OHIO et al v. ENVIRONMENTAL PROTECTION AGENCY et al on applications for stay No. 23A349. Argued February 21, 2024—Decided June 27, 2024* The Clean Air Act envisions a collaborative effort between States and the federal government to regulate air quality. When the Environmental Protection Agency sets standards for common air pollutants, States must submit a State Implementation Plan, or SIP, providing for the “implementation, maintenance, and enforcement” of those standards in their jurisdictions. See 42 U. S. C. § 7410(a)(1). Because air currents can carry pollution across state borders, States must also design their plans with neighboring States in mind. Under the Act's “Good Neighbor Provision,” state plans must prohibit emissions “in amounts which will . . . contribute signifcantly to nonattainment in, or interfere with maintenance by, any other State” of the relevant air-quality standard. § 7410(a)(2)(D)(i)(I). Only if a SIP fails to satisfy the “applicable requirements” of the Act may EPA issue a Federal Implementation Plan, or FIP, for the noncompliant State that fails to correct the defciencies in its SIP. §§ 7410(k)(3), (c)(1).

In 2015, EPA revised its air-quality standards for ozone, thus triggering a requirement for States to submit new SIPs. Years later, EPA announced its intention to disapprove over 20 SIPs because the agency believed they had failed to address adequately obligations under the Good Neighbor Provision. During the public comment period for the proposed SIP disapprovals, EPA issued a single proposed FIP to bind all those States. EPA designed its proposed FIP based on which emissions-control measures would maximize cost effectiveness in improving ozone levels downwind and on the assumption the FIP would apply to all covered States. Commenters warned that the proposed SIP disapprovals were fawed and that a failure to achieve all the SIP disapprovals as EPA envisioned would mean that EPA would need to reassess the measures necessary to maximize cost-effective ozone-level *Together with No. 23A350, Kinder Morgan, Inc., et al. v. Environmen tal Protection Agency et al., No. 23A351, American Forest & Paper Assn. et al. v. Environmental Protection Agency, et al., and No. 23A384, United States Steel Corp. v. Environmental Protection Agency et al., also on applications for stay.

Page Proof Pending Publication improvements in light of a different set of States. EPA proceeded to issue its fnal FIP without addressing this concern. Instead, EPA announced that its plan was severable: Should any jurisdiction drop out, the plan would continue to apply unchanged to the remaining jurisdictions. Ongoing litigation over the SIP disapprovals soon vindicated at least some of the commenters' concerns. Courts stayed 12 of the SIP disapprovals, which meant EPA could not apply its FIP to those States. A number of the remaining States and industry groups challenged the FIP in the D. C. Circuit. They argued that EPA's decision to apply the FIP after so many other States had dropped out was “arbitrary” or “capricious,” and they asked the court to stay any effort to enforce the FIP against them while their appeal unfolded. The D. C. Circuit denied relief, and the parties renewed their request in this Court. Held: The applications for a stay are granted; enforcement of EPA's rule against the applicants shall be stayed pending the disposition of the applicants' petition for review in the D. C. Circuit and any petition for writ of certiorari, timely sought. Pp. 290–300.

(a) When deciding an application for a stay, the Court asks (1) whether the applicant is likely to succeed on the merits, (2) whether it will suffer irreparable injury without a stay, (3) whether the stay will substantially injure the other parties interested in the proceedings, and (4) where the public interest lies. Nken v. Holder, 556 U. S. 418, 434. When States and other parties seek to stay the enforcement of a federal regulation against them, often “the harms and equities [will be] very weighty on both sides.” Labrador v. Poe, 601 U. S. –––, ––– (Kavalution of applicants' stay requests ultimately turns on the frst question: Who is likely to prevail at the end.

See Nken, 556 U. S., at 434.

Pp. 290–292.

(b) Applicants are likely to prevail on their arbitrary-or-capricious claim. An agency action qualifes as “arbitrary” or “capricious” if it is not “reasonable and reasonably explained.” FCC v. Prometheus Radio Project, 592 U. S. 414, 423. Thus, the agency must offer “a satisfactory explanation for its action[,] including a rational connection between the facts found and the choice made,” and cannot simply ignore “an important aspect of the problem.” Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 43. EPA's plan rested on an assumption that all the upwind States would adopt emissions-reduction measures up to a uniform level of costs to the point of diminishing returns. Commenters posed their concerns that if upwind States fell out of the planned FIP, the point at which emissions- control measures maximize cost-effective downwind air-quality imPage Proof Pending Publication provements might shift. To this question, EPA offered no reasoned response. As a result, the applicants are likely to prevail on their argument that EPA's fnal rule was not “reasonably explained,” Prometheus Radio Project, 592 U. S., at 423, and that it instead ignored “an important aspect of the problem” before it, State Farm Mut. Automobile Ins. Co., 463 U. S., at 43. Pp. 292–294.

(c) EPA's alternative arguments are unavailing. First, EPA argues that adding a “severability” provision to its fnal rule—i. e., providing the FIP would “continue to be implemented” without regard to the number of States remaining—responded to commenters' concerns. But EPA's response did not address those concerns so much as it sidestepped them. Nothing in the fnal rule's severability provision actually addressed whether and how measures found to maximize cost effectiveness in achieving downwind ozone air-quality improvements with the participation of all the upwind States remain so when many fewer States might be subject to the agency's plan. Second, EPA insists that no one raised that concern during the public comment period. The Act's “reasonable specifcity” requirement, however, does not mean a party must rehearse the identical argument made before the agency. Here, EPA had notice of the objection, and its own statements and actions confrm the agency appreciated the concern. Third, EPA argues that applicants must return to EPA and fle a motion asking it to reconsider its fnal rule before presenting their objection in court because the “grounds for [their] objection arose after the period for public comment.” § 7607(d)(7)(B). Nothing requires the applicants to return to EPA to raise (again) a concern EPA already had a chance to address. Pp. 294–298.

Applications for stay granted.

C. J., and Thomas, Alito, and Kavanaugh, JJ., joined. Barrett, J., fled a dissenting opinion, in which Sotomayor, Kagan, and Jackson, JJ., joined, post, p. 300.

Mathura J. Sridharan, Deputy Solicitor General of Ohio, argued the cause for state applicants in No. 23A349. With her on the application for stay and reply were Dave Yost, Attorney General, Zachery P. Keller, Deputy Solicitor General, Theodore E. Rokita, Attorney General of Indiana, James A. Barta, Deputy Solicitor General, Patrick Morri sey, Attorney General of West Virginia, Lindsay S. See, SoCounsel licitor General, and Michael Williams, Principal Deputy Solicitor General.

Catherine E. Stetson argued the cause for industry applicants in all cases. With her on the application for stay and reply in No. 23A350 were Ana M. Gutiérrez, Michael D.

Miller, Brittany M. Pemberton, and Eric D. McArthur.

Jonathan Y. Ellis, Allison D. Wood, Makram B. Jaber, Mi chael B. Schon, Mithun Mansinghani, David M. Flannery, and Kathy G. Beckett fled an application for stay and reply in No. 23A351. John D. Lazzaretti fled an application for stay and reply in No. 23A384.

Deputy Solicitor General Stewart argued the cause for federal respondents in all cases. Solicitor General Prelogar fled a response in opposition to the applications for stay in all cases.

Judith N. Vale, Deputy Solicitor General of New York, argued the cause for state respondents in all cases. With her on the briefs were Letitia James, Attorney General, Barbara D. Underwood, Solicitor General, Elizabeth A. Brody and Stephen J. Yanni, Assistant Solicitors General, and Morgan A. Costello and Claiborne E. Walthall, Assistant Attorneys General, and Sylvia Hinds-Radix, Corporation Counsel for New York City, and by the Attorneys General for their respective jurisdictions as follows: William Tong of Connecticut, Kathleen Jennings of Delaware, Brian L. Schwalb of the District of Columbia, Kwame Raoul of Illinois, Anthony G. Brown of Maryland, Andrea Joy Campbell of Massachusetts, Matthew J. Platkin of New Jersey, Michelle A. Henry of Pennsylvania, and Joshua L. Kaul of Wisconsin. Megan M. Herzog, Sean H. Donahue, David T. Goldberg, Vickie L. Patton, Shaun A. Goho, Neil Gormley, and Kathleen L.

Riley fled a response in opposition to the applications for stay for public interest respondents in all cases.† †Jeremy C. Marwell and Eric Groten fled a brief for the Energy Infrastructure Council as amicus curiae urging grant of application for stay in No. 23A350.

Page Proof Pending Publication

The Clean Air Act envisions States and the federal government working together to improve air quality. Under that law's terms, States bear “primary responsibility” for developing plans to achieve air-quality goals. 42 U. S. C. § 7401(a)(3). Should a State fail to prepare a legally compliant plan, however, the federal government may sometimes step in and assume that authority for itself. § 7410(c)(1). Here, the federal government announced its intention to reject over 20 States' plans for controlling ozone pollution. In their place, the government sought to impose a single, uniform federal plan. This litigation concerns whether, in adopting that plan, the federal government complied with the terms of the Act.

I

A

“The Clean Air Act regulates air quality through a federal- state collaboration.” EME Homer City Generation, L.P. v. EPA, 795 F. 3d 118, 124 (CADC 2015). Periodically, the Environmental Protection Agency (EPA) sets standards for common air pollutants, as necessary to “protect the public health.” §§ 7409(a)(1), (b)(1). Once EPA sets a new standard, the clock starts ticking: States have three years to design and submit a plan—called a State Implementation Plan, or SIP—providing for the “implementation, maintenance, and enforcement” of that standard in their jurisdictions. § 7410(a)(1); see EPA v. EME Homer City Genera tion, L. P., 572 U. S. 489, 498 (2014). Under the Act, States decide how to measure ambient air quality. § 7410(a)(2)(B). States pick “emission limitations and other control measures.” § 7410(a)(2)(A). And States provide for the enforcement of their prescribed measures. § 7410(a)(2)(C).

At the same time, States must design these plans with their neighbors in mind. Because air currents can carry pollution across state borders, emissions in upwind States Page Proof Pending Publication Page Proof Pending Publication sometimes affect air quality in downwind States. See EME Homer, 572 U. S., at 496. To address that externality problem, under the Act's “Good Neighbor Provision,” state plans must prohibit emissions “in amounts which will . . . contribute signifcantly to nonattainment in, or interfere with maintenance by, any other State” of the relevant air-quality standard. § 7410(a)(2)(D)(i)(I).

