United States Court of Appeals
         FOR THE DISTRICT OF COLUMBIA CIRCUIT



Argued October 21, 2019                Decided March 13, 2020

                         No. 15-1015

      CHESAPEAKE CLIMATE ACTION NETWORK, ET AL.,
                     PETITIONERS

                              v.

           ENVIRONMENTAL PROTECTION AGENCY,
                     RESPONDENT

            ALABAMA POWER COMPANY, ET AL.,
                    INTERVENORS


            Consolidated with 16-1169, 16-1349


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


    Patton Dycus argued the cause for petitioners. With him
on the briefs were Neil Gormley, James S. Pew, and Eric
Schaeffer. Sanjay Narayan entered an appearance.

     Meghan E. Greenfield, Trial Attorney, U.S. Department of
Justice, argued the cause for respondent. With her on the brief
were Jeffrey Bossert Clark, Assistant Attorney General, and
Jonathan D. Brightbill, Principal Deputy Assistant Attorney
                                    2
General. Norman L. Rave Jr. and Stephanie J. Talbert,
Attorneys, entered appearances.

    Lauren E. Freeman, Makram B. Jaber, Andrew D.
Knudsen, C. Grady Moore, III, Margaret C. Campbell, and
Emily Gerhardt were on the brief for industry intervenor-
respondents.

     Before: TATEL, PILLARD, and WILKINS, Circuit Judges.

     Opinion for the Court filed by Circuit Judge WILKINS.

     WILKINS, Circuit Judge: Congress charged the EPA with
establishing emission regulations under the Clean Air Act. In
a 2014 final rule challenged here, EPA exempted coal- and oil-
burning power plant utility boilers’ startup periods from
numerical limits on hazardous air pollutants. Instead, EPA
imposed qualitative “work practice” standards during these
periods of time. This consolidated action challenges that final
rule on two fronts. First, Petitioners challenge EPA’s denial of
their petition for reconsideration of the final rule as
procedurally improper. Second, Petitioners challenge the final
rule itself as arbitrary and capricious.

     For the reasons that follow, we conclude that EPA erred in
denying Petitioners’ petition for reconsideration. We grant the
petition in No. 16-1349 because it was impracticable for
Petitioners to raise their two objections during the notice-and-
comment period and the objections were of central relevance
to the final rule. Because we remand the final rule for EPA’s
reconsideration, we do not reach the merits arguments set forth
in No. 15-1015.1
1
 Petitioners brought a third challenge to a separate final action by EPA on
April 6, 2016, entitled “National Emission Standards for Hazardous Air
Pollutants From Coal- and Oil-Fired Electric Utility Steam Generating
                                      3

                                     I.

     The Clean Air Act (“CAA”), 42 U.S.C. § 7401 et seq., was
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,” § 7401(b)(1). In
order to regulate the emission of toxic pollutants considered
hazardous to public health, the CAA created a list of hazardous
air pollutants (HAPs) and required EPA to promulgate
restrictions on their emission by various sources. See U.S.
Sugar Corp. v. EPA, 

, 593 (D.C. Cir. 2016) (per
curiam). This case involves such restrictions on an important
category of existing major sources, specifically utility boilers
at coal-fired and oil-fired power plants. See § 7412(a)(10)
(defining “existing source” as “any stationary source other than
a new source”); U.S. Sugar 

(describing
major sources 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
of any HAP combination” (quoting § 7412(a)(1))).

     EPA must set HAP emission limits in the form of
numerical limits whenever “feasible,” § 7412(d)(2), (h)(4), and
limits for major sources must be capped at the “the maximum
degree of reduction in emissions” that EPA deems

Units and Standards of Performance for Fossil-Fuel-Fired Electric Utility,
Industrial-Commercial-Institutional, and Small Industrial-Commercial-
Institutional Steam Generating Units; Technical Correction.” 81 Fed. Reg.
20,172 (April 6, 2016). We consolidated that challenge, contained in
Petitioners’ Petition for Review at No. 16-1169, with the two petitions
discussed herein. Petitioners acknowledged both in their opening brief and
at oral argument that they do not “specifically challeng[e] any portion of the
final action at issue” with respect to that April 6, 2016, final action, so we
dismiss the petition in No. 16-1169. Pet’rs’ Opening Br. 3 n.1.
                                    4
“achievable,” § 7412(d)(2). EPA’s determination of what is
“achievable” is often referred to as a “MACT” standard, as in
“maximum achievable control technology.” U.S. Sugar 

.

