Board of County Commissioners of Weld County, CO v. EPA

United States Court of Appeals

FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued December 7, 2022 Decided June 30, 2023

No. 21-1263

BOARD OF COUNTY COMMISSIONERS OF WELD COUNTY,

COLORADO

PETITIONER

v.

ENVIRONMENTAL PROTECTION AGENCY,

RESPONDENT

BOARD OF COUNTY COMMISSIONERS OF BOULDER COUNTY, ET

AL.,

INTERVENORS FOR RESPONDENT

Consolidated with 21-1013

On Petitions for Review of an Action of the

United States Environmental Protection Agency

Ethan G. Shenkman argued the cause for petitioner Board

of County Commissioners of Weld County, Colorado. With

him on the briefs were Charles Birkel, John R. Jacus, Shannon

Stevenson, and Kathleen Pritchard. Bill Davis, Deputy

Solicitor General, Office of the Attorney General of the State

of Texas, argued the cause for petitioners State of Texas and

Texas Commission on Environmental Quality. With him on

2

the briefs were Ken Paxton, Attorney General, Brent Webster,

First Assistant Attorney General, Judd E. Stone II, Solicitor

General, and Michael R. Abrams, Assistant Solicitor General.

Alexandra L. St. Romain, Attorney, U.S. Department of

Justice, argued the cause for respondent. With her on the briefs

were Todd Kim, Assistant Attorney General, Laura J.

Glickman, Attorney, U.S. Department of Justice, and Seth

Buchsbaum, Attorney, U.S. Environmental Protection Agency.

Elliot Higgins, Attorney, U.S. Department of Justice, also

argued the cause for respondent.

David Baake and Ryan Maher argued the cause for Board

of County Commissioners of Boulder County, et al. With them

on the brief were Robert Ukeiley and Joshua D. Smith.

Before: SRINIVASAN, Chief Judge, KATSAS, Circuit Judge,

and ROGERS, Senior Circuit Judge.

Opinion for the Court by Circuit Judge KATSAS.

KATSAS, Circuit Judge: This case involves regulation of

ozone levels under the Clean Air Act. In August 2018, the

Environmental Protection Agency designated northern Weld

County, Colorado and El Paso County, Texas as areas that had

already attained a 2015 ozone pollution standard. But EPA

reversed course after Clean Wisconsin v. EPA, 964 F.3d 1145

(D.C. Cir. 2020), remanded these designations. In November

2021, EPA folded northern Weld and El Paso Counties into

areas previously designated as not having attained the standard.

Weld County contends that EPA improperly relied on data

available in 2018, rather than updated data, and that the data do

not support its adverse designation. We hold that EPA

3

reasonably relied on the same data it had used to make the

original designation and that the data support the revised one.

Texas argues that El Paso’s 2021 nonattainment

designation was impermissibly retroactive because EPA made

it effective as of the 2018 attainment designation. As a result,

a statutory deadline for El Paso to attain the governing standard

passed some three months before EPA made the nonattainment

designation. And missing the deadline triggered adverse legal

consequences. We therefore agree with Texas that El Paso’s

revised designation, backdated to the date of the original one,

was impermissibly retroactive.

I

A

The Clean Air Act establishes a comprehensive scheme to

reduce the atmospheric concentration of various air pollutants.

The scheme works in three relevant steps.

First, EPA must establish and periodically revise national

ambient air quality standards (NAAQS) for pollutants that may

endanger public health or welfare. These standards set forth

the maximum permissible concentration of the pollutant in the

atmosphere. 42 U.S.C. §§ 7408(a)(1)(A), 7409(a)–(b).

Second, EPA must divide the country into geographic

areas and designate them according to whether they satisfy the

new standard. 42 U.S.C. § 7407(d)(1)(B)(i)–(iii). EPA marks

an area as “attainment” when local atmospheric concentration

of the pollutant—the area’s so-called “design value”—falls

below the relevant NAAQS. However, an area must be

designated as “nonattainment” if its design value exceeds that

4

level or if the area “contributes” to nonattainment in a “nearby

area.” Id. § 7407(d)(1)(A)(i).

EPA works with the States to make these designations.

Within a year of a new NAAQS, each State must make “initial

designations” suggesting appropriate areas and attainment

designations. 42 U.S.C. § 7407(d)(1)(A)–(B). If EPA

proposes to modify an initial designation, it must notify the

State in advance and allow it to contest the proposal. Id.