Because the States bear “primary responsibility” for developing compliance plans, § 7401(a)(3), EPA has “no authority to question the wisdom of a State's choices of emission limitations.” Train v. Natural Resources Defense Council, Inc., 421 U. S. 60, 79 (1975). So long as a SIP satisfes the “applicable requirements” of the Act, including the Good Neighbor Provision, EPA “shall approve” it within 18 months of its submission. § 7410(k)(3); see §§ 7410(k)(1)(B), (k)(2). If, however, a SIP falls short, EPA “shall” issue a Federal Implementation Plan, or FIP, for the noncompliant State—that is, “unless” the State corrects the defciencies in its SIP frst. § 7410(c)(1); EME Homer, 572 U. S., at 498. EPA must also ensure States meet the new air-quality standard by a statutory deadline. See § 7511.

B

A layer of ozone in the atmosphere shields the world from the sun's radiation. See Natural Resources Defense Coun cil v. EPA, 464 F. 3d 1, 3 (CADC 2006). But closer to earth, ozone can hurt more than it helps. Forming when sunlight interacts with a wide range of precursor pollutants, ground- level ozone can trigger and exacerbate health problems and damage vegetation. 80 Fed. Reg. 65299, 65302, 65370 (2015). To mitigate those and other problems, in 2015 EPA revised its air-quality standards for ozone from 75 to 70 parts per billion. Id., at 65293–65294. That change triggered a requirement for States to submit new SIPs. Id., at 65437.

Along the way, EPA issued a guidance document advising States that they had “fexibility” in choosing how to address their Good Neighbor obligations. See EPA, Memorandum, Information on the Interstate Transport State Implementation Plan Submissions for the 2015 Ozone National Ambient Air Quality Standards 3 (Mar. 27, 2018). With that and other guidance in hand, many (though not all) States submitted SIPs. See 84 Fed. Reg. 66612 (2019). And many of the States that did submit SIPs said that they need not adopt emissions-control measures to comply with the Good Neighbor Provision because, among other things, they were not linked to downwind air-quality problems or they could identify no additional cost-effective methods of controlling the emissions beyond those they were currently employing.

See, e. g., 87 Fed. Reg. 9798, 9810 (2022); 87 Fed. Reg. 9545, 9552 (2022); see generally 88 Fed. Reg. 9336, 9354–9361 (2023).

For over two years, EPA did not act on the SIPs it received.

See, e. g., 87 Fed. Reg. 9838, 9845–9851 (2022).

Then, in February 2022, the agency announced its intention to disapprove 19 of them on the ground that the States submitting them had failed to address adequately their obligations under the Good Neighbor Provision.¹ A few months later, the agency proposed disapproving four more SIPs.² Pursuant to the Act, the agency issued its proposed SIP disapprovals for public comment before fnalizing them. See § 7607(d)(3).

C

During that public comment period, the agency proposed a single FIP to bind all 23 States.³ 87 Fed. Reg. 20036, ¹See 87 Fed. Reg. 9463 (2022) (Maryland); 87 Fed. Reg. 9484 (2022) (New York, New Jersey); 87 Fed. Reg. 9498 (2022) (Kentucky); 87 Fed. Reg. 9516 (2022) (West Virginia); 87 Fed. Reg. 9533 (2022) (Missouri); 87 Fed. Reg. 9545 (2022) (Alabama, Mississippi, Tennessee); 87 Fed. Reg. 9838 (2022) (Illinois, Indiana, Michigan, Minnesota, Ohio, Wisconsin); 87 Fed. Reg. 9878 (2022) (Arkansas, Louisiana, Oklahoma, Texas).

²See 87 Fed. Reg. 31443 (California); 87 Fed. Reg. 31470 (2022) (Utah); 87 Fed. Reg. 31485 (2022) (Nevada); 87 Fed. Reg. 31495 (2022) (Wyoming). ³EPA also added three more States: Pennsylvania and Virginia, which had not submitted SIPs, and Delaware, whose SIP, EPA said, it had approved in “error.” 87 Fed. Reg. 20036, 20038 (2022).

Page Proof Pending Publication 20038 (2022). Rather than continue to encourage “ `fexibilit[y]' ” and different state approaches, EPA now apparently took the view that “[e]ffective policy solutions to the problem of interstate ozone transport” demanded that kind of “uniform framework” and “[n]ationwide consistency.” 87 Fed.

Reg. 9841; see 87 Fed. Reg. 20073. The FIP the agency proposed set as its target the reduction of the emissions of one family of ozone precursors in particular: nitrogen oxides. See id., at 20038. And it sought to impose nitrogen- oxide emissions-control measures that “maximized cost- effectiveness” in achieving “downwind ozone air quality improvements.” Id., at 20055; see also id., at 20043.

In broad strokes, here is how EPA's proposed rule worked to eliminate a State's “signifcant contribution” to downwind ozone problems. First, the agency identifed various emissions-control measures and, using nationwide data, calculated how much each typically costs to reduce a ton of nitrogen-oxide emissions. Id., at 20076; see, e. g., id., at 20077–20081. Next, the agency sought to predict how much each upwind State's nitrogen-oxide emissions would fall if emissions-producing facilities in the State adopted each measure. Id., at 20076; see, e. g., id., at 20088–20089; EPA, Ozone Transport Policy Analysis Proposed Rule TSD 22–23 (EPA–HQ–OAR–2021–0668, 2022) (Proposed Ozone Analysis). In making those predictions, EPA often considered data specifc to the emissions-producing facilities in the State, and fed “unit-level and state-level” values into its calculations. See id., at 9–10, 13. Then, the agency estimated how much, on average, ozone levels would fall in downwind States with the adoption of each measure. 87 Fed. Reg.

20076; see, e. g., id., at 20092–20093, 20096–20097; Proposed Ozone Analysis 51–52. In making those estimations, too, EPA calibrated its modeling to each State's features, “determin[ing] the relationship between changes in emissions and changes in ozone contributions on a state-by-state . . . basis.” Id., at 33; see also id., at 40, 42.

Page Proof Pending Publication To pick which measures would “maximiz[e] cost- effectiveness” in achieving “downwind ozone air quality improvements,” 87 Fed. Reg. 20055, EPA focused on what it called the “ `knee in the curve,' ” or the point at which more expenditures in the upwind States were likely to produce “very little” in the way of “additional emissions reductions and air quality improvement” downwind, id., at 20095 (hyphenation omitted). EPA used this point to select a “uniform level” of cost, and so a uniform package of emissions- reduction tools, for upwind States to adopt. Id., at 20076. And EPA performed this analysis on two “parallel tracks”— one for power plants, one for other industries. Ibid. Pursuant to the Clean Air Act, §§ 7607(d)(1)(B), (d)(3)–(6), the agency published its proposed FIP for notice and comment in April 2022, 87 Fed. Reg. 20036.

Immediately, commenters warned of a potential pitfall in the agency's approach.

EPA had determined which emissions-control measures were cost effective at addressing downwind ozone levels based on an assumption that the FIP would apply to all covered States. But what happens if some or many of those States are not covered? As the commenters portrayed the SIPs, this was not an entirely speculative possibility. Many believed EPA's disapprovals of the SIPs were legally fawed. See, e. g., Comments of Missouri Dept. of Natural Resources 3 (June 17, 2022) (referencing “all the technical, legal, and procedural issues” with the proposed SIP disapproval); see also, e. g., Comments of Louisiana Dept. of Environmental Quality 1–3 (June 21, 2022); Comments of Texas Comm'n on Environmental Quality 2–4 (June 21, 2022); EPA, Response to Public Comments on Proposed Rule 9–11 (EPA–HQ–OAR–2021–0668). They added that EPA's FIP was “inextricably linked” to the SIP disapprovals. E. g., Comments of Missouri Dept. of Natural Resources, at 4. Without a SIP disapproval or missing SIP, after all, EPA could not include a State in its FIP. See, e. g., id., at 3; supra, at 284.

Page Proof Pending Publication Commenters added that failing to include a State could have consequences for the proposed FIP. If the FIP did not wind up applying to all 23 States as EPA envisioned, commenters argued, the agency would need “to conduct a new assessment and modeling of contribution and subject those fndings to public comment.” E. g., Comments of Air Stewardship Coalition 13–14 (June 21, 2022); Comments of Portland Cement Association 7 (June 21, 2022). Why? As noted above, EPA assessed “signifcant contribution” by determining what measures in upwind States would maximize cost-effective ozone-level improvements in the States downwind of them. Supra, at 286–287. And a different set of States might mean that the “knee in the curve” would shift. After all, each State differs in its mix of industries, in its pre-existing emissions-control measures, and in the impact those measures may have on emissions and downwind air quality. See 87 Fed. Reg. 20052, 20060, 20071–20073; EPA, Technical Memorandum, Screening Assessment of Potential Emissions Reductions, Air Quality Impacts, and Costs from Non-EGU Emissions Units for 2026, pp. 12–13 (2022).⁴ As it happened, ongoing litigation over the SIP disapprovals soon seemed to vindicate at least some of the comment⁴Commenters pointed out the variance among emissions-producing facilities too. See, e. g., Comments of Indiana Municipal Power Agency 9 (June 20, 2022) (the “cost effectiveness” of one tool “will be highly variable” across different power plants); Comments of Lower Colorado River Authority 21 (June 21, 2022) (power plants that “have already invested” in one emissions-control tool “have already undertaken signifcant costs to achieve [nitrogen-oxide] reductions and have less to gain from additional control installation”); Comments of Air Stewardship Coalition 27 (June 21, 2022) (noting that the “knee in the curve” appeared to be at a different cost depending on which mix of industries were considered); Comments of Wisconsin Paper Council 2 (June 21, 2022) (the air-quality benefts from controlling one industry—pulp and paper mills—had a “maximum estimated improvement” in ozone levels in downwind States of just 0.0117 parts per billion).

Page Proof Pending Publication Page Proof Pending Publication ers' concerns. Two circuits issued stays of EPA's SIP denials for four States. See Order in No. 23–60069 (CA5, May 1, 2023) (Texas and Louisiana); Order in No. 23–1320 (CA8, May 25, 2023) (Arkansas); Order in No. 23–1719 (CA8, May 26, 2023) (Missouri).

Despite those comments and developments, the agency proceeded to issue its fnal FIP. 88 Fed. Reg. 36654 (2023).⁵ In response to the problem commenters raised, EPA adopted a severability provision stating that, should any jurisdiction drop out, its rule would “continue to be implemented as to any remaining jurisdictions.” Id., at 36693. But in doing so, EPA did not address whether or why the same emissions- control measures it mandated would continue to further the FIP's stated purpose of maximizing cost-effective air-quality improvement if fewer States remained in the plan.

D

After EPA issued its fnal FIP, litigation over the agency's SIP disapprovals continued. One court after another issued one stay after another.⁶ Each new stay meant another State to which EPA could not apply its FIP.