     In setting a MACT standard, EPA follows a two-step
process. 

First, it creates a “MACT floor” for each
category of emissions source that “ensures that all HAP sources
at least clean up their emissions to the level that their best
performing peers have shown can be achieved.” 

marks omitted). For sources like those at issue here,
“the MACT floor cannot be less stringent than the average
emissions limits achieved by the best performing 12 percent of
existing sources in that category or subcategory.” 

7412(d)(3)(A) (emphasis added)). Second, EPA may
exercise its discretion “to require an even greater reduction in
emissions, taking into account costs, health effects,
environmental effects, and energy requirements.” Nat. Res.
Def. Council v. EPA, 

, 1079 (D.C. Cir. 2008)
(citing § 7412(d)(2)).2

     Congress recognized that numerical HAP emission limits
for MACT standards may not always be “feasible,” so it
included § 7412(h), which enables EPA to promulgate number-
alternative standards called “work practice” standards.
§ 7412(h); see Sierra Club v. EPA, 

, 1190 (D.C.
Cir. 2018) (“Work practice standards can be thought of as a
statutory Plan B; EPA may resort to them only when using
numeric limits is not feasible.” (internal quotation marks

2
  This second step, often referred to as “beyond-the-floor” limits, “risk-
based” limits, or “health-based” limits after the factors listed in
§ 7412(d)(2), is not at issue here. Cement Kiln Recycling Coal. v. EPA, 

, 858 (D.C. Cir. 2001) (per curiam); see also Nat. Res. Def.

.
                                5
omitted)). Relevant here, numerical MACT standards are not
feasible (and thus “work practice” standards may be used)
when “the application of measurement methodology to a
particular class of sources is not practicable due to
technological and economic limitations.” § 7412(h)(2)(B).
However, work practice standards must be, “in the
Administrator’s judgment,” consistent with numerical MACT
requirements, § 7412(h)(1), i.e., “the maximum degree of
reduction in emissions” that EPA deems “achievable,”
§ 7412(d)(2).

     In addition, although not expressly contemplated by the
CAA, we have held that EPA has the “flexibility” to “regulate
a HAP indirectly, by controlling a proxy, or ‘surrogate,’ instead
of the pollutant itself. . . . so long as the resulting rules are
reasonably calculated to control the relevant HAPs to the extent
the statute demands.” Sierra 

(citing
U.S. Sugar 

).

     Finally, § 7607 sets forth the CAA’s administrative
proceedings and judicial review provisions applicable to
standards promulgated under § 7412. 42 U.S.C. § 7607(b),
(d)(1)(C). Notices of proposed rulemaking are published in the
Federal Register, accompanied by a statement of the basis and
purpose, the period available for public comment, and a
summary of the factual data on which the proposed rule is
based and the methodology used in obtaining and analyzing the
data. § 7607(d)(3). Anyone may submit written comments and
data in response to a proposed rule during the comment period,
§ 7607(d)(3), (d)(5)(i), but one is entitled to reconsideration by
the Administrator after the period for public comment has
passed if that person can show: (1) “it was impracticable to
raise such objection within such time or if the grounds for such
objection arose after the period for public comment (but within
the time specified for judicial review),” and (2) “such objection
                               6
is of central relevance to the outcome of the rule,”
§ 7607(d)(7)(B). EPA’s refusal to convene such a proceeding,
which is what happened here, is subject to judicial review. See

statutory framework in mind, we turn to the
regulatory background that led to the final rule challenged here.

                               II.

      Coal- and oil-fired electric utility steam generating units
(sometimes referred to as “EGUs”) are one of the main sources
of many HAPs emitted into our atmosphere. Air pollution from
utility boilers at coal-fired and oil-fired power plants can be
particularly problematic during a facility’s startup or shutdown
period. According to EPA, an average power plant had
between 9 and 10 startup events annually between 2011 and
2012, but some plants had over 100 startup events in 2011.
Assessment – Revised, EPA-HQ-OAR-2009-0234-20451, at 4
(Nov. 2014). Environmental groups believe the number of
startup events will increase as renewable energy and gas-fired
units replace coal-fired power plants. Thus, EPA promulgated
specific emission standards applicable during these periods.

                               A.

