Ky. Energy and Env't Cabinet v. EPA

• Court: Court of Appeals for the Sixth Circuit
• Decided: December 6, 2024
• Docket: 23-3225
• Status: Published

Opinion

RECOMMENDED FOR PUBLICATION Pursuant to Sixth Circuit I.O.P. 32.1(b) File Name: 24a0265p.06

UNITED STATES COURT OF APPEALS FOR THE SIXTH CIRCUIT

┐ COMMONWEALTH OF KENTUCKY (23-3216); KENTUCKY │ ENERGY AND ENVIRONMENT CABINET (23-3225), │ Petitioners, │ > Nos. 23-3216/3225 │ v. │ │ │ UNITED STATES ENVIRONMENTAL PROTECTION AGENCY; │ MICHAEL S. REGAN, Administrator, United States │ Environmental Protection Agency, │ │ Respondents. ┘

On Petition for Review of a Final Agency Action of the United States Environmental Protection Agency. Nos. EPA-HQ-OAR-2021-0663; EPA-R04-OAR-2021-0841.

Argued: May 8, 2024

Decided and Filed: December 6, 2024

Before: BOGGS, KETHLEDGE, and MURPHY, Circuit Judges. _________________

COUNSEL

ARGUED: Matthew F. Kuhn, OFFICE OF THE KENTUCKY ATTORNEY GENERAL, Frankfort, Kentucky, for Petitioner Commonwealth of Kentucky. Jarrod L. Bentley, KENTUCKY ENERGY AND ENVIRONMENT CABINET, Frankfort, Kentucky for Petitioner Kentucky Energy and Environment Cabinet. Jeffrey T. Hammons, UNITED STATES DEPARTMENT OF JUSTICE, Washington, D.C., for Respondents. Claiborne E. Walthall, OFFICE OF THE ATTORNEY GENERAL OF THE STATE OF NEW YORK, Albany, New York, for Amici Curiae. ON BRIEF: Matthew F. Kuhn, OFFICE OF THE KENTUCKY ATTORNEY GENERAL, Frankfort, Kentucky, for Petitioner Commonwealth of Kentucky. Jarrod L. Bentley, Joseph A. Newberg, II, KENTUCKY ENERGY AND ENVIRONMENT CABINET, Frankfort, Kentucky for Petitioner Kentucky Energy and Environment Cabinet. Jeffrey T. Hammons, UNITED STATES DEPARTMENT OF JUSTICE, Washington, D.C., for Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 2

Respondents. Claiborne E. Walthall, Elizabeth A. Brody, OFFICE OF THE ATTORNEY GENERAL OF THE STATE OF NEW YORK, Albany, New York, Deena Tumeh, Kathleen Riley, Neil Gormley, EARTHJUSTICE, Washington, D.C., Shaun A. Goho, Hayden W. Hashimoto, CLEAN AIR TASK FORCE, Boston, Massachusetts, for Amici Curiae.

MURPHY, J., delivered the opinion of the court in which BOGGS and KETHLEDGE, JJ., joined. MURPHY, J. (pp. 34–41), also delivered a separate concurring opinion. _________________

OPINION _________________

MURPHY, Circuit Judge. After the Environmental Protection Agency (EPA) changed its air-quality standard for ozone under the Clean Air Act, the States needed to amend their state plans to implement the new standard. To help the States with their plan revisions, the EPA issued two guidance memoranda. It told the States that they could use specific modeling to identify their emissions that cross state lines. And it told them that they presumptively need not worry about any interstate emissions that fall below a specific minimum threshold. This guidance led Kentucky to propose a plan that did not reduce its emissions further. But the EPA sat on Kentucky’s proposed plan for some two years—well past the Clean Air Act’s deadline for the agency to act. It then belatedly disapproved the plan. To Kentucky’s surprise, this disapproval rested on different modeling that came out after the EPA’s deadline and on a lower threshold than the one the EPA told Kentucky it could use. Kentucky petitioned our court to vacate the EPA’s disapproval. In response, the EPA sought to transfer Kentucky’s challenge to the D.C. Circuit because the EPA had disapproved Kentucky’s plan in a rule that also rejected 20 other state plans.

We deny the EPA’s motion to send this case to the D.C. Circuit. Kentucky properly sued in this court because the EPA’s disapproval was not a “nationally applicable . . . final action” or one “based on a determination of nationwide scope or effect[.]” 42 U.S.C. § 7607(b)(1). The EPA’s view that a rule touching less than half the country can qualify as “nationally applicable” conflicts with that phrase’s ordinary meaning and would create much confusion over where to sue. At any rate, the EPA cannot turn the quintessential “local” action (a state-plan decision) into a national one merely by combining it with others. To fend off legal challenges on the Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 3

merits, the EPA’s final rule also clarified that its decision to disapprove Kentucky’s plan rested on Kentucky’s unique facts. So its action was not “based on a determination of nationwide scope or effect.”

We next hold that the EPA’s disapproval of Kentucky’s plan violated the Administrative Procedure Act (APA). The EPA acted in an “arbitrary” way by telling Kentucky one thing and then doing another. 5 U.S.C. § 706(2)(A). It recommended that Kentucky use certain modeling and a certain threshold. Yet it denied Kentucky’s plan using different modeling and a different threshold. Ralph Waldo Emerson might have approved of this approach, since he once opined that “[a] foolish consistency is the hobgoblin of little minds[.]” Antonin Scalia, Assorted Canards of Contemporary Legal Analysis, 40 Case W. Res. L. Rev. 581, 587 (1989) (quoting Ralph Waldo Emerson, Self-Reliance, in Essays and English Traits 66 (C.W. Eliot ed. 1909)). But this “canard” has no place in legal reasoning because “[c]onsistency is the very foundation of the rule of law.” Id. at 588. Congress thus kept it out of our administrative law by passing the APA. Because the EPA has not justified its inconsistencies here, we vacate its disapproval of Kentucky’s plan.

I

A

Like most laws, the Clean Air Act represents a compromise of competing interests. Cf. Luna Perez v. Sturgis Pub. Schs., 598 U.S. 142, 150 (2023). Congress sought to reduce pollution. 42 U.S.C. § 7401(a)(1)–(2). But it also sought to respect state authority. Id. § 7401(a)(3)–(4). The Act thus seeks to improve air quality in a specific way: through “a model of cooperative federalism.” Sierra Club v. Korleski, 681 F.3d 342, 343 (6th Cir. 2012) (quoting Ellis v. Gallatin Steel Co., 390 F.3d 461, 467 (6th Cir. 2004)).

The cooperative process begins at the federal level. The EPA must impose and periodically revise “national ambient air quality standards” for each air pollutant. 42 U.S.C. § 7409(a)(1), (d). The EPA should set these standards at a level that will “protect the public health.” Id. § 7409(b)(1). Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 4

When the EPA adopts a revised air-quality standard for a pollutant, the regulatory process shifts to the States. Each State must develop an “implementation plan” (or “[s]tate plan” for short) that will maintain and enforce this standard within its borders. Id. §§ 7407(a); 7410(a). A State must submit its plan to the EPA for approval within three years of the revised air-quality standard. Id. § 7410(a)(1). Its plan must satisfy many statutory criteria. See id. § 7410(a)(2). The plan, for example, must contain the emissions limits required to meet the new standard. Id. § 7410(a)(2)(A).

Once the State submits its plan, the EPA must approve or disapprove it within 18 months. Id. § 7410(k)(1)–(3). The EPA must evaluate the plan only against the statutory criteria. If the plan meets the criteria, the EPA cannot reject it on the ground that the State has chosen an unwise method to attain the air-quality standard. See Ohio v. EPA, 603 U.S. 279, 284 (2024).

If the EPA disapproves a state plan, the agency must issue a “[f]ederal implementation plan” that meets the revised air-quality standard within the State. 42 U.S.C. § 7410(c)(1)(B). The agency may issue this federal plan “at any time within 2 years” from its disapproval of the state plan. Id. § 7410(c)(1). But if the State timely revises its plan to correct the original’s deficiencies and the EPA approves that revised plan, the EPA need not issue its own federal plan. Id. § 7410(c)(1)(B); see also EPA v. EME Homer City Generation, L.P., 572 U.S. 489, 507–10 (2014).

B

This case concerns a specific requirement that state plans must meet: the “Good Neighbor Provision.” EME Homer, 572 U.S. at 495. Pollutants emitted from a source (say, a power plant) in one State often do not stop at that State’s borders. Rather, the wind can carry the pollutants “over hundreds of miles” to one or more States. Id. at 496. This fact creates a classic “negative externality”: an upwind State can shift some of the cost of its activities (such as energy production) to downwind States. See id. at 495. Without national regulation, upwind States would have little incentive to take these external costs into account. See id. So downwind States might have to impose greater limits on their own activities to achieve acceptable pollution levels. See id. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 5

Congress adopted the Good Neighbor Provision to create the national coordination required for States to “internalize” these harms. See 42 U.S.C. § 7410(a)(2)(D). This provision requires a state plan to include terms that prohibit “any source or other type of emissions activity within the State from emitting any air pollutant in amounts which will . . . contribute significantly to nonattainment in, or interfere with maintenance by, any other State with respect to” an air-quality standard. Id. § 7410(a)(2)(D)(i). How will States know whether their emissions will have these interstate effects? To help them identify emissions that they may have to reduce, the EPA must classify all areas as “nonattainment” (if an area does not meet an air- quality standard) or “attainment” (if it does). See id. § 7407(d); EME Homer, 572 U.S. at 498.

The Supreme Court has addressed the Good Neighbor Provision once when evaluating the “Cross-State Air Pollution Rule” (or “Transport Rule”) that the EPA issued in 2011. EME Homer, 572 U.S. at 500. The Transport Rule adopted a “two-step approach” for determining whether pollution sources in an upwind State “contribute[d] significantly to nonattainment” in a downwind State. Id. (citation omitted). First, the EPA asked whether the upwind State had sources linked to a downwind State’s “receptor[s]” (places where the EPA measures air quality). Id. At this “screening” step, if an upwind State’s sources contributed only a small part of the total pollutant at a downwind receptor (less than 1% of the air-quality standard for that pollutant), the Transport Rule decided that the State did not “contribute[] significantly” to the downwind State’s pollution problems. Id. at 501. Second, the Transport Rule relied on a cost- benefit analysis to control emissions from States whose pollution contributions at a receptor exceeded this 1% threshold. Id. at 500–01. Ultimately, the Court relied on the now-overruled review scheme from Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984), to hold that the Good Neighbor Provision’s text permitted this approach. See EME Homer, 572 U.S. at 495–96; cf. Loper Bright Enters. v. Raimondo, 144 S. Ct. 2244, 2273 (2024).

C

This case also concerns a specific pollutant: ozone. Although ozone in the atmosphere helps protect us “from the sun’s radiation,” the pollutant at ground levels can harm human health (for example, by inflaming our lungs) and the earth’s vegetation (for example, by reducing crop yields). Ohio, 603 U.S. at 284; National Ambient Air Quality Standards for Ozone, 80 Fed. Reg. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 6

65292, 65302–11, 65369–73 (Oct. 26, 2015). In October 2015, the EPA lowered the air-quality standard for ozone from 75 to 70 parts per billion (or ppb). 80 Fed. Reg. at 65330, 65365. This revision required the States to submit revised implementation plans within three years. See 42 U.S.C. § 7410(a). To help States develop plans that would meet the Good Neighbor Provision, the EPA released two memoranda in 2018.

