Full opinion text
PER CURIAM: In these consolidated petitions for review, we address approximately thirty challenges to three regulations promulgated by the United States Environmental Protection Agency (EPA or Agency): (1) the “Major Boilers Rule,” (2) the “Area Boilers Rule,” and (3) the “Commercial/Industrial Solid Waste Incinerators (CISWI) Rule.” Collectively, these rules — all promulgated under the Clean Air Act (CAA or Act), 42 U.S.C. §§ 7401 et seq. — set emissions limits on certain combustion machinery known to release hazardous air pollutants (HAPs). Roughly one-half of the challenges are advanced by a group of municipal-electric organizations, industrial-trade associations, oil-and-gas industry representatives, and other entities that own and operate boilers, process heaters, and incinerators (Industry Petitioners). The other one-half are pressed by organizations interested in safeguarding the environment (Environmental Petitioners). I. BACKGROUND The three rules at issue address a common phenomenon: when combustion occurs, emissions result. The emissions include numerous materials, some of which pose risks to the environment in general and to human health in particular. Because combustion is an inevitable occurrence in the machinery that helps to power modern society, the Congress has authorized the EPA to provide for a regulatory framework that minimizes the deleterious effects of the incineration industry while simultaneously allowing it to operate. In 2013, the EPA finalized its efforts to do so for discrete types of combustion machinery: boilers, process heaters, and incinerators. Two of the three rules at issue — the Major Boilers Rule and the Area Boilers Rule — govern boilers and process heaters. The former are enclosed devices that use a controlled flame to heat water and convert it into steam or hot-water. 40 C.F.R. § 63.11237. The latter are also enclosed devices that use a controlled flame but, instead of generating steam, they indirectly heat a “process material,” whether liquid, gas, or solid, or a “heat transfer material” like glycol or a mixture of glycol and water. Id. For simplicity, our use of “boilers” covers both machinery types. The two boiler-specific rules further divide the machinery into three categories: industrial, commercial, and institutional. See 2011 Area Boilers Rule, 76 Fed. Reg. at 15,557. Industrial boilers are used for manufacturing, processing, mining, refining, and other similar operations. See id. Commercial boilers are used by shopping malls, laundromats, apartment complexes, restaurants, and hotels. See id. And institutional boilers include those used by, e.g., medical centers, schools, churches, prisons, and courthouses. See id. Collectively, over 200,000 boilers at over 100,000 separate facilities must comply with the standards set out in the Major Boilers Rule or the Area Boilers Rule. The third rule that we address — the CI-SWI Rule — governs combustion machinery known as “solid waste incineration unit[s].” 42 U.S.C. § 7429. The Act defines an incinerator as a “distinct operating unit of any facility” that burns solid waste from either commercial establishments, industrial establishments, or the general public. Id. § 7429(g)(1). An incinerator subjects “waste material” to “high temperatures until it is reduced to ash.” Incinerator, New OxfoRD AmericaN Dictionary 853 (2d ed. 2005). Incinerators fall into different subcategories and, in the past, the EPA has issued rules governing many of them, including, e.g., municipal solid-waste incinerators, medical-waste incinerators, and sewage-sludge incinerators. At issue in the CISWI Rule are incinerators located in commercial or industrial facilities that combust solid waste as defined in the Resource Conservation Recovery Act (RCRA), 42 U.S.C. §§ 6901 et seq. See 2011 CISWI Rule, 76 Fed. Reg. at 15,706. A. The Clean Air Act, 42 U.S.C. §§ 7401 ET SEQ. Enacted “to protect and enhance the quality of the Nation’s air resources so as to promote the public health and welfare and the productive capacity of its population,” 42 U.S.C. § 7401(b)(1), the Act has been amended several times since the Congress first attempted to control air pollution via legislation in 1963. In 1970, the Congress required the EPA to identify and publish a list of HAPs, which the CAA defined as substances that increase “mortality,” “serious irreversible” illness, or “incapacitating reversible” illness. Clean Air Amendments of 1970, Pub. L. No. 91-, 604, § 4(a), 84 Stat. 1676, 1685 (1970). The EPA had to set emission limits for every HAP based on the risk it posed to human health. See Sierra Club v. EPA (Sierra Club I), 353 F.3d 976, 979 (D.C. Cir. 2004). In other words, the EPA was to “consider[ ] levels of HAPs at which health effects are observed, factor[ ] in an ample margin of safety to protect the public health, and set emission restrictions accordingly.” Id. (quotation marks omitted). The risk-focused approach to capping HAP emissions left something to be desired. “In light of unrealistic time frames and scientific uncertain[t]y over which substances posed a threat to public health,” the EPA “only listed eight pollutants as hazardous between 1970 and 1990,” Nat Res. Def. Council v. EPA (NRDCII), 529 F.3d 1077, 1079 (D.C. Cir. 2008), and set “emission standards for [only] seven of them,” Sierra Club I, 353 F.3d at 979; see also S. Rep. No. 101-228, at 3 (1989) (“Very little has been done since the passage of the 1970 Act to identify and control hazardous air pollutants.”). After twenty years of the risk-based approach, the Congress went back to the drawing board and, via the 1990 CAA Amendments, Pub. L. No. 101-549, 104 Stat. 2399 (1990), established the technology-based approach that governs today. See Sierra Club I, 353 F.3d at 979. 1. 42 U.S.C. § 7412 — “Hazardous Air Pollutants” The 1990 CAA Amendments overhauled the Act’s “Hazardous Air Pollutants” provision, codified at 42 U.S.C. § 7412. Although earlier iterations of the Act had assigned HAPs-identification responsibility to the EPA, the slow pace at which the EPA discharged its duty prompted the Congress to create a list of pollutants itself. See Sierra Club I, 353 F.3d at 979-80 (citing 42 U.S.C. § 7412(b)). After identifying nearly two hundred HAPs that warranted emissions restrictions, see 42 U.S.C. § 7412(b)(1), the Congress directed the EPA, first, to identify the sources of each HAP, see id. § 7412(c). The Agency then was to set emissions limits for each source that result in HAPs reduction to the greatest extent achievable by current technology. See generally Nat’l Ass’n for Surface Finishing v. EPA, 795 F.3d 1, 4 (D.C. Cir. 2015) (citing 42 U.S.C. § 7412(b)(1), (c), (d)). a. Identifying and Categorizing HAP Sources The EPA’s first task is to create HAP-source categories and subcategories. See 42 U.S.C. § 7412(c). The Act distinguishes “major” from “area” sources, defining the former as “any stationary source or group of stationary sources” that neighbor each other, share common control, and emit (or have the potential to emit) either ten tons per year or more of any single HAP or twenty-five tons per year or more of any HAP combination. Id. § 7412(a)(1). The latter are sources that do not emit enough HAPs to qualify as “major.” Id. § 7412(a)(2). Although the EPA must set stringent restrictions on major sources, it has discretion to set more lenient emissions caps on area sources. See id. § 7412(d)(5). Apart from the