      In February 2012, EPA promulgated the “MATS Rule,”
which set national emission standards for HAPs emitted from
utility boilers at coal-fired and oil-fired power plants. MATS
Rule, 77 Fed. Reg. 9,304, 9,380-83 (Feb. 16, 2012). Not to be
confused with MACT standards, MATS is shorthand for
“Mercury and Air Toxics Standards.” 

The MATS
Rule requires a power plant to comply with established
numerical emission limits at all times except during periods of
startup and shutdown. 

The numerical emission
                                7
limits in the MATS Rule are “production-based,” so EPA
believed that startup periods – in which production is by
definition non-existent – required separate emissions limits. 

The MATS Rule defined the period of “startup” as

       either the first-ever firing of fuel in a boiler for
       the purpose of producing electricity, or the
       firing of fuel in a boiler after a shutdown event
       for any purpose. Startup ends when any of the
       steam from the boiler is used to generate
       electricity for sale over the grid or for any other
       purpose (including on-site use).

EPA rationalized the definition “based on the fact
that EGUs function to provide electricity primarily for sale to
the grid but also at times for use on-site; therefore, EGUs
should be considered to be operating normally at all times
electricity is generated.” 

EPA acknowledged that
it previously had proposed numerical emission standards for
periods of startup, but it received a number of comments, with
“[m]any commenters point[ing] to the lack of data in the record
concerning emissions that occur during periods of startup.” 

Rule explained that of various data collected

       there were almost no HAP data for startup and
       shutdown periods and almost all of the data
       failed to meet our data quality requirements.
       Thus, we do not have sufficient data on
       emissions that occur during startup and
       shutdown on which to set emission standards.

MATS Rule promulgated non-numerical
work practice standards for periods of startup and shutdown.

practice standards set forth in the MATS Rule
required, among others, EGUs to use clean fuels for ignition.
                               8

3. The MATS Rule does not specifically cite to
§ 7412(h), which governs when work practice standards may
be implemented in the place of numerical standards.

                              B.

     Following the promulgation of the MATS Rule, “[t]he
EPA received petitions [for reconsideration] asserting that the
public lacked an opportunity to comment on the startup and
shutdown provisions in the final MATS [Rule].”
Reconsideration of Certain New Source and Startup/Shutdown
Issues, 77 Fed. Reg. 71,323, 71,330 (Nov. 30, 2012). EPA
acknowledged that it had previously “proposed numerical
standards for startup and shutdown periods, and in response to
comments on the proposed rule [it] changed those standards in
the final MATS to work practice standards.” 

opened the period for comment on the nature of the work
practice standards during the startup period. Contrary to
Intervenors’ assertion, EPA did not propose altering the
endpoint of startup, which remained defined as the point in
which the power plant is able to generate electricity.

                              C.

     On June 25, 2013, EPA issued a Proposed Rule and
Reopening of Comment Period. Reconsideration of Certain
Startup/Shutdown Issues, 78 Fed. Reg. 38,001, 38,002
(June 25, 2013) (“2013 Proposed Rule”). EPA explained that
in the preceding comment period, “comments raised several
significant issues regarding the definition of startup,” among
other issues. 

new startup endpoints
based not on when the source first generated electricity, but on
the number of hours after an EGU hits certain operational
benchmarks because then air pollution control devices
                               9
(“APCDs”) would be operational. 

EPA noted
that

       The commenters asserted that an EGU remains
       in “startup” mode beyond the first generation of
       electricity because, according to the
       commenters, at that point in time many of the
       APCDs needed to comply with the
       requirements of this subpart may not be
       technically or safely capable of operation and
       those that are may be operating far from design
       conditions because the requisite temperature(s)
       and/or flow conditions have not been achieved.

EPA sought comments on these suggested
definitions “so that the public can review the industry-provided
information and data and comment on the suggested revisions
to the startup and shutdown provisions.” 

     “In addition, the EPA request[ed] comment on the
additional technical analyses it conducted in response to the
above comments concerning the end of startup.” 

(citing Assessment of Startup Period at Coal-fired Electric
Generating Units, EPA-HQ-OAR-2009-0234-20224 (June 17,
2013)). We refer to this technical support document, which
accompanied the 2013 Proposed Rule, as the “2013 TSD.”
According to the 2013 Proposed Rule, the 2013 TSD
“examined several indicators that can aid in assessing the time
required to achieve operating benchmarks.” 78 Fed. Reg. at
38,005.

