The State of Texas, numerous energy companies, power plants, steel mills, consumer organizations, state regulators, and a labor union in Texas (collectively, “Petitioners”) challenge the Environmental Protection Agency’s action disapproving Oklahoma’s and Texas’s plans for controlling regional haze and imposing EPA’s own plans instead. Petitioners contend that EPA has acted outside its statutory authority and seek a stay pending review of the rule on the merits. EPA moves to dismiss or transfer the petition because it asserts this court lacks jurisdiction over the petition. Because the Clean Air Act gives jurisdiction over petitions for review to the courts of appeal generally and because the Act’s forum selection clause designates the regional circuit as the appro*411priate venue for this challenge, we DENY EPA’s motion to dismiss or transfer. Because Petitioners have demonstrated a strong likelihood of success on the merits, because they are likely to suffer irreparable injury in the absence of a stay while EPA has not shown similar injury from the issuance of a stay, and because the public interest weighs in favor of a stay, we GRANT the motion for a stay pending resolution of the petitions for review on the merits.
I.
A. The Clean Air Act’s Regulatory Process
The Clean Air Act is “an experiment in cooperative federalism.” Michigan v. EPA, 268 F.3d 1075, 1083 (D.C. Cir. 2001). It “establishes a comprehensive program for controlling and improving the nation’s air quality through state and federal regulation.” BCCA Appeal Grp. v. EPA, 355 F.3d 817, 821-22 (5th Cir. 2003). The Act requires the states and the federal government to set and seek to achieve targets for visibility in protected national parks and wildlife areas by modifying regulations that control air pollutants in ambient air. 42 U.S.C. §§ 7410, 7491, 7492(e)(2). While the federal government has the primary responsibility for identifying air pollutants and setting standards, the states “bear ‘the primary responsibility’ for implementing those standards” by promulgating state implementation plans (“SIPs”). Luminant Generation Co. v. EPA, 675 F.3d 917, 921 (5th Cir. 2012) (quoting BCCA Appeal Grp., 355 F.3d at 822).
The Clean Air Act gives each state “wide discretion in formulating its plan” for achieving the air quality standards set by EPA. Union Elec. Co. v. EPA, 427 U.S. 246, 250, 96 S.Ct. 2518, 49 L.Ed.2d 474 (1976). “[S]o long as the ultimate effect of a State’s choice of emission limitations is compliance with the national standards for ambient air, the State is at liberty to adopt whatever mix of emission limitations it deems best suited to its particular situation.” Train v. Nat. Res. Def. Council, Inc., 421 U.S. 60, 79, 95 S.Ct. 1470, 43 L.Ed.2d 731 (1975).
The Clean Air Act confínes EPA’s role in implementing air quality standards “to the ministerial function of reviewing SIPs for consistency with the Act’s requirements.” Luminant, 675 F.3d at 921. The statute mandates that “the administrator shall approve such [a state implementation plan] as a whole if it meets all of the applicable requirements of this chapter.” 42 U.S.C. § 7410(k)(3) (emphasis added); see also Fla. Power & Light Co. v. Costle, 650 F.2d 579, 587 (5th Cir. 1981) (“The great flexibility accorded the states under the Clean Air Act is ... illustrated by the sharply contrasting, narrow role to be played by EPA.”); Michigan, 268 F.3d at 1083 (EPA’s “overarching role is in setting standards, not in implementation.”). “This division of • responsibility between the states and the federal government ‘reflects the balance of state and federal rights and responsibilities characteristic of our federal system of government.’ ” Luminant, 675 F.3d at 921 (quoting Fla. Power & Light, 650 F.2d at 581). The structure of the Clean Air Act indicates a congressional preference that states, not EPA, drive the regulatory process. As our sister circuit recently observed, “[disagreeing with Congress’s expressly codified policy choices isn’t a luxury administrative agencies enjoy.” Central United Life Ins. Co. v. Burwell, 827 F.3d 70, 73, No. 15-5310, 2016 WL 3568084, at *2 (D.C. Cir. July 1, 2016).
*412Only if the state has not complied with the requirements of the Clean Air Act does EPA assume the role of primary regulator by drafting a state-specific plan. At that point, after disapproving a state implementation plan, EPA has two years to promulgate a federal implementation plan (“FIFO. 42 U.S.C. § 7410(c)(1). EPA promulgates the federal implementation plan “to fill all or a portion of a gap ... in a State implementation plan.” 42 U.S.C. § 7602. As a result, EPA’s obligations and authority to promulgate the federal implementation plan are the same the state had when promulgating its implementation plan. See, e.g., 77 Fed. Reg. 40,150, 40,164 (July 6, 2012) (“At the point EPA becomes obligated to promulgate a FIP, EPA steps into the State’s shoes, and must meet the same requirements.... ”).
Within this framework, one provision of the Clean Air Act requires EPA and the states to jointly act to improve visibility at certain protected federal lands. 42 U.S.C. § 7491. EPA’s obligations under this provision begin with identifying the federal lands that need improved visibility. 42 U.S.C. § 7491(a)(2); 40 C.F.R. §§ 81.400-81.437. After EPA has identified areas for targeted haze reduction, the Act requires EPA to write regulations providing the guidelines that states will use to design state implementation plans to reduce haze in the affected areas. 42 U.S.C. § 7491(b)(1), (2). In 1999, EPA promulgated the Regional Haze Rule. 40 C.F.R. § 51.308; 64 Fed. Reg. 35,714 (July 1, 1999). The Regional Haze Rule established the guidelines for state compliance with the air visibility requirements of § 7491.1
The Regional Haze Rule requires five elements in a state implementation plan. For each affected wilderness and national park, the plan must: (1) set “reasonable progress goals” toward achieving natural visibility conditions that ensure improvements in visibility on the most impaired days over the period of the implementation plan; (2) calculate baseline visibility and natural visibility conditions; (3) devise a long-term strategy with enforceable emissions limitations, compliance schedules, and other measures necessary to achieve the reasonable progress goals; (4) develop a monitoring strategy for measuring and reporting visibility; and (5) list the best available retrofit technology (“BART”) that emission sources in the state will have to adopt to achieve the visibility goals, along with a schedule for implementing BART. 40 C.F.R. § 51.308(d), (e).
The Regional Haze Rule also prescribes how states may calculate their reasonable progress goals. A state begins by calculating the steady linear rate of decreasing emissions that would achieve natural visibility in the covered wildernesses and national parks by the year 2064. 40 C.F.R. § 51.308. If a state determines that the linear rate would result in unreasonable regulations, it must propose an alternative set of reasonable progress goals and demonstrate both that the linear rate is unreasonable and that the alternative goals are reasonable. Id. § 51.308(d)(l)(ii). The Clean Air Act and the Regional Haze Rule require a state to consider four factors when setting reasonable progress goals: “the costs of compliance, the time necessary for compliance, and the energy and nonair quality environmental impacts of compliance, and the remaining useful life of any existing source subject to such re*413quirements.” 42 U.S.C. § 7491(g)(1); see also 40 C.F.R. § 61.308(d)(l)(i)(A) (repeating the factors listed in § 7491(g)(1)).
BART is the only portion of the implementation plan that is enforced against emission sources in a state. See 64 Fed. Reg. at 35,733 (“Once a State has adopted a reasonable progress goal and determined what progress will be made toward that goal over a 10-year period, the goal itself is not enforceable. All that is ‘enforceable’ is the set of control measures which the State has adopted to meet that goal.”).
If emissions in one state are anticipated to impact the visibility conditions in protected areas in another state, the Regional Haze Rule requires the states to consult with one another and develop a coordinated emission strategy. 40 C.F.R. § 51.308(d)(3). States may participate in regional planning organizations that jointly approve the technical analyses assessing the cross-border impact of emissions. Id.
[[Image here]]
B. The Texas and Oklahoma SIPs
The rulemaking under challenge here concerns visibility in two national parks and one federal wildlife refuge.2 The Wich-' ita Mountains Wildlife Refuge in southwestern Oklahoma was established in 1901 and protected by statute in 1905. 16 U.S.C. § 684. The refuge is home to scores of bird, mammal, and fish species, including reintroduced populations of turkey, bison, and elk.3 Big Bend National Park was established in 1935 with lands in southwest Texas donated by the state to the federal government. 16 U.S.C. § 156. Today the park spans over 801,000 acres of desert, mountain, and river valley and is home to cultural, geological, and biological treasures of the United States.4 Guadalupe Mountains National Park in west Texas is the world’s premier example of a fossilized reef from the Permian Era, more than 250 million years ago.5 Now it is home to 60 species of mammals, 289 species of birds and 55 species of reptiles including black bear, grey fox, porcupine, mountain short-horned lizard, and mountain lion.6
The Regional Haze Rule requires states to develop an implementation plan for the *414period from 2009-2018 and to submit revised plans for each ten-year period thereafter. 40 C.F.R. § 51.308(b), (f).7 Initial state implementation plans were due December 17, 2007. 40 C.F.R. § 51.308(b). In January 2009, EPA found that thirty-seven states, including Texas and Oklahoma, had missed the deadline. 74 Fed. Reg. 2392-01 (Jan. 15, 2009). This finding triggered a two-year deadline for EPA to promulgate a federal implementation plan. 42 U.S.C. § 7410(c)(1). Before EPA drafted its federal implementation plan, Texas submitted a state implementation plan on March 31, 2009. More than three years later — and three years into the ten-year window for this round of implementation plans — EPA issued a limited disapproval of Texas’s plan. 77 Fed. Reg. 33,642, 33,653 (June 7, 2012).8 Oklahoma first submitted its state implementation plan in February 2010. EPA partially disapproved its plan in 2011. 76 Fed. Reg. 81,728 (Dec. 28, 2011).
