoride, chlorine, and heavy metals such as
mercury. 

During the regulatory process, the
                               4
EPA identifies categories of sources that generate hazardous
air pollutants, and then sets emissions limits for each major
source category. 

(d)(1).

        The EPA found that kilns emit hazardous acid gases,
primarily hydrogen chloride (HCl), hydrogen fluoride (HF),
and chlorine (Cl2). 80 Fed. Reg. at 65,473. Each of these acid
gas pollutants causes health issues, such as asthma, respiratory
problems, skin irritation, burns, low blood pressure, and, in
severe cases, death. Regulatory Impact Analysis: Final Brick
and Structural Clay Products NESHAP at 4-28 to 4-30, Docket
# EPA-HQ-OAR-2013-0291-0665 (July 2015). Kilns also
emit heavy metal pollutants, such as mercury, lead, arsenic, and
other particulate matter. 80 Fed. Reg. at 65,473. Heavy metal
pollutants also cause health issues, such as neurological
damage, respiratory harm, and cancer, and threaten the natural
environment. Regulatory Impact Analysis at 4-27 to 4-33,
Docket # EPA-HQ-OAR-2013-0291-0665.

        Once the EPA listed brick and ceramic kilns as a major
source of hazardous air pollutants, it was required to regulate
them. 42 U.S.C. § 7412(e)(1)(E). In 2003, the EPA first
undertook the regulation of kiln emissions under the CAA. 68
Fed. Reg. 26,690 (May 16, 2003) (“2003 Rule”). In 2007, this
Court held that the 2003 Rule did not comply with the CAA
and vacated it. Sierra Club v. EPA, 

, 876 (D.C.
Cir. 2007) (per curiam) (“Sierra I”). In 2012, Sierra Club
brought suit against the EPA to force it to promulgate
regulations to replace the vacated 2003 Rule. Sierra Club v.
EPA, 

, 301 (D.D.C. 2012). In 2014, after
years of data collection, the EPA proposed a new rule to replace
the vacated standards for kiln emissions. 79 Fed. Reg. 75,622
(Dec. 18, 2014). On October 26, 2015, the EPA published the
final Brick/Clay Rule. 80 Fed. Reg. at 65,470. Industry
Petitioners then submitted a petition for reconsideration of the
                               5
Brick/Clay Rule, which the EPA denied in relevant part. 81
Fed. Reg. 31,234 (May 18, 2016).

        The Brick/Clay Rule and the EPA’s partial denial of
reconsideration are the subjects of the petitions in this
consolidated case. The Brick/Clay Rule applies to brick, clay,
and tile kilns. The emissions standards for brick and structural
clay products, such as clay pipe and roof tile, are published in
the Brick/Clay Rule and codified at 40 C.F.R. pt. 63, subpt.
JJJJJ. 80 Fed. Reg. at 65,520. The emissions standards for clay
ceramic products, such as pressed tile and sanitaryware (e.g.,
toilets and sinks), are published within the Brick/Clay Rule,
titled the National Emissions Standards for Hazardous Air
Pollutants for Clay Ceramic Manufacturing (“Clay
NESHAP”), codified at 40 C.F.R. pt. 63, subpt. KKKKK. 80
Fed. Reg. at 65,543.

        Now, the Brick Industry Association petitions for
review of the Brick/Clay Rule as applicable to the brick
industry and for review of the EPA’s denial of reconsideration
of the Brick/Clay Rule. The Kohler Company petitions for
review of the Clay NESHAP. The Tile Council of North
America petitions for review of certain provisions of the Clay
NESHAP applicable to the subcategory of ceramic tile
manufacturing (“Tile NESHAP”). Finally, the Environmental
Petitioners petition for review of the Brick/Clay Rule as
applicable to all the Industry Petitioners. Industry Petitioners
intervened on behalf of the EPA in response to the
Environmental Petitioners’ petition; the Environmental
Petitioners intervened on behalf of the EPA in response to the
Industry Petitioners’ petitions.
                                6
       A. Clean Air Act Framework for Emissions
       Standards

        The CAA governs the setting of emissions standards
using technology and health thresholds. 42 U.S.C. § 7412(d).
When these types of limits are not feasible, the EPA may
substitute alternative methods to limit emissions such as
operational controls. 

           1. Maximum Achievable Control Technology
           Emissions Standards

        The 1990 amendments to the CAA directed the EPA to
issue emissions limits using technology-based standards, called
“Maximum Achievable Control Technology” (“MACT”).
MACT standards require the “maximum degree of [emissions]
reductions” that the EPA “determines is achievable.” 42
U.S.C. § 7412(d)(2); see Sierra 

. The
emissions standards for new sources must be no “less stringent
than the emission control that is achieved in practice by the best
controlled similar source.” 42 U.S.C. § 7412(d)(3). Congress
adopted this MACT-based scheme because the EPA’s previous
use of health-based standards had been “problematic,” because
of uncertainty over which pollutants pose a health risk. NRDC
v. EPA, 

, 1079 (D.C. Cir. 2008).

        In establishing a MACT emissions standard, the EPA
defines a minimum stringency requirement, or “floor,” based
on emissions levels achieved by existing sources. 42 U.S.C.
§ 7412(d)(3)(A), (B); Sierra 

. For categories
and subcategories of existing emissions sources composed of
thirty or more individual sources, the EPA sets the MACT floor
using “the average emission limitation achieved by the best
performing 12 percent of the existing sources (for which the
Administrator has emissions information).”           42 U.S.C.
                              7
§ 7412(d)(3)(A). If there are fewer than thirty individual
sources, the EPA sets the MACT floor using “the average
emission limitation achieved by the best performing 5
sources.” 

The EPA may set more
stringent standards than the MACT floor if the more stringent
standard is achievable considering cost and other factors. 

U.S. Sugar Corp. v. EPA, 

, 594-95
(D.C. Cir. 2016) (per curiam) (“U.S. Sugar”), cert. denied sub
nom. American Mun. Power v. EPA, 

(2017).

           2. Alternatives to MACT Standards

        Alternatively, the EPA may use a health threshold
rather than MACT standards for “pollutants for which a health
threshold has been established.” 42 U.S.C. § 7412(d)(4); see
U.S. 