statutory distinction between major and area sources, the EPA has discretion to differentiate “among classes, types, and sizes of sources within a category or subcategory.” Id. § 7412(d)(1). Once the EPA finalizes HAPs-source categories and subcategories, the CAA mandates that it draw one final dividing line— between “new” sources and “existing” sources. See id. § 7412(d)(3). “New” sources are those “on which construction begins after EPA publishes emission standards,” Cement Kiln Recycling Coal. v. EPA, 255 F.3d 855, 858 (D.C. Cir. 2001); most of the others are “existing” sources, see 42 U.S.C. § 7412(a)(10). But if an existing source experiences either a physical change or a change in operation method and the change increases HAP emissions by more than a de minimis amount, the Act mandates that the source meet the standards set for new sources. See id. § 7412(a)(5), (g). b. Setting Emission Standards for Major Sources — the “MACT” Standard After the EPA identifies HAP-source categories and subcategories, it then sets emissions limits for each. See id. § 7412(d)(2). “[Whenever ... feasible,” the caps must use numeric HAPs limits. Id. § 7412(h)(4). The size of the source— either “major” or “area” — dictates whether the EPA must set the numeric limit at the most stringent level that current technology allows or at the level set by “generally available control technologies.” Id. § 7412(d)(5). For major sources, the CAA directs the EPA to establish emissions caps that result in the “the maximum degree of reduction in emissions” that the EPA determines is “achievable.” Id. § 7412(d)(2). We refer to an emissions cap that reflects the current “maximum achievable control technology” as a “MACT” standard. See NRDCII, 529 F.3d at 1079. Setting a MACT standard is a two-step process. First, the EPA establishes a “MACT floor” for each category or subcategory. Sierra Club I, 353 F.3d at 980. The MACT floor ensures that all HAPs sources “at least clean up their emissions to the level that their best performing peers have shown can be achieved.” Id. For new sources — those built after promulgation of a HAPs limit, see 42 U.S.C. § 7412(a)(4)— the MACT floor cannot be less stringent than the emissions levels achieved by the best performing similar source. Id. § 7412(d)(3). For existing sources in categories or subcategories that have thirty or more sources, the MACT floor cannot be less stringent than the average emissions limits achieved by the best performing 12 per cent of existing sources in that category or subcategory. Id. § 7412(d)(3)(A). And for existing sources in categories or subcategories with fewer than thirty sources, the MACT floor cannot be less stringent than the average emissions achieved by the best performing five sources. Id. § 7412(d)(3)(B). When setting the MACT floor, the EPA considers only the performance of the cleanest sources in a category or subcategory; it does not take into account other factors, including the cost of putting a source in line with its better-performing counterparts. See Cement Kiln, 255 F.3d at 857-58 (citing Nat’l Lime Ass’n v. EPA, 233 F.3d 625, 629 (D.C. Cir. 2000), as amended on denial of reh’g, No. 99-1325 (D.C. Cir. Feb. 14, 2001)). Second, the EPA must determine whether current technology makes it possible for a source to perform even better than the best performing similar source or sources. In other words, the CAA directs the EPA to consider whether it should set a “beyond-the-floor” MACT standard. Nat’l Lime Ass’n, 233 F.3d at 629. In determining whether a beyond-the-floor standard is “achievable,” the Agency must consider additional factors like “the cost of achieving such emission reduction,” “any non-air quality health and environmental impacts” and “energy requirements.” 42 U.S.C. § 7412(d)(2). It has broad discretion in its determination. See id.; cf. Nat’l Ass’n of Clean Water Agencies v. EPA (NACWA), 734 F.3d 1115, 1157 (D.C. Cir. 2013) (noting, in section 7429 case, that “Congress gave EPA broad discretion in considering whether to go beyond-the-floor”). c. Setting Emission Standards for Area Sources — the “GACT” Standard Although the EPA must cap HAP emissions from major sources at the “maximum degree of reduction,” see 42 U.S.C. § 7412(d)(2), it has discretion to set less stringent caps on emissions from area sources. Indeed, the EPA need not list categories of area sources at all unless: (A) it finds that the sources in that category or subcategory “present[] a threat of adverse effects” to the environment or human health, see id. § 7412(c)(1), (3); or (B) control of a particular area source category or subcategory is necessary to ensure that sources accounting for at least 90 per cent of the aggregate emissions of the thirty HAPs the EPA believes “present the greatest threat to public health in the largest number of urban areas” are subject to CAA control, id. § 7412(c)(3), (k)(3)(B). If it finds that controlling emissions from a particular area source subcategory is necessary to achieve a 90 per cent reduction in the aggregate emissions of any of seven CAA-enumerated HAPs, section 7412(c)(6) requires the Agency to impose MACT caps on that subcategory. See id. § 7412(c)(6). With the exception of section 7412(c)(6)’s MACT-standard requirement, the EPA need not cap emissions from area sources at the MACT level. Instead, it may set more lenient emissions limits based on “generally available control technologies.” Id. § 7412(d)(5). We refer to these caps as GACT standards. The Act provides no guidance for setting GACT standards but the legislative history of the 1990 CAA Amendments describes GACT “as methods, practices and techniques [that] are commercially available and appropriate for application by the sources in the category considering economic impacts and the technical capabilities of the firms to operate and maintain the emissions control systems.” S. Rep. No. 101-228, at 171 (1989). According to the EPA, it can and will consider the following in setting a GACT standard: • “costs and economic impacts ..., which [are] particularly important when developing regulations for source categories that may have many small businesses ... ”; • “the control technologies and management practices that are generally available to the area sources in the source category”; • “the standards applicable to major sources in the analogous source category to determine if the control technologies and management practices are transferable and generally available to area sources”; and • “technologies and practices at area and major sources in similar categories to determine whether such technologies and practices could be considered generally available for the area source categories at issue.” 2011 Area Boilers Rule, 76 Fed. Reg. at 15,556. And, unlike the EPA’s duty to consider a beyond-the-floor MACT standard, it need not consider a more stringent GACT standard. d. Work-Practice and Management-Practice Standards Although the CAA requires numeric emission standards where possible, the EPA can “promulgate a design, equipment, work practice, or operational standard, or combination thereof’ if it determines that a numeric limit is “not feasible.” 