SIP Disapprovals Are Locally Applicable Actions; EPA Must Independently Assess State Good Neighbor SIPs But May Not Base Disapprovals on Post‑Submission Modeling Without a Reasoned Explanation

Introduction

This consolidated challenge arose after the Environmental Protection Agency (EPA) disapproved Good Neighbor State Implementation Plan (SIP) submissions from 21 states following the 2015 ozone National Ambient Air Quality Standards (NAAQS) revision. The petitioners here—Texas, Louisiana, Mississippi, their environmental agencies and public utility commissions, and a variety of energy and industrial groups—sought review of EPA’s disapprovals for their states. They argued the disapprovals were arbitrary, capricious, and contrary to the Clean Air Act (CAA), and that the cases belonged in the D.C. Circuit.

In a decision with immediate consequences for venue, cooperative federalism under the CAA, and administrative “hard look” review, the Fifth Circuit:

Held venue in this circuit was proper because the “final action” under 42 U.S.C. § 7607(b)(1) is each state-specific SIP disapproval (a locally/regional action), not the multi-state final rule.
Clarified EPA’s role when reviewing Good Neighbor SIPs: even assuming a state advances a reasonable interpretation of the Good Neighbor Provision, EPA must independently determine whether the SIP actually “meets all applicable requirements” and contains “adequate provisions” to eliminate significant downwind contributions.
Denied the Texas and Louisiana petitions on the ground that, under their own methodologies and record evidence, their SIPs showed linkages and a persistent pattern of contribution that triggered Good Neighbor obligations that the SIPs failed to address.
Granted the Mississippi petition, vacated EPA’s disapproval, and remanded because EPA made the updated (post-submission) 2016-based modeling outcome-determinative without a reasoned explanation for relying on data that did not exist when Mississippi filed its SIP.

Summary of the Opinion

The Fifth Circuit (Judge Richman) resolved three clusters of issues:

Venue: For purposes of § 7607(b)(1), the “final action” is the individual SIP disapproval. Such disapprovals are locally or regionally applicable, so venue lies in the “appropriate circuit,” not exclusively in the D.C. Circuit. The court rejected the Tenth Circuit’s contrary view in Oklahoma v. EPA (2024) and aligned with the Fourth and Sixth Circuits.
EPA’s review role: Without deciding whether states or EPA have primary interpretive control over ambiguous Good Neighbor terms (and without relying on Chevron, which was overruled by Loper Bright), the court held that § 7410(k)(3) requires EPA to independently determine whether a SIP “meets all applicable requirements,” including containing “adequate provisions” to prevent significant downwind contributions—beyond merely checking for a “reasoned analysis.”
Merits and remedy:
- Louisiana and Texas: EPA reasonably disapproved both SIPs because, even under each state’s own “persistent and consistent” significance criterion and evidence, the modeling showed linkages and persistent contributions. The states’ additional meteorological, back-trajectory (HYSPLIT), and other “weight-of-evidence” analyses did not credibly refute those linkages or demonstrate insignificance on the relevant high-ozone days.
- Mississippi: EPA’s disapproval was arbitrary and capricious because EPA made the post-submission 2016-based modeling outcome-determinative while claiming, elsewhere, that updated modeling was not dispositive. Vacatur—not remand without vacatur—was appropriate; EPA cannot implement a federal implementation plan (FIP) in Mississippi without a lawful disapproval.

Analysis

Precedents Cited and Their Influence

EPA v. EME Homer City Generation (2014): Recognized the Good Neighbor Provision’s ambiguity and upheld EPA’s two-step approach in the Transport Rule (screening threshold plus cost-effective allocation). Here, the Fifth Circuit used EME Homer to validate EPA’s reliance on linkages at Step 2 but did not rely on Chevron deference; instead, it focused on whether EPA’s disapprovals comported with the states’ own submissions and the statute.
Ohio v. EPA (2024): Reiterated states’ primary responsibility to implement NAAQS and EPA’s review role. The Fifth Circuit cited Ohio v. EPA for SIP responsibilities and Good Neighbor context.
Alaska Dept. of Environmental Conservation (ADEC) v. EPA (2004): Upheld limited but meaningful EPA supervisory authority over state BACT determinations under different CAA provisions. The Fifth Circuit distinguished ADEC’s BACT context (site-specific, case-by-case determinations) from the Good Neighbor SIP review, emphasizing § 7410(k)(3)’s directive that EPA approve only if a SIP meets all applicable requirements.
Texas v. EPA (2012) and Texas v. EPA (2016): Confirmed states’ wide discretion over the means to achieve air-quality goals and EPA’s “ministerial” duty to review SIPs for consistency with the Act. In Texas 2016, the court stayed an EPA disapproval because EPA demanded a source-specific analysis not found in the statute; the Fifth Circuit here harmonized that case by reaffirming EPA cannot impose extra-statutory requirements, but may independently assess compliance with statutory requirements.
Calumet Shreveport Refining v. EPA (5th Cir. 2023): Provided the two-step venue test under § 7607(b)(1) (national vs. local applicability; and “based on” nationwide scope or effect with a published finding). The Fifth Circuit relied on Calumet to hold that the “legal effect”—not practical reach—controls and that these SIP disapprovals are locally applicable.
North Carolina v. EPA and Westar Energy, Inc. v. EPA: Reinforced the Good Neighbor architecture and EPA’s authority to decide whether SIPs satisfy statutory requirements, respectively.
Oklahoma v. EPA (10th Cir. 2024): Reached the opposite venue conclusion, treating the combined rule as the “final action.” The Fifth Circuit expressly disagreed, deepening an inter-circuit conflict on venue in multi-state SIP disapproval rules.
Prometheus Radio Project (2021) and State Farm (1983): Anchored the arbitrary-and-capricious standard: reasoned decision-making within a zone of reasonableness, and attention to relevant factors.
Loper Bright (2024): Overruled Chevron. The Fifth Circuit flagged it but found no need to resolve deference questions because EPA’s disapprovals stood or fell on the SIPs’ own terms and the adequacy of EPA’s explanations.