See § 7410(c)(1). Ultimately, EPA recognized that it could not ⁵The fnal FIP covered 23 States. 88 Fed. Reg. 36654, 36656 (2023). That plan included Pennsylvania and Virginia, but EPA declined to cover Tennessee or Wyoming at the time, even though it had announced its intention to disapprove those States' SIPs. Ibid.; see also supra, at 285, and nn. 1–2. EPA has since proposed a plan for Tennessee and several other States. 89 Fed. Reg. 12666 (2024).

⁶See, e. g., Order in No. 23–60069 (CA5, June 8, 2023) (Mississippi); Order in No. 23–682 (CA9, July 3, 2023) (Nevada); Order in No. 23–1776 (CA8, July 5, 2023) (Minnesota); Order in No. 23–3216 (CA6, July 25, 2023) (Kentucky); Order in No. 23–9520 etc. (CA10, July 27, 2023) (Utah and Oklahoma); Order in No. 23–11173 (CA11, Aug. 17, 2023) (Alabama); see also Order in No. 23–1418 (CA4, Aug. 10, 2023) (West Virginia, pending oral argument on preliminary motions to stay and to transfer); Order in No. 23–1418 (CA4, Jan. 10, 2024) (West Virginia, after oral argument and pending merits review of petition).

apply its FIP to 12 of the 23 original States.⁷ Together, these 12 States accounted for over 70 percent of the emissions EPA had planned to address through its FIP. See Application for Ohio et al. in No. 23A349, p. 1 (States' Application); see also 88 Fed. Reg. 36738–36739.⁸ A number of the remaining States and industry groups challenged the remnants of the FIP in the D. C. Circuit.

They pointed to the Act's provisions authorizing a court to “reverse any . . . action” taken in connection with a FIP that is “arbitrary” or “capricious.” § 7607(d)(9)(A). And they argued that EPA's decision to apply the FIP to them even after so many other States had dropped out met that standard. As part of their challenge, they asked that court to stay any effort to enforce the FIP against them while their appeal unfolded. After that court denied relief, the applicants renewed their request here. The Court has received and reviewed over 400 pages of briefng and a voluminous record, held over an hour of oral argument on the applications, and engaged in months of postargument deliberations as we often do for the cases we hear.

II

A

Stay applications are nothing new. They seek a form of interim relief perhaps “as old as the judicial system of the nation.” Scripps-Howard Radio, Inc. v. FCC, 316 U. S. 4, 17 (1942). Like any other federal court faced with a stay request, we must provide the applicants with an answer— ⁷See 88 Fed. Reg. 49295 (2023) (Arkansas, Kentucky, Louisiana, Mississippi, Missouri, and Texas); 88 Fed. Reg. 67102 (2023) (Alabama, Minnesota, Nevada, Oklahoma, Utah, and West Virginia). EPA has since proposed settling the litigation over the Nevada SIP disapproval. 89 Fed. Reg. 35091 (2024).

⁸Of course, this could change again as litigation over the SIP denials progresses past preliminary stay litigation and toward fnal decisions on the merits.

Page Proof Pending Publication Page Proof Pending Publication “grant or deny.” Labrador v. Poe, 601 U. S. –––, ––– (2024) (Kavanaugh, J., concurring in grant of stay).

In deciding whether to issue a stay, we apply the same “sound . . . principles” as other federal courts. Nken v. Holder, 556 U. S. 418, 434 (2009) (internal quotation marks omitted). Specifcally, in this litigation, we ask (1) whether the applicant is likely to succeed on the merits, (2) whether it will suffer irreparable injury without a stay, (3) whether the stay will substantially injure the other parties interested in the proceedings, and (4) where the public interest lies. Ibid.; States' Application 13; Response in Opposition for Respondent EPA in No. 23A349 etc., p. 16 (EPA Response).⁹ When States and other parties seek to stay the enforcement of a federal regulation against them, often “the harms and equities [will be] very weighty on both sides.” Labra dor, 601 U. S., at ––– (opinion of Kavanaugh, J.). That is certainly the case here, for both sides have strong arguments with respect to the latter three Nken factors. On one side of the ledger, the federal government points to the air- quality benefts its FIP offers downwind States. EPA Response 48–50. On the other side, the States observe that a FIP issued unlawfully (as they contend this one was) necessarily impairs their sovereign interests in regulating their own industries and citizens—interests the Act expressly recognizes. See Part I–A, supra; States' Application 24–26; Maryland v. King, 567 U. S. 1301, 1303 (2012) (Roberts, C. J., in chambers). The States observe, too, that having to comply with the FIP during the pendency of this litigation risks placing them at a “competitive disadvantage” to their exempt peers. States' Application 21. The States and the private applicants also stress that complying with ⁹Approaching the applications before us like any other stay request both accords with the Clean Air Act's text, see 42 U. S. C. § 7607(d), and usual practice in this feld, see, e. g., Texas v. EPA, 829 F. 3d 405, 424 (CA5 2016); West Virginia v. EPA, 90 F. 4th 323, 331 (CA4 2024); In re Murray Energy Corp., 788 F. 3d 330, 335 (CADC 2015).

the FIP during the pendency of this litigation would require them to incur “hundreds of millions[,] if not billions of dollars.” Tr. of Oral Arg. 96. Those costs, the applicants note, are “nonrecoverable.” Thunder Basin Coal Co. v. Reich, 510 U. S. 200, 220–221 (1994) (Scalia, J., concurring in part and concurring in judgment); see, e. g., States' Application 24; Application for American Forest & Paper Association et al. 25; see also Alabama Assn. of Realtors v. Department of Health and Human Servs., 594 U. S. 758, 765 (2021) (per curiam).

Because each side has strong arguments about the harms they face and equities involved, our resolution of these stay requests ultimately turns on the merits and the question who is likely to prevail at the end of this litigation. See Nken, 556 U. S., at 434; Labrador, 601 U. S., at ––– (opinion of Kavanaugh, J.).

B

When it comes to that question, the parties agree on the rules that guide our analysis. The applicants argue that a court is likely to hold EPA's fnal FIP “arbitrary” or “capricious” within the meaning of the Act and thus enjoin its enforcement against them. 42 U. S. C. § 7607(d)(9)(A); see, e. g., States' Application 15–16; Application for American Forest & Paper Association et al. 14; see also 5 U. S. C. § 706(2)(A). An agency action qualifes as “arbitrary” or “capricious” if it is not “reasonable and reasonably explained.” FCC v. Prometheus Radio Project, 592 U. S. 414, 423 (2021). In reviewing an agency's action under that standard, a court may not “ `substitute its judgment for that of the agency.' ” FCC v. Fox Television Stations, Inc., 556 U. S. 502, 513 (2009). But it must ensure, among other things, that the agency has offered “a satisfactory explanation for its action[,] including a rational connection between the facts found and the choice made.” Motor Vehicle Mfrs. Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 43 (1983) (internal quotation marks Page Proof Pending Publication Page Proof Pending Publication omitted). Accordingly, an agency cannot simply ignore “an important aspect of the problem.” Ibid. We agree with the applicants that EPA's fnal FIP likely runs afoul of these long-settled standards. The problem stems from the way EPA chose to determine which emissions “contribute[d] signifcantly” to downwind States' diffculty meeting national ozone standards. 42 U. S. C. § 7410(a)(2)(D)(i)(I). Recall that EPA's plan rested on an assumption that all 23 upwind States would adopt emissions- reduction tools up to a “uniform” level of “costs” to the point of diminishing returns. 87 Fed. Reg. 20076, 20095; 88 Fed. Reg. 36661, 36683–36684, 36719; see Part I–C, supra. But as the applicants ask: What happens—as in fact did happen— when many of the upwind States fall out of the planned FIP and it may now cover only a fraction of the States and emissions EPA anticipated? See, e. g., States' Application 16–21; Application for American Forest & Paper Association et al. 14–15, 19–20. Does that affect the “knee in the curve,” or the point at which the remaining States might still “maximiz[e] cost-effectiv[e]” downwind ozone-level improvements? 87 Fed. Reg. 20055. As “the mix of states changes, . . . and their particular technologies and industries drop out with them,” might the point at which emissions-control measures maximize cost-effective downwind air-quality improvements also shift? Tr. of Oral Arg. 6.

Although commenters posed this concern to EPA during the notice and comment period, see Part I–C, supra, EPA offered no reasoned response. Indeed, at argument the government acknowledged that it could not represent with certainty whether the cost-effectiveness analysis it performed collectively for 23 States would yield the same results and command the same emissions-control measures if conducted for, say, just one State. Tr. of Oral Arg. 58–59. Perhaps there is some explanation why the number and identity of participating States does not affect what measures maximize cost-effective downwind air-quality improvements. But if there is an explanation, it does not appear in the fnal rule. As a result, the applicants are likely to prevail on their argument that EPA's fnal rule was not “reasonably explained,” Prometheus Radio Project, 592 U. S., at 423, that the agency failed to supply “a satisfactory explanation for its action[,]” State Farm Mut. Automobile Ins. Co., 463 U. S., at 43, and that it instead ignored “an important aspect of the problem” before it, ibid. The applicants are therefore likely to be entitled to “revers[al]” of the FIP's mandates on them.

§ 7607(d)(9).¹⁰

III

A

Resisting this conclusion, EPA advances three alternative arguments.

First, the government insists, the agency did offer a reasoned response to the applicants' concern, just not the one they hoped. When fnalizing its rule in response to public comments, the government represents, “the agency did consider whether the [FIP] could cogently be applied to a subset of the 23 covered States.” EPA Response 27; see also post, at 316–317 (Barrett, J., dissenting). And that consideration, the government stresses, led EPA to add a “severability” provision to its fnal rule in which the agency announced that the FIP would “ `continue to be implemented' ” without regard to the number of States remaining, even if just one State remained subject to its terms. EPA Response 27 (quoting 88 Fed. Reg. 36693). In support of its severability provision, EPA cited, among other things, its intent to address “ `important public health and environmental benefts' ” and encourage reliance by others “ `on th[e] fnal rule in their planning.' ” Ibid. ¹⁰Various applicants offer various other reasons why they believe they are likely to succeed in challenging EPA's FIP. Having found that they are likely to succeed on the basis discussed above, however, we have no occasion to address those other arguments.