In 2014, five years after receiving Texas’s SIP and four years after receiving Oklahoma’s SIP, EPA proposed a federal implementation plan to replace the parts of the Texas and Oklahoma state implementation plans that EPA found deficient. 79 Fed. Reg. 74,818 (Dee. 16, 2014). Finally, in 2016 — nearly seven years after Texas submitted its implementation plan and nearly six years after Oklahoma submitted its implementation plan — EPA promulgated a final rule (“the Final Rule”) partially approving and partially disapproving the Texas and Oklahoma plans and replacing portions of them with a federal implementation plan. 81 Fed. Reg. 296 (Jan. 5, 2016). The Final Rule imposes federal reasonable progress goals for wildlife refuges and national parks in Texas and Oklahoma but only requires emission controls in Texas. No Oklahoma power plants or emission sources are affected.
Texas’s state implementation plan included each of the five elements required by the Regional Haze Rule. Texas concluded that the linear rate required to achieve natural visibility by 2064 was unreasonable and set an alternative series of reasonable progress goals. It coordinated with eight other states — Louisiana, Oklahoma, Arkansas, Kansas, Minnesota, Missouri, Nebraska, and Iowa — through the Central Regional Air Planning Association (“CEN-RAP”) and used CENRAP analysis to assess the impact of Texas emissions on protected areas in other states. CENRAP modeled visibility estimates for 2018 (the close of the current SIP window) and compared the 2018 estimates to the linear rate of progress. Texas determined that cur*415rent actual visibility conditions in the covered areas were already better than the reasonable progress goals it set for 2018. 81 Fed. Reg. at 341; 79 Fed. Reg. at 74,887 (finding that measured visibility already exceeds reasonable progress goals under both Texas’s state implementation plan and EPA’s alternative federal implementation plan). Texas considered various emission control technologies and concluded that any additional technologies would impose more costs than benefits.
EPA partially approved and partially disapproved Texas’s proposed plan. 81 Fed. Reg. at 298-99. EPA approved Texas’s measurements of current visibility conditions in Big Bend, Guadalupe Mountains, and Wichita Mountains. These measurements show that visibility at all three wildernesses is already better than the reasonable progress goals in both the disapproved Texas plan and the replacement federal implementation plan.
[[Image here]]
EPA agreed that the linear rate necessary to achieve natural visibility by 2064 was unreasonable but disapproved Texas’s alternative reasonable progress goals on the grounds that Texas’s analysis “was not appropriately refined, targeted, or focused on those sources having the most significant and potentially cost-effective visibility benefits.” 81 Fed. Reg. at 299. EPA also disagreed with Texas’s calculation of natural visibility conditions while approving Texas’s calculation of current baseline visibility. EPA based its disapproval on a disagreement over the amount of dust that is naturally occurring in the protected regions. 81 Fed. Reg. at 300. EPA further disapproved Texas’s long-term strategy, arguing that Texas’s failure to conduct source-specific analysis also invalidated this portion of the Texas plan. Id. at 302.
EPA also, in the same rulemaking, disapproved Oklahoma’s plan. EPA disapproved only the reasonable progress goals in the Oklahoma plan and did so exclusively because of the effects of Texas emissions on the Wichita Mountains Wildlife Refuge in Oklahoma. Id. EPA argued that Oklahoma’s consultation with Texas was “flawed” because Texas “denied [Oklahoma] the knowledge it needed — the extent to which cost-effective controls were available for those sources or groups of sources in Texas with the greatest potential to impact visibility at the Wichita Mountains — in order to properly construct its reasonable progress goal for the Wichita Mountains.” Id. Essentially, the only flaw in Oklahoma’s plan springs from EPA’s conclusion that Texas was required to conduct a source-specific analysis and impose restrictions on specific sources in *416Texas.11
EPA then imposed a federal implementation plan for Texas and Oklahoma. 81 Fed. Reg. at 303-07. It analyzed emissions individually by source and selected fifteen electrical generating units with the greatest impact on the protected areas. The federal implementation plan required specific emission controls only for these specific sources without consideration of other controls at other sources. 81 Fed. Reg. at 304-05. EPA demanded scrubber upgrades at eight facilities and scrubber retrofits at an additional seven facilities with installation deadlines of 2019 and 2021. 81 Fed. Reg. at 305.
By the time EPA promulgated the Final Rule, only two years remained in the 2008-18 regulatory window. After promulgating the Final Rule, EPA issued a notice of proposed rulemaking indicating that it would amend the Regional Haze Rule and change the governing standards for the second and all subsequent ten-year planning periods. 81 Fed. Reg. 26,942 (May 4, 2016) (proposing revisions to 40 C.F.R. § 51.308(f)).
The power companies, labor unions, consumer groups, state regulatory agencies, steel manufacturers, and state of Texas petitioned for review of the Final Rule.12 Petitioners argue that EPA, under the guise of requiring imperceptible haze reductions, has actually targeted coal-fired power plants. According to Petitioners, the proposed changes at the targeted power plants would cost $2 billion, rendering them uneconomical and forcing the plants to close. The association that manages the Texas power grid, ERCOT, concluded that EPA’s proposal, which became the Final Rule, would close plants and remove 3,000 MW to 8,400 MW of generating capacity in Texas. Petitioners further argue that it would be cost-prohibitive to reopen the affected plants even if courts ultimately conclude that EPA acted unlawfully. Highlighting this installment deadlines of 2019 and 2021, Petitioners argue that EPA’s plan produces no benefits in the period it is intended to .cover (2009-2018) while imposing significant compliance costs.
Petitioners also argue that closure of the plants would substantially threaten grid reliability in Texas. Electricity rates would increase for Texas consumers and businesses. Power companies in the state would need to undertake costly construction projects to create new transmission infrastructure that would supply power to central Texas. Until infrastructure projects are completed or new electrical generation facilities are opened in the state, Petitioners argue that Texas could face power shortages and grid failures because the *417federal implementation plan does not include any exemption for grid reliability.
While the petitions for review are pending, Petitioners request a stay of the Final Rule to avoid the irreparable damage they assert the federal implementation plan would impose. EPA has filed a motion to dismiss or transfer this petition for review to the D.C. Circuit pursuant to the Clean Air Act’s mandate that an EPA ruling “based on a determination of nationwide scope or effect” can only be brought in the D.C. Circuit. 42 U.S.C. § 7607(b)(1). Petitioners argue, by contrast, that the Final Rule is only “locally or regionally applicable” and therefore cannot be brought in the D.C. Circuit. Id.
II.
EPA moved to dismiss or transfer this petition for review arguing that jurisdiction lies only in the D.C. Circuit and that this court lacks the power to consider Petitioners’ challenge. Petitioners respond that jurisdiction and venue are appropriate in this court.
A.
Section 7607(b)(1) of the Clean Air Act delineates the appropriate forum for petitions for review. It provides, in relevant part, that:
A petition for review of ... any nationally applicable regulations promulgated, or final action taken, by the Administrator under this chapter may be filed only in the United States Court of Appeals for the District of Columbia. A petition for review of the Administrator’s action in approving or promulgating any implementation plan ... or any other final action of the Administrator under this chapter ... which is locally or regionally applicable may be filed only in the United States Court of Appeals for the appropriate circuit. Notwithstanding the preceding sentence a petition for review of any action referred to in such sentence may be filed only in the United States Court of Appeals for the District of Columbia if such action is based on a determination of nationwide scope or effect and if in taking such action the Administrator finds and publishes that such action is based on such a determination.
42 U.S.C. § 7607(b)(1); see also Texas v. EPA, No. 10-60961, 2011 WL 710598, at *3 (5th Cir. Feb. 24, 2011). Because “the determination of our jurisdiction is exclusively for the court to decide,” we do not defer to the agency’s interpretation of this section. Lopez-Elias v. Reno, 209 F.3d 788, 791 (5th Cir. 2000); see also Exelon Wind 1, L.L.C. v. Nelson, 766 F.3d 380, 392 (5th Cir. 2014) (“‘[T]he courts, however, have to make their own determination [about] jurisdiction, rather than defer to the [federal agency] in the first instance.’ ” (first alteration added) (quoting Reeb v. Econ. Opportunity Atlanta, Inc. 516 F.2d 924, 926 (5th Cir. 1975))).13 Nor *418do we defer to the agency’s interpretation when determining venue. See Smith v. Aegon Co. Pension Plan, 769 F.3d 922, 928 (6th Cir. 2014).
Section 7607(b)(1) divides challenges into three general categories. Petitions for review of nationally applicable actions may only be filed in the D.C. Circuit. Petitions for review of locally or regionally applicable actions may only be filed in the regional circuit courts of appeal.14 Petitions for review of locally or regionally applicable actions based on a determination that has nationwide scope or effect may only be filed in the D.C. Circuit. Texas, 2011 WL 710598, at *3.
EPA argues that the division between the three groups is jurisdictional. In prior cases, we have treated § 7607(b)(1) as a venue provision without deciding whether it is jurisdictional. Texas, 2011 WL 710598, at *3 n.28. Here, the parties continue to dispute whether this provision governs jurisdiction or venue and, therefore, dispute whether we have power to consider the motion for a stay. We conclude that § 7607(b)(1) is a two-fold provision. First, it is a “conferral of jurisdiction upon the courts of appeals.” Harrison v. PPG Indus., Inc., 446 U.S. 578, 593, 100 S.Ct. 1889, 64 L.Ed.2d 525 (1980). Second, its allocation of petitions between the regional circuits and the D.C. Circuit delineates the appropriate venue for challenges.15
This reading accords with the text of the statute, which empowers both the D.C. Circuit and the regional circuits. The statute is not framed as a limitation on the power of the courts but as an instruction to petitioners. Our two-fold reading comports with that of our sister circuit. Dalton Trucking, Inc. v. EPA, 808 F.3d 875, 878-80 (D.C. Cir. 2015) (holding that § 7607(b)(1) confers jurisdiction on all courts of appeal and divides venue among them).
B.
Having concluded that § 7607(b)(l)’s three categories delineate venue, we now conclude that venue is appropriate in the Fifth Circuit because the present challenge addresses a “locally or regionally applicable” action, which is not based on a determination that has nationwide scope or effect.