. Such a health-based standard
must include an “ample margin of safety.” 42 U.S.C.
§ 7412(d)(4). Additionally, if the EPA determines it is “not
feasible” to prescribe either a health- or technology-based
emissions standard the agency may promulgate a design,
equipment, work-practice, or operational standard.        

       B. Brick/Clay Rule

       In the Brick/Clay Rule, the EPA set MACT standards
to regulate heavy metal emissions from kilns and health
thresholds to regulate acid gas emissions. 80 Fed. Reg. at
65,471.

           1. Heavy Metal Emissions

       The EPA chose to regulate heavy metal emissions in the
Brick/Clay Rule using MACT standards under § 7412(d)(1).
80 Fed. Reg. at 65,471. In setting the brick MACT floor, the
                                8
EPA set separate standards for particulate matter (which it used
as a surrogate for nonmercury hazardous metals) and mercury,
with subcategories for large tunnel and small tunnel brick kilns.

Additionally, the EPA provided “alternative
equivalent limits” for heavy metal emissions from brick kilns.

65,485. There were more than thirty individual
sources for each category, so the best twelve percent of
performers were used to set the brick MACT floor. 

        The Clay NESHAP sets separate MACT standards for
subcategories of floor tile, wall tile, and sanitaryware that are
heavy metal emissions sources. 

In response to
an EPA information request, Kohler activated a
decommissioned emissions scrubber at its South Carolina Kiln
10 (“Kiln 10”), and that data was used as one of the best
performing sources in setting the corresponding MACT floor.
See 

see also 81 Fed. Reg. at 31,235. The Tile
NESHAP regulates emissions for a subcategory of ceramic tile
plants and has separate dioxin/furan and mercury emissions
standards. 80 Fed. Reg. at 65,554. Because fewer than thirty
sources exist for each category of clay and tile kilns, the
average of the five best performers was used to set the clay and
tile MACTs. See 

n.102, 65,510.

        Both Industry and Environmental Petitioners challenge
the EPA’s methods of setting various MACT floors as unlawful
and arbitrary. We will address the background of those
methods in further detail with respect to the individual petitions
for review.

           2. Acid Gas Emissions

      The EPA regulates hazardous acid gas emissions for
new and existing brick tunnel kilns and all ceramics kilns
                               9
except for sanitaryware shuttle kilns in the Brick/Clay Rule
using health thresholds under § 7412(d)(4). 80 Fed. Reg. at
65,471, 65,474, 65,478. Environmental Petitioners contend
that EPA’s choice of health thresholds over a MACT standard
means that kiln facilities largely will not need to add more
pollution controls. Environmental Petitioners challenge EPA’s
use of health thresholds and the methods it used to set the
thresholds as contrary to the CAA. We will address the
background of the EPA’s reasoning in setting the health
thresholds in the context of the Environmental Petitioners’
petition for review.

II.   Standard of Review

        EPA’s final rule is subject to judicial review under 42
U.S.C. § 7607(b)(1). We will set aside an EPA action under
the CAA if it is “arbitrary, capricious, an abuse of discretion,
or otherwise not in accordance with law.” 42 U.S.C.
§ 7607(d)(9)(A). We review EPA’s interpretation of the CAA
under the two-step framework of Chevron U.S.A., Inc. v.
NRDC, Inc., 

(1984). We first determine if
Congress has “directly spoken to the precise question at issue.”

If so, then we must “give effect to the
unambiguously expressed intent of Congress.” 

If,
however, “the statute is silent or ambiguous with respect to the
specific issue,” we defer to the EPA’s interpretation of the
CAA so long as it “is based on a permissible construction of
the statute.” 

standards, we review in turn the
Environmental Petitioners’ and the Industry Petitioners’
petitions for review.
                                10

III. Environmental Petitioners’ Petition

        The Environmental Petitioners petition this Court to
review the Brick/Clay Rule, claiming that it is unlawful and
arbitrary. They argue that the EPA erred in its use of health
thresholds to regulate acid gas pollutants, in its use of the upper
prediction limit method in setting MACT floors, and in its
provision of alternative MACT floors for brick kilns.

        A. Health Thresholds for Acid Gas Emissions

        The Environmental Petitioners mount a three-part
challenge to the EPA’s use of health thresholds to regulate acid
gas emissions. First, the Environmental Petitioners argue that
the EPA violated the CAA by concluding that the acid gas
pollutants do not pose a cancer risk. Second, Environmental
Petitioners argue that the EPA violated the CAA and was
arbitrary in its determination of the health thresholds for the
acid gas pollutants’ noncarcinogenic risks. Finally, they
contend that the EPA violated the CAA by failing to include an
ample margin of safety in the health threshold.

            1. Cancer Risks

       Petitioners contend that the EPA violated the CAA by
concluding that the acid gas pollutants do not pose a cancer
risk. Under the CAA, the EPA may implement a health
threshold-based emissions limit for “pollutants for which a
health threshold has been established.”           42 U.S.C.
§ 7412(d)(4). The EPA and Environmental Petitioners agree,
though for differing reasons, that the EPA should not
implement health thresholds in lieu of MACT standards for
carcinogenic pollutants.
                              11
        Under Chevron step one, Environmental Petitioners
argue that the EPA acted contrary to the clear meaning of the
word “established” in 42 U.S.C. § 7412(d)(4)—“to put beyond
a reasonable doubt: prove,” WEBSTER’S NEW COLLEGIATE
DICTIONARY 284 (7th ed. 1971)—by concluding based on mere
lack of evidence that hydrogen chloride, hydrogen fluoride, and
chlorine do not cause cancer. Likewise, the Environmental
Petitioners argue that the term “health threshold” in
§ 7412(d)(4) unambiguously requires the EPA to identify the
specific amount of a pollutant that lacks any health risk.
Because the EPA did not rely on conclusive proof that the acid
gas pollutants are noncarcinogenic, the Environmental
Petitioners assert that the EPA violated the plain language of
the CAA requiring it to find an “established” “health threshold”
when formulating health-based standards under § 7412(d)(4).