Page Proof Pending Publication Page Proof Pending Publication None of this, however, solves the agency's problem. True, the severability provision highlights that EPA was aware of the applicants' concern. But awareness is not itself an explanation. The severability provision highlights, too, the agency's desire to apply its rule expeditiously and “ `to the greatest extent possible,' ” no matter how many States it could cover. Ibid. But none of that, nor anything else EPA said in support of its severability provision, addresses whether and how measures found to maximize cost effectiveness in achieving downwind ozone air-quality improvements with the participation of 23 States remain so when many fewer States, responsible for a much smaller amount of the originally targeted emissions, might be subject to the agency's plan. Put simply, EPA's response did not address the applicants' concern so much as sidestep it.^{11 11}As the applicants conceded at oral argument, see Tr. of Oral Arg. 25– 26, EPA did not need to address every possible permutation when it sought to adopt a multi-State FIP. Our conclusion is narrower: When faced with comments like the ones it received, EPA needed to explain why it believed its rule would continue to offer cost-effective improvements in downwind air quality with only a subset of the States it originally intended to cover. To be sure, after this Court heard argument, EPA issued a document in which it sought to provide further explanations for the course it pursued. See 89 Fed. Reg. 23526 (2024). But the government has not suggested that we should consult this analysis in assessing the validity of the fnal rule. See Letter from E. Prelogar, Solicitor General, to S. Harris, Clerk of Court 1 (Mar. 28, 2024). Nor could it, since the Clean Air Act prevents us (and courts that may in the future assess the FIP's merits) from consulting explanations and information offered after the rule's promulgation. See 42 U. S. C. §§ 7607(d)(6)(C) (“The promulgated rule may not be based (in part or whole) on any information or data which has not been placed in the docket as of the date of such promulgation”), 7607(d)(7)(A) (restricting the “record for judicial review”). We therefore look to only “the grounds that the agency invoked when it” promulgated the FIP.

Michigan v. EPA, 576 U. S. 743, 758 (2015).

Should the applicants show the FIP was arbitrary or capricious on the existing record, as we have concluded is likely, the Clean Air Act entitles them to “revers[al]” of that rule's mandates on them. §7607(d)(9)(A). Second, the government pivots in nearly the opposite direction. Now, it says, if its fnal rule lacks a reasoned response to the applicants' concern, it is because no one raised that concern during the public comment period. And, the agency stresses, a litigant may pursue in court only claims premised on objections frst “ `raised with reasonable specifcity' ” before the agency during the public comment period. Id., at 19–20 (quoting § 7607(d)(7)(B)); see also post, at 307–310. We cannot agree. The Act's “reasonable specifcity” requirement does not call for “a hair-splitting approach.” Ap palachian Power Co. v. EPA, 135 F. 3d 791, 817 (CADC 1998). A party need not “rehears[e]” the identical argument made before the agency; it need only confrm that the government had “notice of [the] challenge” during the public comment period and a chance to consider “in substance, if not in form, the same objection now raised” in court. Id., at 818; see also, e. g., Bahr v. Regan, 6 F. 4th 1059, 1070 (CA9 2021).

Here, EPA had notice of the objection the applicants seek to press in court. Commenters alerted the agency that, should some States no longer participate in the plan, the agency would need to return to the drawing board and “conduct a new assessment and modeling of contribution” to determine what emissions-control measures maximized cost effectiveness in securing downwind ozone air-quality improvements. Comments of Air Stewardship Coalition, at 13–14; see also Part I–C, supra (noting examples of other comments). And, as we have just seen, EPA's own statements and actions confrm the agency appreciated that concern. In preparing the fnal rule in response to public comments, the agency emphatically insists, it “did consider whether the [r]ule could cogently be applied to a subset of the 23 covered States.” EPA Response 27. And as a result of that consideration, the agency observes, it opted to add a severability provision to its fnal rule. Ibid. By its own words and actions, then, the agency demonstrated that it was Page Proof Pending Publication on notice of the applicants' concern. Yet, as we have seen, it failed to address the concern adequately.¹² Third, the government pursues one more argument in the alternative. As the agency sees it, the applicants must return to EPA and fle a motion asking it to reconsider its fnal rule before presenting their objection in court. They must, the agency says, because the “grounds for [their] objection arose after the period for public comment.” § 7607(d)(7)(B); see EPA Response 20–21. As just discussed, however, EPA had the basis of the applicants' objection before it during the comment period. It chose to respond with a severability provision that in no way grappled with their concern. Nothing requires the applicants to return to EPA to raise (again) a concern EPA already had a chance to address.

Taking the government's argument (much) further, the dissent posits that every “objection that [a] fnal rule was not reasonably explained” must be raised in a motion for reconsideration. Post, at 306 (internal quotation marks omitted; emphasis deleted). But there is a reason why the government does not go so far. The Clean Air Act opens the court¹²The dissent resorts to a “hair-splitting approach” to the public comments. Post, at 307–310. It stresses, for example, that some comments highlighted variances among specifc emissions-producing facilities and industries, “not States.” Post, at 307 (emphasis deleted). But the dissent fails to acknowledge that, for purposes of the FIP, States are a sum of their emissions-producing facilities. See, e. g., Ozone Transport Policy Analysis Final Rule TSD 12 (EPA–HQ–OAR–2021–0668, 2023) (Final Ozone Analysis). Similarly, the dissent characterizes the comment indicating EPA would need to “conduct a new assessment and modeling” if States dropped out of the FIP as a complaint about the “sequencing” of the proposed SIP disapprovals and the FIP. Post, at 309. But why would the sequencing matter? Because the FIP cannot apply to a State if its SIP is not disapproved. See Part I–A, supra. And why would EPA need to perform a “new assessment and modeling of contribution”? Because it may be that “the math . . . wouldn't necessarily turn out the same” if some States were not covered by the FIP. Tr. of Oral Arg. 59. Fairly on notice of the concern, EPA needed to, and by its own admission sought to, “consider” whether its FIP could apply to a subset of States. EPA Response 27. Page Proof Pending Publication house doors to those with objections the agency already ignored. If an “objection [is] raised with reasonable specifcity during the period for public comment” but not reasonably addressed in the fnal rule, the Act permits an immediate challenge. § 7607(d)(7)(B). A person need not go back to the agency and insist on an explanation a second time. Tellingly, the case on which the dissent relies involves an entirely different situation: a “ `logical outgrowth' challeng[e].” Post, at 306. There, the objection was that EPA had supposedly “ `signifcantly amend[ed] the [r]ule between the proposed and fnal versions,' ” making it impossible for people to comment on the rule during the comment period. Ibid. (quoting EME Homer, 795 F. 3d, at 137). That is nothing like the challenge here, where EPA failed to address an important problem the public could and did raise during the comment period.

B

With the government's theories unavailing, the dissent advances others of its own. It begins by suggesting that the problem the applicants raise was not “ `important' ” enough to warrant a reasoned reply from the agency because the methodology EPA employed in its FIP “appear[s] not to depend on the number of covered States.” Post, at 311–316, 317–318. Then, coming at the same point from another direction, the dissent seeks to excuse the agency's lack of a reasoned reply as “harmless” given, again, “the apparent lack of connection between the number of States covered and the FIP's methodology.” Post, at 319.

The trouble is, if the government had arguments along these lines, it did not make them. It did not despite its ample resources and voluminous briefng. See supra, at 290. This Court “normally decline[s] to entertain” arguments “forfeited” by the parties. Kingdomware Technologies, Inc. v. United States, 579 U. S. 162, 173 (2016). And we see no persuasive reason to depart from that rule here.

Page Proof Pending Publication If anything, we see one reason for caution after another. Start with the fact the dissent itself expresses little confdence in its own theories, contending no more than it “ap pear[s]” EPA's methodology did not depend on the number of covered States. Post, at 313 (emphasis added). Add to that the fact that, at oral argument, even the government refused to say with certainty that EPA would have reached the same conclusions regardless of which States were included in the FIP. See Tr. of Oral Arg. 59. Combine all that with the further fact that, in developing the FIP, EPA said it used the “same regulatory framework” this Court described in EME Homer City Generation, L. P. v. EPA, 572 U. S. 489. E. g., EPA Response 7–8. And, at least as the Court described that framework, state-level analyses play a signifcant role in EPA's work.¹³ Finally, observe that, while the Act seems to anticipate, as the dissent suggests, that the agency's “procedural determinations” may be subject to harmless-error review, § 7607(d)(8), the Act also seems to treat separately challenges to agency “actions” like the FIP before us, authorizing courts to “reverse any . . . action” found to be “arbitrary” or “capricious,” § 7607(d)(9)(A). With so many reasons for caution, we think sticking to our normal course of declining to consider forfeited arguments the right course here.^{14 13}The agency, we said, “frst calculated, for each upwind State, the quantity of emissions the State could eliminate at each of several cost [levels]”; next, it “conducted complex modeling to establish the combined effect the upwind reductions projected at each cost [level] would have on air quality in downwind States”; and only after all that did the agency “then identif[y] `signifcant cost [levels]' ” to use in setting its emissions budgets. EME Homer City, 572 U. S., at 501–502.

¹⁴Admittedly, the dissent points to some statements in the FIP suggesting EPA considered nationwide data in parts of its analysis. See, e. g., post, at 313; see also, e. g., 88 Fed. Reg. 36721, 36727. But other statements in that rule and supporting documents also seem to suggest EPA considered state-specifc information. See Part I–C, supra. If, as the Page Proof Pending Publication * The applications for a stay in Nos. 23A349, 23A350, 23A351, and 23A384 are granted. Enforcement of EPA's rule against the applicants shall be stayed pending the disposition of the applicants' petitions for review in the United States Court of Appeals for the D. C. Circuit and any petition for writ of certiorari, if such writ is timely sought. Should the petition for certiorari be denied, this order will terminate automatically. If the petition is granted, this order shall terminate upon the sending down of the judgment of this Court.

It is so ordered.

The Court today enjoins the enforcement of a major Environmental Protection Agency rule based on an underdeveloped theory that is unlikely to succeed on the merits. In so doing, the Court grants emergency relief in a fact-intensive and highly technical case without fully engaging with both the relevant law and the voluminous record. While the Court suggests that the EPA failed to explain itself suffciently in response to comments, this theory must surmount sizable procedural obstacles and contrary record evidence. dissent posits, only nationwide data informed EPA's analysis, why would EPA say that, “for purposes of identifying the appropriate level of control,” it focused on “the 23 upwind states that were linked” to the downwind States, rather than, say, “all states in the contiguous U. S.”? Final Ozone Analysis 3 (footnote omitted). Why would EPA explain that its “fndings regarding air quality improvement” downwind were a “central component” of picking the appropriate cost levels and so defning a State's signifcant contribution? 88 Fed. Reg. 36741. And why would EPA bother to “determine the relationship between changes in emissions and changes in ozone contributions on a state-by-state . . . basis” and “calibrat[e]” that relationship “based on state-specifc source apportionment”? Final Ozone Analysis 43. In asking these questions, we do not profess answers; we simply highlight further reasons for caution. Page Proof Pending Publication Applicants therefore cannot satisfy the stringent conditions for relief in this posture.

