Respondents .

Petition for Review of a Final Rule of the Environmental Protection Agency 88 Fed. Reg. 9336-9384Page 2 Before Stewart and Richman, Circuit Judges , and Scholer, District Judge . *

The opinion issued on March 25, 2025 1 was WITHDRAWN on March 9, 2026. The following opinion is SUBSTITUTED :

The State of Texas and state entities, together referred to as the Texas State Petitioners, as well as industries collectively referred to as Texas Industry Petitioners, filed petitions for rehearing en banc after we issued our March 25, 2025 opinion. While those petitions were pending, the Environmental Protection Agency ( EPA ) issued, on January 30, 2026, a notice of a proposed rule and reconsideration of a final rule. 2

In our opinion that is now withdrawn, we found persuasive EPA ’s statement in its briefing that it did not “disapprove the state submissions based on the 2016-based modeling [submitted after Texas filed its SIP ]” but instead that the newer modeling “only confirmed EPA 's conclusion.'³ For reasons explained below, including EPA ’s recent notice, we withdraw our prior opinion. The petitions for review of EPA ’s disapproval of Texas’s SIP are granted, the disapproval is vacated, and the matter is remanded to EPA . Louisiana did not seek en banc rehearing or reconsideration by the panel, and no party sought rehearing or reconsideration of our disposition of Page 3Mississippi's challenge to the denial of its SIP . Accordingly, our prior dispositions of those petitions remain unchanged. The petition for review of EPA ’s disapproval of Louisiana’s SIP is therefore denied. The petitions for review of EPA ’s disapproval of Mississippi’s and Texas’s SIP s are granted, those disapprovals are vacated, and those matters are remanded to EPA .

This case concerns the obligations of states under a provision of the Clean Air Act ( CAA ) known as the Good Neighbor Provision.⁴ Congress has directed EPA to establish national ambient air quality standards ( NAAQS ) for pollutants at levels that “in [its] judgment, cause or contribute to air pollution which may reasonably be anticipated to endanger public health or welfare.'⁵ 'Once EPA settles on a NAAQS , the Act requires the Agency to designate 'nonattainment' areas, i.e. , locations where the concentration of a regulated pollutant exceeds the NAAQS .'⁶ The establishment or revision of NAAQS also 'shifts the burden to States to propose plans adequate for compliance with the NAAQS .'⁷ States must develop state implementation plans ( SIP s) to comply with any new NAAQS.⁸ Every SIP must comply with the Good Neighbor Provision. Since 1990, the CAA has required SIP s to “contain adequate provisions . . . prohibiting . . . any source or other type of emissions activity Page 4within 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 any . . . [ NAAQS ].'⁹ This provision is concerned with two separate problems: (1) states that “contribute significantly to [ NAAQS ] nonattainment” in another state, and (2) states that “interfere with maintenance [of the NAAQS ] by' another state.¹⁰

The Good Neighbor Provision was meant to address the reality that “[a]ir pollution is transient, heedless of state boundaries.'¹¹ 'Left unregulated, the emitting or upwind State reaps the benefits of the economic activity causing the pollution without bearing all the costs.'¹² 'Conversely, downwind States to which the pollution travels are unable to achieve clean air'-and more specifically, the prescribed NAAQS -'because of the influx of out-of-state pollution they lack authority to control.'¹³

“If EPA determines that a State has failed to submit an adequate SIP . . . the Act requires [ EPA ] to promulgate a Federal Implementation Plan, or FIP , within two years . . . ‘ unless the State corrects the deficiency,''¹⁴ 'and [ EPA ] approves the plan or plan revision, before thePage 5 [ EPA ] promulgates' a FIP . 15 A FIP corrects 'an inadequacy' in a SIP “and provides for attainment of the relevant” NAAQS . 16

This case involves EPA ’s 2015 revision of the ozone NAAQS . “A layer of ozone in the atmosphere shields the world from the sun's radiation.'¹⁷ However, '[f]orming when sunlight interacts with a wide range of precursor pollutants, ground-level ozone can trigger and exacerbate health problems and damage vegetation.'¹⁸ In October 2015, EPA decreased the allowable concentration of ozone in the ambient air from 75 parts per billion (ppb) to 70 ppb.¹⁹ That triggered the states' respective obligations to submit SIP s within three years to achieve compliance with the new ozone NAAQS . 20

Texas timely submitted its SIP . 21 Louisiana and Mississippi filed the Good-Neighbor portion of their SIP s approximately one year beyond the deadline.²² The CAA provides that EPA has up to eighteen months Page 6following each submission to act on a SIP . 23 After considerable delay, EPA proposed disapproval of Texas’s, Louisiana’s, and Mississippi’s SIP s about three years after they were submitted.²⁴ In a final rule, EPA disapproved the SIP s of twenty-one states.²⁵ The final rule incorporated by reference the reasons for disapproval given in the proposed disapprovals that EPA had previously issued.²⁶ EPA ’s disapprovals precipitated over twenty lawsuits in at least seven other circuits.²⁷ The petitioners in the case before us are three of those states—Louisiana, Mississippi, Texas—and energy-industry members.

A threshold issue is whether the petitions for review must be transferred to the D.C. Circuit. Our prior opinion addressed the venue question at length.²⁸ We concluded that venue is proper in this circuit.²⁹

8-Hour Ozone National Ambient Air Quality Standards (AL, MS, TN Proposed Disapproval), 87 Fed. Reg. 9545, 9554 (Feb. 22, 2022) (noting Mississippi Good-Neighbor SIP was submitted September 3, 2019).

Page 7Subsequently, the Supreme Court decided Oklahoma v. EPA , 30 which held “that EPA ’s disapprovals [of Oklahoma’s and Utah’s SIP s] are locally or regionally applicable actions reviewable in a regional Circuit.'³¹ We are satisfied that the Supreme Court's reasoning applies equally to the disapprovals of the Louisiana, Mississippi and Texas SIP s.

Another initial matter pertains to the fact that EPA ’s disapproval of Texas’s SIP was issued more than two years after Texas had submitted it. The Texas State Petitioners assert in a sentence of their initial brief that “[i]f EPA does not issue a completeness finding within six months, the SIP is deemed to meet the minimum criteria by operation of law [citing 42 U.S.C. § 7410(k)(1)(B)]. Upon its issuance of a completeness finding, EPA has 12 months to approve or disapprove the SIP [citing § 7410(k)(2)-(3)].'³² Texas does not assert that its SIP was approved by 'operation of law' when EPA failed to act before the 12 month-period had passed. Such an argument would likely fail.

As JUDGE MURPHY of the Sixth Circuit observed in his concurring opinion in Kentucky v. EPA , 33 the 'operation of law' language applies only to a determination, or lack thereof, regarding a “completeness finding” as to the 'minimum criteria' specified in subsection (k).³⁴ Congress did not specify that a SIP would be approved by “operation of law” if EPA failed to act within a subsequent 12-month period. JUDGE MURPHY's concurring Page 8opinion also notes that '[t]he 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.'³⁵

There are significant disagreements between the petitioners and EPA regarding the proper construction of the CAA . The Texas State Petitioners contended in their initial brief that ' EPA 's disapproval [] held Texas to requirements found nowhere in the Act” by treating “any ozone contribution by Texas to a downwind receptor at greater than one percent of the NAAQS as a presumptively ‘significant’ contribution.” 36 Texas maintains that '[t]he fundamental problem here is that the ‘one -percent threshold does not appear in the text of' the good-neighbor provision.'³⁷ Therefore, Texas argues, “ EPA lacked statutory authority to treat that threshold as a presumptive finding of 'significance' that States must 'rebut' to secure approval of a SIP .'³⁸ Texas acknowledges that ' EPA has repeatedly claimed that any State whose contribution meets or exceeds the one-percent threshold does not automatically violate its good-neighbor obligations; instead, such a State is ‘screen[ed] in . . . for further evaluation of emissions control Page 9opportunities.''³⁹ 'But,' Texas argues, 'in practice, EPA deploys the onepercent threshold as a presumption of significant contribution that a State must rebut on pain of SIP disapproval.'⁴⁰ Intertwined among these arguments was Texas's complaint that EPA used modeling and data published after Texas had submitted its SIP to disapprove that SIP .