§ 7407(d)(1)(B)(ii). EPA must finalize its designations within

two years of promulgating the new standard—a deadline

extendable for at most one year. Id. § 7407(d)(1)(B)(i).

Third, States must ensure that their designated areas

achieve or maintain attainment status. To that end, a State must

prepare a State Implementation Plan (SIP) specifying how each

of its areas will do so. 42 U.S.C. § 7407(a). EPA sets the SIP

deadline, which must be within three years of any

nonattainment designation. Id. § 7502(b).

A nonattainment designation triggers more stringent

regulation. For attainment areas, the SIP need only set forth

measures “to prevent significant deterioration of air quality.”

42 U.S.C. § 7471. But for nonattainment areas, the SIP must

impose “all reasonably available” measures to achieve

attainment “as expeditiously as practicable.” Id. § 7502(c)(1).

In addition, the Clean Air Act imposes deadlines for

nonattainment areas to achieve attainment, which are called

“attainment dates.” For ozone standards, EPA must designate

nonattainment areas as marginal, moderate, serious, severe, or

extreme. Areas designated as marginal nonattainment have

three years to attain, while areas with worse designations have

correspondingly longer deadlines. 42 U.S.C. § 7511(a)(1).

5

A worse nonattainment designation triggers more stringent

regulation. For moderate nonattainment areas, SIPs must

undertake to significantly reduce emissions. 42 U.S.C.

§ 7511a(b)(1)(A)(i). And for serious, severe, or extreme

nonattainment areas, SIPs must undertake even more. Id.

§ 7511a(c)–(e).

Failing to achieve attainment by the attainment date also

has consequences. Within six months of that deadline, EPA

must determine whether the area achieved attainment. 42

U.S.C. § 7511(b)(2)(A). In general, an area that missed the

deadline—i.e., failed to timely achieve attainment—must be

“reclassified by operation of law” into a worse nonattainment

status. Id. § 7511(b)(2)(A)(i)–(ii). And in some

circumstances, EPA may sanction a State by taking away

federal highway funds or by imposing further environmental

regulations. Id. § 7509(a)–(b). So as a practical matter, States

with nonattainment areas must “implement potentially

expensive technology or expensive process changes to reduce

pollution levels over a relatively short period of time.” Miss.

Comm’n on Env’t Quality v. EPA, 790 F.3d 138, 146 (D.C. Cir.

2015) (per curiam).

B

Ground-level ozone forms when its precursors, nitrogen

oxides and volatile organic compounds, react with sunlight.

Power plants, motor vehicles, and combustion engines emit the

precursors. Because ozone and its precursors travel easily

through the atmosphere, nonattainment can occur hundreds of

miles away from where the precursors were emitted.

In 2015, EPA reduced the NAAQS for ozone from 0.075

to 0.070 parts per million. National Ambient Air Quality

Standards for Ozone, 80 Fed. Reg. 65,292 (Oct. 26, 2015). For

6

this standard, EPA calculates design values based on three

years of certified data. Id. at 65,294.

At the same time, EPA issued a guidance memo on how to

designate areas under the 2015 ozone standard. The memo

flagged five primary considerations: air quality, emissions,

weather, topography, and jurisdictional boundaries. J.A. 152.

In 2018, EPA promulgated its designations. Additional

Air Quality Designations for the 2015 Ozone National Ambient

Air Quality Standards, 83 Fed. Reg. 25,776 (June 4, 2018).

The agency relied primarily on data from 2014 to 2016, which

was “the most recent data that states were required to certify at

the time the EPA notified the states of its intended

modifications to their recommendations.” Id. at 25,779. The

designations went into effect in August 2018, starting the clock

for nonattainment areas to attain. See 40 C.F.R. § 51.1303(a).

C

In Clean Wisconsin, this Court held that EPA had acted

arbitrarily in designating northern Weld County and El Paso

County as attainment areas.

Weld is a large Colorado county located north of Denver.

EPA concluded that the southern part of Weld County, but not

the northern part, contributed to ozone pollution in the Denver

metropolitan area. So it folded the southern part into a

nonattainment area encompassing greater Denver, and it

designated the northern part as a standalone attainment area.

