United States Court of Appeals
         FOR THE DISTRICT OF COLUMBIA CIRCUIT



Argued May 7, 2020                     Decided July 14, 2020

                        No. 19-1231

                STATE OF NEW YORK, ET AL.,
                       PETITIONERS

                             v.

  ENVIRONMENTAL PROTECTION AGENCY AND ANDREW
 WHEELER, IN HIS OFFICIAL CAPACITY AS ADMINISTRATOR OF
    THE U.S. ENVIRONMENTAL PROTECTION AGENCY,
                      RESPONDENTS

               ADIRONDACK COUNCIL, ET AL.,
                     INTERVENORS


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


    Steven C. Wu, Deputy Solicitor General, Office of the
Attorney General for the State of New York, argued the cause
for petitioners. With him on the briefs were Letitia James,
Attorney General for the State of New York, Barbara D.
Underwood, Solicitor General, Morgan A. Costello and
Claiborne E. Walthall, Assistant Attorneys General, Gurbir S.
Grewal, Attorney General for the State of New Jersey, Lisa
Morelli, Deputy Attorney General, and Christopher G. King,
Senior Counsel, New York City Law Department.
                               2
     Joshua A. Berman argued the cause for petitioners-
intervenors Sierra Club, et al. With him on the briefs were Sean
H. Donahue, Graham G. McCahan, Vickie L. Patton, and
Liana James.

    Samara M. Spence, Attorney, U.S. Department of Justice,
argued the cause for respondent. With her on the brief were
Jonathan Brightbill, Principal Deputy Assistant Attorney
General, and Abirami Vijayan and Stephanie L. Hogan,
Counsel, U.S. Environmental Protection Agency. Sarah A.
Buckley, Attorney, U.S. Department of Justice, entered an
appearance.

    David M. Flannery, Kathy G. Beckett, Edward L. Kropp,
Samuel B. Boxerman, Samina M. Bharmal, David M.
Friedland, Laura K. McAfee, E. Carter Chandler Clements,
Norman W. Fichthorn, Steven P. Lehotsky, and Michael B.
Schon were on the brief for respondents-intervenors Midwest
Ozone Group, et al. Laura M. Goldfarb, Amy M. Smith and
Peter Tolsdorf entered appearances.

    Joseph A. Newberg II and Mary Ann Lee were on the brief
for amicus curiae Commonwealth of Kentucky, Energy and
Environment Cabinet in support of respondents.

   Before: SRINIVASAN, Chief Judge, and GRIFFITH and
MILLETT, Circuit Judges.

    Opinion for the Court filed by Circuit Judge MILLETT.

    Concurring opinion filed by Circuit Judge GRIFFITH.

    MILLETT, Circuit Judge: Air pollutants do not stay still.
Nor do they respect state borders. That has created a “complex
problem”—namely that “air pollution emitted in one State[]
                               3
[can] caus[e] harm in other States.” EPA v. EME Homer City
Generation, L.P., 

, 495 (2014).

     This case involves a challenge to the Environmental
Protection Agency’s asserted failure to address cross-border
pollution under the Clean Air Act’s Good Neighbor Provision,
42 U.S.C. § 7410(a)(2)(D)(i). The State of New York
petitioned the EPA to find that power-generating and other
facilities in nine different States were violating the Good
Neighbor Provision by producing emissions that contributed
significantly to New York’s difficulty attaining or maintaining
compliance with the 2008 and 2015 National Ambient Air
Quality Standards for ozone.

     The EPA denied New York’s petition on the ground that it
failed to meet the agency’s standard for establishing a violation
of the Good Neighbor Provision and, in particular, for
demonstrating that cost-effective controls could be imposed on
the pollution sources. The State of New York, the State of New
Jersey, and the City of New York petitioned this court for
review.

    We grant the petition for review. The EPA offered
insufficient reasoning for the convoluted and seemingly
unworkable showing it demanded of New York’s petition. In
addition, the EPA’s finding that New York did not have an air
quality problem under the 2008 National Ambient Air Quality
Standards for ozone relied on two faulty interpretations of the
Clean Air Act that have since been invalidated. See
Maryland v. EPA, No. 18-1285, slip op. at 25–34 (D.C. Cir.
May 19, 2020). For those reasons, we vacate the EPA’s
decision and remand for further proceedings not inconsistent
with this opinion.
                                4
                                I

                                A

     The Clean Air Act, 42 U.S.C. §§ 7401 et seq., directs the
EPA to establish and periodically revise National Ambient Air
Quality Standards, or NAAQS, that set the maximum allowable
concentrations for various air pollutants, including ozone. 42
U.S.C. §§ 7408(a), 7409. To measure compliance with the
NAAQS, the EPA, “in coordination with state governments,
divides the country geographically into ‘air quality control
regions.’” Natural Res. Defense Council v. EPA, 

,
458 (D.C. Cir. 2014) (formatting modified) (quoting 42 U.S.C.
§ 7407). While some air quality control regions “lie within a
single state[,] * * * others encompass portions of two or more
states.” Maryland, slip op. at 6 (quoting Delaware Dep’t of
Natural Res. & Environmental Control v. EPA, 

,
94 (D.C. Cir. 2018)).

     Once new air quality standards go into effect, each State
must develop an implementation plan to ensure the standards
are met within the State’s air quality control region. See 42
U.S.C. § 7410(a)(1); see also

In addition,
those plans must prohibit “any source or * * * emissions
activity within the State from emitting any air pollutant in
amounts which will * * * contribute significantly to
nonattainment in, or interfere with maintenance by, any other
State with respect to” the NAAQS.

That subpart is known as the “Good Neighbor Provision.” See
Wisconsin v. EPA, 

, 309–319 (D.C. Cir. 2019).