I

I will start by setting the record straight with respect to some important background.

First, the Court downplays EPA's statutory role in ensuring that States meet air-quality standards. Ante, at 283– 284. The Clean Air Act directs EPA to “establish national ambient air quality standards (NAAQS) for pollutants at levels that will protect public health.” EPA v. EME Homer City Generation, L. P., 572 U. S. 489, 498 (2014); see 42 U. S. C. §§ 7408, 7409. States must create State Implementation Plans (SIPs) to ensure that their air meets these standards. § 7410(a)(1). But States also face an externality problem: “Pollutants generated by upwind sources are often transported by air currents . . . to downwind States,” relieving upwind States “of the associated costs” and making it diffcult for downwind States to “maintain satisfactory air quality.” EME, 572 U. S., at 496. So the Act's Good Neighbor Provision requires SIPs to “prohibi[t]” the State's emissions sources from “emitting any air pollutant in amounts which will . . . contribute signifcantly to nonattainment in, or interfere with maintenance by, any other State with respect to any [NAAQS].” § 7410(a)(2)(D)(i)(I).

Given the incentives of upwind States to underregulate the pollution they send downwind, the Act requires EPA to determine whether a State “has failed to submit an adequate SIP.” EME, 572 U. S., at 498; see § 7410(c)(1). If a SIP does not prevent the State's polluters from signifcantly contributing to nonattainment in downwind States, EPA “shall” promulgate a Federal Implementation Plan (FIP) that does. § 7410(c)(1). And EPA must stop the State's signifcant contributions by the statutory deadline for the affected downwind States to achieve compliance. See Wis consin v. EPA, 938 F. 3d 303, 313–314 (CADC 2019) (per curiam); § 7511.

Page Proof Pending Publication Second, the Court fails to recognize that EPA's SIP disapprovals may, in fact, be valid. EPA justifed its fndings that 23 States had failed to submit adequate SIPs. It found that these States all signifcantly contributed to ozone pollution in downwind States. See 88 Fed. Reg. 36656 (2023). But 21 of these States, including applicants, proposed to do noth ing to reduce their ozone-precursor (i. e., NOx) emissions— arguing that they did not actually contribute to downwind nonattainment or that there were no other cost-effective emissions-reduction measures they could impose. See 88 Fed. Reg. 9354–9361 (2023). The other two States failed to submit a SIP at all. See 84 Fed. Reg. 66614 (2019). While 12 of EPA's SIP disapprovals have been temporarily stayed, no court yet has invalidated one. So EPA's replacement FIP—the Good Neighbor Plan—may yet apply to all 23 original States. Indeed, EPA and the plaintiffs who challenged Nevada's SIP disapproval have proposed a settlement that would lift that stay. 89 Fed. Reg. 35091 (2024).

Third, the Court claims that commenters on the proposed FIP warned that its emissions limits might change if it covered fewer States, but EPA failed to respond. Ante, at 287– 289. Not exactly. As I will elaborate below, commenters merely criticized EPA's decision to propose a FIP before its SIP disapprovals were fnal. EPA responded that this sequencing was “consistent with [its] past practice in [its] efforts to timely address good neighbor obligations”: Given the August 2024 deadline for certain States to comply with the 2015 ozone NAAQS, EPA was “obligated” to start the years- long process of promulgating a FIP so that one could be effective in time. EPA, Response to Public Comments on Proposed Rule 149–150 (EPA–HQ–OAR–2021–0668–1127, June 2023) (Response to Comments); see Wisconsin, 938 F. 3d, at 313–314.

Finally, the Court repeatedly characterizes the FIP as relying on an “assumption that [it] would apply to all covered States.” Ante, at 287; see ante, at 293. But try as it might, Page Proof Pending Publication the Court identifes no evidence that the FIP's emissions limits would have been different for a different set of States or that EPA's consideration of state-specifc inputs was anything but confrmatory of the limits it calculated based on nationwide data. See ante, at 286–287, 299–300, n. 14. The Court leans on the fact that EPA “considered data specifc to the emissions-producing facilities in [each] State” to calculate “how much each upwind State's [NOx] emissions would fall” if the State's emitters “adopted each [emissions-control] measure.” Ante, at 286 (citing EPA, Ozone Transport Policy Analysis Proposed Rule TSD 9–10, 13, 22–23 (EPA–HQ– OAR–2021–0668–0133, Feb. 2022) (Proposed Ozone Analysis)). But the Proposed Ozone Analysis makes clear that EPA did these state-specifc calculations to determine each State's “emissions budget.” Proposed Ozone Analysis 7–13. A State's budget consists of the “emissions that would remain” after the State's power plants meet the emissions limits that EPA independently calculated. 88 Fed. Reg. 36762; see Proposed Ozone Analysis 13 (“adjust[ed]” “unit-level emissions are summed up to the state level”); n. 6, infra. Of course each State's emissions budget will depend on the emitters in that State. What matters is whether the limits the FIP imposes on each emitter depend on the number of States the FIP covers. Tellingly, the Court does not identify any NOx limit for any industry that relied on state- specifc data.

On the contrary, as I will explain in Part II–B, the fnal rule and its supporting documents suggest that EPA's methodology for setting emissions limits did not depend on the number of States in the plan, but on nationwide data for the relevant industries—and the FIP contains many examples of emissions limits that EPA created using nationwide inputs. Moreover, EPA has now confrmed this interpretation.

During this litigation, EPA received petitions seeking reconsideration of the FIP on the ground that it should not be implemented in just a subset of the original States. EPA Page Proof Pending Publication denied these petitions on April 4, 2024. 89 Fed. Reg. 23526. It thoroughly explained how its “methodology for defning” each State's emissions obligations is “independent of the number of states included in the Plan” because it “relies on a determination regarding what emissions reductions each type of regulated source can cost-effectively achieve.” EPA, Basis for Partial Denial of Petitions for Reconsideration on Scope 1 (EPA–HQ–OAR–2021–0668–1255, Apr. 2024) (Denial). The “control technologies and cost-effectiveness fgures the EPA consider[ed] . . . do not depend in any way on the number of states included.” Id., at 2. So “[s]ources in the remaining upwind states currently regulated by the Plan . . . would bear the same actual emission reduction obligations” regardless of the number of covered States. Id., at 3–4.

II

To obtain emergency relief, applicants must, at a minimum, show that they are likely to succeed on the merits, that they will be irreparably injured absent a stay, and that the balance of the equities favors them. Nken v. Holder, 556 U. S. 418, 425–426 (2009). Moreover, we should grant relief only if we would be likely to grant certiorari were the applicants' case to come to us in the usual course. See Does 1–3 v. Mills, 595 U. S. –––, ––– (2021) (Barrett, J., concurring in denial of application for injunctive relief); Hollingsworth v. Perry, 558 U. S. 183, 190 (2010) (per curiam). In my view, the applicants cannot satisfy the stay factors. Most signifcantly, they have not shown a likelihood of success on the merits.

The Court holds that applicants are likely to succeed on a claim that the Good Neighbor Plan is “arbitrary” or “capricious.”

42 U. S. C. § 7607(d)(9).

The “arbitrary-andcapricious standard requires that agency action” be both “[1] reasonable and [2] reasonably explained.” FCC v. Prometh eus Radio Project, 592 U. S. 414, 423 (2021). The Court's theory is that EPA did not “ `reasonably explai[n]' ” “why Page Proof Pending Publication the number and identity of participating States does not affect what measures maximize cost-effective downwind air- quality improvements.” Ante, at 293–294 (quoting only the second part of Prometheus Radio's formulation (emphasis added)). So to be clear, the Court does not conclude that EPA's actions were substantively unreasonable—e. g., that the FIP cannot rationally be applied to fewer States because a change in the number of participants would undermine its rationale or render it ineffective. Nor could it, given the signifcant evidence in the record (not to mention EPA's denial of reconsideration) that the covered States did not, in fact, affect the plan's emissions-reduction obligations. See Part II–B, infra. Thus, the only basis for the Court's decision is the argument that EPA failed to provide “ `a satisfactory explanation for its action' ” and a “reasoned response” to comments. Ante, at 293–294 (quoting Motor Vehicle Mfrs.

Assn. of United States, Inc. v. State Farm Mut. Automobile Ins. Co., 463 U. S. 29, 43 (1983)). There are at least three major barriers to success on such a claim.

A

The Clean Air Act imposes a procedural bar on the challenges that a plaintiff can bring in court: Only objections that were “raised with reasonable specifcity during the period for public comment . . . may be raised during judicial review.” § 7607(d)(7)(B). If it was “impracticable to raise such objection within such time or if the grounds for such objection arose after the period for public comment,” the challenger may petition for reconsideration of the rule and can obtain judicial review only if EPA refuses. Ibid. While EPA has now separately denied petitions for reconsideration of the Good Neighbor Plan, this case came to us directly; we are assessing applicants' likelihood of success in challenging the plan itself, not the denial of reconsideration. So the procedural bar on objections not raised in the comments presents a signifcant obstacle—in two ways.

Page Proof Pending Publication Page Proof Pending Publication First, consider the Court's basic theory: that EPA offered “no reasoned response” to comments allegedly questioning whether the plan's emissions limits depend on the States covered. Ante, at 293. That EPA failed to adequately explain its fnal rule in response to comments is “an objection to the notice and comment process itself,” which applicants “obviously did not and could not have raised . . . during the period for public comment.” EME Homer City Generation, L. P. v. EPA, 795 F. 3d 118, 137 (CADC 2015) (Kavanaugh, J.). No one could have raised during the proposal's comment period the objection that the “fnal rule was not `reasonably explained.' ” Ante, at 294 (emphasis added).