We were skeptical because northern Weld County produced

emissions that “approached or exceeded those of several entire

counties in the nonattainment area.” 964 F.3d at 1168. And

we found EPA’s analysis of the local weather and topography

to be shallow and inconsistent. Id. at 1169.

7

El Paso County lies in western Texas and borders New

Mexico. In 2018, EPA designated it as an attainment area. But

when the Clean Wisconsin petitioners argued that El Paso

contributed to nonattainment in Doña Ana County, New

Mexico, EPA asked us to remand the designation for further

explanation. We obliged but instructed the agency to revise its

analysis “as expeditiously as practicable.” 964 F.3d at 1176.

We remanded the northern Weld and El Paso designations

without vacating either one. In declining to vacate, we

perceived a “realistic possibility” that EPA would be able to

justify the original designations on remand. 964 F.3d at 1177.

EPA promulgated the revised designations in November

2021. Additional Revised Air Quality Designations for the

2015 Ozone National Ambient Air Quality Standards: El Paso

County, Texas and Weld County, Colorado, 86 Fed. Reg.

67,864 (Nov. 30, 2021) (Final Rule). These designations rest

only on data that was available to EPA when it promulgated its

original designations. Id. at 67,868–69. Based on this data, the

agency concluded that the disputed areas contribute to nearby

nonattainment. So EPA folded northern Weld County into the

greater Denver marginal nonattainment area, and it folded El

Paso County into a marginal nonattainment area that includes

Doña Ana County. Id. at 67,873.

In doing so, EPA declined to recognize new attainment

dates running from the date of the revised designations. Final

Rule, 86 Fed. Reg. at 67,869. Because EPA designated Doña

Ana County as a marginal nonattainment area in August 2018,

its attainment date passed in August 2021—three months

before the Final Rule folded El Paso County into that area. And

because EPA recognized no new deadline, El Paso had no

opportunity to meet its attainment date and thus timely attain.

8

EPA did extend one other deadline: Because Texas “had no

notice that it should prepare a marginal area SIP submission”

for the expanded nonattainment area, EPA gave Texas one

more year to prepare a SIP with planning requirements for a

marginal nonattainment area. Id. EPA later concluded that the

El Paso-Doña Ana County area had not attained the 2015 ozone

standard as of its August 2021 attainment date. Determination

of Attainment by the Attainment Date But for International

Emissions for the 2015 Ozone National Ambient Air Quality

Standard; El Paso-Las Cruces, Texas-New Mexico, 88 Fed.

Reg. 14,095 (Mar. 7, 2023).

Weld County and Texas seek review of the revised

designations. We have jurisdiction under 42 U.S.C.

§ 7607(b)(1).

II

Weld County offers two reasons for why EPA acted

arbitrarily in designating the entire county as marginal

nonattainment. First, EPA failed to consider the most current

available data. Second, the older data do not support the

designation. Neither argument persuades.

A

On remand, EPA faced a choice about what data to use in

considering whether northern Weld County contributes to

nonattainment in Denver. One option was to use only the

certified data, gathered primarily from 2014 to 2016, that EPA

had considered in making the original designations. Another

option was to consider the most recent certified data, which

was gathered from 2015 to early 2021. Weld contends that it

was arbitrary for EPA to use the original data. We disagree.

9

The Clean Air Act allows us to reverse rules that are

“arbitrary, capricious, an abuse of discretion, or otherwise not

in accordance with law.” 42 U.S.C. 7607(d)(9)(a). Under this

familiar standard, copied from the Administrative Procedure

Act, we uphold a rule if the agency “considered all relevant

factors and articulated a rational connection between the facts

found and the choice made.” Miss. Comm’n, 790 F.3d at 150

(cleaned up).

EPA reasonably explained its decision to use only the data

at its disposal while making the original designations. First,

using the same data for localized redesignations would

standardize its analysis and thus facilitate consistent treatment

of all affected counties. J.A. 676. Second, using the original

data would streamline the process and thus comply with our

instruction in Clean Wisconsin to make any redesignations “as

expeditiously as practicable.” 964 F.3d at 1176; see J.A. 677.

Mississippi Commission bolsters EPA’s choice. In that

case, EPA used older data to designate a tristate nonattainment

area despite possessing more recent, certified data from two of

the three states. We declined “to declare irrational the EPA’s

conclusion that comparing data from the same time period

would be more appropriate than analyzing data from different

time periods in the same evaluation process.” 790 F.3d at 160.

