With its 1990 amendments to the Clean Air Act (the "Act"), Congress established a program for addressing acid rain, for the first time seeking to cut air pollution by means of marketable permits. See 42 U.S.C. §§ 7651-7651%io; see generally Indianapolis Power & Light Co. v. U.S. EPA, 58 F.3d 643 (D.C.Cir.1995). The basic idea of such a regulatory device is that if polluters for which cutbacks are relatively costly can buy pollution entitlements from ones for which cutbacks are relatively cheap, the nation can achieve a much greater overall cutback for a given expenditure of resources (or achieve a given cutback for a lower expenditure).

We deal here with a necessary aspect of any such program--the allocation of initial entitlements. Congress intended that for "Phase II" of the general emissions trading scheme (Phase I targeted specific heavy polluters), entitlements would add up to emissions of 8.95 million tons of sulphur dioxide (SO sub2 ) per year by affected utilities, 10 million tons below 1980 levels. See 42 U.S.C. § 7651b(a)(1) (basic allowances of 8.9 million tons); id. at § 7651d(a)(3) (providing for an additional 50,000 tons). It charged the Environmental Protection Agency ("EPA") with allocating these entitlements among 2,200 electricity plants throughout the country. In most cases, the EPA was to calculate the allowance by multiplying the unit's "baseline"--i.e., the average of the annual amount of fossil fuel it consumed during 1985, 1986, and 1987--by the lesser of (1) a plant's actual emissions rate (usually for 1985), or (2) the regulatory ceiling on that rate. 41 U.S.C. §§ 7651a, 7651d. To accomplish this, the EPA was to construct what it calls a "National Allowance Database" or "NADB" of fuel consumption ("baseline") data and emissions data using existing government data sources. The "baseline" data was to be drawn from reports by plants on a standard Department of Energy form, known as Energy Information Administration ("EIA") Form 767. 42 U.S.C. § 7651a(4)(A). If a unit did not file an EIA form during the relevant period, the baseline fuel consumption was to be the level specified for the unit in the 1985 National Acid Precipitation Assessment Program ("NAPAP") Emissions Inventory, Version 2, National Utility Reference File ("NURF") or in a corrected database established by EPA. Id.; Brief of Respondent at 8-9 & n.5. Actual 1985 emissions rates--distinct from "baseline" or fuel consumption rates--were to be the emissions rates reported in NURF. 42 U.S.C. § 7651a(16). Once the allowances determined by the formula were aggregated for the 2,200 units in question, they could then be reduced across the board so that they would add up to only the permissible 8.95 million tons per year.

The petitioners here are utilities asserting that they were shortchanged in the allocation process in the following ways. (1) American Municipal Power-Ohio, Inc. ("AMP-Ohio") asserts that EPA was unjustified in using the average sulphur content of fuel burned by Ohio utilities in 1985 in calculating the actual 1985 SO sub2 emission rates for AMP-Ohio's Gorsuch plant. EPA says that in doing so it was bending over backwards on AMP-Ohio's behalf, as the latter had failed to provide proper data for the Gorsuch plant and could have been denied any allocation for the plant. (2) Indiana Municipal Power Association ("IMPA") and Wyandotte Municipal Service Commission ("Wyandotte") own units that started operation between October 1, 1990 and December 31, 1992. Because they failed to submit data about these plants in accordance with statutory and administrative deadlines (as read by the EPA), the EPA denied the owners any entitlements for these plants. (3) Texas Municipal Power Authority ("TMPA") contends that the EPA misinterpreteds 402(4)(A) of the Act, 42 U.S.C. § 7651a(4)(A), which tells the EPA how to treat plant outages, and as a result wrongly denied TMPA an adjustment in its fuel consumption in the measuring period for a 33-day shutdown caused by mechanical difficulties. (4) Nebraska Public Power District, Southwestern Public Service Company and Arco Coal Company (collectively, "NPPD") say that the EPA misconstrued § 402(18) of the Act, 42 U.S.C. § 7651a(18), which governs the conversion of disparate emissions limitations rates into annualized limits on emissions of pounds of SO sub2 per million British thermal units ("mmBtu") (a measure of the heat content of the fossil fuel consumed). According to NPPD, the statute calls for no conversion at all when the pre-existing limitation is expressed in pounds per million Btu, without an express statement of the time period over which it is to be measured.

Before we reach the merits, we must consider two preliminary issues. First, we raised on our own the question whether 42 U.S.C. § 7607(b)(1), which allocates petitions for review of certain "locally or regionally applicable" actions of the EPA to "the appropriate circuit" (conceived in geographic terms), is jurisdictional or merely a matter of venue. Second, the EPA asserts that § 402(4)(C) of the Act, 42 U.S.C. § 7651a(4)(C), barring judicial review of certain "corrections" in the process by which the EPA transforms material from various databases into final allocations, precludes our consideration of the first three claims.

We conclude that § 7607(b)(1) is a matter of venue, not jurisdiction; since EPA raised no objection, the provision is no bar to our review. We also find that § 402(4)(C) does not preclude review of any of the petitioners' claims. Finally, on the merits, we reject each of those claims.

EPA contends that we cannot review the so-called database challenges of AMP-Ohio, IMPA and Wyandotte, and TMPA under § 402(4)(C)'s preclusion of judicial review clause. Section 402(4)(C) provides:

The Administrator shall, upon application or on his own motion, by December 31, 1991, supplement data needed in support of this subchapter and correct any factual errors in data from which affected ... units' baselines or actual 1985 emission rates have been calculated. Corrected data shall be used for purposes of issuing allowances under the subchapter. Such corrections shall not be subject to judicial review, nor shall the failure of the Administrator to correct an alleged factual error in such reports be subject to judicial review.

42 U.S.C. § 7651a(4)(C) (1994). It is argued that this language, read in light of the statutory scheme and objectives and the legislative history, overcomes the presumption in favor of reviewability of agency action and precludes any challenge to a database allocation. See Block v. Community Nutrition Inst., 467 U.S. 340, 345, 349, 104 S. Ct. 2450, 2453-54, 2455-56, 81 L. Ed. 2d 270 (1984).

Petitioners claim, however, that § 402(4)(C)'s scope is narrow. Its language precludes review only of changes to the NURF-NAPAP data, not the NADB. Since the language refers to "data from which" baselines or actual emission rates are calculated--the NURF-NAPAP--and allows corrections "in such reports," it does not extend to the NADB or the emissions allowances. And it only bars review of "corrections" of data, not "additions" to data. Here, there were no data in the historical data sets for EPA to correct. In a similar vein, TMPA argues that § 402(4)(C) bars review only of the original determination of the baseline emission rates, not of subsequent "adjustments" to established baselines pursuant to § 402(4)(A), under which it requested a change; the two actions are distinct, as are the statutory sections.1 And if subsection (C)'s preclusion of review already applied, subsection (A)'s language giving the Administrator "sole discretion" to exclude shutdown periods from the baseline calculation would be superfluous.