    The 2013 TSD identified those “several indicators” to be
the removal efficacy of sulfur dioxide and nitrogen oxides
emission APCDs, analyzing their average time for engagement
across several categories. 2013 TSD, at 2. Although sulfur
                                    10
dioxide and nitrogen oxides are not HAPs, “EPA believes that
the removal efficacy of [these] APCDs, as evidenced by hourly
emission rates well below uncontrolled levels, may be used as
an indicator of the end of the startup period for the purpose of
the MATS rule.” 

TSD identified a specific set of data3 on
emissions in order to identify all startup events at different
types of boilers. 

when (in terms of hours)
after generation sulfur dioxide and nitrogen oxides emissions
were reduced by APCDs, and then categorized those results by
combustion technologies, APCD type, and boiler type. 

“found no significant difference in
performance related to startup between the different groups
assessed in this analysis,” and “could support defining the end
of startup at coal-fired EGUs as occurring at [a specific
capacity threshold] plus 3 hours or the start of electricity
generation plus 6 hours, whichever comes first.” 78 Fed. Reg.
at 38,005; see also 2013 TSD, at 19.

     Neither the 2013 TSD nor the 2013 Proposed Rule
referenced the work practice standards provision, § 7412(h), or
analyzed its requirements for application.




3
  The data used in the 2013 TSD came from EPA’s Clean Air Markets
Database, which included data from EPA’s Acid Rain Program. Petitioners
base their arbitrary-and-capricious argument on EPA’s alleged failure to
explain how it may treat emissions post-generation as immeasurable based
on EPA’s measured emission data from another program. In other words,
Petitioners argue that EPA failed to reasonably explain how it could treat
similar situations differently. But as already noted, we need not reach that
issue here.
                               11
                               D.

     Several environmental groups, including Petitioners,
submitted comments in response to the 2013 Proposed Rule.
Petitioners commented on the feasibility of numerical
measurability after generation, pointing out the broader
problem that EPA failed to make any finding that numerical
emission standards during this extended time frame were
infeasible under § 7412(h) such that EPA could lawfully
promulgate the extended period of work practice standards.
Petitioners argued in their timely comment that “feasible
‘measurement methodolog[ies]’” were available in order to
demonstrate that EPA could not lawfully conclude that work
practice standards were needed.           J.A. 103 (quoting
§ 7412(h)(2)(B) (alteration in original)).        Additionally,
Petitioners pointed out that EPA’s work practice standards
“cannot be less stringent than ‘the average emission limitation
achieved by the best performing 12 percent of the existing
sources, respectively.’” J.A. 163 (quoting § 7412(d)(3)).

                               E.

     Ultimately, EPA adopted two alternative definitions of
“startup,” each with its own end point, in the final rule at issue
here. Reconsideration of Certain Startup/Shutdown Issues, 79
Fed. Reg. 68,777 (Nov. 19, 2014) (codified at 40 C.F.R.
§ 63.10042) (“Final Rule”). Power plants may choose between
the definitions because EPA “believe[s] that they both meet the
requirements of [§ 7412] to reduce HAP emissions during this
time period . . . .” 

Once startup ends (regardless
of which definition the plant uses), the plant must then comply
with the MATS Rule’s numerical emission requirements. 

                                12
     The first startup definition mirrors the original definition
from the 2012 MATS Rule and the November 2012 action:
startup begins with “the first-ever firing of fuel in a boiler for
the purpose of producing electricity, or the firing of fuel in a
boiler after a shutdown event for any purpose,” and it ends at
electricity generation. 

This portion of the Final
Rule is not challenged here.

     The alternative definition evolved from the proposed
extended definition tested by the 2013 TSD. It defines the
beginning of startup similarly to the first definition, but defines
the end of startup as occurring at a later juncture: “Startup
ends 4 hours after the EGU generates electricity . . . or 4 hours
after the EGU makes useful thermal energy (such as heat or
steam) for industrial, commercial, heating, or cooling purposes,
whichever is earlier.” 

(internal citations omitted).

     EPA acknowledged that the 2013 TSD “did not attempt to
identify the EGUs that were the best performing sources, but
instead simply looked at a category-wide average time for
engagement of APCDs.” 