In its petition for rehearing, Texas maintained that 'whereas EPA treats any contribution exceeding one percent of the NAAQS as a ‘ significant’ contribution that triggers a state’s obligations under the CAA , Texas treats such a linkage as merely a starting point for further detailed analysis.'⁴¹ Texas says, 'The panel did not determine whether Texas or EPA had the better interpretation of the statutory phrase ‘ contribute significantly.''⁴² Citing Loper Bright Enterprises v. Raimondo , 43 Texas argues that 'courts 'need not and under the APA may not defer to an agency interpretation of the law simply because a statute is ambiguous,’” and “[t]hat remains true ‘ even if an ambiguity happens to implicate a technical matter.''⁴⁴ The Texas State Petitioners again argued strenuously in their petition for rehearing en banc that EPA had impermissibly relied on data and Page 10modeling published after Texas filed its SIP that was markedly different from the guidance it had given to States before Texas submitted its SIP .

The Texas Industry Petitioners make similar arguments in their petition for rehearing en banc. They additionally assert that “ EPA premised its disapproval on the agency’s claim that it is ‘empowered to interpret ' the statutory term ‘ contribute significantly' in the Clean Air Act.'⁴⁵ They observe, “ Chevron is gone.'⁴⁶ The Texas Industry Petitioners assert that our panel’s decision, which we have now withdrawn, “squarely conflicts with the Supreme Court’s intervening decision in Loper Bright . The central question in this case is the meaning of the undefined statutory term 'contribute significantly' in the Good Neighbor Provision of the Clean Air Act, 42 U.S.C. § 7410(a)(2)(D).'⁴⁷

The Texas State Petitioners also made clear in their briefing that they contend the CAA gives the authority to the States, not EPA , to interpret the Good Neighbor Provision. Texas’s Reply Brief argues, “The good-neighbor provision delegates interpretive authority to the States in the first instance at the SIP stage.'⁴⁸

Although Texas and the other petitioners did to some extent assert arguments regarding their interpretation of the CAA , their primary complaint in their initial briefing to our panel was that EPA disapproved Texas’s SIP based on data and modeling published after Texas submitted its SIP . The Final Order under review in this appeal considered in some detail Page 11deficiencies EPA found in Texas's own modeling and Texas's analysis of its modeling results submitted in support of its SIP . 49 The gaps and deficiencies EPA found in Texas's modeling and methodology had been considered in greater detail in EPA 's earlier proposed disapproval of SIP submissions.⁵⁰ EPA ’s briefing in our court said, “Nor did EPA , as Petitioners erroneously contend, disapprove the state submissions based on the 2016-based modeling [submitted after Texas filed its SIP ], which only confirmed EPA ’s conclusion.'⁵¹ In our prior, now withdrawn decision, we concluded it was unnecessary to decide “the question of interpretive authority or the effect of the Supreme Court’s recent decision in Loper Bright Enterprises v. Raimondo regarding deference to agencies’ interpretations of the statutes they administer' because EPA determined that Texas's SIP submission was “flawed on [its] own terms,” and “[a]ccordingly, it is sufficient for us to hold that, at the very least, the CAA directs EPA to assess independently whether a SIP satisfies a state's chosen reasonable interpretation of the Good Neighbor Provision.'⁵² We concluded that though EPA 's Final Rule disapproving Texas's SIP referred to data and modeling EPA published after Texas submitted its SIP , ' EPA provided more than one reason for disapproving Texas's SIP : both EPA 's and Texas's modeling showed emissions that EPA concluded triggered Good-Neighbor obligations,' and “[t]he smattering of places where EPA mentioned its own data during its Page 12discussion of the data in the Texas SIP does not undermine' the conclusion that EPA independently assessed and rejected Texas's SIP on its own merits.⁵³

However, as we will explain in greater detail below, EPA ’s January 30, 2026 “notice of a proposed rule and reconsideration of final rule” states that EPA did utilize data published after Texas submitted its SIP in disapproving Texas’s and a number of other States’ SIP s.⁵⁴ We are now vacating the disapproval of Texas's SIP and remanding to EPA , in part because, in light of the January 30, 2026 notice, we can no longer conclude with confidence that EP A’s disapproval was based on its technical review of the data and reasoning Texas offered in support of what Texas called a “multifactor ‘weight -of-the-evidence' analysis.''⁵⁵

Accordingly, we now proceed to address the arguments the Texas State Petitioners and Texas Industry Petitioners have asserted regarding the proper construction of the Good Neighbor Provision in the CAA . We begin with Loper Bright .

The Supreme Court explained in Loper Bright that section 706 of the Administrative Procedure Act 56 ( APA ) 'delineates the basic contours of judicial review' of agency action.⁵⁷ The APA 'specifies that courts, not agencies, will decide ' all relevant questions of law' arising on review of agency action-even those involving ambiguous laws-and set aside any such action inconsistent with the law as they interpret it.'⁵⁸ By contrast, § 706 ' does mandate that judicial review of agency policymaking and factfinding be deferential. See § 706(2)(A) (agency action to be set aside if 'arbitrary, capricious, [or] an abuse of discretion'); § 706(2)(E) (agency factfinding in formal proceedings to be set a side if 'unsupported by substantial evidence').'⁵⁹ The Loper Bright decision is emphatic that the APA 'makes clear that agency interpretations of statutes-like agency interpretations of the Constitution-are not entitled to deference,' and “[u]nder the APA, it thus ‘remains the responsibility of the court to decide whether the law means what the agency says.''⁶⁰

While “courts must exercise independent judgment in determining the meaning of statutory provisions,' they 'may-as they have from the start—seek aid from the interpretations of those responsible for implementing particular statutes.'⁶¹ The Loper Bright decision explains thatPage 14 “[s]uch interpretations 'constitute a body of experience and informed judgment to which courts and litigants may properly resort for guidance' consistent with the APA .'⁶² “And interpretations issued contemporaneously with the statute at issue, and which have remained consistent over time, may be especially useful in determining the statute’s meaning.'⁶³

Loper Bright also recognizes that “[i]n a case involving an agency, of course, the statute’s meaning may well be that the agency is authorized to exercise a degree of discretion.'⁶⁴ One example the Court gave was when statutes “empower an agency to prescribe rules to ‘fill up the details’ of a statutory scheme.'⁶⁵ The Court cited Wayman v. Southard 66 as an illustration.⁶⁷ Another category of statutes authorizing an agency to exercise a degree of discretion the Court identified in Loper Bright was statutes that 'regulate subject to the limits imposed by a term or phrase that ‘ leaves Page 15agencies with flexibility,’ . . . such as ‘ appropriate’ or ‘ reasonable.''⁶⁸ The Court's supporting case citation 69 was Michigan v. EPA , 70 which observed that “Congress instructed EPA to add power plants to the program if (but only if) the Agency finds regulation ‘ appropriate and necessary.''⁷¹ The Court said in Michigan , “One does not need to open up a dictionary in order to realize the capaciousness of this phrase. In particular, ‘appropriate’ is ‘the classic broad and all-encompassing term that naturally and traditionally includes consideration of all the relevant factors.''⁷² The Court in Michigan further explained that “Although this term leaves agencies with flexibility, an agency may not ‘ entirely fai[l] to consider an important aspect of the problem' when deciding whether regulation is appropriate.'⁷³ The Court held in Michigan that 'Read naturally in the present context, the phrase ‘appropriate and necessary’ requires at least some attention to cost.” 74 The Michigan decision continued, 'One would not say that it is even rational, never mind ‘appropriate,’ to impose billions of dollars in economic costs in return for a few dollars in health or environmental benefits.'⁷⁵