42 U.S.C. § 7412(h)(1).. In other words, the EPA can require that all sources in a given category or subcategory take a certain action (e.g., conduct a periodic tune-up) or install certain emissions-control technology (e.g., install a fabric filter). Although the EPA has discretion to impose a work-practice standard, the Act limits it by defining the operative phrase “not feasible” narrowly to mean: (A) a hazardous air pollutant or pollutants cannot be emitted through a conveyance designed and constructed to emit or capture such pollutant, or that any requirement for, or use of, such a conveyance would be inconsistent with any Federal, State or local law, or (B) the application of measurement methodology to a particular class of sources is not practicable due to technological and economic limitations. Id. § 7412(h)(2). Similarly, for area sources, the EPA can impose a “management-practice standard” in lieu of a numeric GACT standard. See id. § 7412(d)(5). A management-practice GACT standard is like a work-practice MACT standard in all ways but one — the EPA need not consider feasibility when setting management-practice standards. Compare id. § 7412(d)(2), with id. § 7412(d)(5). 2. 42 U.S.C. § 742 — “Solid Waste Combustion” In addition to amending the Act’s “Hazardous Air Pollutants” provision, see id. § 7412, the 1990 CAA Amendments added to the U.S. Code section 7429, titled “Solid Waste Combustion.” Section 7429 regulates “solid waste incineration units” generally, see id. § 7429(a)(1)(A), and CI-SWI specifically, see id. § 7429(a)(1)(D). Although section 7412 requires the EPA to control emissions of nearly two hundred HAPs, see id. § 7412(d)(1), section 7429 mandates that the EPA control emissions from only nine specific'pollutants (as well as opacity, where appropriate), none of which the Congress included on its initial section 7412 list, see id. § 7429(a)(4); see also Nat. Res. Def. Council v. EPA (NRDC I), 489 F.3d 1250, 1256 (D.C. Cir. 2007). We have held that this difference “makes promulgating ... standards under [section .7412] and [section 7429] mutually exclusive.” NACWA, 734 F.3d at 1119. In other words, if a source (or facility) is considered a CISWI and, therefore, regulated under section 7429, it cannot be regulated under section 7412.. See id. Whether a source falls under section 7412 or section 7429, “the statutory directive on setting MACT standards is virtually identical.” Id.; see also Nat’l Lime Ass’n, 233 F.3d at 631. That said, regulation under one section instead of the other “has practical consequences.” NACWA, 734 F.3d at 1120. For example, section 7412 allows the EPA to impose a GACT standard for area sources only but section 7429 requires the EPA to impose MACT standards for all covered units, regardless of their size. Compare 42 U.S.C. § 7412(d)(1), (5), with id. § 7429(a)(1)(A); see also NRDC I, 489 F.3d at 1256. Moreover, section 7412 mandates that the EPA control HAP emissions from “major souree[s],” which the Act defines broadly to include “group[s] of stationary sources located within a contiguous area and under common control.” 42 U.S.C. § 7412(a)(1) (emphasis added). Section 7429, in contrast, mandates that the EPA control emissions from “solid waste incineration unit[s],” which the Act defines more narrowly as “a distinct operating unit of any facility which combusts any solid waste material,” 42 U.S.C. § 7429(g)(1) (emphases added). And finally, section 7429 does not provide for work-practice standards. 3. 42 U.S.C. §§ 7661 et seq.— “Title V Permits” Finally, the 1990 CAA Amendments added a provision to Title V of the Act that requires all owners and operators of HAP sources to obtain operating permits. See id. § 7661a. Title V does no more than consolidate “existing air pollution requirements into a single document, the Title V permit, to facilitate compliance monitoring” without imposing any new substantive requirements. Sierra Club v. Leavitt, 368 F.3d 1300, 1302 (11th Cir. 2004). The legislative history of the 1990 CAA Amendments indicates that the Congress required the “Title V permits” so that the public might “better determine the requirements to which the source is subject, and whether the source is meeting those requirements.” S. Rep. No. 101-228, at 347. Although owners and operators of all major HAP sources must obtain Title V permits, see generally 42 U.S.C. § 7661a(a), the EPA has discretion to exempt certain area source categories if it “finds that compliance with such requirements is impracticable, infeasible, or unnecessarily burdensome,” id. B. The Major Boilers, Area Boilers, and CISWI Rules On March 21, 2011, the.EPA issued the first iteration of all three rules under review. That same day, however, the EPA announced that it intended to reconsider certain aspects of each rule. Not long after, multiple parties filed the petitions for review that we now address. Earlier, the EPA had concluded its reconsideration and issued the most recent iteration of the three rules. Because of this procedural quirk, each “rule” we address is in fact two separate rules — the EPA’s “final” 2011 version and its “final” 2013 version. The EPA’s analyses remained mostly consistent from 2011 to 2013 and we indicate, where necessary, the instances in which the EPA changed course in a significant way. 1. The Major Boilers Rule The Major Boilers Rule sets HAPs emission caps for all industrial, commercial, and institutional boilers that emit a large volume of HAPs. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,611. The EPA further divided the major boiler categories into subcategories based on the primary fuel combusted by the boilers in the subcategory (e.g., coal, biomass, gas, etc.) and, for some subcategories, based on the method used to “feed” the fuel into the boiler. See 2013 Major Boilers Rule, 78 Fed. Reg. at 7,144. For most of the subcategories, the EPA set a numeric MACT standard for four different HAPs: pártieu-late matter (PM); hydrogen chloride (HC1); mercury (Hg); and carbon monoxide (CO). See id. at 7,142 tbl.3; No. 11-1108 EPA Br. 9. The EPA used some of these HAPs — particularly CO — as a surrogate (or proxy) to set emissions limits for others on the section 7412(b) HAPs list. See 2013 Major Boilers Rule, 78 Fed. Reg. at 7,144-45. For the other major boiler subcategories, the EPA set a work-practice standard (specifically, a tune-up requirement) in lieu of numeric MACT standards. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613. The EPA also established a tune-up work-practice standard to control for dioxin/furan emissions across all major boiler subcategories. 2013 Major Boilers Rule, 78 Fed. Reg. at 7,138. In addition to these emission standards, the Major Boilers Rule includes several other provisions relevant to the current petitions for review. a. The “Upper Prediction Limit” Several factors complicate the process of setting MACT floors. The first is the CAA itself, which mandates that all MACT floors (1) must be achievable, see 42 U.S.C. § 7412(d)(2); (2) must ensure continuous regulation of the covered sources, see id. § 7602(k); and (3) must be no less stringent than the emissions levels being achieved by the best-controlled sources, see id. § 7412(d)(3). The second is that no source emits any HAP at a constant level; rather, HAP emissions fluctuate over time and for many reasons, including, e.g., “operation of control technologies, variation in combustion materials and combustion conditions, variation in operation of the unit itself, and variation associated with the emission measurement techniques.” Memorandum from Stephen D. Page, EPA Director of Air Quality Planning and Standards, EPA’s Response to Remand of the Record for Commercial and Industrial Solid Waste Incineration Units (Page Mem.) (July 14, 2014), at 3 (No. 11-1125 J.A. 1316). Finally, most sources do not measure their HAP emissions at all times and under all conditions. Id. at 6. Instead, data are usually gathered when a source conducts a “three-run stack test.” Id. This test provides three “snapshots” of a source’s emissions in a limited set of conditions and, accordingly, it fails to demonstrate accurately a source’s emissions during all times and under all conditions. Id. To compensate for the lack of adequate emissions data, the EPA uses a statistical tool known as the “upper prediction limit” (UPL) to account for the expected variability in emissions levels. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,630. The UPL, in turn, allows the Agency to set a MACT floor that is continuously achievable. Id. We discuss the UPL mechanics at greater length below, see infra § IV.C, but, in short, the EPA: (1) ranks all sources in a given category based on their three-run stack-test data; (2) determines the HAP emissions level of the “best controlled similar source” to establish standards for new sources, 42 U.S.C. § 7412(d)(3), and determines the average HAP emissions levels of the best performing 12 per cent of sources to establish standards for existing sources, id. § 7412(d)(3)(A); and then (3) applies the UPL methodology to provide the cushion necessary to account for the expected peaks and valleys in HAP emissions not reflected in the three-run stack-test “snapshots.” See Page Mem. 4, 6. b. The “Pollutant-By-Pollutant” Approach In identifying the best performing sources in a given category, often the EPA could not identify a single source that controlled all HAPs better than all other sources. Instead, the EPA found that one source effectively controlled emissions from one HAP but was nonetheless one of the worst-performing sources at controlling emissions from a different HAP. For this reason, the EPA adopted a “pollutant-by-pollutant” approach in setting MACT floors for major boiler subcategories. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,622-23. That is, instead of identifying the one source that, on balance, best controlled all HAPs in the aggregate, the EPA used one source to set the MACT floor for, e.g., PM, and used a different source to set the MACT floor for, e.g., HC1. For at least two subcategories of major boilers- — new heavy oil-fired units and existing stoker coal-fired units — the EPA’s pollutant-by-pollutant approach resulted in MACT floors that no source had achieved in toto. c. Startups, Shutdowns, and Malfunctions The EPA found it difficult to account for HAP emissions when sources start up, shut down, and malfunction. All three occurrences alter HAP emissions and, historically, the EPA exempted sources from normal numeric MACT-standard compliance when these events occurred. See, e.g., Standards of Performance for New Stationary Sources, 42 Fed. Reg. 57,125 (Nov. 1, 1977). Nevertheless, concluding that the Act “require[s] that there must be continuous section [7412]-compliant standards” and observing that the exemption meant that “no section [7412] standard governs these events,” in 2008 we vacated the exemption for startups, shutdowns,- and malfunctions when the issue arose in a case challenging a different rule. Sierra Club v. EPA (Sierra Club III), 551 F.3d 1019, 1027-28 (D.C. Cir. 2008) (emphasis added). In response to the Sierra Club III vaca-tur, the EPA established a work-practice standard in lieu of a numeric MACT standard during startup and shutdown periods (but not during malfunctions) when it promulgated the Major Boilers Rule. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,-613. It did so after determining that the “physical limitations and the short duration of startup and shutdown periods” made it technologically infeasible to conduct the requisite- testing for numeric emissions limits. Id. A work-practice standard sufficed, in the EPA’s view, because “[pjeriods of startup, normal operations, and shutdown are all predictable and routine aspects of a source’s operations.” Id. But because a malfunction is “sudden, infrequent, and not reasonably preventable,” id. (quoting 40 C.F.R. § 63.2), the EPA declined to treat a malfunction as a “distinct operating mode,” id. As a result, the EPA did not account for malfunctions when it set the MACT floors and it required sources to comply with all MACT floors even during periods of malfunction. Id. At the same time and recognizing that even the best equipment can fail and that such failure can spike emissions, the EPA added to the Major Boilers Rule “an affirmative defense to civil penalties for excee-dances of numerical emission limits that are caused by malfunctions.” Id. In reviewing a challenge to a different EPA rule, however, we vacated a materially identical affirmative-defense provision and held that the EPA has no power under the CAA to create a defense to civil liability. See Natural Res. Def. Council v. EPA (NRDC III), 749 F.3d 1055, 1062-64 (D.C. Cir. 2014). Here, the EPA defends its decision not to address malfunctions by asserting that it will use its enforcement discretion regarding malfunctions on a case-by-case basis. d. The One-Time Energy Assessment The EPA also promulgated a “beyond-the-floor” requirement for all facilities with existing major boilers. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613. Specifically, the Major Boilers Rule mandates a “a one-time energy assessment ... on the affected boilers and facility to identify any cost-effective energy conservation measures,” id., which assessment includes, inter alia, a review of fuel usage, energy management practices, and conservation measures, see 2013 Major Boilers Rule, 78 Fed. Reg. at 7,198-99. In some respects, the energy assessment is limited: it (1) need occur only one time, see 40 C.F.R pt. 63, subpt. DDDDD tbl.3; (2) is “based on energy use by discrete segments of a facility and not by a total aggregation of all individual energy using elements of a facility,” 2013 Major Boilers Rule, 78 Fed. Reg. at 7,146; and (3) does not require an owner or operator to implement any of the energy-saving findings the assessment makes. In one respect, however, it is expansive — it requires owners and operators to assess not only the boilers themselves but also other components “located on the site of the affected boiler that use energy provided by the boiler,” including “compressed air systems” as well as “facility heating, ventilation, and air conditioning systems.” 40 C.F.R. § 63.11237. e. The Health-Based Emissions Limits for HC1 Although the EPA set numeric MACT standards to control HC1 emissions, see 2013 Major Boilers Rule, 78 Fed. Reg. at 7,193-98 tbls.l & 2, in an earlier iteration of the Major Boilers Rule, the EPA did not set MACT standards for HC1. See National Emission Standards for Hazardous Air Pollutants for Industrial, Commercial, and Institutional Boilers and Process Heaters (2004 Boilers Rule), 69 Fed. Reg. 55,218, 55,227 (Sept. 13, 2004). Instead, the Agency opted for a less stringent health-based emissions limit under section 7412(d)(4). See id. The EPA changed course after concluding that HC1 emissions posed health concerns the Agency had not previously considered — in particular, the EPA feared the “potential cumulative public health and environmental effects” of HC1 emissions, 2011 Major Boilers Rule, 76 Fed. Reg. at 15,643-44 (emphasis added) — and after recognizing that it did not have the requisite data to weigh adequately the newly identified health risks. 