        We cannot resolve this case at step one as
Environmental Petitioners urge. In other contexts, we have
held that the “EPA is not obligated to conclusively resolve
every scientific uncertainty before it issues regulation.” White
Stallion Energy Ctr., LLC v. EPA, 

, 1245 (D.C.
Cir. 2014) (per curiam), rev’d on other grounds by Michigan v.
EPA, 

(2015). The statutory term “established”
does not unambiguously require that the EPA prove its
scientific conclusions beyond all possible doubt. Nor does the
term “health threshold” require that the EPA find a specific
threshold that lacks uncertainty. With respect to scientific
conclusions, “established” and “health threshold” are
ambiguous terms and we give deference to the EPA to the
extent its interpretations fall within the bounds of
reasonableness.

        The Environmental Petitioners separately argue that the
EPA acted unreasonably in concluding that it was “established”
that the acid gases present no cancer risk. The Environmental
                                12
Petitioners contend that the EPA’s reliance on the lack of
expert data on the cancer risks of these pollutants is insufficient
to demonstrate that their non-carcinogenic nature was
“established.”

        The EPA counters that it is not required to prove a
negative and that its interpretation of “established” is based on
its expert judgment that it need not find “universal agreement”
that the acid gases pose no cancer risk. To reach its conclusion,
the EPA reviewed toxicity assessments in several databases
and opinions from scientific bodies and found that none
classified the three acid gas pollutants as carcinogens or as
“suggestive of the potential to be carcinogenic.” 80 Fed. Reg.
at 65,488. Further, the EPA searched for existing studies
regarding the cancer risk of the acid gases and found that,
though the studies are limited in scope, none proved any
carcinogenic potential. 

        The EPA has not provided a sufficient record to
determine that there is no cancer risk. The EPA noted that for
hydrogen chloride “[l]ittle research has been conducted on its
carcinogenicity.” 79 Fed. Reg. at 75,639-41; see also 80 Fed.
Reg at 65,488-90. In addressing the cancer risk of chlorine, the
EPA reviewed limited studies of workers in the chemical
industry. 80 Fed. Reg. at 65,489. In addition to reviewing
limited studies of workers, EPA also reviewed studies
involving rodents and non-human primates. 

to hydrogen fluoride, the EPA observed there was a “limited
number of studies investigating the carcinogenic potential of
[Hydrogen Fluoride]” and “[t]hese studies are unreliable on the
issue of possible carcinogenicity of [Hydrogen Fluoride]
and/or fluorides, in general, because of many confounding
factors . . . and because no breakdown was done by type of
fluoride exposure.” 

This is not merely a situation
in which the EPA relies on the results of scientific studies that
                               13
were unable to demonstrate a cancer risk to “prove a negative.”
Rather, the EPA relies on the lack of any significant studies.
The EPA failed to explain how it was able to reach the
conclusion that the acid gas pollutants do not pose a cancer risk
despite the admittedly limited body of research. Because of
this acknowledged lack of evidence, the EPA did not base its
findings on substantial evidence.

        Therefore, the EPA acted unreasonably by concluding
that it is “established” that the acid gas pollutants pose no
cancer risk. Accordingly, we grant the Environmental
Petitioners’ petition on this issue.

            2. Noncarcinogenic Risks

        Second, Environmental Petitioners argue that the EPA
violated the CAA and was unreasoned and arbitrary by relying
on low confidence evidence for the health risks of hydrogen
chloride, by declining to use the California Environmental
Protection Agency’s reference concentrations for hydrogen
chloride, and by not shielding the public from acute exposure
to hydrogen fluoride.

               a.   Hydrogen Chloride

        With respect to the noncarcinogenic health threshold
for hydrogen chloride, Environmental Petitioners argue that the
EPA violated the CAA’s requirement that the health threshold
be “established” by relying on a low confidence reference
concentration value in setting the emissions limit. The
Environmental Petitioners assert that any deference to the
EPA’s interpretation of “established” does not extend to
allowing the EPA to use a single low-confidence, low-quality
risk assessment to conclude that a threshold is “established.”
Further, the California EPA uses a more conservative reference
                               14
concentration for the amount of hydrogen chloride that does
not pose a health risk, and the EPA failed to explain why it used
a different reference concentration.

        The EPA explained that the “low confidence” label
means that it has low confidence in the derivation of the
hydrogen chloride reference concentration below which no
adverse health risks were expected to occur because the
existing data were incomplete and the reference concentration
is subject to change as new data are developed. 80 Fed. Reg.
at 65,489-90. Further, the EPA asserted that though reference
concentrations are assigned high, medium, and low confidence
values based on the supporting database, even a low confidence
value is reliable enough for regulatory use. 

The
EPA stated that it did not use the California EPA’s hydrogen
chloride reference concentration because of its general
preference “favoring EPA benchmarks.” 

        The EPA’s statement that “low confidence” reference
concentrations are suitable for regulatory purposes lacks any
supporting reasoning. The EPA did not explain how the health
threshold could be established if low confidence reference
concentrations are subject to change. This lack of reasoning is
compounded by the EPA’s failure to explain why low
confidence data were enough to demonstrate an “established”
limit less strict than the California EPA threshold. Because the
EPA acted unreasonably in finding that the noncarcinogenic
health threshold for hydrogen chloride was established, we
grant the Environmental Petitioners’ petition as to this issue.

               b. Hydrogen Fluoride

       With respect to the hydrogen fluoride health threshold,
the Environmental Petitioners argue that the emissions limits
do not protect the public from the health effects of acute
                              15
exposure, as required for health thresholds. See 42 U.S.C.
§ 7412(d)(4). In both the brick and ceramics risk assessments,
the EPA found that some kilns had the potential to exceed the
health threshold for acute exposures, but concluded there was
a “low potential” for health effects. The EPA calculated a
hazard quotient of two for these acute exposures, where a
hazard quotient less than one indicates low potential for acute
health risk. 80 Fed. Reg. at 65,503.

        EPA argues that the hazard quotient calculation is
conservative. First, the hazard quotient calculation uses the
conservative California EPA acute reference concentration to
calculate the exposure level and if it had applied its own
reference level, there would have been a hazard quotient less
than one. However, the EPA does not offer a calculation or
explanation to confirm this assertion. Second, the EPA’s acute
exposure scenario conservatively assumes that someone is
present at a time with both worst-case emissions and worst-
case weather conditions. Finally, the EPA reasoned that it was
unlikely a facility would emit only hydrogen fluoride at the
maximum permitted concentration rather than a combination
of other gases.