Page 16The Loper Bright decision explained a court’s role when discretionary authority has been delegated: “When the best reading of a statute is that it delegates discretionary authority to an agency, the role of the reviewing court under the APA is, as always, to independently interpret the statute and effectuate the will of Congress subject to constitutional limits.'⁷⁶ 'The court fulfills that role by recognizing constitutional delegations, 'fix[ing] the boundaries of [the] delegated authority,''⁷⁷ and 'ensuring the agency has engaged in ‘reasoned decisionmaking’ within those boundaries .'⁷⁸

In Loper Bright , the Court jettisoned Chevron ’s directive that when a statute is ''silent or ambiguous with respect to the specific issue’ at hand,” a court must 'defer to the agency's interpretation if it ‘ is based on a permissible construction of the statute.''⁷⁹ When a statute is ambiguous “[i]n an agency case as in any other . . . even if some judges might (or might not) consider the statute ambiguous, there is a best reading all the same— 'the reading the court would have reached' if no agency were involved.'⁸⁰ “It therefore makes no sense to speak of a ‘permissible’ interpretation that is not the one the court, after applying all relevant interpretive tools, concludes is best. In the business of statutory interpretation, if it is not the Page 17best, it is not permissible.'⁸¹ 'The very point of the traditional tools of statutory construction-the tools courts use every day-is to resolve statutory ambiguities. That is no less true when the ambiguity is about the scope of an agency’s own power—perhaps the occasion on which abdication in favor of the agency is least appropriate.'⁸²

Loper Bright reasons that 'even when an ambiguity happens to implicate a technical matter, it does not follow that Congress has taken the power to authoritatively interpret the statute from the courts and given it to the agency. Congress expects courts to handle technical statutory questions.'⁸³ ' ' [M]any statutory cases’ call upon ‘ courts [to] interpret the mass of technical detail that is the ordinary diet of the law.'⁸⁴ 'In an agency case in particular, the court will go about its task with the agency’s ‘body of experience and informed judgment,' among other information, at its disposal.'⁸⁵ '[A]lthough an agency’s interpretation of a statute ‘cannot bind a court,’ it may be especially informative ‘to the extent it rests on factual premises within [the agency's] expertise.''⁸⁶ 'Such expertise has always been one of the factors which may give an Executive Branch interpretation particular ‘power to persuade, if lacking power to control.’” 87

Page 18'The view that interpretation of ambiguous statutory provisions amounts to policymaking suited for political actors rather than courts is especially mistaken, for it rests on a profound misconception of the judicial role.'⁸⁸ 'That is not to say that Congress cannot or does not confer discretionary authority on agencies. Congress may do so, subject to constitutional limits, and it often has.'⁸⁹ When 'discretionary policymaking [is] left to the political branches, judges need only fulfill their obligations under the APA to independently identify and respect such delegations of authority, police the outer statutory boundaries of those delegations, and ensure that agencies exercise their discretion consistent with the APA.'⁹⁰

We turn to application of Loper Bright to the CAA and its Good Neighbor Provision.

We begin with the text of the Good Neighbor Provision. It says:

1. (1) Each State shall . . . submit to the Administrator, within 3 years (or such shorter period as the Administrator may prescribe) after the promulgation of a national primary ambient air quality standard (or any revision thereof) under section 7409 of this title for any air pollutant, a plan which provides for implementation, maintenance, and enforcement of such primary standard . . . within such State. In addition, such State shall adopt and submit to the Administrator . . . within 3 years (or such shorter period as the Administrator may prescribe)

Page 19after the promulgation of a national ambient air quality secondary standard (or revision thereof), a plan which provides for implementation, maintenance, and enforcement of such secondary standard . . . within such State. . . .

1. (2) Each implementation plan submitted by a State . . . shall— . . .

(D) contain adequate provisions— (i) prohibiting, consistent with the provisions of this subchapter, any source or other type of emissions activity within the State from emitting any air pollutant in amounts which will— (I) contribute significantly to nonattainment in, or interfere with maintenance by, any other State with respect to any such national primary or secondary ambient air quality standard, or (II) interfere with measures required to be included in the applicable implementation plan for any other State under part C to prevent significant deterioration of air quality or to protect visibility, (ii) insuring compliance with the applicable requirements of sections 7426 and 7415 of this title (relating to interstate and international pollution abatement); . . . . 91

As discussed earlier, each state must submit a SIP that includes provisions that comply with § 7410(2)(D) three years after EPA promulgates or revises a NAAQS . The Good Neighbor requirement is forward-looking.

Page 20It requires that a SIP must prohibit emissions of an air pollutant that 'will' contribute or interfere during the life of the SIP , as specified in § 7410(2)(D)(i)(II) or (II). Before a State can propose SIP provisions, EPA must establish the NAAQS for the particular pollutant at issue. Other key terms in the Good Neighbor Provision are 'nonattainment' in and “maintenance” by any State with respect to such NAAQS .

We have no difficulty in concluding that in 42 U.S.C. § 7409, Congress delegated to EPA the responsibility for promulgating NAAQS . No party to this appeal questions that. The statute directs EPA to establish primary and secondary NAAQS . Primary NAAQS 'shall be ambient air quality standards the attainment and maintenance of which in the judgment of the Administrator, based on such criteria and allowing an adequate margin of safety, are requisite to protect the public health.'⁹² Secondary NAAQS “shall specify a level of air quality the attainment and maintenance of which in the judgment of the Administrator, based on such criteria, is requisite to protect the public welfare from any known or anticipated adverse effects associated with the presence of such air pollutant in the ambient air.'⁹³

The CAA requires EPA to review each NAAQS (primary and secondary, for each pollutant) at five-year intervals, but the CAA also gives EPA the discretion to 'review and revise criteria or promulgate new standards earlier or more frequently.'⁹⁴ EPA first established the NAAQSPage 21 for ozone in 1971 as 0.08 parts per million (ppm) averaged over one hour.⁹⁵ Over the years, EPA performed its statutory duties, reevaluating and at times revising the ozone NAAQS . 96 In 2015, EPA revised the then-existing ozone NAAQS of 0.075 ppm to 0.070 ppm.⁹⁷ The standard was to be measured as the annual fourth-highest daily maximum 8 hour average concentration, averaged across three consecutive years.⁹⁸ EPA promulgated the 0.70 ppm standard in a final rule following a decision of the D.C. Circuit 99 that had remanded, in part, for further consideration.¹⁰⁰

The D.C. Circuit’s decision in Mississippi related some of the history of ozone NAAQS . That court recounted that EPA had at one point in time set the secondary NAAQS as 0.08 ppm measured over an 8-hour period, that “several parties challenged these revisions,” and “the Supreme Court reversed this court’s conclusion that the Clean Air Act unconstitutionally delegated Congress's legislative authority.'¹⁰¹ JUSTICE SCALIA, writing for the Court in American Trucking Ass’ns , held that “the text of § 109(b)(1) of the CAA [42 U.S.C. § 7409(b)(1)] at a minimum requires that '[f]or a discrete set of pollutants and based on published air quality criteria that Page 22reflect the latest scientific knowledge, [the] EPA must establish uniform national standards at a level that is requisite to protect public health from the adverse effects of the pollutant in the ambient air,’” and that “[r]equisite, in turn, ‘mean[s] sufficient, but not more than necessary.’” 102

In Mississippi , the D.C. Circuit also explained that the CAA requires primary NAAQS to be “set at levels ‘the attainment and maintenance of which in the judgment of the [ EPA ], . . . allowing an adequate margin of safety, are requisite to protect the public health.''¹⁰³ The court further explained that “secondary NAAQS ‘shall specify a level of air quality the attainment and maintenance of which in the judgment of the Administrator . . . is requisite to protect the public welfare from any known or anticipated adverse effects.''¹⁰⁴ As the D.C. Circuit noted, there are different considerations in promulgating secondary NAAQS : 'The Act provides that the public welfare protected by secondary NAAQS includes 'effects on soils, water, crops, vegetation, manmade materials, animals, wildlife, weather, visibility, and climate, damage to and deterioration of property, and hazards to transportation, as well as effects on economic values and on personal comfort and well-being.''¹⁰⁵

The terms “nonattainment” and “maintenance” as used in the Good Neighbor Provision must be considered in light of other provisions of the Page 23CAA , including 42 U.S.C. §§ 7407, 7409, 7511 and 7511a. For each area designated nonattainment for ozone pursuant to § 7407(d), Congress set forth, in § 7511, precise design values to be used in classifying the degree of nonattainment.

