United States Sugar Corporation v. EPA

113 F.4th 984

• Court: Court of Appeals for the D.C. Circuit
• Decided: September 3, 2024
• Docket: 22-1271
• Status: Published

Opinion

United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued March 21, 2024 Decided September 3, 2024

No. 22-1271

UNITED STATES SUGAR CORPORATION, PETITIONER

v.

ENVIRONMENTAL PROTECTION AGENCY, RESPONDENT

SIERRA CLUB, INTERVENOR

Consolidated with 22-1302, 22-1303

On Petitions for Review of a Final Administrative Action of the Environmental Protection Agency

Timothy S. Bishop argued the cause for Industry Petitioners. With him on the briefs were Shannon S. Broome and Charles H. Knauss. Avi Kupfer entered an appearance.

James S. Pew argued the cause and filed the briefs for Environmental Petitioners. 2

Perry M. Rosen, Attorney, U.S. Department of Justice, argued the cause for respondent. With him on the brief were Todd Kim, Assistant Attorney General, and Lucas May, Attorney, U.S. Environmental Protection Agency.

James S. Pew argued the cause and filed the brief for respondent-intervenor Sierra Club.

Charles H. Knauss, Shannon S. Broome, Elliott Zenick, Kevin M. Dempsey, and Lisa M. Jaeger were on the brief for intervenor-respondents American Forest and Paper Association, et al.

Before: WILKINS, KATSAS, and WALKER, Circuit Judges.

Opinion for the Court filed PER CURIAM.

The Clean Air Act directs the Environmental Protection Agency to establish emission standards for new and existing sources of hazardous air pollutants. See 42 U.S.C. §§ 7401, 7412; Michigan v. EPA, 576 U.S. 743, 747 (2015).

The Act says a source is “new” if it is built after EPA proposes an applicable emission standard for that source. 42 U.S.C. § 7412(a)(4). A source is “existing” if it is built before then. Id. § 7412(a)(10). This distinction matters because the required standards for new sources are generally stricter than the required standards for existing sources. See id. § 7412(d)(3).

In a 2022 rule, EPA classified some industrial boilers as “new” sources of hazardous air pollutants even though they were built before the applicable emission standards were proposed in 2020. See National Emission Standards for 3

Hazardous Air Pollutants for Major Sources: Industrial, Commercial, and Institutional Boilers and Process Heaters, 87 Fed. Reg. 60816, 60830, 60841 (Oct. 6, 2022). Because that classification conflicts with the Clean Air Act’s definitions of “new” and “existing” sources, we grant the petitions brought by the U.S. Sugar Corporation and trade groups representing other operators of industrial boilers.

In the same 2022 rule, EPA drew its data from the same 2013-era dataset it had used for other, still-valid emission standards promulgated for industrial boilers back in 2013. See 87 Fed. Reg. at 60820–21. It intentionally excluded more recent data because it wanted to regulate similar sources similarly, so these new 2022 standards would be consistent with the still-valid 2013 standards. Id. Because that decision did not violate the Clean Air Act, we deny the petition brought by four environmental organizations.

I

Boilers burn materials like coal, paper, and agricultural waste to create heat, electricity, and other forms of energy. U.S. Sugar Corp. v. EPA, 830 F.3d 579, 592, 597 (D.C. Cir. 2016) (per curiam) (“U.S. Sugar Corp. I”). In doing so, they emit hazardous air pollutants like mercury, carbon monoxide, hydrogen chloride, and particulate matter. Those hazardous air pollutants can cause adverse health effects. Id.

The emission of hazardous air pollutants by boilers is regulated by EPA at the direction of the Clean Air Act. See generally 42 U.S.C. § 7412. 4

A

Section 112 of the Clean Air Act requires EPA to create technology-based emission standards for stationary sources of hazardous air pollutants. See generally id. Because these standards generally require sources to be at least as clean as their peers—and so must often use technology that is the same as, or better than, the technology of comparable sources—these emission limits are called the “maximum achievable control technology” or “MACT.” See U.S. Sugar Corp. I, 830 F.3d at 594 (cleaned up). See also id. (Congress wanted sources to pollute no worse than “their best performing peers”) (cleaned up).

To set MACT standards, “EPA’s first task is to create … categories and subcategories” of sources. Id. at 593. Once EPA has created those categories and subcategories, the Clean Air Act requires that it draw “one final dividing line—between ‘new’ sources and ‘existing’ sources.” Id. at 594 (citing 42 U.S.C. § 7412(d)(3)). A “new” source is “a stationary source the construction or reconstruction of which is commenced after the Administrator first proposes regulations under this section establishing an emission standard applicable to such source.” 42 U.S.C. § 7412(a)(4). An “existing” source is “any stationary source other than a new source.” Id. § 7412(a)(10). 1

1 The Clean Air Act also separates boilers based on the amount of hazardous air pollutants they emit. A boiler is a “major” boiler if it annually emits (a) more than 10 tons of any single hazardous air pollutant, or (b) more than 25 tons of any combination of hazardous pollutants. 42 U.S.C. § 7412(a)(1). If a boiler does neither, it is an “area” boiler. Id. at § 7412(a)(2). This case concerns only major boilers. 5

Once EPA has identified categories and subcategories and drawn the necessary distinctions between new and existing sources, EPA must set emission standards applicable to each category and subcategory. See id. § 7412(c)(2), (d)(3). A “new” source in a given category or subcategory must meet a standard at least as strict as “the emission control that is achieved in practice by the best controlled similar source.” Id. § 7412(d)(3). For an “existing” source in a given category or subcategory, the MACT standard must be at least as stringent as “the average emission limitation achieved by the best performing 12 percent of the existing sources (for which the Administrator has emissions information).” Id. 2 § 7412(d)(3)(A). Though this statutory text is somewhat technical, it means at least this—the standard for a “new” source is stricter than the standard for an “existing” source.