March 2018 Memorandum. The EPA issued its first guidance in March 2018. It told the States that it planned to determine each State’s good-neighbor obligations using the same approach from the Transport Rule (but the agency now divided this approach into four steps). Mar. Mem., J.A. 76–77. At Step 1, the relevant federal or state regulator should “identify downwind” locations that will struggle to attain or maintain the air-quality standard. Id. At Step 2, the regulator should identify the upwind States whose pollution sources “contribute enough to those downwind air quality problems to warrant further review[.]” Id. at 76. At Step 3, the regulator should rely on cost considerations (among other things) to calculate the amount of emissions reductions (if any) that an upwind State must make so that it does not contribute to the “downwind air quality problems[.]” Id. at 77. At Step 4, the regulator should identify “permanent and enforceable measures” to achieve these reductions. Id. The EPA told States that their plans could follow this approach or “alternative frameworks” that comported with the Good Neighbor Provision. Id.

This March 2018 Memorandum also included “modeling” data to predict ozone problems in downwind States in 2023 and to identify the upwind contributors. Id. The EPA suggested that the States could “consider using this national modeling to develop” their plans. Id. at 80. The modeling relied on data centered on pollution from 2011, so we will call it the “2011 modeling.” Id. at 77–80. It identified two types of downwind “receptors”: those showing that a location would not attain the ozone standard in 2023 (“nonattainment receptors”) and those showing that a location would struggle to maintain that standard in 2023 (“maintenance receptors”). Id. at 78.

August 2018 Memorandum. Five months later, the EPA issued more “recommendations” for state plans. Aug. Mem., J.A. 98. These recommendations addressed “Step 2”—the “screening” step at which the EPA eliminates the upwind States that contribute only a small Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 7

amount to downwind receptors. Id. at 99. As with the Transport Rule, the EPA historically screened out States that contributed less than 1% of an air-quality standard. See EME Homer, 572 U.S. at 500 & n.3. If the EPA applied this rule to the ozone standard of 70 ppb, it would set a .7 ppb “threshold” that a State’s contributions must exceed. Aug. Mem., J.A. 100. The August 2018 Memorandum compared this 1% rule (.7 ppb) to a larger threshold: 1 ppb. Id. It found that “the amount of upwind collective contribution” under both thresholds was “generally comparable[.]” Id. at 101. The EPA thus told States that they could use the higher 1 ppb threshold. Id. According to the EPA, if a State’s ozone contributions fell below 1 ppb at a receptor, the State could find that it did not contribute enough pollutants to warrant further review at Steps 3 and 4. Id. at 99. Yet the EPA also refused to guarantee approval of a state plan that followed this recommendation because the recommendation might “not apply to the facts and circumstances” of all States. Id. at 98.

D

Kentucky relied on these memoranda to draft its plan. It used the 2011 modeling from the March 2018 Memorandum and the 1 ppb threshold from the August 2018 Memorandum. State Plan, J.A. 151–52. At Step 2, these choices showed that Kentucky exceeded the screening threshold at only one maintenance receptor in Harford County, Maryland. Id. at 152; Mar. Mem., J.A. 89. Kentucky contributed 1.52 ppb of ozone to this location. State Plan, J.A. 152.

Turning to Step 3, Kentucky decided that it need not impose further emissions reductions despite the link to the Maryland receptor. Id. Kentucky gave several reasons for this conclusion. To begin with, the Maryland receptor was a “maintenance” receptor, so the pollution there was less severe than the pollution at a “nonattainment” receptor. Id. Kentucky’s ozone-producing emissions also would continue to decrease in future years because of other regulations and because of the expected “retirement of several” power plants. Id. at 163–66, 177–78. Kentucky also opined that “local” emissions in Maryland (such as local car traffic) contributed far more than Kentucky emissions to ozone in the area. Id. at 173–77. The Commonwealth believed that the EPA should compel “local sources” to reduce their emissions before requiring far-away sources to “over-control” their emissions. Id. at 173. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 8

During Kentucky’s notice-and-comment process, the EPA commented on its state plan. EPA Comments, J.A. 124–28. Among other things, Kentucky’s plan used an “intricate combination” of the 2011 modeling from the March 2018 Memorandum and additional modeling from a Kentucky contractor. Id. at 126. The EPA recommended that Kentucky simply use the EPA’s 2011 modeling and its 1 ppb screening threshold as an “alternative, more straightforward approach[.]” Id. The agency added that this approach would still connect Kentucky only to the Maryland monitor. Id. The EPA separately noted that Kentucky should specifically identify the power-plant closures that EPA’s 2011 modeling failed to capture. Id.

After this feedback, Kentucky formally submitted its plan in January 2019. State Plan, J.A. 129. When discussing the EPA’s comments, Kentucky “concur[red] with EPA’s” simplified proposal. Id. at 454. Despite the link to the Maryland monitor, Kentucky also continued to find that it need not reduce emissions. Id. at 178. In that respect, it identified the planned power- plant “shutdown” that the 2011 modeling had not considered. Id. at 454.

Under the Act, the EPA had until July 2020 to rule on Kentucky’s plan. 42 U.S.C. § 7410(k)(1)(B), (2). That date came and went without a ruling.

Nearly 20 months after the deadline, the EPA proposed a rule to disapprove Kentucky’s plan. Air Plan Disapproval, 87 Fed. Reg. 9498, 9498 (Feb. 22, 2022). This rule included two surprises. As for the first, the EPA used different modeling than the modeling in its March 2018 Memorandum. See id. at 9500–01. The new modeling centered on pollution in 2016, so we will call it the “2016v2 modeling.” Id. As for the second, the EPA rejected the 1 ppb threshold that its August 2018 Memorandum had recommended to screen out small upwind contributions. Id. at 9502–03. The agency proposed returning to the lower threshold that the Transport Rule had used: 1% of the air-quality standard (.7 ppb for the ozone standard). Id. These changes affected Kentucky in different ways. On the one hand, the new modeling no longer linked Kentucky to the Maryland receptor because that receptor was now adequately attaining the air-quality standard. Id. at 9507, 9509. On the other hand, the .7 ppb threshold linked Kentucky to three nonattainment receptors in Connecticut and one maintenance receptor in Pennsylvania. Id. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 9

A year later, the EPA disagreed with Kentucky’s objections and issued a final rule disapproving its plan. Air Plan Disapprovals, 88 Fed. Reg. 9336, 9356 (Feb. 13, 2023). This rule combined the EPA’s denial of Kentucky’s plan with its denial of 20 others. See id. at 9336– 38 & n.8. The rule contained two fresh surprises. The EPA again switched to new modeling: the “2016v3 modeling.” See id. at 9339, 9344. Although this modeling contained similar data as the 2016v2 modeling, it included “models, methods, and source datasets that became available in calendar years 2020 through 2022[.]” Id. at 9345. Further, the EPA now identified a third type of “receptor”—what it called a “violating monitor”—in addition to maintenance and nonattainment receptors. Id. at 9342. It chose this third group as a cautionary matter. Fresh information from 2021 and 2022 showed that certain locations were not attaining the ozone standard even though the EPA’s modeling predicted that these areas would meet the standard. Id. at 9349. At the same time, the EPA used these new monitors only on a “confirmatory basis” and chose not to rely on them as a sole ground to deny a state plan. Id.

How did these fresh changes affect Kentucky? The 2016v3 modeling and 1% threshold continued to link Kentucky to three Connecticut receptors. 2016v3 Modeling, J.A. 577. The EPA also linked Kentucky to four new violating monitors in Connecticut, New York, Michigan, and Ohio. Id. at 578; see 88 Fed. Reg. at 9356. Relying on these links at Step 2, the EPA found Kentucky’s proposal inadequate at Step 3. See 88 Fed. Reg. at 9356.

E

The EPA’s disapproval of these 21 state plans triggered a flood of litigation. Many States sought review of the EPA’s denials in their regional circuit courts. Kentucky and its Energy and Environment Cabinet (collectively, Kentucky) filed petitions for review in our court, asking us to vacate the EPA’s disapproval of their plan. The EPA responded by moving to transfer Kentucky’s case to the D.C. Circuit. Meanwhile, the Commonwealth moved to stay the EPA’s disapproval of its plan pending our full review. In May 2023, a motions panel granted a short administrative stay while it considered these motions. The same month, other circuit courts also stayed the denials of other state plans. See Ohio, 603 U.S. at 288–89. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 10

A few days after we granted our administrative stay, the EPA imposed its own federal implementation plan on Kentucky and 22 other States for the 2015 ozone standard. Federal “Good Neighbor Plan” for the 2015 Ozone National Ambient Air Quality Standards, 88 Fed. Reg. 36654, 36654 (June 5, 2023). States challenged that plan in the D.C. Circuit. See Ohio, 603 U.S. at 290.

As litigation over the federal plan progressed, regional circuits continued to stay the EPA’s disapproval of state plans. See id. In July 2023, we turned our administrative stay into a stay pending our full review. Kentucky v. EPA, 2023 WL 11871967, at *5 (6th Cir. July 25, 2023) (order). We also denied the EPA’s motion to transfer the suit to the D.C. Circuit. See id. at *1–3.

A short time later, the EPA stayed its federal plan for the States that had obtained judicial stays. Federal “Good Neighbor Plan” for the 2015 Ozone National Ambient Air Quality Standards, 88 Fed. Reg. 49295, 49295 (July 31, 2023). These stays grew to cover “12 of the 23” States subject to the federal plan. Ohio, 603 U.S. at 289–90. Yet this plan’s emissions reductions rested on the premise that all States would participate. See id. at 287–88, 293–94. The Supreme Court thus stayed the federal plan pending judicial review of that plan’s validity. See id. at 300.

* * *

The Supreme Court’s decision completes the backdrop against which we must review Kentucky’s petitions here. Those petitions raise three questions. Did Kentucky sue in the right court? If so, did the EPA lawfully deny Kentucky’s plan? If not, what is the proper remedy? Our short answers: Yes, No, and Vacatur. We will fill in the details to each answer in turn.

II. Did Kentucky Sue in the Right Court?

The EPA has renewed its request to transfer Kentucky’s petitions to the D.C. Circuit. Kentucky responds that we should reject this request based solely on the motions panel’s earlier order denying the EPA’s motion to transfer. See Kentucky, 2023 WL 11871967, at *1–3. According to Kentucky, we must “defer[]” to “the decisions of motions panels” and “cannot Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 11

simply” reject them. Wallace v. FedEx Corp., 764 F.3d 571, 583 (6th Cir. 2014). Yet we have never explained the amount of deference due these decisions. Indeed, why should we defer at all? We can (and sometimes do) depart from unpublished opinions. See Bell v. Johnson, 308 F.3d 594, 611 (6th Cir. 2002). And other circuits do not defer to similar interlocutory rulings. See Tex. Democratic Party v. Abbott, 978 F.3d 168, 176 (5th Cir. 2020); United States v. Lopez- Armenta, 400 F.3d 1173, 1175 (9th Cir. 2005); Am. Fed’n of Grain Millers, Loc. 24 v. Cargill Inc., 15 F.3d 726, 727 (7th Cir. 1994); 16AA Charles A. Wright et al., Federal Practice and Procedure § 3973.3, at 256–60 (5th ed. 2023). Still, we may save that question for another day. Even reviewing the EPA’s motion de novo, we agree with the motions panel that it lacks merit.