*419Section 7607(b)(1) is not a model of statutory clarity. Interpreting it requires close attention to detail. We begin by defining the significant statutory terms. Section 7607(b)(1) categorizes petitions for review according to the “action” under challenge. The “action” is the rule or other final action taken by the agency that the petitioner seeks to prevent or overturn. Here, the “action” in question is EPA’s final rule disapproving portions of the Texas and Oklahoma SIPs and imposing a FIP. Section 7607(b)(1) then looks to the “determination” that the challenged action is “based on.” These determinations are the justifications the agency gives for the action and they can be found in the agency’s explanation of its. action. They are the reason the agency takes the action that it does. For example, because EPA determined that Texas failed to appropriately calculate reasonable progress goals, EPA disapproved Texas’s reasonable progress goals and imposed its own. EPA’s determination about Texas’s goals provided the basis for EPA’s action. Because the statute speaks of the determinations the action “is based on,” the relevant determinations are those that lie at the core of the agency action. Merely peripheral or extraneous determinations are not relevant — the agency should identify the core determinations in the action. Last, § 7607(b)(1) speaks of a “finding]” and publication by the Administrator. This finding is an independent, post hoc, conclusion by the agency about the nature of the determinations; the finding is not, itself, the determination. The finding must be published as part of the action itself.
Section 7607(b)(1) categorizes petitions for review according to the nature of the action that the petition challenges. The statute separates petitions for review of nationally applicable actions from petitions for review of locally or regionally applicable actions. The question of applicability turns on the legal impact of the action as a whole. See, e.g., Texas, 2011 WL 710598, at *3 (action disapproving SIPs from thirteen widely dispersed states and issuing SIP call requiring states to submit new SIPs is nationally applicable); Am. Road & Transp. Builders Ass’n v. EPA, 705 F.3d 458, 455-56 (D. C. Cir. 2013) (action disapproving California SIP is locally applicable and must be filed in Ninth Circuit); ATK Launch Sys., Inc. v. EPA, 651 F.3d 1194, 1199 (10th Cir. 2011) (“[SIPs are] an undis-putably regional action [and] the nature of the regulation ... controls.”). The parties agree that the Final Rule under review is a locally or regionally applicable action.16
Next, § 7607(b)(1) subdivides challenges to locally or regionally applicable actions. The default presumption is that petitions for review of locally or regionally applicable actions “may only be filed in the United States Court of Appeal for the appropriate circuit.” 42 U.S.C. § 7607(b)(1). The statute creates an exception, however, for actions “based on a determination of nationwide scope or effect.” Id. If a challenged action is based on such a determination, § 7607(b)(1) gives the Administrator the discretion to move venue *420to the D.C. Circuit by publishing a finding declaring the Administrator’s belief that the action is based on a determination of nationwide scope or effect. EPA argues that the Administrator’s finding-alone is conclusive, but this does not comport with the statutory text. Section 7607(b)(l)’s exception lists two criteria: “(1) if such action is based on a determination of nationwide scope or effect and (2) if in taking such action the Administrator finds and publishes that such action is based on such a determination.” 42 U.S.C. § 7607(b)(1) (emphasis and numerals added). Because these criteria are listed “if ... and if ...,” both criteria must be satisfied to transfer venue from the appropriate regional circuit to the D.C. Circuit.
To determine whether the exception applies, we must answer two questions: (1) is the action based on a determination that has nationwide scope or effect; and (2) did the Administrator publish an adequate finding?17 EPA argues that the court has no role in assessing whether the action is based on a determination of nationwide scope or effect.18 EPA, however, cannot point to any part of § 7607(b)(1) giving it such exclusive authority.19 Rather, the statute provides" a clear metric by which a court can assess the scope or effect of the relevant determinations. The reviewing court merely asks whether the scope20 or effect21 of the determinations is nationwide.22
Numerous cases demonstrate that courts must assess the “applicability” of the action. See, e.g., Texas, 2011 WL 710598, at *3 (challenged action is nationally applicable); Am. Road & Transp. Builders Ass’n v. EPA, 705 F.3d 453, 455-56 (D. C. Cir. 2013) (SIP is locally or regionally applicable); ATK Launch Sys., 651 F.3d at 1199 (SIP call is nationally applicable); Madison Gas & Elec. Co. v. EPA, 4 F.3d 529, 530-31 (7th Cir. 1993) (EPA regulation allocating emissions allowances to listed power plants is locally or *421regionally applicable). Courts make this decision by asking whether the action is “nationally applicable” or “locally or regionally applicable.” Assessing whether a determination’s scope or effect is “nationwide” requires a similar judicial inquiry.
We consider whether the first condition is satisfied by assessing the scope or effect of the determinations underlying the challenged action. Petitioners suggest that we must make our inquiry as to the scope or effect of the determinations independent from EPA’s finding because the inquiry governs the powers of the court rather than those of the agency. EPA, by contrast, argues that we review whether its finding is arbitrary or capricious. We agree with Petitioners. The exception for locally or regionally applicable actions based on a determination of nationwide scope or effect has two conditions. First the action must be based on such a determination and second the agency must so find and publish. The statute gives EPA discretion to transfer venue only if the first condition is also satisfied. Because the answer to the first condition controls the role of the court, we are persuaded that we must make an independent assessment of the scope of the determinations just as we make an independent assessment of the applicability of the action. See Exelon Wind, 766 F.3d at 392; Texas, 2011 WL 710598, at *3-*4 (conducting an independent inquiry into the applicability of SIP call); ATK Launch Sys., 651 F.3d at 1196—97 (same).
Addressing the first condition de novo, we conclude that the Final Rule is not based on any determinations that have nationwide scope or effect. EPA based its disapproval of the Texas and Oklahoma SIPs and its FIP on a number of intensely factual determinations.23 These determinations all related to the particularities of the emissions sources in Texas and the confluence of factors impacting visibility at two locations in Texas and one in southwest Oklahoma.24
*422Even assuming, arguendo, that our task is to review the agency’s conclusion about the scope or effect of the determinations— rather than to make our own independent inquiry — and ask only whether the agency’s finding is arbitrary or capricious,25 we are not persuaded that the Final Rule is based on a determination of nationwide scope or effect.26 EPA could not identify with particularity the determinations with nationwide scope or effect that formed the basis of the Final Rule. EPA advanced two general arguments in support of its position that the Final Rule is based on a determination of nationwide scope or effect. These arguments are legal in nature and do not rely on EPA’s factual conclusions or expertise. First, EPA argues that because the Final Rule disapproves SIPs from states in two different circuits, it has nationwide scope or effect. Second, EPA argues that its rejection of the Texas and Oklahoma reasonable progress goals and long-term strategies relied on novel interpretations of the Regional Haze Rule and that the novel interpretation will guide other states designing future SIPs.
EPA’s first argument is unpersuasive and improperly focuses on the nature of the rule as a whole and not on the determinations on which the Final Rule is based. It speaks to applicability of the rule, not to the scope or effect of the relevant determinations.27
*423EPA’s prior actions also call into question EPA’s assertion here that the involvement of two circuits gives nationwide scope or effect to the relevant determinations. While § 7607(b)(1) gives EPA discretion to make and publish a finding in certain actions while declining' to do so in other similar actions, general principles of administrative law demand that EPA explain its basis for doing so. See FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515, 129 S.Ct. 1800, 173 L.Ed.2d 738 (2009) (requiring agency to “provide reasoned explanation for its action”).
In another SIP assessment for Michigan and Minnesota — neighboring states falling in different circuits — EPA did not argue that determinations of nationwide scope or effect existed even though the plans regulated facilities in both states. Approval and Promulgation of Regional Haze Implementation Plan for States of Minnesota and Michigan, 78 Fed. Reg. 8706, 8733 (Feb. 6, 2013) (petitions for judicial review of federal implementation plans “must be filed in the United States Court of Appeals for the appropriate circuit.”); see also Disapproval of Interstate Transport Requirements for the 2008 Ozone NAAQS, 81 Fed. Reg. 38,957-01 (June 15, 2016) (disapproving SIPs for Ohio (6th Cir.) and Indiana (7th Cir.) without any suggestion of nationwide scope or effect). EPA may treat different actions differently but EPA must provide an explanation for its varying treatment. While EPA’s briefing advanced explanations for treating the Michigan and Minnesota approval differently from the Texas and Oklahoma disapproval, no such explanation appears in the Final Rule.
Nor are we persuaded by EPA’s second argument that whatever precedential ef-feet the Final Rule has shows that it is based on determinations with nationwide scope or effect. Specifically, EPA argues that its determination that Oklahoma and Texas conducted insufficient discussions about the impact of Texas emissions on the Wichita Mountains Wildlife Refuge in Oklahoma will provide guidance to future interstate consultations. This argument focuses on the appropriate level — the scope or effect of the determinations that are the basis of the Final Rule — but it too is unconvincing. To begin with, the argument sweeps too broadly. All SIPs are likely informed by EPA’s assessment of SIPs from other states. It would reverse § 7607(b)(l)’s presumption that review of implementation plans should take place in regional circuits if the guidance one SIP approval provides another state necessarily gave nationwide scope or effect to EPA’s determinations. See Am. Road & Transp. Builders, 705 F.3d at 456.
Furthermore, EPA’s argument that a SIP disapproval like the Final Rule provides interpretive guidance with nationwide scope or effect is undermined by EPA’s own actions and arguments. In the Final Rule itself, EPA responded to com-menters who alleged inconsistency with other SIP approvals by arguing that its regulations “do not require uniformity between ... actions in all circumstances and instead ‘allow for some variation’ in actions taken in different regions.” 81 Fed. Reg. at 326 (quoting Amendments to Regional Consistency Requirements, 80 Fed. Reg. 50,250, 50,258 (Aug. 19, 2015)). Some variation is to be expected because SIP approvals or disapprovals are highly fact-dependent actions. As. EPA itself insisted during oral argument, the agency’s particular in*424terpretations of the Regional Haze Rule in the Final Rule cannot be divorced from the particular factual context of the Texas and Oklahoma plans.