The D. C. Circuit, on remand in EME Homer, considered a similar objection that EPA had “violated the Clean Air Act's notice and comment requirements”: EPA had “signifcantly amend[ed] the Rule between the proposed and fnal versions without providing additional opportunity for notice and comment.” 795 F. 3d, at 137. But because this procedural objection could not have been raised during the comment period, “the only appropriate path for petitioners” under §7607(d)(7)(B) was to raise it “through an initial petition for reconsideration to EPA.” Ibid. So the D. C. Circuit lacked “authority at th[at] time to reach this question.” Ibid. While such “logical outgrowth” challenges typically are cognizable under the Administrative Procedure Act, see Shell Oil Co. v. EPA, 950 F. 2d 741, 747 (CADC 1991), the Clean Air Act channels these challenges through reconsideration proceedings. This Court's failure-to-explain objection may face the same problem: It is not judicially reviewable in its current posture.^{1 1}The Court offers a feeble response to this application of § 7607(d)(7)(B)'s procedural bar. Ante, at 297–298. It simply quotes § 7607(d)(7)(B) and asserts without support that it means that a plaintiff “need not go back to the agency and insist on an explanation a second time.” Ante, at 298. The Court fails to engage with the logic of this argument: The objection that the fnal rule did not contain suffcient explanation was not and could not have Second, even putting aside this aspect of § 7607(d)(7)(B), it is not clear that any commenter raised with “reasonable specifcity” the underlying substantive issue: that the exclusion of some States from the FIP would undermine EPA's cost-effectiveness analyses and resulting emissions controls. § 7607(d)(7)(B); see ante, at 293–294. The Court concludes otherwise only by putting in the commenters' mouths words they did not say. It frst cites a bevy of comments arguing that EPA's “disapprovals of the SIPs were legally fawed” and noting the obvious point that EPA cannot “include a State in its FIP” unless it validly disapproves the State's SIP. Ante, at 287. These comments do not address the continued effcacy of a FIP that applies to a subset of the originally covered States.

Another collection of the Court's inapposite comments relates to the inclusion of specifc sources, emissions controls, and industries in the proposed plan—not States. See ante, at 288, n. 4. For example, one commenter argued that the “cost effectiveness of the requirement to employ SNCR will be highly variable, and is unlikely to meet EPA expectations in even the most optimistic case.” Comments of Indiana Municipal Power Agency 9 (June 20, 2022). That is a challenge to EPA's endorsement of a particular emissions-control technology; it says nothing about the FIP's dependence on a particular number of States. See also, e. g., Comments of Lower Colorado River Authority 21–22 (June 21, 2022).

been raised during the comment period, so it must be raised in a petition for reconsideration. EME Homer, 795 F. 3d, at 137. The Court claims that its theory is different from the logical-outgrowth challenge the D. C. Circuit considered in EME Homer. Ante, at 298. But the Court ignores the fact that its failure-to-explain challenge and logical-outgrowth challenges are both “objection[s] to the notice and comment process itself” that depend on the content of the fnal rule. EME Homer, 795 F. 3d, at 137. Even if the public raised an “important problem . . . during the comment period,” ante, at 298, the Court's basis for enjoining the FIP's enforcement is not that the alleged problem is real, but that the fnal rule did not address it.

Page Proof Pending Publication Similarly, another commenter argued that pulp and paper mills should not be included because the “maximum estimated improvement” in ozone levels from controlling their emissions would be “too small to even measure.” Comments of Wisconsin Paper Council 2 (June 21, 2022).² An argument that the maximum benefits from regulating an industry are too small is not an argument that those benefts would become too small if fewer States were covered.³ The closest comment that the Court can fnd—which it quotes repeatedly—is one sentence that obliquely refers to some “new assessment and modeling of contribution” that EPA might need to perform. Comments of Air Stewardship Coalition 13–14 (June 21, 2022). The Court dresses up this comment by characterizing it as a warning about what might happen “[i]f the FIP did not wind up applying to all 23 States” and responding to the concern that a “different set of States might mean that the `knee in the curve' would shift” and change the cost-effective “emissions-control meas²The Court claims that in distinguishing comments about particular industries from comments that question whether the plan depends on a number of States, I “fai[l] to acknowledge” that the FIP treats States as the “sum of their emissions-producing facilities.” Ante, at 297, n. 12. But in reality, it is the Court that ignores how the FIP works. The FIP determines emissions limits for particular sources based on their industries; the total NOx emissions limit for each State is simply the sum of the limits the plan imposes on each of the State's sources. See, e. g., 88 Fed. Reg. 36678, 36762; Part II–B, infra. So comments critiquing a particular industry-specifc emissions limit or technology assumption say nothing about the FIP's dependence on a certain number of States. ³Nor did the Air Stewardship Coalition's comment about the “knee in the curve” raise concerns about which States are included. See ante, at 288, n. 4. Rather, this comment questioned EPA's proposed average cost- effectiveness threshold of $7,500 per ton for non-power-plant sources; it argued that EPA should use different thresholds for different industries. Comments of Air Stewardship Coalition 27 (June 21, 2022) (Air Stewardship Comments). It did not link its concern about cost thresholds to the States covered by the plan.

Page Proof Pending Publication ures.” Ante, at 288. But those words are the Court's, not the commenter's.

The commenter's actual objection was to EPA's sequencing of its actions—proposing a FIP before it fnalized its SIP disapprovals. The commenter titled this section “EPA Step Two Screening is Premised on the Premature Disapproval of 19 Upwind States['] Good Neighbor SIPs.” Air Stewardship Comments 13 (boldface omitted). And the relevant sentence reads in full: “The proposed FIP essentially prejudges the outcome of those pending SIP actions and, in the event EPA takes a different action on those SIPs than contemplated in this proposal, it would be required to conduct a new assessment and modeling of contribution and subject those fndings to public comment.” Id., at 14.

This sentence says nothing about what would be required if after EPA fnalizes its SIP disapprovals and issues a fnal FIP, some States drop out of the plan. Nor does it suggest that the plan's cost-effectiveness thresholds or emissions controls would change with a different number of States. Nor is it clear what the comment means by its bare reference to a “new assessment and modeling of contribution”: Would EPA be required to perform a new evaluation of which upwind States cause pollution in downwind States? A new analysis of how much pollution each source must eliminate? A new assessment of the plan's impact on downwind States? It is therefore diffcult to see how this comment raised with “reasonable specifcity” the objection that the removal of some States from the fnal plan would invalidate EPA's cost-effectiveness thresholds and chosen emissions-control measures.⁴ That is not how EPA understood it. EPA char⁴So too with Portland Cement's comment. See ante, at 288. That comment simply echoes the quoted sentence from the Air Stewardship Coalition almost word for word, also in the context of objecting to EPA's decision to propose a FIP before fnalizing its SIP disapprovals. Comments of Portland Cement Association 7 (June 21, 2022).

Page Proof Pending Publication acterized this comment as arguing that “by taking action before considering comments on the proposed disapprovals, the EPA is presupposing the outcome of its proposed rule- makings on the SIPs.” Response to Comments 147 (noting this comment's ID number, 0518). And EPA explained that it “disagree[d]” with the argument that the “sequence” of its actions was “improper, unreasonable, or bad policy”; EPA had a statutory obligation to promulgate a FIP by the August 2024 NAAQS attainment deadline. Id.,at150. Ifa commenter had said with reasonable specifcity what the Court says today—that “a different set of States might mean that the `knee in the curve' would shift,” ante, at 288—EPA could have responded with more explanation of why its methodology did not depend on the number of covered States—as it has recently explained. But EPA cannot be penalized if it did not have reasonable notice of this objection.<sup>5</sup> In sum, § 7607(d)(7)(B)'s procedural bar likely forecloses both the failure-to-explain objection that the Court credits and any substantive challenge to the reasonableness of applying the FIP to a subset of the originally covered States. <sup>5</sup>The Court concludes to the contrary only by building out the comment's bare reference to a “new assessment and modeling” with its own inferences about the possible effect of different numbers of States on “the math.” Ante, at 297, n. 12 (internal quotation marks omitted). But as explained above, the comment itself said nothing about States dropping out of the fnal plan or the possible impact of different numbers of States on the FIP's cost thresholds or emissions limits. It is hard to believe that a single sentence with no elaboration or explanation of the potential issue—in a sea of thousands of pages of comments—gave EPA reasonable notice that it should have included in its fnal rule a detailed explanation of why the FIP's emissions limits did not depend on the number of States. Cf. Public Citizen, Inc. v. FAA, 988 F. 2d 186, 197 (CADC 1993) (An “agency need not respond at all to comments that are `purely speculative and do not disclose the factual or policy basis on which they rest' ”).

Page Proof Pending Publication

B

Even if applicants clear § 7607(d)(7)(B)'s procedural bar, they face an uphill battle on the merits. To prevail on the Court's theory, applicants must show that EPA's actions were “arbitrary” or “capricious.” §§ 7607(d)(9)(A), (D).

“The scope of review under the `arbitrary and capricious' standard is narrow and a court is not to substitute its judgment for that of the agency.” State Farm, 463 U. S., at 43. A rule is arbitrary and capricious if the agency “entirely failed to consider an important aspect of the problem.”

Ibid. (emphasis added). But we will “ `uphold a decision of less than ideal clarity if the agency's path may reasonably be discerned.' ” Ibid. (quoting Bowman Transp., Inc. v. Arkansas-Best Freight System, Inc., 419 U. S. 281, 286 (1974)). Given the explanations and state-agnostic methodology apparent in the fnal rule and its supporting documentation—and the paucity of comments specifcally raising the issue—EPA may well have done enough to justify its plan's severability.

To begin, the rule and its supporting documents arguably make clear that EPA's methodology for calculating cost- effectiveness thresholds and imposing emissions controls did not depend on the number of covered States. The rule applied EPA's longstanding “4-step interstate transport framework” to create emissions limits that will prevent NOx sources in upwind States from signifcantly contributing to ozone pollution in downwind States. 88 Fed. Reg. 36659; see 42 U. S. C. § 7410(a)(2)(D). Under that framework, EPA (1) identifes “downwind receptors that are expected to have problems attaining or maintaining the NAAQS”; (2) identifes which upwind States are “ `link[ed]' ” to those downwind receptors because they contribute at least 1% of a receptor's ozone; (3) determines which NOx sources in the linked upwind States “signifcantly contribute” to downwind nonattainment or interference; and (4) implements emissions limPage Proof Pending Publication its to stop those sources' signifcant contributions. 88 Fed. Reg. 36659; see EME, 572 U. S., at 500–501 (describing similar approach used in earlier FIP). The frst two steps determine which States the FIP must cover. The rubber meets the road at steps 3 and 4: How much do sources in those States “signifcantly contribute” to downwind pollution, and what must they do about it?