EPA argues, instead, that § 402(4)(C) includes changes to both the previously-filed NURF-NAPAP data and the NADB. The section refers to "[c]orrected data ... used for purposes of issuing allowances," which must mean, according to EPA, the NADB. Moreover, the "data" referred to logically must include the NADB; since the NADB is composed of and calculated from the NURF-NAPAP data, a change in the latter necessarily requires a change in the former. The provision's subsequent reference to "such reports" therefore includes the NADB. EPA also claims that an "addition" to an incomplete data set is a "correction" of that data set. Or, at minimum, it is a "supplement" to that data, and § 402(4)(C) expressly includes any "supplement [to] data needed in support of this subchapter." It would be anomalous for Congress to allow review of such changes while precluding review of any change to existing data. As to TMPA's § 402(4)(A) argument, EPA contends that "adjustments" under subsection (A) are simply another type of "correction" that is unreviewable under § 402(4)(C). The data adjustments that subsection (A) addresses are part of the NADB--subsection (A) refers to the baseline specified in "a corrected data base as established by the Administrator" under subsection (C). And subsection (C) refers broadly to all supplements and corrections to the NADB and expressly includes corrections to data used to calculate baselines, which would seem to include adjustments to that data.

Petitioners also contend that their claims are presumptively reviewable, even under § 402(4)(C)'s preclusion language, because they raise statutory interpretation issues. See Bowen v. Michigan Academy of Family Physicians, 476 U.S. 667, 670-72, 106 S. Ct. 2133, 2135-36, 90 L. Ed. 2d 623 (1986). And EPA has failed to point to any language that would overcome this presumption. The statutory language precludes review only of "factual errors," which are then narrowly defined by the reference to the historical data sets, so legal and procedural challenges remain reviewable. Lindahl v. Office of Personnel Mgmt., 470 U.S. 768, 779-80, 105 S. Ct. 1620, 1627-28, 84 L. Ed. 2d 674 (1985). Thus, AMP-Ohio characterizes its challenge as whether EPA's decision to use a statewide average for fuel sulfur content is a permissible construction of the statute. And TMPA emphasizes that its challenge to EPA's interpretation of "accidents that caused prolonged outages" does not turn on a factual dispute (although the parties argue over whether this was an "explosion") or on EPA's application of its rule, but rather on whether EPA's definitions of "prolonged" and "accident" are reasonable statutory interpretations.

EPA objects that this reading eviscerates § 402(4)(C). Since almost any "alleged factual error" can, to some extent, be characterized as an issue of statutory interpretation, allowing review on that basis would undermine significantly § 402(4)(C)'s preclusion. In the case of AMP-Ohio and IMPA and Wyandotte, the gravamen of their claims is that they submitted data to EPA which EPA improperly failed to include in the NADB: AMP-Ohio challenges the use of a statewide average fuel sulfur content instead of AMP-Ohio's submitted figure, and IMPA and Wyandotte contest the straightforward exclusion of all data pertaining to their units because not timely submitted. TMPA similarly seeks a change in its baseline emission rate. EPA asserts that these are precisely the type of factual claims--turning on the application of statutory and regulatory criteria--that § 402(4)(C) is intended to preclude, regardless of the source of the error.

Finally, both EPA and petitioners argue that the legislative history and congressional purpose favor their interpretation. Thus, petitioners assert that their more restricted view of § 402(4)(C) accords with Congress' purpose. The legislative history is claimed to show that the provision was intended to allow EPA to correct only ministerial, book-keeping mistakes in the previously-filed data. See 136 Cong. Rec. S2,988 (daily ed. March 22, 1990) (statement by sponsor Senator Levin). And since Congress precluded review of these corrections based on their straightforward nature, there is no indication that allowing review of other data claims will undermine the statutory objectives. EPA, by contrast, argues that its reading of § 402(4)(C) is proper because the statutory scheme and legislative history show that Congress intended to preclude judicial review of all changes to the data. Congress imposed short statutory deadlines--requiring EPA rapidly to compile enormous amounts of data and allocate allowances to 2,200 utilities--in order to allow utilities to know their allowances and to be able to trade well in advance of the allowance market's effective date. Such advance knowledge and minimal changes were thought necessary to the market's proper functioning. Judicial review, including review of outage adjustments under § 402(4)(A), would undermine this goal by introducing uncertainty and delay into the allocation process. H.R. REP. NO . 490, 101st Cong., 2d Sess. 371 (1990).

We agree with EPA that "additions" are "corrections" within the meaning of § 402(4)(C) and that the sectioncovers changes in the NADB as well as changes in theNURF-NAPAP data. But this does not resolve the issue of what type of claim is reviewable, i.e., what is a "correction[ to] ... an alleged factual error."2 It became clear at oral argument that EPA is defining such a correction as the subject of any challenge made under § 402(4)--what it calls a "database challenge." The distinction EPA is drawing therefore is not between "factual" corrections and "legal" issues, but between the various sections of the statute. EPA in effect is asserting that any challenge under § 402(4), regardless of its primary thrust, necessarily implicates a factual correction, while any other challenge does not. Accordingly, EPA claims that TMPA's petition challenging the definition of "accidents causing prolonged outages" under § 402(4)(A) cannot be reviewed, while it concedes NPPD's claim that EPA erred in determining its allowable emission rate when it annualized the permit limit under § 402(18) can be.

EPA's construction appears to provide a clear demarcation between reviewable and unreviewable claims.3 But it may preclude review of claims that, although they will inevitably affect NADB figures and allowance allocations, focus primarily on an interpretive issue. On the other hand, confining the scope of preclusion to mere correction of ministerial errors, as petitioners suggest, does not accord with the provision's language--e.g., its references to "supplement" as well as "correct"--or Congress' apparent purpose in enacting it; and such a narrow interpretation certainly is not compelled by legislative history. We are uncomfortable with both positions.

While the statutory language does make a distinction between challenges to factual errors and legal/statutory interpretation issues, that distinction--at least as shown in these cases--is not easily applied, and, in any event, EPA's interpretation or application of that concept does not accord with traditional understandings of the distinction. Many factual corrections to the NADB would seem to depend on a process and a choice of a particular methodology, allowing it plausibly to be characterized as a legal issue--an interpretation of what process the statute requires or permits. For example, AMP-Ohio challenges the inclusion of a particular data element in the NADB, representing the actual amount of sulfur dioxide its unit emitted, yet EPA's decision to include that number rather than AMP-Ohio's figure likely involves an interpretation of what the statutory requirement of "actual emission rate" entails. Similarly, TMPA seeks a correction (or adjustment) of the data representing its unit's emissions for the baseline years to reflect an outage, but the requested change seems to turn on the definition of "accidents causing prolonged outages." As these examples show, attempting to distinguish claims that are factual versus legal appears difficult, without any obvious standards or principles.