     Under Section 110 of the Clean Air Act, the EPA must
review each State’s implementation plan and ensure its
compliance with statutory requirements, including the Good
Neighbor provision. See 42 U.S.C. § 7410(k)(1)–(4). If a State
fails to timely correct a deficiency in its plan, then the EPA will
                                 5
promulgate a federal implementation plan for the relevant
region(s).

     Section 126(b) of the Clean Air Act, 42 U.S.C. § 7426(b),
creates an additional mechanism for enforcing the Good
Neighbor Provision. It authorizes affected States or local
subdivisions to petition the EPA to make a “finding that any
major source or group of stationary sources emits or would
emit any air pollutant in violation of the prohibition of [the
Good Neighbor Provision.]” Id.1

     Under Section 126(b), the EPA must generally respond to
the petition “[w]ithin 60 days after receipt of [such] petition
* * * and after public hearing[.]” 42 U.S.C. § 7426(b). The
agency may, however, grant itself an extension of up to six
months “upon a determination that such extension is necessary
to afford the public, and the agency, adequate opportunity to
carry out the purposes of th[e] subsection.”

(d)(10).

     If an existing pollution source in another jurisdiction is
found to be in violation of the Good Neighbor Provision, that
source generally must cease operation within three months. 42
U.S.C. § 7426(c). But the EPA may allow continued operation
if the “source complies with such emission limitations and
compliance schedules * * * as may be provided by the
Administrator to bring about compliance * * * as expeditiously


    1
      Section 126(b) cross-references Section 110(a)(2)(D)(ii) of the
Clean Air Act, 42 U.S.C. § 7410(a)(2)(D)(ii). But that is understood
to be a scrivener’s error. Appalachian Power Co. v. EPA, 

, 1040–1044 (D.C. Cir. 2001). For present purposes, the proper
cross-reference is the Good Neighbor Provision, 42 U.S.C.
§ 7410(a)(2)(D)(i). See Appalachian 

–1044.
                                  6
as practicable, but in no case later than three years after the date
of such finding.”

time, the EPA has promulgated increasingly
stringent ozone standards.2 As relevant here, in 2008, the EPA
lowered the acceptable ozone level, measured over eight hours,
from 80 parts per billion to 75 parts per billion. 40 C.F.R.
§ 50.15. And in 2015, it promulgated an even more restrictive
ozone standard of 70 parts per billion.

Both the
2008 NAAQS and the more stringent 2015 NAAQS remain in
effect with differing deadlines for compliance.

     Depending on the degree of nonattainment, the Clean Air
Act provides a deadline by which each air quality control
region must achieve compliance. See 42 U.S.C. § 7511(a)(1),
(b)(1). The more severe the noncompliance, the more time the
region has to remedy the problem. See

region fails to
meet the compliance deadline, the EPA will reclassify it to a
higher severity level. See

That, in turn,
automatically extends the deadline for compliance to the
attainment date for that higher level.

    As relevant here, areas in “serious” nonattainment of the
2008 NAAQS have a statutory attainment deadline of 2021.
See Determination of Attainment and Reclassification for 2008
Ozone NAAQS, 84 Fed. Reg. 44,238, 44,244 (Aug. 23, 2019).
Areas in “moderate” nonattainment of the 2015 NAAQS have

     2
      See, e.g., 44 Fed. Reg. 8202, 8217 (Feb. 8, 1979) (setting the
primary ozone standard at 120 parts per billion); 62 Fed. Reg.
38,856, 38,885 (July 18, 1997) (at 80 parts per billion); 73 Fed. Reg.
16,436, 16,483 (March 27, 2008) (at 75 parts per billion); 80 Fed.
Reg. 65,292, 65,362 (Oct. 26, 2015) (at 70 parts per billion).
                                 7
a 2024 deadline for compliance. Response to Section 126(b)
Petition from New York, 84 Fed. Reg. 56,058, 56,072 n.48
(Oct. 18, 2019).

                                 C

     The New York-Northern New Jersey-Long Island, New
York-New Jersey-Connecticut Area (“New York Metropolitan
Area” or “Area”) is a multistate air quality control region. It is
currently in “serious” nonattainment of the 2008 ozone
NAAQS, having twice failed to meet previously applicable
statutory deadlines for attainment. See Determination of
Attainment and Reclassification for 2008 Ozone NAAQS, 84
Fed. Reg. at 44,238, 44,244 (reclassifying seven areas,
including the New York Metropolitan Area, to serious
nonattainment); see also

tbl.2.                The
reclassification to serious nonattainment triggered a July 2021
attainment deadline.

    The Area is also in “moderate” nonattainment of the 2015
NAAQS, with a 2024 deadline for attainment. Additional Air
Quality Designations for 2015 Ozone NAAQS, 83 Fed. Reg.
25,776, 25,821 (June 4, 2018); see 84 Fed. Reg. at 56,072 n.48.

     In March 2018, New York filed a Section 126(b) petition
(“Petition”) that asked the EPA to find that approximately 350
sources of nitrogen oxides in nine States were contributing
significantly to nonattainment in the New York Metropolitan
Area under the 2008 and 2015 NAAQS. J.A. 58, 60, 76.3 The



    3
       The Petition also alleged that these out-of-state sources were
interfering with attainment in Chautauqua County, New York.
J.A. 60.    New York’s and the Intervenor Environmental
Associations’ arguments before this court focus exclusively on the
                                 8
Petition pointed to Illinois, Indiana, Kentucky, Maryland,
Michigan, Ohio, Pennsylvania, Virginia, and West Virginia as
the sources of infiltrating ozone pollution. J.A. 60. New
York’s modeling projected that the nine States would
contribute at least one percent of the 2008 NAAQS (that is, at
least 0.75 parts per billion) to at least one nonattaining ozone
monitor in the New York Metropolitan Area. J.A. 60, 69.
Within those nine States, the Petition focused the need for
regulation on facilities that emit at least 400 tons of nitrogen
oxides per year. See J.A. 60, 68, 76.4

     Rather than resolving the Petition within the 60-day
statutory deadline, 42 U.S.C. § 7426(b), the EPA granted itself
a six-month extension of time. Extension of Deadline, 83 Fed.
Reg. 21,909, 21,910–21,912 (May 11, 2018); see also 42
U.S.C. § 7607(d)(1)(N), (d)(10) (authorizing the EPA to grant
itself an extension under certain circumstances).