In addition, new and existing sources have different deadlines for compliance. See id. § 7412(i). An existing source has three years to comply with a stricter new standard for existing sources. See id. § 7412(i)(3). In contrast, a new source must immediately comply with a stricter new standard for new sources, unless EPA promulgates the standard while the new source is under construction and the promulgated standard is more onerous than the proposed one. See id. § 7412(i)(1), (2). In that case, the new source has three years to comply. See id. § 7412(i)(2).

2 Or, if the subcategory is so small that it has fewer than thirty sources, the standard for existing sources must be at least as stringent as “the average emission limitation achieved by the best performing 5 sources (for which the Administrator has or could reasonably obtain emissions information).” Id. § 7412(d)(3)(B). 6

B

EPA promulgated MACT emission standards for boilers in 2004. See 69 Fed. Reg. 55218 (Sept. 13, 2004). But for reasons that are not relevant to today’s case, this Court vacated EPA’s rule. See NRDC v. EPA, 489 F.3d 1250, 1257, 1261 (D.C. Cir. 2007).

In 2010, EPA tried again. It proposed new standards for boilers, based on data drawn from a new dataset. Then, in 2011, it promulgated a final rule. See 76 Fed. Reg. 15608 (Mar. 21, 2011). This rule, as amended on reconsideration in subsequent years, resulted in 202 emission standards for new and existing major boilers.

Industry and environmental groups challenged the new standards on a variety of grounds. See U.S. Sugar Corp. I, 830 F.3d at 591. Most of those challenges failed. See id. at 667. But, as relevant here, one succeeded—we held that when EPA based new standards on the practices and technologies of existing boilers, EPA had sometimes excluded certain boilers in violation of the law. See id. at 631–32. That mistake did not affect most of the 202 standards, but it did render 34 of the standards invalid.

In U.S. Sugar Corp. I, the Court vacated the invalid standards. See id. at 667. But on rehearing, we reconsidered that remedy and instead remanded without vacatur, thus allowing the invalid standards from the 2011 Rule to remain in place while EPA “revise[d]” the invalid standards “consistent with … [the] opinion in [that] case.” U.S. Sugar Corp. v. EPA, 844 F.3d 268, 270 (D.C. Cir. 2016) (per curiam) (“U.S. Sugar Corp. II”). We “expect[ed] the EPA to complete this rulemaking promptly.” Id. 7

C

In response to U.S. Sugar Corp. I and U.S. Sugar Corp. II, EPA proposed a new rule on August 24, 2020, and promulgated it as a final rule in 2022. See 87 Fed. Reg. at 60819–21. The rule replaced the 34 invalid MACT standards. See id. at 60817. It included two decisions that today’s petitioners challenge. 3

First, the Industry Petitioners. Leading a group of industry trade groups, the U.S. Sugar Corporation challenges the 2022 Rule’s definition of “new” boilers. In the 2022 Rule, EPA classified boilers built after June 4, 2010 as “new” boilers— even if the applicable standards for those boilers were not proposed until 2020. 4 Id. at 60830.

One such boiler is Boiler No. 9 at a sugar facility owned by the U.S. Sugar Corporation and located in Clewiston, Florida. By burning bagasse—the pulp that remains after juice is extracted from sugarcane—this boiler creates steam and heat to help power the facility.

The U.S. Sugar Corporation began building Boiler No. 9 in 2016 at a cost of $65 million to replace three older and higher-polluting boilers. Upon its completion in 2019, it complied with the strictest standards in effect at the time—the

3 There is an additional argument raised by the Industry Petitioners regarding the new-source MACT standard for Hydrogen Chloride. But we decline to address that argument because it is not necessary to our resolution of the petitions. See infra Section II. 4 The Industry Petitioners have standing because the challenged rule imposes direct costs on their businesses. See Chamber of Com. of U.S. v. EPA, 642 F.3d 192, 199–200 (D.C. Cir. 2011). 8

standards for “new” boilers that were enacted by EPA in 2011. The Industry Petitioners say that because Boiler No. 9 was constructed “with state-of-the-art control technology,” it is “the best-controlled bagasse-fueled boiler in the country.” Industry Petitioners Br. at 10.

When EPA proposed 34 revised standards in 2020, Boiler No. 9 already surpassed the revised standards for “existing” boilers.

But EPA’s redefinition of “new” boilers treats Boiler No. 9 as a “new” boiler, subject to the standards for “new” boilers proposed in 2020—even though construction on Boiler No. 9 began and ended before 2020. Under this regime—whose logic suggests that boilers built after June 4, 2010 are forever “new”—the U.S. Sugar Corporation must spend tens of millions of dollars retrofitting Boiler No. 9. 5

Second, the Environmental Petitioners. When EPA calculated the 34 new MACT standards proposed in 2020, EPA decided not to use additional data it had collected after 2013. See 87 Fed. Reg. at 60820–22. Instead, it relied on data from the 2013-era dataset, which still serves as the basis for the 168 standards that were left undisturbed by U.S. Sugar Corp. I and U.S. Sugar Corp. II. See id. EPA believed this approach would treat similar sources similarly and ensure consistency across all 202 standards, regardless of when the standards were proposed.

5 The other Industry Petitioners—the American Forest & Paper Association, the American Wood Council, and the Council of Industrial Boiler Owners—represent members who also built boilers between 2010 and 2020. These boilers also already satisfy the revised standards if they are treated as “existing” boilers, but will also require expensive retrofitting if treated as “new” boilers. 9

See id. According to EPA, that approach avoided a “potentially inequitable outcome,” conserved agency resources, and demonstrated fidelity to the “limited nature” of this Court’s remand in U.S. Sugar Corp. II. Id. at 60822.