EPA’s practices confirm that SIP analy-ses are contextual and variable. Two years ago, EPA approved a New Mexico SIP with a less stringent reasonable progress goal for the monitor at the Guadalupe Mountains than the reasonable progress goal in the disapproved Texas SIP.28 Had Texas looked to New Mexico’s SIP for guidance, it would have been misled precisely because - the SIP process is fact-intensive and will vary from state to state.
Finally, as a practical matter, the determinations in the Final Rule will have no nationwide precedential scope or effect because every other state has already submitted its SIP for the 2008-18 round. Nor will the Final Rule impact SIPs considered in the next round because EPA has proposed revisions of the exact portions of the Regional Haze Rule that EPA claims to have definitively interpreted in the Final Rule. Protection of Visibility: Amendments to Requirements for State Plans, 81 Fed. Reg. 26,942, 26,952 (May 4, 2016). If the Final Rule has the impact on other SIPs that EPA argues it has, no revision of the Regional Haze Rule would be needed.
Because the Final Rule is a locally or regionally applicable action, the default presumption of § 7607(b)(1) requires review in this circuit. Because the action is not based on any determinations that have nationwide scope or effect, the exception to the default presumption does not apply. Therefore, review is appropriate here, in the regional circuit.
III.
We consider four factors when deciding whether to grant a stay pending appeal: “(1) whether the stay applicant has made a strong showing that he is likely to succeed on the merits; (2) whether the applicant will be irreparably injured absent a stay; (3) whether issuance of the stay will substantially injure the other parties interested in the proceeding; and (4) where the public interest lies.” Nken v. Holder, 556 U.S. 418, 426, 129 S.Ct. 1749, 173 L.Ed.2d 550 (2009) (quoting Hilton v. Braunskill, 481 U.S. 770, 776, 107 S.Ct. 2113, 95 L.Ed.2d 724 (1987)). “A stay is not a matter of right, even if irreparable injury might otherwise result” to the appellant. Id. at 433, 129 S.Ct. 1749 (quoting Va. Ry. Co. v. United States, 272 U.S. 658, 672, 47 S.Ct. 222, 71 L.Ed. 463 (1926)). Similarly situated petitioners challenging other federal implementation plans under the Regional Haze Rule have often obtained stays. See, e.g., Oklahoma v. EPA, 723 F.3d at 1206-07 (staying implementation of rule imposing $1.2 billion in costs); Wyoming v. EPA, Nos. 14-9529, 14-9530, 14-9533, 14-9534 (10th Cir. Sept. 9, 2014) (staying and tolling compliance deadlines for rules imposing $700 million in costs); Cliffs Nat. Res. Inc. v. EPA, Nos. 13-1758, 13-1761 (8th Cir. June 14, 2013) (staying and tolling compliance deadlines for rules imposing $200 million in costs).
A.
To show a strong likelihood of success on the merits, Petitioners must demonstrate that EPA acted arbitrarily, capri*425ciously, or unlawfully.29 Because the BART requirements&emdash;the portion of the Final Rule imposing injury on Petitioners&emdash;flow from the federal implementation plan, and because EPA only has the power to promulgate a federal implementation plan if it disapproves the state implementation plan, Petitioners can demonstrate a strong likelihood of success on the merits either by showing that EPA acted unlawfully when it disapproved the Texas and Oklahoma SIPs or that EPA acted unlawfully when it drafted the FIP.
The Clean Air Act permits a reviewing court to invalidate any action taken by EPA that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with the law ... in excess of statutory jurisdiction, authority, or limitations, or short of statutory right; or without observance of procedure required by law....” 42 U.S.C. § 7607(d)(9); see Luminant, 675 F.3d at 925 (standard of review of Clean Air Act actions tracks standards provided by Administrative Procedure Act, 5 U.S.C. § 706). Agency action:
is arbitrary and capricious “if the agency has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.”
Tex. Oil & Gas Ass’n v. EPA, 161 F.3d 923, 933 (5th Cir. 1998) (quoting Motor Vehicle Mfrs. Ass’n v. State Farm, 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983)). “We must disregard any post hoc rationalizations of EPA’s action and evaluate it solely on the basis of the agency’s stated rationale at the time of its decision.” Luminant, 675 F.3d at 925; see Motor Vehicle Mfrs. Ass’n, 463 U.S. at 50, 103 S.Ct. 2856 (“It is well-established that an agency’s action must be upheld, if at all, on the basis articulated by the agency it-sélf.”); Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168-69, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962) (“The courts may not accept appellate counsel’s post hoc rationalizations for agency action; Chenery requires that an agency’s discretionary order be upheld, if at all, on the same basis articulated in the order by the agency itself.” (citing SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1760, 91 L.Ed. 1995 (1947))).
We review factual findings to determine if they are supported by substantial evidence, but legal conclusions are reviewed de novo. Bd. of Miss. Levee Comm’rs v. EPA, 674 F.3d 409, 417 (5th Cir. 2012). Because EPA’s disapproval of Texas’s plan occurred in the form of a rulemaking and has the force of law, our assessment of EPA’s disapproval is deferential to EPA’s interpretation of the Clean Air Act if the statute is susceptible to multiple reasonable interpretations. See United States v. Mead Corp., 533 U.S. 218, 226-27, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001); Chevron, U.S.A., Inc. v. Nat. Res. Def. Council, 467 U.S. 837, 844, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984).
*426If EPA’s action is not permitted by the statute, we must overturn the action. 5 U.S.C. § 706(2) (“[T]he reviewing court shall ... hold unlawful and set aside agency action ... found to be ... arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”); Food and Drug Admin. v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 125, 120 S.Ct. 1291, 146 L.Ed.2d 121 (“Regardless of how serious the problem an administrative agency seeks to address, ... it may not exercise its authority ‘in a manner that is inconsistent with the administrative structure that Congress enacted into law.’” (quoting ETSI Pipeline Project v. Missouri, 484 U.S. 495, 517, 108 S.Ct. 805, 98 L.Ed.2d 898 (1988))).
Petitioners assert two grounds on which EPA’s disapproval of Texas’s state implementation plan' was unlawful: (1) that EPA exceeded its powers when it disapproved Texas’s reasonable progress goals and the resulting long-term strategy despite their compliance with the Clean Air Act; (2) that EPA acted arbitrarily and capriciously when it disapproved Texas’s consultation with Oklahoma. Petitioners also assert three independent grounds on which EPA’s alternative federal implementation plan is unlawful; (3) that the federal plan impermissibly relied on effects outside the ten-year regulatory window in requiring emission controls; (4) that the federal plan failed to adequately consider costs as required by Michigan v. EPA, — U.S. -, 135 S.Ct. 2699, 2707, 192 L.Ed.2d 674 (2015); and (5) that the federal plan failed to consider the effects on grid reliability in Texas.
We now turn to the particular challenges Petitioners raise regarding EPA’s disapproval of Texas’s implementation plan and EPA’s alternative federal implementation and consider whether Petitioners have a strong likelihood of success on the merits.
1. Reasonable Progress Goals
Petitioners have a strong likelihood of success in showing that EPA exceeded its statutory authority by disapproving the Texas and Oklahoma reasonable progress goals even though the goals complied with the Clean Air Act’s standards. EPA cannot base disapproval on any requirements other than those listed in the Clean Air Act because EPA has “no authority to question the wisdom of a State’s choices of emission limitations if they are part of a plan which satisfies the standards of [§ 7410(a)(2)].” Train, 421 U.S. at 79, 95 S.Ct. 1470. “[T]he Agency may devise and promulgate a specific plan of its own only if a State fails to submit an implementation plan which satisfies those standards.” Id.
The Clean Air Act imposes a multi-step process for setting visibility targets. States begin by estimating natural visibility conditions at the protected federal lands. The state then assesses the changes necessary to achieve natural visibility by 2064. If those changes impose unreasonable costs, the state must devise alternative reasonable progress goals for the close of the current regulatory window.
The Regional Haze Rule grants states considerable flexibility when they estimate natural conditions, requiring that they “estimate] the degree of visibility impairment existing under natural conditions for the most impaired days and least impaired days, based on available monitoring information and appropriate data analysis techniques....” 40 C.F.R. § 51.308(d)(2)(iii). EPA’s natural visibility guidance expressly permits states to use refined approaches for the calculation and to “identify [other approaches] that are more appropriate for *427their own situations.”30 Yet EPA disapproved Texas’s refined estimates for Big Bend and Guadalupe Mountains because Texas assumed that natural factors such as dust storms and wildfires contributed 100% of coarse mass and soil to the air on the 20% of days with the most visibility impairment. EPA devised its own estimate using an 80% contribution from natural factors. Petitioners argue that EPA provided scant justification for its substitution other than to allege that Texas’s estimate was inadequately supported. 81 Fed. Reg. at 325.31
Once a state has estimated natural visibility conditions, the Regional Haze Rule requires the state to calculate the changes necessary to achieve natural visibility by 2064. If those changes are unreasonable, a state is required to set reasonable progress goals that ensure more gradual progress toward natural visibility conditions. Both EPA and Texas agree that the regulatory changes necessary to achieve natural visibility by 2064 are unreasonable. 81 Fed. Reg. 299. As required by the Clean Air Act and the Regional Haze Rule, Texas’s state implementation plan proposed an alternative set of reasonable progress goals.