A. Background Law

The Clean Air Act allows a party injured by an EPA action to file a “petition for review” challenging the action in a circuit court of appeals. 42 U.S.C. § 7607(b)(1). The Act’s (lengthy) judicial-review provision contains three basic parts. First, the provision gives the D.C. Circuit the authority to review the EPA’s “nationally applicable” “regulations” or “final action”:

A petition for review of action of the Administrator in promulgating any national primary or secondary ambient air quality standard, any emission standard or requirement under section 7412 of this title, any standard of performance or requirement under section 7411 of this title,[] any standard under section 7521 of this title (other than a standard required to be prescribed under section 7521(b)(1) of this title), any determination under section 7521(b)(5) of this title, any control or prohibition under section 7545 of this title, any standard under section 7571 of this title, any rule issued under section 7413, 7419, or under section 7420 of this title, or any other nationally applicable regulations promulgated, or final action taken, by the Administrator under this chapter may be filed only in the United States Court of Appeals for the District of Columbia.

Id. Second, the provision gives the appropriate regional circuit court the authority to review the EPA’s “locally or regionally applicable” actions:

A petition for review of the Administrator’s action in approving or promulgating any implementation plan under section 7410 of this title or section 7411(d) of this title, any order under section 7411(j) of this title, under section 7412 of this title, under section 7419 of this title, or under section 7420 of this title, or his action under section 1857c-10(c)(2)(A), (B), or (C) of this title (as in effect before August 7, 1977) or under regulations thereunder, or revising regulations for enhanced monitoring and compliance certification programs under section Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 12

7414(a)(3) of this title, or any other final action of the Administrator under this chapter (including any denial or disapproval by the Administrator under subchapter I) which is locally or regionally applicable may be filed only in the United States Court of Appeals for the appropriate circuit.

Id. Third, the provision redirects even a locally or regionally applicable action to the D.C. Circuit if the action is “based on a determination of nationwide scope or effect” and the EPA notes that the action has this scope or effect:

Notwithstanding the preceding sentence a petition for review of any action referred to in such sentence may be filed only in the United States Court of Appeals for the District of Columbia if such action is based on a determination of nationwide scope or effect and if in taking such action the Administrator finds and publishes that such action is based on such a determination.

Id.

Our court has yet to decide on the nature of this provision. Does it impose a (nonwaivable) jurisdictional ban on the wrong circuit’s adjudication of a petition or instead a (waivable) limit on the proper venue in which to sue? See Wachovia Bank v. Schmidt, 546 U.S. 303, 315–16 (2006). Other courts have read § 7607(b) as granting jurisdiction to all circuit courts, and they have added that its circuit-specific rules identify only the proper venue. See Texas v. EPA, 829 F.3d 405, 418 (5th Cir. 2016); Dalton Trucking, Inc. v. EPA, 808 F.3d 875, 879–80 (D.C. Cir. 2015); Clean Water Action Council of Ne. Wis., Inc. v. EPA, 765 F.3d 749, 751–52 (7th Cir. 2014). But the parties did not brief this distinction. And our holding that Kentucky sued in the right court eliminates the need to decide it. Cf. Jaber v. Gonzales, 486 F.3d 223, 228 n.3 (6th Cir. 2007).

B. Application

The EPA offers two reasons why this suit belongs in the D.C. Circuit. It first argues that its disapproval of the 21 state plans qualifies as a “nationally applicable . . . final action.” And even if the disapproval of Kentucky’s plan was “locally or regionally applicable,” the EPA next argues that it was “based on a determination of nationwide scope or effect.” These claims have created a circuit split. Apart from Kentucky, several States have challenged the disapprovals of their plans in regional circuits. And most circuit courts have held that they (not the D.C. Circuit) Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 13

represent the proper tribunal for the suits. See West Virginia v. EPA, 90 F.4th 323, 331 (4th Cir. 2024) (citing cases). The Tenth Circuit, by contrast, has sent challenges to the denial of Oklahoma’s plan to the D.C. Circuit. See Oklahoma ex rel. Drummond v. EPA, 93 F.4th 1262, 1266–69 (10th Cir. 2024), cert. granted, 2024 WL 4529798 (U.S. Oct. 21, 2024). We agree with the courts that have rejected the EPA’s two arguments for D.C. Circuit review.

1. Did the EPA take a “nationally applicable” action when denying Kentucky’s plan?

The EPA argues that its rule disapproving the state plans qualifies as a single “nationally applicable . . . final action” rather than many “locally or regionally applicable” actions. 42 U.S.C. § 7607(b)(1). This claim misreads the phrases “nationally applicable” and “final action.”

Nationally Applicable. The EPA first misconstrues the phrase “nationally applicable” in § 7607(b)(1). As with any case, we start with the “ordinary meaning” of the words Congress used. Engine Mfrs. Ass’n v. S. Coast Air Quality Mgmt. Dist., 541 U.S. 246, 252 (2004) (quoting Park ‘N Fly, Inc. v. Dollar Park & Fly, Inc., 469 U.S. 189, 194 (1985)); see Nat’l Ass’n of Mfrs. v. Dep’t of Def., 583 U.S. 109, 127 (2018). The key words—“nationally applicable regulations promulgated, or final action taken,” by the EPA—convey that the challenged regulations or action must apply to the entire country. 42 U.S.C. § 7607(b)(1). Because the phrasal adjective “nationally applicable” modifies the nouns “regulations” and “action,” the regulations or action must “have reference to” (“applicable”) the “nation as a whole” (“nationally”). 1 Oxford English Dictionary 575 (2d ed. 1989); 10 Oxford English Dictionary, supra, at 235; Webster’s Third New Int’l Dictionary 105, 1505 (1976). We doubt that we need dictionaries for this point. Would anyone say that Congress passed a “nationally applicable” law if the law formally applied to only 40% of the country? No, the word “‘[n]ational’ contemplates an activity with a nationwide scope.” Black’s Law Dictionary 923 (5th ed. 1979). And here, all agree that the EPA’s rule formally applies to just 21 States—not the whole country.

Two canons of construction confirm this reading. The sentence in § 7607(b)(1) that contains the “nationally applicable” clause triggers the “ejusdem generis” canon. That canon applies to a catchall clause following a list of items. See Bissonnette v. LePage Bakeries Park Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 14

St., LLC, 601 U.S. 246, 252 (2024). It instructs courts to interpret the catchall as falling within the same class as the specific items that precede it. See id. So, for example, the phrase “seamen, railroad employees, or any other class of workers” covers only transportation workers—not all workers. 9 U.S.C. § 1; see Bissonnette, 601 U.S. at 252–53. The catchall clause here—“any other nationally applicable regulations promulgated, or final action taken, by the” EPA—also follows a list of specific actions that parties must challenge in the D.C. Circuit. 42 U.S.C. § 7607(b)(1). And as far as we can tell, those actions have nationwide applicability. See id. For example, challengers must sue in the D.C. Circuit over “any national primary or secondary ambient air quality standard[.]” Id. Likewise, they must sue in the D.C. Circuit over the EPA’s “standards” for stationary sources, hazardous pollutants, motor vehicles, fuels, and aircraft under 42 U.S.C. §§ 7411, 7412, 7521, 7545, and 7571. Id. § 7607(b)(1). And they must sue in that court over the general “rules” that govern enforcement (or similar) proceedings under §§ 7413, 7419, and 7420. Id. § 7607(b)(1). Given that the list in § 7607(b)(1) includes several national regulations, we read the catchall to have a similar breadth. See Bissonnette, 601 U.S. at 252.

Next, courts presume that Congress means to adopt “clear boundaries” in “jurisdictional statutes” to avoid wasteful litigation over the proper forum. Direct Mktg. Ass’n v. Brohl, 575 U.S. 1, 11 (2015); see Miss. ex rel. Hood v. AU Optronics Corp., 571 U.S. 161, 172–73 (2014); Hertz Corp. v. Friend, 559 U.S. 77, 94 (2010). Even if § 7607(b)(1) is a mere venue provision, the Supreme Court would likely extend this “clear boundaries” principle to § 7607(b)(1). See New York v. EPA, 133 F.3d 987, 990 (7th Cir. 1998). And our reading has “administrative simplicity” because it asks an objective question: Does the action formally apply to the nation? Hertz, 559 U.S. at 94. The EPA’s view, by contrast, would engender perpetual litigation over the actions that are “nationally applicable.” The agency says its rule here falls within that phrase because the rule covers 21 States in different parts of the country. Would this logic reach an action that covered only California, Maine, and Alabama because they fall in different parts of the country? If not, how many more States are needed? The EPA does not even try to answer these questions.

Final Action. The EPA also misconstrues the phrase “final action” in § 7607(b)(1). It argues that its disapprovals of the 21 state plans all qualify as a single consolidated action rather Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 15

than 21 separate actions. It thus advocates for a “rulemaking approach” that treats as a single “action” any regulatory decisions that the EPA combines into a single rule in the Federal Register. According to Kentucky, by contrast, the phrase “final action” adopts a “statutory approach” that treats each activity the Clean Air Act allows the EPA to take as a distinct “action”—whether the EPA announces these activities in a single rule or separate rules in the Federal Register.

We side with Kentucky in this debate. This time, though, the ordinary meaning of “action” does not help all that much. That word means simply a “thing done” or “something done.” Webster’s Third, supra, at 21; Black’s, supra, at 26. So these (circular) definitions leave the critical question unanswered: Did the EPA do 21 “things” or a single “thing” in the rule?

That said, Kentucky’s approach better fits the “structure” of the judicial-review provision. Engine Mfrs. Ass’n, 541 U.S. at 252. As a general matter, § 7607(b)(1) focuses on the statute (not the rulemaking) to distinguish the EPA actions that parties must challenge in the D.C. Circuit from those they must challenge in regional circuits. See Texas v. EPA, 2023 WL 7204840, at *4 (5th Cir. May 1, 2023) (per curiam). The provision ties the proper tribunal to the activity taken “under this chapter” (that is, under the Clean Air Act). 42 U.S.C. § 7607(b)(1). For example, § 7409(a)(1) requires the EPA to issue a national air-quality standard “for each air pollutant,” and § 7607(b)(1) sends a challenge to “any” such “standard” to the D.C. Circuit. On the other hand, § 7419(a)(1)(A) allows the EPA to issue a “primary nonferrous smelter order” for a specific smelter, and § 7607(b)(1) sends a challenge to such an “order” to a regional circuit. Nothing in § 7607(b)(1)’s text would allow the EPA to obtain D.C. Circuit review of the smelter order simply by combining it with the air-quality standard. Rather, the standard and the order represent distinct “actions” even if the EPA publishes them in the same rule in the Federal Register.

Congress also knows how to refer to rules in the Federal Register when it wants to. Section 7607(b)(1) makes the “notice” of an EPA action “in the Federal Register” relevant to the 60-day limit for suing. Id. Congress also required a “promulgated rule” to go through several procedures. Id. § 7607(d)(6)(A). Yet Congress used “final action” (not “promulgated rule”) when deciding where parties must sue. Id. § 7607(b)(1), (d)(6)(A). And courts normally Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 16

presume Congress acts intentionally with such different word choices. See Burlington N. & Santa Fe Ry. Co. v. White, 548 U.S. 53, 63 (2006); Gen. Motors Corp. v. United States, 496 U.S. 530, 541 (1990).