EPA’s exclusion of post-2013 data made some of the 34 new standards less strict than they might have otherwise been had EPA been relying on data from 2020. In its petition for review, a group of environmentalists says the exclusion of this newer data both “contravenes the statute” and was “arbitrary and capricious.” 6 Environmentalists Br. at 16–17.

* * *

In 2023, the U.S. Sugar Corporation filed a motion to stay enforcement of the 2022 Rule with regard to Boiler No. 9. A special panel of this Court granted the motion because it “satisfied the stringent requirements for a stay pending court review,” see Order, U.S. Sugar Corp. v. EPA, No. 22-1271, at *1–2 (March 10, 2023), including “a strong showing that” the U.S. Sugar Corporation was “likely to succeed on the merits,” Nken v. Holder, 556 U.S. 418, 434 (2009).

For the reasons explained below, we now grant the Industry Petitioners’ petitions and deny the Environmental Petitioners’ petition. 6 The Environmental Petitioners are the California Communities Against Toxics, the Coalition For A Safe Environment, the Sierra Club, and the Utah Physicians for a Healthy Environment. Under Supreme Court precedent, these groups have standing because the challenged rule exposes their members to higher levels of pollution than their interpretation of the Clean Air Act allows. See Friends of the Earth v. Laidlaw Env’t. Servs., 528 U.S. 167, 180–83 (2000). 10

II

The Industry Petitioners present an issue of statutory interpretation that we review de novo. See Loper Bright Enters. v. Raimondo, 144 S. Ct. 2244, 2261 & n.4, 2273 (2024). Instead of deferring to EPA’s interpretation of the Clean Air Act, we must apply what we regard as the statute’s “best” reading. See id. at 2266. 7

Before interpreting the statute, however, we begin with EPA’s argument that the Industry Petitioners’ challenge to its interpretation of the new-source definition is untimely. In 2011, EPA adopted regulations defining a “new source” in the boiler context as ones built after June 4, 2010. According to EPA, that was the time for challenging its view of which boilers are “new.”

EPA is mistaken. Regardless of whether the Industry Petitioners could and should have challenged the 2011 regulation, the “reopening” doctrine applies here and allows us to hear their current challenge. See Pub. Emps. for Env’t Resp. v. EPA, 77 F.4th 899, 911 (D.C. Cir. 2023). That doctrine provides that when an agency explicitly or implicitly revisits a 7 Loper Bright held that the Administrative Procedure Act requires courts to construe statutes de novo, without deference to the views of agencies entrusted to administer the statutes. See 144 S. Ct. at 2261 & n.4. That analysis controls EPA interpretations of the Clean Air Act reviewed under its judicial-review provision, 42 U.S.C. § 7607(d)(9), because judicial review under the Clean Air Act is “essentially the same” as judicial review under the APA, see Ethyl Corp. v. EPA, 51 F.3d 1053, 1064 (D.C. Cir. 1995). 11

prior position, “its renewed adherence is substantively reviewable.” Id. (cleaned up). In applying the doctrine, we examine the “entire context” of the rulemaking, including “relevant proposals and reactions of the agency,” to determine whether the agency undertook a “serious, substantive reconsideration” of an existing rule. Growth Energy v. EPA, 5 F.4th 1, 21 (D.C. Cir. 2021) (per curiam) (cleaned up).

Here, EPA clearly took a serious look at what date to use when classifying boilers as “new” or “existing.” It spent over a full page in the Federal Register defending its decision to retain the June 4, 2010 cutoff date. See 87 Fed. Reg. at 60830– 31. EPA first explained its view that the Clean Air Act and existing regulations permitted use of that date in this rulemaking. Id. at 60830. Then, EPA detailed its policy rationale for using that date, including the fact that the emission standards here were based on the same dataset as those used to support the 2011 regulation. Id. at 60830–31. The dataset justification was not (and could not have been) given when EPA promulgated the 2011 regulation, and its choice to provide a new rationale for the old date confirms its serious reconsideration of the question. Given that reconsideration, the Industry Petitioners’ challenge can go forward. 8

8 EPA further contends that the Industry Petitioners waived their argument by not citing 40 C.F.R. § 63.7490(b), which sets forth the June 4, 2010 cutoff date for classifying boilers as new or existing. But the entire thrust of their argument was that EPA had to update its definition of what counts as a “new source.” Indeed, the Industry Petitioners’ first argument heading asserts that “the 2022 rule unlawfully applies revised new source standards to boilers that commenced construction before the rule was first proposed in 2020.” Industry Pet. Br. at 19 (cleaned up). The Industry Petitioners were clearly arguing that the existing cutoff date was no longer valid, so 12

On the merits, the question presented by the Industry Petitioners is whether Section 112 of the Clean Air Act allowed EPA, in promulgating the 2022 emission standards for boilers, to categorize all boilers constructed after June 4, 2010 as new sources of pollution. Recall that the minimum emission standard for “new” boilers “shall not be less stringent than the emission control that is achieved in practice by the best controlled similar source.” 42 U.S.C. § 7412(d)(3). In simple terms, all “new” boilers must meet an exacting standard pegged to the cleanest similar model on the market. On the other hand, “existing” boilers need only meet the “average emission limitation achieved by the best performing 12 percent of the existing sources.” Id. § 7412(d)(3)(A). In sum, new boilers must satisfy more stringent emissions requirements than existing ones.

Section 112 defines a “new source” as “a stationary source the construction or reconstruction of which is commenced after the Administrator first proposes regulations under this section establishing an emission standard applicable to such source.” Id. § 7412(a)(4). EPA concluded that the “regulations” under Section 112 “establishing an emission standard” were “first propose[d]” in the 2011 rulemaking, id., so any boiler constructed after June 4, 2010 meets the definition of a “new source,” 87 Fed. Reg. at 60830; EPA Br. at 25. The Industry Petitioners counter that “an emission standard” is “first propose[d]” when EPA first proposes each consecutive standard, not when EPA first proposes any emission standard

EPA cannot claim any unfair “surprise.” See NRDC v. EPA, 25 F.3d 1063, 1071–72 n.4 (D.C. Cir. 1994). 13

for entire categories of sources. 42 U.S.C. § 7412(a)(4); Industry Pet. Br. at 19–20. On this view, the proper date to determine whether a boiler is “new” is August 24, 2020, when this rulemaking proposed new emission standards for boilers.