Legal Reasoning

1) Venue under § 7607(b)(1): the “final action” is each SIP disapproval

The court parsed § 7607(b)(1) through the lens of § 7410. Although EPA bundled 21 SIP disapprovals into one Federal Register rule, the CAA repeatedly refers to “any implementation plan” in the singular and directs EPA to approve or disapprove “such submittal.” Therefore, the court held the relevant “final action” is each individual disapproval—prototypically a locally or regionally applicable action. Because the legal effect of each disapproval is confined to a single state, venue belongs in the respective regional circuit unless the action is both (a) based on nationwide scope/effect determinations and (b) accompanied by a contemporaneous “finds and publishes” statement.

EPA did publish a nationwide-scope finding in the multi-state rule, satisfying (b). But the court held (a) was not met: the core justifications were state-specific, “intensely factual” determinations about the contents and sufficiency of each SIP. General, cross-cutting commentary in the rule’s response-to-comments section could not convert those state-specific grounds into “nationwide” determinations. Venue in the Fifth Circuit was therefore proper.

2) EPA’s role reviewing Good Neighbor SIPs: independent compliance determination

The court carefully avoided staking out a broad post–Loper Bright allocation of interpretive authority between states and EPA for the Good Neighbor Provision. Instead, it held narrowly that § 7410(k)(3) requires EPA to independently determine whether a SIP “meets all applicable requirements” before approving it. This means EPA:

Cannot approve a SIP merely because the state offered a “reasoned analysis” under some arguably permissible interpretation;
Must ensure the SIP actually contains “adequate provisions” prohibiting emissions that significantly contribute to nonattainment or interfere with maintenance in downwind states;
Still may not disapprove a SIP for policy disagreements or extra-statutory demands—the federal role is to ensure compliance with statutory criteria, not to dictate the state’s chosen control strategies.

The court drew support from the text and structure of the CAA: § 7410(k)(3) (EPA shall approve only if the submission meets “all applicable requirements”); § 7410(k)(5) (SIP Call authority when a plan is “substantially inadequate”); and § 7426(b) (downwind petition process requiring EPA to determine Good Neighbor violations). Taken together, these provisions contemplate EPA making substantive compliance judgments—not simply second-guessing the state’s reasoning but ensuring the plan actually works under the statute’s terms.

3) Applying the APA standard to each state

Louisiana

Louisiana used a three-step approach akin to EPA’s framework but adopted a 1 ppb Step 2 screen and defined “significant” as a “persistent and consistent pattern” of contribution on several high-ozone days. Louisiana’s own modeling showed contributions at or above 1 ppb at multiple Texas receptors (1.71–4.72 ppb), but the state discounted those linkages using:

HYSPLIT back trajectories (28% of 99 exceedance-day trajectories over Louisiana; 8% originated in Louisiana, with 35% over south Louisiana);
Generalized seasonal wind patterns (southerly flow during much of the ozone season);
Wind roses for broader regions and periods.

EPA explained, with technical specificity, why these additional analyses did not rebut the modeled linkages on the relevant high-ozone days:

Back trajectories are indicative but not quantitative (they do not capture chemistry, dispersion, deposition) and the centerline understates the contributing geographic envelope;
General wind patterns and year-round wind roses are poor proxies for transport on the handful of high-ozone days that drive the design value; surface winds may not reflect upper-level transport; and temporal/spatial variability matters; and
A Step 2 “linkage” already demonstrates the “persistent and consistent” pattern Louisiana’s definition requires because the modeling averages contributions over the five to ten highest projected ozone days.