In order to “most
closely follow[] the requirements” of § 7412(h), EPA “revised”
its technical support documents to determine which EGUs
“were able to most quickly engage their [sulfur dioxide and
nitrogen oxides] APCDs because [EPA] determined that the
best performing EGUs for purposes of defining the end of
startup are those that are able to most efficiently engage their
controls after the start of electricity generation.” 

to this “revised” technical support document as the “Final
TSD.” Assessment – Revised, EPA-HQ-OAR-2009-0234-
20447.

    The 2013 Proposed Rule connected the removal efficacy
of APCDs regulating sulfur dioxide and nitrogen oxides
emissions to “achiev[ing] operating benchmarks,” 78 Fed. Reg.
                              13
at 38,005, specifically the point where “APCDs needed to
comply with the requirements of this subpart [became]
technically or safely capable of operation,” 

The
Final Rule filled in some gaps. According to the Final Rule,
sulfur dioxide and nitrogen oxides emissions can be proxies of
other emissions and their APCDs aid in the engagement of
HAP controls “that industry commenters stated required
additional time to engage after the start of generation of
electricity or useful thermal energy.” 79 Fed. Reg. at 68,781.
The takeaway in the Final Rule was that EPA believed the
removal efficacy of APCDs regulating sulfur dioxide and
nitrogen oxides emissions was reliable evidence of when EGUs
can begin to not only remove pollutants but measure HAP
emissions. See 

                              F.

     Petitioners submitted a petition for reconsideration of the
Final Rule that objected to the alternative extended startup
definition. Specifically, Petitioners objected to EPA’s late-
breaking selection of the best performing power plants which
formed the basis of its Final TSD analysis and objected to
EPA’s conclusion that work practice standards were lawful
under § 7412(h) beyond electricity generation. EPA denied the
petition on August 8, 2016. Reconsideration on the [MATS]
and the Utility New Source Performance Standards Startup and
Shutdown Provisions; Final Action, 81 Fed. Reg. 52,346,
52,346 (Aug. 8, 2016). EPA explained why it was denying the
petition for reconsideration in a separate document
accompanying the Federal Register notice. See Denial of
Petitions for Reconsideration of Certain Startup/Shutdown
Issues: MATS, EPA-HQ-OAR-2009-0234-20581 (July 29,
2016).
                                    14
                                   III.

     Although EPA and Intervenors do not challenge
Petitioners’ standing under Article III to seek judicial review
of both the denial of the petition for reconsideration and the
Final Rule, the Court has an independent obligation to ensure
standing exists. The Court concludes that because at least one
Petitioner – Sierra Club – demonstrates standing, this case may
proceed to the merits. See Ctr. for Biological Diversity v. EPA,

, 182 (D.C. Cir. 2017) (“When more than one
association brings suit, we need only find one party with
standing to satisfy the requirement.” (internal quotation marks
and citations omitted)).

     Article III of the Constitution provides an association with
standing to sue “only if (1) at least one of its members would
have standing to sue in his own right; (2) the interest it seeks to
protect is germane to its purpose; and (3) neither the claim
asserted nor the relief requested requires the member to
participate in the lawsuit. ” Am. Trucking Ass’ns v. Fed. Motor
Carrier Safety Admin., 

, 247 (D.C. Cir. 2013)
(internal quotation marks and citations omitted). Sierra Club’s
member’s standing here, as demonstrated in the declaration
attached to Petitioners’ Reply,4 is clear for the same reasons we
articulated in previous cases involving these parties:

         [Its] members “unquestionably live[d] within
         zones they claim are exposed to” regulated air
         pollutants and “our vacatur [would] require

4
 Petitioners submitted declarations from other members with their opening
brief, but Petitioners later “realized that the power plants harming
Petitioners’ opening-brief standing declarants have not elected to use the
extended startup period.” Pet’rs’ Reply Br. 3. We accept the declaration
submitted with Petitioners’ reply. See Ctr. for Sustainable Econ. v. Jewell,

, 599 (D.C. Cir. 2015).
                              15
       EPA . . . to entertain and respond to the
       [Petitioners’] claims about the necessary scope
       and stringency of the standards” for regulating
       those pollutants. Sierra Club has “shown its
       members’ . . . concrete interest” of a type that
       its asserted procedural interest is “plainly
       designed to protect” and that its injury “is
       potentially redressable” by further agency
       action on remand.