EPA implicitly recognizes this difficulty in trying to draw the reviewability line in accordance with sections of the statute. It emphasizes, quite correctly, § 402(4)(C)'s importance in minimizing challenges to the database and the clear congressional intent to preclude at least some challenges. But the distinction EPA suggests is simply not supported by the statutory reference to "factual errors." Its practical effect is equally untenable: it would preclude review of claims, such as TMPA's, that appear primarily legal, but allow those brought under other statutory sections, such as NPPD's petition challenging the annualization of its permit emissions limit, even if the goal of the suit were to change the NADB data in order to obtain more allowances. Congress clearly could preclude any challenge to the NADB, but we cannot fairly conclude that it has done so on the terms EPA suggests. Thus, while we are highly sympathetic to EPA's claim, we are presented with no principled way to implement § 402(4)(C)'s preclusion language in these cases. We are therefore obliged to reach the merits of these petitions.4

We reach a similar conclusion with regard to EPA's contention that under § 402(4)(C) petitioners' procedural claims relating to changes in the database are also unreviewable. EPA argues that this follows logically from the preclusion of review of corrections; allowing review of the lack of a response to a comment on a correction would provide backdoor review of EPA's decision, undermining the statutory scheme in precisely the same manner by causing delay and disruption. This is said to accord with congressional intent to limit procedural review. See § 307(d)(8). Petitioners note, however, that statutory provisions precluding judicial review of factual determinations generally do not extend to procedural or legal questions unless expressly stated. Lindahl, 470 U.S. at 778-81, 105 S.Ct. at 1626-28. That EPA has discretion in deciding whether to correct factual errors does not, therefore, bar review of EPA's compliance with procedural requirements. And compliance would not, as EPA implicitly suggests, be futile; having to articulate an explanation may lead the agency to reach a different result, even if there is no review of the substance of that result. We think petitioners likely correct, and having determined that EPA's premise--that § 402(4)(C) bars any database challenge--is faulty, we can hardly apply § 402(4)(C)'s preclusion to procedural claims.

We were also troubled--although the issue was not raised by the parties--by the potential application of § 307(b)(1) to these petitions. We asked the parties to address its application at oral argument. Both parties agree that the case is properly before us, but disagree on the effect of § 307(b)(1). That section provides that

[a] petition for review of ... any other 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 ... 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.

42 U.S.C. § 7607(b)(1) (1994). EPA contends that § 307(b)(1) determines subject matter jurisdiction--not venue--and therefore we should consider it sua sponte. Nevertheless, we have jurisdiction over these petitions, it is argued, since the NADB and the allocation of emissions allowances are nationally applicable, or at minimum have nationwide scope and effect. Since there is a national cap on the number of allowances, a change for one unit or for a category of units, even though it may appear local or regional in character, affects the overall allocation.5 And treating the NADB as a nationally applicable rule does not undermine § 307(b)(1)'s "locally applicable" clause, which is relevant to other EPA actions, such as state implementation plans.

The national scope of the NADB, EPA asserts, is similar to that of the regulation at issue in NRDC v. Thomas, 838 F.2d 1224, 1249 (D.C.Cir.1988). There, we determined that a challenge brought by industry petitioners to the EPA's refusal to exempt previously compliant sources from demonstrating compliance with new source requirements was to a nationally applicable action that could only be brought in the D.C. Circuit. We noted that "the clearly nationwide scope of the regulation is controlling.... If the jurisdictional provision turns on the de facto scope of the regulation, choice of the correct forum might raise complex factual and line-drawing problems." Id. In other words, the nationwide scope of the database allocation rule should control, rather than the local incidence of any one allowance allocation.

However, the case most directly on point, Madison Gas & Electric Co. v. EPA, 4 F.3d 529, 530 (7th Cir.1993), reached the opposite conclusion. In Madison Gas, the petitioner challenged its allocation of emission allowances in the Seventh Circuit, claiming its allowances were based on an incorrect determination of its generating capacity. EPA moved for dismissal under § 307(b)(1), claiming this was a challenge of national application that must be brought in D.C. The court noted that the challenge was neither to "a national feature of the acid-rain program" nor to "a state implementation plan or some other regulation avowedly local or regional," but rather was an "intermediate case": a challenge to an element of a national program, based on an entirely local factor. This did not require petitioner to file in the D.C. Circuit, the court determined, because the only potentially national effect--the concern with piercing the national ceiling of allowances--was too speculative.

EPA does not ask us to disagree with Madison Gas, but would have us confine its holding to the facts of the case. Petitioners agree, asserting that if § 307(b)(1) is a jurisdictional provision, jurisdiction in the D.C. Circuit is appropriate since their claims are not based on entirely local factors. Both parties, then, believe that Thomas is distinguishable from Madison Gas. (Madison Gas characterized Thomas as dealing with a "national feature" of the acid rain program.)6 But we think this distinction between "entirely local" and "local but having national effect," like the one between fact and law under § 402(4)(C), rather elusive. The petitions here--e.g., the dispute over the actual emissions at one of AMP-Ohio's units--appear no more "national" than the one at issue inMadison Gas.

We do not have to resolve this issue, however, since we agree with petitioners' alternative contention that § 307(b)(1) is a venue provision, the application of which can be waived.7 Section 307(b)(1) can be read as prescribing the choice among circuits and not the power of a particular federal circuit court to hear a claim. This reading is supported by the provision's reference to where a petitioner may "file," and by its unequivocal characterization in the legislative history as a venue provision. See H.R. REP. NO . 294, 95th Cong., 1st Sess. 323-24 (1977). Since parties may normally consent to be sued in a court that would otherwise be an improper venue, EPA's failure to object waives the issue. See FED.R.CIV.P. 12(h)(1); C.A. WRIGHT, FEDERAL COURTS 257 (1994) (citing Neirbo Co. v. Bethlehem Shipbuilding Corp., 308 U.S. 165, 60 S. Ct. 153, 84 L. Ed. 167 (1939)).

EPA objects to characterizing this as a venue provision. It notes that § 307(b) is entitled "Judicial review," that the language is mandatory rather than providing a "choice" of circuits, and that the provision also includes a clearly jurisdictional 60-day limit for filing petitions for review. While there is some "jurisdictional" language, as EPA asserts, that is not determinative; we think it more significant that federal court power to entertain petitions is clear, that the provision refers to where a petitioner must file, and that the apparent congressional purpose was to place nationally significant decisions in the D.C. Circuit. Given the less than clear language, the structure of the section--dividing cases among the circuits--and the legislative history indicate that § 307(b)(1) is framed more as a venue provision. Therefore, EPA's failure to object waives § 307(b)(1)'s venue requirements, and the petitions are appropriately before us.

Turning to the merits, we reject all substantive claims of these petitioners. Specifically, we hold that it was within EPA's authority to rely on an emission rate calculated from a state-wide average in determining the allowances to be received by a utility unit whose emission rate was not included in the relied-upon database, to deny Wyandotte's and IMPA's efforts to request allowances after the published deadline for such requests, and to define "prolonged," as used in 42U.S.C. § 7651a(4)(A), as meaning at least three months in duration. We also reject NPPD's argument that EPA improperly annualized its relevant allowable emission rate. Finally, we deny all procedural claims raised by these petitioners because they have not first presented their claims before the agency, 42 U.S.C. § 7607(d)(7)(B), because they have not demonstrated an error on the part of the EPA, or because they have not demonstrated sufficient prejudice from the alleged procedural deficiencies to justify reversal. 42 U.S.C. § 7607(d)(8).

Section 405(c)(2) of the 1990 Amendments to the Clean Air Act directs the EPA to grant certain fossil-fueled utility units a number of emission allowances equal to "the product of the unit's baseline multiplied by the lesser of its actual 1985 emissions rate or its allowable 1985 emissions rate, divided by 2,000." 42 U.S.C. § 7651d(c)(2). Section 402(16) of these amendments defines the term "actual 1985 emission rate" for an electric utility unit to mean "the annual sulfur dioxide ... emission rate in pounds per million Btu as reported in the NAPAP Emissions Inventory, Version 2, National Utility Reference File." 42 U.S.C. § 7651a(16). No statutory provision, however, suggests how the EPA should define an existing unit's actual emission rate should the unit not appear in the NURF-NAPAP database named in § 402(16).