Despite their differences, the parties agree on one key point: Each thinks that the other’s reading of the statute is semantically plausible. As EPA explained, the new-source definition “could refer to … the first time the agency proposes any standards for the source category … or the first time the Agency proposes a particular standard.” EPA Br. at 29 (cleaned up). The agency had good reason to admit as much because it repeatedly has adopted the Industry Petitioners’ proposed interpretation in setting Section 112 emission standards for sources other than boilers. See, e.g., National Emission Standards for Coke Oven Batteries, 70 Fed. Reg. 19992, 20009–10 (Apr. 15, 2005) (defining “new sources” as “those constructed after the date of proposal” of regulatory “amendments”); National Emission Standards for Hazardous Air Pollutants: Carbon Black Production and Cyanide Chemicals Manufacturing Residual Risk and Technology Reviews, and Carbon Black Production Area Source Technology Review, 86 Fed. Reg. 66096, 66102–03 (Nov. 19, 2021) (similar).

We agree that the phrase “after the Administrator first proposes … an emission standard,” read in a vacuum, can be interpreted either way. This provision is indeterminant because it uses the indefinite article “an,” which refers to a standard that is “unidentified or not immediately identifiable.” See Indefinite, Webster’s Third New International Dictionary: Unabridged 1147 (1993). As the Supreme Court just reiterated, Congress’s choice between a definite and indefinite article matters when determining statutory meaning. See 14

Corner Post, Inc. v. Bd. of Governors of Fed. Res. Sys., 144 S. Ct. 2440, 2455 (2024).

It is true, as EPA points out, that the standard must be the one “first propose[d],” but that requirement does not tell us whether the provision means “first propose[d]” for the original emission standard or for the updated one. 42 U.S.C. § 7412(a)(4). Given this semantic ambiguity, we turn to “the remainder of the statutory scheme” to determine who has the better reading of the new-source definition. United Sav. Ass’n of Tex. v. Timbers of Inwood Forest Assocs., 484 U.S. 365, 371 (1988); see also Pulsifer v. United States, 144 S. Ct. 718, 731– 35 (2024).

Statutory structure and context persuade us that the Industry Petitioners’ interpretation is correct. When Section 112 references the date “an emission standard” is “first propose[d],” it means the first proposal of each consecutive standard. 42 U.S.C. § 7412(a)(4). This reading allows all parts of Section 112 to function as a “harmonious whole,” Roberts v. Sea-Land Servs., Inc., 566 U.S. 93, 100 (2012) (cleaned up), whereas EPA’s contrary reading does not.

Start with how the definition of “new source” works with other provisions of Section 112. Subsection (i) establishes compliance deadlines that make sense only on the Industry Petitioners’ proposed reading. It requires all new sources to comply with emission standards immediately upon their “effective date,” 42 U.S.C. § 7412(i)(3)(A), which is the date when the standard is “promulgat[ed],” id. § 7412(d)(10). In contrast, existing sources have up to three years to comply with their lower emission standards. Id. § 7412(i)(3)(A). As EPA 15

explained in a different Section 112 rulemaking, “new sources know from the beginning of the construction effort what [emission] controls will be required, and do not have to incur the higher costs and the time-consuming disruptions normally associated with control retrofits.” National Emission Standards for Hazardous Air Pollutants from the Portland Cement Manufacturing Industry, 71 Fed. Reg. 76518, 76541 & n.16 (Dec. 20, 2006) (Portland Cement Rule).

If the proper date to determine whether a source is “new” is when EPA first proposed the specific “emission standards” under review, the statute’s pieces work in harmony. Sources built before the amended standards were first proposed must meet less-stringent standards tied to the emissions from other already-operating sources, and they have some time to retrofit their boilers. 42 U.S.C. § 7412(d)(3)(A), (i)(3)(A). This makes sense because it is more difficult to retrofit old boilers to meet modern, state-of-the-art standards than it is to construct new boilers to meet those standards from the beginning. See Portland Cement Rule, 71 Fed. Reg. at 76541 & n.16. EPA itself has explained that retrofitting older sources to comply with increasingly stringent modern standards may be “draconian” if not “impossible.” Id. And we should not lightly assume that a statute is “draconian,” Snyder v. United States, 144 S. Ct. 1947, 1957 (2024), or “demands the impossible,” Anniston Mfg. Co. v. Davis, 301 U.S. 337, 350–52 (1937).

A special statutory compliance rule confirms that already- constructed boilers need not meet the new-source standards. Under the special rule, if construction of a source begins after an emission standard is proposed but before it is finalized, and if the final rule requires a “more stringent” standard than the proposed one, a new source need not comply with the final rule “until the date 3 years after the date of promulgation.” 42 16

U.S.C. § 7412(i)(2). In essence, if a company starts building a source in reliance on a proposed standard and then EPA makes the actual standard stricter, the company gets a grace period to retrofit the source, just like existing sources get. Id.

EPA’s interpretation of the new-source definition would render the “special rule” almost nonsensical. Companies that build emission sources right after a new standard has been proposed would be entitled to a grace period to retrofit their sources if the final standard turns out to be stricter than the proposed one, but companies that built sources much earlier would be required to comply immediately with the updated standard. EPA does not explain why a statute that takes pains to give regulated parties enough time to meet changing emission standards would create this type of inconsistency.