2. The Area Boilers Rule In the Area Boilers Rule, the EPA set ■emissions limits for the same three boiler categories it controlled in the Major Boilers Rule, see supra § T.B.l: industrial, commercial, and institutional boilers. See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488. It further split the categories into seven subcategories, see id., and set emissions limits for three of them, see id. at 7,517-18 tbls.l & 2. These include: (1) coal-fired boilers (ie., “any boiler that burns any solid fossil fuel and no more than 15 percent biomass,” 40 C.F.R. § 63.11237); (2) oil-fired boilers (ie., “any boiler that burns any liquid fuel and is not in either the biomass or coal subcategories,” id.); and (3) biomass-fired boilers (ie., “any boiler that burns any” “biomass-based solid fuel that is not a solid waste” and “is not in the coal subcategory,” id.). See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,517-18 tbls.l & 2. For these subcategories, the EPA set emissions limits for three HAPs: Hg, PM, and CO, with PM functioning as a surrogate for non-Hg urban metals and CO functioning as a surrogate for polycyclic organic matter (POM). See 2011 Area Boilers Rule, 76 Fed. Reg. at 15,586. Because Hg and POM are both listed in section 7412(c)(6), the EPA had to set MACT standards for Hg and for CO (as surrogate for POM) for any area source category that, in the EPA’s view, required MACT control to assure a 90 per cent reduction in the aggregate emissions of these two HAPs. See 42 U.S.C. § 7412(c)(6). The Agency complied, setting numeric MACT standards for Hg and CO emissions from large coal-fired boilers and a MACT work-practice standard (specifically, a tune-up requirement) for emissions from small coal-fired boilers. See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488, 7,517-lS. It did not, however, set MACT standards for Hg and POM emissions from biomass or oii-fired boilers, finding it unnecessary to assure a 90 per cent reduction in aggregate emissions of those two HAPs. See 2011 Area Boilers Rule, 76 Fed. Reg. at 15,566. Thus, with the exception of Hg and CO emissions from coal-fired boilers, the EPA had discretion to promulgate GACT standards for all other HAPs in all other source subcategories. See 42 U.S.C. § 7412(d)(5). Exercising this discretion resulted in the following standards: 2013 Area Boilers Rule, 78 Fed. Reg. at 7,488-89, 7,517-19. The Area Boilers Rule shares many of the same features as the Major Boilers Rule; for example, the Area Boilers Rule treats startups, shutdowns, and malfunctions in the same fashion as the Major Boilers Rule, see supra § I.B.l.c — i.e., the Area Boilers Rule creates work-practice (or management-practice) standards for startup and shutdown periods but does not account for malfunctions at all, save for the Agency’s commitment to consider malfunctions on a case-by-case basis. See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,496; 2011 Area Boilers Rule, 76 Fed. Reg. at 15,560-61. Additionally, the Area Boilers Rule imposes the same one-time energy-assessment requirement for existing large area boilers that the Major Boilers Rule imposes for existing major boilers. See supra § I.B.l.d; see also 2013 Area Boilers Rule, 78 Fed. Reg. at 7,500; 2011 Area Boilers Rule, 76 Fed. Reg. at 15,560, 15,-567-68. There are, however, two unique features of the Area Boilers Rule that warrant brief discussion. a. Exclusion of “Temporary Boilers” After the EPA promulgated the 2011 Area Boilers Rule but before it promulgated the 2018 version, it proposed an amendment to 40 C.F.R. § 63.11195 that added temporary boilers to the list of those boilers not regulated by section 7412. See National Emission Standards for Hazardous Air Pollutants for Area Sources: Industrial, Commercial, and Institutional Boilers (2011 Proposed Area Boilers Rule on Reconsideration), 76 Fed. Reg. 80,532, 80,535 (Dec. 23, 2011). The EPA created the exclusion because, in its view, temporary boilers are “insignificant sources! ] and were not included in the EPA’s analysis of the source category.” Id. The Agency eventually defined “temporary boiler” as “any gaseous or liquid fuel boiler that is designed to, and is capable of, being carried or moved from one location to another by means of, for example, wheels, skids, carrying handles, dollies, trailers, or platforms.” See 2013 Area Boilers Rule, 78 Fed. Reg. at 7,491 (quoting 40 C.F.R. § 63.11237). b. Title V Permit Exemption for Synthetic Area Sources As noted, see supra § I.A.3, Title V of the CAA imposes a permit requirement on all owners and operators of major and area HAP sources. See 42 U.S.C. § 7661a. The EPA, however, can exempt an area source subcategory if it finds “that compliance with such requirements is impracticable, infeasible, or unnecessarily burdensome on such categories.” Id. § 7661a(a). When it proposed the Area Boilers Rule in 2010, the EPA considered exempting some area sources because, in its view, the existing restrictions on those sources made Title V duplicative. See 2010 Proposed Area Boilers Rule, 75 Fed. Reg. at 31,910-13. At the time, the EPA announced that it did not intend to exempt “synthetic” area sources (ie., area sources that, but for existing air-pollution controls, would be considered major sources).. Id. at 31,913. In so doing, the EPA reasoned that synthetic area sources: (1) more closely resemble major sources than area sources, (2) are often located in populous areas, and (3) have high HAP emissions potential when uncontrolled. Id. But in the 2011 Area Boilers Rule, the EPA changed course and exempted synthetic area sources from the Title V permitting requirement. See 76 Fed. Reg. at 15,578. It reasoned that the “observations and data ... relied upon in other rulemak-ings for distinguishing between sources that became synthetic area sources due to controls and other synthetic and natural area sources did not necessarily apply to this source category.” Id. In its view, it no longer had “sufficient information” to distinguish synthetic area sources from the others it exempted and, accordingly, “the rationale for exempting most area sources subject to this rule ... is also now relevant for” synthetic area sources. Id.; see also 2013 Area Boilers Rule, 78 Fed. Reg. at 7,497. 3. The CISWI Rule «In the CISWI Rule, the EPA created four CISWI subcategories: (1) incinerators (ie., “units designed to burn [solid] waste materials for the purpose of disposal”); (2) small, remote incinerators (“SRIs”) (ie., units that burn small waste batches); (3) energy recovery, units (“ERUs”) (ie., units that would be classified as boilers but for the fact they com-bust solid waste); and (4) waste-burning kilns (ie., units that would be classified as cement kilns if they did not burn solid waste). 