Given that Louisiana’s own modeling crossed its chosen threshold and its rebuttal evidence did not seriously undermine those linkages, the court upheld EPA’s disapproval as reasonable and reasonably explained.

Texas

Texas conducted its own photochemical modeling (2012 base year) and used a “weight-of-evidence” Step 3 analysis under the same “persistent and consistent” definition. Texas identified linkages to multiple receptors in Colorado and California, but concluded its contributions were not significant after considering design value trends, monitored exceedance-day trends, HYSPLIT, a custom averaging method across all >70 ppb days, collective interstate contributions, a Direct Decoupled Method (DDM) sensitivity analysis, and southern California’s geography/meteorology.

EPA’s key findings—accepted by the court—were:

A Step 2 “linkage” itself demonstrates a persistent and consistent pattern across the highest 5–10 days, satisfying Texas’s own significance definition.
Texas’s weight-of-evidence factors did not credibly refute those linkages:
- Trend data did not support the magnitude of reductions Texas projected by 2023; EPA’s own modeling suggested Texas’s approach likely underestimated future ozone levels and may have missed additional receptors.
- HYSPLIT parameters and counting were flawed; even with valid parameters, trajectories remain non-quantitative and the centerline undercounts contributing air parcels.
- Texas’s averaging method diluted contributions by including many days above 70 ppb instead of focusing on the top 5–10 days that drive attainment.
- Collective interstate contributions could not be used to negate linkages (especially for Colorado, where EPA has previously found upwind impacts to be substantial); and
- DDM’s “limited responsiveness” finding did not contradict significant contribution for high-ozone days.

The court rejected arguments that EPA’s reliance on a “ballpark” trend check rendered its analysis arbitrary: that heuristic was only one piece supporting a broader conclusion that Texas’s modeling likely understated future ozone levels; in any event, EPA had already concluded Texas’s own record showed significant contributions that Texas failed to address. Nor did Texas demonstrate inconsistent treatment with past approvals (e.g., Arizona’s 2008 ozone SIP), because the Arizona approval involved California-only receptors with unusually low total upwind contributions; Texas, by contrast, was linked to receptors beyond California.

Mississippi

Mississippi relied on EPA’s 2011-based modeling, showing a maximum contribution of 0.79 ppb, and used a 1 ppb screen; it found no linkages and thus no Step 3 obligation. In proposing disapproval, EPA explicitly stated that Mississippi’s choice of a 1 ppb screen was “inconsequential,” and then relied on EPA’s updated 2016-based modeling (developed after Mississippi submitted its SIP) to find that Mississippi contributed above both 0.7 ppb (1% of the NAAQS) and 1 ppb to certain 2023 receptors.

The court held this was arbitrary and capricious: EPA had elsewhere represented that updated modeling was not outcome-determinative and that it evaluated each SIP “on its own terms,” yet for Mississippi it made the updated modeling dispositive. Without a reasoned explanation for relying on post-submission data to change the outcome, EPA’s disapproval could not stand. On remedy, the court vacated and remanded; allowing EPA to implement a FIP absent a lawful SIP disapproval would invert § 7410(c)(1)’s structure.

Impact

Venue and multi-state rules: The decision entrenches a growing circuit split with the Tenth Circuit over whether bundled SIP disapprovals are nationally applicable. The Fifth, Fourth, and Sixth Circuits treat each SIP disapproval as the “final action,” keeping challenges in regional circuits unless the statutory “based on a determination of nationwide scope or effect” test is truly met.
EPA’s review duties: The court reaffirms that EPA’s task is not to defer to any “reasonable” state interpretation and analysis in the abstract, but to ensure the SIP actually contains adequate controls to meet the Good Neighbor requirements. States retain broad discretion over how to control emissions; EPA must approve if the SIP complies, but may disapprove if—on the state’s own premises—the plan fails to address significant linkages.
Use of updated modeling: EPA is not categorically barred from considering new modeling, but when updated data did not exist at the time of a SIP’s submission and becomes outcome-determinative, EPA must acknowledge that fact and supply a reasoned explanation. Absent that, the disapproval risks vacatur.
Immediate regulatory consequences: For Mississippi, the FIP cannot be implemented unless and until EPA lawfully disapproves the SIP. For Texas and Louisiana, EPA’s disapprovals stand, with the attendant federal planning consequences unless the states submit approvable revisions.
Technical expectations for states: Back-trajectory and generalized wind-pattern analyses, standing alone, will rarely overcome modeled Step 2 linkages that are averaged over the highest ozone days. Robust Step 3 analyses must directly engage the days that drive design values and quantify control strategy feasibility and downwind benefits.
Administrative law: The case exemplifies post–Loper Bright review focused on text, structure, and reasoned explanation. Courts will scrutinize whether an agency’s generalized statements match state-by-state application and whether outcome-determinative reliance on new data is adequately justified.