The rule under review attempts to make up for the harshness of treating decade-old boilers as “new source[s]” by granting them a three-year grace period to comply with the new-source emission standards. See 87 Fed. Reg. at 60832. But Section 112(d) states that “[e]mission standards or other regulations promulgated under this subsection shall be effective upon promulgation.” 42 U.S.C. § 7412(d)(10). Section 112(i) prohibits any person from operating a “new” emissions source “in violation of” an emission standard after its “effective date.” Id. § 7412(i)(3)(A). And it further permits EPA to establish a delayed “compliance date” for any existing- source emission standard, which may fall up to “3 years after the effective date of such standard.” Id. In sum, “new” sources must comply with emission standards immediately, but “existing” sources may be afforded a three-year grace period to bring themselves into compliance. In other words, EPA adopted a sweepingly broad understanding of what counts definitionally as a “new” source, then proceeded to treat the 17

sources at issue here similar to “existing” ones for purposes of delayed compliance dates. This need to mix-and-match in addressing different parts of the statute casts further doubt upon EPA’s broad interpretation of the definition itself. See Util. Air Regul. Grp. v. EPA, 573 U.S. 302, 319–20 (2014).

Classifying boilers constructed after June 4, 2010 as eternally “new” also does not accord with the Clean Air Act’s iterative process for regulating pollution sources. Section 112 requires EPA to reevaluate existing emission standards—and promulgate new ones if facts on the ground have changed— every eight years, as technology allows for cleaner-burning sources. 42 U.S.C. § 7412(d)(6); see Nat’l Ass’n for Surface Finishing v. EPA, 795 F.3d 1, 4 (D.C. Cir. 2015). Yet EPA’s interpretation of “new source” would short-circuit this process by locking in which sources are “new” for all time, rather than having the definition apply to each discrete rulemaking. For example, in the year 2034, a boiler constructed in 2011 would be considered “new” and would have to comply with standards for the “best controlled” 2034 model. 42 U.S.C. § 7412(d)(3). Needless to say, a two-decade-old boiler is not “new” within any normal understanding of that term, and “it is not unusual to consider the ordinary meaning of a defined term” when interpreting an ambiguity in a definition. Bond v. United States, 572 U.S. 844, 861 (2014).

Finally, EPA’s interpretation of the “new source” definition is internally inconsistent. Under that interpretation, the proper time to determine whether a boiler is “new” should have been 2003, not 2010, because that was when EPA first proposed emission standards for boilers. See NRDC, 489 F.3d at 1261. EPA resists this conclusion because this Court vacated the 2003 standards. See id. at 1262. But Section 112 references the time when standards are “first propose[d],” and the fact that 18

a final, promulgated standard was later vacated does not change that date. 42 U.S.C. § 7412(a)(4). EPA’s own logic thus would require setting the date in 2003, which even EPA consistently has refused to do.

D

We are unmoved by EPA’s incantation that “first proposed” means “first proposed.” As we have explained, the statutory reference to “an emission standard” (emphasis added) does not answer the question which emission standard— original or amended—the definition is referring to. And in context, “first” references the first time EPA proposes each emission standard for the source at issue. Id. Each regulation “establish[es]” a new “emission standard applicable to such source,” so “an emission standard” refers to each of the increasingly stringent standards that EPA must propose over time. See id.

Under this interpretation, the word “first” still plays a significant role in the statutory scheme. For example, if EPA proposes an emission standard and then decides, in response to comments, to revise the proposal and reopen the comment period, the date demarcating “new” and “existing” sources would still be the date the rule was first proposed, not the date of the re-proposal. Id. This is no fanciful possibility, as EPA sometimes “reproposes” regulations after an original proposal generates significant criticism. Ethyl Corp. v. EPA, 541 F.2d 1, 49 (D.C. Cir. 1976) (en banc). In that circumstance, any re- proposal containing the same standard—despite generating a new comment period and delaying the promulgation of a final standard—would not push back the date on which the standard was “first proposed.” In contrast, if the definition of “new source” omitted the word “first,” as it did before the 1990 19

amendments, then the date of the most-recent proposal would likely control the distinction between new and existing sources.

EPA invokes the definition of “new source” in Section 111 of the Clean Air Act, which is keyed to when regulations “prescribing a standard of performance under this section” were “proposed.” 42 U.S.C. § 7411(a)(2). Invoking the meaningful-variation canon, see, e.g., Southwest Airlines Co. v. Saxon, 596 U.S. 450, 457–58 (2022), EPA contends that “proposed” under Section 111 cannot mean the same thing as “first proposed” under Section 112. Fair enough, at least if we assume that the respective definitions are similar enough to trigger application of the canon. But as shown in the example above, the Industry Petitioners’ interpretation of Section 112 does not collapse “first proposed” into “proposed,” so neither the meaningful-variation canon, nor the related consideration of avoiding surplusage, cut in favor of EPA here.

E

EPA offers an alternative justification for labeling as “new” boilers constructed over a decade ago. According to EPA, because the standards under review here rely on the same dataset as did the 2013 standards, the date to determine whether boilers are “new” or “existing” should still be June 4, 2010.

Section 112 provides no support for the proposition that the vintage of the dataset is relevant to determining whether a source is “new” or “existing.” Under that provision, the critical date turns on when an “emission standard” is “first propose[d]”—not when supporting data for the standard was generated. 42 U.S.C. § 7412(a)(4). EPA points to no textual source for its argument. Nor is it clear how EPA could do so. An “emission standard” is a “requirement” placed on sources 20

to “limit[] the[ir] quantity, rate, or concentration of emissions of air pollutants.” Id. § 7602(k). A dataset used to calculate the standard is not a standard itself. Under Section 112, when the dataset was compiled simply has no bearing on when the emission standard was “first propose[d].” Id. § 7412(a)(4).