        EPA’s reasoning is flawed. As we held in the hydrogen
chloride analysis, the EPA failed to explain why it would be
appropriate to use a less conservative standard than the
California EPA level. Thus, even assuming that use of EPA’s
reference level would have resulted in below-one hazard
quotients, EPA fails to show that the hazard quotient
calculation performed protects from acute exposure. Likewise,
the EPA failed to explain why “no adverse health effects are
expected to occur,” 80 Fed. Reg. at 65,491, in the worst-case
scenario of peak emissions during worst-case weather. The
EPA does not adequately explain how a hydrogen fluoride
emissions limit based on an “established health threshold” can
                              16
permit potential health risks from acute exposure. Therefore,
we grant the Environmental Petitioners’ petition as to this
issue.

            3. Ample Margin of Safety

        Third, Environmental Petitioners argue that the EPA
violated the CAA by failing to include an ample margin of
safety. For health-based standards, the CAA directs the EPA
to set a “threshold level, with an ample margin of safety.” 42
U.S.C. § 7412(d)(4). Environmental Petitioners argue that, by
failing to do so, the EPA violated the plain language of the
CAA.

       EPA responds that it is entitled to deference in
determining “how” to provide the ample margin of safety
because the statute does not provide unambiguous direction.
According to EPA, we should defer to its interpretation that an
ample margin of safety can be built into a health threshold by
using “conservative” modeling assumptions that provide the
margin of safety “at multiple points in the process.” The
Environmental Petitioners counter that the word “with”
requires that the ample margin of safety be added to the health
threshold.

        The EPA offered a limited description of how it built a
margin into its model by adopting conservative assumptions.
80 Fed. Reg at 65,501. First, the EPA identified the source that
produced the highest modeled ambient concentration of
hazardous air pollutants. 

EPA modeled that
source, with worst-case weather and worst-case population
proximity, and scaled up its emissions until it reached a hazard
quotient of one, at which point it would be expected to affect
human health. The EPA used that scaled-up emissions value
as the health-based emissions limit. 

is unlikely
                              17
that any source would ever experience both worst-case weather
and population proximity when maximum emissions occur, the
EPA concluded that its emissions limit was “health protective”
and provides an ample margin of safety. 

conservative model, however, the EPA still found cases that
exceeded the hazard quotient of one, indicating a human health
risk, but nonetheless concluded that there was a low potential
for health impacts because it is unlikely that a person will be
present at the place and time when maximum risk occurs. 80
Fed. Reg. at 65,502-03.

        Though we agree that the EPA is entitled to deference
in determining how to include an ample margin of safety in the
health threshold, the question is whether the EPA provided any
margin of safety at all.           The EPA’s model actually
demonstrates a situation in which a source could emit up to the
emissions limit and, together with worst-case weather or
population proximity, the source could meet or exceed the
health threshold. Therefore, the EPA has not offered a
sufficient explanation of how its model includes an ample
margin of safety to the health threshold. Because the EPA did
not meet the CAA requirement to include an ample margin of
safety in the health threshold, we grant the Environmental
Petitioners’ petition as to this issue.

        For the reasons stated above, we grant the
Environmental Petitioners’ petition for review of the EPA’s
setting of a health threshold for the acid gases.

       B. Use of the Upper Prediction Limit to Set MACT
       Floors

       The Environmental Petitioners next contend that the
EPA’s methodology in setting the MACT floors is inconsistent
with the CAA. The EPA used a method called the “upper
                               18
prediction limit,” which is a statistical tool that allows the EPA
to estimate the best performing source from a limited data set
and account for the variability inherent to emissions and
emissions testing. See generally National Ass’n of Clean
Water Agencies v. EPA (“NACWA”), 

, 1122,
1144-45 (D.C. Cir. 2013). The EPA used the upper prediction
limit for twenty-two of the brick MACT floors and twenty-
seven of the clay MACT floors. A number of the brick and
clay MACT floors were based on datasets that EPA concedes
were “limited.”

         The EPA relies on “stack tests” to sample actual kiln
emissions. See Use of the Upper Predicition Limit for
Calculating MACT Floors (“HA for Calculating MACT
Floors”) at 2, Docket # EPA-HQ-OAR-2013-0291-0128 (Sept.
9, 2014). The EPA generally collects and tests emissions from
a source three times, each for a discrete period of time,
capturing snapshots of the emissions. See 

. However, emissions vary over time and there are
variations in testing equipment and laboratories; stack tests do
not measure the emissions at all times and under all conditions.

test emissions results are variable, the upper
prediction limit applies statistical methods to the results in
order to derive an emissions limit that accounts for that
variability within a specified degree of confidence. See 

prediction limit equation factors in the average of the
best performing source’s stack test results, the distribution of
the results, the variance of the results, and the total number of
tests in order to calculate a MACT floor. UPL for Calculating
MACT Floors at 4-5.

        We “accord[] Chevron deference to [the EPA’s]
interpretation of [the CAA] as allowing it to estimate MACT
floors.” U.S. 

(third alteration in
original). The EPA has “wide latitude” in its data gathering to
                               19
set the MACT floor. 

has discussed the EPA’s
application of the upper prediction limit method to calculate
MACT floors.

        In NACWA, our key concern was whether the EPA had
provided a reasoned basis to conclude that the upper prediction
limit represents the CAA’s “average emissions limitation
achieved.” 

, 1139-45; see also U.S.

. We held that the EPA “ha[d] not
clearly explained how the upper prediction limit itself operates
to predict [the best performing sources’ upper limit] with
sufficient accuracy” or “how the upper prediction limit can
actually predict the upper limit EPA expects the best
performing unit or units to achieve.” 

.

        In U.S. Sugar, the EPA provided the explanation that
had been lacking in NACWA, describing the limitations of
stack-test data, why the agency chose to use the upper
prediction limit, and how and why the upper prediction limit
accounts for variability in test results. U.S. 