     When the EPA missed that extended deadline, New York
filed suit to compel a decision. The United States District
Court for the Southern District of New York ordered the EPA
to grant or deny the Petition by September 2019. See New York




New York Metropolitan Area. So we do not address the EPA’s
findings with respect to Chautauqua County.
    4
        New Jersey, like the nine listed States, was projected to
contribute at least 0.75 parts per billion to ozone levels in the New
York Metropolitan Area. J.A. 69. The Petition nevertheless did not
list New Jersey as a potential violator because New York’s modeling
indicated that the 400-ton-per-year sources in New Jersey did “not
significantly contribute to any nonattainment or maintenance
monitors.” J.A. 71.
                                9
v. Wheeler, No. 19-CV-3287, 

, at *2
(S.D.N.Y. July 25, 2019).

    After undertaking notice and comment procedures and
conducting a public hearing, the EPA issued a final decision
denying the Petition on September 20, 2019. 84 Fed. Reg. at
56,093. The decision was published in the Federal Register the
next month.

the Petition, the EPA applied a four-step
framework derived from prior rulemakings on the interstate
transport of ozone. 84 Fed. Reg. at 56,058, 56,062–56,063.
Those steps are: (1) identifying downwind areas that have
trouble attaining or maintaining the NAAQS; (2) determining
which upwind States’ emissions are “linked” to downwind air
quality problems; (3) ascertaining which of those linked States’
upwind sources “significantly contribute” to nonattainment or
interfere with maintenance of the NAAQS in a downwind area;
and (4) implementing emission reductions/budgets within the
upwind States.

     The EPA imposed the burden of satisfying each of those
steps on New York as the Section 126(b) petitioner. See 84
Fed. Reg. at 56,069–56,070. The EPA also construed
Section 126(b) as allowing States to challenge interstate
transport of pollution only when it impacted downwind
receptors “within their geographical borders,” even if the
upwind pollutants impede attainment in the air quality region
of which the State is a part.

see also

& n.70.

    With respect to Step 1 of the four-part framework, the EPA
found an air quality problem in the New York Metropolitan
Area under the 2015 NAAQS. 84 Fed. Reg. at 56,080–56,081.
But it found no such attainment problem under the 2008
NAAQS.
 Id. The EPA 
reached that conclusion by treating
                              10
2023 as the relevant year for evaluating the existence of an air
quality problem under the 2008 NAAQS. Id.; see also
 id. at
56,074 
(“The EPA disagrees that it is inappropriate to rely on
the 2023 modeling because it does not align with a particular
attainment date.”). On that basis, the EPA denied the portion
of the Petition seeking to enforce the 2008 NAAQS. Rather
than project air quality in 2021—the year by which attainment
was legally required—the EPA found that “New York has not
demonstrated that there will be a nonattainment or maintenance
problem” in 2023. The EPA’s own analysis also projected no
air quality problems under the 2008 ozone NAAQS by 2023.

Id. at 56,080–56,081.
     The EPA agreed with New York, though, that the New
York Metropolitan Area would likely be in nonattainment of
the 2015 NAAQS in 2023. See 84 Fed. Reg. at 56,080–56,081.

     At Step 2, the EPA “assum[ed], without deciding” that the
emissions in the nine States identified in the Petition were
“linked” to air quality problems in the New York Metropolitan
Area. 84 Fed. Reg. at 56,082.

     At Step 3, the EPA denied the Petition in full based on
New York’s failure to carry its assigned burden of establishing
significant contributions from upwind sources under either the
2008 or 2015 NAAQS. In particular, the EPA decided that the
Petition’s “assessment of whether the sources” could be
“further controlled through implementation of cost-effective
controls [was] insufficient[.]” 84 Fed. Reg. at 56,059; see also

id. at 56,088–56,089.
     The EPA reasoned that New York could have met its
evidentiary burden of demonstrating the availability of cost-
effective controls by producing “one or more of the following”
analyses:
                                11
      (i) Verifying that the named sources whose emissions
      are those from the most recent emissions inventory
      continue to emit [nitrogen oxides] at the same rate or
      continue to operate; (ii) describing or quantifying
      potentially available emissions reductions from the
      named        sources       (i.e.,       the      control
      technologies/techniques and the costs of those control
      technologies/techniques);       (iii) describing     the
      downwind air quality impacts of controlling the
      named sources relative to other sources; or
      (iv) providing information on the relative cost of the
      available emissions reductions and whether they are
      less expensive than other reductions from other
      sources.

84 Fed. Reg. at 56,088–56,089.

     The EPA then added that it could “[]not determine whether
it would be appropriate to regulate any of the hundreds of”
named sources unless they were all “compared to one another
or * * * compared to other, unnamed sources in the same
upwind states or in other states.” 84 Fed. Reg. at 56,090. In
the EPA’s view, this comparison must involve

      identifying the current operating status of each named
      facility, the magnitude of emissions from each
      emitting unit within each named facility, the existing
      controls on each of these emissions units, additional
      control options on each emissions unit, the cost of
      each potential control option, the emissions
      reductions potential resulting from the installation of
      controls, and potential air quality impacts of
      emissions reductions.