So too here. If EPA could choose a matched dataset to classify

a nonattainment area spanning multiple states, then it can also

choose a matched dataset to classify a nonattainment area

spanning multiple counties within a state.

We recognize that an agency generally must base its

decisions on the best available data. But the question here is

whether EPA was required to use one data set (the most recent

certified data) in assessing northern Weld County’s

contribution to ozone pollution in greater Denver even though

10

it had used another data set (the certified data available at the

time of the original designations) in assessing the contribution

of at least eight other counties in the same area. In these

circumstances, EPA plausibly explained why the benefits of a

matched dataset—greater parity among counties and faster

turnaround—make the original data a better choice than partial

updating.

Weld County objects that EPA failed to act consistently.

Weld notes that EPA refused to consider certain air quality data

from 2014 to 2016. But this data was not made available to

EPA until 2020 and 2021, so its exclusion was consistent with

EPA’s overall approach to stay within the record available

when it made the original designations. Weld further notes that

EPA, in making those designations, did consider some data

from as late as 2017. EPA did so to the extent that some States

chose to certify air quality data ahead of schedule. But because

Colorado did not avail itself of this option, EPA based its 2018

designation on Colorado’s certified data from 2014 to 2016.

None of this suggests that EPA acted inconsistently or

otherwise arbitrarily.

B

Weld County further argues that the certified data do not

support including its northern part in the greater Denver

nonattainment area. As noted above, EPA uses five factors to

define areas and determine their attainment status. The last

four factors (emissions, weather, topography, and jurisdictional

boundaries) bear on the appropriate boundaries for a particular

area. The first factor (air quality) bears on its appropriate

designation. If EPA determines that one area contributes to

another’s nonattainment, it will combine those areas into a

single nonattainment area.

11

Weld County does not challenge this overall framework

for making the designations. Instead, it contends that EPA

unreasonably applied the framework to conclude that northern

Weld County contributes to Denver’s nonattainment. We see

no reason to disturb this highly technical judgment.

1. Air Quality. EPA assesses air quality by considering

whether local monitors report NAAQS violations—i.e., a

design value above 0.070 parts per million. The presence of a

single violating monitor justifies a nonattainment designation.

In the Denver metropolitan area, EPA found five of them.

Weld contends that EPA erred by using outlier data

associated with wildfires and stratospheric intrusions. EPA

may disregard data that arises from an “exceptional event.” 40

C.F.R. § 50.14. But to exclude data on this ground, a State

must prove to EPA that an exceptional event “caused a specific

air pollution concentration at a particular air quality monitoring

location.” Id. § 50.14(a)(1)(ii). Colorado made no effort to

link the Denver monitor readings to any exceptional event. To

the contrary, in opposing the Final Rule, Colorado

acknowledged its failure to submit any “exceptional event

demonstrations.” J.A. 631.

2. Emissions. This factor primarily considers the origin

and quantity of precursor emissions. Data showed that Weld

County produces more than three times the emissions of the

next-highest-emitting county in the Denver metropolitan area.

And although northern Weld County accounts for only a small

fraction of the County’s overall emissions, we noted in Clean

Wisconsin that a small fraction of a large number can still be a

large number. 964 F.3d at 1168.

On remand, EPA concluded that northern Weld County

produces significant emissions because (a) the County

12

produces far more emissions than any nearby county, (b) most

of its emissions come from oil-and-gas wells, and (c) nearly

eight percent of the County’s 36,682 wells are in its northern

portion. Furthermore, northern Weld County has three

individual sources that each emit over 100 tons of ozone

precursors per year. In our view, these facts support EPA’s

revised conclusion.

Weld again claims inconsistency. It objects that EPA

failed to reevaluate emissions from the northern part of nearby

Larimer County, which EPA excluded from the Denver

nonattainment area. But no data in the relevant set compares

the emissions of northern Weld and northern Larimer counties.

Instead, Weld flags data comparing the combined emissions of

northern Weld and northern Larimer counties to emissions

from the Denver nonattainment area. We cannot infer from this

that northern Larimer’s emissions exceed northern Weld’s.

Weld continues that northern Larimer County is a stronger

candidate for inclusion in the nonattainment area because it has

a higher population density and more vehicle miles travelled

than does northern Weld County. But as EPA explained,

northern Larimer and northern Weld Counties differ in other

important respects, such as topography, that cut in favor of

designating only northern Weld County. To establish

arbitrariness based on inconsistency, Weld must show that

EPA “treated genuinely similar counties dissimilarly.” Miss.