2013 CISWI Rule, 78 Fed. Reg. at 9,118. Initially, the EPA proposed a fifth subcategory — burn-off ovens — but eliminated burn-off ovens after comments revealed that it had greatly underestimated the number of units in that subcategory (36 versus 15,000) and that it lacked the requisite data to set limits for the units. See 2011 CISWI Rule, 76 Fed. Reg. at 15,734. Of the four CISWI subcategories, the EPA further divided the ERU subcategory (for CO emissions only) into coal-fired, biomass-fired and oil/gas-fired ERUs and it further divided the waste-burning kiln subcategory (again, for CO emissions only) into long and preheater/precalcinator kilns. See 2013 CISWI Rule, 78 Fed. Reg. at 9,118 tbl.2. The EPA then set numeric MACT limits for the section 7429(a)(4) pollutants. See 2011 CISWI Rule, 76 Fed. Reg. at 15,709-10 tbl.l. Unlike the Major Boilers Rule and the Area Boilers Rule, the CI-SWI Rule contains no beyond-the-floor MACT standards. The EPA also declined to promulgate work-practice standards, concluding that it had no authority to do so because section 7429 includes no work-practice standard provision similar to that in section 7412. See id. at 15,721. The CISWI Rule shares several features with the Major Boilers Rule. In the CI-SWI Rule, for instance, the EPA also used the UPL, see id. at 15,722-27, as well as the pollutant-by-pollutant approach, see id. at 15,719-21, in setting MACT floors. Based in part on the differences between section 7412 and section 7429, the CISWI rule has four unique characteristics we briefly describe. a. Startups, Shutdowns, and Malfunctions As discussed, see supra § ILB.l.c, the EPA imposed a work-practice standard for major and area source boilers during periods of startup and shutdown but declined to make any regulatory modification for malfunctions. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613; 2011 Area Boilers Rule, 76 Fed. Reg. at 15,560-61. The CI-SWI Rule, however, makes no modification for any of these periods, mandating instead that the numeric MACT standards “apply at all times,” even when CISWI units are starting up or shutting down. 2011 CISWI Rule, 76 Fed. Reg. at 15,711, 15,737-38. The Agency concluded that it had no legal authority under section 7429 to impose anything but a numeric MACT standard on CISWI units. See id. at 15,709 tbl.l; see also id. at 15,737-38. b. The Record-Keeping Requirement Whether the EPA considers a combustion unit to be a boiler (and thus subject to section 7412) or a CISWI (and thus subject to section 7429) turns entirely on whether the unit combusts “solid waste.” See id. at 15,709. The term “solid waste” is defined in RCRA, 42 U.S.C. §§ 6901 et seq., and clarified by EPA regulation, see Identification of NonHazardous Secondary Materials that Are Solid Waste (NHSM Rule), 76 Fed. Reg. 15,456, 15,457 (Mar. 21, 2011). See also 2011 CISWI Rule, 76 Fed. Reg. at 15,709. If the unit combusts solid waste, it is a CISWI. Id. The source owner or operator initially decides whether the material its combustion unit burns meets the definition of solid waste.' See id. at 15,740. For this reason, the CISWI rule requires that the owner or operator of a combustion unit that burns materials “not clearly listed as traditional fuels” keep records explaining how the materials meet the regulatory definition of “non-solid waste.” Id.; see also 40 C.F.R. § 60.2175(v). Failure to do so means, for the purposes of the EPA, that “the operating unit is a CISWI unit.” 40 C.F.R. § 60.2265; see also 2013 CISWI Rule, 78 Fed. Reg. at 9,188. c. Emissions Averaging • During the notice-and-eomment period, certain industry entities urged the EPA to allow a facility containing more than one CISWI unit to demonstrate compliance with the CISWI MACT standards by averaging the HAP emissions of all units in the facility. See Commercial and Industrial Solid Waste Incineration Units: Reconsideration and Proposed Amendments; NonHazardous Secondary Materials that Are Solid Waste (2011 Proposed CISWI Rule on Reconsideration), 76 Fed. Reg. 80,452, 80,463 (Dec. 23, 2011). Although it allowed facility-wide averaging in the Major Boilers Rule, the Agency declined to allow it for facilities with CISWI units. See id. The EPA explained, first, that “[t]he applicability of CISWI is such that each unit is an affected facility.” Id. In response to further comments, the EPA subsequently explained that it did “not believe [it had] the legal authority to allow emissions averaging in CISWI or under section [7429] generally because each individual unit is an affected facility.” Summary of Public Comments and Responses for Commercial and Industrial Solid Waste Incineration Units (CISWI Rule — Responses to Comments), EPA-HQ-OAR-2003-0119-2638-A2 (Dec. 2012), at 195. d. Treatment of Units that Begin Combusting Solid Waste Finally, in the preamble to the 2011 CISWI Rule, the EPA stated broadly that “[u]nits that begin combusting solid waste are considered existing sources under CI-SWI.” 76 Fed. Reg. at 15,714 (emphasis added). This categorical pronouncement drew objections from commentators who insisted that, if such units experienced an increase in HAP emissions, the units would meet the statutory definition of “modified solid waste incineration unit[s],” see 42 U.S.C. § 7429(g)(3), and would, accordingly, be subject to the MACT standards -for new units, see id. § 7429(g)(2). In the subsequent proposed CISWI Rule, the EPA clarified that “[a]n existing source will not be considered a new source solely due to a combustion material switch. Assuming new source applicability is not triggered, existing sources that change fuels or materials are considered existing sources.... ” 2011 Proposed CISWI Rule on Reconsideration, 76 Fed. Reg. at 80,459. II. STANDARD OF REVIEW For each issue, the Petitioners argue that the EPA'either misinterpreted the CAA, acted arbitrarily and capriciously, or both. We review the EPA’s construction of the statute under the two-part framework established in Chevron, U.S.A., Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). At Chevron step 1, we ask whether the Congress “has directly spoken to the precise question at issue”; if it has, we “must give effect to [its] unambiguously expressed intent.” Id. at 842-43, 104 S.Ct. 2778. In so doing, we examine the CAA’s text, structure, purpose, and legislative history to determine if the Congress has expressed its intent unambiguously. See Bell Atl. Tel. Co. v. FCC, 131 F.3d 1044, 1047 (D.C. Cir. 1997). If the statute is “silent or ambiguous with respect to the specific issue,” we proceed to Chevron step 2 and defer to the EPA’s interpretation so long as it is “based on -a permissible construction of the statute.” Chevron, 467 U.S. at 842-43, 104 S.Ct. 2778. The CAA authorizes the Court to “reverse any [EPA] action found to be ... arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 42 U.S.C. § 7607(d)(9)(A). Our review is “narrow” and we will “not ... substitute [our] judgment for that of the agency.” Motor Veh. Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co. (State Farm,), 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). We “must uphold an agency’s action where [the agency] ‘has considered the relevant factors and articulated a rational connection between the facts found and the choice made,’ and has not ‘relied on [improper] factors.’” Nat’l Ass’n of Clean Air Agencies v. EPA (NACAA), 489 F.3d 1221, 1228 (D.C. Cir. 2007) (citations omitted) (quoting Allied Local & Reg’l Mfrs. Caucus v. EPA, 215 F.3d 61, 68 (D.C. Cir. 2000), and State Farm, 463 U.S. at 43, 103 S.Ct. 2856). A rule is arbitrary and capricious if the agency: (1) “has relied on factors which Congress has not intended it to consider,” (2) “entirely failed to consider an important aspect of the problem,” (3) “offered an explanation for its decision that runs counter to the evidence before the agency,” or (4) “is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.” State Farm, 463 U.S. at 43, 103 S.Ct. 2856. We review the EPA’s factual determinations for substantial evidence. 