The EPA addressed this gap in the statutory scheme through a notice published along with the second version of the National Allowance Data Base ("NADB"), the national database that catalogs the number of allowances granted to a particular unit. See Acid Rain Provisions, 56 Fed.Reg. 33,278 (July 19, 1991). This "July 1991 Notice" allowed a utility to submit a claim for allowances for a unit that was not listed in the NURF-NAPAP database, but it required that the utility also submit certain information, including the unit's "actual emission rate" and some supporting data, including the sulfur content and ash retention of the fuel burned at the unit, that could be used to verify the submitted rate. See id. at 33,283-85.

In response to this July 1991 Notice, AMP-Ohio submitted to the EPA a request for allowances for one of its facilities--the Richard Gorsuch generating station--accompanied by various information about that unit. This information included an "actual emission rate" for the facility, but did not include the other data required by the July Notice. Though AMP-Ohio later supplemented its submission with additional information, it again omitted the critical supporting data. Nonetheless, EPA included the Gorsuch facility in its third proposed version of the NADB. See Acid Rain Allowance Allocations and Reserves, 57 Fed.Reg. 29,940, 30,000 (July 7, 1992). In this version, the EPA granted some allowances to the Gorsuch facility, but far fewer than AMP-Ohio had expected.

AMP-Ohio protested this result during the appropriate comment period, and alleged a number of errors that the EPA made in calculating its allowances. The EPA corrected several computational and other errors. AMP-Ohio, however, still did not submit the supporting data that was necessary to confirm the emission rate it had asserted in its application, and the EPA did not specifically encourage AMP-Ohio to submit this information.

In its final version of the NADB, the EPA granted the Gorsuch facility a total of almost 20,000 allowances, see Acid Rain Allowance Allocations and Reserves, 58 Fed.Reg.15,634, 15,684 (Mar. 23, 1993), which was roughly half of the number of allowances AMP-Ohio expected from the data it had submitted. EPA subsequently informed AMP-Ohio that it had received fewer allowances because the EPA had based the allowances, not on the emission rate submitted byAMP-Ohio, but on an emission rate calculated using the average sulfur content of the fuel burned by utilities in Ohio in 1985. The EPA stated that it used this calculated rate because it did not have the supporting data necessary to verify the accuracy of AMP-Ohio's claimed emission rate. The EPA has since explained that this rate was calculated according to a technique used to compute emission rates for other databases, including the NURF-NAPAP database that § 402(16) specified was to be used for defining a unit's actual emission rate. See 42 U.S.C. § 7651a(16).

AMP-Ohio challenges this result on both substantive and procedural grounds.

Substantively, AMP-Ohio protests the EPA's decision to use an emission rate that the EPA calculated in part from the statewide average of sulfur content in utility fuel instead of the emission rate that AMP-Ohio submitted (albeit without the necessary supporting data). AMP-Ohio contends that this substitution of a calculated emission rate for the submitted rate was an impermissible construction of the applicable statutory sections, or, alternatively, was unlawfully arbitrary and capricious. See 5 U.S.C. § 706.

a. Did the EPA's interpretation of the term "actual emission rate" run contrary to the statute or to reason under Chevron?

Because the EPA is the agency that administers the relevant elements of the Clean Air Act, we review its interpretation of § 402(16) under the familiar two-step analysis of Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-44, 104 S. Ct. 2778, 2781-83, 81 L. Ed. 2d 694 (1984). Under Chevron, we first must determine " 'whether Congress has directly spoken to the precise question at issue.' " Nuclear Information Resource Serv. v. NRC, 969 F.2d 1169, 1173 (D.C.Cir.1992) (in banc) (quoting Chevron, 467 U.S. at 842, 104 S.Ct. at 2781.) If Congress has unambiguously expressed its intent in the statute, our inquiry ends. See id. If the statute is instead silent or ambiguous, we must then " 'defer to the agency's interpretation of the statute if it is reasonable and consistent with the statute's purpose.' " Id. (quoting Chemical Mfrs. Ass'n v. EPA, 919 F.2d 158, 162-63 (D.C.Cir.1990)).

In this case, we have little difficulty concluding that the statute does not address the "precise question" at issue. The statute makes no mention of how the EPA should treat a utility unit whose emission rates were not included in the NURF-NAPAP database. Although courts have, on rare occasion, managed to divine some meaning from silence, see, e.g., Chisom v. Roemer, 501 U.S. 380, 396 & n. 23, 111 S. Ct. 2354, 2364 & n. 23, 115 L. Ed. 2d 348 (1991) (rejecting an interpretation of a statute because "if Congress had such an [unorthodox] intent, congress would have made it explicit in the statute"), a silent statute cannot preclude its reasonable interpretation by the agency that administers it. See Nuclear Information Resource Serv., 969 F.2d at 1173 (quoting Chevron, 467 U.S. at 843, 104 S.Ct. at 2781-82). In view of its silence on the point at issue, we must hold the statute ambiguous.

Having held the statute ambiguous, we must examine whether the agency's interpretation was reasonable. In its July 1991 Notice, the EPA announced that it understood the statute to permit the EPA to allocate allowances to a utility unit missing from the NURF-NAPAP database specified in § 402(16). In this Notice, the EPA also declared that it would not necessarily allocate allowances to these units unless a utility submitted a "missing" unit's 1985 emission rate and certain information so that the EPA might verify this rate. See 56 Fed.Reg. at 33,283-85. The Notice, however, did not describe the EPA's proposed methodology for determining the emission rate in case all of the required information was not submitted.

That omission, however, does not make the EPA's subsequent conduct irrational. In the absence of a rule or stated policy on point, the EPA must possess a reasonable amount of discretion to implement its own regulations. See, e.g., NRDC, Inc. v. EPA, 22 F.3d 1125, 1148 (D.C.Cir.1994); cf., Pension Benefit Guaranty Corp. v. LTV Corp., 496 U.S. 633, 653-56, 110 S. Ct. 2668, 2679-81, 110 L. Ed. 2d 579 (1990) (holding that agency must have discretion to adopt procedures to execute tasks left to it). In this case, it would have made little sense for the EPA to allow AMP-Ohio, which failed to submit the supporting data, to benefit from its sin of omission and simply receive all the allowances that AMP-Ohio claimed the Gorsuch units should receive. On the other hand, it may have seemed unduly harsh for the EPA to deny all allowances to the Gorsuch facility, given that the statutory scheme probably intended units like Gorsuch to receive some fair number of allowances. Instead, navigating between these extremes, the EPA calculated an emission rate for the Gorsuch facility that it could confirm had at least some basis in a verifiable fact: the statewide average of the sulfur content of fuel burned by Ohio utilities. This approach was not unprecedented--the EPA had previously relied on such calculated rates in compiling other databases, including the NURF-NAPAP database that § 402(16) explicitly indicates should be used to determine the "actual emission rate" of a facility included in that database. And though the EPA did not explain its precise method for calculating a rate based on a statewide average that was used in this case until after the close of general proceedings before the agency, the failure of an agency to identify every detail of a process before it is used does not automatically require judicial interference in matters that must be thought to lie within the agency's expertise. See American Meat Institute v. EPA, 526 F.2d 442, 457 & n. 28 (7th Cir.1975). In light of these considerations, we conclude that the EPA's decision to use the calculated rate for determining the allowances allotted to the Gorsuch unit was reasonable and consistent with the statute.

b. Did the EPA otherwise act in an arbitrary or capricious manner?