Comm’n, 790 F.3d at 169 (cleaned up). Given the various

cross-cutting considerations we have noted, Weld has not made

that showing.

3. Weather. EPA considers how meteorological

conditions affect the movement of ozone and its precursors

through the atmosphere. EPA uses a model to determine this

movement from data about wind speed and direction,

13

temperature, humidity, and air pressure. The model predicts

the paths, known as “back trajectories,” traveled by air parcels

that reach a violating monitor. Clean Wisc., 964 F.3d at 1155.

If the model’s projections show air parcels moving from a

region to the violating monitor, that supports including the

region in the nonattainment area. According to EPA, the model

indicates that emissions from northern Weld County move into

the Denver nonattainment area.

Weld contends that EPA ignored certain model projections

suggesting otherwise. But EPA did consider these projections,

and it offered three sound reasons for discounting them: First,

the projections missed all back trajectories from one of the

violating monitors. Second, they missed back trajectories from

2016. Third, they missed back trajectories from each day when

a monitor registered an above-NAAQS design value. In any

event, EPA further explained that even the County’s preferred

data suggest that air flow from northern Weld County “affect[s]

violating monitors” in the Denver area. J.A. 682.

Weld also highlights supposed flaws in EPA’s data. First,

it complains that the data tracks air parcels arriving at a monitor

only during a single hour of the day. But Weld did not raise

this argument below, and we thus need not consider it. In any

event, EPA sensibly focused on the time of day when ozone

concentrations were highest. Plus, data from other times could

only expand the possible source regions; they could not change

the critical fact that some air parcels travel from northern Weld

County to greater Denver. Second, Weld objects that EPA did

not try to predict how terrain near violating monitors might

influence particle trajectory. But the model seeks to measure

the paths traveled by air parcels to a violating monitor; it is not

concerned with how terrain near violating monitors impacts

particle trajectory.

14

4. Topography. In originally excluding northern Weld

County from the Denver nonattainment area, EPA claimed that

its boundary line tracked an elevation called the Cheyenne

Ridge, which assertedly blocks local emissions from reaching

Denver. But in Clean Wisconsin, we found that the ridge was

in the northernmost part of Weld County, along the boundary

between Colorado and Wyoming. See 964 F.3d at 1168.

On remand, EPA agreed with our assessment and then

concluded that the local topography funneled, rather than

impeded, the flow of air from Northern Weld County to greater

Denver. Weld objects that EPA again misplaced the ridgeline,

but it provides no convincing response to the evidence already

credited by this Court.

5. Jurisdictional boundaries. The last main factor that

EPA considers is existing jurisdictional boundaries. All else

equal, administration is easier when area designations track

preexisting boundaries such as county lines. So here, EPA

reasoned, such boundaries support combining northern and

southern Weld County.

The County objects that EPA failed to quantify the

emissions contributions of its northern portion. But we have

allowed EPA to designate nonattainment areas without

isolating and quantifying the exact emissions from each subpart

of a jurisdiction. Catawba Cnty., N.C. v. EPA, 571 F.3d 20, 40

(D.C. Cir. 2009). Weld notes considerations such as its size

and uneven elevation. But we cannot conclude that EPA was

legally compelled to subdivide the County, particularly given

some affirmative evidence that northern Weld County does

contribute to Denver’s nonattainment.

15

III

Texas argues that the Final Rule is impermissibly

retroactive because, in December 2021, it folded El Paso

County into a nonattainment area for which the August 2021

attainment date had already passed. We agree.

A

Agencies cannot promulgate retroactive rules without

express statutory authorization. See, e.g., Bowen v.

Georgetown Univ. Hosp., 488 U.S. 204, 224 (1988) (Scalia, J.,

concurring). And this Court has held that nothing in the Clean

Air Act gives EPA “the unusual ability to implement rules

retroactively.” Sierra Club v. Whitman, 285 F.3d 63, 68 (D.C.

Cir. 2002). So, if the Final Rule operates retroactively as

applied to El Paso, then it cannot stand.

A rule operates retroactively when it “would impair rights

a party possessed when he acted, increase a party’s liability for

past conduct, or impose new duties with respect to transactions

already completed.” Landgraf v. USI Film Prods., 511 U.S.