Congress’s choice not to tie the definitions of “new” and “existing” sources to the underlying dataset makes sense given the extent of EPA’s discretion to fashion standards out of underlying data. In particular, EPA could drastically increase the stringency of a standard based on the same underlying data. This case proves the point: Although EPA used the same data here as in the 2013 reconsideration, the HCl emission standard became 100 times more stringent based on EPA’s choice to designate a different boiler as the best-performing source. 87 Fed. Reg. at 60823.

Despite the semantic possibility that “after the Administrator first proposes … an emission standard” references when the original emission standard was first proposed, we conclude that this clause refers to when each sequential emission standard was first proposed. The structure of the Clean Air Act makes clear that boilers constructed before each individual standard was first proposed are “existing,” and boilers constructed after each individual standard was first proposed are “new.” We therefore set aside EPA’s 2022 Rule to the extent that it defines sources constructed or reconstructed before August 24, 2020—that is, the date the 2022 Rule was proposed by EPA—as “new source[s].” 9

9 The Industry Petitioners also ask us to rule that EPA arbitrarily set the HCl limit for new sources. Once their boilers are properly 21

III

We turn next to the Environmental Petitioners’ challenges to the 2022 Rule. Under the Clean Air Act, this Court “may reverse” the EPA’s 2022 Rule if it is “in excess of statutory jurisdiction, authority, or limitations,” or “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 42 U.S.C. § 7607(d)(9). To determine “whether an agency has acted within its statutory authority,” we use “the traditional tools of statutory construction.” Loper Bright Enters., 144 S. Ct. at 2268, 2273.

As we have explained, when EPA recalculated the emission standards in the 2022 Rule in response to this Court’s remand, EPA chose to rely upon the same dataset that it had used to calculate the emission standards in the 2013 Rule. The Environmental Petitioners contend that EPA’s choice to use that original dataset, rather than updated data EPA had in its possession, was contrary to law. The Environmental Petitioners also argue that even if EPA’s decision not to use updated data was not unlawful, it was arbitrary and capricious.

We disagree. We hold that EPA’s decision to rely upon the same dataset it used to calculate the emission standards in the 2011 and 2013 Rules was neither unlawful nor arbitrary or capricious. We accordingly deny the Environmental Petitioners’ petition for review.

classified, however, they will no longer be subjected to the new- source HCl limit. Therefore, we decline to reach that issue. 22

The Environmental Petitioners contend that EPA’s decision to rely on the original dataset was unlawful because, under Environmental Petitioners’ reading of Section 112(d), Section 112 requires EPA to use any available data it has in making floor calculations. The Environmental Petitioners rely on two portions of the statutory text in support of their argument that EPA’s decision to rely on its original dataset violated Section 112(d). First, Section 112(d)(3)(A) provides that emission standards for existing sources cannot be less stringent than “the average emission limitation achieved by the best performing 12 percent of the existing sources (for which the Administrator has emissions information).” 42 U.S.C. § 7412(3)(A) (emphasis added). Second, and similarly, Section 112(d)(3)(B) provides that emission standards for existing source subcategories with fewer than thirty sources cannot be less stringent than “the average emission limitation achieved by the best performing 5 sources (for which the Administrator has or could reasonably obtain emissions information).” Id. § 7412(d)(3)(B) (emphasis added). The Environmental Petitioners argue that Section 112(d)(3)’s references to the information the Administrator has in its possession or could reasonably obtain, see id. § 7412(d)(3)(A), (B), requires EPA to use whatever information it possesses when calculating emission standards. EPA is not, in other words, “free to exclude from its floor calculations emissions information that it has in its possession.” Environmental Pet. Br. 25. The Environmental Petitioners explain that Section 112’s statement that existing source standards must be based on the sources “for which the Administrator has emissions information” effectively identifies the group of sources EPA must consider when setting a standard for a given source category or subcategory. So, when EPA “has” information for 23

a given source in a category, Section 112 requires it to consider the available information for that source when setting the relevant category’s emission standards.

The Environmental Petitioners thus argue that EPA ran afoul of Section 112 by choosing to ignore the post-2013 emissions data in its possession when it promulgated the standards in the 2022 Rule. Because EPA chose not to use that post-2013 emissions data in calculating emission standards, the Environmental Petitioners contend that its standards do not reflect the “average emission limitation achieved by the best performing 12 percent of the existing sources (for which the Administrator has emissions information).” 42 U.S.C. § 7412(d)(3)(A); see also id. § 7412(d)(3)(B).

EPA responds that the Environmental Petitioners have misunderstood the statutory text. Section 112’s language that emission standards must be based on the performance of the sources “for which the Administrator has emissions information” does not require EPA to take into account all information that it literally “has” in its possession. Rather, the statutory language’s location in a parenthetical clause shows that EPA has no obligation to “obtain emission data from 100 percent of the source category or subcategory in order to identify the best performing 12 percent” of sources. 87 Fed. Reg. at 60821; see Becerra v. Empire Health Found., 597 U.S. 424, 440 (2022) (“[A] parenthetical is typically used to convey an aside or afterthought.”) (cleaned up). In effect, then, the statutory language does not constrain EPA—to the contrary, it allows EPA to “prevent delay” in promulgating emission standards by moving forward with the emission standard- setting process, even where it does not have emissions data from every single source in a category. 87 Fed. Reg. at 60821. 24

Whether and to what extent Section 112(d)(3) imposes restrictions on what data EPA must use when calculating emissions floors is a tricky question, but we think the Environmental Petitioners’ suggested answer to that question cannot possibly be correct. As we have explained, under the Environmental Petitioners’ expansive interpretation of the statutory text, whenever EPA “has” emissions information for a source, see 42 U.S.C. § 7412(d)(3)(A), (B) (or can reasonably obtain emissions information, see id. § 7412(d)(3)(B)), it must use that information in its calculations. Reading the statutory text this way gives rise to a natural question: At what point in the rulemaking process must EPA ensure that it has used the data it “has” in its possession? At oral argument, counsel for the Environmental Petitioners clarified that, in Environmental Petitioners’ view, EPA must use the data it “has” in its possession up until the promulgation of the final rule. Oral Argument Tr. at 47; see also Oral Argument Tr. at 48–50.