Once EPA promulgates or revises a NAAQS for ozone, 'states and EPA work within the Clean Air Act's structure of cooperative federalism to implement the new standards.'¹⁰⁶ 'Governors must ‘ submit to the Administrator a list of all areas (or portions thereof) in the State, designating [each area] as . . . nonattainment, . . . attainment, . . . or unclassifiable.''¹⁰⁷ A process then ensues, as our court explained in Texas 2020 , and ultimately, EPA designates whether an area is in attainment or nonattainment.¹⁰⁸ A state may challenge whether EPA correctly construed the CAA in designating an area as nonattainment, as Texas did in Texas 2020 . 109 'Once a county has been designated nonattainment, the state has 'the primary responsibility for assuring air quality within’ its borders.” 110 'The state must develop a state implementation plan ( SIP ) that ‘ provides for implementation, maintenance, and enforcement’ of the unattained standard.'¹¹¹ 'At that point, ‘ the Administrator shall approve such submittal as a whole if it meets all of the applicable requirements of this chapter.''¹¹²

Page 24With regard to state SIP s and NAAQS , the Supreme Court held more than 50 years ago that “so long as the ultimate effect of a State’s choice of emission limitations is compliance with the national standards for ambient air, the State is at liberty to adopt whatever mix of emission limitations it deems best suited to its particular situation.'¹¹³ That does not mean, however, that EPA has no substantive role to play in evaluating whether a state’s SIP is in 'compliance with the national standards for ambient air.' EPA is required by the CAA to ensure that a state SIP complies with all aspects of the Good Neighbor Provision.¹¹⁴ Our court has recognized that “ EPA cannot approve a SIP revision ‘if the revision would interfere with any applicable requirement concerning attainment' of NAAQS 'or any other applicable requirement’ of the CAA .'¹¹⁵ JUSTICE SCALIA explained in EPA v. EME Homer City Generation, L.P. , 116 that '[d]etermining the overall level of air pollutants allowed to be emitted in a State is comparable to determining [ NAAQS ] , which the courts have recognized as EPA ’s responsibility, and is distinguishable from determining the particular mix of controls among individual sources to attain those standards, which the caselaw identifies as a State responsibility.'¹¹⁷

Page 25This reading is confirmed by other CAA provisions, including 42 U.S.C. § 7410( l ), which governs review of a revision of a previously approved SIP . It directs EPA not to approve a revision 'if the revision would interfere with any applicable requirement concerning attainment . . . or any other applicable requirement of this chapter.'¹¹⁸ Like § 7410(k)(3), § 7410( l ) directs EPA to determine a submission's compliance with the CAA .

Similarly, 42 U.S.C. § 7410(k)(5) obligates EPA to call for SIP revisions if EPA finds that an area's implementation plan is deficient. It provides:

Whenever the Administrator finds that the applicable implementation plan for any area is substantially inadequate to attain or maintain the relevant [ NAAQS ], to mitigate adequately the interstate pollutant transport described in section 7506a of this title or section 7511c of this title, or to otherwise comply with any requirement of this chapter, the Administrator shall require the State to revise the plan as necessary to correct such inadequacies.¹¹⁹

This provision unambiguously grants EPA authority to determine that a SIP is substantially inadequate to achieve statutory requirements.

The Good Neighbor Provision has two related but separate requirements relevant to the present case. The state's SIP must “prohibit[] . . . any source or other type of emissions activity within the State from emitting any air pollutant in amounts which will- [1] contribute significantly to nonattainment in, or [2] interfere with maintenance by, any Page 26other State with respect to any such national primary or secondary ambient air quality standard.'¹²⁰

The Supreme Court held in EME Homer that 'the Good Neighbor Provision delegates authority to EPA at least as certainly as the CAA provisions involved in Chevron .'¹²¹ With regard to “contribute significantly,' the Supreme Court said in EME Homer , '[t]he statute requires States to eliminate those 'amounts' of pollution that 'contribute significantly to nonattainment ’ in downwind States.'¹²² 'Thus,' the Supreme Court said, “ EPA ’s task is to reduce upwind pollution, but only in ‘ amounts’ that push a downwind State’s pollution concentrations above the relevant NAAQS.'¹²³ The Court was concerned with a FIP in EME Homer , but the statutory language it construed is that of 42 U.S.C. § 7410(a)(2)(D)(1), precisely the same provision that is at issue in the appeal before us and that governs SIP s. The language does not and cannot mean one thing for a federally promulgated plan and another for a state-proposed plan. The Supreme Court further held in EME Homer that 'EPA cannot require a State to reduce its output of pollution by more than is necessary to achieve attainment in every downwind State.'¹²⁴ 'If EPA requires an upwind State to reduce emissions by more than the amount necessary to achieve attainment in every downwind State to which it is linked, the Agency will have overstepped its authority, under the Good Neighbor Provision, to Page 27eliminate those ‘ amounts [that] contribute . . . to nonattainment.''¹²⁵ However, the Court discussed over-control and under-control in EME Homer . It said that “while EPA has a statutory duty to avoid over-control, the Agency also has a statutory obligation to avoid ‘under -control,’ i.e. , to maximize achievement of attainment downwind.'¹²⁶ The Court continued, “a degree of imprecision is inevitable in tackling the problem of interstate air pollution. Slight changes in wind patterns or energy consumption, for example, may vary downwind air quality in ways EPA might not have anticipated.'¹²⁷ The Supreme Court reasoned, '[t]he Good Neighbor Provision requires EPA to seek downwind attainment of NAAQS notwithstanding the uncertainties. Hence, some amount of over-control, i.e. , emission budgets that turn out to be more demanding than necessary, would not be surprising. Required to balance the possibilities of under-control and over-control, EPA must have leeway in fulfilling its statutory mandate.'¹²⁸

Even postLoper Bright , we conclude that the Supreme Court's construction of the Good Neighbor Provision is authoritative and precedential in at least the following respects: (1) “ EPA ’s task is to reduce upwind pollution, but only in 'amounts' that push a downwind State's pollution concentrations above the relevant NAAQS ,'; 129 (2) '[i]f EPA requires an upwind State to reduce emissions by more than the amount necessary to achieve attainment in every downwind State to which it is linked, the Agency will have overstepped its authority, under the Good Page 28Neighbor Provision, to eliminate those ‘ amounts [that] contribute . . . to nonattainment,''; 130 and (3) 'while EPA has a statutory duty to avoid overcontrol, the Agency also has a statutory obligation to avoid ‘under -control,’ i.e. , to maximize achievement of attainment downwind,'¹³¹ and therefore, “required to balance the possibilities of under-control and over-control, EPA must have leeway in fulfilling its statutory mandate.'¹³²