As a specific matter, § 7607(b)(1) clarifies that the disapproval of each state plan qualifies as a distinct “action” that falls on the local side of this divide. The Clean Air Act requires the EPA to “act” on each State’s “submission” on a plan-by-plan basis within a specific time. 42 U.S.C. § 7410(k)(2); see West Virginia, 90 F.4th at 330. The judicial-review provision then indicates that the EPA’s “action in approving . . . any implementation plan” (in the singular) belongs in a regional circuit. 42 U.S.C. § 7607(b)(1). And while the EPA denied Kentucky’s plan, this decision falls within the catchall for “locally or regionally applicable” actions (which includes a “denial or disapproval”). Id. As then-Judge Kavanaugh explained, EPA action on a state plan qualifies as the “prototypical” decision that belongs in a regional circuit because of its local character. Am. Rd. & Transp. Builders Ass’n v. EPA, 705 F.3d 453, 455 (D.C. Cir. 2013). So each state-plan denial in the EPA’s rule qualifies as a distinct “action” under the Clean Air Act. This reading makes this case easy: the EPA’s disapproval of Kentucky’s plan is a locally applicable action.

The EPA’s contrary arguments do not convince us otherwise. The agency offers no textual theory for its view that the phrase “nationally applicable” can reach an action applying to a hodgepodge of States that make up less than half the nation. It instead jumps to precedent, suggesting that courts have treated EPA actions as “nationally applicable” if they affect a large (unidentified) number of States. See S. Ill. Power Coop. v. EPA, 863 F.3d 666, 671 (7th Cir. 2017); ATK Launch Sys., Inc. v. EPA, 651 F.3d 1194, 1197 (10th Cir. 2011); Texas v. EPA, 2011 WL 710598, at *3–4 (5th Cir. Feb. 24, 2011); W. Va. Chamber of Com. v. Browner, 1998 WL 827315, at *2, *6–8 (4th Cir. Dec. 1, 1998) (per curiam). But the Fifth Circuit has since clarified that the EPA misread its precedent. See Calumet Shreveport Refin., L.L.C. v. EPA, 86 F.4th 1121, 1131 (5th Cir. 2023), cert. granted, 2024 WL 4529794 (U.S. Oct. 21, 2024). True, a regulation can be “nationally applicable” if its “legal effect” applies to the whole country—even if its practical effect is felt only by a subset of States. Id. The EPA could, for example, adopt a universal regulation imposing requirements that practically apply to only some States Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 17

(say, because of their different geographies or because the remaining States already satisfy the regulation). But the EPA seeks to go well beyond that valid principle because it makes no claim that a rule disapproving 21 state plans formally applies everywhere.

As for the remaining circuits, we disagree with their view that a regulation need not regulate the nation as a whole to be “nationally applicable.” Take the Tenth Circuit’s decision in ATK Launch Systems, 651 F.3d at 1197. It did not try to interpret the phrase “nationally applicable.” Rather, it merely concluded that a regulation’s reach “from coast to coast and beyond” suffices even if the regulation does not formally apply to the whole nation. Id. But the court did not say why. And it left many questions unanswered. How many States must a rule cover? How spread out must the covered States be? Nothing in the statute can answer these questions.

Next, the EPA suggests that we must consider the “face of [its] rule”—not the rule’s “practical effect” as applied to Kentucky—to decide whether it is “nationally applicable.” Respondents’ Br. 32 (quoting ATK Launch Sys., 651 F.3d at 1197). This argument conflates the rule issued in the Federal Register (the EPA’s words) with the “final action” that the EPA takes (the statute’s words). 42 U.S.C. § 7607(b)(1). Those phrases do not mean the same thing. And the relevant final action (the denial of Kentucky’s plan) applies only to Kentucky on its face.

The EPA also argues that its disapprovals of the 21 plans represent a single “nationally applicable” action because the disapprovals rested on a “uniform legal interpretation and common, nationwide analytical methods[.]” Respondents’ Br. 35 (quoting 88 Fed. Reg. at 9380– 81). This theory conflicts with § 7607(b)(1)’s text and structure. Starting with text, the phrase “nationally applicable” modifies the action that the EPA takes—not the reasons for it. And here, the action was a “prototypical” local decision. Am. Rd. & Transp. Builders Ass’n, 705 F.3d at 455. Turning to structure, Congress recognized that local actions might rest on national decisions. So it allowed D.C. Circuit review if the EPA showed that a local action was “based on a determination of nationwide scope or effect[.]” 42 U.S.C. § 7607(b)(1). We would unravel this layered scheme if we looked to an action’s justification at the start to decide whether the action is national or local. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 18

The EPA ends with policy. It suggests that Kentucky’s view inefficiently requires many courts to “concurrently” review actions that rest on the “same legal interpretations, policy decisions, and analytical methodology,” which could generate “inconsistent” results. Respondents’ Br. 37. Yet we see competing “wisdom” in “allowing difficult issues to mature through full consideration” by different courts. E.I. du Pont de Nemours & Co. v. Train, 430 U.S. 112, 135 n.26 (1977). This expanded review increases the likelihood that the Supreme Court will have all arguments before it when it resolves the issue. See id. Regardless, the axiom that “no legislation pursues its purposes at all costs” applies just as much to procedural provisions as it does to substantive ones. CTS Corp. v. Waldburger, 573 U.S. 1, 12 (2014) (quoting Rodriguez v. United States, 480 U.S. 522, 525–26 (1987) (per curiam)). We thus must enforce the judicial-review provision as written even if it conflicts with the EPA’s views of “sound policy.” Nat’l Ass’n of Mfrs., 583 U.S. at 132 (quoting Florida Power & Light Co. v. Lorion, 470 U.S. 729, 746 (1985)). And the written text makes clear that the EPA took a locally applicable action when disapproving Kentucky’s plan.

2. Is the denial of Kentucky’s plan “based on a determination of nationwide scope or effect”?

The EPA alternatively claims that, even if its disapproval of Kentucky’s plan was “locally or regionally applicable,” that disapproval was “based on a determination of nationwide scope or effect” under § 7607(b)(1). 88 Fed. Reg. at 9380. The EPA said as much in the final rule itself, so it met the separate requirement that it “publish” this finding. Id.; 42 U.S.C. § 7607(b)(1). The agency reasoned that all 21 disapprovals were “based on a common core of nationwide policy judgments and technical analysis concerning the interstate transport of pollutants throughout the continental U.S.” 88 Fed. Reg. at 9380. The EPA then listed three examples. First, the agency used the same four-step “framework” when disapproving all 21 plans. Id. Second, the disapprovals rested on the 2016v3 modeling and a uniform approach to identifying maintenance and nonattainment receptors. Id. at 9380–81. Third, the EPA considered requests to use “alternative approaches or alternative sets of data” under the same standards. Id. at 9381. This interpretation misreads two other phrases in § 7607(b)(1): “determination” and “based on.” Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 19

Determination. The EPA argues that each analytical step in the chain of reasoning to its ultimate decision for a “final action” qualifies as an independent “determination” that triggers review in the D.C. Circuit if at least one step (considered in isolation) has a “nationwide scope or effect.” 42 U.S.C. § 7607(b)(1). To give the obvious example, the EPA reasons that every approval or disapproval of a state plan includes the choice to follow its four-step “framework” for applying the Good Neighbor Provision. 88 Fed. Reg. at 9380. Because this framework applies everywhere, the EPA says that every one of its actions on a state plan will be “based on a determination of nationwide scope or effect” about the Good Neighbor Provision’s meaning.

This broad reading of “determination” lacks merit. We again start with the “ordinary meaning” of the text. Engine Mfrs. Ass’n, 541 U.S. at 252. Admittedly, “determination” in the abstract could mean what the EPA says because it can refer to any academic resolution of a “question by argument or reasoning.” Webster’s Third, supra, at 616. But the EPA’s reading does not fit the context. When used to describe a ruling from an “administrative agency,” Black’s Law Dictionary, supra, at 405, “determination” has a more precise “legal meaning” that refers to the agency’s ultimate decision—not each preliminary step on the road to that decision. Smith v. Spizzirri, 601 U.S. 472, 477 (2024); United States v. Hansen, 599 U.S. 762, 774 (2023). In other words, the word “implies an ending or finality of a controversy or suit,” Black’s Law Dictionary, supra, at 405, and directs us to look to the “authoritative decision” of the “matter at issue,” 4 Oxford English Dictionary, supra, at 548; Webster’s Third, supra, at 616; see also McQuillin v. Hartford Life & Accident Ins. Co., 36 F.4th 416, 420 (2d Cir. 2022). This understanding shows that courts should ask whether the ultimate decision underlying the EPA’s “final action” has a “nationwide scope or effect.” 42 U.S.C. § 7607(b)(1); see West Virginia, 90 F.4th at 328–29.

The EPA’s “unbounded interpretation,” by contrast, would upset Congress’s “careful delineation” of the actions that parties should challenge in the D.C. Circuit and those that they should challenge in regional circuits. Fischer v. United States, 603 U.S. 480, 493 (2024). If “determination” covered each discrete choice the EPA makes en route to a final decision, every “final action” rests on a nationwide “determination.” After all, the EPA always must ground its actions in a “national rule or standard” in the Clean Air Act or its regulations. West Virginia, 90 Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 20

F.4th at 328. The agency cannot haphazardly regulate parties by adopting one view of the law for some States and the opposite view for others. The EPA’s reading thus would allow it to send every action to the D.C. Circuit despite Congress’s presumptive choice to keep local actions local.

To be sure, the EPA’s reading would not make Congress’s choice entirely “superfluous.” Fischer, 603 U.S. at 493. Even if all local actions rest on nationwide determinations (as the EPA implies), the agency still must “publish[]” a finding to that effect. 42 U.S.C. § 7607(b)(1). If it opts not to, parties could sue in regional circuits. The EPA thus asks us to give it total discretion to decide where parties should sue. But this fact confirms its error because § 7607(b)(1) does not grant it this discretion. Recall that the relevant sentence sends a local action to the D.C. Circuit only if both the action “is based on a determination of nationwide scope or effect and” the EPA “publishes that such action is based on such a determination.” Id. (emphasis added). So the action must rest on a nationwide determination (the first requirement) independent of the EPA’s published views (the second requirement). Congress would not have written this sentence in the conjunctive if it meant to give the EPA complete discretion. It would have simply said that a local action belongs in the D.C. Circuit if the EPA finds that it rests on a national determination. All told, then, “determination” refers to the EPA’s “authoritative” answer to the question it resolves when taking an action. 4 Oxford English Dictionary, supra, at 548.

This definition proves the EPA’s error here. The agency suggests that its preliminary choices (such as its use of the four-step framework for the Good Neighbor Provision) were independent “determination[s]” under § 7607(b)(1). But those choices did not “end[]” the “controversy”: whether the EPA should approve Kentucky’s plan. 4 Oxford English Dictionary, supra, at 548. Instead, the “determination” underlying the EPA’s disapproval was its ultimate decision that Kentucky did not satisfy the Good Neighbor Provision. See 88 Fed. Reg. at 9356.

We thus must ask whether this decision about Kentucky’s plan had a “nationwide scope or effect” under § 7607(b)(1). It did not. Like “nationally applicable,” the word “nationwide” shows that a determination’s “scope or effect” must exist “throughout [the] entire nation.” Webster’s Third, supra, at 1505; West Virginia, 90 F.4th at 328. And the phrase “scope or effect” shows that this provision reaches EPA decisions that apply to the entire country as a legal Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 21

matter (de jure) or as a practical one (de facto). That is, a determination might have a nationwide “scope” if its formal “area” of operation covers the country. 14 Oxford English Dictionary, supra, at 672; see Webster’s Third, supra, at 2035. And it might have a nationwide “effect” if its “operative influence” is felt everywhere. 5 Oxford English Dictionary, supra, at 79; see Webster’s Third, supra, at 724.