244, 280 (1994). In other words, “retroactive rules alter the

past legal consequences of past actions.” Arkema, Inc. v. EPA,

618 F.3d 1, 7 (D.C. Cir. 2010) (quoting Bowen v. Georgetown

Univ. Hosp., 488 U.S. 204, 219 (1988) (Scalia, J., concurring))

(cleaned up); see Landgraf, 511 U.S. at 269–70 (“the court

must ask whether the new provision attaches new legal

consequences to events completed before its enactment”). On

the other hand, it is not enough that a rule “upsets expectations

based in prior law,” id. at 269, or “draws upon antecedent facts

for its operation,” id. at n.24 (cleaned up).

16

The Final Rule is impermissibly retroactive. Recall that a

marginal nonattainment designation gives rise to an attainment

date that is three years after the designation. 42 U.S.C.

§ 7511(a). If a State misses the deadline, EPA must reclassify

the designated area to a worse nonattainment status “by

operation of law.” Id. § 7511(b)(2)(A)(i)–(ii). And the

reclassification triggers the various additional burdens that

come with the downgraded status. Id. § 7511a(b)(1). By

design, this scheme provides strong incentives for States with

nonattainment areas to use the three-year runway to achieve

attainment. And 17 of the 36 areas designated as marginal

nonattainment for the 2008 ozone NAAQS did timely attain.

See Determinations of Attainment by the Attainment Date, 81

Fed. Reg. 26,697, 26,700 (May 4, 2016).

Here, Texas never had the requisite opportunity to reach

timely attainment. In August 2018, EPA classified El Paso as

an attainment area. Texas thus had no reason to plan for

improving El Paso’s air quality at that time. Yet in November

2021, EPA folded El Paso into an existing nonattainment

area—three months after that area’s August 2021 attainment

deadline had passed. And despite considering the question at

length, EPA refused to recognize an attainment date for El Paso

running from the date of its new nonattainment designation.

J.A. 661–63. Thus, despite designating El Paso as a

nonattainment area in November 2021, EPA effectively

backdated to August 2018 the start of its three-year runway for

reaching attainment. The Final Rule thereby imposed

liabilities on Texas’s inaction between August 2018 and

August 2021—i.e., it imposed on Texas the consequences of

missing a compliance deadline that passed before the

underlying legal obligation was imposed.

Our caselaw confirms that the Final Rule operates

retroactively. We have made clear that because EPA lacks

17

statutory authority to promulgate retroactive rules, it cannot

impose on States new obligations with compliance deadlines

already in the past. Three decisions prove this point.

Start with Sierra Club v. Whitman. In 1991, EPA

designated St. Louis as a moderate nonattainment area,

triggering a 1996 attainment date. After EPA missed its own

1997 deadline for determining whether St. Louis had timely

attained, the agency refused to backdate to that deadline its

later determination that the city had not timely attained.

Upholding the refusal to backdate, we invoked the principle

that EPA cannot engage in “retroactive rulemaking.” 285 F.3d

at 68. Likewise, we explained that the requested backdating

would have “likely impose[d] large costs on the States, which

would face fines and suits for not implementing air pollution

prevention plans in 1997, even though they were not on notice

at the time” of any legal obligation to do so. Id.

The same logic guided our decision in Sierra Club v. EPA,

356 F.3d 296 (D.C. Cir. 2004). In that case, EPA downgraded

the District of Columbia from serious to severe nonattainment.

But because the deadline for submitting a severe nonattainment

SIP had already passed, EPA gave the District a new deadline

for doing so. We again rejected a contention that the original

deadline should control—which, we said, would make the

reclassification retroactive “by holding the States in default of

their submission obligations before the events necessary to

trigger that obligation (reclassification) occurred.” Id. at 309

(cleaned up).

Last is WildEarth Guardians v. EPA, 830 F.3d 529 (D.C.

Cir. 2016). After we held that EPA had been using the wrong

statutory scheme to regulate fine particulate matter, the agency

adjusted the SIP and attainment deadlines under the correct

scheme “to avoid treating states as having already missed

18

deadlines of which they were never aware.” Id. at 531. We

rejected a contention that EPA should have assumed the correct

framework had been applied all along. In doing so, we

described the adjustments as necessary to avoid imposing

“retroactive consequences on states.” Id. at 540. We also

rejected a proposed distinction between “present findings of

noncompliance” and the sort of “backdated findings” in the

Sierra Club cases. Id. In either instance, we reasoned, “States

would be held to long-passed deadlines of which they were

unaware, with meaningful legal consequences.” Id. at 541.