-37. EPA explained how the upper prediction limit “allows
a reasonable inference” regarding the emissions of top
performing units, and upon review we held that the EPA “ha[d]
conducted reasoned decision making.” 

Thus, we
accepted the upper prediction limit as consistent with the CAA,
but with the caveat that whether it could be reasonably applied
to small datasets remained an open question. 

n.25.

        In this case, EPA has sufficiently explained the
application of the upper prediction limit to small datasets, but
in five cases failed to explain adjustments made to individual
upper prediction limit calculations. The EPA defined a
“limited dataset” as having between three and six data points,
and explained why it was possible to use datasets with as few
                               20
as three data points, though no fewer so long as the EPA
engages in further evaluation. Approach for Applying the
Upper Prediction Limit to Limited Datasets (“Brick Limited
Datasets Memo”) at 5-6, Docket # EPA-HQ-OAR-2013-0291-
0661 (Sept. 24, 2015); Approach for Applying the Upper
Prediction Limit to Limited Datasets (“Clay Limited Datasets
Memo”) at 5-6, Docket # EPA-HQ-OAR-2013-0290-0295
(Sept. 24, 2015). Except where adjustments were made to
results, Environmental Petitioners provide no reason to think
that any of the results reached by applying the upper prediction
limit to limited datasets are unreasonable or inadequately
explained. EPA failed, however, to adequately explain
adjustments it made to five of the upper prediction limit results.

        The EPA argues that it was sufficient for it to evaluate
the “reasonableness” of the results produced by the upper
prediction limit for small datasets on a case-by-case basis and
then adjust the results to best estimate the reductions in
emissions achieved by the best performing sources. Brick
Limited Datasets Memo at 6-10. EPA made no adjustments to
its upper prediction limits results for the Clay MACT floors.
Clay Limited Datasets Memo at 8. For brick plants, the EPA
made five adjustments to MACT floors because it concluded
that the upper prediction limit produced unreasonable results.
Brick Limited Datasets Memo at 8-10. In two of those cases,
EPA rejected datasets that were in fact from the best
performing source, resulting in a MACT floor that was derived
from the second best performing unit rather than the best
performing. 

         Although EPA adequately explains the general bases
for its evaluation and adjustment, 

the agency offered
only cursory explanations of the specific choices it made in
adjusting five of the floors. EPA also did not explain how using
data from the second best performing unit instead of the best
                               21
performing unit comports with the CAA’s requirements that
MACT floors be calculated based on the best performing unit
or units. 

EPA must better explain its reasoning
when performing such adjustments. For that we reason we
grant Environmental Petitioners’ petition as to the five adjusted
floors. We deny the petition as to the other applications of the
upper prediction limit to limited datasets.

       C. Provision of Alternative Emissions Floors for
       Brick Plants

        The Environmental Petitioners argue that the EPA’s
decision to allow brick tunnel kilns to comply with
“alternative” emissions floors is not authorized by the CAA.
Rather than a single limit for each brick tunnel kiln
subcategory, the EPA gave the brick industry a choice of three
emissions limits expressed in different units of measurement
for mercury (mass of pollutant per ton of bricks produced,
pounds per hour, or concentration) and non-mercury emissions
(pounds per hour and options that limit particulate matter as a
surrogate). 80 Fed. Reg. at 65,474, 65,530-31.

        The Environmental Petitioners argue that EPA’s
provision of alternative emissions limits violates the statute’s
plain language requiring the MACT floor to be set based on
“the best” performing sources in a category. 42 U.S.C.
§ 7412(d)(3). Thus, it is contrary to the plain meaning of “best”
to define several different groups of sources as “the best.” The
EPA counters that its interpretation of “best” should receive
deference because Congress did not specify which unit of
measurement the EPA should use to identify the best
performing source. Different sources perform differently
depending on the unit of measurement—for instance, a small
kiln might produce a smaller volume of pollutants overall than
a larger source but pollutes at a higher rate when emissions are
                               22
measured in pounds per hour. See Final [MACT] Floor
Analysis for Brick and Structural Clay Products
Manufacturing, at A-2, A-5, A-9, Docket # EPA-HQ-OAR-
2013-0291-0660 (Sept. 24, 2015). The EPA argues it was
reasonable for it to allow the industry to select which emissions
limit unit of measurement their kilns must meet. We disagree.

        The EPA has the authority to “distinguish among
classes, types, and sizes” of emissions sources and set separate
MACT floors for each. See 42 U.S.C. § 7412(d)(1). But once
the EPA identifies a source in a category it must set the MACT
floor based on the “best” performing sources. 

EPA has the discretion to determine what metric to use in
defining the “best” source, so long as it is reasonable. Cf.
Mossville Env. Action Now v. EPA, 

, 1241 (D.C.
Cir. 2004). However, the EPA’s discretion does not extend to
defining several different “best” metrics within the same
category and allowing emitters to comply with the most
favorable standard. The alternative standards allow a kiln to
avoid complying with a MACT floor defined by the best
emissions achieved by other sources in its category, as required
by the CAA. Because the EPA’s provision of alternate
emissions standards is contrary to the statutory requirement of
a standard based on the “best” performing sources, we grant the
Environmental Petitioners’ petition as to this issue also.

IV. Industry Petitioners’ Petition

        In four of the consolidated cases, Industry Petitioners
petition for review of the Brick/Clay Rule by challenging the
EPA’s methodology in setting the MACT floor for their
respective categories and subcategories of kilns. We address
in turn the Brick Industry Association’s petition for review of
the brick MACT floor, the Tile Council’s petition for review of
                               23
the Tile NESHAP, and Kohler’s petition for review of the Clay
NESHAP.

       A. Brick MACT Floor

        With respect to the portion of the Brick/Clay Rule in
which the EPA set the MACT floor for large and small tunnel
brick kilns, the Brick Industry argues that: (1) EPA violated
the CAA because it used synthetic minor sources to set the
MACT floor for brick plant major sources; (2) EPA acted
arbitrarily and capriciously because it did not explain its use of
allegedly suspicious data in setting the particulate matter and
nonmercury hazardous metal emissions standards; and
(3) EPA violated the intent of Congress by setting mercury
emissions standards which require raw material substitution.