Nothing about the EPA’s decision to deny Kentucky’s plan has this scope or effect. To the contrary, the agency decided that Kentucky’s plan did not satisfy the Good Neighbor Provision due to “circumstances . . . unique” to that plan. West Virginia, 90 F.4th at 328. To start, the EPA found that Kentucky wrongly suggested at Step 2 of its framework that the Commonwealth’s emissions sources were linked only to a Maryland receptor. See 87 Fed. Reg. at 9509. The EPA decided instead that Kentucky sources were linked to three Connecticut receptors. 88 Fed. Reg. at 9356; 2016v3 Modeling, J.A. 577. Next, the EPA disagreed with Kentucky’s “particular arguments” for using the higher 1 ppb threshold (rather than the lower 1% threshold) to weed out the receptors with which it had an insufficient connection. 88 Fed. Reg. at 9373 (emphasis added). Because Kentucky’s connection to the Connecticut receptors fell in between these thresholds, the EPA proceeded to Step 3. See id. at 9356. It lastly found that Kentucky did not adequately show at that step why it need not reduce any emissions to help the Connecticut receptors attain the ozone standard. Id. This logic applied only to Kentucky and had, at most, regional effects. So the EPA does not even try to argue that its final “determination” had any sort of nationwide scope or effect.

Based On. Apart from the EPA’s overbroad reading of “determination,” it disregards that the final action must be “based on” the identified determinations. 42 U.S.C. § 7607(b)(1). In ordinary English, one would say that a person’s action (say, taking an umbrella on a trip) is “based on” a factor (say, the high chance of rain) if the person has “ground[ed]” the action in that factor. McGraw-Hill’s Dictionary of American Idioms and Phrasal Verbs 34 (2005). When a statute uses a phrase like “based on” to connect one thing to another in this way, the Supreme Court has relied on a “traditional background principle[]”—but-for causation—to eliminate factors that an action could not have been based on. Burrage v. United States, 571 U.S. 204, 213–14 (2014); Safeco Ins. Co. of Am. v. Burr, 551 U.S. 47, 63 (2007); see also Univ. of Tex. Sw. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 22

Med. Ctr. v. Nassar, 570 U.S. 338, 350 (2013). Under this test, the person’s decision to take an umbrella could not have been based on the chance of rain unless the person would have left the umbrella at home “but for” this chance. Burrage, 571 U.S. at 212. That is, the chance of rain must have mattered to that decision. See id. at 211–12. Here, then, if the EPA would have taken the same “action” without the alleged “determination,” nobody would say the action was “based on” the determination.

This reading of “based on” independently dooms the EPA’s reliance on two of the alleged “determinations” identified in the final rule. The agency suggested that its denial of Kentucky’s plan rested on its decision to rely on the 2016v3 modeling and its use of “uniform” criteria to evaluate each State’s request to use a 1 ppb threshold rather than a threshold of 1% of the ozone standard (.7 ppb). 88 Fed. Reg. at 9380. In this court, however, the EPA claims that it would have denied Kentucky’s plan even if it had allowed Kentucky to use the 2011 modeling and 1 ppb threshold. Respondents’ Br. 56–57. Under that older modeling and higher threshold, Kentucky sources were still linked to a Maryland receptor. State Plan, J.A. 152. And the EPA argues that Kentucky did not adequately explain why it need not cut emissions to allow this receptor to meet the ozone standard. Respondents’ Br. 57–68. Yet if the EPA would have denied Kentucky’s plan anyway, that denial could not have been “based on” the agency’s use of the later modeling and lower threshold. So these decisions cannot allow the EPA to take this case to the D.C. Circuit.

The EPA’s responses again lack merit. It makes two textual arguments to support its reading of “determination.” The agency first points out that the statute uses “a” to modify “determination” (“based on a determination of nationwide scope or effect”). This indefinite article allegedly implies that an action can rest on multiple determinations and that only one need have a “nationwide scope or effect.” But the EPA places too much weight on this article. Because “determination” is a “countable noun[],” basic grammar rules required Congress to use the article (nobody would say “based on determination”). Niz-Chavez v. Garland, 593 U.S. 155, 162–63 (2021). And the use of “a” says nothing about whether “determination” should reach each of the EPA’s nonfinal interpretive choices or just its final decision. If anything, the Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 23

decision to use a singular noun could just as well convey that the final action must rest on a “single” decision—not on many separate decisions. Id. at 163.

The EPA also suggests that our reading of “determination” suffers from the same problem as its reading: we allegedly strike this clause out of the statute because the final “applying-the-law-to-the-facts” decision that underlies any local action will never have the required nationwide scope or effect. Yet caselaw existing before Congress amended the key language in § 7607(b)(1) offers examples of state-plan “determinations” that might have a “nationwide scope or effect.” Take Dayton Power & Light Co. v. EPA, 520 F.2d 703 (6th Cir. 1975). There, the EPA followed a “unitary rule-making procedure” to adopt regulations that had “the effect of amending every state’s air quality implementation plan in precisely the same way.” Id. at 705. Interpreting an earlier version of the statute, we held that the challenges to these uniform state-plan amendments belonged in the D.C. Circuit. Id. at 706–09. Thus, when the “automatic application of standard, nation-wide guidelines to all plans simultaneously preordains wholesale” actions, the EPA “determination” may well have a nationwide scope or effect. Nat. Res. Def. Council, Inc. v. EPA, 465 F.2d 492, 494 (1st Cir. 1972) (per curiam). But the EPA makes no such claim here.

Moving away from the text, the EPA argues that the legislative history supports its view. When discussing this judicial-review provision, a House Report agreed “with the comments” in a statement accompanying recommended amendments that had been proposed by the Administrative Conference of the United States. H.R. Rep. 95-294, at 324 (1977) (discussing Admin. Conf. of the U.S., Recommendations on Judicial Review Under the Clean Air Act, 41 Fed. Reg. 56767, 56769 (Dec. 30, 1976) (Statement of G. William Frick)). These comments suggested that Congress should send “‘national’ [state-plan] issues” to the D.C. Circuit. 41 Fed. Reg. at 56769. Yet this meaningless statement says nothing about the key question: Which state- plan issues are “national”? In fact, this legislative history may well cut the other way. The statement relied on by the EPA cites decisions like Dayton Power & Light as examples of the uniform regulations that belong in the D.C. Circuit. See 41 Fed. Reg. at 56769 & n.3. And again, the EPA’s denial of Kentucky’s plan here looks nothing like the regulations in that case. In all events, we cannot use (at best) ambiguous legislative history to muddy the unambiguous Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 24

meaning of statutory phrases like “nationwide.” See Food Mktg. Inst. v. Argus Leader Media, 588 U.S. 427, 437 (2019). Nothing about the disapproval of Kentucky’s plan has a national sweep.

The EPA also claims that we should use a “deferential standard” to review whether an agency action rests on “a determination of nationwide scope or effect” under § 7607(b)(1). Respondents’ Br. 46. But the Supreme Court has recently told courts that we should no longer defer to an agency’s answers to legal questions. See Loper Bright, 144 S. Ct. at 2273. And the EPA resolves a pure question of law when it interprets the key terms in § 7607(b)(1) (such as “nationally applicable” or “determination”). After Loper Bright, we must review (and correct) the agency’s mistaken interpretation of those terms without giving it deference.

The EPA lastly makes a big-picture point. It says that any decision about the Good Neighbor Provision will always involve nationwide issues because this part of the Clean Air Act concerns pollution that crosses state lines. This argument all but confirms its error: it reads words like “nationally applicable” and “nationwide” to mean “regionally applicable” or “regional.” The Good Neighbor Provision regulates pollution as long as it is emitted from one “State” and travels to “any other State[.]” 42 U.S.C. § 7410(a)(2)(D)(i). It thus would cover pollution emitted from St. Louis that landed just across the Mississippi River in East St. Louis. Yet nobody would say that the EPA made a “nationwide” determination when it decided whether Missouri adequately accounted for this Illinois pollution problem. In the end, the EPA’s fact- specific denial of Kentucky’s plan belongs in our court, not the D.C. Circuit.

III. Did the EPA Lawfully Disapprove Kentucky’s Plan?

Because Kentucky sued in the right tribunal, we proceed to the merits. Our court has evaluated the EPA’s disapprovals of state plans using the APA’s general standards. See Navistar Int’l Transp. Corp. v. EPA, 941 F.2d 1339, 1341 (6th Cir. 1991); Dressman v. Costle, 759 F.2d 548, 555 (6th Cir. 1985); cf. 42 U.S.C. § 7607(d)(1), (9). Under those standards, a court must “hold unlawful and set aside agency action” that the court finds “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A). Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 25

We need not proceed past § 706(2)(A)’s ban on “arbitrary” and “capricious” actions to resolve this case. That ban requires agencies to make reasonable decisions in a reasonable way. See FCC v. Prometheus Radio Project, 592 U.S. 414, 423 (2021); Dep’t of Homeland Sec. v. Regents of the Univ. of Cal., 591 U.S. 1, 16 (2020). Under this deferential “reasonableness” test, a court cannot set aside an action simply because it would have made a different choice. See Prometheus Radio Project, 592 U.S. at 423. But the court must ensure that the agency considered each “important aspect of the problem” and issued a decision rooted in the law and facts. Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983).

When an agency departs from its prior position, the change creates unique “aspect[s] of the problem” that it must confront. Id.; Encino Motorcars, LLC v. Navarro, 579 U.S. 211, 221– 22 (2016). To start, the agency must “display awareness” that it has, in fact, jettisoned its past views. FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009). This requirement does not compel the agency to meet a “heightened” reasonableness test as compared to when it adopts a policy from scratch. Id. at 514. Still, the agency must acknowledge the change and explain why it sees things differently. See Encino Motorcars, 579 U.S. at 221. So while the agency can make a “U-turn,” it cannot make an unexplained one. Casey v. Berryhill, 853 F.3d 322, 329 (7th Cir. 2017). Courts thus have found agency action arbitrary and capricious when it rested on an “[u]nexplained inconsistency” with past practice. Encino Motorcars, 579 U.S. at 222 (citation omitted).

Next, the agency must address how a change will affect those who have relied on its prior position. See Fox, 556 U.S. at 515; Smiley v. Citibank (S.D.), N.A., 517 U.S. 735, 742 (1996). The agency should identify these reliance interests and weigh them against the “policy” reasons supporting the change. Regents of the Univ. of Cal., 591 U.S. at 33. The Supreme Court has held, for example, that an agency arbitrarily adopted a more employee-friendly view of the wage-and-hour laws when the agency ignored that employers had structured their payment arrangements based on the agency’s past interpretation. See Encino Motorcars, 579 U.S. at 222– 24. And this mandate to address reliance interests applies just as much to an agency’s departure Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 26

from informal guidance as it does to its departure from formal regulations. See Perez v. Mortg. Bankers Ass’n, 575 U.S. 92, 105–06 (2015).

The EPA failed to live up to these standards here. The agency’s denial of Kentucky’s plan departed from prior guidance in three respects. For one, the EPA’s March 2018 Memorandum told the States (including Kentucky) that they “may consider using” the 2011 modeling to develop their plans for the first two steps of the EPA’s four-step approach to the Good Neighbor Provision. Mar. Mem., J.A. 80. For another, the EPA’s August 2018 Memorandum told the States that “it may be reasonable and appropriate for [them] to use a 1 ppb contribution threshold” when identifying the links to downwind receptors that require more scrutiny at the agency’s later steps. Aug. Mem., J.A. 101. The EPA issued this memo to “provide[] recommendations” to States but added that this presumption of a 1 ppb threshold may not fit “the facts” of a particular State. Id. at 98. For a third, the EPA saw nothing about Kentucky’s unique “facts” that warranted a departure from these recommendations. Id. To the contrary, the EPA told Kentucky during the Commonwealth’s notice-and-comment process that it could “rely entirely” on the 2011 modeling and 1 ppb threshold as a “more straightforward approach” at Step 2. EPA Comments, J.A. 126.