B

EPA’s responses are unpersuasive. EPA objects that

Texas did not preserve its retroactivity argument below. But in

opposing the proposed Final Rule, Texas could not have been

much clearer. It argued that the El Paso County area “should

not be tied retroactively to implementation deadlines that

existed prior to the area being designated as nonattainment.”

J.A. 614. And it warned that “[a]ny attempt to ‘link’ El Paso

County to the [Doña Ana] nonattainment designation

implementation dates would exceed” EPA’s statutory

authority. J.A. 615. The Texas Association of Manufacturers

echoed these concerns. It objected that EPA lacked statutory

authority to eliminate Texas’s three-year attainment runway

“[b]y retroactively attaching Dona Ana County’s attainment

date to El Paso.” J.A. 411. This comment also preserved the

retroactivity issue. See Ne. Md. Waste Disposal Auth. v. EPA,

358 F.3d 936, 948 n.12 (D.C. Cir. 2004) (“It is sufficient that

an issue was raised by any commenter; the party petitioning for

judicial review need not have done so itself.”).

On the merits, EPA contends that its classification of El

Paso as a nonattainment area imposed only prospective

obligations on Texas—the requirement to submit a new SIP,

19

for which EPA imposed a new deadline, and the future

planning requirements associated with nonattainment status.

But as discussed, EPA refused to set an attainment date keyed

to the November 2021 designation. Texas thus found itself in

the unenviable position of learning in November 2021 that El

Paso County either had to have reached attainment by its

August 2021 attainment date or would suffer the consequences

flowing “by operation of law” from having missed that

deadline. 42 U.S.C. § 7511(b)(2)(A). To be sure, a downgrade

does not happen by itself, but only after EPA determines that

the area has missed the deadline. See id. But EPA’s decision

to backdate El Paso’s nonattainment designation retroactively

adjusted Texas’s legal rights by increasing the State’s exposure

to the harsh consequences that follow from failing to meet an

already past deadline.

We recognize that El Paso may yet avoid the additional

burdens flowing from a downgraded attainment classification.

The Clean Air Act provides an exception to the mandatory

downgrade if a State proves to EPA that the nonattainment area

would have met its deadline “but for emissions emanating from

outside of the United States.” 42 U.S.C. § 7509a(a)(2); see

Implementation of the 2015 National Ambient Air Quality

Standards for Ozone, 83 Fed. Reg. 62,998, 63,009 & n.24 (Dec.

6, 2018). Several months after oral argument, EPA proposed a

rule reflecting its tentative judgment that the El Paso-Doña Ana

area would have timely attained but for emissions from

Mexico. Determination of Attainment by the Attainment Date

But for International Emissions for the 2015 Ozone National

Ambient Air Quality Standard, 88 Fed. Reg. at 14,101. This

proposed rule does not change our analysis. For one thing, it

is not final and thus currently lacks the force of law. In any

event, the nonattainment designation still created substantial

legal exposure for Texas based on its inaction between August

2018 and August 2021. The fact that a distinct affirmative

20

defense might extinguish it does not change the retroactive

character of the rule creating the exposure in the first place.

The intervenors press two further retroactivity points,

which we reject. First, they claim Texas knew all along that El

Paso’s status could change. For support, they invoke Treasure

State Resource Industry Association v. EPA, 805 F.3d 300

(D.C. Cir. 2015), which held that a NAAQS designation is not

impermissibly retroactive just because it relies on old data. Id.

at 305–06; see also Landgraf, 511 U.S. at 269 n.24. But Texas

does not contend that El Paso’s nonattainment designation in

November 2021 was impermissibly retroactive because it was

based on air quality data from earlier years. Instead, Texas

contends that the designation was impermissibly backdated to

August 2018. Moreover, agencies always may prospectively

change their regulations, just as legislatures always may

prospectively amend their statutes. If that possibility were

enough to vitiate retroactivity concerns, the presumption

against retroactive statutes and rules would amount to nothing.