Id.
                              12
     In 
addition to finding that New York failed to carry its
burden with respect to Step 3, the EPA concluded that a prior
Cross-State Air Pollution Rule Update, which we shall refer to
as the 2008 Update Rule, had fully addressed any Good
Neighbor Provision violations arising in the nine named States.
84 Fed. Reg. at 56,089 (“[T]he EPA has now determined * * *
that the emissions reductions required under the * * * [2008]
Update [Rule] fully address the good neighbor requirements
with respect to the 2008 ozone NAAQS for all the States named
in the [P]etition.”); see Cross-State Air Pollution Rule Update
for the 2008 Ozone NAAQS, 81 Fed. Reg. 74,504 (Oct. 26,
2016).

     The State of New York, the State of New Jersey, and the
City of New York petitioned this court for review. Three
environmental organizations intervened in support of the
petitioning States.5 Several parties (collectively, “Industry
Intervenors”) separately intervened in support of the EPA.6 On
December 20, 2019, this court granted expedited review.

                              II

    This court has jurisdiction under Section 307(b)(1) of the
Clean Air Act. See 42 U.S.C. § 7607(b)(1); see also Sierra
Club v. EPA, 
955 F.3d 56
, 61 (D.C. Cir. 2020).




    5
     The intervening environmental organizations are Adirondack
Counsel, Environmental Defense Fund, and Sierra Club.
    6
      The Industry Intervenors are: Midwest Ozone Group, the Air
Stewardship Coalition, GenOn Holdings, LLC, the National
Association of Manufacturers, and the Chamber of Commerce of the
United States of America.
                               13
      We may set aside the EPA’s decision under Section 126 if
it is “arbitrary, capricious, an abuse of discretion, or otherwise
not in accordance with law.” 42 U.S.C. § 7607(d)(1)(N),
(d)(9); see also Maryland, slip op. at 18 (“[W]e apply the same
standard of review under the Clean Air Act as we do under the
Administrative Procedure Act.”) (quoting Allied Local & Reg’l
Mfrs. Caucus v. EPA, 
215 F.3d 61
, 68 (D.C. Cir. 2000)).

                               III

     The EPA’s reasons for rejecting New York’s Petition were
arbitrary and capricious in two respects.

     First, the EPA failed to provide a reasoned explanation for
why, under Step 3 of its framework, the Petition failed to show
that the named sources contributed significantly to downwind
nonattainment. The EPA’s test, at best, was a moving target
and, at worst, demanded likely unattainable standards of proof.

    Second, binding circuit precedent flatly rejects the two
grounds on which the EPA relied in deciding, under Step 1, that
the New York Metropolitan Area did not have a cognizable air
quality problem under the 2008 NAAQS.

                                A

    The EPA denied the Petition in full, as to compliance with
both the 2008 and 2015 NAAQS, under Step 3. Specifically,
the EPA pointed to perceived inadequacies in New York’s
evidence that cost-effective emission reductions could be
imposed at the sources of the offending contamination. 84 Fed.
Reg. at 56,059, 56,088–56,089.

    The central problem is that the standard by which the EPA
deemed New York’s cost-effectiveness showing to be
insufficient is impossible to discern because the explanation
                               14
kept shifting. And if the standard truly means what the EPA’s
decision at times says, it would be nigh impossible to meet.

     First, the EPA’s decision denying the Petition said that
New York could have carried its burden by undertaking “one
or more of” four possible analyses. 84 Fed. Reg. at 56,088–
56,089. By way of reminder, those were: (i) “[v]erifying that
the named sources * * * continue to emit [nitrogen oxides] at
the same rate or continue to operate”; (ii) “describing or
quantifying potentially available emissions reductions from the
named sources”; (iii) “describing the downwind air quality
impacts of controlling the named sources relative to other
sources”; and (iv) evaluating “the relative cost of the available
emissions reductions and whether they are less expensive than
other reductions from other sources.”
 Id. at 56,088–56,089.
     Taking the EPA at its word, the Petition’s satisfactory
demonstration of any “one” of those prongs should have
sufficed. 84 Fed. Reg. at 56,088.

     Yet the EPA denied the Petition without any reasoned
explanation as to how New York failed to satisfy the first of the
four analyses, in particular with respect to sources that are
electric generating units, or “EGUs.” J.A. 76 (“Appendix B
includes average emission rates by EGU facility for the 2014
to 2016 period (these data are unavailable for non-EGUs)[.]”);
see also J.A. 90–92 (listing EGU emission rates). The agency’s
decision never offered a coherent explanation for why it
nonetheless rejected the Petition in full at Step 3. It simply
went on to discuss more potential hurdles for New York’s
Petition to clear. 84 Fed. Reg. at 56,089–56,090.

    At oral argument, the agency could not say whether New
York had satisfied the first of the four listed analyses. Counsel
simply said it was “questionable” whether the first analytical
option had been met. Oral Arg. Tr. 41:22–24. But the EPA
                              15
cannot sensibly reject a petition on the ground that it has not
yet figured out if the information provided is sufficient.