The EPA threw out all this guidance when it disapproved Kentucky’s plan. The EPA relied on the 2016v3 modeling—not the 2011 modeling that it had recommended—to identify Kentucky’s links to downwind receptors. 88 Fed. Reg. at 9356. Even worse, this new modeling showed that Kentucky’s highest contribution to a maintenance or nonattainment receptor was only “0.84 ppb,” which is below the 1 ppb threshold that the EPA had recommended. Id. Yet the EPA denied Kentucky’s plan by using a lower threshold: 1% of the ozone standard (.7 ppb). See id.; see also id. at 9342. The EPA acted arbitrarily because it did not adequately explain either change.

Threshold Change. The EPA did not even “display awareness” that its switch from a 1 ppb threshold to a .7 ppb threshold changed anything. Fox, 556 U.S. at 515. The EPA’s final rule reasoned that its August 2018 Memorandum had cautioned that the recommended 1 ppb threshold might not fit the “facts and circumstances” of all States. 88 Fed. Reg. at 9372. It then invoked this disclaimer to find that no State “made a sufficient showing” to use this threshold. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 27

Id. at 9373. Yet the EPA’s logic flipped the key presumption. The August 2018 Memorandum treated the 1 ppb threshold as presumptively acceptable unless a State’s unique facts made the threshold improper, but the final rule treated the 1 ppb threshold as presumptively unacceptable unless that threshold fit a State’s facts. The August 2018 Memorandum had presumptively allowed States to use the 1 ppb threshold because it was “generally comparable” to a .7 ppb threshold in capturing upwind contributors to downwind receptors. Aug. Mem., J.A. 101. When departing from this presumption, did the EPA disagree that the thresholds were generally comparable in this way? Did it change for another reason? The EPA did not even acknowledge the change, let alone explain it. See Encino Motorcars, 579 U.S. at 221–22.

The EPA instead accused States like Kentucky of “misunderstanding” the August 2018 Memorandum by accepting at face value what it said. 88 Fed. Reg. at 9373. “That’s a bit rich.” Mingo Logan Coal Co. v. EPA, 829 F.3d 710, 740 (D.C. Cir. 2016) (Kavanaugh, J., dissenting). It is especially rich in Kentucky’s case. The EPA specifically told Kentucky that it could “rely” on the 1 ppb threshold. EPA Comments, J.A. 126. So by “disagree[ing] with Kentucky’s arguments for” that threshold, the agency was disagreeing with itself. 88 Fed. Reg. at 9356.

Modeling Change. When deciding to switch to the 2016v3 modeling, by comparison, the EPA ignored the “serious reliance interests” that arose from its earlier assurance that States could use the 2011 modeling. Fox, 556 U.S. at 515. To justify the newer modeling, the EPA reasoned that it should be able to use “the best information available to it at the time it” decides on a plan’s validity. 88 Fed. Reg. at 9366. This logic ignored an “important aspect of the problem”: Kentucky’s reliance on the March 2018 Memorandum. State Farm, 463 U.S. at 43. Indeed, the EPA’s briefing cites nothing in the rule that even acknowledged this reliance interest.

If the EPA had “considered” Kentucky’s reliance, it might have chosen an alternative to its flat disapproval. Regents of the Univ. of Cal., 591 U.S. at 32. As Kentucky suggested, the EPA could have announced in advance that it would use newer data and given States the option to update their plans. Ky. Comments, J.A. 513–14. In response to comments like this one, the EPA argued that the Clean Air Act did not require it to give States “specific metrics” to help them comply with the Good Neighbor Provision. 88 Fed. Reg. at 9363 (quoting EME Homer, 572 U.S. at 510). Yet this generic response ignores the key question: May the EPA affirmatively Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 28

give the States one set of “metrics” to draft their plans and then use another set of metrics to grade them? The EPA’s response thus confirms its utter disregard of Kentucky’s reliance interest.

The EPA’s defenses of these changes lack merit. First, the EPA rests on the disclaimer in the August 2018 Memorandum that its recommended 1 ppb threshold might not fit a State’s facts. But this disclaimer does the EPA no good as applied to Kentucky. Again, the memorandum proposed the 1 ppb threshold because it captured a “generally comparable” amount of upwind contributions to downwind receptors as would a .7 ppb threshold. Aug. Mem., J.A. 101. So if the 1 ppb threshold was not “comparable” to a .7 ppb threshold for a particular State and receptor, this approach might give way. But the EPA has made no receptor- specific argument for Kentucky. Indeed, the memorandum itself listed all receptors to which Kentucky contributed over .7 ppb when finding the two thresholds comparable. Compare Aug. Mem., J.A. 102–03, with State Plan, J.A. 152. The EPA instead argues that Kentucky did not provide an adequate “technical justification” for the higher threshold. Respondents’ Br. 70. Yet the EPA does not explain what “technical” justification Kentucky needed beyond the one in the EPA’s own prior memorandum.

Second, the EPA suggests that its comments about Kentucky’s draft plan did not promise that Kentucky could rely on the 2011 modeling or 1 ppb threshold “without conducting further analysis and offering technical justification.” Respondents’ Br. 70. That is false. When discussing the modeling data in Kentucky’s plan, the EPA stated: “An alternative, more straightforward approach would be to rely entirely upon the EPA’s [2011 modeling] and apply the 1 ppb screening threshold” from the August 2018 Memorandum. EPA Comments, J.A. 126 (emphasis added). The EPA did not then warn that it might still reject the state plan despite this recommendation.

Third, the EPA says that Kentucky could not have relied on the two memoranda because the EPA did not have a “longstanding policy” of using only modeling data in existence when a State submitted a plan or a 1 ppb threshold. Respondents’ Br. 72, 88 (quoting Breeze Smoke, LLC v. FDA, 18 F.4th 499, 507 (6th Cir. 2021) (order)). This argument misunderstands the law. While “longstanding policies” suffice to create reliance interests, Encino Motorcars, 579 U.S. at Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 29

222, they are not necessary to do so. Such interests can arise in other ways—such as when an agency tells a party that it can do something during the rulemaking process. See EPA Comments, J.A. 126. A contrary holding would allow agencies to “pull a surprise switcheroo” by issuing a final action that flatly contradicts prior guidance, as long as the action does not depart from some traditional practice. Env’t Integrity Project v. EPA, 425 F.3d 992, 996 (D.C. Cir. 2005).

The EPA’s support for this alleged “longstanding policy” requirement—our decision in Breeze Smoke—is not to the contrary. There, we recognized that an agency’s earlier guidance was not “longstanding,” so a party could not rely on caselaw refusing to defer to agency actions that inexplicably departed from traditional practice. Breeze Smoke, 18 F.4th at 507. But we did not uphold the agency action in Breeze Smoke on this basis alone. Rather, we also rejected the regulated party’s other claim that the agency had told the party one thing during the regulatory process and then done something different when taking the final action. See id. at 506–07. We reasoned that the final action comported with the earlier guidance, so the party could not have relied on anything in that guidance to believe that the agency would act differently. See id. Here, by contrast, a clear conflict exists. The EPA told Kentucky that it could use the 2011 modeling and 1 ppb threshold and then denied Kentucky’s plan in large part because Kentucky had done what the EPA told it to do. Because the EPA did not adequately consider Kentucky’s “reliance interests” when changing course in these ways, it acted arbitrarily. Fox, 556 U.S. at 515.

Fourth, the EPA argues that Kentucky did not prove its reliance because it did not incur “substantial” monetary “costs” using the 2011 modeling. Respondents’ Br. 93. But Kentucky wasted the costs it incurred drafting its plan because the EPA evaluated that plan using a different yardstick from the one the agency promised. Regardless, as a sovereign entity, Kentucky possesses the “primary responsibility” to regulate pollution emitted from its borders and the discretion to decide how best to meet air-quality standards. 42 U.S.C. § 7401(a)(3); Ohio, 603 U.S. at 284. Kentucky placed this important regulatory interest on the line when it chose to rely on the EPA’s 2011 modeling. And it lost that interest when the EPA used newer data to disapprove its plan and to impose a federal plan that took the Commonwealth’s Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 30

regulatory discretion. See 42 U.S.C. § 7410(c). Given the Clean Air Act’s “cooperative federalism” design, Sierra Club, 681 F.3d at 343 (quoting Ellis, 390 F.3d at 467), this type of sovereign reliance interest warrants protection under the APA, see Ohio, 603 U.S. at 291.

Even if the EPA acted arbitrarily and capriciously in the ways that we have explained, the agency claims that we should find its mistakes harmless. The EPA opines that it would have denied Kentucky’s plan even if it had judged the plan using Kentucky’s 2011 modeling and 1 ppb threshold. Under those datapoints, Kentucky still contributed to a Maryland monitor at Step 2. State Plan, J.A. 152. And the EPA now argues that Kentucky gave inadequate reasons at Step 3 as to why it need not cut omissions to keep this monitor in attainment. Respondents’ Br. 57–68. The problem? Under basic principles of administrative law, agencies may invoke in court only the grounds that they relied on at the time they took the challenged action. See Regents of the Univ. of Cal., 591 U.S. at 20. If those relied-upon reasons cannot sustain an agency action, neither a court nor an agency may salvage the action by invoking “better” reasons for it in litigation. See id. at 22–23; see also Calcutt v. FDIC, 598 U.S. 623, 628–29 (2023) (per curiam).

This principle renders the EPA’s harmless-error arguments premature. The EPA did not rely on its current Step 3 claims to deny Kentucky’s state plan. To be sure, its rule did state that Kentucky “did not conduct an adequate Step 3 analysis.” 88 Fed. Reg. at 9356. But the EPA’s four-part approach raises distinct “steps” for a reason. The validity of a State’s decision on what emissions to cut at Step 3 depends on the State’s links at Step 2. And the EPA’s rule used the 2016v3 modeling at Step 2 to link Kentucky to Connecticut receptors. See id.; 2016v3 Modeling, J.A. 577. At Step 3, then, the rule analyzed these receptors. The agency noted, for example, that Kentucky was still linked to them even after the agency considered Kentucky’s planned power-plant shutdowns. See 88 Fed. Reg. at 9356. The rule, by contrast, did not suggest that Kentucky’s Step 3 analysis would fall short for the Maryland receptor linked to Kentucky under the 2011 modeling. Because the EPA abandoned this modeling, it did not even ask that question. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 31

Nor do we find it a forgone conclusion that the EPA would deny Kentucky’s plan on remand. See Calcutt, 598 U.S. at 629–30. Suppose the EPA kept its 2016v3 modeling but retained the higher 1 ppb threshold. In that scenario, Kentucky would not be linked to any maintenance or nonattainment receptors because its “highest-level contribution” to any such receptor was .84 ppb under that newer modeling. 88 Fed. Reg. at 9356. Or suppose the EPA relied on the older modeling showing only a single Kentucky link to the Maryland receptor. It might not make much sense to find Kentucky’s Step 3 analysis inadequate for this receptor. Kentucky’s plan predicted that its then-existing emissions reductions would allow the Maryland receptor to maintain the new ozone standard. State Plan, J.A. 177–78. And the EPA’s newer data may well have proved Kentucky right because Kentucky is no longer linked to this receptor. 2016v3 Modeling, J.A. 577.