Sierra Club v. EPA, 

, 849 (D.C. Cir. 2019)
(quoting Sierra Club v. EPA, 

, 533 (D.C. Cir.
2012)). Because the interest Sierra Club seeks to protect is
germane to its purpose and neither the claim asserted nor the
relief requested requires the member to participate in the
lawsuit, Sierra Club has established standing.

                              IV.

     As we noted at the outset, Petitioners challenge both
EPA’s denial of their petition for reconsideration and the Final
Rule itself. The portions of the petition for reconsideration at
issue here attack EPA’s determination in the Final Rule that the
“best performing” power plants cannot reliably measure
emissions until four hours after they generate electricity. This
objection is really two-fold. First, Petitioners argue that EPA
did not analyze which power plants were the “best performers”
until the Final TSD, thus depriving Petitioners of the
opportunity to timely comment. Second, Petitioners argue
EPA’s best-performers analysis was based on an underlying
assumption that the point at which power plants begin to
operate their APCDs and remove pollutants can be used as a
proxy for when it becomes feasible to measure HAP emissions,
an assumption not disclosed prior to the Final Rule.
                                16
     Petitioners seek relief in the form of an order compelling
EPA to reconsider the Final Rule in light of their objections.
While Petitioners ask the Court to vacate the denial of the
petition for reconsideration, they do not ask the Court to vacate
the Final Rule itself.          At oral argument, Petitioners
acknowledged that if we remand for reconsideration under
§ 7607(d)(7)(B), we need not reach their additional arguments
that the Final Rule, as it stands now, is arbitrary and capricious.
See Oral Arg. Rec. at 22:00-26. Because we conclude that
Petitioners’ reconsideration petition raised objections that were
both impracticable to raise during the comment period and
centrally relevant to the outcome of the Final Rule, see
§ 7607(d)(7)(B), we vacate EPA’s denial of that petition and
remand for EPA’s reconsideration on these matters.

     In Clean Air Council v. Pruitt, we declined to resolve a
dispute about whether our review of EPA’s treatment of the
two § 7607(d)(7)(B) elements is subject to de novo review or
arbitrary-and-capricious review, concluding that EPA erred
even under the more deferential arbitrary-and-capricious
standard. 

, 10 (D.C. Cir. 2017) (per curiam). Here,
Petitioners seek de novo review, arguing that “EPA has no
greater expertise than this Court in determining whether the
requirements of § 7607(d)(7)(B) have been satisfied.” Pet’rs’
Opening Br. 35. EPA does not directly respond to this
argument, only articulating that the arbitrary-and-capricious
standard governs its “action in promulgating a rule.” Resp’t’s
Br. 23. Further complicating matters, we have previously
applied the abuse-of-discretion standard in cases where we are
asked to “reverse an agency’s denial of reconsideration.”
AT&T Corp. v. FCC, 

, 509 (D.C. Cir. 2004); see
North Carolina v. EPA, 614 F. App’x 517, 519 (D.C. Cir. 2015)
(unpublished decision applying abuse-of-discretion standard,
citing to AT&T Corp., to review of EPA’s denial of a petition
for reconsideration under § 7607(d)(7)(B)).
                              17

     We need not resolve this issue here, as the same conclusion
in Clean Air Council can be made here. Even under the more
deferential arbitrary-and-capricious standard, EPA erred in
concluding that the two identified issues from Petitioners’
petition for reconsideration did not meet the two requirements
for reconsideration under § 7607(d)(7)(B).

                              A.

     We conclude that the first element of § 7607(d)(7)(B) is
met because it was impracticable to raise either of the two
objections during the period for public comment. The first
element’s impracticability prong – rather than the “arising
after” prong – is met “when the final rule was not a logical
outgrowth of the proposed rule.” Alon Refining Krotz Springs,
Inc. v. EPA, 

, 648 (D.C. Cir. 2019) (per curiam).

       A final rule is the “logical outgrowth” of a
       proposed rule if “interested parties should have
       anticipated that the change was possible, and
       thus reasonably should have filed their
       comments on the subject during the notice-and-
       comment period.” A final rule “fails the logical
       outgrowth test” if “interested parties would
       have had to divine the agency’s unspoken
       thoughts, because the final rule was surprisingly
       distant from the proposed rule.”

Clear Air 

(quoting CSX Transp., Inc. v.
Surface Transp. Bd., 

, 1080 (D.C. Cir. 2009)).