Complex Concepts Simplified

NAAQS: National Ambient Air Quality Standards set maximum allowable concentrations for pollutants (e.g., ozone). The 2015 ozone NAAQS is 70 parts per billion (ppb).
SIP vs. FIP: States write SIPs showing how they will meet NAAQS; EPA approves if the SIP meets all applicable requirements. If a SIP is missing or inadequate, EPA must disapprove and then may promulgate a federal implementation plan (FIP) to fill the gap.
Good Neighbor Provision: Requires each state’s SIP to include “adequate provisions” prohibiting in-state emissions from significantly contributing to nonattainment or interfering with maintenance of NAAQS in other states.
Design value: A statistic reflecting air quality used to assess NAAQS attainment; for ozone, it averages the 4th-highest daily maximum 8-hour concentration over three consecutive years.
EPA’s four-step transport framework (simplified):
1. Identify downwind receptors projected to have attainment/maintenance problems in the analytic year.
2. Quantify upwind state contributions to those receptors; screen out de minimis contributions (often 1% of NAAQS, i.e., 0.7 ppb).
3. For “linked” states, perform a multifactor analysis of feasible emission controls to determine whether contributions are “significant.”
4. Adopt control strategies needed to eliminate significant contributions.
“Linkage” (Step 2): A modeled contribution from an upwind state to a downwind receptor at or above the screening threshold; EPA models average contributions across the highest 5–10 ozone days.
HYSPLIT back trajectories: A tool that traces the path of an air parcel backward in time; useful for understanding potential transport pathways, but not for quantifying pollutant formation, chemistry, or total contribution magnitudes.
Direct Decoupled Method (DDM): A sensitivity analysis within photochemical models that estimates how responsive ozone is to changes in precursor emissions; informative but must be tied to the high-ozone days that matter for attainment.

What the Court Did Not Decide

Whether EPA can categorically require the 1% of NAAQS (0.7 ppb) screen in SIPs (the court affirmed on state-specific grounds and did not reach that question).
Who has ultimate interpretive primacy over the Good Neighbor Provision post–Loper Bright (the court assumed without deciding that states may advance reasonable interpretations but held EPA must still independently ensure compliance).
Whether reliance on post-submission modeling is per se impermissible (the defect here was EPA’s unreasoned, outcome-determinative reliance against Mississippi, not the mere use of new data).

Practice Takeaways

States:
- Assume EPA will treat a Step 2 linkage as strong evidence of a “persistent and consistent” pattern on the highest ozone days—especially if the state adopts that concept as its significance criterion.
- Weight-of-evidence showings should directly address the top 5–10 ozone days, tie meteorology/trajectories to those days, and quantify why contributions are de minimis or rendered insignificant by practical controls.
- If adopting a screen other than 1% of NAAQS, build a record showing why that screen is consistent with the statute and does not mask significant impacts.
EPA:
- When relying on updated modeling, acknowledge whether it is outcome-determinative and explain why reliance is appropriate notwithstanding state filing dates.
- Anchor disapprovals in the state’s own record and interpretive choices where possible; that approach withstands APA review even without Chevron deference.
Litigants:
- Venue challenges in multi-state SIP disapproval rules will turn on whether the reviewing court treats the “final action” as the state-specific disapproval (Fifth, Fourth, Sixth Circuits) or the combined rule (Tenth Circuit).
- Be prepared to address whether determinations are state-specific (not nationwide in scope/effect) and to parse the agency’s core justifications rather than peripheral comment responses.

Conclusion

This decision sets two important markers. First, for venue: state-by-state SIP disapprovals bundled in a single rule remain locally or regionally applicable actions, placing litigation in regional circuits absent a true nationwide-scope determination that the agency both makes and publishes. Second, for EPA’s review role: cooperative federalism preserves state discretion over the means, but EPA must independently determine that a SIP contains the “adequate provisions” required by the Good Neighbor Provision—not merely that the state offered a reasonable rationale.

On the merits, the court sustained EPA’s disapprovals of Texas and Louisiana because their own modeling and record supported linkages and a persistent pattern of contribution that their SIPs failed to remedy. Conversely, it vacated the Mississippi disapproval because EPA relied in an outcome-determinative way on post-submission modeling without a reasoned explanation, and reaffirmed that a FIP cannot be imposed without a lawful SIP disapproval. Taken together, the ruling sharpens administrative law’s demand for reasoned, record-rooted explanations while clarifying how cooperative federalism allocates responsibility: states choose how to control, and EPA ensures the plan, as submitted, truly meets the Act’s requirements for downwind neighbors.