Texas considered emissions from a broad range of sources in Texas and conducted a holistic analysis of emissions controls for this range of sources. EPA’s disapproval asserted that this approach was unreasonable and instead substituted its own source-specific analysis examining the particular costs and benefits of emissions from particular power plants. 81 Fed. Reg. 298-99. Texas set 2018 reasonable progress goals of 16.6 dv, 16.3 dv, and 21.47 dv for Big Bend, Guadalupe Mountains, and Wichita Mountains, respectively. EPA’s replacement reasonable progress goals instead demanded 16.57 dv, 16.26 dv, and 21.33 dv. 81 Fed. Reg. at 306-07; 79 Fed. Reg. at 74,887 tbl. 43. EPA’s reasonable progress targets are less than 1% lower than the Texas goals that EPA found inadequate and current visibility conditions are already better than the targets set in either Texas’s or EPA’s reasonable progress goals. EPA’s reasonable progress goals would require a number of costly changes including: installation of sulfur dioxide scrubbers at seven electrical generating units and upgrades of existing scrubbers at seven other electrical generating units. 81 Fed. Reg. at 298.
EPA disapproved both Texas’s and Oklahoma’s goals by arguing that Texas incorrectly weighed the four statutory factors that govern the development of reasonable progress goals. See 42 U.S.C. § 7491(g)(1) (“costs of compliance, the time necessary for compliance, and the energy and nonair quality environmental impacts of compliance, and the remaining useful life of any existing source subject to such requirements”). EPA argues that it had several grounds for disapproving the Texas and Oklahoma goals and suggests each alone provides a sufficient basis for the disapproval. Most of these “independent” grounds boil down to EPA’s insistence that Texas should have conducted a source-specific requirement. Other grounds for disapproval were asserted in *428the proposed rule but were not finalized in the Final Rule. Compare 79 Fed. Reg. at 74,842-43 (proposing disapproval because of Texas’s cost threshold, weighing of factors for individual sources, reliance on CAIR reductions, assumptions about efficiency of S02 scrubbers, evaluation of potential improvements, order of magnitude estimate, and scrubber upgrade estimates), with 81 Fed. Reg. at 298-300 (finalizing disapproval because of lack of source-specific analysis and estimation of natural visibility conditions).
EPA’s requirement that Texas conduct a source-specific analysis is not supported by the Clean Air Act or the Regional Haze Rule. As our sister circuit held, “[n]either the Clean Air Act nor the Regional Haze Rule requires source-specific analysis in the determination of reasonable progress goals.” Wildearth Guardians v. EPA, 770 F.3d 919, 944 (10th Cir. 2014). Indeed, EPA itself has repeatedly argued that states are not required to use a source-specific analysis. See, e.g., Wildearth, 770 F.3d at 944; Arizona, 815 F.3d at 539-40 (upholding EPA decision not to conduct source-specific analysis). If the Clean Air Act empowered EPA to draft reasonable progress goals on a blank slate, EPA’s action may be permissible, but the Clean Air Act limits EPA to a deferential role. EPA must defer to Texas’s goals so long as the Texas goals comply with the Act. See 42 U.S.C. § 7410(k)(3) (EPA “shall approve” a state implementation plan that satisfies the requirements of the Act); Luminant, 675 F.3d at 921. EPA’s lack of deference to the state inverts the agency’s “ministerial function” in this system of “cooperative federalism.” Luminant, 675 F.3d at 921.
Petitioners are likely to establish that EPA improperly failed to defer to Texas’s application of the statutory factors and improperly required a source-specific analysis not found in the Act or Regional Haze Rule.
2. Consultation between Texas and Oklahoma
The Regional Haze Rule requires states to “consult with the other State(s) in order to develop coordinated emission management strategies” if emissions in one state affect visibility at a protected area in another state. 40 C.F.R. § 51.308(d)(3)(I). The Regional Haze Rule adds that states may collect and project visibility data using a regional planning process and that a state implementation plan must document its compliance with any agreements that the regional planning process produces. 40 C.F.R. § 51.308(d)(3)(h), (hi).
Texas and Oklahoma consulted through CENRAP, the regional planning association. CENRAP assessed the impact each state’s emissions had on visibility in other member states. Texas and Oklahoma relied on the CENRAP process to ensure they satisfied the requirement to control emissions causing visibility impairment in downwind states.
EPA disapproved Oklahoma’s consultation with Texas because EPA disagreed with Oklahoma’s decision not to demand further emissions controls at plants located in Texas. 81 Fed. Reg. at 302-03. EPA’s disapproval seems to stem in large part from its assertion that Texas had to conduct a source-specific analysis and provide Oklahoma with that source-specific analysis.32
*429Neither the Regional Haze Rule nor the Clean Air Act explicitly requires upwind states to provide downwind states with source-specific emission control analysis. The Regional Haze Rule only requires that “[w]here the State has emissions that are reasonably anticipated to contribute to visibility in [protected areas] in another State or States, the State must consult with the other State(s) in order to develop coordinated emission management strategies.” 40 C.F.R. § 51.308(d)(3)(I). The Clean Air Act makes no mention of interstate consultation at all.
Given the absence of a regulation or statute requiring source-specific consultations, the extent of negotiations between CENRAP states, the volume of analysis produced by CENRAP, and the fact that EPA has never before disapproved the consultation between states under the Regional Haze Rule, Petitioners have a strong likelihood of success in showing that EPA’s disapproval of the consultation between Oklahoma and Texas was arbitrary and capricious.
3. Effective Date of Emissions Controls
Petitioners also have a strong likelihood of success in establishing that EPA exceeded its statutory authority by imposing emissions controls that go into effect years after the period of time covered by the current round of implementation plans.
The Regional Haze Rule requires states to “consider ... the emission reduction measures needed to achieve [the reasonable progress goal] for the period covered by the implementation plan,” and to impose “enforceable emissions limitations, compliance schedules, and other measures, as necessary to achieve the reasonable progress goals.” 40 C.F.R. § 51.308(d)(l)(i)(B), (d)(3) (emphasis added). The Regional Haze Rule provides that each implementation plan will cover a ten-year period; before the close of each ten-year period, the state must submit a comprehensive revision to cover the next ten-year period. 40 C.F.R. § 51.308(b), (f) (first implementation plan due December 2007; first “comprehensive periodic revision” due July 31, 2018, and every ten years thereafter).
The emissions controls included in a state implementation plan, therefore, must be those designed to achieve the reasonable progress goal for the period covered by the plan. 40 C.F.R. § 51.308(d)(l)(i)(B). When the EPA disapproves a SIP and proposes a FIP, it stands in the position of the state with all the same requirements and powers the state had in initially drafting its SIP. Here, the state implementation plans under review only cover the period up to 2018.33 Yet EPA’s federal implementation plan requires power plants in Texas to meet reasonable progress goals by installing scrubbers in 2019 and 2021. 81 Fed. Reg. at 347. Petitioners persuasively argue that this exceeds the power granted by the Regional Haze Rule.
EPA responds that it has the statutory authority to impose emission control requirements outside the ten-year window because the Clean Air Act gave EPA the flexibility to require revised implementation plans at ten- to fifteen-year intervals. 42 U.S.C. § 7491(b)(2)(B). The minimal deference owed to an agency interpretation first raised during the course of litigation is insufficient to persuade us that *430EPA’s interpretation of the Clean Air Act and Regional Haze Rule is reasonable.34 The Clean Air Act may grant EPA the authority to require new state implementation plans with fifteen-year coverage periods, but EPA bound states (and accordingly bound itself) to a ten-year window when it promulgated the Regional Haze Rule.
As Petitioners observe, if EPA wishes to extend the ten-year regulatory window, it may do so by amending the Regional Haze Rule. EPA, apparently recognizing this, has already proposed amendments to the Regional Haze Rule to remove the language tying emissions controls to the reasonable progress controls “for the period covered by the implementation plan.” 81 Fed. Reg. at 26,972.35 The regulations in effect when Texas and Oklahoma submitted their plans, however, require states to set the emissions controls necessary to achieve the reasonable progress goal for 2018. Agency actions must be assessed according to the statutes and regulations in effect at the time of the relevant activity. See Caring Hearts Personal Home Servs., Inc. v. Burwell, No. 14-3243, 2016 WL 3064870 (10th Cir. May 31, 2016) (vacating CMS sanctions imposed under regulations that came into effect years after the relevant claims were filed). “[I]t is elementary that an agency must adhere to its own rules and regulations. Ad hoc departures from those rules, even to achieve laudable aims, cannot be sanctioned.” Reuters Ltd. v. FCC, 781 F.2d 946, 950 (D.C. Cir. 1986).
EPA also defends the 2019 and 2021 deadlines by arguing that Texas ought not to benefit from Texas’s delay in promulgating a state implementation plan. Were there some evidence that Texas’s alleged intransigence caused the delay in the promulgation of the Final Rule, we might be inclined to consider this argument. But Texas submitted its implementation plan for approval in 2009. EPA waited seven years before finalizing its disapproval in 2016. It does not seem that Texas created inordinate delay in order to obstruct EPA. EPA may not use its own delay as an excuse for imposing burdens on Texas that the Regional Haze Rule does not permit.
Petitioners have a strong likelihood of showing that EPA acted in excess of its statutory power when it disapproved the Texas state implementation plan for failing to require scrubbers that will not be installed until the state implementation plan is no longer in effect.
4. Consideration of Costs
Petitioners further challenge EPA’s federal implementation plan, arguing that EPA did not adequately consider the costs of the power plant changes — mostly installation of sulfur dioxide scrubbers — when it imposed the plan.