    It is undisputed that the 2013 Proposed Rule was the first
time that EPA introduced the idea of extending work practice
standards beyond the point of generation to several hours after
                              18
generation. It is also undisputed that the 2013 Proposed Rule
and 2013 TSD did not identify which sources were considered
the “best performing,” and neither referenced the section
governing when work practice standards may replace
numerical standards – § 7412(h).

                               1.

     The Final Rule’s reliance on an identified list of best
performing power plants was not a logical outgrowth of the
2013 Proposed Rule. Given the Proposed Rule’s lack of any
mention of best performing sources or § 7412(h), Petitioners
simply could not have anticipated during the comment period
that EPA would convert the 2013 Proposed Rule – based on
data of when all power plants engage pollutant controls – into
an analysis of when then-unidentified best performers could
engage their controls.

     It is true that Petitioners did comment on the lack of best
performing source analysis during the comment period. EPA
argues that this forecloses relief, relying on Portland Cement
Ass’n v. EPA, for the premise that a final rule is a logical
outgrowth of the proposed rule where a petitioner commented
on the issue. 

, 189 (D.C. Cir. 2011) (per curiam).
But in Portland Cement, the Court noted that EPA had sought
comment on the matter and the petitioner had commented. 

does not stand for the proposition that an
individual’s comment in and of itself demonstrates sufficient
notice from EPA to the individual. To the contrary, in
Petitioners’ cited CSX Transportation, this Court recognized
that “notice must come from the [Notice of Proposed
Rulemaking,]” not the comments arising out of 

. Because nothing in the 2013 Proposed Rule indicated
EPA was setting its standards based on the best performing
sources, the Final Rule’s reliance on its newly selected “best
                                     19
performers” cannot be considered a logical outgrowth of the
2013 Proposed Rule. Thus, it was impracticable for Petitioners
to have raised this challenge during the comment period.
Besides, even if we did consider the Petitioners’ mention of a
best-performer analysis as relevant to the logical outgrowth
test, their comment raised that analysis in a different context.
Namely, Petitioners’ comment argued that EPA was required
to apply a best-performer analysis in setting the work-practice
standards; they said nothing of employing a best-performer
analysis in order to determine the duration of startup.

     Even if reliance on any “best performing sources” could
be considered a logical outgrowth, EPA’s process for
identifying those best performing sources was certainly not. To
hold otherwise would place the unreasonable burden on
commenters not only to identify errors in a proposed rule but
also to contemplate why every theoretical course of correction
the agency might pursue would be inappropriate or incorrect.
It was simply impracticable for Petitioners to predict how EPA
would cure the missing “best performer” component and then
submit preemptive attacks on such hypothetical solutions. For
this reason, we find unavailing EPA’s argument that mere
similarities in analytical approaches of the 2013 TSD and the
Final TSD render the latter a logical outgrowth of the former.
EPA gave no notice that it would analyze any best performing
stringency requirements, so EPA cannot now claim that
Petitioners were on notice of how EPA would ultimately
analyze such issues.5 There was simply no opportunity for

5
  Furthermore, EPA’s argument that it applied the “same analytical
approach” in the Final TSD as it did in the 2013 TSD is blatantly erroneous.
Resp’t’s Br. 30 (emphasis added). The process by which EPA went about
discerning which sources were the best performing required additional
calculations and data tinkering beyond what EPA did in the 2013 TSD. For
example, the Final TSD eliminated data from 563 startup events in which
“electricity generation lasted less than 4 hours before fossil fuel combustion
                                 20
Petitioners to weigh in on whether additional factors beyond
what was considered in the 2013 TSD should be considered
when determining which sources are the best performing.

     To sum up, Petitioners were not given the opportunity to
comment on, propose revisions to, or otherwise challenge the
process for selecting the “best performing” power plants that
EPA first unveiled in the Final TSD. In this respect, because
the final rule was not a logical outgrowth of the 2013 Proposed
Rule, this first objection satisfies the first element of
§ 7607(d)(7)(B).

                                      2.

     Turning to their second objection, Petitioners argue that
EPA did not disclose “the critical reasoning behind its
measurability analysis until its [F]inal [R]ule.” Pet’rs’
Opening Br. 42. It was only then, say Petitioners, that EPA
asserted “that the point at which power plants begin to operate
their pollution controls can be used as a proxy for when they
can measure emissions,” 

technologically and
economically feasible methodology exists to measure HAP
emissions up until that point, see § 7412(h)(2)(B).