5 U.S.C. § 706(2)(E). We also “owe[ ] particular deference to EPA when its rulemakings rest upon matters of scientific and statistical judgment within [its] sphere of special competence and statutory jurisdiction.” Am. Coke & Coal Chems. Inst. v. EPA, 452 F.3d 930, 941 (D.C. Cir. 2006). But “[w]e are hesitant to rubber-stamp EPA’s invocation of statistics without some explanation of the underlying principles or reasons why its formulas would produce an accurate result.” NACWA, 734 F.3d at 1145. III. INDUSTRY PETITIONERS’ CHALLENGES A. STARTUPS, Shutdowns, and Malfunctions Industry Petitioners raise two sets of challenges to startup, shutdown, and malfunction periods: (1) a challenge to the EPA’s failure to take malfunctions into account in the Major Boilers and Area Boilers Rules and (2) a challenge to EPA’s failure to take into account periods of startup, shutdown, and malfunction in the CISWI Rule. For the reasons that follow, we reject all of the Industry Petitioners’ claims related to startups, shutdowns, and malfunctions. 1. Periods of Malfunction in the Major Boilers and Area Boilers Rules First, Industry Petitioners challenge the Major Boilers and Area Boilers Rules’ failure to take malfunctions into account in setting MACT floors. See 2011 Major Boilers Rule, 76 Fed. Reg. at 15,-613; 2011 Area Boilers Rule, 76 Fed. Reg. at 15,560-61. The EPA defends its refusal to account for malfunctions on the basis of (1) the impracticability of accounting for events that are necessarily unpredictable, and (2) the EPA’s assertion that it will use its prosecutorial discretion to determine on a ease-by-case basis whether an exceedance of emission standards is attributable to an excusable malfunction or whether applicable regulatory penalties should be imposed instead. See No. 11-1108 EPA Br. 38; No. 11-1141 EPA Br. 29. Both sides agree that malfunctions are inevitable in the operation of area and major boilers. According to the EPA, “even equipment that is properly designed and maintained can sometimes fail and ... such failure can sometimes cause an excee-dance of the relevant emission standard.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613; 2011 Area Boilers Rule, 76 Fed. Reg. at 15,561. Thus, the EPA defined a malfunction as a “sudden, infrequent, and not reasonably preventable failure of air pollution control and monitoring equipment, process equipment or a process to operate in a normal or usual manner.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613 (citing 40 C.F.R. § 63.2); 2011 Area Boilers Rule, 76 Fed. Reg. at 15,560 (same). In attempting to write rules to account for emissions, however, the EPA faced an intractable problem: how to account for a malfunction which is, by definition, unpredictable in terms of timing, duration, magnitude, and effect. While the existence of malfunctions is entirely predictable, the nature of those malfunctions is not, and it is the malfunction’s nature that affects emissions and thus is relevant to the application of emission limits. At first glance, the EPA’s chosen approach to malfunctions may seem counter-intuitive, as the Agency appears to have several reasonable alternatives: it could exempt periods of malfunction entirely from the application of the emission standards; or it could apply the standards to malfunctions while giving boiler owners the opportunity to defend against a penalty by demonstrating they were not at fault for the malfunction. But the EPA has previously been stymied in its attempts to implement either of these solutions, as this court has concluded neither approach is consistent with the Agency’s enabling statutes. For instance, in Sierra Club III, the EPA attempted to exempt major sources from complying with emission standards during start up, shut down, and malfunction. See 551 F.3d at 1027-28. This court rejected that approach because the Congress “required that there must be continuous section 112-compliant standards” and so the EPA lacked discretion to exempt certain periods from compliance, regardless of their unpredictability. Id. at 1027. In NRDC III, this court considered a challenge to the affirmative defense provision the EPA adopted for persons defending against civil suits under 42 U.S.C. § 7604(a), which allows “any person” to “commence a civil action on his own behalf’ against any entity alleged to be in violation of an emission standard or limitation. The affirmative defense provision was meant to shield alleged violators from liability for certain emissions violations caused by “unavoidable” malfunctions; under the provision, therefore, “the district court [could] assess penalties only if violators fail[ed] to meet [their] burden of proving all of the requirements in the affirmative defense.” NRDC III, 749 F.3d at 1062 (internal quotation omitted). The court rejected this provision as an impermissible intrusion on the judiciary’s role. See id. at 1063 (“[U]nder this statute, deciding whether penalties are ‘appropriate’ in a given private civil suit is a job for the courts, not for EPA.”). Faced with an obvious dilemma, the EPA arrived at the approach it defends today. Malfunctions receive no special treatment and the EPA instead exercises “its enforcement discretion to address ex-ceedances of emission limits that may be caused by such uncertain, unpredictable events, on a case-by-case basis.” No. 11-1108 EPA Br. 38; see also No. 11-1141 EPA Br. 29. The EPA’s current treatment of malfunctions thus differs from its invalid affirmative defense provision because the Agency is exercising its oum regulatory enforcement power on an ad hoc basis outside the context of citizen suits. When an exceedance occurs during a malfunction, the EPA determines what enforcement action — if any — it should take by considering “the good faith efforts of the source to minimize emissions during malfunction periods, including preventative and corrective actions, as well as root cause analyses to ascertain and rectify excess emissions.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613; see also 2011 Area Boilers Rule, 76 Fed. Reg. at 15,561 (same). The EPA also considers whether the exceedance was in fact “not reasonably preventable” or whether it was “caused in part by poor maintenance or careless operation.” 2011 Major Boilers Rule, 76 Fed. Reg. at 15,613 (citing 40 C.F.R. § 63.2); see also 2011 Area Boilers Rule, 76 Fed. Reg. at 15,561 (same). For our purposes, we need not (indeed, must not) evaluate the policy implications of the EPA’s regulatory choice because our review is confined to determining whether the EPA’s regulation reflects a ' permissible reading of the applicable statute under Chevron. Here, we conclude that it does. The relevant statute requires only that the EPA set “achievable” standards, 42 U.S.C. § 7412(d)(2), and it defines ac-hievability to be no less “than the emission control that is achieved in practice by the best controlled similar source,” 42 U.S.C. § 7412(d)(3). The “best controlled similar source,” however, is unlikely to be a malfunctioning source, and the EPA is bound to enact a standard in keeping with emission limits achieved by that “best controlled similar source.” If anything, then, the statutory language on its face prevents the EPA from taking into account the effect of potential malfunctions when setting MACT emission standards. At the very least, the language permits the EPA to ignore malfunctions in its standard-setting and account for them