State of New York v. EPA

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., 572 U.S. 489, 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, 777 F.3d 456, 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, 895 F.3d 90, 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 id. § 7407(b)–(e). 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. Id. § 7410(a)(2)(D)(i). That subpart is known as the “Good Neighbor Provision.” See Wisconsin v. EPA, 938 F.3d 303, 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). Id. § 7410(c)(1).

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.” Id. § 7607(d)(1)(N), (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, 249 F.3d 1032, 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 Power, 249 F.3d at 1040–1044.

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

B

Over 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. Id. § 50.19. 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 id. If the region fails to meet the compliance deadline, the EPA will reclassify it to a higher severity level. See id. § 7511(b)(2). 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 id. at 44,243 tbl.2. The reclassification to serious nonattainment triggered a July 2021 attainment deadline. Id. at 44,244.

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, 2019 WL 3337996, 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. Id.

In evaluating 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. Id. at 56,062.

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. Id. at 56,080; see also id. at 56,081 & 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.” Id. at 56,090. 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.

Id. at 56,090.

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. Id. at 56,090.

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.