     As we demonstrate above, the Final Rule is the first time
EPA connects an EGU’s ability to remove pollutants with its
ability to measure HAP emissions. 

EPA is
correct that the 2013 TSD articulated that it was using certain
removal efficacies as a proxy for the end of startup. The end
of startup is clearly the beginning of when emissions are to be
numerically measured for purposes of complying with existing
numerical limits set forth in the MATS Rule. But nowhere in

ended” and then “calculated the 2-hour rolling average emission rate.”
Final TSD, at 20. The 2013 TSD does not indicate similar methods.
                                 21
the 2013 TSD or the 2013 Proposed Rule does EPA express its
belief that emissions are immeasurable from the point of
electricity generation until the APCDs become operational.
This key link is a critical statutory requirement for EPA to
lawfully deviate from numerical standards and impose
alternative, non-numerical work practice standards. Simply
put, the 2013 Proposed Rule failed to disclose that it was using
APCDs’ effectiveness as a proxy for measurability.

     Again, recognizing that the 2013 Proposed Rule seriously
lacked compliance with § 7412(h)’s requirements, Petitioners
timely commented on the measurability of pollutants.
Petitioners pointed out that EPA failed to make any finding that
numerical standards remained infeasible after the generation of
electricity (the startup end point in the first definition) to justify
work practice standards during the extended period of time in
the alternative definition. Petitioners’ comment also argued
that feasible measurement methodologies existed. EPA again
pounces on this comment, arguing that Petitioners’ timely
comment on feasibility is sufficient to show that it was not
impracticable for Petitioners to timely object to the use of
APCDs as a proxy for measurability.

     We conclude that EPA clearly erred in faulting Petitioners’
failure to divine from the 2013 TSD that EPA would later
conclude under § 7412(h)(2)(B) that specific emission
measurements during startup show that HAP emissions cannot
be measured in a technologically and economically feasible
way.

     Despite “conced[ing] that the 2013 Proposal could have
been more explicit” on this point, EPA doubles down, arguing
that the challenged premise – “that it was not feasible to
establish numeric limitations” for HAPs until APCDs became
operational – “was central” to the 2013 Proposal, Resp’t’s Br.
                                     22
32, even though the 2013 Proposal lacks any reference to
measurement feasibility or § 7412(h).6 But commenters do not
have to be mind readers, and this is exactly the kind of
“divin[ation]” of “unspoken thoughts” that the logical
outgrowth test rejects. Clear Air 

(quoting CSX Transp., 

). This second
objection also passes the first § 7607(d)(7)(B) element.

                                      B.

     The parties agree that § 7607(d)(7)(B)’s second element,
central relevance, asks whether the objections provide
substantial support for the argument that the regulation should
be revised. Both of Petitioners’ objections surpass that hurdle,
as they go to the very legality of the Final Rule’s § 7412(h)
work practice standards for the extended startup period. See
Kennecott Corp. v. EPA, 

, 1019 (D.C. Cir. 1982)
(“Because the reasonableness and accuracy of the forecast data
is critical to whether a smelter can qualify for an NSO,
[petitioners’] objections to that data, if well-founded, would
clearly have been ‘of central relevance.’” (citing
§ 7607(d)(7)(B))).

     EPA argues that Petitioners’ objections cannot be centrally
relevant because they lack merit. We are not so convinced.
What is clear is that if different best performers are selected,
the extended startup definition’s end point would have to be

6
   Intervenors make the same argument, claiming “measurement
impracticability was a primary issue underlying the entire reconsideration
proceeding.” Intervenors’ Br. at 10. This may indeed be true for the MATS
Rule’s definition of startup – and Petitioners do not challenge the feasibility
of measurability up until generation – but EPA’s radio silence on the matter
after proposing a new, extended endpoint of startup deprived Petitioners of
the opportunity to comment on this issue as it applied to an extended period
of startup.
                              23
recalculated, and unless EPA demonstrates that measurability
is infeasible until APCDs become operational, work practice
standards cannot be applied. These issues certainly meet the
“central relevance” requirement. § 7607(d)(7)(B).

                             VI.

     For the foregoing reasons, the Court grants Petition No.
16-1349, vacates EPA’s denial of the petition for
reconsideration, and remands to the agency for reconsideration.
Petition Nos. 15-1015 and 16-1169 are dismissed.

                                                   So ordered.