The issue in EME Homer was whether, in promulgating a FIP , EPA could consider the cost associated with eliminating “amounts that contribute significantly to nonattainment' in deciding how to apportion emission reduction requirements among more than one state rather than apportioning emission reductions based on the proportionate amount of emissions each state contributed to another state's nonattainment. The Supreme Court explained the issue this way: 'As EPA interprets the statute, upwind emissions rank as ‘ amounts [that] . . . contribute significantly to nonattainment' if they (1) constitute one percent or more of a relevant NAAQS in a nonattaining downwind State and (2) can be eliminated under the cost threshold set by the Agency. In other words, to identify which emissions were to be eliminated, EPA considered both the magnitude of upwind States' contributions and the cost associated with eliminating them.'¹³³ The Court held that EPA could consider the cost, over a strenuous dissenting opinion by JUSTICE SCALIA joined by JUSTICE THOMAS. (The dissenting opinion maintained that the Good Neighbor Provision 'specified quite precisely the responsibility of an upwind State under the Page 29Good Neighbor Provision: to eliminate those amounts of pollutants that it contributes to downwind problem areas,'¹³⁴ and that '[i]t would be extraordinary for Congress, by use of the single word 'significantly,' to transmogrify a statute that assigns responsibility on the basis of amounts of pollutants emitted into a statute authorizing EPA to reduce interstate pollution in the manner that it believes most efficient.'¹³⁵ )

The Supreme Court's decision in EME Homer referred to EPA's “one-percent” threshold repeatedly. The so-called “Transport Rule” was under scrutiny in EME Homer . 136 The Supreme Court recounted that '[u]nder the Transport Rule, EPA employed a 'two -step approach' to determine when upwind States ‘contribute[d] significantly to nonattainment,’ and therefore in ‘amounts’ that had to be eliminated.” 137 'At step one, called the 'screening' analysis, the Agency excluded as de minimis any upwind State that contributed less than one percent of the three NAAQS to any downwind State ‘ receptor,' a location at which EPA measures air quality.'¹³⁸ The Supreme Court explained that under the Transport Rule, “[i]f all of an upwind State’s contributions fell below the one-percent threshold, that State would be considered not to have ‘ contribute[d] significantly’ to the nonattainment of any downwind Page 30State.'¹³⁹ The Supreme Court noted that 'the one-percent threshold corresponded to . . . 0.8 parts per billion (ppb) for 8-hour ozone.'¹⁴⁰ The Supreme Court explained in EME Homer that the one-percent threshold was “the Agency’s interpretation of the Good Neighbor Provision,” specifically the words 'contribute significantly.'¹⁴¹

No party in EME Homer challenged EPA 's use of the one-percent (0.80 ppb) threshold as the agency's interpretation of when a state's emission would not 'contribute significantly' to nonattainment. The Supreme Court therefore accepted that threshold for purposes of construing the Good Neighbor Provision without any analysis of whether EPA ’s conclusion that a threshold measured on the basis of ppb’s at a single or even within a range of values comported with the text of the CAA. The Court said in EME Homer that ' EPA cannot require a State to reduce its output of pollution . . . at odds with the one-percent threshold the Agency has set.'¹⁴² The Court similarly said, ' EPA [cannot] demand reductions that would drive an upwind State’s contribution to every downwind State to which it is linked below one percent of the relevant NAAQS . Doing so would be counter to step one of the Agency’s interpretation of the Good Neighbor Provision.'¹⁴³

In the Final Rule at issue in the present appeal, EPA described a fourstep framework for evaluating SIP s.¹⁴⁴ Step 2 provided that '[i]f a state's contribution value does not equal or exceed the threshold of 1 percent of the NAAQS (i.e., 0.70 ppb for the 2015 ozone NAAQS ), the upwind state is not ‘ linked' to a downwind air quality problem, and the EPA , therefore, concludes that the state does not contribute significantly to nonattainment or interfere with maintenance of the NAAQS in the downwind states.'¹⁴⁵ The Final Rule further explained, “However, if a state’s contribution equals or exceeds the 1 percent threshold, the state’s emissions are further evaluated in Step 3, considering both air quality and cost as part of a multi-factor analysis, to determine what, if any, emissions might be deemed ‘ significant’ and, thus, must be eliminated pursuant to the requirements of CAA section 110(a)(2)(D)(i)(I).'¹⁴⁶

The Texas State Petitioners assert that EPA ’s use of one percent of the ozone NAAQS as part of its analysis of SIP s is impermissible under the CAA . 147 But an examination of their arguments reveals that the actual points of contention between Texas and EPA boil down to two matters. The first is that EPA used data and modeling that it published after Texas submitted Page 32its SIP on August 17, 2018, and after the statutory deadline for SIP revisions.¹⁴⁸ We consider that contention in more detail below. The second matter in dispute is not actually the one-percent threshold; it's whether Texas's data and methodology were scientifically sound. EPA applied a screening threshold of one percent of the ozone NAAQS (i.e., 0.7 ppb), concluding that contributions below that amount were considered de minimis and therefore discounted.¹⁴⁹ If Texas's contribution to a receptor was equal to or more than one percent of the NAAQS , then EPA considered it linked to that receptor.¹⁵⁰ But Texas also used one percent of the ozone NAAQS at step 2 of its approach in its SIP submission.¹⁵¹

When Texas applied the one percent of ozone NAAQS (0.7 ppb) threshold to its data, it projected that in 2023, there would be sixteen nonattainment or maintenance receptors at or above that level.¹⁵² These receptors were in Arizona, California, and Colorado.¹⁵³ All but one of the maintenance receptors were also nonattainment receptors.¹⁵⁴

Texas argues that “ EPA treated any ozone contribution by Texas to a downwind receptor at greater than one percent of the NAAQS as a presumptively ‘significant’ contribution,” and that “Congress imposed no such threshold.'¹⁵⁵ But as Texas explained in its initial brief, Texas itselfPage 33 “identified nonattainment and maintenance receptors with projected Texas contributions to ozone design values of greater than one percent of the NAAQS and selected those receptors for further review at step 3' of its approach.¹⁵⁶ It is at step 3 in Texas's analysis that the true nature of Texas's complaint becomes evident.

At step 3, Texas asserts that it is entitled to use a 'weight-ofevidence” analysis “to determine whether Texas’s ozone contributions to downwind States were 'significant.''¹⁵⁷ Texas says that its 'ozone contributions would be deemed ‘significant’ ‘only if there [were] a persistent and consistent pattern of contribution on several days with elevated ozone.''¹⁵⁸ Of course, the text of the Good Neighbor Provision does not define 'significant contribution' in terms of 'a persistent and consistent pattern of contribution on several days with elevated ozone.” Texas makes no attempt whatsoever to quantify its “elevated ozone” standard or specify the number of days that constitutes “several days.” Equally importantly, Texas says that '[t]o determine whether such a pattern existed, TCEQ weighed a variety of factors for each linked receptor, including trends in the design values, the meteorological conditions that lead to high ozone formation at the downwind receptor, and the number of days with elevated ozone.'¹⁵⁹ It should be readily apparent that whether Texas's ozone contributions would be 'significant' turned not on the text of the Good Page 34Neighbor Provision, but on highly technical, factand policy-laden determinations. Such determinations are subject to deferential review under the APA , as Loper Bright made clear, 160 and as our court has long recognized.¹⁶¹

EPA engaged in an extensive technical analysis of the data and methodology underlying Texas’s SIP in both its proposed disapproval of the SIP 162 and the final rule disapproving the SIP . 163 Texas's SIP submission looked at the number of monitored days when ozone levels exceeded the 70 ppb NAAQS standard at linked monitors in Colorado in the previous ten years and at linked monitors in California in the previous five years.¹⁶⁴ Texas noted in its SIP submission that '[t]rends in elevated ozone days overall appear to be decreasing” at the linked monitors in Colorado and “[t]rends in elevated ozone days overall appear to be flat' at the linked monitors in California.¹⁶⁵ EPA responded that '[w]hile this data supports that the number of ozone exceedance days is improving, neither the analysis of the number of high ozone days in Colorado or California provide any evidence to Page 35refute [Texas's] photochemical modeling results that show these areas should be considered nonattainment and/or maintenance receptors.'¹⁶⁶