The Clean Air Act requires EPA to consider “costs of compliance” when it develops its reasonable progress goals and sets the emission controls necessary to obtain them. 42 U.S.C. § 7491(g)(1). EPA set a cost threshold for emissions controls in terms of cost per ton of emissions re*431duction. Commenters suggested that because the purpose of the Regional Haze Rule is to improve visibility, EPA should instead assess costs in terms of cost per deciview of visibility improvement. EPA declined to do so. 81 Fed. Reg. at 319. Notably, even though the designated areas have all achieved better visibility than their targets for 2018,36 the federal implementation plan imposes emissions controls that will cost $2 billion without achieving any visibility changes in the time period covered by the plan. Because, as discussed infra, Petitioners have a strong likelihood of establishing other flaws in EPA’s federal implementation plan, we need not decide whether EPA’s use of $/ton metrics instead of $/dv metrics fell short of its obligation to consider the costs of its regulations, or whether the costs imposed are unreasonable as a whole in light of the minimal visibility benefits the FIP would achieve in the relevant time period. See Michigan v. EPA, 135 S.Ct. at 2712; Entergy Corp. v. Riverkeeper, Inc., 556 U.S. 208, 225-26, 129 S.Ct. 1498, 173 L.Ed.2d 369 (2009) (“[W]hether it is ‘reasonable’ to bear a particular cost may well depend on the resulting benefits; if the only relevant factor was the feasibility of the costs, their reasonableness would be irrelevant.”).
5. Grid Reliability
The Clean Air Act requires EPA to consider “the energy ... impacts of compliance” with the emission controls in a SIP or FIP. 42 U.S.C. § 7491(g)(1). Petitioners have a strong likelihood of success in showing that EPA failed to do so when it devised its FIP because the Final Rule would render several of the affected electrical generating units uneconomical and cause the closure of 3,000 to 8,400 MW of generating capacity in Texas.
In its electrical grid, as in so many things, Texas stands alone. While all the other states in the Union have extensive interconnections with neighboring states, nearly 90% of Texas is covered by a single isolated grid with limited connections to external power supplies. This grid shares, the name of its governing board, the Electric Reliability Council of Texas (ERCOT). Pub. Utility Comm’n of Tex. v. City Pub. Serv. Bd. of San Antonio, 53 S.W.3d 310, 312 (Tex. 2001). ERCOT’s independence makes the Texas electrical grid uniquely vulnerable to sudden power shortages when power plants in the state unexpectedly close because each power plant provides a larger fraction of the grid’s total power than individual power plants in either the Western or Eastern Interconnections.
EPA, responding to concerns ERCOT submitted in the notice-and-comment period, attempted to address all questions about the impact of the Final Rule on power capacity in Texas in half of one page of the Federal Register. 81 Fed. Reg. at 345. EPA criticized ERCOT’s rules for failing to require “meaningful notice” from producers planning to close plants and suggested that any reliability concerns arose not from the Final Rule but from ERCOT’s regulatory system.37 Id. EPA also criticized ERCOT’s comment for understating new gas turbine capacity in the state and for overstating the likelihood of plant closures (although the power company petitioners in the present case agree *432with ERCOT’s projections for plant closures). Id. EPA summarily dismissed concerns about grid reliability in Texas and relied on a report prepared by a private expert outside the agency to bolster its conclusion. 81 Fed. Reg. at 345.
[[Image here]]
The Final Rule, other than pointing to the report of EPA’s outside expert, does not detail why the emissions controls in question would not endanger reliability or cause the closure of up to 8,400 MW of generating capacity, as ERCOT’s studies suggest. While the agency is free to rely on outside experts to support its conclusions, the level of deference owed to an agency’s conclusions is substantially diminished when the subject matter in question lies beyond the agency’s expertise. Cf. Brown & Williamson Tobacco Corp., 529 U.S. at 132, 120 S.Ct. 1291 (deference is justified by “the agency’s greater familiarity with the ever-changing facts and circumstances surrounding the subjects regulated.”). As EPA’s reliance on an outside expert demonstrates, EPA has no expertise on grid reliability — its sister agency FERC, uninvolved in this regulatory scheme or this rulemaking, is the federal expert in that area. Therefore the deference owed to EPA’s assertions about grid reliability are diminished and the agency must support its arguments more thoroughly than in those areas in which it has considerable expertise and knowledge. *433Particularly when contrasted with the expert report of ERCOT, the group with the greatest knowledge regarding questions of grid reliability in Texas, EPA’s truncated discussion of grid reliability indicates that the agency may not have fulfilled its statutory obligation to consider the energy impacts of the FIP.
Given the exceptional complexity of grid reliability concerns in Texas, EPA’s limited authority to dictate how ERCOT should run the Texas grid, and the explicit directive in the Clean Air Act that implementation plans “take[ ] into consideration ... the energy ... impacts of compliance,” 42 U.S.C. § 7491(g)(1), it is noteworthy that the Final Rule provides neither an exemption from compliance when necessary to preserve the power supply nor a more rigorous exploration of the impact of the Final Rule on grid reliability. Petitioners have a strong likelihood of success in showing that the Final Rule’s failure to include either may render it arbitrary and capricious.
B.
Petitioners have demonstrated several irreparable injuries if the Final Rule is not stayed. They argue that compliance with the Final Rule would impose $2 billion in costs on power companies, businesses, and consumers. Because plant emission controls take several years to install, the regulated companies will have to begin installation almost immediately. The costs of compliance would not only increase rates for consumers but would also endanger the reliability of power in ERCOT if plant operators close facilities rather than install or upgrade uneconomical emissions controls. These closures would permanently shut down plants with up to 8,400 MW of generating capacity.39 The petitioner steel mills and business associations allege they would also suffer injury as their input costs rise substantially.40 The petitioner unions argue that their members would lose their employment at the various power and industrial plants that are threatened by the rule. Petitioners also state that the absence of a stay would require the Public Utility Commission of Texas to spend significant resources enforcing compliance with a voided federal implementation plan rather than enforcing a valid state implementation plan. Finally, Petitioners assert that allowing the Final Rule to stand pending the appeal would disrupt the system of cooperative federalism enshrined in the Clean Air Act. See Michigan v. EPA, 268 F.3d at 1083.
The losses Petitioners allege are sufficient to satisfy the irreparable injury prong of the stay test. The tremendous costs of the emissions controls impose a substantial financial injury on the petitioner power companies which, in this circuit, “may also be sufficient to show irreparable injury.” Enter. Int’l Inc. v. Corp. Estatal Petrolera Ecuatoriana, 762 F.2d 464, 472-73 (5th Cir. 1985). Indeed “complying with a regulation later held invalid almost always produces the irreparable harm of nonrecoverable compliance costs.” Thunder Basin Coal Co. v. Reich, 510 U.S. 200, 220-21, 114 S.Ct. 771, 127 L.Ed.2d 29 (1994) (Sealia, J., concurring in part and in the judgment). When determining whether injury is irreparable, “it is not so much the magnitude but the irreparability that *434counts.... ” Enter. Int’l, 762 F.2d at 472. No mechanism here exists for the power companies to recover the compliance costs they will incur if the Final Rule is invalidated on the merits.41
EPA relies on precedent from other circuits to argue that all the alleged injuries are purely financial, and cannot satisfy the irreparable injury prong. See Am. Hosp. Ass’n v. Harris, 625 F.2d 1328, 1331 (7th Cir. 1980); Freedom Holdings, Inc. v. Spitzer, 408 F.3d 112, 115 (2d Cir. 2005); Mexichem Specialty Resins, Inc. v. EPA, 787 F.3d 544 (D.C. Cir. 2015). These same cases, however, recognize Petitioners’ injury because our sister circuits categorize financial losses as irreparable injury “where no ‘adequate compensatory or other corrective relief will be available at a later date, in the ordinary course of litigation,’ ” Mexichem, 787 F.3d at 555 (quoting Wis. Gas, 758 F.2d at 674), or “where the loss threatens the very existence of the movant’s business,” Wis. Gas Co., 758 F.2d at 674. The plant closures here threaten the very existence of some of Petitioners’ businesses and, even assuming, arguendo, that the plant operators could recover their costs from ERCOT or their consumers, this would not be a recovery made in the course of the litigation.
Here Petitioners have raised threatened harms — including unemployment and the permanent closure of plants — that would arise during the litigation if a stay is not granted, that are irreparable, and that are great in magnitude. Even setting aside the costs of compliance for the power company petitioners, if the Final Rule causes plant closures, the threat of grid instability and potential brownouts alone constitute irreparable injury to Texans. Similarly, the institutional injury to Texas from the inversion of the federalism principles enshrined in the Clean Air Act may constitute irreparable injury. In sum, Petitioners have shown irreparable injury.
C.
The third and fourth factors ask whether the absence of a stay will injure other parties and whether the public interest favors or disfavors a stay. EPA asserts that a stay would injure the public by delaying the achievement of natural visibility at covered areas under the Clean Air Act. But EPA’s asserted injury from a stay is unconvincing as it acknowledges that its proposed implementation plan would not reduce emissions for at least three years, after the next revision window has opened. Given the miniscule difference between the 2018 visibility goals in the federal and state implementation plans, given that current visibility already exceeds even the federal goals for 2018, and given that the major emissions controls will not take effect until 2019 and 2021 (although producers would incur significant costs imminently), we are not persuaded that a stay would injure EPA or the intervenors.42 Petitioners, who them*435selves represent a broad spectrum of the public including labor unions, manufacturers, power providers, and consumer groups, argue that because of the increase in electricity costs and the threat to grid reliability, the public interest favors a stay.
We agree with Petitioners that the public’s interest in ready access to affordable electricity outweighs the inconsequential visibility differences that the federal implementation plan would achieve in the near future. See, e.g., Sierra Club v. Ga. Power Co., 180 F.3d 1309, 1311 (11th Cir. 1999) (“[A] steady supply of electricity during the summer months, especially in the form of air conditioning to the elderly, hospitals and day care centers, is critical.”); Tri-State Generation & Transmission Ass’n v. Shoshone River Power, Inc., 805 F.2d 351, 357 (10th Cir. 1986) (holding that public interest favored an injunction pending appeal when necessary to preserve power supply to the public).
D.