Here is how EPA explains that methodology. A source “signifcantly contributes” to downwind pollution if there are cost-effective measures it could implement to reduce its emissions: It must halt those emissions that can be eliminated at a cost “under the cost threshold set by the Agency” for sources in that industry. EME, 572 U. S., at 518 (upholding this approach). So the “ `amount' of pollution” that sources must eliminate is “that amount . . . in excess of the emissions control strategies the EPA has deemed cost effective.” 88 Fed. Reg. 36676. EPA calculates for each type of source a “uniform level of NOx emissions control stringency” expressed as a “cost per ton of emissions reduction.” Id., at 36719. This cost-effectiveness threshold is based on the point “at which further emissions mitigation strategies become excessively costly on a per-ton basis while also delivering far fewer additional emissions reductions.” Id., at 36683 (describing this “ `knee in the curve' ” analysis). The plan requires sources in each covered State to reduce their emissions accordingly.^{6 6}For power plants, EPA implements these requirements by allocating each State an “ `emissions budget' . . . representing the EPA's quantifcation of the emissions that would remain” if plants in that State eliminated all the emissions that EPA determines can be eliminated for less than the cost threshold. 88 Fed. Reg. 36762. EPA then allocates tradeable “ `allowances' ” proportionally among the State's sources, creating a marketplace for emissions. Ibid. With respect to other sources, EPA determined that nine industries in the covered States produced the most signifcant emissions. Id., at 36817. The rule requires sources in each of those industries to meet specifc emissions limits that were calculated based on the reductions that they can cost-effectively achieve. Ibid. Page Proof Pending Publication Page Proof Pending Publication Crucially, the fnal rule suggests that EPA calculated cost- effectiveness thresholds based on the likely cost and impact of available emissions-reduction technology given national, industry-wide data. Contrary to the Court's speculations, ante, at 293, these thresholds and the FIP's resulting emissions limits appear not to depend on the number of covered States. Consider the plan's approach to power plants (“electric generating units,” or EGUs). EPA assessed the cost and impact of different NOx mitigation strategies that EGUs could implement. One strategy was to fully operate “selective catalytic reduction” (SCR) technology. 88 Fed. Reg.

36655; see id., at 36720. EPA estimated that a “representative marginal cost” for this strategy would be $1,600 per ton, and a “reasonable level of performance” would be 0.08 lb/ mmBtu—based on “nationwide” power plant “emissions data.” Id., at 36720–36721. EPA thus determined that SCR optimization was a “viable mitigation strategy for the 2023 ozone season” and built this assumption into the plan's emissions limits. Id., at 36720. In other words, EPA relied on nationwide industry data to select cost thresholds that corresponded to how much it would cost to use particular emissions-reduction technologies, and it applied that “uniform control stringency to EGUs within the covered upwind states.” Id., at 36680.^{7 7}EPA ultimately selected (based on nationwide data) a cost threshold of $1,800 per ton of NOx reduction that would apply in earlier years, and a cost threshold of $11,000 per ton that would apply in later years. 88 Fed. Reg. 36749, 36846. These cost thresholds corresponded to the cost of different emissions-control measures: For example, EGUs can “[r]etrof[t] state-of-the-art combustion controls” and “[o]ptimiz[e] idled SCRs” for less than $1,800 per ton, and they can “[i]nstal[l new] SCR[s]” for less than $11,000 per ton. Ozone Transport Policy Analysis Final Rule TSD 5 (EPA–HQ–OAR–2021–0668–1080, Mar. 2023) (Final Ozone Analysis). EPA then calculated each State's emissions budget based on the assumption that the State's EGUs would implement the emissions-reductions strategies that cost less than the chosen thresholds. See n. 6, supra; 88 Fed. Reg. 36762; Final Ozone Analysis 6, 9. Given the likelihood that In fact, some commenters criticized EPA's reliance on a “nationwide data set” to calculate emissions limits, arguing that EPA should “limit the dataset to . . . just the covered states”—an approach that would have made the cost- effectiveness thresholds depend on which States were covered.

Id., at 36723.

But EPA expressly defended its approach based on its “intention to identify a technology- specifc representative emissions rate” and its interest in “the performance potential of a technology”—which were best served by the “largest dataset possible (i. e., nation wide).” Id., at 36723–36724 (emphasis added). EPA explained that it used the same approach it had successfully applied in previous rulemakings: It “derive[d] technology performance averages” based on nationwide data. Id., at 36724. Then it applied the relevant industry standard “on a uniform basis” to each emitter across the covered States. Id., at 36817.⁸ The Court, perhaps recognizing the problem that the FIP's seemingly state-agnostic methodology poses for its theory, throws at the wall a cherry-picked assortment of EPA state- EPA selected cost thresholds based on nationwide data, each State's budget would be the same even if the FIP covered different States. ⁸While EPA's methodology with respect to other industrial sources (non-EGUs) was more complicated, it also seems to have relied on nationwide data. EPA chose a “$7,500 marginal cost-per-ton threshold,” 88 Fed. Reg. 36740, which corresponded to the point of diminishing returns (the “knee in the curve”) when EPA assessed the impact of emissions controls in the highest impact industries and in “all industries” on the total “[o]zone [s]eason NOx [r]eductio[n]” potential, Technical Memorandum, Screening Assessment of Potential Emissions Reductions, Air Quality Impacts, and Costs From Non-EGU Emissions Units for 2026, p. 4 (EPA–HQ–OAR– 2021–0668–0150, Feb. 2022) (boldface omitted). This fgure thus appears to have been determined based on industry-wide cost and emissions data rather than state-specifc calculations. So too with the specifc emissions limits EPA decided could be implemented for less than that cost. See, e. g., 88 Fed. Reg. 36825 (“EPA based the proposed emissions limits for cement kilns on the types of limits being met across the nation”). Page Proof Pending Publication ments mentioning state data. See ante, at 286–287, 299– 300, n. 14. None stick. The fundamental problem with the Court's citations is that they discuss analyses that EPA performed after it chose cost thresholds and emissions limits based on nationwide industry data. EPA did assess the impact on downwind States if particular upwind States met the proposed emissions limits, and that impact depended on the States included in the modeling. Ante, at 286, 299–300, n. 14. But EPA said that these “ `fndings regarding air quality improvement,' ” ante, at 300, n. 14 (quoting 88 Fed. Reg. 36741), served only to “cement EPA's identifcation of the selected . . . mitigation measures as the appropriate control stringency,” 88 Fed. Reg. 36741 (emphasis added); see Denial 18. EPA explained that the statutory requirement to “eliminate signifcant contribution” depends on the implementation of cost-effective emissions controls at individual “industrial sources,” not some overall impact on “downwind areas' nonattainment and maintenance problems.” 88 Fed. Reg.

36741. EPA assessed the FIP's impact assuming the participation of particular States primarily to ensure that its emissions limits did not result in “overcontrol”—i. e., more reductions than necessary to help downwind States comply with the NAAQS. Ibid.; see EME Homer City, 572 U. S., at 521.

The technical document that the Court cites, ante, at 286, makes this point clear: “The downwind air quality impacts are used to inform EPA's assessment of potential overcontrol.” Proposed Ozone Analysis 31.

EPA's analysis confrmed that its chosen emissions limits would not result in overcontrol if they were implemented in the States originally covered by the FIP. 88 Fed. Reg.

36741. Importantly, implementing the FIP “in fewer upwind states does not (and cannot possibly) result in overcontrol” given that “there was no overcontrol even when more states, making more emission reductions, were included.”

Denial 22. So the fact that EPA used state-specifc data in Page Proof Pending Publication its overcontrol analysis does not mean that the FIP's emissions limits depended on the number of States it covered. And the inclusion of fewer States in that analysis logically could not have affected the results.

Thus, EPA generally characterized the FIP's emissions limits as dependent on nationwide data, not on any particular set of States.⁹ Confrming this interpretation, the fnal rule contemplates its application to a different number of States. It recognizes that “states may replace FIPs with SIPs if EPA approves them,” and several sections explain how States may exit this FIP. 88 Fed. Reg. 36753, 36838–36843. And the rule's severability provision explains that EPA views the plan as “severable along . . . state and/or tribal jurisdictional lines.” Id., at 36693.

Moreover, EPA justifed the FIP's severability: EPA “must address good neighbor obligations as expeditiously as practicable and by no later than the next applicable attainment date”; severability serves “important public health and environmental benefts” and ensures that stakeholders can “rely on this fnal rule in their planning.” Ibid. These rationales align with EPA's response to critics of its decision to propose a FIP before fnalizing its SIP disapprovals: Quickly proposing a FIP—just like keeping the FIP in place even if some States drop out—“is a reasonable and prudent means ⁹The Court argues that EPA equated the framework it used here with the one that we described in EME Homer City. Ante, at 299, and n. 13. But even if EME described an approach that selected cost thresholds “only after” conducting downwind air-quality assessments, ante, at 299, n. 13, it is not clear that the Good Neighbor Plan adopted this aspect of the EME framework. In fact, there are other key similarities between this FIP and EME's approach. For example, the fnal FIP refers to EME in order to note that EPA's “uniform framework of policy judgments”—i. e., applying the same cost thresholds with “[n]ationwide consistency”—was upheld in that case. 88 Fed. Reg. 36673. And the fnal FIP identifes cost thresholds that EPA chose based on nationwide data—like $1,800 based on the cost of EGUs “optimiz[ing] . . . existing SCRs and SNCRs.” Id., at 36749.

Page Proof Pending Publication of assuring that [EPA's] statutory obligation to reduce air pollution affecting the health and welfare of people in downwind states is implemented without delay.” Response to Comments 151.

Given these justifcations and the state-agnostic methodology apparent in the fnal rule, EPA's “ `path may reasonably be discerned.' ” State Farm, 463 U. S., at 43. The FIP's cost thresholds and emissions limits did not depend in any signifcant way on the number of States included, so the drawbacks of severability were minimal.

On the other hand, severability was necessary so that EPA could fulfll, to the greatest extent possible, its statutory obligation to eliminate the significant ozone contributions of upwind States and reduce harmful pollution in downwind States in time to meet the attainment deadlines. See Response to Comments 150 (noting the August 2024 ozone-NAAQS attainment deadline). If the FIP were not severable, EPA would have to go back to the drawing board for all States whenever a single State is removed—thwarting its mission for little reason.¹⁰ Finally, it is unlikely that EPA's response to comments evinces a “fail[ure] to consider an important aspect of the problem.” State Farm, 463 U. S., at 43 (emphasis added).