Texas performed a back trajectory analysis to model whether air parcels at linked monitors came from Texas on days when ozone was measured to exceed the NAAQS . 167 With respect to the linked monitors in Colorado, Texas looked at “trajectories during elevated ozone episodes, that start within Colorado’s mixing layer, that do not hit the surface, and that have endpoints within Texas’[s] mixing layer” in order to “find a clear case where emissions in Texas would affect the ozone in Colorado.'¹⁶⁸ Texas found:

Those filters showed that 6% of elevated ozone days in Colorado had trajectories that reached the mixing layer in Texas. Further analysis of the trajectories by year showed that 66% of days where trajectories reached the Texas mixing layer occurred in 2011 and 2012. There are many years where no trajectories reach Texas from Colorado. In the years where no trajectories reached Texas, the tagged monitors still observed a high number of elevated ozone days and fourth-highest eighthour ozone concentrations above 70 ppb.¹⁶⁹

From this, Texas concluded that '[a]lthough air from Texas can reach Colorado, the air from Texas does not appear to significantly affect the ozone Page 36concentrations.'¹⁷⁰ Texas undertook a similar analysis with respect to the linked California receptors.¹⁷¹

EPA discussed this analysis at length and noted “several concerns” with Texas’s performance of the back trajectories, including:

start time and heights, length (number of hours) of the back trajectory, inappropriate removal of some back trajectories based on start height, center-line height touch down, and trajectory center-line height when over Texas, and inappropriate counting of trajectories by not considering that the center-line represents the centerline of a much wider area of air parcels that could have reached the monitor/receptor.¹⁷²

“Due to these concerns,” EPA continued, it found 'the results of [Texas's] back trajectory and endpoint analysis [were] flawed (underestimate[d] back trajectories that reach Texas) and [did] not provide evidence that refute[d] [Texas's] photochemical modeling analysis results' that showed linked receptors.¹⁷³ EPA elaborated that 'even valid back trajectories are of limited use” because they “simply estimate[] the path [of] a parcel of air backward in hourly steps for a specified length of time.'¹⁷⁴ The back trajectory model, HYSPLIT , “estimates the central path in both the vertical and horizontal planes. The HYSPLIT central path represents the centerline with the understanding that there are areas on each side horizontally and vertically that also contribute to the concentrations at the end point.'¹⁷⁵ BecausePage 37 “[t]he horizontal and vertical areas that potentially contribute to concentrations at the endpoint (monitor) grow wider from the centerline the further back in time the trajectory goes,' EPA emphasized that 'a HYSPLIT centerline does not have to pass directly over emissions sources or emission source areas, but merely relatively near emission source areas for those areas, to contribute to concentrations at the trajectory endpoint.'¹⁷⁶

EPA noted that it 'relies on back trajectory analysis as a corollary analysis along with observation-based meteorological wind fields at multiple heights to examine the general plausibility of the photochemical model ‘linkages.’” 177 But '[s]ince the back trajectory calculations do not account for any air pollution formation, dispersion, transformation, or removal processes as influenced by emissions, chemistry, deposition, etc., the trajectories cannot be used to develop quantitative contributions.'¹⁷⁸ 'Therefore,' in EPA ’s view, 'back trajectories cannot be used to quantitatively evaluate the magnitude of the existing photochemical contributions from upwind states to downwind receptors.'¹⁷⁹

Texas modeled its average contributions to linked receptors on days in the (then-future) year 2023 when the receptors were projected to measure ozone concentrations above 70 ppb.¹⁸⁰ For the Colorado receptors, Texas projected that its average contribution would be between 0.71% and 1.21% of the ozone concentration.¹⁸¹ For the California receptors, Texas projected an Page 38average contribution between 0.00% and 0.73%.¹⁸² Texas therefore concluded that its 'expected impact [was] not significant.'¹⁸³ Critiquing Texas’s analysis, EPA noted that EPA ’s technique for evaluating a state’s projected contribution focused on the five to ten days in 2023 with the highest projected ozone concentration, whereas Texas used every day projected to be above 70 ppb.¹⁸⁴ EPA explained that 'this meant many more days could be included in the average which had the effect of showing a smaller estimated contribution.'¹⁸⁵ EPA 'believe[d] it [was] appropriate to focus on the highest values as these [were] the ones that ultimately [would] have to be reduced for the [ NAAQS ] standard to be attained.'¹⁸⁶ EPA concluded: ' EPA 's review of [Texas's] alternate contribution method analysis for California and Colorado receptors is that it does not provide substantial evidence that refutes [Texas’s] photochemical modeling analysis results.'¹⁸⁷

Texas evaluated the collective ozone contribution from all upwind states to the monitors to which Texas was linked.¹⁸⁸ For the Colorado receptors, Texas assessed that the total interstate contributions ranged from 9.32% to 10.27% of the total ozone concentrations.¹⁸⁹ For the California Page 39receptors, the range was between 3.20% and 4.58%.¹⁹⁰ Texas asserted that this factor supported 'the conclusion that interstate transport does not contribute significantly to nonattainment at these monitors.'¹⁹¹

EPA made two responses. First: 'As an initial matter, the EPA is not solely relying on [Texas’s] findings of linkages to Colorado and California but is also relying on its own findings of linkages to areas in the Midwest Region.'¹⁹² Accordingly, EPA stated that Texas's 'analysis of relative contributions to Colorado and California does not provide justification for not addressing downwind impacts.'¹⁹³ EPA also concluded:

Nonetheless, EPA has found in the past that certain California receptors are so heavily impacted by local emissions, and total upwind contribution is so low, that those receptors may not be considered to be affected by interstate ozone transport. However, this is a narrow circumstance that does not apply in the vast majority of cases and has never been applied outside of California. EPA has previously found, for instance, that receptors in Colorado are heavily impacted by upwind-state contribution. . . . EPA affirms, contrary to [Texas’s] suggestion, that the Colorado receptors [Texas] analyzed are impacted by upwind state contributions.¹⁹⁴

Texas used a Direct Decoupled Method ( DDM ) analysis as another way to evaluate its contributions to its linked Colorado receptors.¹⁹⁵ Texas explained that DDM “is a probing tool . . . that estimates the responsiveness Page 40of ozone formation to small changes in any input parameter.'¹⁹⁶ For its analysis, Texas examined the responsiveness of ozone formation at the linked Colorado monitors to changes in Texas's emissions of ozone precursors.¹⁹⁷ It concluded that although the DDM analysis exhibited “a limited responsiveness,” “the instances where this occurs are infrequent and rarely coincide with' ozone concentration exceeding 70 ppb.¹⁹⁸ Evaluating this analysis, EPA noted that “[t]he DDM modeling does show some response to Texas [ozone-precursor] emissions but from the scale it is hard to discern the level of response.'¹⁹⁹ Still, EPA explained that 'the results of the DDM tool showing only a relatively small response to reductions is not inconsistent with the finding that Texas emissions contribute significantly to elevated readings in Colorado.'²⁰⁰

With respect to the California receptors, Texas took into account the “persistent nonattainment issues” caused by the topological and meteorological features of southern California.²⁰¹ In EPA 's view, 'this information does not refute [Texas’s] modeling” because “photochemical modeling is the most sophisticated tool available to estimate future ozone levels.'²⁰²

However, as discussed earlier, we are no longer confident that the foregoing analyses by EPA formed the basis for its decision to disapprove Page 41Texas’s SIP . If instead, EPA ’s modeling and data published after Texas filed its SIP was outcome determinative or was the primary basis for denying approval of Texas’s SIP , Texas would be similarly situated to Mississippi. In our now-withdrawn opinion, we vacated the disapproval of Mississippi’s SIP in the Final Rule and remanded to EPA . We reaffirm that disposition as to Mississippi.