We have the power to stay the agency’s action “to the extent necessary to prevent irreparable injury[.]” 5 U.S.C. § 705. Petitioners request a stay of the Final Rule in its entirety. EPA, in passing, requests that any stay be “narrowly tailored.” Because EPA offers nothing beyond this cursory comment, it has waived any argument about the scope of the stay. See United States v. Green, 964 F.2d 365, 371 (5th Cir. 1992) (“Failure to prosecute an issue on appeal constitutes waiver of the issue.”). Although it is our understanding that the emissions controls in the Final Rule are the only portion of the Final Rule with practical effect and therefore a stay of those requirements effectively operates as a stay of the Final Rule, neither party has briefed how we might craft a limited stay. Therefore, we stay the Final Rule in its entirety, including the emissions control requirements, pending the outcome of this petition for review.
IV.
Section 7607(b)(1) directs that challenges to EPA’s assessment of a state implementation plan may only be filed in the appropriate regional circuit. Because the Final Rule is not based on a determination that has nationwide scope or effect, the narrow exception in § 7607(b)(1) does not apply. Venue for this challenge is appropriate in this court.
Petitioners have demonstrated a strong likelihood of success in establishing that EPA acted arbitrarily, capriciously, and in excess of its statutory authority when it disapproved the Texas and Oklahoma implementation plans and imposed a federal implementation plan. Petitioners have also shown a threat of irreparable injury if a stay is not granted and have demonstrated that EPA will not suffer injury if a stay is granted. Finally, Petitioners have shown that the balance of public interests weigh in favor of a stay. Therefore, Petitioners’ *436motion for a stay of the Final Rule is GRANTED.
. The guidelines were updated in 2005 after the 1999 version was partially vacated. Am. Corn Growers Ass'n v. EPA, 291 F.3d 1 (D.C. Cir. 2002) (vacating 1999 Haze Rule because it imposed technology requirements stricter than those permitted by the Clean Air Act); 70 Fed. Reg. 39,104 (July 6, 2005) (replacing vacated 1999 Haze Rule).
. Changes in visibility are measured in deci-views. A higher deciview measurement indicates more haze and less visibility. 40 C.F.R. § 51.301. A single deciview is around the increment that the average person can perceive with the naked eye. Nat’l Parks Conservation Ass’n v. EPA, 788 F.3d 1134, 1139 n.2 (9th Cir. 2015).
. Wichita Mountains Wildlife Refuge: About the Refuge, U.S. Fish & Wildlife Serv., http:// www.fws.gov/refuge/Wichita_Mountains/ about.html (last visited May 3, 2016).
. Michael Welsh, Landscape of Ghosts, River of Dreams: A History of Big Bend National Park 2-6 (2002).
. Guadalupe Mountains National Park: Geologic Formations, Nat’l Park Serv., https:// www.nps.gov/gumoAearn/nature/geologic formations.htm (last visited May 4, 2016).
. Guadalupe Mountains: Animals, Nat’l Park Serv., https://www.nps.gov/gumoAearn/ nature/animals.htm (last visited May 4, 2016).
. The Clean Air Act gives EPA flexibility to determine the length of time between revisions to implementation plans and the length of time each implementation plan will cover. See 42 U.S.C. § 7491(b)(2)(B) (requiring EPA to draft regulations requiring states to develop "a long-term (ten to fifteen years) strategy for making reasonable progress toward meeting the national goal-"). When it promulgated the Regional Haze Rule, EPA elected to bind states to a ten-year revision period. 40 C.F.R. § 51.308(f). When EPA steps into the shoes of a state to develop a federal implementation plan, that period is binding on EPA as it was on the state. See 77 Fed. Reg. at 40,164 ("At the point EPA becomes obligated to promulgate a FIP, EPA steps into the State’s shoes, and must meet the same requirements. ...”).
. State implementation plans for haze reduction need not impose additional requirements if other emissions controls will achieve the necessary visibility improvements. Texas relied on the requirements of the Clean Air Interstate Rule (which imposed National Ambient Air Quality Standards) to meet its haze obligations. After the D.C. Circuit vacated the Clean Air Interstate Rule, North Carolina v. EPA, 531 F.3d 896, 929 (D.C. Cir. 2008), modified by 550 F.3d 1176 (remanding for reconsideration without vacatur), EPA disapproved Texas’s haze plan because it had relied on the vacated Rule's requirements.
. The alleged inadequacy of the consultation and the reasonable progress goals that Oklahoma set as a result of the consultation seem to be Oklahoma’s only involvement in the Final Rule. All of the additional emission controls in EPA’s federal implementation plan affect only electrical generating units located in Texas. 81 Fed. Reg. at 298. Oklahoma has not sought to participate in any of the petitions for review.
. After filing this challenge here, Petitioners also filed for review of the same rule in the D.C. Circuit, Texas v. EPA, No. 16-1078 (filed Mar. 4, 2016), consolidated with Nos. 16-1086, 16-1087, 16-1083, 16-1091, 16-1084, 16-1085, and in the Tenth Circuit, Luminant Generation Co. v. EPA, No. 16-9508 (filed Mar. 2, 2016), consolidated with Nos. 16-9509, 16-9511, 16-9512. The D.C. Circuit has suspended the filing deadlines in the consolidated petitions and has yet to set a briefing schedule. Texas v. EPA, No. 16-1078, (D.C. Cir. Apr. 6, 2016) (order suspending filing deadlines). The Tenth Circuit has ordered the petitions abated pending resolution of EPA’s motion to dismiss or transfer by this court. Luminant Generation Co. v. EPA, No. 16-9508 (10th Cir. Jun. 16, 2016) (order granting motions to intervene, consolidate, and abate).
. Our sister circuits also determine the jurisdiction of the court without deference. Shweika v. Dep't of Homeland Sec., 723 F.3d 710, 718 (6th Cir. 2013) (“Chevron deference does not apply to an agency’s interpretation of a federal court’s jurisdiction.”); Lindstrom v. United States, 510 F.3d 1191, 1195 n.3 (10th Cir. 2007) ("Determining federal court jurisdiction is exclusively the province of the courts regardless of what an agency may say.” (internal quotation marks omitted)); Bechtel v. Competitive Techs., Inc., 448 F.3d 469, 478 (2d Cir. 2006) (Leval, J., concurring in the judgment) ("[BJecause the statutory interpretation at issue concerns the scope of federal court jurisdiction, it is not a proper subject of deference under Chevron."); Verizon Md., Inc. v. Global NAPS, Inc., 377 F.3d 355, 383 (4th Cir. 2004) (Niemeyer, J., concurring in part and dissenting in part) ("Chevron deference is not required when the ultimate question is about federal jurisdiction.”); *418Murphy Exploration & Prod. Co. v. Dep't of Interior, 252 F.3d 473, 478-80 (D.C. Cir. 2001) ("Chevron does not apply to statutes that ... confer jurisdiction on the federal courts.”), modified on denial of petition for reh’g, 270 F.3d 957 (D.C. Cir. 2001), cited with approval in NetCoalition v. SEC, 715 F.3d 342, 348 (D.C. Cir. 2013).
. State implementation plans under the Regional Haze Rule are the subject of frequent litigation in the regional courts of appeal. See, e.g., Arizona v. EPA, 815 F.3d 519 (9th Cir. 2016) (upholding EPA’s partial disapproval of Arizona’s state implementation plan and the replacement federal implementation plan); Wildearth Guardians v. EPA, 770 F.3d 919 (10th Cir. 2014) (upholding EPA’s approval of state implementation plans for Wyoming, New Mexico, and Utah); North Dakota v. EPA, 730 F.3d 750 (8th Cir. 2013) (vacating in part and upholding in part EPA’s disapproval of North Dakota's implementation plan and the replacement federal implementation plan); Oklahoma v. EPA, 723 F.3d 1201 (10th Cir. 2013) (upholding EPA's partial disapproval of Oklahoma's state implementation plan and the replacement federal implementation plan).
. In reading the allocation of petitions between circuit courts as a venue provision, we are mindful of the Supreme Court's consistent instruction that courts should not infer jurisdictional limitations when a statute does not expressly limit jurisdiction. See, e.g., V.L. v. E.L., - U.S. -, 136 S.Ct. 1017, 1021-22, 194 L.Ed.2d 92 (2016); Arbaugh v. Y&H Corp., 546 U.S. 500, 513-16, 126 S.Ct. 1235, 163 L.Ed.2d 1097 (2006).
. Section 7607(b)(1) directs that "[a] petition for review of the Administrator's action ■ in approving or promulgating any implementation plan ... or any other final action ... which is locally or regionally applicable” may only be filed in the regional circuit. 42 U.S.C. § 7607(b)(1). Regardless of whether "which is locally or regionally applicable” modifies only "any other final action” or both “any other final action” and "action in approving or promulgating any implementation plan,” see Barnhart v. Thomas, 540 U.S. 20, 27-28, 124 S.Ct. 376, 157 L.Ed.2d 333 (2003) (describing the last-antecedent canon), the statutory text places review of SIP approvals or disapprov-als in the regional circuits while providing -an exception for review of a small subset of those actions in the D.C. Circuit.
. EPA concedes that courts of appeals can and should independently consider whether the Administrator has published a suitable finding. If a circuit court holds that the Administrator made no publication, or made an inadequate publication, the exception transferring venue to the D.C. Circuit does not apply. See, e.g., Nat’l Parks Conservation Ass'n v. McCarthy, 816 F.3d 989, 993 (8th Cir. 2016) (EPA made no publication); Lion Oil Co. v. EPA, 792 F.3d 978, 981-82 (8th Cir. 2015) (EPA publication was defective because EPA notified regulated party privately rather than publishing finding in Federal Register).
. In the alternative, EPA argues that only the D.C. Circuit can assess the scope or effect of the determinations. This argument is unsupported by any statutory text and is directly contrary to the familiar maxim that ''[w]hen judicial review depends on a particular fact or legal conclusion, then a court may determine whether that condition exists.” Okoro v. INS, 125 F.3d 920, 925 n.10 (5th Cir. 1997) (internal quotation marks omitted).