In short, we have no idea how the EPA would have acted if it had not committed the legal errors that we have identified. The agency thus has failed to show that the errors did not prejudice Kentucky. Yet nothing we have said in this opinion would prohibit the EPA from properly raising any new concerns in additional administrative proceedings on remand.

IV. What Is the Proper Remedy?

Despite the errors underlying its action, the EPA lastly asks us to remand to the agency without vacating its disapproval of Kentucky’s plan. Substantial out-of-circuit caselaw underlies this request. Several courts have held that they may remand an illegal action to an agency without vacating the action—allowing the action to continue to have binding force in the meantime. See Black Warrior Riverkeeper, Inc. v. U.S. Army Corps of Eng’rs, 781 F.3d 1271, 1290 (11th Cir. 2015) (collecting cases); see also 33 Charles A. Wright et al., Federal Practice and Procedure § 8382, at 289–93 (2d ed. 2018). Yet the parties cite only one of our own decisions that has remanded to an agency without vacating its action. See Sierra Club v. EPA, 60 F.4th 1008, 1021–23 (6th Cir. 2023). There, we did not find the action unlawful and instead remanded for additional proceedings at the agency’s request. See id. at 1020–21. So the case says nothing about whether we may refuse to vacate illegal actions. The EPA asks us to break new ground. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 32

We need not do so to resolve this case. Even assuming that the APA permits this remand-without-vacatur remedy, it would not apply to the EPA’s errors. The courts that have permitted remand-without-vacatur relief consider two primary factors when deciding on the propriety of that relief. See id. at 1022. They first ask how serious of an error the agency made. See id. If the agency committed a technical error that it could easily fix on remand, courts are more likely to keep its action in place. See id. If, by contrast, the agency committed a “fundamental” error—such as taking a substantively illegal action or ignoring notice-and- comment requirements—these courts will not let the action stand. See Long Island Power Auth. v. FERC, 27 F.4th 705, 717 (D.C. Cir. 2022); Allina Health Servs. v. Sebelius, 746 F.3d 1102, 1110–11 (D.C. Cir. 2014).

The courts that permit a remand-without-vacatur remedy next ask how disruptive their vacatur would be. See Sierra Club, 60 F.4th at 1022. If vacatur would, say, upend years of transactions entered in reliance on the agency’s action, courts are more likely to keep it in place. See Am. Great Lakes Ports Ass’n v. Schultz, 962 F.3d 510, 518–19 (D.C. Cir. 2020). If, however, an agency can point to just the normal uncertainty that follows every vacatur of agency action, courts are more likely to vacate the action. See Sierra Club, 60 F.4th at 1023; Nat. Res. Def. Council v. Wheeler, 955 F.3d 68, 85 (D.C. Cir. 2020).

Apart from these two factors, courts treat vacatur as the default and remand without vacatur as the “rare” remedy. United Steel, Paper & Forestry, Rubber, Mfg., Energy, Allied Indus. & Serv. Workers Int’l Union v. Mine Safety & Health Admin., 925 F.3d 1279, 1287 (D.C. Cir. 2019); see Am. Great Lakes Ports Ass’n, 962 F.3d at 518. And the two factors merely structure the inquiry because the proper remedy will depend on all the equities. See Sierra Club, 60 F.4th at 1022.

The EPA has failed to establish that the equities justify remand without vacatur here. For starters, while the EPA may try to rely on distinct grounds to fix its errors on remand, its current disapproval contains a “fundamental” defect. Allina Health Servs., 746 F.3d at 1110 (citation omitted). The agency’s bait-and-switch tactics left Kentucky with the type of “deficient notice” that has justified vacatur in other cases. Id. If anything, this case warrants that relief even more because the EPA undercut the Clean Air Act’s “cooperative federalism” structure. Sierra Club, Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 33

681 F.3d at 343 (quoting Ellis, 390 F.3d at 467). If we did not vacate its disapproval, the agency would have every “incentive” to take similar shortcuts in the future. Env’t Def. Fund v. FERC, 2 F.4th 953, 976 (D.C. Cir. 2021).

The EPA has also not shown that vacatur would have any unusually disruptive effects. See Sierra Club, 60 F.4th at 1023. It asserts that vacatur will bar it from imposing its federal implementation plan on Kentucky and thus will delay that plan’s emissions reductions. Yet the Supreme Court has already indefinitely stayed the federal plan. See Ohio, 603 U.S. at 300. So the agency could not enforce its emissions reductions even if we refused to vacate the disapproval of Kentucky’s plan. And the EPA’s long delay in acting on this state plan undercuts any claim that an urgent need exists to compel Kentucky to reduce its emissions. See Texas, 2023 WL 7204840, at *11. Besides, the EPA’s arguments simply assume the conclusion on the merits: that Kentucky’s current plan falls short of meeting its good-neighbor obligations. But “it is far from certain” that the EPA will reach that same conclusion on remand. Env’t Def. Fund, 2 F.4th at 976.

All told, we deny the EPA’s motion to transfer. We also grant Kentucky’s petitions for review, vacate the EPA’s disapproval of Kentucky’s state implementation plan, and remand to the agency for further proceedings consistent with this opinion. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 34

_________________

CONCURRENCE _________________

MURPHY, Circuit Judge, concurring. Our holding that the Environmental Protection Agency (EPA) acted arbitrarily and capriciously and our assumption that the remand-without- vacatur remedy exists allow us to fully resolve this case. I write this separate concurrence to highlight two issues that we need not conclusively decide. First, the EPA’s failure to timely act on Kentucky’s state implementation plan may have provided an independent ground to vacate the EPA’s disapproval under the Administrative Procedure Act (APA). Second, the validity of the remand-without-vacatur remedy may depend on the nature of this “vacatur” relief.

I. Did the EPA’s Untimely Disapproval of Kentucky’s Plan Violate the APA?

Apart from barring arbitrary and capricious agency action, the APA also requires agencies to act “in accordance with law[.]” 5 U.S.C. § 706(2)(A). Kentucky and the EPA have debated whether the EPA violated this mandate by disapproving Kentucky’s plan in an untimely manner using data generated after its deadline to act. I think it may well have done so.

Start with some statutory background. The Clean Air Act imposes two deadlines on the EPA after a State submits a plan. The EPA must first decide whether the plan contains the “information necessary to enable [the agency] to determine whether the plan submission complies with the” Act. 42 U.S.C. § 7410(k)(1)(A). The EPA has no duty to “act on” a plan until it contains this information. Id. Yet the EPA should make this “[c]ompleteness” finding within 60 days. Id. § 7410(k)(1)(B). And if the EPA has failed to make the finding “6 months after” the State submitted the plan, the plan “shall on that date be deemed by operation of law to meet [the] minimum criteria” necessary for the EPA’s review. Id. The second deadline then identifies when the EPA must approve or disapprove the plan: “[w]ithin 12 months” of the EPA’s completeness finding (whether made expressly or “by operation of law”). Id. § 7410(k)(2)–(3).

The parties agree on how this law applied here. Kentucky submitted its plan in January 2019. But the EPA never expressly found that this plan contained all the required information. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 35

By operation of law, therefore, that finding occurred in July 2019. See id. § 7410(k)(1)(B). This date, in turn, required the EPA to act on Kentucky’s plan by July 2020. See id. § 7410(k)(2)–(3). But the EPA missed this deadline. It disapproved the plan more than two years later in February 2023. What consequences should follow from the missed deadline?

Background principles help answer this question. Many statutes command public officials to act within a certain time. See United States v. James Daniel Good Real Prop., 510 U.S. 43, 63 (1993). The Supreme Court has repeatedly held that officials who miss these statutory deadlines do not automatically forfeit the power to act belatedly if the statute itself does not impose that penalty. See id.; Nielsen v. Preap, 586 U.S. 392, 411 (2019) (plurality opinion); Barnhart v. Peabody Coal Co., 537 U.S. 149, 158–63 (2003); Brock v. Pierce County, 476 U.S. 253, 259–62 (1986); see also McIntosh v. United States, 601 U.S. 330, 337–42 (2024). When a statute does not identify the consequence for a violation, the Court has explained, courts may not simply pick the remedy they think best as a common-law matter. See James Daniel Good, 510 U.S. at 63; Gen. Med., P.C. v. Azar, 963 F.3d 516, 526 (6th Cir. 2020) (Murphy, J., concurring).

These principles provide the first step in the analysis: The Clean Air Act says nothing about what should happen if the EPA misses its “[d]eadline for action” on a state plan. 42 U.S.C. § 7410(k)(2). And we cannot create our “own coercive sanction” (say, an automatic approval of the state plan) to remedy the EPA’s untimely action on the plan. Barnhart, 537 U.S. at 159 (quoting James Daniel Good, 510 U.S. at 63). Indeed, the Act’s silence stands out compared to the remedy that applies when the EPA misses the initial deadline to decide whether the plan contains all required information. In that event, the Act “deem[s]” the plan complete “by operation of law[.]” 42 U.S.C. § 7410(k)(1)(B). And Congress presumably acted intentionally with its inclusion of this express remedy for the first deadline and its omission of the remedy for the second one. See State Farm Fire & Cas. Co. v. United States ex rel. Rigsby, 580 U.S. 26, 34 (2016).

As other background principles show, however, this conclusion does not allow the EPA to violate the Clean Air Act’s timelines with impunity. Rather, additional sources of law can allow courts to enforce a statute’s requirements even when the statute itself lacks its own Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 36

remedy. As one example from the Supreme Court’s caselaw, Article III gives courts the “inherent power” to sanction parties for violating court rules—even if no statute gives them this power. Id. at 37 (citing Chambers v. NASCO, Inc., 501 U.S. 32, 43–46 (1991)). So when a statute required plaintiffs to file complaints under seal, this background power allowed courts to punish plaintiffs who violated the sealing requirement despite the lack of a statutory remedy. See id. If courts rely on these other remedies, however, they must also respect the limitations that come with the remedies. In criminal cases, for instance, district courts must overlook all errors that do not harm the defendant. See Fed. R. Crim. P. 52(a); McIntosh, 601 U.S. at 338. This harmless-error rule covers the failure of courts or prosecutors to meet various statutory deadlines in criminal cases. See Dolan v. United States, 560 U.S. 605, 617 (2010); United States v. Montalvo-Murillo, 495 U.S. 711, 722 (1990).

These principles provide the second step in the analysis: Although the Clean Air Act does not explain what should happen if the EPA acts on a state plan after its deadline, the APA applies to this untimely action. And this second legal source does allow us to “set aside” (and issue an “injunction” against) an EPA “action” if it is “not in accordance with” the Clean Air Act’s timelines. 5 U.S.C. §§ 703, 706(2)(A). The EPA also seemingly violated the Clean Air Act’s clear command that it “shall act on” Kentucky’s plan within 12 months of its completeness finding. 42 U.S.C. § 7410(k)(2). The word “shall” suggests that the EPA lacks discretion to miss this deadline. See State Farm, 580 U.S. at 33–34. In short, courts need not rely on their “own coercive sanction” to remedy violations of the Clean Air Act’s timelines. James Daniel Good, 510 U.S. at 63. They need only rely on the APA’s sanction for this violation.