Though Texas asserts that 'the 'one -percent threshold does not appear in the text of' the good neighbor provision,'²⁰³ Texas posits that “[m]aybe EPA could proceed this way if it ‘were instead defending a FIP,’ where the agency is ‘entitled to exercise far more discretion.’” 204 The fact of the matter is that the section of the CAA at issue is the same provision for both SIPs and FIPs-§ 7410(a)(2)(D).²⁰⁵ A FIP, just like a SIP, must “contain adequate provisions . . . prohibiting . . . 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 any such [NAAQS].'²⁰⁶

The question is how to determine what 'amounts' 'contribute significantly to nonattainment . . . or interfere with maintenance.”

Page 42A natural reading of “amounts” suggests that Congress had in mind some measurable amount, just as NAAQS are expressed in measurable amounts, such as 0.70 ppb. The fact that the CAA directs EPA to re-evaluate NAAQS at least every five years 207 and to monitor whether a state's approved SIP continues to meet the Good Neighbor Provision in between periodic revisions to NAAQS 208 indicates that Congress thought that circumstances might, or even would, change and evolve, and that the meaning of “amounts which will . . . contribute significantly” would not be static but would have to be re-evaluated from time to time. That evaluation is delegated to EPA, much as the Supreme Court held in Michigan v. EPA 209 that “Congress instructed EPA to add power plants to [a] program if (but only if) the Agency finds regulation ‘appropriate and necessary.’” 210

In order to evaluate whether a SIP complies with the Good Neighbor Provision, EPA has derived a standard that employs measurable amounts, at least as a starting point, that are tethered to NAAQS. It has done so in at least two notice-and-comment rulemaking proceedings 211 prior to the Final Rule at issue in this appeal. In the Final Rule that is before us, EPA expressly and extensively addressed comments about the continued use of the 1% of NAAQS standard.²¹² EPA's Final Rule addressed, in detail, comments about Page 43the 'Technical Merits of a 1 Percent of the NAAQS Contribution Threshold.'²¹³ It then addressed in detail 'Justification of a 1 Percent of the NAAQS Contribution Threshold.'²¹⁴ Indeed, a subsection of that discussion is prefaced: “Commenters contend that the EPA cannot use the 1 percent threshold as a determination for significance.'²¹⁵ The Final Rule also summarized the history of the 1% standard and EPA's rationale in adopting it.²¹⁶

The initial notice-and-comment rulemaking in 2011 that employed a 1% standard for ozone explained what other standards EPA considered and why it chose the 1% standard.²¹⁷ EPA ultimately concluded, 'Considering the combined downwind impact of multiple upwind states, the health effects of low levels of . . . ozone pollution, and EPA’s previous use of a 1 percent threshold for PM 2.5 in CAIR , EPA’s judgment is that the 1 percent threshold is a reasonable choice.'²¹⁸ EPA had considered .05%, 1%, and 5%.²¹⁹ It explained its choice of 1%.²²⁰

Page 44However, as discussed below, EPA is reconsidering its use of 1% of the 2015 ozone NAAQS as a screening threshold in evaluating whether a SIP meets the requirements of the Good Neighbor Provision. We therefore express no opinion on whether EPA’s exercise of discretion in arriving at 0.7 ppb (1% of the 2015 NAAQS) was within the “outer statutory boundaries of [that] delegation[]' or 'consistent with the APA.'²²¹

The Texas Industry Petitioners argue that two aspects of the agency’s decisionmaking process were arbitrary and capricious. First, they identify a portion of the technical support document that EPA created to augment its Page 45disapproval of Texas's SIP . 222 In that document, EPA discussed why it believed that Texas's ozone modeling underestimated future ozone levels.²²³ EPA identified two nonattainment areas in Texas for which the state had previously submitted SIP s.²²⁴ Those prior SIP s projected that ozone design values would decrease by about 1 to 1.2 ppb per year-which was approximately what EPA eventually measured the trend to be.²²⁵ Based on this trend, EPA hypothesized that there would be approximately a 3 to 4 ppb decrease in ozone concentrations over the (then-future) period 2020 to 2023.²²⁶ 'To assess if [Texas's] ozone transport modeling [was] potentially underestimating future year 2023 modeled [design values] in [the monitor] areas, the EPA compared 2020 monitored [design values] at several of the typically higher ozone monitors in the [relevant areas] with [Texas's] projected 2023 [design values].'²²⁷ EPA explained that '[w]hile not as exact as developing new modeling of emission changes from 2020 to 2023 to project 2023 [design values] (nor appropriate for use in any other context), using the general 3-4 ppb approximation provides a ballpark estimate to evaluate whether [Texas’s] modeling might be underestimating 2023 future [design values].'²²⁸ Noting that many monitors in the relevant areas 'would need a drop in ozone [design values] over the next 3 years of [more than] 5Page 46 ppb . . . to reach [Texas’s] 2023 modeled projected levels,” EPA concluded that '[t]his analysis supports a finding that [Texas's] modeling is likely underestimating future ozone [design values] in the two nonattainment areas in Texas.'²²⁹

In the view of the Texas Industry Petitioners, this critique was unreasonable. They point out that EPA conceded it found 'nothing that was a clear cause of the much lower 2023 [design values] that [Texas’s] modeling [was] projecting.'²³⁰ The Texas Industry Petitioners object that rather than identifying an error in the modeling, EPA relied on a “ballpark estimate” of 2023 ozone conditions.²³¹ They emphasize that EPA said its estimation was “not as exact as developing new modeling,” “not usable in any other CAA action,' and not 'a defensible basis on which to reach any conclusions regarding future air quality conditions.'²³² The Texas Industry Petitioners conclude that “[b]y basing its disapproval on a ‘ballpark estimate’ that was admittedly not ‘a defensible basis’ for ‘ any conclusions regarding future air quality conditions,’ EPA utterly failed to meet” the arbitrary and capricious standard.²³³

The Texas Industry Petitioners overstate the significance of this aspect of EPA ’s analysis to the agency’s conclusion that Texas’s modeling may have underestimated future ozone levels. Based on 2020 design values measured by EPA , the agency noted that for Texas's modeling to be Page 47accurate, many monitors would need a decrease in ozone design values of 5 ppb or more over three years.²³⁴ The decrease would have to be even greater for areas where the ozone concentration is higher.²³⁵ EPA explained that it used its 3 to 4 ppb “ballpark figure” “only as a bounding assumption relevant to an analytical exercise establishing why the [Texas] modeling submitted in its transport SIP submittal likely under-projects future ozone concentrations.'²³⁶ In other words, the 3 to 4 ppb figure was a heuristic meant to highlight why a 5 ppb or greater decrease in ozone concentrations was implausible. That observation contributed to EPA 's conclusion that Texas’s modeling “may understate anticipated ozone levels at high ozone monitors' in some areas.²³⁷

More fundamentally, EPA 's criticisms of Texas's modeling were separate from EPA ’s assessment and disapproval of Texas’s SIP on its own merits. True, EPA noted that it thought Texas's 'modeling likely underestimates' Texas's impacts on Colorado and California.²³⁸ But that was after expressing that Texas’s data showed that the state had a “persistent and consistent pattern of contribution on several days with elevated ozone” and that the agency did 'not find the additional weight of evidence evaluations' to provide 'compelling reasons to discount the impacts indicated in Colorado and California by the [ Texas ] modeling .'²³⁹ We cannot say that EPA ’s approach to critiquing Texas’s modeling evinced that it failed Page 48to 'reasonably consider[] the relevant issues and reasonably explain[] the decision.'²⁴⁰

Second, the Texas Industry Petitioners contend that EPA acted arbitrarily and capriciously by failing to treat Texas’s SIP like previous SIP s the agency had approved.²⁴¹ Because 'an agency changing its course must supply a reasoned analysis,' we have recognized that '[i]t is a bedrock principle of administrative law that an agency must 'treat like cases alike.''²⁴² In the view of the Texas Industry Petitioners, EPA treated Texas’s SIP inconsistently with a SIP submitted by Arizona in response to the 2008 ozone NAAQS update.²⁴³ In particular, they emphasize that EPA approved Arizona's SIP despite the state having had a greater than one percent contribution to problem receptors in California.²⁴⁴ EPA ’s approval concluded that Arizona lacked Good-Neighbor obligations because the collective interstate contribution to the California receptors was between 2.5% and 4.4%, which EPA considered 'negligible.'²⁴⁵