     Perhaps recognizing the problem, the EPA sidestepped the
issue by claiming that the four analyses proposed by the agency
are not “a specific test,” and instead simply “lay[] out the
categories of things [the agency] is looking for.” Oral Arg. Tr.
42:9–11; see also
 id. at 44:5–11. 
So, we are told, when the
EPA said “one or more” in its decision, it actually meant more
than one but maybe not all.
 Id. at 44:10–12.
     We are at a loss. Nowhere does the decision explain which
of these four analyses are necessary or sufficient. Instead, the
EPA faulted New York for failing to provide “this or any such
similar analyses[.]” 84 Fed. Reg. at 56,089. But the decision
never explains what “this” analysis is or why the Petition did
not meet it. Nor did it shed light on what “similar analysis”
would suffice. The EPA’s decision just left the court and future
Section 126(b) petitioners to guess at the agency’s meaning.
The reasoned agency decisionmaking that the Clean Air Act
demands, 42 U.S.C. § 7607(d)(1)(N), (d)(9), does not allow the
EPA to keep moving the finish line.

     Second, the EPA’s decision sent contradictory messages
about whether, or to what extent, New York had to produce a
global comparative analysis of potential emission reductions at
listed and unnamed sources within each of the nine States.

     The EPA lists “describing the downwind air quality
impacts of controlling the named sources relative to other
sources” as one of the four analyses that would have allowed
New York to meet its burden of proof. 84 Fed. Reg. at 56,089
(emphasis added). The EPA then explained that it “cannot
determine whether it would be appropriate to regulate any of
the hundreds of” named sources unless those sources are first
“compared to one another or * * * compared to other, unnamed
                              16
sources in the same upwind states or in other states.”
 

This suggests that a comprehensive comparative
analysis of all sources—named and unnamed—within each
designated State is strictly required.

     In response to commenters’ concerns that such a universal
source comparison requirement was unworkable, the EPA
stated that such “[a]pportioning” of “responsibility for
emissions reductions across many sources in many states is a
key outcome of applying the four-step interstate transport
framework * * * under step 3[.]” 84 Fed. Reg. at 56,089–
56,090 (emphasis added). The EPA then elaborated that the
critical “source comparison necessarily involves” the
petitioning State

    identifying the current operating status of each named
    facility, the magnitude of emissions from each
    emitting unit within each named facility, the existing
    controls on each of these emissions units, additional
    control options on each emissions unit, the cost of
    each potential control option, the emissions
    reductions potential resulting from the installation of
    controls, and potential air quality impacts of
    emissions reductions.

     The EPA concluded that, without such detailed
comparative information about individual sources’
technological and operational capabilities, the agency “cannot
determine whether the sources named in the [Petition] have
available or cost-effective emissions reductions either as
compared to one another or as compared to other, unnamed
sources in the same upwind states or in other states.” 84 Fed.
Reg. at 56,090 (emphasis added). Without that broad swath of
comparative data, the decision said, the “EPA cannot determine
                                17
whether it would be appropriate to regulate any” of the sources
identified in New York’s Petition.
 Id.

     But, despite 
comments flagging the concern, 84 Fed. Reg.
at 56,089, the EPA left entirely unexplained how States are
supposed to obtain the required detailed and technically
particularized internal information from some unknown
number of unnamed and unidentified sources. On top of the
crushing breadth of the demand for information from unnamed
sources across each State, the EPA directed that the analysis
must “necessarily” identify each individual source’s
“magnitude of emissions from each emitting unit within each
named facility,” as well as “the existing controls [and]
additional control options” for each unit, and “the emissions
reductions potential resulting from the installation of controls”
on each unit.
 

     Those analyses—especially determining the emission
reductions that would result from installing a particular control
technology on each emitting unit—would require detailed and
intricate inside knowledge of each facility’s equipment and
operations. Such information is frequently not publicly
available, especially for non-EGUs. See Oral Arg. Tr. 34:13–
17, 59:5–9. Nor did the EPA explain why sources charged with
polluting would hand such information out at the asking.

     At oral argument, the EPA backed away from the plain
language of its decision, insisting that “[i]t is not EPA’s
position that a petitioning state would have to do a comparative
analysis.”     Oral Arg. Tr. 46:6–8.7         Rather, the EPA

    7
       But see 84 Fed. Reg. at 56,088–56,089 (EPA directing New
York to conduct “one or more” of four possible analyses, including
“describing the downwind air quality impacts of controlling the
named sources relative to other sources[,]” and analyzing “the
relative cost of the available emissions reductions and whether they
                                18
characterized such a comparative analysis as simply “one way”
for a petitioning State to show cost-effectiveness.
 Id. at 46:13.
The EPA also insisted that the statement in its decision that
States can demonstrate cost-effectiveness by “describing the
downwind air quality impacts of controlling the named sources
relative to other sources[,]” 84 Fed. Reg. at 56,089 (emphasis
added), did not suggest a comparative analysis. See Oral Arg.
Tr. 46:14–23.

    This is all quite mystifying. If New York did not have to
undertake the comparative analysis flagged in two of the four
proposed analyses and discussed over two pages of the Federal
Register, and if the EPA cannot definitively say whether New
York has satisfied “one or more” of the preferred analyses, 84
Fed. Reg. at 56,088, then we are left with no coherent
explanation of what was missing from New York’s Petition.
The required analysis seems to be a constantly moving target,
with the words of explanation from the agency variously
meaning and not meaning what they say.

    At bottom, the EPA’s Delphic explanation of New York’s
purported failure to carry its burden of proof—and of even
what that burden is—falls far short of reasoned
decisionmaking. See Environmental Defense Fund v. EPA,
922 F.3d 446
, 454 (D.C. Cir. 2019) (“An agency acts arbitrarily




are less expensive than other reductions from other sources”)
(emphasis added);
 id. at 56,090 
(EPA stating that New York has
provided insufficient information to allow it to “determine whether
the sources named in the New York [P]etition have available or cost-
effective emissions reductions either as compared to one another or
as compared to other, unnamed sources in the same upwind states or
in other states”) (emphasis added).
                               19
and capriciously when it offers * * * unreasoned justifications
for a decision.”).