           1. Use of Synthetic Minor Sources in Setting
           MACT Floors

       First, the petitioner argues that the EPA violated the
CAA because it used emissions data from synthetic minor
sources to set the MACT floor for particulate matter and
nonmercury hazardous metal emissions for brick plant major
sources. The EPA uses particulate matter emissions as a
surrogate for all nonmercury hazardous metal emissions. See
80 Fed. Reg. at 65,484.

         As an initial matter, the EPA argues that the Brick
Industry Association’s petition for review is untimely because
it failed to raise this issue when the EPA used synthetic minor
sources to set the MACT floors in the 2003 Rule. We lack
jurisdiction to review a claim unless the petition for review is
brought within sixty days of the challenged action appearing in
the Federal Register. Medical Waste Inst. & Energy Recovery
Council v. EPA, 

, 427 (D.C. Cir. 2011) (citing 42
                              24
U.S.C. § 7607(b)(1)). But this Court vacated the 2003 Rule.
Sierra 

. The existence of a prior but now-
vacated rule does not bar a later petition to review the
replacement rule. See Motor & Equip. Mfrs. Ass’n v. Nichols,

, 460-61 (D.C. Cir. 1998). We now turn to the
merits of the petition for review.

        Because there are more than thirty sources in the brick
plant source category, the EPA sets the MACT floor by
determining “the average emission limitation achieved by the
best performing 12 percent of the existing sources . . . in the
category.” 42 U.S.C. § 7412(d)(3)(A). The petitioner argues
that “synthetic minor sources” (also called “synthetic area
sources”) are not major sources. Thus, the petitioner contends
that the EPA’s inclusion of synthetic minor sources in the list
of best performing major sources violates the statutory
requirement that the EPA use data from sources “in the
category.”

       The EPA admits that it used emissions data from
synthetic minor sources to set the MACT floor for brick plant
major sources. A “major source” is defined as one that “emits
or has the potential to emit considering controls, in the
aggregate, 10 tons per year or more of any hazardous air
pollutant or 25 tons per year or more of any combination of
hazardous air pollutants.” 42 U.S.C. § 7412(a)(1) (emphasis
added). An “area source” is defined as “any stationary source
of hazardous air pollutants that is not a major source.” 

The CAA does not define “synthetic minor
source” or “synthetic area source.”

       The EPA explained that a synthetic minor source emits
lower quantities of hazardous air pollutants than a major source
“because they use some emission control device(s), pollution
prevention techniques or other measures . . . .” Summary of
                               25
Public Comments and Responses, at 46-47, Docket # EPA-
HQ-OAR-2013-0291-0685 (Sept. 2015). However, “[i]f not
for the enforceable controls they have implemented, synthetic
minor sources would be major sources under . . . the CAA.” 

The EPA’s definition of a synthetic minor source
conforms to the plain text of the CAA, which states that major
sources include sources with “the potential to emit considering
controls.” 42 U.S.C. § 7412(a)(1).

        Next, the Brick Industry argues that synthetic minor
sources do not qualify as “in the category” of major brick plant
sources, because treating them as in the same category deprives
the phrase of any meaning. We disagree. In this case, the EPA
has defined categories of small and large tunnel brick kiln
sources, while excluding tile and other nonbrick sources,
meaning the phrase “in the category” still has meaning. We
defer to the EPA’s reasonable interpretation that “the category”
should be defined as major brick sources, including those with
the potential to emit to that level considering controls, such as
synthetic minor sources, as required by the CAA.

        Accordingly, we deny the Brick Industry Association’s
petition for review because the EPA did not violate the CAA
by using emissions data from synthetic minor sources to set the
MACT floor for particulate matter and nonmercury metal
emissions for brick plant major sources.

           2. Data Used to Set Particulate Matter MACT
              Floors

        Second, the Brick Industry Association argues that the
EPA acted arbitrarily and capriciously because it did not
correct, supplement, or reconcile suspicious data and used that
data to set the MACT floor for particulate matter and
nonmercury hazardous metal emissions from brick plants.
                               26

       Brick plants may incorporate air pollution control
devices, such as fabric filters or dry lime absorbers. See 79
Fed. Reg. 75,625. The Brick Industry Association argues that
the EPA relied on erroneous data from kilns that use dry lime
absorbers as the best performing sources without explaining
why those sources performed unexpectedly better than the kilns
equipped with fabric filters.

         In the proposed Brick/Clay Rule, the EPA
acknowledged that data showed that fabric filters were
generally the best pollution control technology for particulate
matter emissions. 79 Fed. Reg. at 75,635. Accordingly, the
EPA proposed to determine the MACT floor based on the
twenty-seven best performing fabric filter-based kilns, which it
expected would represent the top twelve percent of the 255
operating major source and synthetic minor source brick tunnel
kilns. 

the proposal, the EPA requested
additional information to address whether low emissions data
from nonfabric-filter kilns showing high performance were
reliable. 

noted that it could, in the alternative, set
the particulate matter MACT floor based on the top twelve
percent of all brick plants in each kiln size subcategory. 

n.1. The EPA explained that “the reliability of the data
showing low emissions from some kilns without a [fabric
filter]-based [pollution control] is a key factor in the EPA’s
determination of which approach is appropriate.” 

final Brick/Clay Rule, the EPA adopted this
alternative approach, setting the particulate matter MACT floor
based on the best performing twelve percent of kilns in each
size subcategory without regard to the type of pollution
controls used. 80 Fed. Reg. at 65,485. This resulted in a
significantly more stringent control than expected in the
proposed rule based on data from fewer sources because the
                               27
EPA had not collected as much data from nonfabric filter
sources. See 

65,530. The Brick Industry
Association objects to the EPA’s adoption of this method,
arguing that dry lime absorbers are “only incidentally
effective” in controlling particulate matter emissions, therefore
any data that shows these kilns are performing better than those
with fabric filters is “technically not plausible.” Therefore, the
Brick Industry Association contends, the EPA relied on
anomalous results and acted arbitrarily by failing to resolve the
anomalies. Further, the Brick Industry Association argues that
the EPA compounded its error by removing data points from
kilns utilizing fabric filters.