An agency must respond to “ `relevant' and `signifcant' public comments,” and that requirement is not “particularly demanding”; the “agency need not respond at all to comments that are `purely speculative and do not disclose the factual or policy basis on which they rest.' ” Public Citizen, Inc. v. ¹⁰The Court claims that the severability clause is evidence that EPA “had notice of the objection the applicants seek to press in court,” yet EPA's justifcations for it did not address (alleged) concerns about how the cost-effectiveness thresholds would change with fewer States. Ante, at 296–297. But as explained above, commenters did not raise that issue with specifcity; they simply pointed out that some SIP disapprovals might be invalid. The severability clause is evidence that EPA was aware of that possibility, and the clause was EPA's response to it. Page Proof Pending Publication FAA, 988 F. 2d 186, 197 (CADC 1993) (quoting Home Box Offce, Inc. v. FCC, 567 F. 2d 9, 35, and n. 58 (CADC 1977) (per curiam); emphasis added); see § 7607(d)(6)(B) (EPA must respond to “signifcant” comments). EPA received hundreds of comments, and its response numbered nearly 1,100 pages. Given the likelihood that the FIP's emissions limits did not depend on the covered States, the risk of it applying to fewer States may not be “important,” and comments purportedly raising that possibility might not be “relevant” and “signifcant.” Moreover, the one comment that vaguely referred to a need for a “new assessment and modeling,” Air Stewardship Comments 14, was “purely speculative” and “disclose[d]” no “factual or policy basis”; it likely merited no response, Home Box Offce, 567 F. 2d, at 35, n. 58. Requiring more from EPA risks the “sort of unwarranted judicial examination of perceived procedural shortcomings” that might “seriously interfere with that process prescribed by Congress.” Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U. S. 519, 548 (1978).¹¹

C

Applicants face one more impediment: the Clean Air Act's stringent harmless-error rule. A court “reviewing alleged procedural errors . . . may invalidate [an EPA] rule only if the errors were so serious and related to matters of such central relevance to the rule that there is a substantial like lihood that the rule would have been signifcantly changed if such errors had not been made.” § 7607(d)(8) (emphasis added). This provision appears “tailor-made to undo” any “rigid presumption of vacatur” that might apply in other con¹¹Despite the Court's suggestion of forfeiture, ante, at 298, EPA could not have forfeited the argument that the comments the Court cites were too insubstantial to merit a response. The Court relies on comments that were not raised until the applicants' reply briefs or that were uncovered later by the Court itself. See, e. g., Reply Brief in No. 23A351, p. 11 (raising the Air Stewardship Coalition “modeling” comment for the frst time). Page Proof Pending Publication texts. N. Bagley, Remedial Restraint in Administrative Law, 117 Colum. L. Rev. 253, 291 (2017).

The alleged error here plausibly is subject to § 7607(d)(8)'s harmless-error rule. As explained above, the Court does not suggest that it is substantively “[un]reasonable” to apply the FIP to fewer States, only that EPA did not “reasonably explai[n]” the FIP's severability in response to comments. Prometheus, 592 U. S., at 423. That is arguably an “alleged procedural error” within the meaning of § 7607(d)(8). In fact, the Act contemplates that at least some “arbitrary or capricious” challenges allege failures to “observ[e] . . . procedure required by law,” and such challenges may only succeed if § 7607(d)(8)'s “condition is . . . met.” § 7607(d)(9)(D). If the Act's harmless-error rule applies, applicants are unlikely to prevail. Given the apparent lack of connection between the number of States covered and the FIP's methodology for determining cost thresholds and emissions limits, it is diffcult to imagine a “substantial” likelihood that the rule would have been “signifcantly” different had EPA just responded more thoroughly. In fact, applicants seem to have conceded as much. See Tr. of Oral Arg. 6 (“[W]ith full candor to the Court, [the cost threshold] could be the same or even be more expensive”); id., at 9 (“I can't tell you what that looks like, whether there is a difference in the obligations or not”). And EPA, the Court says, had “notice” of the alleged concern that the cost thresholds might change with different States. Ante, at 296. Yet EPA still chose to make the FIP severable because of its statutory obligation to reduce downwind pollution—an obligation it repeatedly referenced. See, e. g., 88 Fed. Reg. 36693; Response to Comments 149–151.

Would that same EPA have “signifcantly changed” the FIP had it just explained more thoroughly why the plan did not depend on the States covered? ¹² And on top of all this, ¹²The Court faults EPA for “refus[ing] to say with certainty” at oral argument that it would have reached the same conclusions if different States were included in the FIP. Ante, at 299. But § 7607(d)(8) does not Page Proof Pending Publication EPA has in fact refused to reconsider the FIP now that it applies to fewer States, explaining in detail why its methodology was unaffected by the States it covered.¹³ * * * With little to say in response to the FIP's apparent state- agnostic methodology for setting emissions limits and the Clean Air Act's stringent harmless-error rule, the Court resorts to raising forfeiture. Ante, at 298–299. But it is the Court that goes out of its way to develop a failure-to-explain theory largely absent from applicants' briefs.

One can search diligently in the hundreds of pages of applicants' opening briefs for the Court's theory—that EPA failed to explain in its fnal rule why the FIP's cost-effectiveness thresholds for imposing emissions limits do not shift with a different mix of States—and be left wondering where the Court found it. That theory appears not to have crystal- require the Government to show that the rule would be the same; it is most naturally read to require the challenger to demonstrate a “substantial likelihood that the rule would have been signifcantly changed if such errors had not been made.” (Emphasis added.) And though § 7607(d)(9)(A) appears to allow reversal of “ `any' ” arbitrary or capricious “ `action,' ” ante, at 299, § 7607(d)(8)'s more specifc harmless-error rule and § 7607(d)(9)(D)'s more specifc requirements for reversal based on arbitrary or capricious procedural errors would seem to control.

¹³The Court claims that the Clean Air Act prevents us from considering EPA's denial of reconsideration. Ante, at 295, n. 11. But it is not obvious that the relevant provision of the Act—§ 7607(d)(7)(A)'s defnition of the “record for judicial review”—bars consideration of later developments for purposes of the Act's stringent harmless-error rule, § 7607(d)(8). Even assuming that the denial of reconsideration itself cannot count as evidence of harmlessness, we are judging applicants' likelihood of success on the merits. On the merits, we can expect EPA to make just the sort of arguments it made in its denial: EPA likely will explain why the covered States did not matter by citing and interpreting material in the record. See, e. g., Denial 11 (“Record Basis Establishing Why the Plan Functions Independently by State”); id., at 15, and nn. 16–18 (citing the fnal rule and technical support documents on the rulemaking docket).

Page Proof Pending Publication Page Proof Pending Publication lized until oral argument, during which counsel for the state applicants struggled to locate it in the States' brief. Tr. of Oral Arg. 11–12. Consider just one illustrative example.

Given the importance to the Court's theory of how the “knee in the curve” might change with different States, see ante, at 287, 288, and n. 4, 293, one might expect to fnd some mention of that idea in applicants' briefs. One would be wrong.

Given that applicants' theory has evolved throughout the course of this litigation, we can hardly fault EPA for failing to raise every potentially meritorious defense in its response brief.

That is particularly true given the compressed briefng schedule in this litigation's emergency posture: The Court gave EPA less than two weeks to respond to multiple applications raising a host of general and industry-specifc technical challenges, fled less than a week earlier. Even still, EPA raised § 7607(d)(7)(B)'s procedural bar. Brief for Respondents 19. And on the merits, EPA expressly argued that the FIP's “viability and validity do not depend on the number of jurisdictions it covers”; the “Rule need not apply to any minimum number of States in order to operate coherently.” Id., at 24. EPA could also have demonstrated how the FIP's state-agnostic methodology for selecting cost thresholds was apparent in the fnal rule. But EPA cannot have forfeited that more specifc point because applicants did not raise it to begin with.

Because EPA did not forfeit these responses to the merits of applicants' arbitrary-or-capricious challenge, there is no need to consider whether a departure from our typical approach to forfeited arguments is justifed. See ante, at 298– 299. It remains applicants' burden to show that the FIP's alleged dependence on the covered States likely was an “important” problem that EPA “entirely failed to consider.”

State Farm, 463 U. S., at 43. And that is on top of their burden to overcome § 7607(d)(7)(B)'s procedural bar and the lack of “signifcant,” specifc comments raising this issue. § 7607(d)(6)(B).

Finally, I would exercise our discretion to consider § 7607(d)(8)'s harmless-error rule. Even putting aside the expedited briefng schedule and the limited discussion of the Court's theory in applicants' briefs, applicants bear the burden in seeking emergency relief to show a likelihood of success on the merits. In other words, we must predict whether applicants will overcome every barrier to relief at the end of the day, after full merits briefng and argument in the lower courts and, potentially, again in this Court. Section 7607(d)(8)'s harmless-error rule is one such important obstacle, and EPA has already signaled that it will raise it as litigation progresses. See Denial 35, n. 38 (arguing that any failure to more fully explain “how the Rule is not interdependent” is harmless error under § 7607(d)(8)). I see no reason not to consider it now.

III

Given the emergency posture of this litigation, my views on the merits of the failure-to-explain objection and the application of the Clean Air Act's procedural bar and harmless- error rule are tentative. But even a tentative adverse conclusion can undermine applicants' likelihood of success. And applicants, to prevail, must run the table; they face the daunting task of surmounting all of these signifcant obstacles. They are unlikely to succeed.

The Court, seizing on a barely briefed failure-to-explain theory, grants relief anyway. It enjoins the Good Neighbor Plan's enforcement against any state or industry applicant pending review in the D. C. Circuit and any petition for certiorari. Ante, at 300. Given the number of companies included and the timelines for review, the Court's injunction leaves large swaths of upwind States free to keep contributing signifcantly to their downwind neighbors' ozone problems for the next several years—even though the temporarily stayed SIP disapprovals may all be upheld and the FIP may yet cover all the original States. The Court justifes Page Proof Pending Publication this decision based on an alleged procedural error that likely had no impact on the plan. So its theory would require EPA only to confrm what we already know: EPA would have promulgated the same plan even if fewer States were covered. Rather than require this years-long exercise in futility, the equities counsel restraint.

Our emergency docket requires us to evaluate quickly the merits of applications without the beneft of full briefng and reasoned lower court opinions. See Does, 595 U. S., at ––– (opinion of Barrett, J.). Given those limitations, we should proceed all the more cautiously in cases like this one with voluminous, technical records and thorny legal questions. I respectfully dissent.

Page Proof Pending Publication Page Proof Pending Publication Reporter’s Note The attached opinion has been revised to refect the usual publication and citation style of the United States Reports. The revised pagination makes available the offcial United States Reports citation in advance of publication. The syllabus has been prepared by the Reporter of Decisions for the convenience of the reader and constitutes no part of the opinion of the Court. A list of counsel who argued or fled briefs in this case, and who were members of the bar of this Court at the time this case was argued, has been inserted following the syllabus. Other revisions may include adjustments to formatting, captions, citation form, and any errant punctuation. The following additional edits were made: p. 284, line 13 from bottom: “National” is replaced with “Natural” p. 286, line 8: “family of” is inserted before “ozone”; “nitrous oxide” is replaced with “nitrogen oxides” p. 286, line 9: “nitrous oxide” is replaced with “nitrogen-oxide“ p. 286, line 18: “nitrous” is replaced with “nitrogen” p. 286, line 20: “nitrous” is replaced with “nitrogen” p. 288, n. 4, line 7: “nitrous oxide” is replaced with “nitrogen-oxide” p. 308, line 18: “might” is replaced with “would” p. 310, line 12: “might” is replaced with “would”