     Third, in addition to dismissing New York’s cost-
effectiveness analysis as insufficient to support a Step 3
finding, the decision claims “that the emissions reductions
required under the * * * [2008] Update [Rule] fully address the
good neighbor requirements with respect to the 2008 ozone
NAAQS for all the States named in the [P]etition.” 84 Fed.
Reg. at 56,089. The EPA added that the electric generating unit
control technologies identified by New York and by certain
commenters had already been accounted for in the 2008 Update
Rule’s trading scheme.
 Id. at 56,092.
     The EPA has abandoned the first contention—that the
2008 Update Rule fully satisfies the Good Neighbor
requirements under the 2008 NAAQS. And for good reason.
This court has held that the 2008 Update Rule did not satisfy
any States’ Good Neighbor obligations. See New York v. EPA,
781 F. App’x 4, 6–7 (D.C. Cir. 2019) (vacating the EPA’s
Determination Regarding Good Neighbor Obligations for the
2008 Ozone National Ambient Air Quality Standard, 83 Fed.
Reg. 65,878 (Dec. 21, 2018)); see also 
Wisconsin, 938 F.3d at
309
, 313–318 (holding that the 2008 Update Rule violated the
Clean Air Act by allowing upwind States to continue
contributing to downwind air quality problems “beyond the
statutory deadlines by which downwind States must
demonstrate their attainment”).

    But the EPA continues to press its second rationale—that
the 2008 Update Rule’s emissions trading scheme fully
addressed any Good Neighbor Provision obligations associated
with electric generating units. To be sure, this court’s decision
in Wisconsin upheld the reasonableness of the Update’s cap-
                               20
and-trade scheme and its specific emissions budgets.
Wisconsin, 938 F.3d at 329
–335.

     But the EPA is incorrect to argue that Wisconsin also held
that the 2008 Update Rule comprehensively addressed all Good
Neighbor Provision obligations associated with electric
generating units. To the contrary, Wisconsin described the
2008 Update Rule as only a “first, partial step to addressing a
given upwind State’s significant 
contribution.” 938 F.3d at 313
(quoting 2008 Update Rule, 81 Fed. Reg. 74,522). Indeed, the
EPA itself announced that full resolution of the Good Neighbor
Provision obligations would require consideration of “further
EGU reductions[.]”
 Id. (quoting 2008 
Update Rule, 81 Fed.
Reg. at 74,522); see also 2008 Update Rule, 81 Fed. Reg. at
74,521 (“To evaluate full elimination of a state’s significant
contribution to nonattainment or interference with
maintenance, * * * further EGU reductions that are achievable
after 2017 should be considered.”). So the 2008 Update Rule
does nothing to salvage the EPA’s Step 3 rejection of New
York’s Petition.

     Finally, the Industry Intervenors urge this court to override
the EPA’s finding that the New York Metropolitan Area is
likely to face compliance issues in 2023 with respect to the
2015 NAAQS, and to uphold the EPA’s denial of the Petition,
at least in part, based on this alternative ground. Industry
Intervenors Br. 36–39. The Industry Intervenors also argue
that the court should uphold the EPA’s decision on the alternate
ground that the sources listed in the Petition do not qualify as a
“group” under Section 126. Industry Intervenors Br. 7–17.
Because the agency did not rest its decision on either of those
bases, we reject both arguments. See USPS v. NLRB, 
969 F.2d
1064
, 1069 (D.C. Cir. 1992) (“[W]e reject [the intervenor’s]
endeavor to achieve disposition of this case on a rationale [not]
set forth by the agency itself.”) (internal quotation marks
                              21
omitted); see also SEC v. Chenery Corp., 
318 U.S. 80
, 93–95
(1943).

                               B

     The EPA’s denial of the Petition at Step 1 with respect to
the 2008 NAAQS is equally unsustainable. Recall that Step 1
involves identifying downwind areas that have trouble
attaining or maintaining the NAAQS. 84 Fed. Reg. at 56,062.
The EPA found that the Petition failed to demonstrate that the
New York Metropolitan Area had an air quality problem under
the 2008 NAAQS. 84 Fed. Reg. at 56,080–56,081.

    That decision was legally flawed in two ways.

     For starters, the EPA’s decision erroneously treated 2023
as the relevant Step 1 reference year for the 2008 NAAQS,
even though the New York Metropolitan Area is subject to a
2021 nonattainment deadline. 84 Fed. Reg. at 56,080. This
court rejected that very proposition in Wisconsin, holding that
the 2008 Update Rule violated the Clean Air Act by allowing
“upwind States to continue their significant contributions to
downwind air quality problems beyond the statutory deadlines
by which downwind States must demonstrate their
attainment[.]” 938 F.3d at 309
; see also
 id. at 315–316 
(The
Clean Air Act “cannot reasonably be understood to enable
upwind States to continue their significant contributions
outside of the statutory timeframe by which downwind areas
must achieve attainment, much less continue those
contributions with no deadline at all.”). This means that the
agency “must evaluate downwind air quality at [the downwind
States’ statutory] deadline”—here, 2021—and “not at some
later date[,]” like 2023, as the EPA did in this case. Maryland,
slip op. at 33.
                               22
     The second problem with the EPA’s analysis of the 2008
NAAQS portion of New York’s Petition is that it unreasonably
“interpreted * * * [S]ection 126(b)’s petition authority as
limited to states * * * seeking to address interstate transport of
pollution impacting downwind receptors within their
geographical borders.” 84 Fed. Reg. at 56,080 (emphasis
added); see also
 
id. at 56,081 
& n.70.