Reading Section 112(d)(3) to require EPA to use the data it “has” in its possession until the moment a rule is promulgated would frustrate the statutory purposes of the Clean Air Act. See 42 U.S.C. § 7401(b) (noting that the purpose of the Clean Air Act is, among other things, to “protect and enhance the quality of the Nation’s air resources”). Calculating emission standards that satisfy the Section 112(d) statutory criteria is a complicated process: EPA must gather and analyze data for boilers in each given category and then arrive at an emissions floor for each category that has “maximum stringency” but is also “continuously achievable.” U.S. Sugar Corp. I, 830 F.3d at 632; see also, e.g., id. (promulgating emission standards that satisfy “the statutory criteria is no easy task”); Nat’l Ass’n of Clean Water Agencies v. EPA, 734 F.3d 1115, 1132–33 (D.C. Cir. 2013) (“[E]stablishing MACT floors is no simple task.”). Under the Environmental Petitioners’ interpretation, that 25

complicated process would seemingly have no end. Consider, for example, what would happen if, following years of calculations to establish emissions floors, EPA were to receive new emissions information from a source the day before the planned promulgation of a rule setting emissions floors. The Environmental Petitioners’ interpretation would require EPA to pause the rulemaking process and recalculate its standards to reflect the new information that EPA newly “has” in its possession. And that same issue would repeat itself if a single boiler sent EPA new information after it had completed that next round of emissions floor calculations. We do not think that Congress intended for Section 112(d) to force EPA into such a never-ending loop. Cf. U.S. Sugar Corp. I, 830 F.3d at 647 (rejecting a proposed interpretation that would complicate EPA’s attempts to control pollutants and noting that “[n]othing in the CAA suggests that the Congress intended to so hamstring the Agency”). And we should not lightly read the statutory text to require EPA to act in a manner that is “self-defeating.” See Quarles v. United States, 587 U.S. 645, 654 (2019).

Moreover, this Court has generally acknowledged that EPA may exercise discretion and utilize its expertise when calculating emission standards for categories of sources. See, e.g., id. at 636. This Court has upheld, for example, EPA’s choice to set emission standards under Section 112(d)(3) by estimating the best performing twelve percent of sources based on the performance of the available technology, rather than solely on “the recorded performance” of sources in a category. See Nat’l Lime Ass’n v. EPA, 233 F.3d 625, 631–32 (D.C. Cir. 2000). In doing so, we explained that Section 112 “says nothing about what data the Agency should use to calculate emission standards.” Id. at 632. Yet the Environmental Petitioners’ reading of Section 112(d)(3) would seemingly contradict this precedent by insisting that EPA has no 26

discretion in choosing what data to use when setting these standards. In fact, under the Environmental Petitioners’ interpretation, EPA would seemingly lack discretion to ignore data even when that data is unusable for some reason. Imagine, for example, that the only data EPA has for a particular source in a subcategory of sources is unusable or unreliable. So EPA, quite reasonably, chooses to ignore that data—and, effectively, that source—when establishing the MACT floor for the relevant subcategory. Under the Environmental Petitioners’ reading of Section 112(d)(3), EPA’s choice to ignore that unusable data would be unlawful: EPA literally “ha[d]” information for a source in a subcategory, so it was required to include that source in its calculations and use the unreliable data. We find it quite implausible that Congress intended for the language to which the Environmental Petitioners point to remove EPA’s widely accepted ability to use its expertise to craft sensible standards.

The upshot of the Environmental Petitioners’ statutory argument is that if EPA “has” information about a source, it must include that source in its calculations—even if the only information it has is unusable or if EPA receives that information a week before a rule is to be proposed. Because that interpretation of Section 112(d) would substantially hamper EPA’s ability to effectively promulgate standards, we reject Environmental Petitioners’ interpretation and hold that EPA’s decision to rely on its original dataset was not unlawful. 10

10 Environmental Petitioners contend that affirming EPA’s decision to rely on its original dataset effectively allows EPA unlimited discretion to “selectively exclude valid emissions data from its floor calculations,” thereby permitting EPA to tailor standards to meet its policy goals. Environmental Pet. Br. 31. But we do not hold that there are no statutory limitations on the data that 27

The Environmental Petitioners next contend that even if EPA’s decision to rely on its original dataset did not contravene Section 112, EPA’s action was arbitrary and capricious in two related ways. First, they contend that even if EPA was not under a statutory obligation to continue collecting newer data, it was still arbitrary for EPA to choose not to use the newer data already in its possession. Second, the Environmental Petitioners explain that as a result of EPA’s refusal to use that newer data, several of the 2022 Rule’s standards are less protective than the prior standards in the 2013 Rule that were remanded by this Court in U.S. Sugar Corp. II. Until the promulgation of the 2022 Rule, every boiler was required to comply with the 2013 standards remanded by this Court in U.S. Sugar Corp. I and U.S. Sugar Corp. II. That many of the 2022 Rule’s standards are less stringent than the 2013 standards, Environmental Petitioners emphasize, effectively means that the 2022 standards are worse than the emissions level every boiler has been required to achieve for years. As we will explain, Environmental Petitioners’ arbitrary and capricious arguments are unpersuasive.