Similarly, applying the Administrative Procedures Act, we cannot find that the EPA acted arbitrarily in substituting an emission rate based on statewide averages for the emission rate submitted, without the necessary supporting data, by AMP-Ohio. The EPA may, within reasonable bounds, specify what data is necessary for it to determine how many allowances a facility should receive. See, e.g., Alabama Power Co. v. Costle, 636 F.2d 323, 351-52 (D.C.Cir.1979) (permitting the EPA to specify the data necessary for a PSD application under the Clean Air Act). The EPA may then refuse to act if the data submitted is of uncertain authenticity. See Kennecott v. EPA, 780 F.2d 445, 458 (4th Cir.1985), cert. denied, Secondary Lead Smelters Ass'n, Inc. v. Thomas, 479 U.S. 814, 107 S. Ct. 67, 93 L. Ed. 2d 25 (1986).

In this case, the EPA published a general notice as to the information it needed to allocate any allowances to a utility unit missing from the then-current version of the NADB, and AMP-Ohio's submission did not comply with these requirements. That AMP-Ohio may not have understood what was required does not excuse its omissions. It is the duty of a party seeking agency action to press its claim; see, e.g., Northside Sanitary Landfill, Inc. v. Thomas, 849 F.2d 1516, 1519-20 (D.C.Cir.1988), cert. denied, 489 U.S. 1078, 109 S. Ct. 1528, 103 L. Ed. 2d 833 (1989); the agency need not repeatedly warn the party that its submission did not satisfy publicly announced guidelines. Nor, as previously discussed, did the EPA improperly substitute an emission rate calculated according to its established method for determining unknown rates in lieu of the unverifiable rate submitted by AMP-Ohio. Because EPA acted within its discretion in specifying what information a utility had to submit to receive allowances and in determining how the EPA should handle a submission that did not fulfill these requirements, the EPA was neither unreasonable nor impermissibly arbitrary in using a rate based on a statewide average in order to calculate the appropriate emission level of AMP-Ohio's Gorsuch facility.

In addition to its substantive challenges, AMP-Ohio makes a number of procedural attacks. The only three that require some discussion are that the EPA failed to respond to significant comments as required by statute, to request additional supporting data from AMP-Ohio soon enough for AMP-Ohio to provide such information, and to explain how it arrived at the emission rate it used to calculate Gorsuch's final allowances. None of these concerns merit reversal.

a. Did EPA err when it did not affirmatively act to correct AMP-Ohio's deficient submission as to the emission rate?

The first two of these significant challenges may be considered together. After the third proposed version of the NADB was published, AMP-Ohio commented that that version did not appear to use the emission rate it had submitted. EPA, however, neither explained its decision not to use the submitted rate nor promptly notified AMP-Ohio of the utility's failure to include the necessary supporting data in its submission. AMP-Ohio argues that the EPA erred when it did not indicate why it chose not to use the submitted emission rate and when it did not more quickly request the necessary supporting information.

Neither claim establishes error. Precedent has established that an agency need not respond to comments if that response would mainly restate "what had already been set forth" in some published notice. See NRDC, Inc. v. EPA, 859 F.2d 156, 188-89 (D.C.Cir.1988). As the EPA had already declared what information was necessary for a submitted emission rate to be considered, see 56 Fed.Reg. at 33,283-85, it did not have to take additional steps to inform AMP-Ohio individually of these requirements. Similarly, the EPA, as noted, does not bear some general duty to seek out information that an applicant has omitted in a submission, especially when the applicant has some notice that the data was essential. Thus, the EPA did not err when it did not remind AMP-Ohio that it should submit all the data required by the July 1991 Notice.

b. Did the EPA err when it did not explain its method of substituting a calculated emission rate for a submitted rate that was not verifiable?

AMP-Ohio's third set of procedural complaints must also be denied. AMP-Ohio argues that the EPA did not explain how it determined the emission rate eventually used to calculate the Gorsuch unit's final allowances. According to the applicable review provision in the Clean Air Act, however, we have no jurisdiction to hear a claim until it has been "raised with reasonable specificity" before the EPA, either "during the period for public comment" or in a separate petition for reconsideration. 42 U.S.C. § 7607(d)(7)(B). AMP-Ohio has not demonstrated that it ever brought this specific complaint to the attention of the agency during an appropriate comment period. Nor did AMP-Ohio ever demonstrate that it was impracticable for it to do so in a petition for reconsideration. In accordance with § 7607(d)(7)(B), then, we must dismiss review of this procedural challenge. For similar reasons, we must also decline review as to whether the EPA's failure to include the statewide average on which it relied in the record violated 42 U.S.C. § 7607(d)(6)(C).

1. Background Specific to IMPA's and Wyandotte's Claims

Section 405(g)(3) of the 1990 Amendments to the Clean Air Act permits the EPA to allocate allowances to utility units that commence operation between October 1, 1990, and December 31, 1992 ("1992 units"). 42 U.S.C. § 7651d(g)(3). Because these units did not have an actual or allowable 1985 emission rate used to calculate allowances for units operating in 1985, however, the section instructed the EPA to determine the unit's allowances based on the unit's "annual fuel consumption" and "allowable sulfur dioxide emission rate." Id. Of course, as these units had not previously existed, the EPA was largely unaware of the proper number of allowances they should receive.

In a notice published along with the second version of the NADB ("July 1991 Notice"), the EPA established guidelines by which utilities could seek allowances for any units that were not included in the EPA's second version of that database. See Acid Rain Provisions, 56 Fed.Reg. 33,278, 33,279 (July 19, 1991). This July 1991 Notice warned all applicants that:

In order to finalize the database in a timely manner, EPA will not accept requests for data corrections following the comment period [which ended September 3, 1991], except when the utility can demonstrate that the correct data were unavailable during the comment period. No data will be changed or added after publication of the final database in December, 1991.... Units eligible for allowances will not be allocated allowances if the final database does not include the information necessary.

Id. at 33,283. The Notice also stated that any unit seeking to add data must submit a data change request form and necessary supporting documentation. See id. The form itself listed 36 items of information that a utility had to supply, ranging from the name and location of the plant, general design specifications of the plant, to the allowable emission rate of the plant. See id. at 33,283-86. The Notice, however, did not explicitly state that it intended to govern allocation requests for 1992 units, though a number of such units were absent from the second version of the database.

On May 13, 1992, IMPA first contacted the EPA about four utility units that were to become "imminently" operational. Though none of the units were in the second version of the NADB, IMPA had not previously notified the EPA about these units. IMPA explained that it had not submitted information earlier, despite the July 1991 Notice's statement that the EPA would not make corrections after the end of that year, because IMPA did not think the July 1991 Notice applied to 1992 units, as some data required by the Notice may not have been available until a unit commenced operations. IMPA thus had not submitted any data about these units, even though IMPA's own correspondence made clear that IMPA had known much basic information about the units prior to the end of 1991.

Another utility that was constructing a new unit slated to become operational in 1992, the Wyandotte Municipal Services Commission ("Wyandotte"), also notified the EPA later than any deadline named in the July Notice, in August 1992. Like IMPA, Wyandotte did not show that the data on its 1992 unit was unavailable in 1991. In fact, Wyandotte's own late submission, like IMPA's, indicated that much information was known before the close of the July 1991 Notice's comment period. Wyandotte, however, explained that the utility had not sent information on its new boiler earlier because it had interpreted the local instructions for data correction as not allowing submissions for units under construction.