     Our recent decision in Maryland firmly closed the door on
that proposition, “at least” with respect to monitors like New
York’s that are “located in a multistate nonattainment area that
includes the petitioning state.” Maryland, slip op. at 27–28
(internal quotation marks omitted). To hold otherwise would
have created an untenable incongruity in the statute—placing
States “in regulatory limbo” where they are subject to
regulatory burdens based on their air quality control region’s
nonattainment, “yet unable to avail [themselves] of the
intended remedy for addressing upwind contributions” to that
nonattainment.
 Id. at 26.
                               IV

     For the foregoing reasons, we grant the petition for review,
vacate the EPA’s denial of the Petition, and remand the case
for further proceedings consistent with this opinion.

     New York asks that we include a 60-day deadline for the
EPA to issue a new decision. See 42 U.S.C. § 7426(b)
(imposing a 60-day deadline for the EPA to act on
Section 126(b) petitions); Natural Res. Defense Council, Inc.
v. Train, 
510 F.2d 692
, 705 (D.C. Cir. 1974) (“The authority to
set enforceable deadlines both of an ultimate and an
intermediate nature is an appropriate procedure for exercise of
the court’s equity powers to vindicate the public interest.”).
                            23
    Although we decline to impose a formal deadline at this
time, we fully expect the EPA to act promptly on remand.

                                               So ordered.
    GRIFFITH, Circuit Judge, concurring: I write separately to
discuss the proper role of section 126 of the Clean Air Act.
Comprehensive remedies for interstate ozone transfer must be
implemented through state implementation plans (SIPs)—or, if
necessary, federal plans—that satisfy the Good Neighbor
provision. By contrast, section 126 is designed for targeted
regulation. New York’s petition, which covers 350 diverse
sources across nine states, is inconsistent with that design.

     Start with the text. Section 126 empowers downwind
jurisdictions to request EPA regulation of “any major source or
group of stationary sources” that causes air-quality problems
under the Good Neighbor provision. 42 U.S.C. § 7426(b). In its
original form, section 126 limited petitioners to a single “major
source.” Pub. L. No. 95-95, § 123(b), 91 Stat. 685, 685 (1977).
Congress added the “group of stationary sources” language
more than a decade later. Clean Air Act Amendments, Pub. L.
No. 101-549, § 109(a)(1)(A), 104 Stat. 2399, 2469 (1990).

     That change “plainly reflected a decision to act against
sources whose emissions, while harmless individually, could
become harmful when combined with others.” Appalachian
Power Co. v. EPA, 
249 F.3d 1032
, 1049 (D.C. Cir. 2001). But
it did not convert section 126 from a rifle to a blunderbuss. The
contrast with the Good Neighbor provision, which applies to
“any source or other type of emissions activity,” remains stark.
42 U.S.C. § 7410(a)(2)(D)(i); cf. Appalachian 
Power, 249 F.3d
at 1049 
(contrasting the Good Neighbor provision’s “broad”
language, which enabled “findings based on aggregate
emissions from within each regulated state,” with section 126,
which “demands that the significant contribution come from a
‘major source or group of stationary sources’”). Had Congress
wished to harmonize the scope of these provisions, it could
have standardized their text. It did not.

     Nor did Congress simply append a plural—“or sources”—
to “any major source.” Instead, it provided that the “stationary
                               2
sources” must constitute a “group.” That choice indicates that
the provision’s intended remedial scope is limited, not
comprehensive. We give an “undefined term its ordinary
meaning,” United States v. Williams, 
836 F.3d 1
, 8 (D.C. Cir.
2016), and dictionaries confirm what common sense suggests:
a “group” is a collection of items that share a common attribute.
Group, WEBSTER’S NEW INTERNATIONAL DICTIONARY 1004
(3d ed. 1981) (defining “group” as “an assemblage of objects
regarded as a unit because of their comparative segregation
from others”); see also Oxford English Dictionary Online,
www.oed.com/view/Entry/81855 (3d ed. 2014) (“[a] number
of things having some related properties or attributes in
common, regarded as forming a unity or classified together
under a general name or description”).

     As we observed in Appalachian Power, this “statutory
language allows the EPA to regulate facilities in upwind states
as a class or category, e.g. all coal-fired power plants in North
Carolina.” 249 F.3d at 1057 
(emphasis added). Petitioners and
EPA presumably enjoy wide latitude when identifying such
commonalities, and we don’t need to set boundaries today. The
crucial point for present purposes is that “group of stationary
sources” describes a set of sources with some unifying
characteristic; it is not merely the plural form of “stationary
source.”

     New York’s petition, which seeks EPA regulation of 350
disparate sources across the Midwest and mid-Atlantic regions,
falls short by any measure. In addition to 130 power plants, the
petition covers oil refineries, natural-gas compressor stations,
chemical plants, steel and paper mills, waste incinerators, and
factories that produce goods ranging from glass to ammunition.
J.A. 90-99. These sources are not united by geography (they
range from Illinois to Maryland), plant technology, industry
                                 3
sector, or any other “class or category.” Appalachian 
Power,
249 F.3d at 1057
.

     The only feature shared by the sources in New York’s
petition is that each emits more than 400 tons of nitrogen
oxides per year. See NYSDEC Detailed Comments at 10, J.A.
495. That arbitrary threshold captures both an Indiana power
plant emitting more than 10,000 tons annually and a Virginia
bottle factory emitting just 412 tons. J.A. 90, 97. If that’s
enough to establish a “group,” the term is all but meaningless.
At oral argument, Petitioners claimed to “stick to a group in the
sense that we are talking just about interstate ozone transport
here, and we are trying to identify all the sources that contribute
to our problem.” Oral Arg. Tr. 11:3-6. In other words,
Petitioners think a petition is sufficiently limited if it targets a
single air pollutant and lists sources that allegedly transgress
the Good Neighbor provision. But that reading equates the duty
to identify a “group of . . . sources” with the substantive
inquiry—effectively erasing the word “group” from the statute.