. Nor does EPA establish that it has unre-viewable discretion under the standard test for reviewability. An agency's conclusions are unreviewable only in limited circumstances when "(1) statutes preclude judicial review ... [or (2) ] the statute is drawn so that a court would have no meaningful standard against which to judge the agency’s exercise of discretion.” Heckler v. Chaney, 470 U.S. 821, 828, 830, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985); see also 5 U.S.C. § 701(a). Although the Administrator may have discretion over the decision to publish a finding, the statutory text does not give it exclusive discretion to assess the scope or effect of the determinations.
. "The area covered by a given activity or subject.” Scope, American Heritage Dictionary (2d. College Ed. 1982).
. "Something brought about by a cause or agent; result.” Effect, Id.
. "Throughout a whole nation.” Nationwide, Id.
. EPA’s determinations that Texas’s and Oklahoma’s SIPs have inadequate reasonable progress goals and long term strategies are the core determinations that are relevant for the § 7607(b)(1) inquiry, but even assuming, arguendo, that the action is "based on” all the determinations EPA made in the course of promulgating the Final Rule, we see none that has nationwide scope or effect. EPA "determined that Texas’ analysis [setting reasonable progress goals] is inadequate because it does not provide the information necessary to determine the reasonableness of controls at those sources....” 81 Fed. Reg. at 301. It "determined that Texas has satisfied [the monitoring] requirement.” Id. It "determined that nine facilities (with 21 units) merited further modeling to assess what the visibility benefits might be....” 81 Fed. Reg. at 303. It "determined the amount the baseline visibility values exceeded the natural visibility conditions to calculate visibility impairment for each area.” 81'Fed. Reg. at 306. EPA "determined that the CAMx photochemical model was best suited” to model visibility impacts at the affected sites on the 20% worst days. 81 Fed. Reg. at 304. EPA “determined] that under a proper assessment of reasonable progress factors, additional controls for some sources in Texas are warranted for the first planning period.” 81 Fed. Reg. at 316. EPA "determined that additional controls on Parish and Welsh were not required for reasonable progress for the first planning period.” 81 Fed. Reg. at 331. Because EPA “determined that the visibility impacts due to [certain] facilities was almost entirely due to their sulfate emissions [EPA] determined that to address the visibility impacts on 20% worst days from these sources, it was only necessary to evaluate sulfate controls for this planning period.” 81 Fed. Reg. at 332. EPA "determined that it was reasonable to focus [its] analysis on point sources of S02 and NOx-” Id.
. Although the SIP process is generally highly fact-bound and particular to the individual state, EPA has made determinations in other SIP approvals that may have nationwide scope or effect. For example, in another rule-making, it determined that CSAPR, a nation*422wide regulation governing NAAQS, "would provide for greater reasonable progress than BART and established regulations that allow certain states to rely on CSAPR to meet the S02 and NOx BART requirements.” 81 Fed. Reg. at 301-02 (discussing a determination made in Regional Haze: Revisions to Provisions Governing Alternatives to Source-Specific Best Available Retrofit Technology Determinations, 77 Fed. Reg. 33642-01 (June 7, 2012)). A determination that a national standard satisfies a particular requirement in each state may be a determination that has nationwide scope or effect. But EPA explicitly declined to make that determination in the Final Rule: "Given the uncertainty arising from the remand of Texas' CSAPR budgets, we have concluded that it would not be appropriate to finalize our proposed determination to rely on CSAPR as an alternative to S02 and NOx BART for EGUs in Texas at this time.” 81 Fed. Reg. at 302.
.[A]n agency must give adequate reasons for its decisions. The agency must examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts and the choice made. That requirement is satisfied when the agency's explanation is clear enough that its path may be • reasonably discerned. But where the agency has failed to provide even that minimal level of analysis, its action is arbitrary and capricious and so cannot carry the force of law. Encino Motorcars, LLC v. Navarro, - U.S. -, 136 S.Ct. 2117, 2125, 195 L.Ed.2d 382 (2016) (internal quotation marks omitted).
. In the Final Rule, EPA found that "our action on the Texas and Oklahoma regional haze SIPs, which includes the promulgation of a partial FIP for each state, is based on a determination of nationwide scope and effect.” 81 Fed. Reg. at 346.
. EPA supports its argument by reference to a single comment in one House Report commenting on • 1977 amendments to § 7607(b)(1). We do not consider passing commentary in the legislative history, however, when the statutory text itself yields a single meaning. United States v. Ron Pair Enters., Inc., 489 U.S. 235, 241, 109 S.Ct. 1026, 103 L.Ed.2d 290 (1989); see also Bank One Chicago, N.A. v. Midwest Bank & Trust Co., 516 U.S. 264, 280, 116 S.Ct. 637, 133 L.Ed.2d 635 (Scalia, J., concurring) ("Legislative history that does not represent the intent of the whole Congress is nonprobative; and legislative history that does represent the intent of the whole Congress is fanciful.”). Furthermore, even if we were to consider the House Report, it would weigh in favor of review in this circuit. The House Report from which EPA extracted its passing reference explicitly adopted the views expressed by the Administrative Conference of the United States, which observed that “available transfer provisions” could prevent any "undue duplication of proceedings.” 41 Fed. Reg. 56,767 (Dec. 30, *4231976). The normal governing transfer provision dictates that venue for a petition for review is proper in the circuit where a petition for review is first filed. 28 U.S.C. § 2112(a). In the present case, Petitioners' petition for review in this circuit was the first filed.
. Compare 77 Fed. Reg. 36,044, 36,071, 36,-078 (June 15, 2012) (setting a visibility goal of 16.92 deciviews for the Carlsbad Caverns National Park measured at "the IMPROVE monitoring site ... located in Guadalupe Mountains National Park, Texas”), with 81 Fed. Reg. at 301, 347 (setting a visibility goal of 16.26 for the Guadalupe Mountains National Park measured using the IMPROVE monitoring site in Guadalupe Mountains).
. Our determination of Petitioners' likelihood of success on the merits is for the purposes of the stay only and does not bind the merits panel. See Mattern v. Eastman Kodak Co., 104 F.3d 702, 704 (5th Cir. 1997), abrogated on other grounds by Burlington N. & Santa Fe Ry. Co. v. White, 548 U.S. 53, 126 S.Ct. 2405, 165 L.Ed.2d 345 (2006); see also Planned Parenthood of Greater Tex. Surgical Health Servs. v. Abbott, 734 F.3d 406, 419 (5th Cir. 2013). By contrast, our conclusions regarding jurisdiction and venue in evaluating the motion to dismiss or transfer are not preliminary.
. Environmental Protection Agency, Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Program, 3-2 (2003), available at https://www3.epa.gov/ttn/ caaa/tl/memoranda/rh_envcurhr_gd.pdf.
. Because EPA’s use of a source-specific analysis provides a sufficient basis for con-eluding that EPA's disapproval of the reasonable progress goals was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with the law,” we do not address whether EPA acted arbitrarily or exceeded its statutory authority by disapproving the natural visibility estimates.
. [W]e find that this additional analysis [that Texas provided to Oklahoma] was inadequate because the large control set Texas selected was not appropriately refined, targeted, or focused on those sources having significant and potentially cost-effective visibility benefits and did not provide the information necessary to determine the rea*429sonableness of controls at those sources in Texas that have the greatest visibility impacts at the Wichita Mountains.
81 Fed. Reg. at 300-01.
. 40 C.F.R. § 51.308(f).
. EPA raised this argument for the first time in its response to Petitioners’ motion for a stay and asks the court to defer to this novel interpretation of the Clean Air Act. As we have stated in prior Clean Air Act cases, Chevron deference does not apply to statutory interpretations an agency advances in documents, like litigation documents, that do not bind with the force of law. Luminant, 675 F.3d at 928; see also Calix v. Lynch, 784 F.3d 1000, 1007 (5th Cir. 2015) (Agency interpretations "warrant Chevron deference so long as they were established prior to the case under consideration.”).
. Under the proposed amendments, Petitioners’ argument would be much less compelling.
. 81 Fed. Reg. at 341 (acknowledging that visibility on the 20% worst days is better than the agency's reasonable progress goals).
. ERCOT rules require a plant operator to provide 90 days’ notice before closing an electrical generating unit in the system. 16 Tex. Admin. Code § 25.502. ERCOT also has authority to compel power plants to stay online in order to guarantee adequate electrical supply. Id.
. Petitioners argue that it is economically infeasible to reopen a plant even if the regulations are invalidated on the merits.
. For petitioner Nucor, for example, electricity costs are the second largest production cost. Substantial rate increases would threaten Nucor’s plants in the state and the livelihoods of the employees who work there.
. EPA suggests that the power companies can request cost recovery from their customers through the state rate recovery process but this does not eliminate the threat of injury to Petitioners. The power company petitioners are not guaranteed any rate recovery would be approved. Approval, if granted, would merely spread the injury more broadly and increase further the damage to the manufacturing petitioners and consumer groups. Furthermore, the costs imposed on parties are irreparable where they cannot be recovered "in the ordinary course of litigation.” Wis. Gas Co. v. PERC, 758 F.2d 669, 674 (D.C. Cir. 1985). Even recoverable costs may constitute irreparable harm "where the loss threatens the very existence of the movant’s business.” Id..-,
. Intervenors Sierra Club and National Parks Conservation Association introduce expert reports to argue that the FIP would produce health benefits for the public. We are not *435persuaded this argument is relevant. Section 7491 is concerned not with health but with the beauty of our nation’s wild spaces. Other provisions of the Clean Air Act which are not relevant here protect public health. See, e.g., 42 U.S.C. § 7410. As EPA itself observed, “for purposes of this action, we are not authorized to specifically consider [health, welfare, and economic benefits, including ecosystem and tourism benefits] under the regional haze program.” 81 Fed. Reg. at 325. Even assuming, arguendo, that the technological controls required by the Final Rule would improve health in a manner relevant to the public interest component of the stay test, because the emissions controls need not be in place until 2019 at the earliest, these health benefits (unlike the costs of initiating the scrubber installation process) would only be realized well after this petition for review concludes regardless of whether a stay is granted.