In a document published in March 1993, the EPA denied both IMPA's and Wyandotte's requests for allowances for these 1992 units because they did not comply with the submission deadlines in the July 1991 Notice. Neither IMPA nor Wyandotte had submitted information clearly available to each utility on their respective 1992 units prior to September 3, 1991, as, according to the EPA, the July Notice required, and neither had submitted all necessary data by the end of December 1991. IMPA and Wyandotte now protest their exclusions.

IMPA contends that the EPA was arbitrary and capricious in denying IMPA's new facility allowances simply because IMPA did not comply with the deadlines established in the July 1991 Notice. IMPA does not challenge the EPA's authority to define when and how a utility must request allowances for its units under the Clean Air Act. Instead, IMPA argues that we should not interpret the July 1991 Notice to have applied to units, like the ones at issue, that were still under construction in 1991.

Although we normally extend great deference to an agency's interpretation of its own regulations, see, e.g., Consarc Corp. v. Office of Foreign Assets Control, 71 F.3d 909, 914 (D.C.Cir.1995), IMPA observes that our precedent has refused to extend this deference to an unclear agency policy that would, if applied, result in a "drastic" penalty to the private party. See General Elec. Co. v. EPA, 53 F.3d 1324, 1328-29 (D.C.Cir.1995); Radio Athens, Inc. v. FCC, 401 F.2d 398, 404 (D.C.Cir.1968). Instead, in these "drastic" cases, we have generally accepted the threatened petitioner's reasonable interpretation of the agency's policy, see, e.g., Satellite Broadcasting Co., Inc., v. FCC, 824 F.2d 1, 3-4 (D.C.Cir.1987); Gates & Fox Co., Inc. v. OSHRC, 790 F.2d 154, 156 (D.C.Cir.1986), unless the petitioner, through a good-faith review of the statements issued by the agency, should have been able to determine, "with ascertainable certainty," the criteria that the agency expected the petitioner to satisfy. General Elec. Co., 53 F.3d at 1329 (citation and internal quotations omitted). Consequently, IMPA contends that we should accept its interpretation that the Notice did not apply to its units because a decision to deny all allowances to its 1992 units would constitute an unfairly severe penalty, because the Notice was not clear, and because IMPA's interpretation, that the Notice did not apply to units not yet in operation, was reasonable.

Assuming that the loss of all allowances would be a drastic penalty, and that the scope of the July 1991 Notice was not clear, we still cannot conclude that IMPA's interpretation of the July 1991 Notice was reasonable in light of the contents and context of that Notice. The text of the Notice specified that the EPA would complete the final version of the allowance database by December 31, 1991. See, e.g., 56 Fed.Reg. at 33,279. The Notice, however, also implied that it contained the last directions the EPA would issue prior to the publication of this final database. See id. A party interpreting the July 1991 Notice at the time of the Notice thus could either conclude that the Notice applied to 1992 units, or that the EPA had decided to offer no guidance at all as to how these units should seek allocations prior to the imminent publication of the finalized database.

The latter reading was not reasonable. No sensible party could have thought that the EPA would specify 36 types of information necessary for an existing unit to add to or correct the data to be used for calculating its allowance without saying one word about what a 1992 unit had to submit. Yet, that is what IMPA claims. Accordingly, we conclude that, given these facts, IMPA was unreasonable to think the July Notice did not apply to 1992 units.

Having concluded that IMPA could not have reasonably thought that the July 1991 Notice did not apply to 1992 units, we must deny IMPA's claim. IMPA clearly did not meet the standards of the July Notice. IMPA did not submit available information on the unit--such as the plant name or certain technical or design data--by September 3, 1991. Nor did IMPA demonstrate that this or other data was unavailable as of September 3, 1991, which, under the July 1991 Notice, would have allowed IMPA to submit the data until December 1991. Nor did IMPA even submit any data to the EPA by December 1991. In short, IMPA did not comply with any aspect or reasonable interpretation of the July 1991 Notice. As that Notice must have applied to IMPA's 1992 units, we cannot overrule the agency's decision not to award these units any allowances.

Having determined that any reasonable interpretation of the July 1991 Notice would apply to 1992 units, we must also reject Wyandotte's claim for allowances for its 1992 unit. Wyandotte, like IMPA, did not submit any information until well after the deadlines specified in the July Notice, even though at least some information on its 1992 units was available well before that date. That Wyandotte wrongly interpreted local instructions on whether and when a utility should submit information on 1992 units also does not excuse the tardiness of its allocation request. The July Notice repeatedly forbade utilities from submitting entirely new claims for allowances after 1991. We thus also conclude that the EPA may deny allowances to Wyandotte's 1992 unit.

Section 402(4)(A) of the Clean Air Act Amendments specifies that the Administrator may adjust a unit's emission allowances for "accidents that caused prolonged outages." 42 U.S.C. § 7651a(4)(A). That provision reads, in pertinent part:

For each utility unit that was in commercial operation prior to January 1, 1985, the baseline shall be the annual average quantity of mmBtu's [heat energy] consumed in fuel during calendar years 1985, 1986 and 1987.... The Administrator, in the Administrator's sole discretion, may exclude periods during which a unit is shutdown for continuous period of four calendar months or longer, and make appropriate adjustments under this paragraph. Upon petition of ... any unit, the Administrator may make appropriate baseline adjustments for accidents that caused prolonged outages.

Id. (emphasis added).

As a response to EPA's general request for applications for outage adjustments in the July 1991 Notice, see 56 Fed.Reg. 33,282 (July 19, 1991), but prior to any establishment of standards guiding EPA review of these applications, TMPA submitted an application for an adjustment to its baseline for a 33-day outage during the "calendar years 1985, 1986 [or] 1987" that was caused by the failure of the plant's cooling bars. Having received several applications for similar adjustments, EPA proposed to dispense with these applications by distributing them among several categories based on the statutory requirements. One proposed category would have allowed adjustments for applications that demonstrated that an outage was "prolonged"--tentatively defined as being "four months or longer"--and caused by an "accident"--that is, "the occurrence of a natural phenomenon ... or an incident unrelated to the operation of the unit that is unpreventable, unforeseeable, and not caused by worker error." Acid Rain Provisions, 57 Fed.Reg. 30,034, 30,036-38 (July 7, 1992).

As TMPA's adjustment application stated that its outage lasted only 33 days in the relevant period, it did not qualify for this proposed category (or any proposed category that would have granted it an adjustment). See id. at 30,037 & 30,038 Table 2. Moreover, because the EPA refused to accept any revised applications lest the EPA be unable to publish a corrected NADB by the statutory deadline, see id., TMPA could not supplement its application to show that its outage might qualify for such a category. Instead, TMPA submitted comments about the suggested categories, and it directed the attention of the EPA to a colloquy between two representatives during the floor discussion of § 402(4)(A)--the Barton-Lent colloquy, see 136 Cong. Rec. H12,876-77 (daily ed. Oct. 26, 1990)--that supported TMPA's position that its outage merited some adjustment.