     Moving from text to context, the Act’s tight deadlines and
harsh remedial scheme confirm that section 126 isn’t designed
to solve comprehensive regional problems. “Congress
specified that the Administrator take final action on a section
126(b) petition very quickly.” New York v. EPA, 
852 F.2d 574
,
578 (D.C. Cir. 1988). EPA must make the requested finding or
issue a denial “[w]ithin 60 days after receipt of any petition . . .
and after public hearing,” 42 U.S.C. § 7426(b), subject to an
optional six-month extension,
 id. § 7607(d)(10). 
In an earlier
case, we found it “reasonable to conclude” that EPA need not
perform the wide-ranging tasks associated with reevaluating
SIPs “in such a short period of time.” New 
York, 852 F.2d at
578
. By the same token, EPA cannot be expected to craft a
definitive solution to a downwind state’s ozone problems
within the timeframe provided by section 126.
                                4
     Compare that quick turnaround to the timelines for SIP
development under section 110. That process affords three
years for states to craft implementation plans, two months for
EPA’s initial “completeness” evaluation, one year for EPA’s
full substantive evaluation, and two years for revisions before
EPA must impose a federal plan. 42 U.S.C. § 7410(a), (c)(1),
(k)(1)-(5). Keep in mind that “the substantive inquiry for
decision is the same in both [section 110 and section 126]
proceedings.” Appalachian 
Power, 249 F.3d at 1047 
(internal
quotation marks omitted). If Congress expected EPA to
conduct the same “substantive inquiry” along such divergent
timelines, then it cannot have intended the scope of the two
inquiries to be the same.

     Next, consider the drastic repercussions of a finding under
section 126(b) that a source “emits or would emit in violation”
of the Good Neighbor provision. 42 U.S.C. § 7426(b).
Petitioners acknowledged at oral argument that section 126, “in
contrast to [section] 110, allows for a . . . more expedited
schedule for compliance.” Oral Arg. Tr. 7:17-19. That’s an
understatement. If EPA makes an affirmative finding, the
offending plant has three months to comply or cease
operations. 42 U.S.C. § 7426(c)(2). Operating beyond that
period is “a violation of this section,”
 id. § 7426(c), 
and
exposes the source to civil penalties,
 id. § 7413(d). 
This swift,
severe, and purely federal intervention is a poor fit for regional
relief and stands in sharp contrast to section 110’s cooperative
approach.

    EPA “may permit the continued operation of a source”
beyond three months, but only if the source “complies with
such emission limitations and compliance schedules” as EPA
may provide.
 Id. § 7426(c). 
But when EPA confronts a petition
seeking regulation of hundreds of diverse sources, that task is
an onerous one for both the regulator and the regulated.
                                5
Consider the petition here. A finding that these sources emit in
violation of the Good Neighbor provision would open a three-
month window for EPA to develop source-specific regulations
for the large power plant in Indiana, the bottle manufacturer in
Virginia, and everything in between. Beyond that three-month
period, each source would face the prospect of shutting down
or risking administrative penalties. Again, these aspects of the
statutory scheme strongly suggest that section 126 authorizes
targeted intervention.

     Section 126 and the Good Neighbor provision differ in
another important respect. The latter obliges a state to regulate
its own sources to prevent harmful interstate transport. See 42
U.S.C. § 7410(a)(2)(D)(i) (“any source or other type of
emissions activity within the State”). EPA steps in only when
state-led efforts fail. See
 id. § 7410(c)(1). 
But section 126
allows a downwind state to request direct federal regulation of
sources beyond its borders. See
 id. § 7426(b) 
(“any major
source or group of stationary sources”). Construed too broadly,
section 126 would make downwind states (with EPA’s help)
the primary regulators of their upwind neighbors. That is
inconsistent with the Clean Air Act’s ethos of cooperative
federalism. I don’t mean to suggest that a state must wait for
the SIP process to conclude before filing a petition. See
Appalachian 
Power, 249 F.3d at 1045
, 1048 (holding that EPA
may “mak[e] [section] 126 findings” and conduct the section
110 process simultaneously because the provisions “operate
independently”). But in the ordinary course, comprehensive
regulation of a state’s polluters should be controlled by the state
itself, not a neighboring government.

     In sum, the textual and contextual evidence convinces me
that section 126 was not designed for petitions of this breadth.
Nor is it simply an alternative route by which a downwind state
can trigger the massive regulatory undertaking associated with
                               6
SIP development under section 110. And I’m not alone in this
assessment: Petitioners call section 126 “a source-specific
tool” for securing “tailored remedies . . . narrower than the
seasonal average ozone budgets established by EPA’s regional
rulemakings under” section 110. Pet’r Br. 57. Quite right. I
likewise agree that “Congress plainly intended for section 126
to provide . . . targeted relief independent of more
comprehensive rulemaking.”
 Id.

     But a 
jurisdiction that invokes this “tailored,” “source-
specific” provision,
 id., should in 
fact tailor its petition to
specific sources—something New York did not even try to do.
Two states took a more appropriate approach in a recent section
126 case decided in our court. See Maryland v. EPA, No. 18-
1285, slip op. at 12-13 (D.C. Cir. May 19, 2020). Maryland’s
petition identified thirty-six sources, all power plants;
Delaware filed four separate petitions, each targeting a single
source.
 Id. Future petitioners 
should follow those examples,
which are more consistent with the text and structure of section
126.

     The opinion for the court properly declines to address
these matters because EPA did not rely on them in its denial of
New York’s petition. But in a future case, EPA may decide to
enforce the textual and structural restrictions on the scope of a
section 126 petition. Prospective petitioners should act
accordingly.