instead through its regulatory discretion. Our Sierra Club III decision confirms this. See 551 F.3d at 1027-28. Because the EPA had no option to exclude these unpredictable periods, its approach is reasonable. We therefore reject Industry Petitioners’ argument that the EPA either misinterpreted the CAA or acted arbitrarily and capriciously in failing to account for malfunctions when setting MACT floors in the Major and Area Boilers Rules. Nor do we agree with the Industry Petitioners’ secondary argument that the EPA acted arbitrarily and capriciously by failing to set a work-practice or a GACT management-practice standard for malfunction periods. First, the statute makes clear that these kinds of standards are to be set at the discretion of the EPA, so it would be difficult to interpret the statute consistently with its text while holding that the text’s permissive language in fact sets out a requirement that the Agency set work-practice or GACT management-practice standards. As to work-practice standards, “[t]he Administrator may, in lieu [of a numeric standard], promulgate a design, equipment, work practice, or operational standard, or combination thereof,” and any such standard set must “in the Administrator’s judgment [be] consistent with the provisions of subsection (d).” 42 U.S.C. § 7412(h)(1). As to GACT management-practice standards, “the Administrator may ... elect to promulgate” such standards with respect to certain “categories and subcategories 'of area sources.” Id. § 7412(d)(5). It should go without saying that “may means may.” McCreary v. Offner, 172 F.3d 76, 83 (D.C. Cir. 1999) (internal quotations omitted). Second, the Petitioners have not demonstrated and the EPA does not concede that setting work-practice or GACT management-practice standards would even be feasible for periods of malfunction. As for work-practice standards, the EPA would have to conceive of a standard that could . apply equally to the wide range of possible boiler malfunctions, ranging from an explosion to minor mechanical defects. Any possible standard is likely to be hopelessly generic to govern such a wide array of circumstances. Similar problems exist for setting GACT management practices. These management practices would also need to apply to the wide range of possible malfunctions, and the EPA would need to determine that the standard would “reduce emissions of hazardous air pollutants,” an evidence-based standard that is difficult (perhaps impossible) to apply to the unpre- • dictable circumstances of malfunctions. 42 U.S.C. § 7412(d)(5). Thus, we reject the Industry Petitioners’ argument that the EPA was required to set a work-practice or GACT management-practice standard for malfunction periods. In doing so, we are mindful that the EPA is not the only entity able to bring enforcement actions under the CAA, but that private citizens are also empowered to enforce emission standards by filing suit in district court. 42 U.S.C. § 7604(a). Assurances that the EPA will use its prosecutorial discretion to account for malfunctions would mean little if private citizens could seek strict enforcement of those same standards. But as we stated in NRDC III, “the Judiciary, not any executive agency, determines ‘the scope’ — including the available remedies — ‘of judicial power vested by’ statutes establishing private rights of action.” 749 F.3d at 1063 (quoting City of Arlington v. FCC, — U.S. -, 133 S.Ct. 1863, 1871, 185 L.Ed.2d 941 (2013)). Accordingly, in citizen suits under the CAA, “the courts determine, on a case-by-case basis, whether civil penalties are ‘appropriate.’ ” Id. Boiler operators can argue that penalties should not be assessed because of an unavoidable malfunction, and they can support that argument with other relevant facts, “such as the defendant’s ‘full compliance history and good faith efforts to comply.’ ” Id. (quoting 42 U.S.C. § 7413(e)(1)). The EPA can also provide supporting argumentation as intervenor or amicus. Id. Courts should not hesitate to exercise their judicial authority to craft appropriate civil remedies in the case of emissions exceedances caused by unavoidable malfunctions. 2. Periods of Startup, Shutdown, and Malfunction in the CISWI Rule In the CISWI Rule, the EPA made no modification for periods of start-up, shutdown, or malfunction. The Industry Petitioners argue that failing to account for these periods violated the EPA’s statutory instruction to set “achievable” standards. Additionally, the Industry Petitioners claim it was arbitrary and capricious for the EPA to set work-practice standards for startup and shutdown periods under the Major Boilers Rule but not under the CISWI Rule. Both arguments are without merit. First, the EPA’s emission standards for small incinerators do take into account periods of shutdown and startup. The EPA based its standards for these machines on “short term stack tests' for pollutants,” in which incinerators are monitored during the course of normal operation, which includes daily startup and shutdown periods. See 2011 CISWI Rule, 76 Fed. Reg. at 15,738. Thus, startup and shutdown times are already incorporated into the standards the EPA set, and what is more, nearly all pollutants are present in smaller numbers during startup and shutdown anyway, when incinerators are burning fuels alone rather than fuels and solid waste. See Standards of Performance for New Stationary Sources and Emission Guidelines for Existing Sources: Commercial and Industrial Solid Waste Incineration Units (2010 Proposed CISWI Rule), 75 Fed. Reg. 31,938, 31,964 (June 4, 2010). Given this reality, the CISWI Rule satisfies the statutory standard of “achievability” and is not arbitrary and capricious. Second, as to periods of malfunctions, the same analysis applies to the CISWI Rule as applies to the Boilers Rules. The EPA adopted a reasonable interpretation of the CAA when it excluded periods of malfunction from its calculations of aehiev-ability given that malfunction periods are by their very nature unpredictable in terms of their effect on emissions. The EPA’s decision to account for malfunctions in its discretion is likewise a reasonable interpretation of 42 U.S.C. § 7412(d)(2) and (3). For these reasons, we reject the Industry Petitioners’ challenges to the EPA’s regulatory choices with regard to periods of startup, shutdown, and malfunction. B. The Pollutant-By-Pollutant Approach The EPA must look to the performance of the best major boilers and CISWI incinerators when setting MACT floors for a pollutant. As described above, for new units, the EPA must set floors at the level achieved by the best similar unit in each subcategory. For existing units, the Agency must set floors at the level achieved by the best 12 per cent of similar units in each subcategory. 42 U.S.C. §§ 7412(d)(3)(A), 7429(a)(2). As a result, the EPA had to identify the best performing units in each subcategory when setting the MACT floors for the Major Boilers and CISWI Rules. But the EPA often could not identify a single unit or set of units that controlled all HAPs better than the other units in the subcategory. Instead, the EPA sometimes found that a unit might rank among the best in its subcategory at controlling emissions of one HAP, but among the worst at controlling emissions of a different HAP. To address this problem, the EPA adopted a “pollutant-by-pollutant” approach in setting the MACT floors: instead of identifying the unit or units that best 'contro