We cannot conclude that EPA treated Texas's SIP inconsistently with Arizona's 2008 ozone NAAQS SIP . When addressing Texas's Page 49collective interstate contribution analysis, EPA acknowledged that the agency “has found in the past that certain California receptors are so heavily impacted by local emissions, and total upwind contribution is so low, that those receptors may not be considered to be affected by interstate ozone transport”—citing specifically its approval of Arizona’s 2008 ozone NAAQS SIP . 246 Consistent with EPA discounting Arizona's link to California receptors in the Arizona SIP approval, EPA explained that it “need not draw any conclusions here regarding whether the California sites [Texas] identified should or should not be considered receptors for ozonetransport purposes.'²⁴⁷ Because Arizona was only linked to California receptors, setting aside those receptors meant that Arizona lacked GoodNeighbor obligations. Here, Texas’s modeling showed that it was linked to receptors outside California.

The Texas Industry Petitioners suggest that collective interstate contributions modeled to the linked Colorado receptors should be similarly discounted.²⁴⁸ They emphasize that the modeled contributions were between 9.32% and 10.27%, which they suggest are similar to the 2.5% to 4.4% contribution range that EPA dismissed as 'negligible' in its approval of Arizona’s 2008 ozone NAAQS SIP . 249 It is not obvious to us that collective interstate contributions that are two to five times as large are necessarily negligible. Regardless, in its proposed disapproval of Texas's SIP , EPA explained that unlike for California, ' EPA has previously found . . . that Page 50receptors in Colorado are heavily impacted by upwind-state contribution.'²⁵⁰ EPA considered Texas's argument that the Colorado receptors should be set aside like California’s, and the agency “affirm[ed], contrary to [Texas]'s suggestion, that the Colorado receptors [Texas] analyzed are impacted by upwind state contributions.'²⁵¹ EPA ’s decision in this regard “is based upon its evaluation of complex scientific data within its technical expertise.'²⁵² EPA 's 'path may reasonably be discerned.'²⁵³ We see no inconsistent treatment.

After we issued our now-withdrawn opinion in this case, the Texas State Petitioners and Texas Industry Petitioners sought rehearing en banc. While those petitions were pending, EPA filed a motion to stay further proceedings in our court in February 2025. EPA stated that it was seeking “to hold the case in abeyance with status reports due every 90 days to allow new Agency leadership to review the underlying final rule.'²⁵⁴ EPA advised in the motion that '[i]t is possible that after its review, EPA could take further action that may obviate the need for judicial resolution of some or all Page 51of the disputed issues. Good cause thus exists for an abeyance.'²⁵⁵ The Texas State Petitioners opposed the motion to stay, as did certain Texas utilities 256 and Texas industry groups.²⁵⁷ We denied the motion and subsequently issued an opinion.²⁵⁸

On January 30, 2026, EPA issued a proposed rule and reconsideration of the Final Rule, 259 part of which is the subject of the present appeal. In light of that notice and reconsideration, we have withdrawn our initial opinion in this case.

First, we have already alluded to indications in the the January 2026 proposed rule and reconsideration that EPA placed considerable reliance on data and modeling published after Texas submitted its SIP . The January 2026 notice states

When acting on certain submissions in 2023, however, the EPA interpreted the March 2018 and August 2018 memoranda as allowing EPA to give greater weight to the EPA 's latest modeling results (referred to as '2016v3') when it showed linkages not identified in the March 2018 memorandum modeling and to apply a 1 percent of the NAAQS contribution threshold. Based on the SIP submissions, the EPA ’s interpretation of its memoranda, and the 2016v3 modeling, the

Page 52EPA disapproved the SIP submissions from Alabama, Kentucky, Minnesota, Mississippi, Nevada, and 16 other States in “Air Plan Disapprovals; Interstate Transport of Air Pollution for the 2015 8-Hour Ozone National Ambient Air Quality Standards” (“ SIP Disapproval Action').²⁶⁰

This seems a strong indication that EPA placed substantial reliance on data that was unavailable to Texas when it submitted its SIP . In the same vein, EPA says in the January 2026 notice, 'The EPA explained that, in our view, we had the authority and responsibility in evaluating interstate transport obligations to consider the best available information.'²⁶¹

Second, it is unclear whether in re-evaluating its denial of SIP s in the Final Rule at issue in this appeal, EPA will apply a 1 ppb standard, one percent of the 2015 ozone NAAQS (0.7 ppb), or some other, less-restrictive standard. The January 2026 notice states, 'The EPA intends to take a subsequent action consistent with this proposal, subject to further public input, to address interstate transport obligations for the 2015 8-hour ozone NAAQS for other States.'²⁶² At one point, EPA says, '[T]he EPA finds it appropriate to presumptively apply a 1-ppb contribution threshold for the consistent treatment of all States.'²⁶³ Yet, EPA also says, 'In addition, as described in Section III.C.4. of this preamble, the EPA ’s proposed approach for evaluating air quality information in this action is to first rely on information provided in the March 2018 memorandum, as included by States Page 53in their SIP submissions, and then consider more recent EPA modeling information only if necessary to determine whether any linkages are still projected to persist.²⁶⁴ This seems to indicate that for a state such as Texas, EPA would use data and modeling that was unavailable to Texas when it submitted its SIP . EPA again says at another juncture in the January 2026 notice that it is proposing to use a 1-ppb threshold at its Step 2 analysis paradigm:

in response to these opinions [our now-withdrawn opinion and a Sixth Circuit decision 265 ] and in light of the 2018 August Memorandum and any reliance interests it may have engendered in State air [sic] agencies, the EPA is proposing to determine that a 1-ppb threshold is the appropriate Step 2 threshold to rely on in the first instance for the 2015 ozone NAAQS for all States in this action and any future actions related to the 2015 ozone NAAQS . As noted in the August 2018 memorandum, nationally the 1-ppb threshold captures a generally comparable amount of total upwind contributions overall (70 percent using 1 ppb versus 77 percent using 1 percent (0.70 ppb))—when considering all receptors. Further, in the EPA 's latest modeling, 2016 Version 3 Emissions Platform Modeling (“2016v3”), the difference in the amount of total upwind contributions captured is even less, identifying a difference of only 5 percentage points.²⁶⁶

But, EPA further says in the January 2026 notice, 'In this proposal, the EPA also solicits comment on the use of thresholds other than the 1-percent or 1- Page 54ppb thresholds discussed in this action, such as a 5-percent threshold or a 2ppb threshold, including a basis for relying on any suggested alternative threshold.'²⁶⁷ It may be that if EPA were to apply a 5-percent threshold in evaluating Texas’s SIP submission, it would approve that SIP . Since EPA has changed course since we issued our initial opinion, and has not settled on how it will evaluate certain SIP s submitted in response to the revised 2015 ozone NAAQS , the disapproval of Texas’s SIP must be vacated, and this matter must be remanded to EPA .

As explained at the outset of this opinion, neither Louisiana nor any other party sought panel rehearing or rehearing en banc with regard to our disposition of either Louisiana's or Mississippi's petitions for review of EPA ’s disapproval of their respective SIP s. We leave our prior determinations regarding their SIP s undisturbed. We reinstate the parts of our prior decision that dealt with Louisiana’s and Mississippi’s SIP s.²⁶⁸

* * *

The petition for review of EPA ’s disapproval of Louisiana’s SIP is DENIED . The petitions for review of EPA ’s disapproval of Mississippi’s and Texas’s SIP s are GRANTED , those disapprovals are VACATED , and those matters are REMANDED to EPA .