In the 2022 Rule, EPA explained that it had chosen to rely on the original dataset underlying the 2013 Rule because of, among other reasons, the “limited nature” of this Court’s remand to EPA in U.S. Sugar Corp. II, 87 Fed. Reg. at 60822;

EPA must take into account or that EPA can arbitrarily remove individual data points from its analysis. Rather, because we reject Environmental Petitioners’ proffered interpretation, we need not reach the question whether Section 112(d)(3) imposes restrictions on EPA’s choice of data (beyond, of course, the general constraint that EPA’s decision cannot be arbitrary and capricious). 28

EPA noted that the U.S. Sugar Corp. II Court had directed EPA to correct its prior standards, not to initiate a new standard- setting process altogether, id. at 60821. Further, EPA explained that if EPA were to revise the relevant standards using the updated emissions information in its possession, that could result in a “potentially inequitable outcome”—some units could be subject to “more stringent standards solely because of the EPA’s error” when it initially calculated the standards in the 2013 Rule. 87 Fed. Reg. at 60822. And, EPA explained, revising all of the standards in the 2013 Rule using the newer data would “require EPA to incur a significant resource burden.” Id.

We do not think EPA’s decision to rely on the original data underlying the 2013 Rule when it promulgated the 2022 standards was arbitrary and capricious. To begin with, EPA’s decision to rely on the original dataset was reasonable considering this Court’s instructions to EPA in U.S. Sugar Corp. II. As EPA correctly recognized, this Court’s remand in U.S. Sugar Corp. II was limited: the Court ordered EPA to “identify those standards for which the MACT floor would have differed” if EPA had not made the error the Court identified in the case, and then to “revise those standards” consistent with this Court’s opinion in U.S. Sugar Corp. I. U.S. Sugar Corp. II, 844 F.3d at 270 (citing U.S. Sugar Corp. I, 830 F.3d at 632). And the U.S. Sugar Corp. II Court also emphasized that EPA should not drag its feet on remand: the Court stated that it “expect[ed] the EPA to complete this rulemaking promptly.” Id. In light of the limited nature of this Court’s remand and the Court’s instruction to complete the rulemaking with haste, EPA’s decision to rely on its original dataset was sensible. So, too, was EPA’s explanation that relying on the original dataset would also allow the agency to ensure consistency across the full suite of standards. 29

In fact, this Court has previously upheld EPA’s decision to rely on a prior dataset, rather than updated data, in rather similar circumstances. In Board of Commissioners of Weld County, Colorado v. EPA, this Court addressed EPA’s actions in response to the Court’s prior remand of EPA’s determinations as to whether certain geographic areas were “attainment areas”—that is, whether pollutants in the relevant areas fell below the natural ambient air quality standards (“NAAQS”). 72 F.4th 284, 289 (D.C. Cir. 2023). When it reexamined its attainment determinations on remand, EPA chose to rely on the data it had used in making its original attainment determinations, rather than more recent data that was available to it. See id. at 288. Weld County, Colorado, petitioned for review of EPA’s revised attainment designations, arguing that EPA’s decision to rely on the original data was arbitrary and capricious. Id. at 289. EPA explained that using the original dataset would “standardize its analysis and thus facilitate consistent treatment of all affected” geographic areas. Id. Using the original dataset would also “streamline the process,” thus allowing EPA to better comply with the Court’s prior instruction that EPA should act “as expeditiously as practicable” on remand. Id. (cleaned up). This Court held that EPA had acted reasonably in declining to use the updated data on remand. The Court explained that EPA had “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.” Id. at 290.

Here, as in Weld County, this Court instructed EPA to complete its rulemaking on remand “promptly.” U.S. Sugar Corp. II, 844 F.3d at 270 (instructing EPA to act “promptly”). And EPA has reasonably explained that relying on its original 30

dataset, rather than engaging in a full-blown data gathering and assessment process, better enabled it to promulgate revised standards quickly. We think EPA’s decision to rely on its original dataset, much like EPA’s decision in Weld County, was reasonable.

The Environmental Petitioners’ contrary arguments are unavailing. The Environmental Petitioners make much of the fact that some of the revised MACT floor standards in the 2022 Rule are less stringent than the emissions limitations required by the 2013 Rule. But we note that, in the 2022 Rule, EPA explained that it had examined the post-2013 data and chosen to promulgate beyond-the-floor standards that require several subcategories to comply with the original 2013 limits. See 87 Fed. Reg. at 60825–26. So, while the MACT floors are in some cases less stringent than the 2013 Rule standards, many of the ultimate standards with which sources must comply are just as stringent. We are similarly unpersuaded by the Environmental Petitioners’ arguments that EPA has, in this case, sought to advance its “policy goals” by “selectively including or excluding sources from its floor analysis.” Environmental Pet. Reply Br. at 25. For one thing, nothing in the record indicates that EPA’s decision to rely on the original dataset represents an EPA policy choice to have less stringent standards. The Environmental Petitioners argue to the contrary, repeatedly quoting from EPA’s statement in the 2022 Rule that using the newer data to revise the affected standards would result in “more stringent” standards. Id.; see also 87 Fed. Reg. at 60822. But the full context of the Environmental Petitioners’ chosen quote tells a different story. EPA did not state that it preferred, as a policy matter, not to have “more stringent” standards—it stated that it would be “potentially inequitable” for some units to be subject to “more stringent” standards than others simply because of EPA’s error in the calculations that led to the 31

standards in the 2013 Rule. See 87 Fed. Reg. at 60822. The Environmental Petitioners’ selective quotation from the record is therefore unpersuasive. And for another thing, the Environmental Petitioners offer no evidence that EPA has, in fact, “selectively” included or excluded sources. The Environmental Petitioners’ qualm with EPA is that EPA chose to use the 2013 dataset, as a whole—not that EPA has removed individual sources from categories in order to arrive at the standards it prefers.

IV

We hold that the 2022 Rule misinterpreted the definition of “new source” and accordingly grant the petitions for review filed by the Industry Petitioners, setting the Rule aside to the extent that it defines sources constructed or reconstructed before August 24, 2020 as new sources. We also hold that EPA’s decision to rely on its original dataset was neither a violation of Section 112(d) nor arbitrary and capricious, and we therefore deny the petition for review filed by the Environmental Petitioners.

So ordered.