        In the final Brick/Clay Rule and upon reconsideration,
the EPA explained that it received no data explaining why
some kilns without fabric filters performed exceptionally well.
80 Fed. Reg. at 65,485. Further, the EPA noted that in some
cases there were issues with probe and filter temperatures
which invalidated the data from some fabric filter kilns,
requiring that data to be thrown out. 

Additionally, the EPA had observed in the proposed rule that
the data from the dry lime absorber kilns showed that the fabric
filters might not be the best performing filters as originally
thought. 79 Fed. Reg. at 75,635.

        In light of this admittedly unexpected result, the EPA
reviewed the data to confirm that the tests were performed in
normal operating conditions and requested additional data from
the Brick Industry Association. 

Industry
Association provided a list of the dry lime absorber kilns that
it recommended should be excluded from the top performing
set of kilns, primarily speculating these kilns performed well
because they had installed new limestone but offering no other
concrete evidence or analysis in support of this speculation.
                               28
Email Re: DLA/Uncontrolled Kilns, Docket # EPA-HQ-OAR-
2013-0291-0614 (June 25, 2015).

        As we have already noted, “EPA typically has wide
latitude in determining the extent of data-gathering necessary
to solve a problem.” Sierra Club v. EPA, 

, 662
(D.C. Cir. 1999). The EPA’s explanation of why it discarded
some of the data from the fabric filter-based kilns and why it
relied on the best twelve percent of all the sources for which it
had data, rather than just the best performing fabric filter-based
sources, was sufficiently reasoned and is entitled to deference.

         Additionally, the Brick Industry Association argues
that the EPA failed to respond to its comment containing the
list of dry lime absorber kilns that it contends should have been
excluded from consideration as best performing sources. The
EPA has an obligation to respond to significant comments to
the regulation. 42 U.S.C. § 7607(d)(6)(B). However, the EPA
determined that the Brick Industry Association’s comment was
not significant because it failed to offer any explanation or data
showing the data from high-performing nonfabric filter kilns to
be unreliable. We agree. Because the comment contained only
speculation, without additional details, it was not “significant.”

         Accordingly, we deny the Brick Industry Association’s
petition for review because the EPA was not arbitrary or
capricious when it used data from kilns that did not use fabric
filters to set the MACT floor for particulate matter emissions
from brick plants.

           3. Improper Requirement to Substitute Raw
              Materials

      Third, the Brick Industry argues that the EPA violated
Congressional intent by setting mercury emissions standards
                               29
that require that the industry make raw material substitutions.
Mercury content varies based on where clay is quarried, which
can affect the resulting mercury emissions from kilns using the
clay. 80 Fed. Reg. at 65,485. Relying on the legislative
history, the Brick Industry Association argues that the 1990
amendment to the CAA prohibited the EPA from requiring
material substitutions when it sets emissions standards. See
H.R. Rep. No. 101-952, at 339 (1990) (Conf. Rep.). Further,
the Brick Industry Association argues that the EPA acted
unreasonably because it did not establish subcategories based
on raw materials to avoid raw material substitutions.

        “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 . . . .”
U.S. 

. This Court has held that the EPA
is required to set a MACT floor that “reflect[s] what the best
performers actually achieve,” and the EPA may not deviate
from that by requiring that “floors must be achievable by all
sources using MACT technology.” Cement Kiln Recycling
Coal. v. EPA, 

, 861 (D.C. Cir. 2001). Likewise,
in addressing the 2003 Rule in Sierra I, we held that setting
MACT floors based on technology controls alone was
insufficient to satisfy the CAA’s 

. Nontechnology factors, such as the source of raw
materials, can affect emissions levels and cannot be ignored.

words, the EPA need not consider whether all
sources can achieve the MACT floor, and must still set the floor
based on the emissions achieved by the best performers even
though they have access to different raw materials than the
other sources. 

the Brick Industry Association’s
argument is misplaced. The EPA did not mandate a raw
material substitution. The EPA responded to a comment by
                               30
stating that it is “not requiring the use of offsite raw materials
in this rule.” Summary of Public Comments and Responses at
26. Furthermore, the EPA considered the effect of raw
materials in defining the MACT floor and used a variability
equation to factor in the variation in mercury content in clay
quarries. 80 Fed. Reg. at 65,485. Additionally, the EPA
considered the existence of potential control technology that
would allow the industry to achieve the MACT floor without a
raw material substitution. Summary of Public Comments and
Responses at 26-27.

        The EPA reasoned that some brick kilns could use an
activated-carbon injection control device in order to achieve
the MACT floor without a raw material substitution. 

Industry Association contends that it is unreasonable for
the EPA to consider this device because the EPA has not shown
it would be effective. However, the EPA cited information
showing that activated-carbon injection devices are effective in
other industries in controlling mercury emissions from similar
sources. Methodology and Assumptions Used to Estimate the
Model Costs and Impacts of BSCP Air Pollution Control
Devices for the Final Rule at 7-9, Docket # EPA-HQ-OAR-
2013-0291-0662 (Sept. 24, 2015). Finally, in response to the
Brick Industry Association’s comment that the EPA should
have created subcategories based on the mercury content of the
raw materials, instead of the size of the brick kilns, the EPA
explained that it did not have enough information to create
those subcategories, because the data did not show a strong
correlation between raw material mercury and mercury
emissions. Summary of Public Comments and Responses at
13.

       The EPA relied on substantial evidence to conclude that
technological controls are available to achieve the MACT floor
without raw material substitution and made a reasoned decision
                               31
not to subcategorize based on the mercury content of raw
materials. Likewise, the EPA is not required to set a standard
that is achievable by all sources. Therefore, we deny the Brick
Industry Association’s petition for review.

       B. Tile NESHAP

        The Tile Council of North America petitions for review
of the Tile NESHAP. The petitioner first argues that the EPA
violated the CAA by listing it as a major source because there
are no longer any major source tile kilns and, further, that it
was denied its right to comment on its listing. Next, the Tile
Council argues that even if the EPA was permitted to
promulgate the Tile NESHAP, the EPA erred by failing to
perform a beyond-the-floor MACT analysis for dioxin/furan
emissions.