Second, the intervenors object that Texas failed to identify

any different steps it would have taken if EPA had designated

El Paso as a nonattainment area in 2018. But Texas need not

make that showing. As explained above, a rule that “attaches

new legal consequences to events completed before its

enactment” is retroactive, see Landgraf, 511 U.S. at 269–70,

and thus “invalid unless specifically authorized.” Nat’l

Petrochemical & Refiners Ass’n v. EPA, 630 F.3d 145, 159

(D.C. Cir. 2010). On the other hand, a rule that merely “upsets

expectations based in prior law” is not retroactive in the same

way; it is instead only “secondarily retroactive,” and thus

“invalid only if arbitrary and capricious.” Id. (quoting

Landgraf, 511 U.S. at 269); see Nat’l Cable & Telecomm.

Ass’n v. FCC, 567 F.3d 659, 670–71 (D.C. Cir. 2009). In

reviewing a rule with such secondary retroactivity, we must

21

balance the harm of “upsetting prior expectations” against any

benefits of applying the rule “to those preexisting interests.”

Nat’l Cable & Telecomm. Ass’n, 567 F.3d at 670; see also

Bowen, 488 U.S. at 220 (Scalia, J., concurring) (“A rule that

has unreasonable secondary retroactivity—for example,

altering future regulation in a manner that makes worthless

substantial past investment incurred in reliance upon the prior

rule—may for that reason be ‘arbitrary’ or ‘capricious.’”). For

this inquiry, the extent of any reliance or expectation interests

is obviously critical. But the intervenors cite no case

suggesting that a statute or regulation exhibiting primary

retroactivity, by changing the past legal consequences of past

actions, is presumptively valid absent a showing of case-

specific reliance by adversely affected parties.

By backdating El Paso’s 2021 nonattainment designation

to 2018, EPA changed the legal consequences of Texas’s

inaction over that past period. The designation thus exhibited

primary retroactivity—and was invalid for that reason.

IV

Our final task is to determine the appropriate remedy. The

Clean Air Act permits us to “reverse” any EPA “action” found

to be arbitrary. 42 U.S.C. 7607(d)(9). Texas asks us to reverse

the Final Rule itself. But regulations—like statutes—are

presumptively severable: If parts of a regulation are invalid

and other parts are not, we set aside only the invalid parts unless

the remaining ones cannot operate by themselves or unless the

agency manifests an intent for the entire package to rise or fall

together. This is true for agency rules in general, e.g., Finnbin,

LLC v. CPSC, 45 F.4th 127, 136 (D.C. Cir. 2022); Carlson v.

Postal Regul. Comm’n, 938 F.3d 337, 351 (D.C. Cir. 2019),

and for EPA rules in particular, e.g., Virginia v. EPA, 116 F.3d

499, 500–01 (D.C. Cir. 1997); Davis Cnty. Solid Waste Mgmt.

22

v. EPA, 108 F.3d 1454, 1459 (D.C. Cir. 1997). Moreover,

judicial remedies should be “no more burdensome to the

defendant than necessary to provide complete relief” to the

plaintiffs or petitioners. Califano v. Yamasaki, 442 U.S. 682,

702 (1979); see also California v. Texas, 141 S. Ct. 2104, 2115

(2021) (remedies “operate with respect to specific parties”

rather than “on legal rules in the abstract”) (cleaned up).

Under these standards, the revised Weld County and El

Paso designations are clearly severable. They adjust the

geographic boundaries, and thereby the attainment status, of

areas hundreds of miles apart. Each revised designation

functions perfectly well on its own, and we have no reason to

think that EPA would want both the revised designations to fall

simply because one of them is invalid. For these reasons, we

decline to disturb the Weld County designation.

A distinct severability question relates to the El Paso

nonattainment designation. As we have explained, its

impermissible retroactivity arises not from the designation

itself, but from the designation combined with EPA’s refusal

to recognize a new attainment date. We could cure the legal

violation by reversing either decision. But since EPA has

strenuously argued that a new attainment date would create

both fairness and administrability concerns, J.A. 662–64, we

are reluctant to force that option on EPA. Instead, we think it

more prudent simply to reverse the nonattainment designation,

leaving EPA free on remand to decide whether to make a new

designation with its own attainment date or simply to let well

enough alone.

V

For these reasons, we deny Weld County’s petition for

review, grant Texas’s petition for review, and reverse the Final

23

Rule insofar as it designates El Paso County to be a marginal

nonattainment area.

So ordered.