That said, like the harmless-error rule in criminal cases, the APA requires us to give “due account” to “the rule of prejudicial error.” 5 U.S.C. § 706. If the EPA’s violation of the deadline when ruling on Kentucky’s plan did not harm Kentucky, the APA would give us no basis to overturn the untimely disapproval. See Shinseki v. Sanders, 556 U.S. 396, 406–07 (2009). For several reasons, however, the illegal delay may well have “had a ‘substantial influence’ on the outcome of the proceeding” in this case. Montalvo-Murillo, 495 U.S. at 722 (quoting Bank of Nova Scotia v. United States, 487 U.S. 250, 256 (1988)); see Shinseki, 556 U.S. at 407–08. Those reasons all follow from the EPA’s decision to base its untimely disapproval on data in the Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 37

2016v3 modeling that postdated the EPA’s deadline to act. See Air Plan Disapprovals, 88 Fed. Reg. 9336, 9345 (Feb. 13, 2023).

For starters, the Clean Air Act notes that the EPA “shall approve” a state plan within the required timeframe if the plan “meets all of the applicable requirements” of the Act. 42 U.S.C. § 7410(k)(3) (emphasis added). So if the information on the date of the EPA’s deadline shows that the plan complies with the Act, the EPA must approve the plan. See id. The Act does not give the EPA residual discretion. See Ohio v. EPA, 603 U.S. 279, 284 (2024). Given this statutory scheme, the EPA’s use of after-the-fact data perhaps shows the required harm to Kentucky.

A related provision confirms this point. The EPA admits that Kentucky could have sued back in July 2020 to obtain a court order requiring the EPA to timely act while Kentucky’s plan remained pending with the EPA. See 42 U.S.C. § 7604(a)(2); 88 Fed. Reg. at 9365; Brock, 476 U.S. at 260 n.7. Yet a court-imposed deadline to rule on the plan in, say, 2021 would likewise have barred the EPA from relying on “datasets” that did not exist until 2022. 88 Fed. Reg. at 9345. As a result, even the EPA’s preferred remedy shows that its delay harmed Kentucky because it allowed the EPA to rely on post-deadline data that Congress did not intend for it to consider. See Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983).

A structural point reinforces the same idea. The Clean Air Act does not leave the EPA powerless to consider new information. It provides a remedy when new facts show that an approved state plan “is substantially inadequate to attain or maintain” an air-quality standard. 42 U.S.C. § 7410(k)(5). In a paragraph entitled “[c]alls for plan revisions,” the Act allows the EPA to “require [a] State to revise” this plan. Id. But such a compelled amendment requires the EPA to jump through several procedural hoops. The EPA must notify the State of its concerns, and the State (not the EPA) gets to take the lead in deciding on the revisions that will address those concerns. See id. This cooperative process also shows that EPA’s unlawful delay likely prejudiced Kentucky. If the EPA had taken this statutorily contemplated path, Kentucky would have retained the authority to devise revisions that accounted for the EPA’s post-deadline modeling. See id. By delaying action on Kentucky’s plan and then disapproving it using this Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 38

modeling, the EPA kicked Kentucky out of this process by triggering the power to impose a federal plan. See Ohio, 603 U.S. at 284–86; Texas v. EPA, 2023 WL 7204840, at *9 (5th Cir. May 1, 2023) (per curiam).

As far as I can tell, nothing in the EPA’s briefing justifies its unlawful delay and prejudicial use of post-deadline data. The agency first cites caselaw holding that agencies might violate the APA’s ban on arbitrary-and-capricious conduct if they “ignore new and better data.” Dist. Hosp. Partners, L.P. v. Burwell, 786 F.3d 46, 56–57 (D.C. Cir. 2015). The EPA thus suggests that it would have acted arbitrarily if it overlooked the 2016v3 modeling. Yet this caselaw has rejected arbitrary-and-capricious challenges when the EPA had good reasons to use older data. See Sierra Club v. EPA, 356 F.3d 296, 308 (D.C. Cir. 2004); see also Bd. of Cnty. Comm’rs of Weld Cnty. v. EPA, 72 F.4th 284, 289–90 (D.C. Cir. 2023). The EPA perhaps had such reasons here: the Act required it to approve or disapprove the plan in a timely manner, not to wait on constantly updated modeling. See Sierra Club, 356 F.3d at 308. To be sure, I agree that this fact does not mean that the EPA should “ignore” this newer modeling. Dist. Hosp. Partners, 786 F.3d at 57. It just means that the EPA should follow the Clean Air Act’s rules for how to use it: by issuing a call for the revision of an approved plan—not by unlawfully postponing its decision on a proposed one.

The EPA next relies on a decision that rejected the claim that the agency must rely on pollution conditions as they existed on the deadline for States to submit their plans. See Wisconsin v. EPA, 938 F.3d 303, 322 (D.C. Cir. 2019) (per curiam). The EPA places undue emphasis on Wisconsin. There, the EPA proposed a federal plan in 2016 using data about expected pollution conditions in 2017. See id. at 312, 321. This data revealed that a downwind State (Delaware) would have no problematic receptors. Id. at 321. But Delaware argued that the Good Neighbor Provision required the EPA to consider pollution conditions as they existed in 2011. Id. The State relied on the fact that the Act required upwind States to submit their plans for meeting the relevant air-quality standard at that time. Id. at 321–22. The D.C. Circuit rejected this view. Because the Good Neighbor Provision uses the future tense (covering States that “will” contribute to downwind nonattainment), the court interpreted the provision as regulating pollution conditions in future years—not as of the plan-submission date. Id. at 322 Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 39

(quoting 42 U.S.C. § 7410(a)(2)(D)(i)). Nothing I say here conflicts with this logic. Both the 2011 modeling that Kentucky used and the 2016v3 modeling that the EPA used predicted pollution in a future year: 2023. See Mar. Mem., J.A. 77; 88 Fed. Reg. at 9345. Neither source modeled pollution existing when Kentucky submitted its plan in 2019. And since Wisconsin involved a federal plan, it did not discuss what should happen if the EPA violates its statutory deadline to rule on a state plan. Wisconsin thus did not suggest that the EPA could deny a state plan using new data generated after its deadline to act.

When the EPA turns to the Clean Air Act’s text, it concedes that the Act imposes a mandatory deadline for the agency to decide on a plan’s validity. But the EPA calls the statutory deadline “procedural” and suggests that it is “not ‘central to the regulatory scheme.’” Respondents’ Br. 80 (quoting Wisconsin, 938 F.3d at 322). The agency adds that it may freely disregard “procedural” rules in pursuit of achieving the “Act’s central object”: attaining the air- quality standard. Id. (quoting Wisconsin, 938 F.3d at 316). I see three problems with this ends- justify-the-means logic. As an initial matter, the Constitution does not give agencies any prerogative power to “dispense” with statutory requirements that they find “unimportant.” See Michael W. McConnell, The President Who Would Not Be King 115–19 (2020). Next, because “no legislation pursues its purposes at all costs,” CTS Corp. v. Waldburger, 573 U.S. 1, 12 (2014) (quoting Rodriguez v. United States, 480 U.S. 522, 525–26 (1987) (per curiam)), the “limitations” on a statute’s main goals are “often [its] price of passage[.]” Henson v. Santander Consumer USA Inc., 582 U.S. 79, 89 (2017). So courts must respect those limitations just as much as the primary provisions. See Kucana v. Holder, 558 U.S. 233, 252 (2010). Finally, Congress’s findings reveal that it viewed respect for state authority as a central goal of the Clean Air Act. See 42 U.S.C. § 7401(a)(3). And the EPA’s unlawful delay here did not just run afoul of a procedural deadline; it also undercut the Act’s primary “cooperative federalism” design. Sierra Club v. Korleski, 681 F.3d 342, 343 (6th Cir. 2012) (quoting Ellis v. Gallatin Steel Co., 390 F.3d 461, 467 (6th Cir. 2004)).

All told, the EPA’s untimely disapproval of Kentucky’s plan may have violated the APA for this independent reason. Given that it acted in an arbitrary and capricious way when disapproving that plan, though, we need not conclusively resolve the issue. Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 40

II. Does the APA Authorize a Remand-Without-Vacatur Remedy?

The EPA has asked us to remand its disapproval of Kentucky’s plan to the agency for reconsideration without vacating that disapproval. In my view, its arguments about the availability of this remedy have broader ramifications. The correct answer might turn on a debate about whether the APA permits vacatur of an agency action as a remedy distinct from an injunction enjoining the action’s enforcement. Compare Corner Post, Inc. v. Bd. of Governors of Fed. Rsrv. Sys., 144 S. Ct. 2440, 2460–70 (2024) (Kavanaugh, J., concurring), with United States v. Texas, 599 U.S. 670, 693–703 (2023) (Gorsuch, J., concurring in the judgment). The APA instructs courts that they “shall” “set aside” agency action found unlawful. 5 U.S.C. § 706(2). What does this text mean? On the one hand, many have read it to create a distinct vacatur remedy against agency actions (rather than an injunction remedy against their enforcement). See Corner Post, 144 S. Ct. at 2462–63 (Kavanaugh, J., concurring) (citing cases); Mila Sohoni, The Power to Vacate a Rule, 88 Geo. Wash. L. Rev. 1121, 1131–37, 1162– 63 (2020). Yet this reading cuts against the notion that courts may refuse to vacate actions found illegal. Section 706(2) uses the word “shall” and suggests that courts must vacate illegal actions if the section creates a distinct judicial remedy. See Milk Train, Inc. v. Veneman, 310 F.3d 747, 757 (D.C. Cir. 2002) (Sentelle, J., dissenting) (quoting Checkosky v. SEC, 23 F.3d 452, 491 (D.C. Cir. 1994) (Randolph, J., concurring)).

On the other hand, others have not read § 706(2)’s “set aside” text as creating a standalone remedy. See Texas, 599 U.S. at 695–97 (Gorsuch, J., concurring in the judgment); Arizona v. Biden, 40 F.4th 375, 396–97 (6th Cir. 2022) (Sutton, C.J., concurring); John Harrison, Section 706 of the Administrative Procedure Act Does Not Call for Universal Injunctions or Other Universal Remedies, 37 Yale J. Reg. Bull. 37, 41–46 (2020). They have instead read it as codifying an idea first made famous by Marbury v. Madison, 5 U.S. 137 (1803): that a court need not give effect to an unlawful rule and may “disregard[]” it when deciding on the parties’ legal rights. Id. at 177–78; see Harrison, supra, at 43. Those on this side of the debate have added that the APA allows courts to grant only traditional remedies like “declaratory judgments or writs of prohibitory or mandatory injunction[.]” 5 U.S.C. § 703; see Harrison, supra, at 37. This differing view might make the remand-without-vacatur remedy more defensible. If the Nos. 23-3216/3225 Commonwealth of Ky., et al. v. EPA Page 41

“vacatur” of an action amounts to nothing more than an injunction, the remedy could trigger the equitable discretion that courts possess when deciding whether to grant that relief. The relief is never “a matter of right” even for parties with valid claims. Hill v. McDonough, 547 U.S. 573, 584 (2006); see Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 24 (2008). And notably, courts holding that they may remand without vacatur have justified this remedy on the ground that vacatur qualifies as an “equitable remedy” subject to their discretion. Black Warrior Riverkeeper, Inc. v. U.S. Army Corps of Eng’rs, 781 F.3d 1271, 1290 (11th Cir. 2015).

At day’s end, our resolution of this case allows us to avoid these tricky issues. We need not decide whether we have the power to keep illegal action in place. Even if we did, the EPA has not justified that remedy on the facts here. So we also need not decide whether “vacatur” qualifies as a new remedy or as an injunction by another name. Nothing turns on that distinction here.