In March 1993, the EPA issued a final interpretation of § 7651a(4)(A). This final rule generally maintained the proposed definition for "accident," but redefined a "prolonged outage" by reducing the necessary minimum from four to three months. Notice of Availability of the National Allowance Data Base, 58 Fed.Reg. 15,720, 15,723 (March 23, 1993). EPA explained this reduction as necessary in order to avoid rendering the statutory language governing accidentally-caused outages mere surplus, as another segment of § 402(4)(A) already provided that a utility that suffered a four-month outage, for whatever reason, could apply for an adjustment. See 42 U.S.C. § 7651a(4)(A). The EPA also concluded that, as three months was longer than the average outage according to a NERC study added to the record by the EPA one day prior to the promulgation of the final rule, three months could reasonably be deemed "prolonged." See id. At that time, the EPA made a final list of applicants that would not receive adjustments, including TMPA.

TMPA challenges the EPA's decisions and its decision-making on several grounds.

Substantively, TMPA challenges EPA's definitions of "prolonged" and "accident." EPA ultimately interpreted "prolonged" to mean longer than three months, which is well longer than the average outage, but shorter than the four months specified in another sentence of § 402(4)(A) required for an adjustment for outages not caused by accidents. See 58 Fed.Reg. at 15,723. TMPA argues that the unambiguous meaning of prolonged denies this definition, and that the term clearly intends to encompass any outage longer than the mean or median outage. TMPA also protests the narrowness of EPA's definition of accident.

a. Did the EPA reasonably interpret the term "prolonged? "

We reject TMPA's challenge to EPA's statutory interpretation of what constitutes prolonged. Again, we perform the "Chevron two-step" in reviewing EPA's interpretations of the Clean Air Act and its amendments. See supra Section III.A.2.b.

Having reviewed § 402(4)(A), we conclude that the term "prolonged," as used in that section, is ambiguous. The common meaning of "prolonged" denotes an extended duration. See, e.g., Webster's Third International Dictionary 1815 (1961). What qualifies as extended, however, is not fixed. At best, the context of the term as used in § 402(4)(A) may suggest that Congress intended any definition of a "prolonged outage[ ]" to be less than a four-month outage, as a unit that suffered an outage lasting four months or more may have been able to receive an adjustment whether or not that outage was caused by an accident. This implication from the statutory context, however, hardly clarifies the meaning of the term sufficiently for us to reach an "unmistakable conclusion that Congress had an intention on the ... question." Ohio v. U.S. Dep't of Interior, 880 F.2d 432, 441 (D.C.Cir.1989).

No other accepted form of statutory interpretation achieves some clearer understanding of "prolonged" as used in the provision. In particular, the presence of the Barton-Lent colloquy does not affect our result. As we have warned in the past, judges must " 'exercise extreme caution before concluding that a statement made in floor debate, or at a hearing, or printed in a committee document may be taken as statutory gospel,' " in light of the " 'endemic interplay, in Congress, of political and legislative consideration[s]' " likely unrelated to the interpretive tasks of a court. Gersman v. Group Health Ass'n, Inc., 975 F.2d 886, 892 (D.C.Cir.1992) (quoting Antolok v. United States, 873 F.2d 369, 377 (D.C.Cir.1989)), cert. denied, --- U.S. ----, 114 S. Ct. 1642, 128 L. Ed. 2d 363 (1994). This caution is especially warranted when, as in this case, it appears that a colloquy was a direct result of "a single member ... attempting to reassure his own constituency or even to create legislative history for citation by courts." Id. In this case, the colloquy in question does not give reason for us to draw the "unmistakable conclusion" that Congress intended "prolonged," as used in § 402(4)(A), to have some fixed meaning.

Undertaking the second step of the Chevron inquiry, we also conclude that EPA's definition of a "prolonged" outage to mean an outage that lasts at least three months is reasonable. A baseline of three months is substantially longer than the average outage, but it is still sufficiently short as not to effectively write the "prolonged outage" language out of the statute. For instance, at least one party described in the record, which actually received an adjustment under another provision, could have sought an adjustment under the "prolonged outage" provision instead. See 58 Fed.Reg. at 15,724 (allowing adjustment for the George Neal plant). EPA thus acted within its authority when it denied TMPA's application for a baseline adjustment under § 402(4)(A). See id. (Table 3).

b. Did the EPA reasonably interpret the term "accident?"

As TMPA conceded at oral argument, a determination that EPA's definition of prolonged is valid moots any claim by TMPA that EPA's interpretation of "accident" is also valid. We thus conclude that TMPA cannot prevail on either of its substantive challenges to EPA's interpretation of § 402(4)(A).

In addition to its substantive challenges, TMPA also raises several procedural arguments. Of these, only three merit any discussion. TMPA contends that the EPA was arbitrary in universally denying any re-application for adjustments after the final rule was issued and that the EPA failed to give TMPA proper notice of the standards it was expected to meet by acting on TMPA's application before the final rule was published. TMPA also claims that the EPA improperly relied on a statistical study that was added to the record after the comment period had closed. Finally, TMPA argues that the EPA was arbitrary because it did not sufficiently discuss TMPA's comments directing the EPA's attention to the Barton-Lent colloquy. We dismiss the first two allegations because they are not properly before us and the last because the EPA did not err in its lack of specific response to the Barton-Lent colloquy.

As noted in the discussion of AMP-Ohio's procedural challenges, 42 U.S.C. § 7607(d) specifies what a petitioner must demonstrate to prevail on a procedural challenge to an EPA action pursuant to subchapter IV, which includes 42 U.S.C. § 7651a(4)(A). See 42 U.S.C. § 7607(d)(1)(T). Under that statutory provision, we may only review claims that have been first raised "with reasonable specificity" before the agency, 42 U.S.C. § 7607(d)(7)(B), and we may reverse only procedural errors that are prejudicial. 42 U.S.C. § 7607(d)(8). Of course, the question of exhaustion precedes any examination of whether any errors were prejudicial. See Cross-Sound Ferry Servs., Inc. v. ICC, 934 F.2d 327, 343-46 (D.C.Cir.1991) (Thomas, J., concurring); cf. UDC Chairs Chapter, American Ass'n of University Professors v. Board of Trustees, 56 F.3d 1469, 1475 (D.C.Cir.1995) (noting that exhaustion is a jurisdictional requirement).

Here, no party suggests that TMPA has properly raised two of these three procedural challenges before the agency. TMPA could have--and, under the statute, should have--addressed its complaints to the agency about the alleged lack of notice and the agency's delayed inclusion of the statistical study either in a specific statement to the agency during an appropriate comment period or in a suggestion for agency reconsideration. 42 U.S.C. § 7607(d)(7)(B). Because TMPA did not do so, however, we cannot review these procedural claims. In particular, we explicitly decline TMPA's invitation to read the "futility" exception present in common-law exhaustion doctrine, see UDC Chairs Chapter, American Ass'n of University Professors, 56 F.3d at 1475, into a statutory provision that permits only objections raised before the agency to be "raised during judicial review." 42 U.S.C. § 7607(d)(7)(B).

We must also reject TMPA's remaining procedural challenge. The EPA simply did not err when it did not more extensively address TMPA's comments regarding the Barton-Lent colloquy. According to our precedent, "[t]he failure to respond to comments is significant only insofar as it demonstrates that the agency's decision was not based on a consideration of the relevant factors." Thompson v. Clark, 741 F.2d 401, 409 (D.C.Cir.1984) (internal quotations and citations omitted). In this case, TMPA has not approached showing that the EPA did not consider the relevant factors in denying TMPA's request for an outage adjustment. Moreover, the EPA's general statement that it would consider all floor colloquies as to specific utilities to the extent appropriate, see 56 Fed.Reg. at 33,282, adequately demonstrates that the agency did not improperly overlook this colloquy, which would be of dubious value as a guide to agency conduct in any case. On this ground, then, we again affirm the EPA.