        The central issue in the Tile Council’s petitions is its
objection to the EPA’s listing of tile kilns as a major source of
hazardous air pollutants. In 2002, the EPA replaced a clay
products manufacturing source category with two source
categories: brick and structural clay products and clay ceramic
products, the latter of which includes tile. 67 Fed. Reg. 47,894
(July 22, 2002). At the time, there were tile-manufacturing
plants that were major sources of hazardous air pollutants. 

Once an industry is listed as a major source of
hazardous air pollutants, the EPA must establish emissions
limits. 42 U.S.C § 7412(d)(1). In 2014, just before the EPA
proposed the Brick/Clay Rule, all tile kilns that that would be
subject to its emissions standards as major sources either closed
or became synthetic area sources. TCNA Letter, Docket
# EPA-HQ-OAR-2013-0290-0131 (Aug. 29, 2014).

       As a threshold matter, we must address the Tile Council
of North America’s standing. Its membership is made up of
                              32
domestic ceramic tile manufacturing companies. To
demonstrate that it has standing, the Tile Council must
establish its injury is actual or imminent, not hypothetical.
Lujan v. Defenders of Wildlife, 

, 560 (1992).
When “standing is not self-evident, . . . the petitioner must
supplement the record to the extent necessary to explain and
substantiate its entitlement to judicial review.” Sierra Club v.
EPA, 

, 900 (D.C. Cir. 2002). The Tile Council
asserts that it has standing because its members operate
ceramic tile kilns that fall within the Tile NESHAP
subcategory and those members have a direct financial and
environmental stake in the Brick/Clay Rule.

       The Tile Council admits that none of its members are
major sources, and so they are not subject to the Tile
NESHAP’s emissions limits. But the Tile Council argues that
its members are constructing new kilns that may be subject to
the major source emissions limits. Further, the Tile Council
claims that its members expend resources to remain area
sources and avoid being subject to emissions limits on major
sources. The EPA argues that the Tile Council lacks standing
because none of its members have suffered harm or face
imminent harm from the emissions limits on major sources.
We agree with the EPA. The Tile Council failed to identify
any member that is injured or will imminently be injured by the
Brick/Clay Rule. The Tile Council’s arguments regarding
potential future injuries are merely speculative.

        Likewise, the Tile Council of North America’s
assertion that it has associational standing also fails. To
establish associational standing, its members must have
individual standing to challenge the Tile NESHAP in their own
right. Hunt v. Wash. State Apple Advert. Comm’n, 

, 342 (1977). Because we have concluded that the Tile
Council of North America has not identified an injury or
                               33
imminent injury to any of its members from the Tile NESHAP,
it has not demonstrated that its members have individual
standing to challenge the rule in their own right. Therefore, the
Tile Council of North America does not have associational
standing.

        Because we find that Tile Council of North America
has no standing to petition for review of the Tile NESHAP, we
need not address the petitioner’s remaining arguments on their
merits.

       C. Clay NESHAP

        With respect to the Clay NESHAP, Kohler argues that
the EPA violated the CAA by using emissions data that was not
representative of any existing operating source to set the
MACT floor for clay sanitaryware kilns. Because there are less
than thirty sources in the clay kiln source category, the EPA
sets the MACT floor based on “the average emission limitation
achieved by the best performing 5 sources.” 42 U.S.C.
§ 7412(d)(3)(B) (emphasis added). Specifically, Kohler argues
that the EPA erred by including data from Kiln 10 as one of the
best performing sources.

        When Kohler constructed Kiln 10 in 2005, it installed a
scrubber to comply with the now-vacated 2003 Rule for new
sources.    80 Fed. Reg. at 65,510.        In 2009, Kohler
decommissioned the scrubber and now operates Kiln 10
without it. 

Kiln 10 scrubber remained
functional and attached to the source. 

data
collection period, the EPA asked Kohler to reactivate the
scrubber and EPA tested Kiln 10 while the scrubber was
operating. 

have repeatedly recognized that section 7412(d)(3)
requires the floors to reflect emissions achieved “in practice.”
See, e.g., Cement 

; Nat’l Lime Ass’n v.
EPA, 

, 632 (D.C. Cir. 2000). It may seem that the
“in practice” requirement supports Kohler’s argument that the
data collected from Kiln 10 while it was using the scrubber was
“artificial” and in performance of a test, not “in practice.”
However, the statute does not specify when emissions achieved
“in practice” occur. Likewise, the statute requires that the EPA
consider the average emissions “achieved,” which
contemplates that the performance could have occurred in the
past. See 42 U.S.C. § 7412(d)(3).

       Kohler operated Kiln 10 with the scrubber between
2005 and 2009. 80 Fed. Reg. at 65,510. The EPA correctly
reasoned that emissions produced with a scrubber were
“achieved” during that time. That the EPA chose to collect the
data through a later test after Kohler decommissioned the
scrubber rather than during its four years of operation is within
the EPA’s discretion.

       This Court considered a similar argument in a case
regarding emissions standards for hospital waste incinerators.
Med. Waste Inst. & Energy Recovery Council v. EPA, 

(D.C. Cir. 2011). In that case, we affirmed the EPA’s use
of emissions data from a facility that had put controls in place
to comply with a remanded, but not vacated, rule. 

424-25. Though in this case the EPA used data from a source
with pollution controls that were put in place to comply with a
rule that was vacated, this distinction is not of “outcome-
changing significance.” 

       We deny Kohler’s petition for review because the EPA
did not violate the CAA when it used data from Kiln 10 as a
                              35
best performing source because the EPA reasonably concluded
that Kiln 10 had achieved those emissions “in practice.”

V.   Conclusion

        For the reasons set forth above, we deny the Industry
Petitioners’ petitions for review and we grant the
Environmental Petitioners’ petition for review as to (1) the
EPA’s use of a health threshold to set the emissions limit for
acid gases; (2) the EPA’s ad hoc adjustments of upper
prediction limit calculations, and (3) the EPA’s provision of
alternative MACT floors for brick plants. We deny the
Environmental Petitioner’s petition for review as to the general
application of the upper prediction limit to limited datasets as
defined by the EPA. We remand the Brick/Clay Rule to the
agency for further proceedings consistent with this opinion.