The acid rain emissions trading system clearly depends on some definition of what, in the end, is to be limited. Congress chose to limit the annual production of tons of SO sub2 . Recall from our introduction that in referring to the initial allocations to be made by the EPA, Congress commonly expressed them as the unit's "baseline"--i.e., the average amount of fossil fuel it consumed in a specified period--multiplied by the lesser of the unit's actual emission rate for 1985 (or another calendar year) or of its "allowable 1985 emissions rate," i.e., the legal limit applicable to the unit. See, e.g., 42 U.S.C. §§ 7651d(b)(3), (b)(4), (c)(2), (c)(3). Because emissions limitations had taken different forms depending on such factors as when the plant was built, it was necessary, to make sure that the EPA was not adding up apples and oranges, to state a rule for conversion into a single, fungible unit. In defining "allowable 1985 emissions rate" Congress in § 402(18) of the Clean Air Act Amendments of 1990 provided for such conversion:

Where the emissions limitation for a unit is not expressed in pounds of emissions per million Btu, or the averaging period of that emissions limitation is not expressed on an annual basis, the Administrator shall calculate the annual equivalent of that emissions limitation in pounds per million Btu to establish the allowable 1985 emissions rate.

§ 402(18), 42 U.S.C. § 7651a(18) (emphasis added). The unit of lbs SO sub2 /mmBtu derives from EPA's regulations promulgated under the 1970 Amendments to the Clean Air Act, in which Congress directed EPA to set performance standards limiting emissions from newly built or modified sources of air pollution. Clean Air Act § 111, 42 U.S.C. § 1857c-6 (1970). EPA responded with a regulation covering large generators whose construction began between 1971 and 1978, setting the limit at 1.2 lbs SO sub2 /mmBtu for coal-burning generators.

The most obvious cases calling for conversion are limits "not expressed in pounds per million Btu" at all, e.g., expressed as pounds per ton of coal on a weekly basis. But as the "or" linking that case in § 402(18) with the next one suggests, the sentence appears also to address limits expressed as pounds per million Btu but "not expressed on an annual basis."

Why so? The answer is that, because of variability in pollution output of a plant over a period of time, a limit must, to be complete, include (implicitly or explicitly) some kind of averaging period. To take a familiar example, a speed limit of 55 MPH means one thing if it is violated by exceeding the limit at one instant (the way such limits are in fact applied), quite another if it were violated only by a motorist who exceeded 55 MPH averaged over an hour. In the latter form, the long-distance commuter could go at 75 where the roads would handle it, and use the crawling pace of center-city traffic to bring his average down to the permitted maximum. Similarly, a plant that must average 1.0 lbs/mmBtu annually is a good deal less constricted than one that must satisfy that standard at every instant in time. Indeed, under the EPA's calculations, which are not contested, a limit of 1.0 lbs/mmBtu measured instantaneously turns into 0.89 lbs/mmBtu averaged annually. The petitioners in this aspect of the case have plants subject to emissions limitations that are simply expressed in pounds per million Btu; but, as they conceded at oral argument, those limitations were to be applied on an instantaneous basis.8 Nonetheless, they object to the EPA's application of its 0.89 conversion factor.

Before addressing petitioners' linguistic arguments, we note briefly a colloquy at oral argument. One of the panel asked counsel "why in the world" Congress would not require conversion for limitations expressed in lbs/mmBtu but not on an annual basis, and counsel could offer no theory. We agree with the view implicit in his candid silence. A trading scheme requires fungible units, as petitioners' brief points out: "[The] variations in the expression of emissions limitations meant that a single, standard form of expression for the 1985 allowable emissions limitation had to be established, and limitations expressed in other forms had to be converted to the standard form"--and there is no reason why the temporal component of the limitation should go unstandardized. Accordingly, in examining the linguistic arguments we start from the unusual perspective that, for petitioners' arguments to prevail, they must be powerful enough to justify a reading that is conceded to make no sense.

To see petitioners' argument most clearly, we repeat the provision, emphasizing the words that petitioners regard as pivotal:

Where the emissions limitation for a unit is not expressed in pounds of emissions per million Btu, or the averaging period of that emissions limitation is not expressed on an annual basis, the Administrator shall calculate the annual equivalent of that emissions limitation in pounds per million Btu to establish the allowable 1985 emissions rate.

§ 402(18), 42 U.S.C. § 7651a(18) (emphasis added). The argument runs that "that emissions limitation" must refer back to the words ahead of the "or," under the so-called "last antecedent" doctrine of statutory construction. Under this view, Congress--despite the "or"--addressed only the problem of pre-existing limitations that were not expressed in pounds per million Btu at all. The second clause, addressing the case where the averaging period "of that emissions limitation" is not expressed on an annual basis, supposedly means that where a limit is not expressed in lbs/mmBtu and has an averaging period of less than a year, the limit is not only converted to lbs/mmBtu but also annualized. (According to the argument, this annualization would not be required at all were it not for this second clause, even though other language expressly directs the Administrator to "calculate the annual equivalent....")

The government responds that the "last antecedent" rule is not absolute, and, indeed, so we have observed. United States v. Pritchett, 470 F.2d 455, 458-59 (D.C.Cir.1972) ("[The] Rule of the Last Antecedent is not an inflexible rule, and is not applied where the context indicates otherwise.") (footnote omitted) (but finding no such indication in that case). Moreover, the government offers a sensible explanation of the phrase "that emissions limitation," namely, that it is a reference back to the phrase "the emissions limitation for a unit" in the opening words of the sentence. The explanation is perhaps a little awkward, in the sense that the second "trigger" for calculating emissions in terms of lbs SO sub2 /mmBtu averaged over a year uses a referent to a phrase embodied in the statement of the first trigger. But this minor awkwardness seems a peccadillo compared to the effect of petitioners' reading, which is to give a windfall to plants whose limitations were expressed as lbs/mmBtu but without a one-year averaging period (and to virtually deny the second phrase any independent force).

Petitioners have another string to their bow. The second clause, even under the government's reading, is confined to lbs/mmBtu limitations "not expressed on an annual basis." Petitioners argue that their limitations were not expressed on any temporal basis at all, so that they are completely outside the scope of § 402(18)'s provision for annualization.

In its use of the 0.89 adjustment factor, the government evidently reasoned--and petitioners do not dispute--that there are no material differences between limits applicable on a strictly instantaneous basis and ones averaged over a period equal to or less than one day. See National Allowance Data Base, Version 2.11, Technical Support Document F-2. The government also says--and again petitioners do not dispute--that limits stated without averaging periods are, as a practical matter under modern measuring technology, measurable over a period of a day or less. Thus, EPA's uncontested assumptions are in full accord with what it did here--treat petitioners as if their limits had expressly provided for averaging over a period of a day or less. See note 1 above. Of course, if the understanding implicit in their limitations had been that they could average over a longer period, they would have had a good case for use of a more favorable conversion factor. But they assert no such implicit understanding. Accordingly, their claim is meritless.

* * *

For the above reasons, the petitions for review are denied.